Compare commits
No commits in common. "c8" and "c9s" have entirely different histories.
7
.gitignore
vendored
7
.gitignore
vendored
@ -1 +1,6 @@
|
||||
SOURCES/glibc-2.28.tar.xz
|
||||
# Release tarballs.
|
||||
/glibc-*.tar.[gx]z
|
||||
# Generated (source) RPMs.
|
||||
/*.rpm
|
||||
# Expanded source trees.
|
||||
/glibc-*/
|
||||
|
||||
@ -1 +0,0 @@
|
||||
ccb5dc9e51a9884df8488f86982439d47b283b2a SOURCES/glibc-2.28.tar.xz
|
||||
File diff suppressed because it is too large
Load Diff
2
README.scripts
Normal file
2
README.scripts
Normal file
@ -0,0 +1,2 @@
|
||||
All of the useful glibc maintainer scripts are located at:
|
||||
https://pagure.io/glibc-maintainer-scripts
|
||||
@ -1,496 +0,0 @@
|
||||
# This file names the currently supported and somewhat tested locales.
|
||||
# If you have any additions please file a glibc bug report.
|
||||
SUPPORTED-LOCALES=\
|
||||
C.UTF-8/UTF-8 \
|
||||
aa_DJ.UTF-8/UTF-8 \
|
||||
aa_DJ/ISO-8859-1 \
|
||||
aa_ER/UTF-8 \
|
||||
aa_ER@saaho/UTF-8 \
|
||||
aa_ET/UTF-8 \
|
||||
af_ZA.UTF-8/UTF-8 \
|
||||
af_ZA/ISO-8859-1 \
|
||||
agr_PE/UTF-8 \
|
||||
ak_GH/UTF-8 \
|
||||
am_ET/UTF-8 \
|
||||
an_ES.UTF-8/UTF-8 \
|
||||
an_ES/ISO-8859-15 \
|
||||
anp_IN/UTF-8 \
|
||||
ar_AE.UTF-8/UTF-8 \
|
||||
ar_AE/ISO-8859-6 \
|
||||
ar_BH.UTF-8/UTF-8 \
|
||||
ar_BH/ISO-8859-6 \
|
||||
ar_DZ.UTF-8/UTF-8 \
|
||||
ar_DZ/ISO-8859-6 \
|
||||
ar_EG.UTF-8/UTF-8 \
|
||||
ar_EG/ISO-8859-6 \
|
||||
ar_IN/UTF-8 \
|
||||
ar_IQ.UTF-8/UTF-8 \
|
||||
ar_IQ/ISO-8859-6 \
|
||||
ar_JO.UTF-8/UTF-8 \
|
||||
ar_JO/ISO-8859-6 \
|
||||
ar_KW.UTF-8/UTF-8 \
|
||||
ar_KW/ISO-8859-6 \
|
||||
ar_LB.UTF-8/UTF-8 \
|
||||
ar_LB/ISO-8859-6 \
|
||||
ar_LY.UTF-8/UTF-8 \
|
||||
ar_LY/ISO-8859-6 \
|
||||
ar_MA.UTF-8/UTF-8 \
|
||||
ar_MA/ISO-8859-6 \
|
||||
ar_OM.UTF-8/UTF-8 \
|
||||
ar_OM/ISO-8859-6 \
|
||||
ar_QA.UTF-8/UTF-8 \
|
||||
ar_QA/ISO-8859-6 \
|
||||
ar_SA.UTF-8/UTF-8 \
|
||||
ar_SA/ISO-8859-6 \
|
||||
ar_SD.UTF-8/UTF-8 \
|
||||
ar_SD/ISO-8859-6 \
|
||||
ar_SS/UTF-8 \
|
||||
ar_SY.UTF-8/UTF-8 \
|
||||
ar_SY/ISO-8859-6 \
|
||||
ar_TN.UTF-8/UTF-8 \
|
||||
ar_TN/ISO-8859-6 \
|
||||
ar_YE.UTF-8/UTF-8 \
|
||||
ar_YE/ISO-8859-6 \
|
||||
ayc_PE/UTF-8 \
|
||||
az_AZ/UTF-8 \
|
||||
az_IR/UTF-8 \
|
||||
as_IN/UTF-8 \
|
||||
ast_ES.UTF-8/UTF-8 \
|
||||
ast_ES/ISO-8859-15 \
|
||||
be_BY.UTF-8/UTF-8 \
|
||||
be_BY/CP1251 \
|
||||
be_BY@latin/UTF-8 \
|
||||
bem_ZM/UTF-8 \
|
||||
ber_DZ/UTF-8 \
|
||||
ber_MA/UTF-8 \
|
||||
bg_BG.UTF-8/UTF-8 \
|
||||
bg_BG/CP1251 \
|
||||
bhb_IN.UTF-8/UTF-8 \
|
||||
bho_IN/UTF-8 \
|
||||
bho_NP/UTF-8 \
|
||||
bi_VU/UTF-8 \
|
||||
bn_BD/UTF-8 \
|
||||
bn_IN/UTF-8 \
|
||||
bo_CN/UTF-8 \
|
||||
bo_IN/UTF-8 \
|
||||
br_FR.UTF-8/UTF-8 \
|
||||
br_FR/ISO-8859-1 \
|
||||
br_FR@euro/ISO-8859-15 \
|
||||
brx_IN/UTF-8 \
|
||||
bs_BA.UTF-8/UTF-8 \
|
||||
bs_BA/ISO-8859-2 \
|
||||
byn_ER/UTF-8 \
|
||||
ca_AD.UTF-8/UTF-8 \
|
||||
ca_AD/ISO-8859-15 \
|
||||
ca_ES.UTF-8/UTF-8 \
|
||||
ca_ES/ISO-8859-1 \
|
||||
ca_ES@euro/ISO-8859-15 \
|
||||
ca_ES@valencia/UTF-8 \
|
||||
ca_FR.UTF-8/UTF-8 \
|
||||
ca_FR/ISO-8859-15 \
|
||||
ca_IT.UTF-8/UTF-8 \
|
||||
ca_IT/ISO-8859-15 \
|
||||
ce_RU/UTF-8 \
|
||||
chr_US/UTF-8 \
|
||||
cmn_TW/UTF-8 \
|
||||
crh_UA/UTF-8 \
|
||||
cs_CZ.UTF-8/UTF-8 \
|
||||
cs_CZ/ISO-8859-2 \
|
||||
csb_PL/UTF-8 \
|
||||
cv_RU/UTF-8 \
|
||||
cy_GB.UTF-8/UTF-8 \
|
||||
cy_GB/ISO-8859-14 \
|
||||
da_DK.UTF-8/UTF-8 \
|
||||
da_DK/ISO-8859-1 \
|
||||
da_DK.ISO-8859-15/ISO-8859-15 \
|
||||
de_AT.UTF-8/UTF-8 \
|
||||
de_AT/ISO-8859-1 \
|
||||
de_AT@euro/ISO-8859-15 \
|
||||
de_BE.UTF-8/UTF-8 \
|
||||
de_BE/ISO-8859-1 \
|
||||
de_BE@euro/ISO-8859-15 \
|
||||
de_CH.UTF-8/UTF-8 \
|
||||
de_CH/ISO-8859-1 \
|
||||
de_DE.UTF-8/UTF-8 \
|
||||
de_DE/ISO-8859-1 \
|
||||
de_DE@euro/ISO-8859-15 \
|
||||
de_IT.UTF-8/UTF-8 \
|
||||
de_IT/ISO-8859-1 \
|
||||
de_LI.UTF-8/UTF-8 \
|
||||
de_LU.UTF-8/UTF-8 \
|
||||
de_LU/ISO-8859-1 \
|
||||
de_LU@euro/ISO-8859-15 \
|
||||
doi_IN/UTF-8 \
|
||||
dsb_DE/UTF-8 \
|
||||
dv_MV/UTF-8 \
|
||||
dz_BT/UTF-8 \
|
||||
el_GR.UTF-8/UTF-8 \
|
||||
el_GR/ISO-8859-7 \
|
||||
el_GR@euro/ISO-8859-7 \
|
||||
el_CY.UTF-8/UTF-8 \
|
||||
el_CY/ISO-8859-7 \
|
||||
en_AG/UTF-8 \
|
||||
en_AU.UTF-8/UTF-8 \
|
||||
en_AU/ISO-8859-1 \
|
||||
en_BW.UTF-8/UTF-8 \
|
||||
en_BW/ISO-8859-1 \
|
||||
en_CA.UTF-8/UTF-8 \
|
||||
en_CA/ISO-8859-1 \
|
||||
en_DK.UTF-8/UTF-8 \
|
||||
en_DK/ISO-8859-1 \
|
||||
en_GB.UTF-8/UTF-8 \
|
||||
en_GB/ISO-8859-1 \
|
||||
en_GB.ISO-8859-15/ISO-8859-15 \
|
||||
en_HK.UTF-8/UTF-8 \
|
||||
en_HK/ISO-8859-1 \
|
||||
en_IE.UTF-8/UTF-8 \
|
||||
en_IE/ISO-8859-1 \
|
||||
en_IE@euro/ISO-8859-15 \
|
||||
en_IL/UTF-8 \
|
||||
en_IN/UTF-8 \
|
||||
en_NG/UTF-8 \
|
||||
en_NZ.UTF-8/UTF-8 \
|
||||
en_NZ/ISO-8859-1 \
|
||||
en_PH.UTF-8/UTF-8 \
|
||||
en_PH/ISO-8859-1 \
|
||||
en_SC.UTF-8/UTF-8 \
|
||||
en_SG.UTF-8/UTF-8 \
|
||||
en_SG/ISO-8859-1 \
|
||||
en_US.UTF-8/UTF-8 \
|
||||
en_US/ISO-8859-1 \
|
||||
en_US.ISO-8859-15/ISO-8859-15 \
|
||||
en_US@ampm/UTF-8 \
|
||||
en_US.UTF-8@ampm/UTF-8 \
|
||||
en_ZA.UTF-8/UTF-8 \
|
||||
en_ZA/ISO-8859-1 \
|
||||
en_ZM/UTF-8 \
|
||||
en_ZW.UTF-8/UTF-8 \
|
||||
en_ZW/ISO-8859-1 \
|
||||
eo/UTF-8 \
|
||||
es_AR.UTF-8/UTF-8 \
|
||||
es_AR/ISO-8859-1 \
|
||||
es_BO.UTF-8/UTF-8 \
|
||||
es_BO/ISO-8859-1 \
|
||||
es_CL.UTF-8/UTF-8 \
|
||||
es_CL/ISO-8859-1 \
|
||||
es_CO.UTF-8/UTF-8 \
|
||||
es_CO/ISO-8859-1 \
|
||||
es_CR.UTF-8/UTF-8 \
|
||||
es_CR/ISO-8859-1 \
|
||||
es_CU/UTF-8 \
|
||||
es_DO.UTF-8/UTF-8 \
|
||||
es_DO/ISO-8859-1 \
|
||||
es_EC.UTF-8/UTF-8 \
|
||||
es_EC/ISO-8859-1 \
|
||||
es_ES.UTF-8/UTF-8 \
|
||||
es_ES/ISO-8859-1 \
|
||||
es_ES@euro/ISO-8859-15 \
|
||||
es_GT.UTF-8/UTF-8 \
|
||||
es_GT/ISO-8859-1 \
|
||||
es_HN.UTF-8/UTF-8 \
|
||||
es_HN/ISO-8859-1 \
|
||||
es_MX.UTF-8/UTF-8 \
|
||||
es_MX/ISO-8859-1 \
|
||||
es_NI.UTF-8/UTF-8 \
|
||||
es_NI/ISO-8859-1 \
|
||||
es_PA.UTF-8/UTF-8 \
|
||||
es_PA/ISO-8859-1 \
|
||||
es_PE.UTF-8/UTF-8 \
|
||||
es_PE/ISO-8859-1 \
|
||||
es_PR.UTF-8/UTF-8 \
|
||||
es_PR/ISO-8859-1 \
|
||||
es_PY.UTF-8/UTF-8 \
|
||||
es_PY/ISO-8859-1 \
|
||||
es_SV.UTF-8/UTF-8 \
|
||||
es_SV/ISO-8859-1 \
|
||||
es_US.UTF-8/UTF-8 \
|
||||
es_US/ISO-8859-1 \
|
||||
es_UY.UTF-8/UTF-8 \
|
||||
es_UY/ISO-8859-1 \
|
||||
es_VE.UTF-8/UTF-8 \
|
||||
es_VE/ISO-8859-1 \
|
||||
et_EE.UTF-8/UTF-8 \
|
||||
et_EE/ISO-8859-1 \
|
||||
et_EE.ISO-8859-15/ISO-8859-15 \
|
||||
eu_ES.UTF-8/UTF-8 \
|
||||
eu_ES/ISO-8859-1 \
|
||||
eu_ES@euro/ISO-8859-15 \
|
||||
fa_IR/UTF-8 \
|
||||
ff_SN/UTF-8 \
|
||||
fi_FI.UTF-8/UTF-8 \
|
||||
fi_FI/ISO-8859-1 \
|
||||
fi_FI@euro/ISO-8859-15 \
|
||||
fil_PH/UTF-8 \
|
||||
fo_FO.UTF-8/UTF-8 \
|
||||
fo_FO/ISO-8859-1 \
|
||||
fr_BE.UTF-8/UTF-8 \
|
||||
fr_BE/ISO-8859-1 \
|
||||
fr_BE@euro/ISO-8859-15 \
|
||||
fr_CA.UTF-8/UTF-8 \
|
||||
fr_CA/ISO-8859-1 \
|
||||
fr_CH.UTF-8/UTF-8 \
|
||||
fr_CH/ISO-8859-1 \
|
||||
fr_FR.UTF-8/UTF-8 \
|
||||
fr_FR/ISO-8859-1 \
|
||||
fr_FR@euro/ISO-8859-15 \
|
||||
fr_LU.UTF-8/UTF-8 \
|
||||
fr_LU/ISO-8859-1 \
|
||||
fr_LU@euro/ISO-8859-15 \
|
||||
fur_IT/UTF-8 \
|
||||
fy_NL/UTF-8 \
|
||||
fy_DE/UTF-8 \
|
||||
ga_IE.UTF-8/UTF-8 \
|
||||
ga_IE/ISO-8859-1 \
|
||||
ga_IE@euro/ISO-8859-15 \
|
||||
gd_GB.UTF-8/UTF-8 \
|
||||
gd_GB/ISO-8859-15 \
|
||||
gez_ER/UTF-8 \
|
||||
gez_ER@abegede/UTF-8 \
|
||||
gez_ET/UTF-8 \
|
||||
gez_ET@abegede/UTF-8 \
|
||||
gl_ES.UTF-8/UTF-8 \
|
||||
gl_ES/ISO-8859-1 \
|
||||
gl_ES@euro/ISO-8859-15 \
|
||||
gu_IN/UTF-8 \
|
||||
gv_GB.UTF-8/UTF-8 \
|
||||
gv_GB/ISO-8859-1 \
|
||||
ha_NG/UTF-8 \
|
||||
hak_TW/UTF-8 \
|
||||
he_IL.UTF-8/UTF-8 \
|
||||
he_IL/ISO-8859-8 \
|
||||
hi_IN/UTF-8 \
|
||||
hif_FJ/UTF-8 \
|
||||
hne_IN/UTF-8 \
|
||||
hr_HR.UTF-8/UTF-8 \
|
||||
hr_HR/ISO-8859-2 \
|
||||
hsb_DE/ISO-8859-2 \
|
||||
hsb_DE.UTF-8/UTF-8 \
|
||||
ht_HT/UTF-8 \
|
||||
hu_HU.UTF-8/UTF-8 \
|
||||
hu_HU/ISO-8859-2 \
|
||||
hy_AM/UTF-8 \
|
||||
hy_AM.ARMSCII-8/ARMSCII-8 \
|
||||
ia_FR/UTF-8 \
|
||||
id_ID.UTF-8/UTF-8 \
|
||||
id_ID/ISO-8859-1 \
|
||||
ig_NG/UTF-8 \
|
||||
ik_CA/UTF-8 \
|
||||
is_IS.UTF-8/UTF-8 \
|
||||
is_IS/ISO-8859-1 \
|
||||
it_CH.UTF-8/UTF-8 \
|
||||
it_CH/ISO-8859-1 \
|
||||
it_IT.UTF-8/UTF-8 \
|
||||
it_IT/ISO-8859-1 \
|
||||
it_IT@euro/ISO-8859-15 \
|
||||
iu_CA/UTF-8 \
|
||||
ja_JP.EUC-JP/EUC-JP \
|
||||
ja_JP.UTF-8/UTF-8 \
|
||||
ka_GE.UTF-8/UTF-8 \
|
||||
ka_GE/GEORGIAN-PS \
|
||||
kab_DZ/UTF-8 \
|
||||
kk_KZ.UTF-8/UTF-8 \
|
||||
kk_KZ/PT154 \
|
||||
kl_GL.UTF-8/UTF-8 \
|
||||
kl_GL/ISO-8859-1 \
|
||||
km_KH/UTF-8 \
|
||||
kn_IN/UTF-8 \
|
||||
ko_KR.EUC-KR/EUC-KR \
|
||||
ko_KR.UTF-8/UTF-8 \
|
||||
kok_IN/UTF-8 \
|
||||
ks_IN/UTF-8 \
|
||||
ks_IN@devanagari/UTF-8 \
|
||||
ku_TR.UTF-8/UTF-8 \
|
||||
ku_TR/ISO-8859-9 \
|
||||
kw_GB.UTF-8/UTF-8 \
|
||||
kw_GB/ISO-8859-1 \
|
||||
ky_KG/UTF-8 \
|
||||
lb_LU/UTF-8 \
|
||||
lg_UG.UTF-8/UTF-8 \
|
||||
lg_UG/ISO-8859-10 \
|
||||
li_BE/UTF-8 \
|
||||
li_NL/UTF-8 \
|
||||
lij_IT/UTF-8 \
|
||||
ln_CD/UTF-8 \
|
||||
lo_LA/UTF-8 \
|
||||
lt_LT.UTF-8/UTF-8 \
|
||||
lt_LT/ISO-8859-13 \
|
||||
lv_LV.UTF-8/UTF-8 \
|
||||
lv_LV/ISO-8859-13 \
|
||||
lzh_TW/UTF-8 \
|
||||
mag_IN/UTF-8 \
|
||||
mai_IN/UTF-8 \
|
||||
mai_NP/UTF-8 \
|
||||
mfe_MU/UTF-8 \
|
||||
mg_MG.UTF-8/UTF-8 \
|
||||
mg_MG/ISO-8859-15 \
|
||||
mhr_RU/UTF-8 \
|
||||
mi_NZ.UTF-8/UTF-8 \
|
||||
mi_NZ/ISO-8859-13 \
|
||||
miq_NI/UTF-8 \
|
||||
mjw_IN/UTF-8 \
|
||||
mk_MK.UTF-8/UTF-8 \
|
||||
mk_MK/ISO-8859-5 \
|
||||
ml_IN/UTF-8 \
|
||||
mn_MN/UTF-8 \
|
||||
mni_IN/UTF-8 \
|
||||
mr_IN/UTF-8 \
|
||||
ms_MY.UTF-8/UTF-8 \
|
||||
ms_MY/ISO-8859-1 \
|
||||
mt_MT.UTF-8/UTF-8 \
|
||||
mt_MT/ISO-8859-3 \
|
||||
my_MM/UTF-8 \
|
||||
nan_TW/UTF-8 \
|
||||
nan_TW@latin/UTF-8 \
|
||||
nb_NO.UTF-8/UTF-8 \
|
||||
nb_NO/ISO-8859-1 \
|
||||
nds_DE/UTF-8 \
|
||||
nds_NL/UTF-8 \
|
||||
ne_NP/UTF-8 \
|
||||
nhn_MX/UTF-8 \
|
||||
niu_NU/UTF-8 \
|
||||
niu_NZ/UTF-8 \
|
||||
nl_AW/UTF-8 \
|
||||
nl_BE.UTF-8/UTF-8 \
|
||||
nl_BE/ISO-8859-1 \
|
||||
nl_BE@euro/ISO-8859-15 \
|
||||
nl_NL.UTF-8/UTF-8 \
|
||||
nl_NL/ISO-8859-1 \
|
||||
nl_NL@euro/ISO-8859-15 \
|
||||
nn_NO.UTF-8/UTF-8 \
|
||||
nn_NO/ISO-8859-1 \
|
||||
nr_ZA/UTF-8 \
|
||||
nso_ZA/UTF-8 \
|
||||
oc_FR.UTF-8/UTF-8 \
|
||||
oc_FR/ISO-8859-1 \
|
||||
om_ET/UTF-8 \
|
||||
om_KE.UTF-8/UTF-8 \
|
||||
om_KE/ISO-8859-1 \
|
||||
or_IN/UTF-8 \
|
||||
os_RU/UTF-8 \
|
||||
pa_IN/UTF-8 \
|
||||
pa_PK/UTF-8 \
|
||||
pap_AW/UTF-8 \
|
||||
pap_CW/UTF-8 \
|
||||
pl_PL.UTF-8/UTF-8 \
|
||||
pl_PL/ISO-8859-2 \
|
||||
ps_AF/UTF-8 \
|
||||
pt_BR.UTF-8/UTF-8 \
|
||||
pt_BR/ISO-8859-1 \
|
||||
pt_PT.UTF-8/UTF-8 \
|
||||
pt_PT/ISO-8859-1 \
|
||||
pt_PT@euro/ISO-8859-15 \
|
||||
quz_PE/UTF-8 \
|
||||
raj_IN/UTF-8 \
|
||||
ro_RO.UTF-8/UTF-8 \
|
||||
ro_RO/ISO-8859-2 \
|
||||
ru_RU.KOI8-R/KOI8-R \
|
||||
ru_RU.UTF-8/UTF-8 \
|
||||
ru_RU/ISO-8859-5 \
|
||||
ru_UA.UTF-8/UTF-8 \
|
||||
ru_UA/KOI8-U \
|
||||
rw_RW/UTF-8 \
|
||||
sa_IN/UTF-8 \
|
||||
sah_RU/UTF-8 \
|
||||
sat_IN/UTF-8 \
|
||||
sc_IT/UTF-8 \
|
||||
sd_IN/UTF-8 \
|
||||
sd_IN@devanagari/UTF-8 \
|
||||
se_NO/UTF-8 \
|
||||
sgs_LT/UTF-8 \
|
||||
shn_MM/UTF-8 \
|
||||
shs_CA/UTF-8 \
|
||||
si_LK/UTF-8 \
|
||||
sid_ET/UTF-8 \
|
||||
sk_SK.UTF-8/UTF-8 \
|
||||
sk_SK/ISO-8859-2 \
|
||||
sl_SI.UTF-8/UTF-8 \
|
||||
sl_SI/ISO-8859-2 \
|
||||
sm_WS/UTF-8 \
|
||||
so_DJ.UTF-8/UTF-8 \
|
||||
so_DJ/ISO-8859-1 \
|
||||
so_ET/UTF-8 \
|
||||
so_KE.UTF-8/UTF-8 \
|
||||
so_KE/ISO-8859-1 \
|
||||
so_SO.UTF-8/UTF-8 \
|
||||
so_SO/ISO-8859-1 \
|
||||
sq_AL.UTF-8/UTF-8 \
|
||||
sq_AL/ISO-8859-1 \
|
||||
sq_MK/UTF-8 \
|
||||
sr_ME/UTF-8 \
|
||||
sr_RS/UTF-8 \
|
||||
sr_RS@latin/UTF-8 \
|
||||
ss_ZA/UTF-8 \
|
||||
st_ZA.UTF-8/UTF-8 \
|
||||
st_ZA/ISO-8859-1 \
|
||||
sv_FI.UTF-8/UTF-8 \
|
||||
sv_FI/ISO-8859-1 \
|
||||
sv_FI@euro/ISO-8859-15 \
|
||||
sv_SE.UTF-8/UTF-8 \
|
||||
sv_SE/ISO-8859-1 \
|
||||
sv_SE.ISO-8859-15/ISO-8859-15 \
|
||||
sw_KE/UTF-8 \
|
||||
sw_TZ/UTF-8 \
|
||||
szl_PL/UTF-8 \
|
||||
ta_IN/UTF-8 \
|
||||
ta_LK/UTF-8 \
|
||||
tcy_IN.UTF-8/UTF-8 \
|
||||
te_IN/UTF-8 \
|
||||
tg_TJ.UTF-8/UTF-8 \
|
||||
tg_TJ/KOI8-T \
|
||||
th_TH.UTF-8/UTF-8 \
|
||||
th_TH/TIS-620 \
|
||||
the_NP/UTF-8 \
|
||||
ti_ER/UTF-8 \
|
||||
ti_ET/UTF-8 \
|
||||
tig_ER/UTF-8 \
|
||||
tk_TM/UTF-8 \
|
||||
tl_PH.UTF-8/UTF-8 \
|
||||
tl_PH/ISO-8859-1 \
|
||||
tn_ZA/UTF-8 \
|
||||
to_TO/UTF-8 \
|
||||
tpi_PG/UTF-8 \
|
||||
tr_CY.UTF-8/UTF-8 \
|
||||
tr_CY/ISO-8859-9 \
|
||||
tr_TR.UTF-8/UTF-8 \
|
||||
tr_TR/ISO-8859-9 \
|
||||
ts_ZA/UTF-8 \
|
||||
tt_RU/UTF-8 \
|
||||
tt_RU@iqtelif/UTF-8 \
|
||||
ug_CN/UTF-8 \
|
||||
uk_UA.UTF-8/UTF-8 \
|
||||
uk_UA/KOI8-U \
|
||||
unm_US/UTF-8 \
|
||||
ur_IN/UTF-8 \
|
||||
ur_PK/UTF-8 \
|
||||
uz_UZ.UTF-8/UTF-8 \
|
||||
uz_UZ/ISO-8859-1 \
|
||||
uz_UZ@cyrillic/UTF-8 \
|
||||
ve_ZA/UTF-8 \
|
||||
vi_VN/UTF-8 \
|
||||
wa_BE/ISO-8859-1 \
|
||||
wa_BE@euro/ISO-8859-15 \
|
||||
wa_BE.UTF-8/UTF-8 \
|
||||
wae_CH/UTF-8 \
|
||||
wal_ET/UTF-8 \
|
||||
wo_SN/UTF-8 \
|
||||
xh_ZA.UTF-8/UTF-8 \
|
||||
xh_ZA/ISO-8859-1 \
|
||||
yi_US.UTF-8/UTF-8 \
|
||||
yi_US/CP1255 \
|
||||
yo_NG/UTF-8 \
|
||||
yue_HK/UTF-8 \
|
||||
yuw_PG/UTF-8 \
|
||||
zh_CN.GB18030/GB18030 \
|
||||
zh_CN.GBK/GBK \
|
||||
zh_CN.UTF-8/UTF-8 \
|
||||
zh_CN/GB2312 \
|
||||
zh_HK.UTF-8/UTF-8 \
|
||||
zh_HK/BIG5-HKSCS \
|
||||
zh_SG.UTF-8/UTF-8 \
|
||||
zh_SG.GBK/GBK \
|
||||
zh_SG/GB2312 \
|
||||
zh_TW.EUC-TW/EUC-TW \
|
||||
zh_TW.UTF-8/UTF-8 \
|
||||
zh_TW/BIG5 \
|
||||
zu_ZA.UTF-8/UTF-8 \
|
||||
zu_ZA/ISO-8859-1 \
|
||||
@ -1,862 +0,0 @@
|
||||
#define _GNU_SOURCE
|
||||
#include <assert.h>
|
||||
#include <dirent.h>
|
||||
#include <errno.h>
|
||||
#include <fcntl.h>
|
||||
#include <locale.h>
|
||||
#include <stdarg.h>
|
||||
#include <stdbool.h>
|
||||
#include <stdio.h>
|
||||
#include <stdlib.h>
|
||||
#include <getopt.h>
|
||||
#include <string.h>
|
||||
#include <sys/mman.h>
|
||||
#include <sys/stat.h>
|
||||
#include <unistd.h>
|
||||
#include "../locale/hashval.h"
|
||||
#define __LC_LAST 13
|
||||
#include "../locale/locarchive.h"
|
||||
#include "../crypt/md5.h"
|
||||
|
||||
const char *alias_file = DATADIR "/locale/locale.alias";
|
||||
const char *locar_file = PREFIX "/lib/locale/locale-archive";
|
||||
const char *tmpl_file = PREFIX "/lib/locale/locale-archive.tmpl";
|
||||
const char *loc_path = PREFIX "/lib/locale/";
|
||||
/* Flags set by `--verbose` option. */
|
||||
int be_quiet = 1;
|
||||
int verbose = 0;
|
||||
int max_locarchive_open_retry = 10;
|
||||
const char *output_prefix;
|
||||
|
||||
/* Endianness should have been taken care of by localedef. We don't need to do
|
||||
additional swapping. We need this variable exported however, since
|
||||
locarchive.c uses it to determine if it needs to swap endianness of a value
|
||||
before writing to or reading from the archive. */
|
||||
bool swap_endianness_p = false;
|
||||
|
||||
static const char *locnames[] =
|
||||
{
|
||||
#define DEFINE_CATEGORY(category, category_name, items, a) \
|
||||
[category] = category_name,
|
||||
#include "../locale/categories.def"
|
||||
#undef DEFINE_CATEGORY
|
||||
};
|
||||
|
||||
static int
|
||||
is_prime (unsigned long candidate)
|
||||
{
|
||||
/* No even number and none less than 10 will be passed here. */
|
||||
unsigned long int divn = 3;
|
||||
unsigned long int sq = divn * divn;
|
||||
|
||||
while (sq < candidate && candidate % divn != 0)
|
||||
{
|
||||
++divn;
|
||||
sq += 4 * divn;
|
||||
++divn;
|
||||
}
|
||||
|
||||
return candidate % divn != 0;
|
||||
}
|
||||
|
||||
unsigned long
|
||||
next_prime (unsigned long seed)
|
||||
{
|
||||
/* Make it definitely odd. */
|
||||
seed |= 1;
|
||||
|
||||
while (!is_prime (seed))
|
||||
seed += 2;
|
||||
|
||||
return seed;
|
||||
}
|
||||
|
||||
void
|
||||
error (int status, int errnum, const char *message, ...)
|
||||
{
|
||||
va_list args;
|
||||
|
||||
va_start (args, message);
|
||||
fflush (stdout);
|
||||
fprintf (stderr, "%s: ", program_invocation_name);
|
||||
vfprintf (stderr, message, args);
|
||||
va_end (args);
|
||||
if (errnum)
|
||||
fprintf (stderr, ": %s", strerror (errnum));
|
||||
putc ('\n', stderr);
|
||||
fflush (stderr);
|
||||
if (status)
|
||||
exit (errnum == EROFS ? 0 : status);
|
||||
}
|
||||
|
||||
void *
|
||||
xmalloc (size_t size)
|
||||
{
|
||||
void *p = malloc (size);
|
||||
if (p == NULL)
|
||||
error (EXIT_FAILURE, errno, "could not allocate %zd bytes of memory", size);
|
||||
return p;
|
||||
}
|
||||
|
||||
static void
|
||||
open_tmpl_archive (struct locarhandle *ah)
|
||||
{
|
||||
struct stat64 st;
|
||||
int fd;
|
||||
struct locarhead head;
|
||||
const char *archivefname = ah->fname == NULL ? tmpl_file : ah->fname;
|
||||
|
||||
/* Open the archive. We must have exclusive write access. */
|
||||
fd = open64 (archivefname, O_RDONLY);
|
||||
if (fd == -1)
|
||||
error (EXIT_FAILURE, errno, "cannot open locale archive template file \"%s\"",
|
||||
archivefname);
|
||||
|
||||
if (fstat64 (fd, &st) < 0)
|
||||
error (EXIT_FAILURE, errno, "cannot stat locale archive template file \"%s\"",
|
||||
archivefname);
|
||||
|
||||
/* Read the header. */
|
||||
if (TEMP_FAILURE_RETRY (read (fd, &head, sizeof (head))) != sizeof (head))
|
||||
error (EXIT_FAILURE, errno, "cannot read archive header");
|
||||
|
||||
ah->fd = fd;
|
||||
ah->mmaped = (head.sumhash_offset
|
||||
+ head.sumhash_size * sizeof (struct sumhashent));
|
||||
if (ah->mmaped > (unsigned long) st.st_size)
|
||||
error (EXIT_FAILURE, 0, "locale archive template file truncated");
|
||||
ah->mmaped = st.st_size;
|
||||
ah->reserved = st.st_size;
|
||||
|
||||
/* Now we know how large the administrative information part is.
|
||||
Map all of it. */
|
||||
ah->addr = mmap64 (NULL, ah->mmaped, PROT_READ, MAP_SHARED, fd, 0);
|
||||
if (ah->addr == MAP_FAILED)
|
||||
error (EXIT_FAILURE, errno, "cannot map archive header");
|
||||
}
|
||||
|
||||
/* Open the locale archive. */
|
||||
extern void open_archive (struct locarhandle *ah, bool readonly);
|
||||
|
||||
/* Close the locale archive. */
|
||||
extern void close_archive (struct locarhandle *ah);
|
||||
|
||||
/* Add given locale data to the archive. */
|
||||
extern int add_locale_to_archive (struct locarhandle *ah, const char *name,
|
||||
locale_data_t data, bool replace);
|
||||
|
||||
extern void add_alias (struct locarhandle *ah, const char *alias,
|
||||
bool replace, const char *oldname,
|
||||
uint32_t *locrec_offset_p);
|
||||
|
||||
extern struct namehashent *
|
||||
insert_name (struct locarhandle *ah,
|
||||
const char *name, size_t name_len, bool replace);
|
||||
|
||||
struct nameent
|
||||
{
|
||||
char *name;
|
||||
struct locrecent *locrec;
|
||||
};
|
||||
|
||||
struct dataent
|
||||
{
|
||||
const unsigned char *sum;
|
||||
uint32_t file_offset;
|
||||
};
|
||||
|
||||
static int
|
||||
nameentcmp (const void *a, const void *b)
|
||||
{
|
||||
struct locrecent *la = ((const struct nameent *) a)->locrec;
|
||||
struct locrecent *lb = ((const struct nameent *) b)->locrec;
|
||||
uint32_t start_a = -1, end_a = 0;
|
||||
uint32_t start_b = -1, end_b = 0;
|
||||
int cnt;
|
||||
|
||||
for (cnt = 0; cnt < __LC_LAST; ++cnt)
|
||||
if (cnt != LC_ALL)
|
||||
{
|
||||
if (la->record[cnt].offset < start_a)
|
||||
start_a = la->record[cnt].offset;
|
||||
if (la->record[cnt].offset + la->record[cnt].len > end_a)
|
||||
end_a = la->record[cnt].offset + la->record[cnt].len;
|
||||
}
|
||||
assert (start_a != (uint32_t)-1);
|
||||
assert (end_a != 0);
|
||||
|
||||
for (cnt = 0; cnt < __LC_LAST; ++cnt)
|
||||
if (cnt != LC_ALL)
|
||||
{
|
||||
if (lb->record[cnt].offset < start_b)
|
||||
start_b = lb->record[cnt].offset;
|
||||
if (lb->record[cnt].offset + lb->record[cnt].len > end_b)
|
||||
end_b = lb->record[cnt].offset + lb->record[cnt].len;
|
||||
}
|
||||
assert (start_b != (uint32_t)-1);
|
||||
assert (end_b != 0);
|
||||
|
||||
if (start_a != start_b)
|
||||
return (int)start_a - (int)start_b;
|
||||
return (int)end_a - (int)end_b;
|
||||
}
|
||||
|
||||
static int
|
||||
dataentcmp (const void *a, const void *b)
|
||||
{
|
||||
if (((const struct dataent *) a)->file_offset
|
||||
< ((const struct dataent *) b)->file_offset)
|
||||
return -1;
|
||||
|
||||
if (((const struct dataent *) a)->file_offset
|
||||
> ((const struct dataent *) b)->file_offset)
|
||||
return 1;
|
||||
|
||||
return 0;
|
||||
}
|
||||
|
||||
static int
|
||||
sumsearchfn (const void *key, const void *ent)
|
||||
{
|
||||
uint32_t keyn = *(uint32_t *)key;
|
||||
uint32_t entn = ((struct dataent *)ent)->file_offset;
|
||||
|
||||
if (keyn < entn)
|
||||
return -1;
|
||||
if (keyn > entn)
|
||||
return 1;
|
||||
return 0;
|
||||
}
|
||||
|
||||
static void
|
||||
compute_data (struct locarhandle *ah, struct nameent *name, size_t sumused,
|
||||
struct dataent *files, locale_data_t data)
|
||||
{
|
||||
int cnt;
|
||||
struct locrecent *locrec = name->locrec;
|
||||
struct dataent *file;
|
||||
data[LC_ALL].addr = ((char *) ah->addr) + locrec->record[LC_ALL].offset;
|
||||
data[LC_ALL].size = locrec->record[LC_ALL].len;
|
||||
for (cnt = 0; cnt < __LC_LAST; ++cnt)
|
||||
if (cnt != LC_ALL)
|
||||
{
|
||||
data[cnt].addr = ((char *) ah->addr) + locrec->record[cnt].offset;
|
||||
data[cnt].size = locrec->record[cnt].len;
|
||||
if (data[cnt].addr >= data[LC_ALL].addr
|
||||
&& data[cnt].addr + data[cnt].size
|
||||
<= data[LC_ALL].addr + data[LC_ALL].size)
|
||||
__md5_buffer (data[cnt].addr, data[cnt].size, data[cnt].sum);
|
||||
else
|
||||
{
|
||||
file = bsearch (&locrec->record[cnt].offset, files, sumused,
|
||||
sizeof (*files), sumsearchfn);
|
||||
if (file == NULL)
|
||||
error (EXIT_FAILURE, 0, "inconsistent template file");
|
||||
memcpy (data[cnt].sum, file->sum, sizeof (data[cnt].sum));
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
static int
|
||||
fill_archive (struct locarhandle *tmpl_ah,
|
||||
const char *fname,
|
||||
size_t install_langs_count, char *install_langs_list[],
|
||||
size_t nlist, char *list[],
|
||||
const char *primary)
|
||||
{
|
||||
struct locarhandle ah;
|
||||
struct locarhead *head;
|
||||
int result = 0;
|
||||
struct nameent *names;
|
||||
struct namehashent *namehashtab;
|
||||
size_t cnt, used;
|
||||
struct dataent *files;
|
||||
struct sumhashent *sumhashtab;
|
||||
size_t sumused;
|
||||
struct locrecent *primary_locrec = NULL;
|
||||
struct nameent *primary_nameent = NULL;
|
||||
|
||||
head = tmpl_ah->addr;
|
||||
names = (struct nameent *) malloc (head->namehash_used
|
||||
* sizeof (struct nameent));
|
||||
files = (struct dataent *) malloc (head->sumhash_used
|
||||
* sizeof (struct dataent));
|
||||
if (names == NULL || files == NULL)
|
||||
error (EXIT_FAILURE, errno, "could not allocate tables");
|
||||
|
||||
namehashtab = (struct namehashent *) ((char *) tmpl_ah->addr
|
||||
+ head->namehash_offset);
|
||||
sumhashtab = (struct sumhashent *) ((char *) tmpl_ah->addr
|
||||
+ head->sumhash_offset);
|
||||
|
||||
for (cnt = used = 0; cnt < head->namehash_size; ++cnt)
|
||||
if (namehashtab[cnt].locrec_offset != 0)
|
||||
{
|
||||
char * name;
|
||||
int i;
|
||||
assert (used < head->namehash_used);
|
||||
name = tmpl_ah->addr + namehashtab[cnt].name_offset;
|
||||
if (install_langs_count == 0)
|
||||
{
|
||||
/* Always intstall the entry. */
|
||||
names[used].name = name;
|
||||
names[used++].locrec
|
||||
= (struct locrecent *) ((char *) tmpl_ah->addr +
|
||||
namehashtab[cnt].locrec_offset);
|
||||
}
|
||||
else
|
||||
{
|
||||
/* Only install the entry if the user asked for it via
|
||||
--install-langs. */
|
||||
for (i = 0; i < install_langs_count; i++)
|
||||
{
|
||||
/* Add one for "_" and one for the null terminator. */
|
||||
size_t len = strlen (install_langs_list[i]) + 2;
|
||||
char *install_lang = (char *)xmalloc (len);
|
||||
strcpy (install_lang, install_langs_list[i]);
|
||||
if (strchr (install_lang, '_') == NULL)
|
||||
strcat (install_lang, "_");
|
||||
if (strncmp (name, install_lang, strlen (install_lang)) == 0)
|
||||
{
|
||||
names[used].name = name;
|
||||
names[used++].locrec
|
||||
= (struct locrecent *) ((char *)tmpl_ah->addr
|
||||
+ namehashtab[cnt].locrec_offset);
|
||||
}
|
||||
free (install_lang);
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
/* Sort the names. */
|
||||
qsort (names, used, sizeof (struct nameent), nameentcmp);
|
||||
|
||||
for (cnt = sumused = 0; cnt < head->sumhash_size; ++cnt)
|
||||
if (sumhashtab[cnt].file_offset != 0)
|
||||
{
|
||||
assert (sumused < head->sumhash_used);
|
||||
files[sumused].sum = (const unsigned char *) sumhashtab[cnt].sum;
|
||||
files[sumused++].file_offset = sumhashtab[cnt].file_offset;
|
||||
}
|
||||
|
||||
/* Sort by file locations. */
|
||||
qsort (files, sumused, sizeof (struct dataent), dataentcmp);
|
||||
|
||||
/* Open the archive. This call never returns if we cannot
|
||||
successfully open the archive. */
|
||||
ah.fname = NULL;
|
||||
if (fname != NULL)
|
||||
ah.fname = fname;
|
||||
open_archive (&ah, false);
|
||||
|
||||
if (primary != NULL)
|
||||
{
|
||||
for (cnt = 0; cnt < used; ++cnt)
|
||||
if (strcmp (names[cnt].name, primary) == 0)
|
||||
break;
|
||||
if (cnt < used)
|
||||
{
|
||||
locale_data_t data;
|
||||
|
||||
compute_data (tmpl_ah, &names[cnt], sumused, files, data);
|
||||
result |= add_locale_to_archive (&ah, primary, data, 0);
|
||||
primary_locrec = names[cnt].locrec;
|
||||
primary_nameent = &names[cnt];
|
||||
}
|
||||
}
|
||||
|
||||
for (cnt = 0; cnt < used; ++cnt)
|
||||
if (&names[cnt] == primary_nameent)
|
||||
continue;
|
||||
else if ((cnt > 0 && names[cnt - 1].locrec == names[cnt].locrec)
|
||||
|| names[cnt].locrec == primary_locrec)
|
||||
{
|
||||
const char *oldname;
|
||||
struct namehashent *namehashent;
|
||||
uint32_t locrec_offset;
|
||||
|
||||
if (names[cnt].locrec == primary_locrec)
|
||||
oldname = primary;
|
||||
else
|
||||
oldname = names[cnt - 1].name;
|
||||
namehashent = insert_name (&ah, oldname, strlen (oldname), true);
|
||||
assert (namehashent->name_offset != 0);
|
||||
assert (namehashent->locrec_offset != 0);
|
||||
locrec_offset = namehashent->locrec_offset;
|
||||
add_alias (&ah, names[cnt].name, 0, oldname, &locrec_offset);
|
||||
}
|
||||
else
|
||||
{
|
||||
locale_data_t data;
|
||||
|
||||
compute_data (tmpl_ah, &names[cnt], sumused, files, data);
|
||||
result |= add_locale_to_archive (&ah, names[cnt].name, data, 0);
|
||||
}
|
||||
|
||||
while (nlist-- > 0)
|
||||
{
|
||||
const char *fname = *list++;
|
||||
size_t fnamelen = strlen (fname);
|
||||
struct stat64 st;
|
||||
DIR *dirp;
|
||||
struct dirent64 *d;
|
||||
int seen;
|
||||
locale_data_t data;
|
||||
int cnt;
|
||||
|
||||
/* First see whether this really is a directory and whether it
|
||||
contains all the require locale category files. */
|
||||
if (stat64 (fname, &st) < 0)
|
||||
{
|
||||
error (0, 0, "stat of \"%s\" failed: %s: ignored", fname,
|
||||
strerror (errno));
|
||||
continue;
|
||||
}
|
||||
if (!S_ISDIR (st.st_mode))
|
||||
{
|
||||
error (0, 0, "\"%s\" is no directory; ignored", fname);
|
||||
continue;
|
||||
}
|
||||
|
||||
dirp = opendir (fname);
|
||||
if (dirp == NULL)
|
||||
{
|
||||
error (0, 0, "cannot open directory \"%s\": %s: ignored",
|
||||
fname, strerror (errno));
|
||||
continue;
|
||||
}
|
||||
|
||||
seen = 0;
|
||||
while ((d = readdir64 (dirp)) != NULL)
|
||||
{
|
||||
for (cnt = 0; cnt < __LC_LAST; ++cnt)
|
||||
if (cnt != LC_ALL)
|
||||
if (strcmp (d->d_name, locnames[cnt]) == 0)
|
||||
{
|
||||
unsigned char d_type;
|
||||
|
||||
/* We have an object of the required name. If it's
|
||||
a directory we have to look at a file with the
|
||||
prefix "SYS_". Otherwise we have found what we
|
||||
are looking for. */
|
||||
#ifdef _DIRENT_HAVE_D_TYPE
|
||||
d_type = d->d_type;
|
||||
|
||||
if (d_type != DT_REG)
|
||||
#endif
|
||||
{
|
||||
char fullname[fnamelen + 2 * strlen (d->d_name) + 7];
|
||||
|
||||
#ifdef _DIRENT_HAVE_D_TYPE
|
||||
if (d_type == DT_UNKNOWN || d_type == DT_LNK)
|
||||
#endif
|
||||
{
|
||||
strcpy (stpcpy (stpcpy (fullname, fname), "/"),
|
||||
d->d_name);
|
||||
|
||||
if (stat64 (fullname, &st) == -1)
|
||||
/* We cannot stat the file, ignore it. */
|
||||
break;
|
||||
|
||||
d_type = IFTODT (st.st_mode);
|
||||
}
|
||||
|
||||
if (d_type == DT_DIR)
|
||||
{
|
||||
/* We have to do more tests. The file is a
|
||||
directory and it therefore must contain a
|
||||
regular file with the same name except a
|
||||
"SYS_" prefix. */
|
||||
char *t = stpcpy (stpcpy (fullname, fname), "/");
|
||||
strcpy (stpcpy (stpcpy (t, d->d_name), "/SYS_"),
|
||||
d->d_name);
|
||||
|
||||
if (stat64 (fullname, &st) == -1)
|
||||
/* There is no SYS_* file or we cannot
|
||||
access it. */
|
||||
break;
|
||||
|
||||
d_type = IFTODT (st.st_mode);
|
||||
}
|
||||
}
|
||||
|
||||
/* If we found a regular file (eventually after
|
||||
following a symlink) we are successful. */
|
||||
if (d_type == DT_REG)
|
||||
++seen;
|
||||
break;
|
||||
}
|
||||
}
|
||||
|
||||
closedir (dirp);
|
||||
|
||||
if (seen != __LC_LAST - 1)
|
||||
{
|
||||
/* We don't have all locale category files. Ignore the name. */
|
||||
error (0, 0, "incomplete set of locale files in \"%s\"",
|
||||
fname);
|
||||
continue;
|
||||
}
|
||||
|
||||
/* Add the files to the archive. To do this we first compute
|
||||
sizes and the MD5 sums of all the files. */
|
||||
for (cnt = 0; cnt < __LC_LAST; ++cnt)
|
||||
if (cnt != LC_ALL)
|
||||
{
|
||||
char fullname[fnamelen + 2 * strlen (locnames[cnt]) + 7];
|
||||
int fd;
|
||||
|
||||
strcpy (stpcpy (stpcpy (fullname, fname), "/"), locnames[cnt]);
|
||||
fd = open64 (fullname, O_RDONLY);
|
||||
if (fd == -1 || fstat64 (fd, &st) == -1)
|
||||
{
|
||||
/* Cannot read the file. */
|
||||
if (fd != -1)
|
||||
close (fd);
|
||||
break;
|
||||
}
|
||||
|
||||
if (S_ISDIR (st.st_mode))
|
||||
{
|
||||
char *t;
|
||||
close (fd);
|
||||
t = stpcpy (stpcpy (fullname, fname), "/");
|
||||
strcpy (stpcpy (stpcpy (t, locnames[cnt]), "/SYS_"),
|
||||
locnames[cnt]);
|
||||
|
||||
fd = open64 (fullname, O_RDONLY);
|
||||
if (fd == -1 || fstat64 (fd, &st) == -1
|
||||
|| !S_ISREG (st.st_mode))
|
||||
{
|
||||
if (fd != -1)
|
||||
close (fd);
|
||||
break;
|
||||
}
|
||||
}
|
||||
|
||||
/* Map the file. */
|
||||
data[cnt].addr = mmap64 (NULL, st.st_size, PROT_READ, MAP_SHARED,
|
||||
fd, 0);
|
||||
if (data[cnt].addr == MAP_FAILED)
|
||||
{
|
||||
/* Cannot map it. */
|
||||
close (fd);
|
||||
break;
|
||||
}
|
||||
|
||||
data[cnt].size = st.st_size;
|
||||
__md5_buffer (data[cnt].addr, st.st_size, data[cnt].sum);
|
||||
|
||||
/* We don't need the file descriptor anymore. */
|
||||
close (fd);
|
||||
}
|
||||
|
||||
if (cnt != __LC_LAST)
|
||||
{
|
||||
while (cnt-- > 0)
|
||||
if (cnt != LC_ALL)
|
||||
munmap (data[cnt].addr, data[cnt].size);
|
||||
|
||||
error (0, 0, "cannot read all files in \"%s\": ignored", fname);
|
||||
|
||||
continue;
|
||||
}
|
||||
|
||||
result |= add_locale_to_archive (&ah, basename (fname), data, 0);
|
||||
|
||||
for (cnt = 0; cnt < __LC_LAST; ++cnt)
|
||||
if (cnt != LC_ALL)
|
||||
munmap (data[cnt].addr, data[cnt].size);
|
||||
}
|
||||
|
||||
/* We are done. */
|
||||
close_archive (&ah);
|
||||
|
||||
return result;
|
||||
}
|
||||
|
||||
void usage()
|
||||
{
|
||||
printf ("\
|
||||
Usage: build-locale-archive [OPTION]... [TEMPLATE-FILE] [ARCHIVE-FILE]\n\
|
||||
Builds a locale archive from a template file.\n\
|
||||
Options:\n\
|
||||
-h, --help Print this usage message.\n\
|
||||
-v, --verbose Verbose execution.\n\
|
||||
-l, --install-langs=LIST Only include locales given in LIST into the \n\
|
||||
locale archive. LIST is a colon separated list\n\
|
||||
of locale prefixes, for example \"de:en:ja\".\n\
|
||||
The special argument \"all\" means to install\n\
|
||||
all languages and it must be present by itself.\n\
|
||||
If \"all\" is present with any other language it\n\
|
||||
will be treated as the name of a locale.\n\
|
||||
If the --install-langs option is missing, all\n\
|
||||
locales are installed. The colon separated list\n\
|
||||
can contain any strings matching the beginning of\n\
|
||||
locale names.\n\
|
||||
If a string does not contain a \"_\", it is added.\n\
|
||||
Examples:\n\
|
||||
--install-langs=\"en\"\n\
|
||||
installs en_US, en_US.iso88591,\n\
|
||||
en_US.iso885915, en_US.utf8,\n\
|
||||
en_GB ...\n\
|
||||
--install-langs=\"en_US.utf8\"\n\
|
||||
installs only en_US.utf8.\n\
|
||||
--install-langs=\"ko\"\n\
|
||||
installs ko_KR, ko_KR.euckr,\n\
|
||||
ko_KR.utf8 but *not* kok_IN\n\
|
||||
because \"ko\" does not contain\n\
|
||||
\"_\" and it is silently added\n\
|
||||
--install-langs\"ko:kok\"\n\
|
||||
installs ko_KR, ko_KR.euckr,\n\
|
||||
ko_KR.utf8, kok_IN, and\n\
|
||||
kok_IN.utf8.\n\
|
||||
--install-langs=\"POSIX\" will\n\
|
||||
installs *no* locales at all\n\
|
||||
because POSIX matches none of\n\
|
||||
the locales. Actually, any string\n\
|
||||
matching nothing will do that.\n\
|
||||
POSIX and C will always be\n\
|
||||
available because they are\n\
|
||||
builtin.\n\
|
||||
Aliases are installed as well,\n\
|
||||
i.e. --install-langs=\"de\"\n\
|
||||
will install not only every locale starting with\n\
|
||||
\"de\" but also the aliases \"deutsch\"\n\
|
||||
and and \"german\" although the latter does not\n\
|
||||
start with \"de\".\n\
|
||||
\n\
|
||||
If the arguments TEMPLATE-FILE and ARCHIVE-FILE are not given the locations\n\
|
||||
where the glibc used expects these files are used by default.\n\
|
||||
");
|
||||
}
|
||||
|
||||
int main (int argc, char *argv[])
|
||||
{
|
||||
char path[4096];
|
||||
DIR *dirp;
|
||||
struct dirent64 *d;
|
||||
struct stat64 st;
|
||||
char *list[16384], *primary;
|
||||
char *lang;
|
||||
int install_langs_count = 0;
|
||||
int i;
|
||||
char *install_langs_arg, *ila_start;
|
||||
char **install_langs_list = NULL;
|
||||
unsigned int cnt = 0;
|
||||
struct locarhandle tmpl_ah;
|
||||
char *new_locar_fname = NULL;
|
||||
size_t loc_path_len = strlen (loc_path);
|
||||
|
||||
while (1)
|
||||
{
|
||||
int c;
|
||||
|
||||
static struct option long_options[] =
|
||||
{
|
||||
{"help", no_argument, 0, 'h'},
|
||||
{"verbose", no_argument, 0, 'v'},
|
||||
{"install-langs", required_argument, 0, 'l'},
|
||||
{0, 0, 0, 0}
|
||||
};
|
||||
/* getopt_long stores the option index here. */
|
||||
int option_index = 0;
|
||||
|
||||
c = getopt_long (argc, argv, "vhl:",
|
||||
long_options, &option_index);
|
||||
|
||||
/* Detect the end of the options. */
|
||||
if (c == -1)
|
||||
break;
|
||||
|
||||
switch (c)
|
||||
{
|
||||
case 0:
|
||||
printf ("unknown option %s", long_options[option_index].name);
|
||||
if (optarg)
|
||||
printf (" with arg %s", optarg);
|
||||
printf ("\n");
|
||||
usage ();
|
||||
exit (1);
|
||||
|
||||
case 'v':
|
||||
verbose = 1;
|
||||
be_quiet = 0;
|
||||
break;
|
||||
|
||||
case 'h':
|
||||
usage ();
|
||||
exit (0);
|
||||
|
||||
case 'l':
|
||||
install_langs_arg = ila_start = strdup (optarg);
|
||||
/* If the argument to --install-lang is "all", do
|
||||
not limit the list of languages to install and install
|
||||
them all. We do not support installing a single locale
|
||||
called "all". */
|
||||
#define MAGIC_INSTALL_ALL "all"
|
||||
if (install_langs_arg != NULL
|
||||
&& install_langs_arg[0] != '\0'
|
||||
&& !(strncmp(install_langs_arg, MAGIC_INSTALL_ALL,
|
||||
strlen(MAGIC_INSTALL_ALL)) == 0
|
||||
&& strlen (install_langs_arg) == 3))
|
||||
{
|
||||
/* Count the number of languages we will install. */
|
||||
while (true)
|
||||
{
|
||||
lang = strtok(install_langs_arg, ":;,");
|
||||
if (lang == NULL)
|
||||
break;
|
||||
install_langs_count++;
|
||||
install_langs_arg = NULL;
|
||||
}
|
||||
free (ila_start);
|
||||
|
||||
/* Reject an entire string made up of delimiters. */
|
||||
if (install_langs_count == 0)
|
||||
break;
|
||||
|
||||
/* Copy the list. */
|
||||
install_langs_list = (char **)xmalloc (sizeof(char *) * install_langs_count);
|
||||
install_langs_arg = ila_start = strdup (optarg);
|
||||
install_langs_count = 0;
|
||||
while (true)
|
||||
{
|
||||
lang = strtok(install_langs_arg, ":;,");
|
||||
if (lang == NULL)
|
||||
break;
|
||||
install_langs_list[install_langs_count] = lang;
|
||||
install_langs_count++;
|
||||
install_langs_arg = NULL;
|
||||
}
|
||||
}
|
||||
break;
|
||||
|
||||
case '?':
|
||||
/* getopt_long already printed an error message. */
|
||||
usage ();
|
||||
exit (0);
|
||||
|
||||
default:
|
||||
abort ();
|
||||
}
|
||||
}
|
||||
tmpl_ah.fname = NULL;
|
||||
if (optind < argc)
|
||||
tmpl_ah.fname = argv[optind];
|
||||
if (optind + 1 < argc)
|
||||
new_locar_fname = argv[optind + 1];
|
||||
if (verbose)
|
||||
{
|
||||
if (tmpl_ah.fname)
|
||||
printf("input archive file specified on command line: %s\n",
|
||||
tmpl_ah.fname);
|
||||
else
|
||||
printf("using default input archive file.\n");
|
||||
if (new_locar_fname)
|
||||
printf("output archive file specified on command line: %s\n",
|
||||
new_locar_fname);
|
||||
else
|
||||
printf("using default output archive file.\n");
|
||||
}
|
||||
|
||||
dirp = opendir (loc_path);
|
||||
if (dirp == NULL)
|
||||
error (EXIT_FAILURE, errno, "cannot open directory \"%s\"", loc_path);
|
||||
|
||||
open_tmpl_archive (&tmpl_ah);
|
||||
|
||||
if (new_locar_fname)
|
||||
unlink (new_locar_fname);
|
||||
else
|
||||
unlink (locar_file);
|
||||
primary = getenv ("LC_ALL");
|
||||
if (primary == NULL)
|
||||
primary = getenv ("LANG");
|
||||
if (primary != NULL)
|
||||
{
|
||||
if (strncmp (primary, "ja", 2) != 0
|
||||
&& strncmp (primary, "ko", 2) != 0
|
||||
&& strncmp (primary, "zh", 2) != 0)
|
||||
{
|
||||
char *ptr = malloc (strlen (primary) + strlen (".utf8") + 1), *p, *q;
|
||||
/* This leads to invalid locales sometimes:
|
||||
de_DE.iso885915@euro -> de_DE.utf8@euro */
|
||||
if (ptr != NULL)
|
||||
{
|
||||
p = ptr;
|
||||
q = primary;
|
||||
while (*q && *q != '.' && *q != '@')
|
||||
*p++ = *q++;
|
||||
if (*q == '.')
|
||||
while (*q && *q != '@')
|
||||
q++;
|
||||
p = stpcpy (p, ".utf8");
|
||||
strcpy (p, q);
|
||||
primary = ptr;
|
||||
}
|
||||
else
|
||||
primary = NULL;
|
||||
}
|
||||
}
|
||||
|
||||
memcpy (path, loc_path, loc_path_len);
|
||||
|
||||
while ((d = readdir64 (dirp)) != NULL)
|
||||
{
|
||||
if (strcmp (d->d_name, ".") == 0 || strcmp (d->d_name, "..") == 0)
|
||||
continue;
|
||||
if (strchr (d->d_name, '_') == NULL)
|
||||
continue;
|
||||
|
||||
size_t d_name_len = strlen (d->d_name);
|
||||
if (loc_path_len + d_name_len + 1 > sizeof (path))
|
||||
{
|
||||
error (0, 0, "too long filename \"%s\"", d->d_name);
|
||||
continue;
|
||||
}
|
||||
|
||||
memcpy (path + loc_path_len, d->d_name, d_name_len + 1);
|
||||
if (stat64 (path, &st) < 0)
|
||||
{
|
||||
error (0, errno, "cannot stat \"%s\"", path);
|
||||
continue;
|
||||
}
|
||||
if (! S_ISDIR (st.st_mode))
|
||||
continue;
|
||||
if (cnt == 16384)
|
||||
{
|
||||
error (0, 0, "too many directories in \"%s\"", loc_path);
|
||||
break;
|
||||
}
|
||||
list[cnt] = strdup (path);
|
||||
if (list[cnt] == NULL)
|
||||
{
|
||||
error (0, errno, "cannot add file to list \"%s\"", path);
|
||||
continue;
|
||||
}
|
||||
if (primary != NULL && cnt > 0 && strcmp (primary, d->d_name) == 0)
|
||||
{
|
||||
char *p = list[0];
|
||||
list[0] = list[cnt];
|
||||
list[cnt] = p;
|
||||
}
|
||||
cnt++;
|
||||
}
|
||||
closedir (dirp);
|
||||
/* Store the archive to the file specified as the second argument on the
|
||||
command line or the default locale archive. */
|
||||
fill_archive (&tmpl_ah, new_locar_fname,
|
||||
install_langs_count, install_langs_list,
|
||||
cnt, list, primary);
|
||||
close_archive (&tmpl_ah);
|
||||
truncate (tmpl_file, 0);
|
||||
if (install_langs_count > 0)
|
||||
{
|
||||
free (ila_start);
|
||||
free (install_langs_list);
|
||||
}
|
||||
char *tz_argv[] = { "/usr/sbin/tzdata-update", NULL };
|
||||
execve (tz_argv[0], (char *const *)tz_argv, (char *const *)&tz_argv[1]);
|
||||
exit (0);
|
||||
}
|
||||
@ -1,259 +0,0 @@
|
||||
From 97700a34f36721b11a754cf37a1cc40695ece1fd Mon Sep 17 00:00:00 2001
|
||||
From: "H.J. Lu" <hjl.tools@gmail.com>
|
||||
Date: Mon, 21 Jan 2019 11:23:59 -0800
|
||||
Subject: [PATCH] x86-64 memchr/wmemchr: Properly handle the length parameter
|
||||
[BZ# 24097]
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
On x32, the size_t parameter may be passed in the lower 32 bits of a
|
||||
64-bit register with the non-zero upper 32 bits. The string/memory
|
||||
functions written in assembly can only use the lower 32 bits of a
|
||||
64-bit register as length or must clear the upper 32 bits before using
|
||||
the full 64-bit register for length.
|
||||
|
||||
This pach fixes memchr/wmemchr for x32. Tested on x86-64 and x32. On
|
||||
x86-64, libc.so is the same with and withou the fix.
|
||||
|
||||
[BZ# 24097]
|
||||
CVE-2019-6488
|
||||
* sysdeps/x86_64/memchr.S: Use RDX_LP for length. Clear the
|
||||
upper 32 bits of RDX register.
|
||||
* sysdeps/x86_64/multiarch/memchr-avx2.S: Likewise.
|
||||
* sysdeps/x86_64/x32/Makefile (tests): Add tst-size_t-memchr and
|
||||
tst-size_t-wmemchr.
|
||||
* sysdeps/x86_64/x32/test-size_t.h: New file.
|
||||
* sysdeps/x86_64/x32/tst-size_t-memchr.c: Likewise.
|
||||
* sysdeps/x86_64/x32/tst-size_t-wmemchr.c: Likewise.
|
||||
---
|
||||
sysdeps/x86_64/memchr.S | 10 ++--
|
||||
sysdeps/x86_64/multiarch/memchr-avx2.S | 8 ++-
|
||||
sysdeps/x86_64/x32/Makefile | 8 +++
|
||||
sysdeps/x86_64/x32/test-size_t.h | 35 ++++++++++++
|
||||
sysdeps/x86_64/x32/tst-size_t-memchr.c | 72 +++++++++++++++++++++++++
|
||||
sysdeps/x86_64/x32/tst-size_t-wmemchr.c | 20 +++++++
|
||||
6 files changed, 148 insertions(+), 5 deletions(-)
|
||||
create mode 100644 sysdeps/x86_64/x32/test-size_t.h
|
||||
create mode 100644 sysdeps/x86_64/x32/tst-size_t-memchr.c
|
||||
create mode 100644 sysdeps/x86_64/x32/tst-size_t-wmemchr.c
|
||||
|
||||
Conflicts:
|
||||
ChangeLog
|
||||
(removed)
|
||||
NEWS
|
||||
(removed)
|
||||
|
||||
diff --git a/sysdeps/x86_64/memchr.S b/sysdeps/x86_64/memchr.S
|
||||
index feef5d4f..cb320257 100644
|
||||
--- a/sysdeps/x86_64/memchr.S
|
||||
+++ b/sysdeps/x86_64/memchr.S
|
||||
@@ -34,12 +34,16 @@ ENTRY(MEMCHR)
|
||||
mov %edi, %ecx
|
||||
|
||||
#ifdef USE_AS_WMEMCHR
|
||||
- test %rdx, %rdx
|
||||
+ test %RDX_LP, %RDX_LP
|
||||
jz L(return_null)
|
||||
- shl $2, %rdx
|
||||
+ shl $2, %RDX_LP
|
||||
#else
|
||||
+# ifdef __ILP32__
|
||||
+ /* Clear the upper 32 bits. */
|
||||
+ movl %edx, %edx
|
||||
+# endif
|
||||
punpcklbw %xmm1, %xmm1
|
||||
- test %rdx, %rdx
|
||||
+ test %RDX_LP, %RDX_LP
|
||||
jz L(return_null)
|
||||
punpcklbw %xmm1, %xmm1
|
||||
#endif
|
||||
diff --git a/sysdeps/x86_64/multiarch/memchr-avx2.S b/sysdeps/x86_64/multiarch/memchr-avx2.S
|
||||
index 5f5e7725..c81da19b 100644
|
||||
--- a/sysdeps/x86_64/multiarch/memchr-avx2.S
|
||||
+++ b/sysdeps/x86_64/multiarch/memchr-avx2.S
|
||||
@@ -40,16 +40,20 @@
|
||||
ENTRY (MEMCHR)
|
||||
# ifndef USE_AS_RAWMEMCHR
|
||||
/* Check for zero length. */
|
||||
- testq %rdx, %rdx
|
||||
+ test %RDX_LP, %RDX_LP
|
||||
jz L(null)
|
||||
# endif
|
||||
movl %edi, %ecx
|
||||
/* Broadcast CHAR to YMM0. */
|
||||
vmovd %esi, %xmm0
|
||||
# ifdef USE_AS_WMEMCHR
|
||||
- shl $2, %rdx
|
||||
+ shl $2, %RDX_LP
|
||||
vpbroadcastd %xmm0, %ymm0
|
||||
# else
|
||||
+# ifdef __ILP32__
|
||||
+ /* Clear the upper 32 bits. */
|
||||
+ movl %edx, %edx
|
||||
+# endif
|
||||
vpbroadcastb %xmm0, %ymm0
|
||||
# endif
|
||||
/* Check if we may cross page boundary with one vector load. */
|
||||
diff --git a/sysdeps/x86_64/x32/Makefile b/sysdeps/x86_64/x32/Makefile
|
||||
index f2ebc24f..7d528889 100644
|
||||
--- a/sysdeps/x86_64/x32/Makefile
|
||||
+++ b/sysdeps/x86_64/x32/Makefile
|
||||
@@ -4,3 +4,11 @@ ifeq ($(subdir),math)
|
||||
# 64-bit llround. Add -fno-builtin-lround to silence the compiler.
|
||||
CFLAGS-s_llround.c += -fno-builtin-lround
|
||||
endif
|
||||
+
|
||||
+ifeq ($(subdir),string)
|
||||
+tests += tst-size_t-memchr
|
||||
+endif
|
||||
+
|
||||
+ifeq ($(subdir),wcsmbs)
|
||||
+tests += tst-size_t-wmemchr
|
||||
+endif
|
||||
diff --git a/sysdeps/x86_64/x32/test-size_t.h b/sysdeps/x86_64/x32/test-size_t.h
|
||||
new file mode 100644
|
||||
index 00000000..78a94086
|
||||
--- /dev/null
|
||||
+++ b/sysdeps/x86_64/x32/test-size_t.h
|
||||
@@ -0,0 +1,35 @@
|
||||
+/* Test string/memory functions with size_t in the lower 32 bits of
|
||||
+ 64-bit register.
|
||||
+ Copyright (C) 2019 Free Software Foundation, Inc.
|
||||
+ This file is part of the GNU C Library.
|
||||
+
|
||||
+ The GNU C Library is free software; you can redistribute it and/or
|
||||
+ modify it under the terms of the GNU Lesser General Public
|
||||
+ License as published by the Free Software Foundation; either
|
||||
+ version 2.1 of the License, or (at your option) any later version.
|
||||
+
|
||||
+ The GNU C Library is distributed in the hope that it will be useful,
|
||||
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
||||
+ Lesser General Public License for more details.
|
||||
+
|
||||
+ You should have received a copy of the GNU Lesser General Public
|
||||
+ License along with the GNU C Library; if not, see
|
||||
+ <http://www.gnu.org/licenses/>. */
|
||||
+
|
||||
+#define TEST_MAIN
|
||||
+#include <string/test-string.h>
|
||||
+
|
||||
+/* On x32, parameter_t may be passed in a 64-bit register with the LEN
|
||||
+ field in the lower 32 bits. When the LEN field of 64-bit register
|
||||
+ is passed to string/memory function as the size_t parameter, only
|
||||
+ the lower 32 bits can be used. */
|
||||
+typedef struct
|
||||
+{
|
||||
+ union
|
||||
+ {
|
||||
+ size_t len;
|
||||
+ void (*fn) (void);
|
||||
+ };
|
||||
+ void *p;
|
||||
+} parameter_t;
|
||||
diff --git a/sysdeps/x86_64/x32/tst-size_t-memchr.c b/sysdeps/x86_64/x32/tst-size_t-memchr.c
|
||||
new file mode 100644
|
||||
index 00000000..29a3daf1
|
||||
--- /dev/null
|
||||
+++ b/sysdeps/x86_64/x32/tst-size_t-memchr.c
|
||||
@@ -0,0 +1,72 @@
|
||||
+/* Test memchr with size_t in the lower 32 bits of 64-bit register.
|
||||
+ Copyright (C) 2019 Free Software Foundation, Inc.
|
||||
+ This file is part of the GNU C Library.
|
||||
+
|
||||
+ The GNU C Library is free software; you can redistribute it and/or
|
||||
+ modify it under the terms of the GNU Lesser General Public
|
||||
+ License as published by the Free Software Foundation; either
|
||||
+ version 2.1 of the License, or (at your option) any later version.
|
||||
+
|
||||
+ The GNU C Library is distributed in the hope that it will be useful,
|
||||
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
||||
+ Lesser General Public License for more details.
|
||||
+
|
||||
+ You should have received a copy of the GNU Lesser General Public
|
||||
+ License along with the GNU C Library; if not, see
|
||||
+ <http://www.gnu.org/licenses/>. */
|
||||
+
|
||||
+#ifndef WIDE
|
||||
+# define TEST_NAME "memchr"
|
||||
+#else
|
||||
+# define TEST_NAME "wmemchr"
|
||||
+#endif /* WIDE */
|
||||
+#include "test-size_t.h"
|
||||
+
|
||||
+#ifndef WIDE
|
||||
+# define MEMCHR memchr
|
||||
+# define CHAR char
|
||||
+# define UCHAR unsigned char
|
||||
+#else
|
||||
+# include <wchar.h>
|
||||
+# define MEMCHR wmemchr
|
||||
+# define CHAR wchar_t
|
||||
+# define UCHAR wchar_t
|
||||
+#endif /* WIDE */
|
||||
+
|
||||
+IMPL (MEMCHR, 1)
|
||||
+
|
||||
+typedef CHAR * (*proto_t) (const CHAR*, int, size_t);
|
||||
+
|
||||
+static CHAR *
|
||||
+__attribute__ ((noinline, noclone))
|
||||
+do_memchr (parameter_t a, parameter_t b)
|
||||
+{
|
||||
+ return CALL (&b, a.p, (uintptr_t) b.p, a.len);
|
||||
+}
|
||||
+
|
||||
+static int
|
||||
+test_main (void)
|
||||
+{
|
||||
+ test_init ();
|
||||
+
|
||||
+ parameter_t src = { { page_size / sizeof (CHAR) }, buf2 };
|
||||
+ parameter_t c = { { 0 }, (void *) (uintptr_t) 0x12 };
|
||||
+
|
||||
+ int ret = 0;
|
||||
+ FOR_EACH_IMPL (impl, 0)
|
||||
+ {
|
||||
+ c.fn = impl->fn;
|
||||
+ CHAR *res = do_memchr (src, c);
|
||||
+ if (res)
|
||||
+ {
|
||||
+ error (0, 0, "Wrong result in function %s: %p != NULL",
|
||||
+ impl->name, res);
|
||||
+ ret = 1;
|
||||
+ }
|
||||
+ }
|
||||
+
|
||||
+ return ret ? EXIT_FAILURE : EXIT_SUCCESS;
|
||||
+}
|
||||
+
|
||||
+#include <support/test-driver.c>
|
||||
diff --git a/sysdeps/x86_64/x32/tst-size_t-wmemchr.c b/sysdeps/x86_64/x32/tst-size_t-wmemchr.c
|
||||
new file mode 100644
|
||||
index 00000000..877801d6
|
||||
--- /dev/null
|
||||
+++ b/sysdeps/x86_64/x32/tst-size_t-wmemchr.c
|
||||
@@ -0,0 +1,20 @@
|
||||
+/* Test wmemchr with size_t in the lower 32 bits of 64-bit register.
|
||||
+ Copyright (C) 2019 Free Software Foundation, Inc.
|
||||
+ This file is part of the GNU C Library.
|
||||
+
|
||||
+ The GNU C Library is free software; you can redistribute it and/or
|
||||
+ modify it under the terms of the GNU Lesser General Public
|
||||
+ License as published by the Free Software Foundation; either
|
||||
+ version 2.1 of the License, or (at your option) any later version.
|
||||
+
|
||||
+ The GNU C Library is distributed in the hope that it will be useful,
|
||||
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
||||
+ Lesser General Public License for more details.
|
||||
+
|
||||
+ You should have received a copy of the GNU Lesser General Public
|
||||
+ License along with the GNU C Library; if not, see
|
||||
+ <http://www.gnu.org/licenses/>. */
|
||||
+
|
||||
+#define WIDE 1
|
||||
+#include "tst-size_t-memchr.c"
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,41 +0,0 @@
|
||||
From ddf0992cf57a93200e0c782e2a94d0733a5a0b87 Mon Sep 17 00:00:00 2001
|
||||
From: Noah Goldstein <goldstein.w.n@gmail.com>
|
||||
Date: Sun, 9 Jan 2022 16:02:21 -0600
|
||||
Subject: [PATCH] x86: Fix __wcsncmp_avx2 in strcmp-avx2.S [BZ# 28755]
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
Fixes [BZ# 28755] for wcsncmp by redirecting length >= 2^56 to
|
||||
__wcscmp_avx2. For x86_64 this covers the entire address range so any
|
||||
length larger could not possibly be used to bound `s1` or `s2`.
|
||||
|
||||
test-strcmp, test-strncmp, test-wcscmp, and test-wcsncmp all pass.
|
||||
|
||||
Signed-off-by: Noah Goldstein <goldstein.w.n@gmail.com>
|
||||
---
|
||||
sysdeps/x86_64/multiarch/strcmp-avx2.S | 10 ++++++++++
|
||||
1 file changed, 10 insertions(+)
|
||||
|
||||
diff --git a/sysdeps/x86_64/multiarch/strcmp-avx2.S b/sysdeps/x86_64/multiarch/strcmp-avx2.S
|
||||
index 156c1949..8fb8eedc 100644
|
||||
--- a/sysdeps/x86_64/multiarch/strcmp-avx2.S
|
||||
+++ b/sysdeps/x86_64/multiarch/strcmp-avx2.S
|
||||
@@ -83,6 +83,16 @@ ENTRY (STRCMP)
|
||||
je L(char0)
|
||||
jb L(zero)
|
||||
# ifdef USE_AS_WCSCMP
|
||||
+# ifndef __ILP32__
|
||||
+ movq %rdx, %rcx
|
||||
+ /* Check if length could overflow when multiplied by
|
||||
+ sizeof(wchar_t). Checking top 8 bits will cover all potential
|
||||
+ overflow cases as well as redirect cases where its impossible to
|
||||
+ length to bound a valid memory region. In these cases just use
|
||||
+ 'wcscmp'. */
|
||||
+ shrq $56, %rcx
|
||||
+ jnz __wcscmp_avx2
|
||||
+# endif
|
||||
/* Convert units: from wide to byte char. */
|
||||
shl $2, %RDX_LP
|
||||
# endif
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,263 +0,0 @@
|
||||
From 23102686ec67b856a2d4fd25ddaa1c0b8d175c4f Mon Sep 17 00:00:00 2001
|
||||
From: Noah Goldstein <goldstein.w.n@gmail.com>
|
||||
Date: Fri, 15 Apr 2022 12:28:01 -0500
|
||||
Subject: [PATCH] x86: Cleanup page cross code in memcmp-avx2-movbe.S
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
Old code was both inefficient and wasted code size. New code (-62
|
||||
bytes) and comparable or better performance in the page cross case.
|
||||
|
||||
geometric_mean(N=20) of page cross cases New / Original: 0.960
|
||||
|
||||
size, align0, align1, ret, New Time/Old Time
|
||||
1, 4095, 0, 0, 1.001
|
||||
1, 4095, 0, 1, 0.999
|
||||
1, 4095, 0, -1, 1.0
|
||||
2, 4094, 0, 0, 1.0
|
||||
2, 4094, 0, 1, 1.0
|
||||
2, 4094, 0, -1, 1.0
|
||||
3, 4093, 0, 0, 1.0
|
||||
3, 4093, 0, 1, 1.0
|
||||
3, 4093, 0, -1, 1.0
|
||||
4, 4092, 0, 0, 0.987
|
||||
4, 4092, 0, 1, 1.0
|
||||
4, 4092, 0, -1, 1.0
|
||||
5, 4091, 0, 0, 0.984
|
||||
5, 4091, 0, 1, 1.002
|
||||
5, 4091, 0, -1, 1.005
|
||||
6, 4090, 0, 0, 0.993
|
||||
6, 4090, 0, 1, 1.001
|
||||
6, 4090, 0, -1, 1.003
|
||||
7, 4089, 0, 0, 0.991
|
||||
7, 4089, 0, 1, 1.0
|
||||
7, 4089, 0, -1, 1.001
|
||||
8, 4088, 0, 0, 0.875
|
||||
8, 4088, 0, 1, 0.881
|
||||
8, 4088, 0, -1, 0.888
|
||||
9, 4087, 0, 0, 0.872
|
||||
9, 4087, 0, 1, 0.879
|
||||
9, 4087, 0, -1, 0.883
|
||||
10, 4086, 0, 0, 0.878
|
||||
10, 4086, 0, 1, 0.886
|
||||
10, 4086, 0, -1, 0.873
|
||||
11, 4085, 0, 0, 0.878
|
||||
11, 4085, 0, 1, 0.881
|
||||
11, 4085, 0, -1, 0.879
|
||||
12, 4084, 0, 0, 0.873
|
||||
12, 4084, 0, 1, 0.889
|
||||
12, 4084, 0, -1, 0.875
|
||||
13, 4083, 0, 0, 0.873
|
||||
13, 4083, 0, 1, 0.863
|
||||
13, 4083, 0, -1, 0.863
|
||||
14, 4082, 0, 0, 0.838
|
||||
14, 4082, 0, 1, 0.869
|
||||
14, 4082, 0, -1, 0.877
|
||||
15, 4081, 0, 0, 0.841
|
||||
15, 4081, 0, 1, 0.869
|
||||
15, 4081, 0, -1, 0.876
|
||||
16, 4080, 0, 0, 0.988
|
||||
16, 4080, 0, 1, 0.99
|
||||
16, 4080, 0, -1, 0.989
|
||||
17, 4079, 0, 0, 0.978
|
||||
17, 4079, 0, 1, 0.981
|
||||
17, 4079, 0, -1, 0.98
|
||||
18, 4078, 0, 0, 0.981
|
||||
18, 4078, 0, 1, 0.98
|
||||
18, 4078, 0, -1, 0.985
|
||||
19, 4077, 0, 0, 0.977
|
||||
19, 4077, 0, 1, 0.979
|
||||
19, 4077, 0, -1, 0.986
|
||||
20, 4076, 0, 0, 0.977
|
||||
20, 4076, 0, 1, 0.986
|
||||
20, 4076, 0, -1, 0.984
|
||||
21, 4075, 0, 0, 0.977
|
||||
21, 4075, 0, 1, 0.983
|
||||
21, 4075, 0, -1, 0.988
|
||||
22, 4074, 0, 0, 0.983
|
||||
22, 4074, 0, 1, 0.994
|
||||
22, 4074, 0, -1, 0.993
|
||||
23, 4073, 0, 0, 0.98
|
||||
23, 4073, 0, 1, 0.992
|
||||
23, 4073, 0, -1, 0.995
|
||||
24, 4072, 0, 0, 0.989
|
||||
24, 4072, 0, 1, 0.989
|
||||
24, 4072, 0, -1, 0.991
|
||||
25, 4071, 0, 0, 0.99
|
||||
25, 4071, 0, 1, 0.999
|
||||
25, 4071, 0, -1, 0.996
|
||||
26, 4070, 0, 0, 0.993
|
||||
26, 4070, 0, 1, 0.995
|
||||
26, 4070, 0, -1, 0.998
|
||||
27, 4069, 0, 0, 0.993
|
||||
27, 4069, 0, 1, 0.999
|
||||
27, 4069, 0, -1, 1.0
|
||||
28, 4068, 0, 0, 0.997
|
||||
28, 4068, 0, 1, 1.0
|
||||
28, 4068, 0, -1, 0.999
|
||||
29, 4067, 0, 0, 0.996
|
||||
29, 4067, 0, 1, 0.999
|
||||
29, 4067, 0, -1, 0.999
|
||||
30, 4066, 0, 0, 0.991
|
||||
30, 4066, 0, 1, 1.001
|
||||
30, 4066, 0, -1, 0.999
|
||||
31, 4065, 0, 0, 0.988
|
||||
31, 4065, 0, 1, 0.998
|
||||
31, 4065, 0, -1, 0.998
|
||||
Reviewed-by: H.J. Lu <hjl.tools@gmail.com>
|
||||
---
|
||||
sysdeps/x86_64/multiarch/memcmp-avx2-movbe.S | 98 ++++++++++++--------
|
||||
1 file changed, 61 insertions(+), 37 deletions(-)
|
||||
|
||||
diff --git a/sysdeps/x86_64/multiarch/memcmp-avx2-movbe.S b/sysdeps/x86_64/multiarch/memcmp-avx2-movbe.S
|
||||
index 16fc673e..99258cf5 100644
|
||||
--- a/sysdeps/x86_64/multiarch/memcmp-avx2-movbe.S
|
||||
+++ b/sysdeps/x86_64/multiarch/memcmp-avx2-movbe.S
|
||||
@@ -429,22 +429,21 @@ L(page_cross_less_vec):
|
||||
# ifndef USE_AS_WMEMCMP
|
||||
cmpl $8, %edx
|
||||
jae L(between_8_15)
|
||||
+ /* Fall through for [4, 7]. */
|
||||
cmpl $4, %edx
|
||||
- jae L(between_4_7)
|
||||
+ jb L(between_2_3)
|
||||
|
||||
- /* Load as big endian to avoid branches. */
|
||||
- movzwl (%rdi), %eax
|
||||
- movzwl (%rsi), %ecx
|
||||
- shll $8, %eax
|
||||
- shll $8, %ecx
|
||||
- bswap %eax
|
||||
- bswap %ecx
|
||||
- movzbl -1(%rdi, %rdx), %edi
|
||||
- movzbl -1(%rsi, %rdx), %esi
|
||||
- orl %edi, %eax
|
||||
- orl %esi, %ecx
|
||||
- /* Subtraction is okay because the upper 8 bits are zero. */
|
||||
- subl %ecx, %eax
|
||||
+ movbe (%rdi), %eax
|
||||
+ movbe (%rsi), %ecx
|
||||
+ shlq $32, %rax
|
||||
+ shlq $32, %rcx
|
||||
+ movbe -4(%rdi, %rdx), %edi
|
||||
+ movbe -4(%rsi, %rdx), %esi
|
||||
+ orq %rdi, %rax
|
||||
+ orq %rsi, %rcx
|
||||
+ subq %rcx, %rax
|
||||
+ /* Fast path for return zero. */
|
||||
+ jnz L(ret_nonzero)
|
||||
/* No ymm register was touched. */
|
||||
ret
|
||||
|
||||
@@ -457,9 +456,33 @@ L(one_or_less):
|
||||
/* No ymm register was touched. */
|
||||
ret
|
||||
|
||||
+ .p2align 4,, 5
|
||||
+L(ret_nonzero):
|
||||
+ sbbl %eax, %eax
|
||||
+ orl $1, %eax
|
||||
+ /* No ymm register was touched. */
|
||||
+ ret
|
||||
+
|
||||
+ .p2align 4,, 2
|
||||
+L(zero):
|
||||
+ xorl %eax, %eax
|
||||
+ /* No ymm register was touched. */
|
||||
+ ret
|
||||
+
|
||||
.p2align 4
|
||||
L(between_8_15):
|
||||
-# endif
|
||||
+ movbe (%rdi), %rax
|
||||
+ movbe (%rsi), %rcx
|
||||
+ subq %rcx, %rax
|
||||
+ jnz L(ret_nonzero)
|
||||
+ movbe -8(%rdi, %rdx), %rax
|
||||
+ movbe -8(%rsi, %rdx), %rcx
|
||||
+ subq %rcx, %rax
|
||||
+ /* Fast path for return zero. */
|
||||
+ jnz L(ret_nonzero)
|
||||
+ /* No ymm register was touched. */
|
||||
+ ret
|
||||
+# else
|
||||
/* If USE_AS_WMEMCMP fall through into 8-15 byte case. */
|
||||
vmovq (%rdi), %xmm1
|
||||
vmovq (%rsi), %xmm2
|
||||
@@ -475,16 +498,13 @@ L(between_8_15):
|
||||
VPCMPEQ %xmm1, %xmm2, %xmm2
|
||||
vpmovmskb %xmm2, %eax
|
||||
subl $0xffff, %eax
|
||||
+ /* Fast path for return zero. */
|
||||
jnz L(return_vec_0)
|
||||
/* No ymm register was touched. */
|
||||
ret
|
||||
+# endif
|
||||
|
||||
- .p2align 4
|
||||
-L(zero):
|
||||
- xorl %eax, %eax
|
||||
- ret
|
||||
-
|
||||
- .p2align 4
|
||||
+ .p2align 4,, 10
|
||||
L(between_16_31):
|
||||
/* From 16 to 31 bytes. No branch when size == 16. */
|
||||
vmovdqu (%rsi), %xmm2
|
||||
@@ -501,11 +521,17 @@ L(between_16_31):
|
||||
VPCMPEQ (%rdi), %xmm2, %xmm2
|
||||
vpmovmskb %xmm2, %eax
|
||||
subl $0xffff, %eax
|
||||
+ /* Fast path for return zero. */
|
||||
jnz L(return_vec_0)
|
||||
/* No ymm register was touched. */
|
||||
ret
|
||||
|
||||
# ifdef USE_AS_WMEMCMP
|
||||
+ .p2align 4,, 2
|
||||
+L(zero):
|
||||
+ xorl %eax, %eax
|
||||
+ ret
|
||||
+
|
||||
.p2align 4
|
||||
L(one_or_less):
|
||||
jb L(zero)
|
||||
@@ -520,22 +546,20 @@ L(one_or_less):
|
||||
# else
|
||||
|
||||
.p2align 4
|
||||
-L(between_4_7):
|
||||
- /* Load as big endian with overlapping movbe to avoid branches.
|
||||
- */
|
||||
- movbe (%rdi), %eax
|
||||
- movbe (%rsi), %ecx
|
||||
- shlq $32, %rax
|
||||
- shlq $32, %rcx
|
||||
- movbe -4(%rdi, %rdx), %edi
|
||||
- movbe -4(%rsi, %rdx), %esi
|
||||
- orq %rdi, %rax
|
||||
- orq %rsi, %rcx
|
||||
- subq %rcx, %rax
|
||||
- jz L(zero_4_7)
|
||||
- sbbl %eax, %eax
|
||||
- orl $1, %eax
|
||||
-L(zero_4_7):
|
||||
+L(between_2_3):
|
||||
+ /* Load as big endian to avoid branches. */
|
||||
+ movzwl (%rdi), %eax
|
||||
+ movzwl (%rsi), %ecx
|
||||
+ bswap %eax
|
||||
+ bswap %ecx
|
||||
+ shrl %eax
|
||||
+ shrl %ecx
|
||||
+ movzbl -1(%rdi, %rdx), %edi
|
||||
+ movzbl -1(%rsi, %rdx), %esi
|
||||
+ orl %edi, %eax
|
||||
+ orl %esi, %ecx
|
||||
+ /* Subtraction is okay because the upper bit is zero. */
|
||||
+ subl %ecx, %eax
|
||||
/* No ymm register was touched. */
|
||||
ret
|
||||
# endif
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,226 +0,0 @@
|
||||
From 8162147872491bb5b48e91543b19c49a29ae6b6d Mon Sep 17 00:00:00 2001
|
||||
From: Wangyang Guo <wangyang.guo@intel.com>
|
||||
Date: Fri, 6 May 2022 01:50:10 +0000
|
||||
Subject: [PATCH] nptl: Add backoff mechanism to spinlock loop
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
When mutiple threads waiting for lock at the same time, once lock owner
|
||||
releases the lock, waiters will see lock available and all try to lock,
|
||||
which may cause an expensive CAS storm.
|
||||
|
||||
Binary exponential backoff with random jitter is introduced. As try-lock
|
||||
attempt increases, there is more likely that a larger number threads
|
||||
compete for adaptive mutex lock, so increase wait time in exponential.
|
||||
A random jitter is also added to avoid synchronous try-lock from other
|
||||
threads.
|
||||
|
||||
v2: Remove read-check before try-lock for performance.
|
||||
|
||||
v3:
|
||||
1. Restore read-check since it works well in some platform.
|
||||
2. Make backoff arch dependent, and enable it for x86_64.
|
||||
3. Limit max backoff to reduce latency in large critical section.
|
||||
|
||||
v4: Fix strict-prototypes error in sysdeps/nptl/pthread_mutex_backoff.h
|
||||
|
||||
v5: Commit log updated for regression in large critical section.
|
||||
|
||||
Result of pthread-mutex-locks bench
|
||||
|
||||
Test Platform: Xeon 8280L (2 socket, 112 CPUs in total)
|
||||
First Row: thread number
|
||||
First Col: critical section length
|
||||
Values: backoff vs upstream, time based, low is better
|
||||
|
||||
non-critical-length: 1
|
||||
1 2 4 8 16 32 64 112 140
|
||||
0 0.99 0.58 0.52 0.49 0.43 0.44 0.46 0.52 0.54
|
||||
1 0.98 0.43 0.56 0.50 0.44 0.45 0.50 0.56 0.57
|
||||
2 0.99 0.41 0.57 0.51 0.45 0.47 0.48 0.60 0.61
|
||||
4 0.99 0.45 0.59 0.53 0.48 0.49 0.52 0.64 0.65
|
||||
8 1.00 0.66 0.71 0.63 0.56 0.59 0.66 0.72 0.71
|
||||
16 0.97 0.78 0.91 0.73 0.67 0.70 0.79 0.80 0.80
|
||||
32 0.95 1.17 0.98 0.87 0.82 0.86 0.89 0.90 0.90
|
||||
64 0.96 0.95 1.01 1.01 0.98 1.00 1.03 0.99 0.99
|
||||
128 0.99 1.01 1.01 1.17 1.08 1.12 1.02 0.97 1.02
|
||||
|
||||
non-critical-length: 32
|
||||
1 2 4 8 16 32 64 112 140
|
||||
0 1.03 0.97 0.75 0.65 0.58 0.58 0.56 0.70 0.70
|
||||
1 0.94 0.95 0.76 0.65 0.58 0.58 0.61 0.71 0.72
|
||||
2 0.97 0.96 0.77 0.66 0.58 0.59 0.62 0.74 0.74
|
||||
4 0.99 0.96 0.78 0.66 0.60 0.61 0.66 0.76 0.77
|
||||
8 0.99 0.99 0.84 0.70 0.64 0.66 0.71 0.80 0.80
|
||||
16 0.98 0.97 0.95 0.76 0.70 0.73 0.81 0.85 0.84
|
||||
32 1.04 1.12 1.04 0.89 0.82 0.86 0.93 0.91 0.91
|
||||
64 0.99 1.15 1.07 1.00 0.99 1.01 1.05 0.99 0.99
|
||||
128 1.00 1.21 1.20 1.22 1.25 1.31 1.12 1.10 0.99
|
||||
|
||||
non-critical-length: 128
|
||||
1 2 4 8 16 32 64 112 140
|
||||
0 1.02 1.00 0.99 0.67 0.61 0.61 0.61 0.74 0.73
|
||||
1 0.95 0.99 1.00 0.68 0.61 0.60 0.60 0.74 0.74
|
||||
2 1.00 1.04 1.00 0.68 0.59 0.61 0.65 0.76 0.76
|
||||
4 1.00 0.96 0.98 0.70 0.63 0.63 0.67 0.78 0.77
|
||||
8 1.01 1.02 0.89 0.73 0.65 0.67 0.71 0.81 0.80
|
||||
16 0.99 0.96 0.96 0.79 0.71 0.73 0.80 0.84 0.84
|
||||
32 0.99 0.95 1.05 0.89 0.84 0.85 0.94 0.92 0.91
|
||||
64 1.00 0.99 1.16 1.04 1.00 1.02 1.06 0.99 0.99
|
||||
128 1.00 1.06 0.98 1.14 1.39 1.26 1.08 1.02 0.98
|
||||
|
||||
There is regression in large critical section. But adaptive mutex is
|
||||
aimed for "quick" locks. Small critical section is more common when
|
||||
users choose to use adaptive pthread_mutex.
|
||||
|
||||
Signed-off-by: Wangyang Guo <wangyang.guo@intel.com>
|
||||
Reviewed-by: H.J. Lu <hjl.tools@gmail.com>
|
||||
|
||||
Conflicts:
|
||||
pthreadP.h
|
||||
(had been moved)
|
||||
nptl/pthread_mutex_lock.c
|
||||
(max_adaptive_count renamed)
|
||||
|
||||
---
|
||||
nptl/pthreadP.h | 1 +
|
||||
nptl/pthread_mutex_lock.c | 16 +++++++--
|
||||
sysdeps/nptl/pthread_mutex_backoff.h | 35 ++++++++++++++++++
|
||||
sysdeps/x86_64/nptl/pthread_mutex_backoff.h | 39 +++++++++++++++++++++
|
||||
4 files changed, 89 insertions(+), 2 deletions(-)
|
||||
create mode 100644 sysdeps/nptl/pthread_mutex_backoff.h
|
||||
create mode 100644 sysdeps/x86_64/nptl/pthread_mutex_backoff.h
|
||||
|
||||
diff --git a/nptl/pthreadP.h b/nptl/pthreadP.h
|
||||
index 7ddc166c..1550e3b6 100644
|
||||
--- a/nptl/pthreadP.h
|
||||
+++ b/nptl/pthreadP.h
|
||||
@@ -33,6 +33,7 @@
|
||||
#include <kernel-features.h>
|
||||
#include <errno.h>
|
||||
#include <internal-signals.h>
|
||||
+#include <pthread_mutex_backoff.h>
|
||||
|
||||
|
||||
/* Atomic operations on TLS memory. */
|
||||
diff --git a/nptl/pthread_mutex_lock.c b/nptl/pthread_mutex_lock.c
|
||||
index d96a9933..c7770fc9 100644
|
||||
--- a/nptl/pthread_mutex_lock.c
|
||||
+++ b/nptl/pthread_mutex_lock.c
|
||||
@@ -133,14 +133,26 @@ __pthread_mutex_lock (pthread_mutex_t *mutex)
|
||||
int cnt = 0;
|
||||
int max_cnt = MIN (MAX_ADAPTIVE_COUNT,
|
||||
mutex->__data.__spins * 2 + 10);
|
||||
+ int spin_count, exp_backoff = 1;
|
||||
+ unsigned int jitter = get_jitter ();
|
||||
do
|
||||
{
|
||||
- if (cnt++ >= max_cnt)
|
||||
+ /* In each loop, spin count is exponential backoff plus
|
||||
+ random jitter, random range is [0, exp_backoff-1]. */
|
||||
+ spin_count = exp_backoff + (jitter & (exp_backoff - 1));
|
||||
+ cnt += spin_count;
|
||||
+ if (cnt >= max_cnt)
|
||||
{
|
||||
+ /* If cnt exceeds max spin count, just go to wait
|
||||
+ queue. */
|
||||
LLL_MUTEX_LOCK (mutex);
|
||||
break;
|
||||
}
|
||||
- atomic_spin_nop ();
|
||||
+ do
|
||||
+ atomic_spin_nop ();
|
||||
+ while (--spin_count > 0);
|
||||
+ /* Prepare for next loop. */
|
||||
+ exp_backoff = get_next_backoff (exp_backoff);
|
||||
}
|
||||
while (LLL_MUTEX_READ_LOCK (mutex) != 0
|
||||
|| LLL_MUTEX_TRYLOCK (mutex) != 0);
|
||||
diff --git a/sysdeps/nptl/pthread_mutex_backoff.h b/sysdeps/nptl/pthread_mutex_backoff.h
|
||||
new file mode 100644
|
||||
index 00000000..5b26c22a
|
||||
--- /dev/null
|
||||
+++ b/sysdeps/nptl/pthread_mutex_backoff.h
|
||||
@@ -0,0 +1,35 @@
|
||||
+/* Pthread mutex backoff configuration.
|
||||
+ Copyright (C) 2022 Free Software Foundation, Inc.
|
||||
+ This file is part of the GNU C Library.
|
||||
+
|
||||
+ The GNU C Library is free software; you can redistribute it and/or
|
||||
+ modify it under the terms of the GNU Lesser General Public
|
||||
+ License as published by the Free Software Foundation; either
|
||||
+ version 2.1 of the License, or (at your option) any later version.
|
||||
+
|
||||
+ The GNU C Library is distributed in the hope that it will be useful,
|
||||
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
||||
+ Lesser General Public License for more details.
|
||||
+
|
||||
+ You should have received a copy of the GNU Lesser General Public
|
||||
+ License along with the GNU C Library; if not, see
|
||||
+ <https://www.gnu.org/licenses/>. */
|
||||
+#ifndef _PTHREAD_MUTEX_BACKOFF_H
|
||||
+#define _PTHREAD_MUTEX_BACKOFF_H 1
|
||||
+
|
||||
+static inline unsigned int
|
||||
+get_jitter (void)
|
||||
+{
|
||||
+ /* Arch dependent random jitter, return 0 disables random. */
|
||||
+ return 0;
|
||||
+}
|
||||
+
|
||||
+static inline int
|
||||
+get_next_backoff (int backoff)
|
||||
+{
|
||||
+ /* Next backoff, return 1 disables mutex backoff. */
|
||||
+ return 1;
|
||||
+}
|
||||
+
|
||||
+#endif
|
||||
diff --git a/sysdeps/x86_64/nptl/pthread_mutex_backoff.h b/sysdeps/x86_64/nptl/pthread_mutex_backoff.h
|
||||
new file mode 100644
|
||||
index 00000000..ec74c3d9
|
||||
--- /dev/null
|
||||
+++ b/sysdeps/x86_64/nptl/pthread_mutex_backoff.h
|
||||
@@ -0,0 +1,39 @@
|
||||
+/* Pthread mutex backoff configuration.
|
||||
+ Copyright (C) 2022 Free Software Foundation, Inc.
|
||||
+ This file is part of the GNU C Library.
|
||||
+
|
||||
+ The GNU C Library is free software; you can redistribute it and/or
|
||||
+ modify it under the terms of the GNU Lesser General Public
|
||||
+ License as published by the Free Software Foundation; either
|
||||
+ version 2.1 of the License, or (at your option) any later version.
|
||||
+
|
||||
+ The GNU C Library is distributed in the hope that it will be useful,
|
||||
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
||||
+ Lesser General Public License for more details.
|
||||
+
|
||||
+ You should have received a copy of the GNU Lesser General Public
|
||||
+ License along with the GNU C Library; if not, see
|
||||
+ <https://www.gnu.org/licenses/>. */
|
||||
+#ifndef _PTHREAD_MUTEX_BACKOFF_H
|
||||
+#define _PTHREAD_MUTEX_BACKOFF_H 1
|
||||
+
|
||||
+#include <fast-jitter.h>
|
||||
+
|
||||
+static inline unsigned int
|
||||
+get_jitter (void)
|
||||
+{
|
||||
+ return get_fast_jitter ();
|
||||
+}
|
||||
+
|
||||
+#define MAX_BACKOFF 16
|
||||
+
|
||||
+static inline int
|
||||
+get_next_backoff (int backoff)
|
||||
+{
|
||||
+ /* Binary expontial backoff. Limiting max backoff
|
||||
+ can reduce latency in large critical section. */
|
||||
+ return (backoff < MAX_BACKOFF) ? backoff << 1 : backoff;
|
||||
+}
|
||||
+
|
||||
+#endif
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,55 +0,0 @@
|
||||
From c6272098323153db373f2986c67786ea8c85f1cf Mon Sep 17 00:00:00 2001
|
||||
From: Noah Goldstein <goldstein.w.n@gmail.com>
|
||||
Date: Tue, 15 Feb 2022 08:18:15 -0600
|
||||
Subject: [PATCH] x86: Fallback {str|wcs}cmp RTM in the ncmp overflow case [BZ
|
||||
#28896]
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
In the overflow fallback strncmp-avx2-rtm and wcsncmp-avx2-rtm would
|
||||
call strcmp-avx2 and wcscmp-avx2 respectively. This would have
|
||||
not checks around vzeroupper and would trigger spurious
|
||||
aborts. This commit fixes that.
|
||||
|
||||
test-strcmp, test-strncmp, test-wcscmp, and test-wcsncmp all pass on
|
||||
AVX2 machines with and without RTM.
|
||||
|
||||
Co-authored-by: H.J. Lu <hjl.tools@gmail.com>
|
||||
---
|
||||
sysdeps/x86_64/multiarch/strcmp-avx2.S | 8 ++------
|
||||
1 file changed, 2 insertions(+), 6 deletions(-)
|
||||
|
||||
Conflicts:
|
||||
sysdeps/x86_64/multiarch/strcmp-avx2.S
|
||||
(split into two patches due to upstream bug differences)
|
||||
|
||||
diff --git a/sysdeps/x86_64/multiarch/strcmp-avx2.S b/sysdeps/x86_64/multiarch/strcmp-avx2.S
|
||||
index 28cc98b6..e267c6cb 100644
|
||||
--- a/sysdeps/x86_64/multiarch/strcmp-avx2.S
|
||||
+++ b/sysdeps/x86_64/multiarch/strcmp-avx2.S
|
||||
@@ -345,10 +345,10 @@ L(one_or_less):
|
||||
movq %LOCALE_REG, %rdx
|
||||
# endif
|
||||
jb L(ret_zero)
|
||||
-# ifdef USE_AS_WCSCMP
|
||||
/* 'nbe' covers the case where length is negative (large
|
||||
unsigned). */
|
||||
- jnbe __wcscmp_avx2
|
||||
+ jnbe OVERFLOW_STRCMP
|
||||
+# ifdef USE_AS_WCSCMP
|
||||
movl (%rdi), %edx
|
||||
xorl %eax, %eax
|
||||
cmpl (%rsi), %edx
|
||||
@@ -357,10 +357,6 @@ L(one_or_less):
|
||||
negl %eax
|
||||
orl $1, %eax
|
||||
# else
|
||||
- /* 'nbe' covers the case where length is negative (large
|
||||
- unsigned). */
|
||||
-
|
||||
- jnbe __strcmp_avx2
|
||||
movzbl (%rdi), %eax
|
||||
movzbl (%rsi), %ecx
|
||||
TOLOWER_gpr (%rax, %eax)
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,60 +0,0 @@
|
||||
From 259a17cc98058d2576511201f85d28cb5d9de2a2 Mon Sep 17 00:00:00 2001
|
||||
From: Stefan Liebler <stli@linux.ibm.com>
|
||||
Date: Mon, 28 Jun 2021 13:01:07 +0200
|
||||
Subject: s390x: Update math: redirect roundeven function
|
||||
|
||||
After recent commit
|
||||
447954a206837b5f153869cfeeeab44631c3fac9
|
||||
"math: redirect roundeven function", building on
|
||||
s390x fails with:
|
||||
Error: symbol `__roundevenl' is already defined
|
||||
|
||||
Similar to aarch64/riscv fix, this patch redirects target
|
||||
specific functions for s390x:
|
||||
commit 3213ed770cbc5821920d16caa93c85e92dd7b9f6
|
||||
"Update math: redirect roundeven function"
|
||||
|
||||
diff --git a/sysdeps/s390/fpu/s_roundeven.c b/sysdeps/s390/fpu/s_roundeven.c
|
||||
index 40b07e054b..0773adfed0 100644
|
||||
--- a/sysdeps/s390/fpu/s_roundeven.c
|
||||
+++ b/sysdeps/s390/fpu/s_roundeven.c
|
||||
@@ -18,6 +18,7 @@
|
||||
<https://www.gnu.org/licenses/>. */
|
||||
|
||||
#ifdef HAVE_S390_MIN_Z196_ZARCH_ASM_SUPPORT
|
||||
+# define NO_MATH_REDIRECT
|
||||
# include <math.h>
|
||||
# include <libm-alias-double.h>
|
||||
|
||||
@@ -31,7 +32,6 @@ __roundeven (double x)
|
||||
__asm__ ("fidbra %0,4,%1,4" : "=f" (y) : "f" (x));
|
||||
return y;
|
||||
}
|
||||
-hidden_def (__roundeven)
|
||||
libm_alias_double (__roundeven, roundeven)
|
||||
|
||||
#else
|
||||
diff --git a/sysdeps/s390/fpu/s_roundevenf.c b/sysdeps/s390/fpu/s_roundevenf.c
|
||||
index d2fbf3d2b6..289785bc4a 100644
|
||||
--- a/sysdeps/s390/fpu/s_roundevenf.c
|
||||
+++ b/sysdeps/s390/fpu/s_roundevenf.c
|
||||
@@ -18,6 +18,7 @@
|
||||
<https://www.gnu.org/licenses/>. */
|
||||
|
||||
#ifdef HAVE_S390_MIN_Z196_ZARCH_ASM_SUPPORT
|
||||
+# define NO_MATH_REDIRECT
|
||||
# include <math.h>
|
||||
# include <libm-alias-float.h>
|
||||
|
||||
diff --git a/sysdeps/s390/fpu/s_roundevenl.c b/sysdeps/s390/fpu/s_roundevenl.c
|
||||
index 29ab7a8616..94b6459ab4 100644
|
||||
--- a/sysdeps/s390/fpu/s_roundevenl.c
|
||||
+++ b/sysdeps/s390/fpu/s_roundevenl.c
|
||||
@@ -18,6 +18,7 @@
|
||||
<https://www.gnu.org/licenses/>. */
|
||||
|
||||
#ifdef HAVE_S390_MIN_Z196_ZARCH_ASM_SUPPORT
|
||||
+# define NO_MATH_REDIRECT
|
||||
# include <math.h>
|
||||
# include <math_private.h>
|
||||
# include <libm-alias-ldouble.h>
|
||||
@ -1,74 +0,0 @@
|
||||
From 1da50d4bda07f04135dca39f40e79fc9eabed1f8 Mon Sep 17 00:00:00 2001
|
||||
From: "H.J. Lu" <hjl.tools@gmail.com>
|
||||
Date: Fri, 26 Feb 2021 05:36:59 -0800
|
||||
Subject: [PATCH] x86: Set Prefer_No_VZEROUPPER and add Prefer_AVX2_STRCMP
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
1. Set Prefer_No_VZEROUPPER if RTM is usable to avoid RTM abort triggered
|
||||
by VZEROUPPER inside a transactionally executing RTM region.
|
||||
2. Since to compare 2 32-byte strings, 256-bit EVEX strcmp requires 2
|
||||
loads, 3 VPCMPs and 2 KORDs while AVX2 strcmp requires 1 load, 2 VPCMPEQs,
|
||||
1 VPMINU and 1 VPMOVMSKB, AVX2 strcmp is faster than EVEX strcmp. Add
|
||||
Prefer_AVX2_STRCMP to prefer AVX2 strcmp family functions.
|
||||
---
|
||||
sysdeps/x86/cpu-features.c | 20 +++++++++++++++++--
|
||||
sysdeps/x86/cpu-tunables.c | 2 ++
|
||||
...cpu-features-preferred_feature_index_1.def | 1 +
|
||||
3 files changed, 21 insertions(+), 2 deletions(-)
|
||||
|
||||
diff --git a/sysdeps/x86/cpu-features.c b/sysdeps/x86/cpu-features.c
|
||||
index 91042505..3610ee5c 100644
|
||||
--- a/sysdeps/x86/cpu-features.c
|
||||
+++ b/sysdeps/x86/cpu-features.c
|
||||
@@ -524,8 +524,24 @@ init_cpu_features (struct cpu_features *cpu_features)
|
||||
cpu_features->preferred[index_arch_Prefer_No_VZEROUPPER]
|
||||
|= bit_arch_Prefer_No_VZEROUPPER;
|
||||
else
|
||||
- cpu_features->preferred[index_arch_Prefer_No_AVX512]
|
||||
- |= bit_arch_Prefer_No_AVX512;
|
||||
+ {
|
||||
+ cpu_features->preferred[index_arch_Prefer_No_AVX512]
|
||||
+ |= bit_arch_Prefer_No_AVX512;
|
||||
+
|
||||
+ /* Avoid RTM abort triggered by VZEROUPPER inside a
|
||||
+ transactionally executing RTM region. */
|
||||
+ if (CPU_FEATURE_USABLE_P (cpu_features, RTM))
|
||||
+ cpu_features->preferred[index_arch_Prefer_No_VZEROUPPER]
|
||||
+ |= bit_arch_Prefer_No_VZEROUPPER;
|
||||
+
|
||||
+ /* Since to compare 2 32-byte strings, 256-bit EVEX strcmp
|
||||
+ requires 2 loads, 3 VPCMPs and 2 KORDs while AVX2 strcmp
|
||||
+ requires 1 load, 2 VPCMPEQs, 1 VPMINU and 1 VPMOVMSKB,
|
||||
+ AVX2 strcmp is faster than EVEX strcmp. */
|
||||
+ if (CPU_FEATURE_USABLE_P (cpu_features, AVX2))
|
||||
+ cpu_features->preferred[index_arch_Prefer_AVX2_STRCMP]
|
||||
+ |= bit_arch_Prefer_AVX2_STRCMP;
|
||||
+ }
|
||||
}
|
||||
/* This spells out "AuthenticAMD". */
|
||||
else if (ebx == 0x68747541 && ecx == 0x444d4163 && edx == 0x69746e65)
|
||||
diff --git a/sysdeps/x86/cpu-tunables.c b/sysdeps/x86/cpu-tunables.c
|
||||
index 3173b2b9..73adbaba 100644
|
||||
--- a/sysdeps/x86/cpu-tunables.c
|
||||
+++ b/sysdeps/x86/cpu-tunables.c
|
||||
@@ -239,6 +239,8 @@ TUNABLE_CALLBACK (set_hwcaps) (tunable_val_t *valp)
|
||||
CHECK_GLIBC_IFUNC_PREFERRED_BOTH (n, cpu_features,
|
||||
Fast_Copy_Backward,
|
||||
disable, 18);
|
||||
+ CHECK_GLIBC_IFUNC_PREFERRED_NEED_BOTH
|
||||
+ (n, cpu_features, Prefer_AVX2_STRCMP, AVX2, disable, 18);
|
||||
}
|
||||
break;
|
||||
case 19:
|
||||
diff --git a/sysdeps/x86/include/cpu-features-preferred_feature_index_1.def b/sysdeps/x86/include/cpu-features-preferred_feature_index_1.def
|
||||
index 17a5cc42..4ca70b40 100644
|
||||
--- a/sysdeps/x86/include/cpu-features-preferred_feature_index_1.def
|
||||
+++ b/sysdeps/x86/include/cpu-features-preferred_feature_index_1.def
|
||||
@@ -32,3 +32,4 @@ BIT (Prefer_ERMS)
|
||||
BIT (Prefer_FSRM)
|
||||
BIT (Prefer_No_AVX512)
|
||||
BIT (MathVec_Prefer_No_AVX512)
|
||||
+BIT (Prefer_AVX2_STRCMP)
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,26 +0,0 @@
|
||||
From 3213ed770cbc5821920d16caa93c85e92dd7b9f6 Mon Sep 17 00:00:00 2001
|
||||
From: "H.J. Lu" <hjl.tools@gmail.com>
|
||||
Date: Wed, 23 Jun 2021 13:29:41 -0700
|
||||
Subject: Update math: redirect roundeven function
|
||||
|
||||
Redirect target specific roundeven functions for aarch64, ldbl-128ibm
|
||||
and riscv.
|
||||
|
||||
Conflicts:
|
||||
sysdeps/aarch64/*
|
||||
(not needed)
|
||||
sysdeps/riscv/*
|
||||
(not supported)
|
||||
|
||||
diff --git a/sysdeps/ieee754/ldbl-128ibm/s_roundevenl.c b/sysdeps/ieee754/ldbl-128ibm/s_roundevenl.c
|
||||
index 6701970f4a..90eecf496b 100644
|
||||
--- a/sysdeps/ieee754/ldbl-128ibm/s_roundevenl.c
|
||||
+++ b/sysdeps/ieee754/ldbl-128ibm/s_roundevenl.c
|
||||
@@ -17,6 +17,7 @@
|
||||
License along with the GNU C Library; if not, see
|
||||
<http://www.gnu.org/licenses/>. */
|
||||
|
||||
+#define NO_MATH_REDIRECT
|
||||
#include <math.h>
|
||||
#include <math_private.h>
|
||||
|
||||
File diff suppressed because it is too large
Load Diff
File diff suppressed because it is too large
Load Diff
@ -1,242 +0,0 @@
|
||||
From 63ad43566f7a25d140dc723598aeb441ad657eed Mon Sep 17 00:00:00 2001
|
||||
From: "H.J. Lu" <hjl.tools@gmail.com>
|
||||
Date: Fri, 5 Mar 2021 06:46:08 -0800
|
||||
Subject: [PATCH] x86-64: Add memmove family functions with 256-bit EVEX
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
Update ifunc-memmove.h to select the function optimized with 256-bit EVEX
|
||||
instructions using YMM16-YMM31 registers to avoid RTM abort with usable
|
||||
AVX512VL since VZEROUPPER isn't needed at function exit.
|
||||
---
|
||||
sysdeps/x86_64/multiarch/Makefile | 1 +
|
||||
sysdeps/x86_64/multiarch/ifunc-impl-list.c | 36 +++++++++++++++++++
|
||||
sysdeps/x86_64/multiarch/ifunc-memmove.h | 21 +++++++++--
|
||||
.../multiarch/memmove-evex-unaligned-erms.S | 33 +++++++++++++++++
|
||||
.../multiarch/memmove-vec-unaligned-erms.S | 24 ++++++++-----
|
||||
5 files changed, 104 insertions(+), 11 deletions(-)
|
||||
create mode 100644 sysdeps/x86_64/multiarch/memmove-evex-unaligned-erms.S
|
||||
|
||||
diff --git a/sysdeps/x86_64/multiarch/Makefile b/sysdeps/x86_64/multiarch/Makefile
|
||||
index 46783cd1..4563fc56 100644
|
||||
--- a/sysdeps/x86_64/multiarch/Makefile
|
||||
+++ b/sysdeps/x86_64/multiarch/Makefile
|
||||
@@ -41,6 +41,7 @@ sysdep_routines += strncat-c stpncpy-c strncpy-c \
|
||||
memset-avx2-unaligned-erms \
|
||||
memset-avx512-unaligned-erms \
|
||||
memchr-evex \
|
||||
+ memmove-evex-unaligned-erms \
|
||||
memrchr-evex \
|
||||
rawmemchr-evex \
|
||||
stpcpy-evex \
|
||||
diff --git a/sysdeps/x86_64/multiarch/ifunc-impl-list.c b/sysdeps/x86_64/multiarch/ifunc-impl-list.c
|
||||
index 082e4da3..6bd3abfc 100644
|
||||
--- a/sysdeps/x86_64/multiarch/ifunc-impl-list.c
|
||||
+++ b/sysdeps/x86_64/multiarch/ifunc-impl-list.c
|
||||
@@ -80,6 +80,12 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
|
||||
IFUNC_IMPL_ADD (array, i, __memmove_chk,
|
||||
CPU_FEATURE_USABLE (AVX),
|
||||
__memmove_chk_avx_unaligned_erms)
|
||||
+ IFUNC_IMPL_ADD (array, i, __memmove_chk,
|
||||
+ CPU_FEATURE_USABLE (AVX512VL),
|
||||
+ __memmove_chk_evex_unaligned)
|
||||
+ IFUNC_IMPL_ADD (array, i, __memmove_chk,
|
||||
+ CPU_FEATURE_USABLE (AVX512VL),
|
||||
+ __memmove_chk_evex_unaligned_erms)
|
||||
IFUNC_IMPL_ADD (array, i, __memmove_chk,
|
||||
CPU_FEATURE_USABLE (SSSE3),
|
||||
__memmove_chk_ssse3_back)
|
||||
@@ -102,6 +108,12 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
|
||||
IFUNC_IMPL_ADD (array, i, memmove,
|
||||
CPU_FEATURE_USABLE (AVX),
|
||||
__memmove_avx_unaligned_erms)
|
||||
+ IFUNC_IMPL_ADD (array, i, memmove,
|
||||
+ CPU_FEATURE_USABLE (AVX512VL),
|
||||
+ __memmove_evex_unaligned)
|
||||
+ IFUNC_IMPL_ADD (array, i, memmove,
|
||||
+ CPU_FEATURE_USABLE (AVX512VL),
|
||||
+ __memmove_evex_unaligned_erms)
|
||||
IFUNC_IMPL_ADD (array, i, memmove,
|
||||
CPU_FEATURE_USABLE (AVX512F),
|
||||
__memmove_avx512_no_vzeroupper)
|
||||
@@ -565,6 +577,12 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
|
||||
IFUNC_IMPL_ADD (array, i, __memcpy_chk,
|
||||
CPU_FEATURE_USABLE (AVX),
|
||||
__memcpy_chk_avx_unaligned_erms)
|
||||
+ IFUNC_IMPL_ADD (array, i, __memcpy_chk,
|
||||
+ CPU_FEATURE_USABLE (AVX512VL),
|
||||
+ __memcpy_chk_evex_unaligned)
|
||||
+ IFUNC_IMPL_ADD (array, i, __memcpy_chk,
|
||||
+ CPU_FEATURE_USABLE (AVX512VL),
|
||||
+ __memcpy_chk_evex_unaligned_erms)
|
||||
IFUNC_IMPL_ADD (array, i, __memcpy_chk,
|
||||
CPU_FEATURE_USABLE (SSSE3),
|
||||
__memcpy_chk_ssse3_back)
|
||||
@@ -587,6 +605,12 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
|
||||
IFUNC_IMPL_ADD (array, i, memcpy,
|
||||
CPU_FEATURE_USABLE (AVX),
|
||||
__memcpy_avx_unaligned_erms)
|
||||
+ IFUNC_IMPL_ADD (array, i, memcpy,
|
||||
+ CPU_FEATURE_USABLE (AVX512VL),
|
||||
+ __memcpy_evex_unaligned)
|
||||
+ IFUNC_IMPL_ADD (array, i, memcpy,
|
||||
+ CPU_FEATURE_USABLE (AVX512VL),
|
||||
+ __memcpy_evex_unaligned_erms)
|
||||
IFUNC_IMPL_ADD (array, i, memcpy, CPU_FEATURE_USABLE (SSSE3),
|
||||
__memcpy_ssse3_back)
|
||||
IFUNC_IMPL_ADD (array, i, memcpy, CPU_FEATURE_USABLE (SSSE3),
|
||||
@@ -623,6 +647,12 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
|
||||
IFUNC_IMPL_ADD (array, i, __mempcpy_chk,
|
||||
CPU_FEATURE_USABLE (AVX),
|
||||
__mempcpy_chk_avx_unaligned_erms)
|
||||
+ IFUNC_IMPL_ADD (array, i, __mempcpy_chk,
|
||||
+ CPU_FEATURE_USABLE (AVX512VL),
|
||||
+ __mempcpy_chk_evex_unaligned)
|
||||
+ IFUNC_IMPL_ADD (array, i, __mempcpy_chk,
|
||||
+ CPU_FEATURE_USABLE (AVX512VL),
|
||||
+ __mempcpy_chk_evex_unaligned_erms)
|
||||
IFUNC_IMPL_ADD (array, i, __mempcpy_chk,
|
||||
CPU_FEATURE_USABLE (SSSE3),
|
||||
__mempcpy_chk_ssse3_back)
|
||||
@@ -654,6 +684,12 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
|
||||
IFUNC_IMPL_ADD (array, i, mempcpy,
|
||||
CPU_FEATURE_USABLE (AVX),
|
||||
__mempcpy_avx_unaligned_erms)
|
||||
+ IFUNC_IMPL_ADD (array, i, mempcpy,
|
||||
+ CPU_FEATURE_USABLE (AVX512VL),
|
||||
+ __mempcpy_evex_unaligned)
|
||||
+ IFUNC_IMPL_ADD (array, i, mempcpy,
|
||||
+ CPU_FEATURE_USABLE (AVX512VL),
|
||||
+ __mempcpy_evex_unaligned_erms)
|
||||
IFUNC_IMPL_ADD (array, i, mempcpy, CPU_FEATURE_USABLE (SSSE3),
|
||||
__mempcpy_ssse3_back)
|
||||
IFUNC_IMPL_ADD (array, i, mempcpy, CPU_FEATURE_USABLE (SSSE3),
|
||||
diff --git a/sysdeps/x86_64/multiarch/ifunc-memmove.h b/sysdeps/x86_64/multiarch/ifunc-memmove.h
|
||||
index 5e5f0299..6f8bce5f 100644
|
||||
--- a/sysdeps/x86_64/multiarch/ifunc-memmove.h
|
||||
+++ b/sysdeps/x86_64/multiarch/ifunc-memmove.h
|
||||
@@ -29,6 +29,10 @@ extern __typeof (REDIRECT_NAME) OPTIMIZE (ssse3_back) attribute_hidden;
|
||||
extern __typeof (REDIRECT_NAME) OPTIMIZE (avx_unaligned) attribute_hidden;
|
||||
extern __typeof (REDIRECT_NAME) OPTIMIZE (avx_unaligned_erms)
|
||||
attribute_hidden;
|
||||
+extern __typeof (REDIRECT_NAME) OPTIMIZE (evex_unaligned)
|
||||
+ attribute_hidden;
|
||||
+extern __typeof (REDIRECT_NAME) OPTIMIZE (evex_unaligned_erms)
|
||||
+ attribute_hidden;
|
||||
extern __typeof (REDIRECT_NAME) OPTIMIZE (avx512_unaligned)
|
||||
attribute_hidden;
|
||||
extern __typeof (REDIRECT_NAME) OPTIMIZE (avx512_unaligned_erms)
|
||||
@@ -59,10 +63,21 @@ IFUNC_SELECTOR (void)
|
||||
|
||||
if (CPU_FEATURES_ARCH_P (cpu_features, AVX_Fast_Unaligned_Load))
|
||||
{
|
||||
- if (CPU_FEATURE_USABLE_P (cpu_features, ERMS))
|
||||
- return OPTIMIZE (avx_unaligned_erms);
|
||||
+ if (CPU_FEATURE_USABLE_P (cpu_features, AVX512VL))
|
||||
+ {
|
||||
+ if (CPU_FEATURE_USABLE_P (cpu_features, ERMS))
|
||||
+ return OPTIMIZE (evex_unaligned_erms);
|
||||
+
|
||||
+ return OPTIMIZE (evex_unaligned);
|
||||
+ }
|
||||
+
|
||||
+ if (!CPU_FEATURES_ARCH_P (cpu_features, Prefer_No_VZEROUPPER))
|
||||
+ {
|
||||
+ if (CPU_FEATURE_USABLE_P (cpu_features, ERMS))
|
||||
+ return OPTIMIZE (avx_unaligned_erms);
|
||||
|
||||
- return OPTIMIZE (avx_unaligned);
|
||||
+ return OPTIMIZE (avx_unaligned);
|
||||
+ }
|
||||
}
|
||||
|
||||
if (!CPU_FEATURE_USABLE_P (cpu_features, SSSE3)
|
||||
diff --git a/sysdeps/x86_64/multiarch/memmove-evex-unaligned-erms.S b/sysdeps/x86_64/multiarch/memmove-evex-unaligned-erms.S
|
||||
new file mode 100644
|
||||
index 00000000..0cbce8f9
|
||||
--- /dev/null
|
||||
+++ b/sysdeps/x86_64/multiarch/memmove-evex-unaligned-erms.S
|
||||
@@ -0,0 +1,33 @@
|
||||
+#if IS_IN (libc)
|
||||
+# define VEC_SIZE 32
|
||||
+# define XMM0 xmm16
|
||||
+# define XMM1 xmm17
|
||||
+# define YMM0 ymm16
|
||||
+# define YMM1 ymm17
|
||||
+# define VEC0 ymm16
|
||||
+# define VEC1 ymm17
|
||||
+# define VEC2 ymm18
|
||||
+# define VEC3 ymm19
|
||||
+# define VEC4 ymm20
|
||||
+# define VEC5 ymm21
|
||||
+# define VEC6 ymm22
|
||||
+# define VEC7 ymm23
|
||||
+# define VEC8 ymm24
|
||||
+# define VEC9 ymm25
|
||||
+# define VEC10 ymm26
|
||||
+# define VEC11 ymm27
|
||||
+# define VEC12 ymm28
|
||||
+# define VEC13 ymm29
|
||||
+# define VEC14 ymm30
|
||||
+# define VEC15 ymm31
|
||||
+# define VEC(i) VEC##i
|
||||
+# define VMOVNT vmovntdq
|
||||
+# define VMOVU vmovdqu64
|
||||
+# define VMOVA vmovdqa64
|
||||
+# define VZEROUPPER
|
||||
+
|
||||
+# define SECTION(p) p##.evex
|
||||
+# define MEMMOVE_SYMBOL(p,s) p##_evex_##s
|
||||
+
|
||||
+# include "memmove-vec-unaligned-erms.S"
|
||||
+#endif
|
||||
diff --git a/sysdeps/x86_64/multiarch/memmove-vec-unaligned-erms.S b/sysdeps/x86_64/multiarch/memmove-vec-unaligned-erms.S
|
||||
index 274aa1c7..08e21692 100644
|
||||
--- a/sysdeps/x86_64/multiarch/memmove-vec-unaligned-erms.S
|
||||
+++ b/sysdeps/x86_64/multiarch/memmove-vec-unaligned-erms.S
|
||||
@@ -48,6 +48,14 @@
|
||||
# define MEMMOVE_CHK_SYMBOL(p,s) MEMMOVE_SYMBOL(p, s)
|
||||
#endif
|
||||
|
||||
+#ifndef XMM0
|
||||
+# define XMM0 xmm0
|
||||
+#endif
|
||||
+
|
||||
+#ifndef YMM0
|
||||
+# define YMM0 ymm0
|
||||
+#endif
|
||||
+
|
||||
#ifndef VZEROUPPER
|
||||
# if VEC_SIZE > 16
|
||||
# define VZEROUPPER vzeroupper
|
||||
@@ -277,20 +285,20 @@ L(less_vec):
|
||||
#if VEC_SIZE > 32
|
||||
L(between_32_63):
|
||||
/* From 32 to 63. No branch when size == 32. */
|
||||
- vmovdqu (%rsi), %ymm0
|
||||
- vmovdqu -32(%rsi,%rdx), %ymm1
|
||||
- vmovdqu %ymm0, (%rdi)
|
||||
- vmovdqu %ymm1, -32(%rdi,%rdx)
|
||||
+ VMOVU (%rsi), %YMM0
|
||||
+ VMOVU -32(%rsi,%rdx), %YMM1
|
||||
+ VMOVU %YMM0, (%rdi)
|
||||
+ VMOVU %YMM1, -32(%rdi,%rdx)
|
||||
VZEROUPPER
|
||||
ret
|
||||
#endif
|
||||
#if VEC_SIZE > 16
|
||||
/* From 16 to 31. No branch when size == 16. */
|
||||
L(between_16_31):
|
||||
- vmovdqu (%rsi), %xmm0
|
||||
- vmovdqu -16(%rsi,%rdx), %xmm1
|
||||
- vmovdqu %xmm0, (%rdi)
|
||||
- vmovdqu %xmm1, -16(%rdi,%rdx)
|
||||
+ VMOVU (%rsi), %XMM0
|
||||
+ VMOVU -16(%rsi,%rdx), %XMM1
|
||||
+ VMOVU %XMM0, (%rdi)
|
||||
+ VMOVU %XMM1, -16(%rdi,%rdx)
|
||||
ret
|
||||
#endif
|
||||
L(between_8_15):
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,254 +0,0 @@
|
||||
From 1b968b6b9b3aac702ac2f133e0dd16cfdbb415ee Mon Sep 17 00:00:00 2001
|
||||
From: "H.J. Lu" <hjl.tools@gmail.com>
|
||||
Date: Fri, 5 Mar 2021 07:15:03 -0800
|
||||
Subject: [PATCH] x86-64: Add memset family functions with 256-bit EVEX
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
Update ifunc-memset.h/ifunc-wmemset.h to select the function optimized
|
||||
with 256-bit EVEX instructions using YMM16-YMM31 registers to avoid RTM
|
||||
abort with usable AVX512VL and AVX512BW since VZEROUPPER isn't needed at
|
||||
function exit.
|
||||
---
|
||||
sysdeps/x86_64/multiarch/Makefile | 1 +
|
||||
sysdeps/x86_64/multiarch/ifunc-impl-list.c | 22 +++++++++++++++++
|
||||
sysdeps/x86_64/multiarch/ifunc-memset.h | 24 +++++++++++++++----
|
||||
sysdeps/x86_64/multiarch/ifunc-wmemset.h | 13 ++++++----
|
||||
.../multiarch/memset-evex-unaligned-erms.S | 24 +++++++++++++++++++
|
||||
.../multiarch/memset-vec-unaligned-erms.S | 20 +++++++++++-----
|
||||
6 files changed, 90 insertions(+), 14 deletions(-)
|
||||
create mode 100644 sysdeps/x86_64/multiarch/memset-evex-unaligned-erms.S
|
||||
|
||||
diff --git a/sysdeps/x86_64/multiarch/Makefile b/sysdeps/x86_64/multiarch/Makefile
|
||||
index 4563fc56..1cc0a10e 100644
|
||||
--- a/sysdeps/x86_64/multiarch/Makefile
|
||||
+++ b/sysdeps/x86_64/multiarch/Makefile
|
||||
@@ -43,6 +43,7 @@ sysdep_routines += strncat-c stpncpy-c strncpy-c \
|
||||
memchr-evex \
|
||||
memmove-evex-unaligned-erms \
|
||||
memrchr-evex \
|
||||
+ memset-evex-unaligned-erms \
|
||||
rawmemchr-evex \
|
||||
stpcpy-evex \
|
||||
stpncpy-evex \
|
||||
diff --git a/sysdeps/x86_64/multiarch/ifunc-impl-list.c b/sysdeps/x86_64/multiarch/ifunc-impl-list.c
|
||||
index 6bd3abfc..7cf83485 100644
|
||||
--- a/sysdeps/x86_64/multiarch/ifunc-impl-list.c
|
||||
+++ b/sysdeps/x86_64/multiarch/ifunc-impl-list.c
|
||||
@@ -160,6 +160,14 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
|
||||
IFUNC_IMPL_ADD (array, i, __memset_chk,
|
||||
CPU_FEATURE_USABLE (AVX2),
|
||||
__memset_chk_avx2_unaligned_erms)
|
||||
+ IFUNC_IMPL_ADD (array, i, __memset_chk,
|
||||
+ (CPU_FEATURE_USABLE (AVX512VL)
|
||||
+ && CPU_FEATURE_USABLE (AVX512BW)),
|
||||
+ __memset_chk_evex_unaligned)
|
||||
+ IFUNC_IMPL_ADD (array, i, __memset_chk,
|
||||
+ (CPU_FEATURE_USABLE (AVX512VL)
|
||||
+ && CPU_FEATURE_USABLE (AVX512BW)),
|
||||
+ __memset_chk_evex_unaligned_erms)
|
||||
IFUNC_IMPL_ADD (array, i, __memset_chk,
|
||||
CPU_FEATURE_USABLE (AVX512F),
|
||||
__memset_chk_avx512_unaligned_erms)
|
||||
@@ -185,6 +193,14 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
|
||||
IFUNC_IMPL_ADD (array, i, memset,
|
||||
CPU_FEATURE_USABLE (AVX2),
|
||||
__memset_avx2_unaligned_erms)
|
||||
+ IFUNC_IMPL_ADD (array, i, memset,
|
||||
+ (CPU_FEATURE_USABLE (AVX512VL)
|
||||
+ && CPU_FEATURE_USABLE (AVX512BW)),
|
||||
+ __memset_evex_unaligned)
|
||||
+ IFUNC_IMPL_ADD (array, i, memset,
|
||||
+ (CPU_FEATURE_USABLE (AVX512VL)
|
||||
+ && CPU_FEATURE_USABLE (AVX512BW)),
|
||||
+ __memset_evex_unaligned_erms)
|
||||
IFUNC_IMPL_ADD (array, i, memset,
|
||||
CPU_FEATURE_USABLE (AVX512F),
|
||||
__memset_avx512_unaligned_erms)
|
||||
@@ -555,6 +571,9 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
|
||||
IFUNC_IMPL_ADD (array, i, wmemset,
|
||||
CPU_FEATURE_USABLE (AVX2),
|
||||
__wmemset_avx2_unaligned)
|
||||
+ IFUNC_IMPL_ADD (array, i, wmemset,
|
||||
+ CPU_FEATURE_USABLE (AVX512VL),
|
||||
+ __wmemset_evex_unaligned)
|
||||
IFUNC_IMPL_ADD (array, i, wmemset,
|
||||
CPU_FEATURE_USABLE (AVX512F),
|
||||
__wmemset_avx512_unaligned))
|
||||
@@ -723,6 +742,9 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
|
||||
IFUNC_IMPL_ADD (array, i, __wmemset_chk,
|
||||
CPU_FEATURE_USABLE (AVX2),
|
||||
__wmemset_chk_avx2_unaligned)
|
||||
+ IFUNC_IMPL_ADD (array, i, __wmemset_chk,
|
||||
+ CPU_FEATURE_USABLE (AVX512VL),
|
||||
+ __wmemset_chk_evex_unaligned)
|
||||
IFUNC_IMPL_ADD (array, i, __wmemset_chk,
|
||||
CPU_FEATURE_USABLE (AVX512F),
|
||||
__wmemset_chk_avx512_unaligned))
|
||||
diff --git a/sysdeps/x86_64/multiarch/ifunc-memset.h b/sysdeps/x86_64/multiarch/ifunc-memset.h
|
||||
index 708bd72e..6f31f4dc 100644
|
||||
--- a/sysdeps/x86_64/multiarch/ifunc-memset.h
|
||||
+++ b/sysdeps/x86_64/multiarch/ifunc-memset.h
|
||||
@@ -27,6 +27,10 @@ extern __typeof (REDIRECT_NAME) OPTIMIZE (sse2_unaligned_erms)
|
||||
extern __typeof (REDIRECT_NAME) OPTIMIZE (avx2_unaligned) attribute_hidden;
|
||||
extern __typeof (REDIRECT_NAME) OPTIMIZE (avx2_unaligned_erms)
|
||||
attribute_hidden;
|
||||
+extern __typeof (REDIRECT_NAME) OPTIMIZE (evex_unaligned)
|
||||
+ attribute_hidden;
|
||||
+extern __typeof (REDIRECT_NAME) OPTIMIZE (evex_unaligned_erms)
|
||||
+ attribute_hidden;
|
||||
extern __typeof (REDIRECT_NAME) OPTIMIZE (avx512_unaligned)
|
||||
attribute_hidden;
|
||||
extern __typeof (REDIRECT_NAME) OPTIMIZE (avx512_unaligned_erms)
|
||||
@@ -56,10 +60,22 @@ IFUNC_SELECTOR (void)
|
||||
|
||||
if (CPU_FEATURE_USABLE_P (cpu_features, AVX2))
|
||||
{
|
||||
- if (CPU_FEATURE_USABLE_P (cpu_features, ERMS))
|
||||
- return OPTIMIZE (avx2_unaligned_erms);
|
||||
- else
|
||||
- return OPTIMIZE (avx2_unaligned);
|
||||
+ if (CPU_FEATURE_USABLE_P (cpu_features, AVX512VL)
|
||||
+ && CPU_FEATURE_USABLE_P (cpu_features, AVX512BW))
|
||||
+ {
|
||||
+ if (CPU_FEATURE_USABLE_P (cpu_features, ERMS))
|
||||
+ return OPTIMIZE (evex_unaligned_erms);
|
||||
+
|
||||
+ return OPTIMIZE (evex_unaligned);
|
||||
+ }
|
||||
+
|
||||
+ if (!CPU_FEATURES_ARCH_P (cpu_features, Prefer_No_VZEROUPPER))
|
||||
+ {
|
||||
+ if (CPU_FEATURE_USABLE_P (cpu_features, ERMS))
|
||||
+ return OPTIMIZE (avx2_unaligned_erms);
|
||||
+
|
||||
+ return OPTIMIZE (avx2_unaligned);
|
||||
+ }
|
||||
}
|
||||
|
||||
if (CPU_FEATURE_USABLE_P (cpu_features, ERMS))
|
||||
diff --git a/sysdeps/x86_64/multiarch/ifunc-wmemset.h b/sysdeps/x86_64/multiarch/ifunc-wmemset.h
|
||||
index eb242210..9290c4bf 100644
|
||||
--- a/sysdeps/x86_64/multiarch/ifunc-wmemset.h
|
||||
+++ b/sysdeps/x86_64/multiarch/ifunc-wmemset.h
|
||||
@@ -20,6 +20,7 @@
|
||||
|
||||
extern __typeof (REDIRECT_NAME) OPTIMIZE (sse2_unaligned) attribute_hidden;
|
||||
extern __typeof (REDIRECT_NAME) OPTIMIZE (avx2_unaligned) attribute_hidden;
|
||||
+extern __typeof (REDIRECT_NAME) OPTIMIZE (evex_unaligned) attribute_hidden;
|
||||
extern __typeof (REDIRECT_NAME) OPTIMIZE (avx512_unaligned) attribute_hidden;
|
||||
|
||||
static inline void *
|
||||
@@ -27,14 +28,18 @@ IFUNC_SELECTOR (void)
|
||||
{
|
||||
const struct cpu_features* cpu_features = __get_cpu_features ();
|
||||
|
||||
- if (!CPU_FEATURES_ARCH_P (cpu_features, Prefer_No_VZEROUPPER)
|
||||
- && CPU_FEATURE_USABLE_P (cpu_features, AVX2)
|
||||
+ if (CPU_FEATURE_USABLE_P (cpu_features, AVX2)
|
||||
&& CPU_FEATURES_ARCH_P (cpu_features, AVX_Fast_Unaligned_Load))
|
||||
{
|
||||
if (CPU_FEATURE_USABLE_P (cpu_features, AVX512F)
|
||||
- && !CPU_FEATURES_ARCH_P (cpu_features, Prefer_No_AVX512))
|
||||
+ && !CPU_FEATURES_ARCH_P (cpu_features, Prefer_No_AVX512)
|
||||
+ && !CPU_FEATURES_ARCH_P (cpu_features, Prefer_No_VZEROUPPER))
|
||||
return OPTIMIZE (avx512_unaligned);
|
||||
- else
|
||||
+
|
||||
+ if (CPU_FEATURE_USABLE_P (cpu_features, AVX512VL))
|
||||
+ return OPTIMIZE (evex_unaligned);
|
||||
+
|
||||
+ if (!CPU_FEATURES_ARCH_P (cpu_features, Prefer_No_VZEROUPPER))
|
||||
return OPTIMIZE (avx2_unaligned);
|
||||
}
|
||||
|
||||
diff --git a/sysdeps/x86_64/multiarch/memset-evex-unaligned-erms.S b/sysdeps/x86_64/multiarch/memset-evex-unaligned-erms.S
|
||||
new file mode 100644
|
||||
index 00000000..ae0a4d6e
|
||||
--- /dev/null
|
||||
+++ b/sysdeps/x86_64/multiarch/memset-evex-unaligned-erms.S
|
||||
@@ -0,0 +1,24 @@
|
||||
+#if IS_IN (libc)
|
||||
+# define VEC_SIZE 32
|
||||
+# define XMM0 xmm16
|
||||
+# define YMM0 ymm16
|
||||
+# define VEC0 ymm16
|
||||
+# define VEC(i) VEC##i
|
||||
+# define VMOVU vmovdqu64
|
||||
+# define VMOVA vmovdqa64
|
||||
+# define VZEROUPPER
|
||||
+
|
||||
+# define MEMSET_VDUP_TO_VEC0_AND_SET_RETURN(d, r) \
|
||||
+ movq r, %rax; \
|
||||
+ vpbroadcastb d, %VEC0
|
||||
+
|
||||
+# define WMEMSET_VDUP_TO_VEC0_AND_SET_RETURN(d, r) \
|
||||
+ movq r, %rax; \
|
||||
+ vpbroadcastd d, %VEC0
|
||||
+
|
||||
+# define SECTION(p) p##.evex
|
||||
+# define MEMSET_SYMBOL(p,s) p##_evex_##s
|
||||
+# define WMEMSET_SYMBOL(p,s) p##_evex_##s
|
||||
+
|
||||
+# include "memset-vec-unaligned-erms.S"
|
||||
+#endif
|
||||
diff --git a/sysdeps/x86_64/multiarch/memset-vec-unaligned-erms.S b/sysdeps/x86_64/multiarch/memset-vec-unaligned-erms.S
|
||||
index 9a0fd818..71e91a8f 100644
|
||||
--- a/sysdeps/x86_64/multiarch/memset-vec-unaligned-erms.S
|
||||
+++ b/sysdeps/x86_64/multiarch/memset-vec-unaligned-erms.S
|
||||
@@ -34,6 +34,14 @@
|
||||
# define WMEMSET_CHK_SYMBOL(p,s) WMEMSET_SYMBOL(p, s)
|
||||
#endif
|
||||
|
||||
+#ifndef XMM0
|
||||
+# define XMM0 xmm0
|
||||
+#endif
|
||||
+
|
||||
+#ifndef YMM0
|
||||
+# define YMM0 ymm0
|
||||
+#endif
|
||||
+
|
||||
#ifndef VZEROUPPER
|
||||
# if VEC_SIZE > 16
|
||||
# define VZEROUPPER vzeroupper
|
||||
@@ -67,7 +75,7 @@
|
||||
ENTRY (__bzero)
|
||||
mov %RDI_LP, %RAX_LP /* Set return value. */
|
||||
mov %RSI_LP, %RDX_LP /* Set n. */
|
||||
- pxor %xmm0, %xmm0
|
||||
+ pxor %XMM0, %XMM0
|
||||
jmp L(entry_from_bzero)
|
||||
END (__bzero)
|
||||
weak_alias (__bzero, bzero)
|
||||
@@ -223,7 +231,7 @@ L(less_vec):
|
||||
cmpb $16, %dl
|
||||
jae L(between_16_31)
|
||||
# endif
|
||||
- MOVQ %xmm0, %rcx
|
||||
+ MOVQ %XMM0, %rcx
|
||||
cmpb $8, %dl
|
||||
jae L(between_8_15)
|
||||
cmpb $4, %dl
|
||||
@@ -238,16 +246,16 @@ L(less_vec):
|
||||
# if VEC_SIZE > 32
|
||||
/* From 32 to 63. No branch when size == 32. */
|
||||
L(between_32_63):
|
||||
- vmovdqu %ymm0, -32(%rdi,%rdx)
|
||||
- vmovdqu %ymm0, (%rdi)
|
||||
+ VMOVU %YMM0, -32(%rdi,%rdx)
|
||||
+ VMOVU %YMM0, (%rdi)
|
||||
VZEROUPPER
|
||||
ret
|
||||
# endif
|
||||
# if VEC_SIZE > 16
|
||||
/* From 16 to 31. No branch when size == 16. */
|
||||
L(between_16_31):
|
||||
- vmovdqu %xmm0, -16(%rdi,%rdx)
|
||||
- vmovdqu %xmm0, (%rdi)
|
||||
+ VMOVU %XMM0, -16(%rdi,%rdx)
|
||||
+ VMOVU %XMM0, (%rdi)
|
||||
VZEROUPPER
|
||||
ret
|
||||
# endif
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,561 +0,0 @@
|
||||
From 91264fe3577fe887b4860923fa6142b5274c8965 Mon Sep 17 00:00:00 2001
|
||||
From: "H.J. Lu" <hjl.tools@gmail.com>
|
||||
Date: Fri, 5 Mar 2021 07:20:28 -0800
|
||||
Subject: [PATCH] x86-64: Add memcmp family functions with 256-bit EVEX
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
Update ifunc-memcmp.h to select the function optimized with 256-bit EVEX
|
||||
instructions using YMM16-YMM31 registers to avoid RTM abort with usable
|
||||
AVX512VL, AVX512BW and MOVBE since VZEROUPPER isn't needed at function
|
||||
exit.
|
||||
---
|
||||
sysdeps/x86_64/multiarch/Makefile | 4 +-
|
||||
sysdeps/x86_64/multiarch/ifunc-impl-list.c | 10 +
|
||||
sysdeps/x86_64/multiarch/ifunc-memcmp.h | 13 +-
|
||||
sysdeps/x86_64/multiarch/memcmp-evex-movbe.S | 440 ++++++++++++++++++
|
||||
sysdeps/x86_64/multiarch/wmemcmp-evex-movbe.S | 4 +
|
||||
5 files changed, 467 insertions(+), 4 deletions(-)
|
||||
create mode 100644 sysdeps/x86_64/multiarch/memcmp-evex-movbe.S
|
||||
create mode 100644 sysdeps/x86_64/multiarch/wmemcmp-evex-movbe.S
|
||||
|
||||
diff --git a/sysdeps/x86_64/multiarch/Makefile b/sysdeps/x86_64/multiarch/Makefile
|
||||
index 1cc0a10e..9d79b138 100644
|
||||
--- a/sysdeps/x86_64/multiarch/Makefile
|
||||
+++ b/sysdeps/x86_64/multiarch/Makefile
|
||||
@@ -41,6 +41,7 @@ sysdep_routines += strncat-c stpncpy-c strncpy-c \
|
||||
memset-avx2-unaligned-erms \
|
||||
memset-avx512-unaligned-erms \
|
||||
memchr-evex \
|
||||
+ memcmp-evex-movbe \
|
||||
memmove-evex-unaligned-erms \
|
||||
memrchr-evex \
|
||||
memset-evex-unaligned-erms \
|
||||
@@ -81,7 +82,8 @@ sysdep_routines += wmemcmp-sse4 wmemcmp-ssse3 wmemcmp-c \
|
||||
wcsncmp-evex \
|
||||
wcsnlen-evex \
|
||||
wcsrchr-evex \
|
||||
- wmemchr-evex
|
||||
+ wmemchr-evex \
|
||||
+ wmemcmp-evex-movbe
|
||||
endif
|
||||
|
||||
ifeq ($(subdir),debug)
|
||||
diff --git a/sysdeps/x86_64/multiarch/ifunc-impl-list.c b/sysdeps/x86_64/multiarch/ifunc-impl-list.c
|
||||
index 7cf83485..c8da910e 100644
|
||||
--- a/sysdeps/x86_64/multiarch/ifunc-impl-list.c
|
||||
+++ b/sysdeps/x86_64/multiarch/ifunc-impl-list.c
|
||||
@@ -56,6 +56,11 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
|
||||
(CPU_FEATURE_USABLE (AVX2)
|
||||
&& CPU_FEATURE_USABLE (MOVBE)),
|
||||
__memcmp_avx2_movbe)
|
||||
+ IFUNC_IMPL_ADD (array, i, memcmp,
|
||||
+ (CPU_FEATURE_USABLE (AVX512VL)
|
||||
+ && CPU_FEATURE_USABLE (AVX512BW)
|
||||
+ && CPU_FEATURE_USABLE (MOVBE)),
|
||||
+ __memcmp_evex_movbe)
|
||||
IFUNC_IMPL_ADD (array, i, memcmp, CPU_FEATURE_USABLE (SSE4_1),
|
||||
__memcmp_sse4_1)
|
||||
IFUNC_IMPL_ADD (array, i, memcmp, CPU_FEATURE_USABLE (SSSE3),
|
||||
@@ -558,6 +563,11 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
|
||||
(CPU_FEATURE_USABLE (AVX2)
|
||||
&& CPU_FEATURE_USABLE (MOVBE)),
|
||||
__wmemcmp_avx2_movbe)
|
||||
+ IFUNC_IMPL_ADD (array, i, wmemcmp,
|
||||
+ (CPU_FEATURE_USABLE (AVX512VL)
|
||||
+ && CPU_FEATURE_USABLE (AVX512BW)
|
||||
+ && CPU_FEATURE_USABLE (MOVBE)),
|
||||
+ __wmemcmp_evex_movbe)
|
||||
IFUNC_IMPL_ADD (array, i, wmemcmp, CPU_FEATURE_USABLE (SSE4_1),
|
||||
__wmemcmp_sse4_1)
|
||||
IFUNC_IMPL_ADD (array, i, wmemcmp, CPU_FEATURE_USABLE (SSSE3),
|
||||
diff --git a/sysdeps/x86_64/multiarch/ifunc-memcmp.h b/sysdeps/x86_64/multiarch/ifunc-memcmp.h
|
||||
index 6c1f3153..3ca1f0a6 100644
|
||||
--- a/sysdeps/x86_64/multiarch/ifunc-memcmp.h
|
||||
+++ b/sysdeps/x86_64/multiarch/ifunc-memcmp.h
|
||||
@@ -23,17 +23,24 @@ extern __typeof (REDIRECT_NAME) OPTIMIZE (sse2) attribute_hidden;
|
||||
extern __typeof (REDIRECT_NAME) OPTIMIZE (ssse3) attribute_hidden;
|
||||
extern __typeof (REDIRECT_NAME) OPTIMIZE (sse4_1) attribute_hidden;
|
||||
extern __typeof (REDIRECT_NAME) OPTIMIZE (avx2_movbe) attribute_hidden;
|
||||
+extern __typeof (REDIRECT_NAME) OPTIMIZE (evex_movbe) attribute_hidden;
|
||||
|
||||
static inline void *
|
||||
IFUNC_SELECTOR (void)
|
||||
{
|
||||
const struct cpu_features* cpu_features = __get_cpu_features ();
|
||||
|
||||
- if (!CPU_FEATURES_ARCH_P (cpu_features, Prefer_No_VZEROUPPER)
|
||||
- && CPU_FEATURE_USABLE_P (cpu_features, AVX2)
|
||||
+ if (CPU_FEATURE_USABLE_P (cpu_features, AVX2)
|
||||
&& CPU_FEATURE_USABLE_P (cpu_features, MOVBE)
|
||||
&& CPU_FEATURES_ARCH_P (cpu_features, AVX_Fast_Unaligned_Load))
|
||||
- return OPTIMIZE (avx2_movbe);
|
||||
+ {
|
||||
+ if (CPU_FEATURE_USABLE_P (cpu_features, AVX512VL)
|
||||
+ && CPU_FEATURE_USABLE_P (cpu_features, AVX512BW))
|
||||
+ return OPTIMIZE (evex_movbe);
|
||||
+
|
||||
+ if (!CPU_FEATURES_ARCH_P (cpu_features, Prefer_No_VZEROUPPER))
|
||||
+ return OPTIMIZE (avx2_movbe);
|
||||
+ }
|
||||
|
||||
if (CPU_FEATURE_USABLE_P (cpu_features, SSE4_1))
|
||||
return OPTIMIZE (sse4_1);
|
||||
diff --git a/sysdeps/x86_64/multiarch/memcmp-evex-movbe.S b/sysdeps/x86_64/multiarch/memcmp-evex-movbe.S
|
||||
new file mode 100644
|
||||
index 00000000..9c093972
|
||||
--- /dev/null
|
||||
+++ b/sysdeps/x86_64/multiarch/memcmp-evex-movbe.S
|
||||
@@ -0,0 +1,440 @@
|
||||
+/* memcmp/wmemcmp optimized with 256-bit EVEX instructions.
|
||||
+ Copyright (C) 2021 Free Software Foundation, Inc.
|
||||
+ This file is part of the GNU C Library.
|
||||
+
|
||||
+ The GNU C Library is free software; you can redistribute it and/or
|
||||
+ modify it under the terms of the GNU Lesser General Public
|
||||
+ License as published by the Free Software Foundation; either
|
||||
+ version 2.1 of the License, or (at your option) any later version.
|
||||
+
|
||||
+ The GNU C Library is distributed in the hope that it will be useful,
|
||||
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
||||
+ Lesser General Public License for more details.
|
||||
+
|
||||
+ You should have received a copy of the GNU Lesser General Public
|
||||
+ License along with the GNU C Library; if not, see
|
||||
+ <https://www.gnu.org/licenses/>. */
|
||||
+
|
||||
+#if IS_IN (libc)
|
||||
+
|
||||
+/* memcmp/wmemcmp is implemented as:
|
||||
+ 1. For size from 2 to 7 bytes, load as big endian with movbe and bswap
|
||||
+ to avoid branches.
|
||||
+ 2. Use overlapping compare to avoid branch.
|
||||
+ 3. Use vector compare when size >= 4 bytes for memcmp or size >= 8
|
||||
+ bytes for wmemcmp.
|
||||
+ 4. If size is 8 * VEC_SIZE or less, unroll the loop.
|
||||
+ 5. Compare 4 * VEC_SIZE at a time with the aligned first memory
|
||||
+ area.
|
||||
+ 6. Use 2 vector compares when size is 2 * VEC_SIZE or less.
|
||||
+ 7. Use 4 vector compares when size is 4 * VEC_SIZE or less.
|
||||
+ 8. Use 8 vector compares when size is 8 * VEC_SIZE or less. */
|
||||
+
|
||||
+# include <sysdep.h>
|
||||
+
|
||||
+# ifndef MEMCMP
|
||||
+# define MEMCMP __memcmp_evex_movbe
|
||||
+# endif
|
||||
+
|
||||
+# define VMOVU vmovdqu64
|
||||
+
|
||||
+# ifdef USE_AS_WMEMCMP
|
||||
+# define VPCMPEQ vpcmpeqd
|
||||
+# else
|
||||
+# define VPCMPEQ vpcmpeqb
|
||||
+# endif
|
||||
+
|
||||
+# define XMM1 xmm17
|
||||
+# define XMM2 xmm18
|
||||
+# define YMM1 ymm17
|
||||
+# define YMM2 ymm18
|
||||
+# define YMM3 ymm19
|
||||
+# define YMM4 ymm20
|
||||
+# define YMM5 ymm21
|
||||
+# define YMM6 ymm22
|
||||
+
|
||||
+# define VEC_SIZE 32
|
||||
+# ifdef USE_AS_WMEMCMP
|
||||
+# define VEC_MASK 0xff
|
||||
+# define XMM_MASK 0xf
|
||||
+# else
|
||||
+# define VEC_MASK 0xffffffff
|
||||
+# define XMM_MASK 0xffff
|
||||
+# endif
|
||||
+
|
||||
+/* Warning!
|
||||
+ wmemcmp has to use SIGNED comparison for elements.
|
||||
+ memcmp has to use UNSIGNED comparison for elemnts.
|
||||
+*/
|
||||
+
|
||||
+ .section .text.evex,"ax",@progbits
|
||||
+ENTRY (MEMCMP)
|
||||
+# ifdef USE_AS_WMEMCMP
|
||||
+ shl $2, %RDX_LP
|
||||
+# elif defined __ILP32__
|
||||
+ /* Clear the upper 32 bits. */
|
||||
+ movl %edx, %edx
|
||||
+# endif
|
||||
+ cmp $VEC_SIZE, %RDX_LP
|
||||
+ jb L(less_vec)
|
||||
+
|
||||
+ /* From VEC to 2 * VEC. No branch when size == VEC_SIZE. */
|
||||
+ VMOVU (%rsi), %YMM2
|
||||
+ VPCMPEQ (%rdi), %YMM2, %k1
|
||||
+ kmovd %k1, %eax
|
||||
+ subl $VEC_MASK, %eax
|
||||
+ jnz L(first_vec)
|
||||
+
|
||||
+ cmpq $(VEC_SIZE * 2), %rdx
|
||||
+ jbe L(last_vec)
|
||||
+
|
||||
+ /* More than 2 * VEC. */
|
||||
+ cmpq $(VEC_SIZE * 8), %rdx
|
||||
+ ja L(more_8x_vec)
|
||||
+ cmpq $(VEC_SIZE * 4), %rdx
|
||||
+ jb L(last_4x_vec)
|
||||
+
|
||||
+ /* From 4 * VEC to 8 * VEC, inclusively. */
|
||||
+ VMOVU (%rsi), %YMM1
|
||||
+ VPCMPEQ (%rdi), %YMM1, %k1
|
||||
+
|
||||
+ VMOVU VEC_SIZE(%rsi), %YMM2
|
||||
+ VPCMPEQ VEC_SIZE(%rdi), %YMM2, %k2
|
||||
+
|
||||
+ VMOVU (VEC_SIZE * 2)(%rsi), %YMM3
|
||||
+ VPCMPEQ (VEC_SIZE * 2)(%rdi), %YMM3, %k3
|
||||
+
|
||||
+ VMOVU (VEC_SIZE * 3)(%rsi), %YMM4
|
||||
+ VPCMPEQ (VEC_SIZE * 3)(%rdi), %YMM4, %k4
|
||||
+
|
||||
+ kandd %k1, %k2, %k5
|
||||
+ kandd %k3, %k4, %k6
|
||||
+ kandd %k5, %k6, %k6
|
||||
+
|
||||
+ kmovd %k6, %eax
|
||||
+ cmpl $VEC_MASK, %eax
|
||||
+ jne L(4x_vec_end)
|
||||
+
|
||||
+ leaq -(4 * VEC_SIZE)(%rdi, %rdx), %rdi
|
||||
+ leaq -(4 * VEC_SIZE)(%rsi, %rdx), %rsi
|
||||
+ VMOVU (%rsi), %YMM1
|
||||
+ VPCMPEQ (%rdi), %YMM1, %k1
|
||||
+
|
||||
+ VMOVU VEC_SIZE(%rsi), %YMM2
|
||||
+ VPCMPEQ VEC_SIZE(%rdi), %YMM2, %k2
|
||||
+ kandd %k1, %k2, %k5
|
||||
+
|
||||
+ VMOVU (VEC_SIZE * 2)(%rsi), %YMM3
|
||||
+ VPCMPEQ (VEC_SIZE * 2)(%rdi), %YMM3, %k3
|
||||
+ kandd %k3, %k5, %k5
|
||||
+
|
||||
+ VMOVU (VEC_SIZE * 3)(%rsi), %YMM4
|
||||
+ VPCMPEQ (VEC_SIZE * 3)(%rdi), %YMM4, %k4
|
||||
+ kandd %k4, %k5, %k5
|
||||
+
|
||||
+ kmovd %k5, %eax
|
||||
+ cmpl $VEC_MASK, %eax
|
||||
+ jne L(4x_vec_end)
|
||||
+ xorl %eax, %eax
|
||||
+ ret
|
||||
+
|
||||
+ .p2align 4
|
||||
+L(last_2x_vec):
|
||||
+ /* From VEC to 2 * VEC. No branch when size == VEC_SIZE. */
|
||||
+ VMOVU (%rsi), %YMM2
|
||||
+ VPCMPEQ (%rdi), %YMM2, %k2
|
||||
+ kmovd %k2, %eax
|
||||
+ subl $VEC_MASK, %eax
|
||||
+ jnz L(first_vec)
|
||||
+
|
||||
+L(last_vec):
|
||||
+ /* Use overlapping loads to avoid branches. */
|
||||
+ leaq -VEC_SIZE(%rdi, %rdx), %rdi
|
||||
+ leaq -VEC_SIZE(%rsi, %rdx), %rsi
|
||||
+ VMOVU (%rsi), %YMM2
|
||||
+ VPCMPEQ (%rdi), %YMM2, %k2
|
||||
+ kmovd %k2, %eax
|
||||
+ subl $VEC_MASK, %eax
|
||||
+ jnz L(first_vec)
|
||||
+ ret
|
||||
+
|
||||
+ .p2align 4
|
||||
+L(first_vec):
|
||||
+ /* A byte or int32 is different within 16 or 32 bytes. */
|
||||
+ tzcntl %eax, %ecx
|
||||
+# ifdef USE_AS_WMEMCMP
|
||||
+ xorl %eax, %eax
|
||||
+ movl (%rdi, %rcx, 4), %edx
|
||||
+ cmpl (%rsi, %rcx, 4), %edx
|
||||
+L(wmemcmp_return):
|
||||
+ setl %al
|
||||
+ negl %eax
|
||||
+ orl $1, %eax
|
||||
+# else
|
||||
+ movzbl (%rdi, %rcx), %eax
|
||||
+ movzbl (%rsi, %rcx), %edx
|
||||
+ sub %edx, %eax
|
||||
+# endif
|
||||
+ ret
|
||||
+
|
||||
+# ifdef USE_AS_WMEMCMP
|
||||
+ .p2align 4
|
||||
+L(4):
|
||||
+ xorl %eax, %eax
|
||||
+ movl (%rdi), %edx
|
||||
+ cmpl (%rsi), %edx
|
||||
+ jne L(wmemcmp_return)
|
||||
+ ret
|
||||
+# else
|
||||
+ .p2align 4
|
||||
+L(between_4_7):
|
||||
+ /* Load as big endian with overlapping movbe to avoid branches. */
|
||||
+ movbe (%rdi), %eax
|
||||
+ movbe (%rsi), %ecx
|
||||
+ shlq $32, %rax
|
||||
+ shlq $32, %rcx
|
||||
+ movbe -4(%rdi, %rdx), %edi
|
||||
+ movbe -4(%rsi, %rdx), %esi
|
||||
+ orq %rdi, %rax
|
||||
+ orq %rsi, %rcx
|
||||
+ subq %rcx, %rax
|
||||
+ je L(exit)
|
||||
+ sbbl %eax, %eax
|
||||
+ orl $1, %eax
|
||||
+ ret
|
||||
+
|
||||
+ .p2align 4
|
||||
+L(exit):
|
||||
+ ret
|
||||
+
|
||||
+ .p2align 4
|
||||
+L(between_2_3):
|
||||
+ /* Load as big endian to avoid branches. */
|
||||
+ movzwl (%rdi), %eax
|
||||
+ movzwl (%rsi), %ecx
|
||||
+ shll $8, %eax
|
||||
+ shll $8, %ecx
|
||||
+ bswap %eax
|
||||
+ bswap %ecx
|
||||
+ movb -1(%rdi, %rdx), %al
|
||||
+ movb -1(%rsi, %rdx), %cl
|
||||
+ /* Subtraction is okay because the upper 8 bits are zero. */
|
||||
+ subl %ecx, %eax
|
||||
+ ret
|
||||
+
|
||||
+ .p2align 4
|
||||
+L(1):
|
||||
+ movzbl (%rdi), %eax
|
||||
+ movzbl (%rsi), %ecx
|
||||
+ subl %ecx, %eax
|
||||
+ ret
|
||||
+# endif
|
||||
+
|
||||
+ .p2align 4
|
||||
+L(zero):
|
||||
+ xorl %eax, %eax
|
||||
+ ret
|
||||
+
|
||||
+ .p2align 4
|
||||
+L(less_vec):
|
||||
+# ifdef USE_AS_WMEMCMP
|
||||
+ /* It can only be 0, 4, 8, 12, 16, 20, 24, 28 bytes. */
|
||||
+ cmpb $4, %dl
|
||||
+ je L(4)
|
||||
+ jb L(zero)
|
||||
+# else
|
||||
+ cmpb $1, %dl
|
||||
+ je L(1)
|
||||
+ jb L(zero)
|
||||
+ cmpb $4, %dl
|
||||
+ jb L(between_2_3)
|
||||
+ cmpb $8, %dl
|
||||
+ jb L(between_4_7)
|
||||
+# endif
|
||||
+ cmpb $16, %dl
|
||||
+ jae L(between_16_31)
|
||||
+ /* It is between 8 and 15 bytes. */
|
||||
+ vmovq (%rdi), %XMM1
|
||||
+ vmovq (%rsi), %XMM2
|
||||
+ VPCMPEQ %XMM1, %XMM2, %k2
|
||||
+ kmovw %k2, %eax
|
||||
+ subl $XMM_MASK, %eax
|
||||
+ jnz L(first_vec)
|
||||
+ /* Use overlapping loads to avoid branches. */
|
||||
+ leaq -8(%rdi, %rdx), %rdi
|
||||
+ leaq -8(%rsi, %rdx), %rsi
|
||||
+ vmovq (%rdi), %XMM1
|
||||
+ vmovq (%rsi), %XMM2
|
||||
+ VPCMPEQ %XMM1, %XMM2, %k2
|
||||
+ kmovw %k2, %eax
|
||||
+ subl $XMM_MASK, %eax
|
||||
+ jnz L(first_vec)
|
||||
+ ret
|
||||
+
|
||||
+ .p2align 4
|
||||
+L(between_16_31):
|
||||
+ /* From 16 to 31 bytes. No branch when size == 16. */
|
||||
+ VMOVU (%rsi), %XMM2
|
||||
+ VPCMPEQ (%rdi), %XMM2, %k2
|
||||
+ kmovw %k2, %eax
|
||||
+ subl $XMM_MASK, %eax
|
||||
+ jnz L(first_vec)
|
||||
+
|
||||
+ /* Use overlapping loads to avoid branches. */
|
||||
+ leaq -16(%rdi, %rdx), %rdi
|
||||
+ leaq -16(%rsi, %rdx), %rsi
|
||||
+ VMOVU (%rsi), %XMM2
|
||||
+ VPCMPEQ (%rdi), %XMM2, %k2
|
||||
+ kmovw %k2, %eax
|
||||
+ subl $XMM_MASK, %eax
|
||||
+ jnz L(first_vec)
|
||||
+ ret
|
||||
+
|
||||
+ .p2align 4
|
||||
+L(more_8x_vec):
|
||||
+ /* More than 8 * VEC. Check the first VEC. */
|
||||
+ VMOVU (%rsi), %YMM2
|
||||
+ VPCMPEQ (%rdi), %YMM2, %k2
|
||||
+ kmovd %k2, %eax
|
||||
+ subl $VEC_MASK, %eax
|
||||
+ jnz L(first_vec)
|
||||
+
|
||||
+ /* Align the first memory area for aligned loads in the loop.
|
||||
+ Compute how much the first memory area is misaligned. */
|
||||
+ movq %rdi, %rcx
|
||||
+ andl $(VEC_SIZE - 1), %ecx
|
||||
+ /* Get the negative of offset for alignment. */
|
||||
+ subq $VEC_SIZE, %rcx
|
||||
+ /* Adjust the second memory area. */
|
||||
+ subq %rcx, %rsi
|
||||
+ /* Adjust the first memory area which should be aligned now. */
|
||||
+ subq %rcx, %rdi
|
||||
+ /* Adjust length. */
|
||||
+ addq %rcx, %rdx
|
||||
+
|
||||
+L(loop_4x_vec):
|
||||
+ /* Compare 4 * VEC at a time forward. */
|
||||
+ VMOVU (%rsi), %YMM1
|
||||
+ VPCMPEQ (%rdi), %YMM1, %k1
|
||||
+
|
||||
+ VMOVU VEC_SIZE(%rsi), %YMM2
|
||||
+ VPCMPEQ VEC_SIZE(%rdi), %YMM2, %k2
|
||||
+ kandd %k2, %k1, %k5
|
||||
+
|
||||
+ VMOVU (VEC_SIZE * 2)(%rsi), %YMM3
|
||||
+ VPCMPEQ (VEC_SIZE * 2)(%rdi), %YMM3, %k3
|
||||
+ kandd %k3, %k5, %k5
|
||||
+
|
||||
+ VMOVU (VEC_SIZE * 3)(%rsi), %YMM4
|
||||
+ VPCMPEQ (VEC_SIZE * 3)(%rdi), %YMM4, %k4
|
||||
+ kandd %k4, %k5, %k5
|
||||
+
|
||||
+ kmovd %k5, %eax
|
||||
+ cmpl $VEC_MASK, %eax
|
||||
+ jne L(4x_vec_end)
|
||||
+
|
||||
+ addq $(VEC_SIZE * 4), %rdi
|
||||
+ addq $(VEC_SIZE * 4), %rsi
|
||||
+
|
||||
+ subq $(VEC_SIZE * 4), %rdx
|
||||
+ cmpq $(VEC_SIZE * 4), %rdx
|
||||
+ jae L(loop_4x_vec)
|
||||
+
|
||||
+ /* Less than 4 * VEC. */
|
||||
+ cmpq $VEC_SIZE, %rdx
|
||||
+ jbe L(last_vec)
|
||||
+ cmpq $(VEC_SIZE * 2), %rdx
|
||||
+ jbe L(last_2x_vec)
|
||||
+
|
||||
+L(last_4x_vec):
|
||||
+ /* From 2 * VEC to 4 * VEC. */
|
||||
+ VMOVU (%rsi), %YMM2
|
||||
+ VPCMPEQ (%rdi), %YMM2, %k2
|
||||
+ kmovd %k2, %eax
|
||||
+ subl $VEC_MASK, %eax
|
||||
+ jnz L(first_vec)
|
||||
+
|
||||
+ addq $VEC_SIZE, %rdi
|
||||
+ addq $VEC_SIZE, %rsi
|
||||
+ VMOVU (%rsi), %YMM2
|
||||
+ VPCMPEQ (%rdi), %YMM2, %k2
|
||||
+ kmovd %k2, %eax
|
||||
+ subl $VEC_MASK, %eax
|
||||
+ jnz L(first_vec)
|
||||
+
|
||||
+ /* Use overlapping loads to avoid branches. */
|
||||
+ leaq -(3 * VEC_SIZE)(%rdi, %rdx), %rdi
|
||||
+ leaq -(3 * VEC_SIZE)(%rsi, %rdx), %rsi
|
||||
+ VMOVU (%rsi), %YMM2
|
||||
+ VPCMPEQ (%rdi), %YMM2, %k2
|
||||
+ kmovd %k2, %eax
|
||||
+ subl $VEC_MASK, %eax
|
||||
+ jnz L(first_vec)
|
||||
+
|
||||
+ addq $VEC_SIZE, %rdi
|
||||
+ addq $VEC_SIZE, %rsi
|
||||
+ VMOVU (%rsi), %YMM2
|
||||
+ VPCMPEQ (%rdi), %YMM2, %k2
|
||||
+ kmovd %k2, %eax
|
||||
+ subl $VEC_MASK, %eax
|
||||
+ jnz L(first_vec)
|
||||
+ ret
|
||||
+
|
||||
+ .p2align 4
|
||||
+L(4x_vec_end):
|
||||
+ kmovd %k1, %eax
|
||||
+ subl $VEC_MASK, %eax
|
||||
+ jnz L(first_vec)
|
||||
+ kmovd %k2, %eax
|
||||
+ subl $VEC_MASK, %eax
|
||||
+ jnz L(first_vec_x1)
|
||||
+ kmovd %k3, %eax
|
||||
+ subl $VEC_MASK, %eax
|
||||
+ jnz L(first_vec_x2)
|
||||
+ kmovd %k4, %eax
|
||||
+ subl $VEC_MASK, %eax
|
||||
+ tzcntl %eax, %ecx
|
||||
+# ifdef USE_AS_WMEMCMP
|
||||
+ xorl %eax, %eax
|
||||
+ movl (VEC_SIZE * 3)(%rdi, %rcx, 4), %edx
|
||||
+ cmpl (VEC_SIZE * 3)(%rsi, %rcx, 4), %edx
|
||||
+ jmp L(wmemcmp_return)
|
||||
+# else
|
||||
+ movzbl (VEC_SIZE * 3)(%rdi, %rcx), %eax
|
||||
+ movzbl (VEC_SIZE * 3)(%rsi, %rcx), %edx
|
||||
+ sub %edx, %eax
|
||||
+# endif
|
||||
+ ret
|
||||
+
|
||||
+ .p2align 4
|
||||
+L(first_vec_x1):
|
||||
+ tzcntl %eax, %ecx
|
||||
+# ifdef USE_AS_WMEMCMP
|
||||
+ xorl %eax, %eax
|
||||
+ movl VEC_SIZE(%rdi, %rcx, 4), %edx
|
||||
+ cmpl VEC_SIZE(%rsi, %rcx, 4), %edx
|
||||
+ jmp L(wmemcmp_return)
|
||||
+# else
|
||||
+ movzbl VEC_SIZE(%rdi, %rcx), %eax
|
||||
+ movzbl VEC_SIZE(%rsi, %rcx), %edx
|
||||
+ sub %edx, %eax
|
||||
+# endif
|
||||
+ ret
|
||||
+
|
||||
+ .p2align 4
|
||||
+L(first_vec_x2):
|
||||
+ tzcntl %eax, %ecx
|
||||
+# ifdef USE_AS_WMEMCMP
|
||||
+ xorl %eax, %eax
|
||||
+ movl (VEC_SIZE * 2)(%rdi, %rcx, 4), %edx
|
||||
+ cmpl (VEC_SIZE * 2)(%rsi, %rcx, 4), %edx
|
||||
+ jmp L(wmemcmp_return)
|
||||
+# else
|
||||
+ movzbl (VEC_SIZE * 2)(%rdi, %rcx), %eax
|
||||
+ movzbl (VEC_SIZE * 2)(%rsi, %rcx), %edx
|
||||
+ sub %edx, %eax
|
||||
+# endif
|
||||
+ ret
|
||||
+END (MEMCMP)
|
||||
+#endif
|
||||
diff --git a/sysdeps/x86_64/multiarch/wmemcmp-evex-movbe.S b/sysdeps/x86_64/multiarch/wmemcmp-evex-movbe.S
|
||||
new file mode 100644
|
||||
index 00000000..4726d74a
|
||||
--- /dev/null
|
||||
+++ b/sysdeps/x86_64/multiarch/wmemcmp-evex-movbe.S
|
||||
@@ -0,0 +1,4 @@
|
||||
+#define MEMCMP __wmemcmp_evex_movbe
|
||||
+#define USE_AS_WMEMCMP 1
|
||||
+
|
||||
+#include "memcmp-evex-movbe.S"
|
||||
--
|
||||
GitLab
|
||||
|
||||
File diff suppressed because it is too large
Load Diff
@ -1,735 +0,0 @@
|
||||
From 4bd660be40967cd69072f69ebc2ad32bfcc1f206 Mon Sep 17 00:00:00 2001
|
||||
From: "H.J. Lu" <hjl.tools@gmail.com>
|
||||
Date: Tue, 23 Feb 2021 06:33:10 -0800
|
||||
Subject: [PATCH] x86: Add string/memory function tests in RTM region
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
At function exit, AVX optimized string/memory functions have VZEROUPPER
|
||||
which triggers RTM abort. When such functions are called inside a
|
||||
transactionally executing RTM region, RTM abort causes severe performance
|
||||
degradation. Add tests to verify that string/memory functions won't
|
||||
cause RTM abort in RTM region.
|
||||
---
|
||||
sysdeps/x86/Makefile | 23 +++++++++++
|
||||
sysdeps/x86/tst-memchr-rtm.c | 54 ++++++++++++++++++++++++++
|
||||
sysdeps/x86/tst-memcmp-rtm.c | 52 +++++++++++++++++++++++++
|
||||
sysdeps/x86/tst-memmove-rtm.c | 53 ++++++++++++++++++++++++++
|
||||
sysdeps/x86/tst-memrchr-rtm.c | 54 ++++++++++++++++++++++++++
|
||||
sysdeps/x86/tst-memset-rtm.c | 45 ++++++++++++++++++++++
|
||||
sysdeps/x86/tst-strchr-rtm.c | 54 ++++++++++++++++++++++++++
|
||||
sysdeps/x86/tst-strcpy-rtm.c | 53 ++++++++++++++++++++++++++
|
||||
sysdeps/x86/tst-string-rtm.h | 72 +++++++++++++++++++++++++++++++++++
|
||||
sysdeps/x86/tst-strlen-rtm.c | 53 ++++++++++++++++++++++++++
|
||||
sysdeps/x86/tst-strncmp-rtm.c | 52 +++++++++++++++++++++++++
|
||||
sysdeps/x86/tst-strrchr-rtm.c | 53 ++++++++++++++++++++++++++
|
||||
12 files changed, 618 insertions(+)
|
||||
create mode 100644 sysdeps/x86/tst-memchr-rtm.c
|
||||
create mode 100644 sysdeps/x86/tst-memcmp-rtm.c
|
||||
create mode 100644 sysdeps/x86/tst-memmove-rtm.c
|
||||
create mode 100644 sysdeps/x86/tst-memrchr-rtm.c
|
||||
create mode 100644 sysdeps/x86/tst-memset-rtm.c
|
||||
create mode 100644 sysdeps/x86/tst-strchr-rtm.c
|
||||
create mode 100644 sysdeps/x86/tst-strcpy-rtm.c
|
||||
create mode 100644 sysdeps/x86/tst-string-rtm.h
|
||||
create mode 100644 sysdeps/x86/tst-strlen-rtm.c
|
||||
create mode 100644 sysdeps/x86/tst-strncmp-rtm.c
|
||||
create mode 100644 sysdeps/x86/tst-strrchr-rtm.c
|
||||
|
||||
diff --git a/sysdeps/x86/Makefile b/sysdeps/x86/Makefile
|
||||
index 59e928e9..5be71ada 100644
|
||||
--- a/sysdeps/x86/Makefile
|
||||
+++ b/sysdeps/x86/Makefile
|
||||
@@ -17,6 +17,29 @@ endif
|
||||
|
||||
ifeq ($(subdir),string)
|
||||
sysdep_routines += cacheinfo
|
||||
+
|
||||
+tests += \
|
||||
+ tst-memchr-rtm \
|
||||
+ tst-memcmp-rtm \
|
||||
+ tst-memmove-rtm \
|
||||
+ tst-memrchr-rtm \
|
||||
+ tst-memset-rtm \
|
||||
+ tst-strchr-rtm \
|
||||
+ tst-strcpy-rtm \
|
||||
+ tst-strlen-rtm \
|
||||
+ tst-strncmp-rtm \
|
||||
+ tst-strrchr-rtm
|
||||
+
|
||||
+CFLAGS-tst-memchr-rtm.c += -mrtm
|
||||
+CFLAGS-tst-memcmp-rtm.c += -mrtm
|
||||
+CFLAGS-tst-memmove-rtm.c += -mrtm
|
||||
+CFLAGS-tst-memrchr-rtm.c += -mrtm
|
||||
+CFLAGS-tst-memset-rtm.c += -mrtm
|
||||
+CFLAGS-tst-strchr-rtm.c += -mrtm
|
||||
+CFLAGS-tst-strcpy-rtm.c += -mrtm
|
||||
+CFLAGS-tst-strlen-rtm.c += -mrtm
|
||||
+CFLAGS-tst-strncmp-rtm.c += -mrtm
|
||||
+CFLAGS-tst-strrchr-rtm.c += -mrtm
|
||||
endif
|
||||
|
||||
ifneq ($(enable-cet),no)
|
||||
diff --git a/sysdeps/x86/tst-memchr-rtm.c b/sysdeps/x86/tst-memchr-rtm.c
|
||||
new file mode 100644
|
||||
index 00000000..e4749401
|
||||
--- /dev/null
|
||||
+++ b/sysdeps/x86/tst-memchr-rtm.c
|
||||
@@ -0,0 +1,54 @@
|
||||
+/* Test case for memchr inside a transactionally executing RTM region.
|
||||
+ Copyright (C) 2021 Free Software Foundation, Inc.
|
||||
+ This file is part of the GNU C Library.
|
||||
+
|
||||
+ The GNU C Library is free software; you can redistribute it and/or
|
||||
+ modify it under the terms of the GNU Lesser General Public
|
||||
+ License as published by the Free Software Foundation; either
|
||||
+ version 2.1 of the License, or (at your option) any later version.
|
||||
+
|
||||
+ The GNU C Library is distributed in the hope that it will be useful,
|
||||
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
||||
+ Lesser General Public License for more details.
|
||||
+
|
||||
+ You should have received a copy of the GNU Lesser General Public
|
||||
+ License along with the GNU C Library; if not, see
|
||||
+ <https://www.gnu.org/licenses/>. */
|
||||
+
|
||||
+#include <tst-string-rtm.h>
|
||||
+
|
||||
+#define LOOP 3000
|
||||
+#define STRING_SIZE 1024
|
||||
+char string1[STRING_SIZE];
|
||||
+
|
||||
+__attribute__ ((noinline, noclone))
|
||||
+static int
|
||||
+prepare (void)
|
||||
+{
|
||||
+ memset (string1, 'a', STRING_SIZE);
|
||||
+ string1[100] = 'c';
|
||||
+ string1[STRING_SIZE - 100] = 'c';
|
||||
+ char *p = memchr (string1, 'c', STRING_SIZE);
|
||||
+ if (p == &string1[100])
|
||||
+ return EXIT_SUCCESS;
|
||||
+ else
|
||||
+ return EXIT_FAILURE;
|
||||
+}
|
||||
+
|
||||
+__attribute__ ((noinline, noclone))
|
||||
+static int
|
||||
+function (void)
|
||||
+{
|
||||
+ char *p = memchr (string1, 'c', STRING_SIZE);
|
||||
+ if (p == &string1[100])
|
||||
+ return 0;
|
||||
+ else
|
||||
+ return 1;
|
||||
+}
|
||||
+
|
||||
+static int
|
||||
+do_test (void)
|
||||
+{
|
||||
+ return do_test_1 ("memchr", LOOP, prepare, function);
|
||||
+}
|
||||
diff --git a/sysdeps/x86/tst-memcmp-rtm.c b/sysdeps/x86/tst-memcmp-rtm.c
|
||||
new file mode 100644
|
||||
index 00000000..e4c8a623
|
||||
--- /dev/null
|
||||
+++ b/sysdeps/x86/tst-memcmp-rtm.c
|
||||
@@ -0,0 +1,52 @@
|
||||
+/* Test case for memcmp inside a transactionally executing RTM region.
|
||||
+ Copyright (C) 2021 Free Software Foundation, Inc.
|
||||
+ This file is part of the GNU C Library.
|
||||
+
|
||||
+ The GNU C Library is free software; you can redistribute it and/or
|
||||
+ modify it under the terms of the GNU Lesser General Public
|
||||
+ License as published by the Free Software Foundation; either
|
||||
+ version 2.1 of the License, or (at your option) any later version.
|
||||
+
|
||||
+ The GNU C Library is distributed in the hope that it will be useful,
|
||||
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
||||
+ Lesser General Public License for more details.
|
||||
+
|
||||
+ You should have received a copy of the GNU Lesser General Public
|
||||
+ License along with the GNU C Library; if not, see
|
||||
+ <https://www.gnu.org/licenses/>. */
|
||||
+
|
||||
+#include <tst-string-rtm.h>
|
||||
+
|
||||
+#define LOOP 3000
|
||||
+#define STRING_SIZE 1024
|
||||
+char string1[STRING_SIZE];
|
||||
+char string2[STRING_SIZE];
|
||||
+
|
||||
+__attribute__ ((noinline, noclone))
|
||||
+static int
|
||||
+prepare (void)
|
||||
+{
|
||||
+ memset (string1, 'a', STRING_SIZE);
|
||||
+ memset (string2, 'a', STRING_SIZE);
|
||||
+ if (memcmp (string1, string2, STRING_SIZE) == 0)
|
||||
+ return EXIT_SUCCESS;
|
||||
+ else
|
||||
+ return EXIT_FAILURE;
|
||||
+}
|
||||
+
|
||||
+__attribute__ ((noinline, noclone))
|
||||
+static int
|
||||
+function (void)
|
||||
+{
|
||||
+ if (memcmp (string1, string2, STRING_SIZE) == 0)
|
||||
+ return 0;
|
||||
+ else
|
||||
+ return 1;
|
||||
+}
|
||||
+
|
||||
+static int
|
||||
+do_test (void)
|
||||
+{
|
||||
+ return do_test_1 ("memcmp", LOOP, prepare, function);
|
||||
+}
|
||||
diff --git a/sysdeps/x86/tst-memmove-rtm.c b/sysdeps/x86/tst-memmove-rtm.c
|
||||
new file mode 100644
|
||||
index 00000000..4bf97ef1
|
||||
--- /dev/null
|
||||
+++ b/sysdeps/x86/tst-memmove-rtm.c
|
||||
@@ -0,0 +1,53 @@
|
||||
+/* Test case for memmove inside a transactionally executing RTM region.
|
||||
+ Copyright (C) 2021 Free Software Foundation, Inc.
|
||||
+ This file is part of the GNU C Library.
|
||||
+
|
||||
+ The GNU C Library is free software; you can redistribute it and/or
|
||||
+ modify it under the terms of the GNU Lesser General Public
|
||||
+ License as published by the Free Software Foundation; either
|
||||
+ version 2.1 of the License, or (at your option) any later version.
|
||||
+
|
||||
+ The GNU C Library is distributed in the hope that it will be useful,
|
||||
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
||||
+ Lesser General Public License for more details.
|
||||
+
|
||||
+ You should have received a copy of the GNU Lesser General Public
|
||||
+ License along with the GNU C Library; if not, see
|
||||
+ <https://www.gnu.org/licenses/>. */
|
||||
+
|
||||
+#include <tst-string-rtm.h>
|
||||
+
|
||||
+#define LOOP 3000
|
||||
+#define STRING_SIZE 1024
|
||||
+char string1[STRING_SIZE];
|
||||
+char string2[STRING_SIZE];
|
||||
+
|
||||
+__attribute__ ((noinline, noclone))
|
||||
+static int
|
||||
+prepare (void)
|
||||
+{
|
||||
+ memset (string1, 'a', STRING_SIZE);
|
||||
+ if (memmove (string2, string1, STRING_SIZE) == string2
|
||||
+ && memcmp (string2, string1, STRING_SIZE) == 0)
|
||||
+ return EXIT_SUCCESS;
|
||||
+ else
|
||||
+ return EXIT_FAILURE;
|
||||
+}
|
||||
+
|
||||
+__attribute__ ((noinline, noclone))
|
||||
+static int
|
||||
+function (void)
|
||||
+{
|
||||
+ if (memmove (string2, string1, STRING_SIZE) == string2
|
||||
+ && memcmp (string2, string1, STRING_SIZE) == 0)
|
||||
+ return 0;
|
||||
+ else
|
||||
+ return 1;
|
||||
+}
|
||||
+
|
||||
+static int
|
||||
+do_test (void)
|
||||
+{
|
||||
+ return do_test_1 ("memmove", LOOP, prepare, function);
|
||||
+}
|
||||
diff --git a/sysdeps/x86/tst-memrchr-rtm.c b/sysdeps/x86/tst-memrchr-rtm.c
|
||||
new file mode 100644
|
||||
index 00000000..a57a5a8e
|
||||
--- /dev/null
|
||||
+++ b/sysdeps/x86/tst-memrchr-rtm.c
|
||||
@@ -0,0 +1,54 @@
|
||||
+/* Test case for memrchr inside a transactionally executing RTM region.
|
||||
+ Copyright (C) 2021 Free Software Foundation, Inc.
|
||||
+ This file is part of the GNU C Library.
|
||||
+
|
||||
+ The GNU C Library is free software; you can redistribute it and/or
|
||||
+ modify it under the terms of the GNU Lesser General Public
|
||||
+ License as published by the Free Software Foundation; either
|
||||
+ version 2.1 of the License, or (at your option) any later version.
|
||||
+
|
||||
+ The GNU C Library is distributed in the hope that it will be useful,
|
||||
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
||||
+ Lesser General Public License for more details.
|
||||
+
|
||||
+ You should have received a copy of the GNU Lesser General Public
|
||||
+ License along with the GNU C Library; if not, see
|
||||
+ <https://www.gnu.org/licenses/>. */
|
||||
+
|
||||
+#include <tst-string-rtm.h>
|
||||
+
|
||||
+#define LOOP 3000
|
||||
+#define STRING_SIZE 1024
|
||||
+char string1[STRING_SIZE];
|
||||
+
|
||||
+__attribute__ ((noinline, noclone))
|
||||
+static int
|
||||
+prepare (void)
|
||||
+{
|
||||
+ memset (string1, 'a', STRING_SIZE);
|
||||
+ string1[100] = 'c';
|
||||
+ string1[STRING_SIZE - 100] = 'c';
|
||||
+ char *p = memrchr (string1, 'c', STRING_SIZE);
|
||||
+ if (p == &string1[STRING_SIZE - 100])
|
||||
+ return EXIT_SUCCESS;
|
||||
+ else
|
||||
+ return EXIT_FAILURE;
|
||||
+}
|
||||
+
|
||||
+__attribute__ ((noinline, noclone))
|
||||
+static int
|
||||
+function (void)
|
||||
+{
|
||||
+ char *p = memrchr (string1, 'c', STRING_SIZE);
|
||||
+ if (p == &string1[STRING_SIZE - 100])
|
||||
+ return 0;
|
||||
+ else
|
||||
+ return 1;
|
||||
+}
|
||||
+
|
||||
+static int
|
||||
+do_test (void)
|
||||
+{
|
||||
+ return do_test_1 ("memrchr", LOOP, prepare, function);
|
||||
+}
|
||||
diff --git a/sysdeps/x86/tst-memset-rtm.c b/sysdeps/x86/tst-memset-rtm.c
|
||||
new file mode 100644
|
||||
index 00000000..bf343a4d
|
||||
--- /dev/null
|
||||
+++ b/sysdeps/x86/tst-memset-rtm.c
|
||||
@@ -0,0 +1,45 @@
|
||||
+/* Test case for memset inside a transactionally executing RTM region.
|
||||
+ Copyright (C) 2021 Free Software Foundation, Inc.
|
||||
+ This file is part of the GNU C Library.
|
||||
+
|
||||
+ The GNU C Library is free software; you can redistribute it and/or
|
||||
+ modify it under the terms of the GNU Lesser General Public
|
||||
+ License as published by the Free Software Foundation; either
|
||||
+ version 2.1 of the License, or (at your option) any later version.
|
||||
+
|
||||
+ The GNU C Library is distributed in the hope that it will be useful,
|
||||
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
||||
+ Lesser General Public License for more details.
|
||||
+
|
||||
+ You should have received a copy of the GNU Lesser General Public
|
||||
+ License along with the GNU C Library; if not, see
|
||||
+ <https://www.gnu.org/licenses/>. */
|
||||
+
|
||||
+#include <tst-string-rtm.h>
|
||||
+
|
||||
+#define LOOP 3000
|
||||
+#define STRING_SIZE 1024
|
||||
+char string1[STRING_SIZE];
|
||||
+
|
||||
+__attribute__ ((noinline, noclone))
|
||||
+static int
|
||||
+prepare (void)
|
||||
+{
|
||||
+ memset (string1, 'a', STRING_SIZE);
|
||||
+ return EXIT_SUCCESS;
|
||||
+}
|
||||
+
|
||||
+__attribute__ ((noinline, noclone))
|
||||
+static int
|
||||
+function (void)
|
||||
+{
|
||||
+ memset (string1, 'a', STRING_SIZE);
|
||||
+ return 0;
|
||||
+}
|
||||
+
|
||||
+static int
|
||||
+do_test (void)
|
||||
+{
|
||||
+ return do_test_1 ("memset", LOOP, prepare, function);
|
||||
+}
|
||||
diff --git a/sysdeps/x86/tst-strchr-rtm.c b/sysdeps/x86/tst-strchr-rtm.c
|
||||
new file mode 100644
|
||||
index 00000000..a82e29c0
|
||||
--- /dev/null
|
||||
+++ b/sysdeps/x86/tst-strchr-rtm.c
|
||||
@@ -0,0 +1,54 @@
|
||||
+/* Test case for strchr inside a transactionally executing RTM region.
|
||||
+ Copyright (C) 2021 Free Software Foundation, Inc.
|
||||
+ This file is part of the GNU C Library.
|
||||
+
|
||||
+ The GNU C Library is free software; you can redistribute it and/or
|
||||
+ modify it under the terms of the GNU Lesser General Public
|
||||
+ License as published by the Free Software Foundation; either
|
||||
+ version 2.1 of the License, or (at your option) any later version.
|
||||
+
|
||||
+ The GNU C Library is distributed in the hope that it will be useful,
|
||||
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
||||
+ Lesser General Public License for more details.
|
||||
+
|
||||
+ You should have received a copy of the GNU Lesser General Public
|
||||
+ License along with the GNU C Library; if not, see
|
||||
+ <https://www.gnu.org/licenses/>. */
|
||||
+
|
||||
+#include <tst-string-rtm.h>
|
||||
+
|
||||
+#define LOOP 3000
|
||||
+#define STRING_SIZE 1024
|
||||
+char string1[STRING_SIZE];
|
||||
+
|
||||
+__attribute__ ((noinline, noclone))
|
||||
+static int
|
||||
+prepare (void)
|
||||
+{
|
||||
+ memset (string1, 'a', STRING_SIZE - 1);
|
||||
+ string1[100] = 'c';
|
||||
+ string1[STRING_SIZE - 100] = 'c';
|
||||
+ char *p = strchr (string1, 'c');
|
||||
+ if (p == &string1[100])
|
||||
+ return EXIT_SUCCESS;
|
||||
+ else
|
||||
+ return EXIT_FAILURE;
|
||||
+}
|
||||
+
|
||||
+__attribute__ ((noinline, noclone))
|
||||
+static int
|
||||
+function (void)
|
||||
+{
|
||||
+ char *p = strchr (string1, 'c');
|
||||
+ if (p == &string1[100])
|
||||
+ return 0;
|
||||
+ else
|
||||
+ return 1;
|
||||
+}
|
||||
+
|
||||
+static int
|
||||
+do_test (void)
|
||||
+{
|
||||
+ return do_test_1 ("strchr", LOOP, prepare, function);
|
||||
+}
|
||||
diff --git a/sysdeps/x86/tst-strcpy-rtm.c b/sysdeps/x86/tst-strcpy-rtm.c
|
||||
new file mode 100644
|
||||
index 00000000..2b2a583f
|
||||
--- /dev/null
|
||||
+++ b/sysdeps/x86/tst-strcpy-rtm.c
|
||||
@@ -0,0 +1,53 @@
|
||||
+/* Test case for strcpy inside a transactionally executing RTM region.
|
||||
+ Copyright (C) 2021 Free Software Foundation, Inc.
|
||||
+ This file is part of the GNU C Library.
|
||||
+
|
||||
+ The GNU C Library is free software; you can redistribute it and/or
|
||||
+ modify it under the terms of the GNU Lesser General Public
|
||||
+ License as published by the Free Software Foundation; either
|
||||
+ version 2.1 of the License, or (at your option) any later version.
|
||||
+
|
||||
+ The GNU C Library is distributed in the hope that it will be useful,
|
||||
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
||||
+ Lesser General Public License for more details.
|
||||
+
|
||||
+ You should have received a copy of the GNU Lesser General Public
|
||||
+ License along with the GNU C Library; if not, see
|
||||
+ <https://www.gnu.org/licenses/>. */
|
||||
+
|
||||
+#include <tst-string-rtm.h>
|
||||
+
|
||||
+#define LOOP 3000
|
||||
+#define STRING_SIZE 1024
|
||||
+char string1[STRING_SIZE];
|
||||
+char string2[STRING_SIZE];
|
||||
+
|
||||
+__attribute__ ((noinline, noclone))
|
||||
+static int
|
||||
+prepare (void)
|
||||
+{
|
||||
+ memset (string1, 'a', STRING_SIZE - 1);
|
||||
+ if (strcpy (string2, string1) == string2
|
||||
+ && strcmp (string2, string1) == 0)
|
||||
+ return EXIT_SUCCESS;
|
||||
+ else
|
||||
+ return EXIT_FAILURE;
|
||||
+}
|
||||
+
|
||||
+__attribute__ ((noinline, noclone))
|
||||
+static int
|
||||
+function (void)
|
||||
+{
|
||||
+ if (strcpy (string2, string1) == string2
|
||||
+ && strcmp (string2, string1) == 0)
|
||||
+ return 0;
|
||||
+ else
|
||||
+ return 1;
|
||||
+}
|
||||
+
|
||||
+static int
|
||||
+do_test (void)
|
||||
+{
|
||||
+ return do_test_1 ("strcpy", LOOP, prepare, function);
|
||||
+}
|
||||
diff --git a/sysdeps/x86/tst-string-rtm.h b/sysdeps/x86/tst-string-rtm.h
|
||||
new file mode 100644
|
||||
index 00000000..d2470afa
|
||||
--- /dev/null
|
||||
+++ b/sysdeps/x86/tst-string-rtm.h
|
||||
@@ -0,0 +1,72 @@
|
||||
+/* Test string function in a transactionally executing RTM region.
|
||||
+ Copyright (C) 2021 Free Software Foundation, Inc.
|
||||
+ This file is part of the GNU C Library.
|
||||
+
|
||||
+ The GNU C Library is free software; you can redistribute it and/or
|
||||
+ modify it under the terms of the GNU Lesser General Public
|
||||
+ License as published by the Free Software Foundation; either
|
||||
+ version 2.1 of the License, or (at your option) any later version.
|
||||
+
|
||||
+ The GNU C Library is distributed in the hope that it will be useful,
|
||||
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
||||
+ Lesser General Public License for more details.
|
||||
+
|
||||
+ You should have received a copy of the GNU Lesser General Public
|
||||
+ License along with the GNU C Library; if not, see
|
||||
+ <https://www.gnu.org/licenses/>. */
|
||||
+
|
||||
+#include <string.h>
|
||||
+#include <x86intrin.h>
|
||||
+#include <sys/platform/x86.h>
|
||||
+#include <support/check.h>
|
||||
+#include <support/test-driver.h>
|
||||
+
|
||||
+static int
|
||||
+do_test_1 (const char *name, unsigned int loop, int (*prepare) (void),
|
||||
+ int (*function) (void))
|
||||
+{
|
||||
+ if (!CPU_FEATURE_USABLE (RTM))
|
||||
+ return EXIT_UNSUPPORTED;
|
||||
+
|
||||
+ int status = prepare ();
|
||||
+ if (status != EXIT_SUCCESS)
|
||||
+ return status;
|
||||
+
|
||||
+ unsigned int i;
|
||||
+ unsigned int naborts = 0;
|
||||
+ unsigned int failed = 0;
|
||||
+ for (i = 0; i < loop; i++)
|
||||
+ {
|
||||
+ failed |= function ();
|
||||
+ if (_xbegin() == _XBEGIN_STARTED)
|
||||
+ {
|
||||
+ failed |= function ();
|
||||
+ _xend();
|
||||
+ }
|
||||
+ else
|
||||
+ {
|
||||
+ failed |= function ();
|
||||
+ ++naborts;
|
||||
+ }
|
||||
+ }
|
||||
+
|
||||
+ if (failed)
|
||||
+ FAIL_EXIT1 ("%s() failed", name);
|
||||
+
|
||||
+ if (naborts)
|
||||
+ {
|
||||
+ /* NB: Low single digit (<= 5%) noise-level aborts are normal for
|
||||
+ TSX. */
|
||||
+ double rate = 100 * ((double) naborts) / ((double) loop);
|
||||
+ if (rate > 5)
|
||||
+ FAIL_EXIT1 ("TSX abort rate: %.2f%% (%d out of %d)",
|
||||
+ rate, naborts, loop);
|
||||
+ }
|
||||
+
|
||||
+ return EXIT_SUCCESS;
|
||||
+}
|
||||
+
|
||||
+static int do_test (void);
|
||||
+
|
||||
+#include <support/test-driver.c>
|
||||
diff --git a/sysdeps/x86/tst-strlen-rtm.c b/sysdeps/x86/tst-strlen-rtm.c
|
||||
new file mode 100644
|
||||
index 00000000..0dcf14db
|
||||
--- /dev/null
|
||||
+++ b/sysdeps/x86/tst-strlen-rtm.c
|
||||
@@ -0,0 +1,53 @@
|
||||
+/* Test case for strlen inside a transactionally executing RTM region.
|
||||
+ Copyright (C) 2021 Free Software Foundation, Inc.
|
||||
+ This file is part of the GNU C Library.
|
||||
+
|
||||
+ The GNU C Library is free software; you can redistribute it and/or
|
||||
+ modify it under the terms of the GNU Lesser General Public
|
||||
+ License as published by the Free Software Foundation; either
|
||||
+ version 2.1 of the License, or (at your option) any later version.
|
||||
+
|
||||
+ The GNU C Library is distributed in the hope that it will be useful,
|
||||
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
||||
+ Lesser General Public License for more details.
|
||||
+
|
||||
+ You should have received a copy of the GNU Lesser General Public
|
||||
+ License along with the GNU C Library; if not, see
|
||||
+ <https://www.gnu.org/licenses/>. */
|
||||
+
|
||||
+#include <tst-string-rtm.h>
|
||||
+
|
||||
+#define LOOP 3000
|
||||
+#define STRING_SIZE 1024
|
||||
+char string1[STRING_SIZE];
|
||||
+
|
||||
+__attribute__ ((noinline, noclone))
|
||||
+static int
|
||||
+prepare (void)
|
||||
+{
|
||||
+ memset (string1, 'a', STRING_SIZE - 1);
|
||||
+ string1[STRING_SIZE - 100] = '\0';
|
||||
+ size_t len = strlen (string1);
|
||||
+ if (len == STRING_SIZE - 100)
|
||||
+ return EXIT_SUCCESS;
|
||||
+ else
|
||||
+ return EXIT_FAILURE;
|
||||
+}
|
||||
+
|
||||
+__attribute__ ((noinline, noclone))
|
||||
+static int
|
||||
+function (void)
|
||||
+{
|
||||
+ size_t len = strlen (string1);
|
||||
+ if (len == STRING_SIZE - 100)
|
||||
+ return 0;
|
||||
+ else
|
||||
+ return 1;
|
||||
+}
|
||||
+
|
||||
+static int
|
||||
+do_test (void)
|
||||
+{
|
||||
+ return do_test_1 ("strlen", LOOP, prepare, function);
|
||||
+}
|
||||
diff --git a/sysdeps/x86/tst-strncmp-rtm.c b/sysdeps/x86/tst-strncmp-rtm.c
|
||||
new file mode 100644
|
||||
index 00000000..236ad951
|
||||
--- /dev/null
|
||||
+++ b/sysdeps/x86/tst-strncmp-rtm.c
|
||||
@@ -0,0 +1,52 @@
|
||||
+/* Test case for strncmp inside a transactionally executing RTM region.
|
||||
+ Copyright (C) 2021 Free Software Foundation, Inc.
|
||||
+ This file is part of the GNU C Library.
|
||||
+
|
||||
+ The GNU C Library is free software; you can redistribute it and/or
|
||||
+ modify it under the terms of the GNU Lesser General Public
|
||||
+ License as published by the Free Software Foundation; either
|
||||
+ version 2.1 of the License, or (at your option) any later version.
|
||||
+
|
||||
+ The GNU C Library is distributed in the hope that it will be useful,
|
||||
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
||||
+ Lesser General Public License for more details.
|
||||
+
|
||||
+ You should have received a copy of the GNU Lesser General Public
|
||||
+ License along with the GNU C Library; if not, see
|
||||
+ <https://www.gnu.org/licenses/>. */
|
||||
+
|
||||
+#include <tst-string-rtm.h>
|
||||
+
|
||||
+#define LOOP 3000
|
||||
+#define STRING_SIZE 1024
|
||||
+char string1[STRING_SIZE];
|
||||
+char string2[STRING_SIZE];
|
||||
+
|
||||
+__attribute__ ((noinline, noclone))
|
||||
+static int
|
||||
+prepare (void)
|
||||
+{
|
||||
+ memset (string1, 'a', STRING_SIZE - 1);
|
||||
+ memset (string2, 'a', STRING_SIZE - 1);
|
||||
+ if (strncmp (string1, string2, STRING_SIZE) == 0)
|
||||
+ return EXIT_SUCCESS;
|
||||
+ else
|
||||
+ return EXIT_FAILURE;
|
||||
+}
|
||||
+
|
||||
+__attribute__ ((noinline, noclone))
|
||||
+static int
|
||||
+function (void)
|
||||
+{
|
||||
+ if (strncmp (string1, string2, STRING_SIZE) == 0)
|
||||
+ return 0;
|
||||
+ else
|
||||
+ return 1;
|
||||
+}
|
||||
+
|
||||
+static int
|
||||
+do_test (void)
|
||||
+{
|
||||
+ return do_test_1 ("strncmp", LOOP, prepare, function);
|
||||
+}
|
||||
diff --git a/sysdeps/x86/tst-strrchr-rtm.c b/sysdeps/x86/tst-strrchr-rtm.c
|
||||
new file mode 100644
|
||||
index 00000000..e32bfaf5
|
||||
--- /dev/null
|
||||
+++ b/sysdeps/x86/tst-strrchr-rtm.c
|
||||
@@ -0,0 +1,53 @@
|
||||
+/* Test case for strrchr inside a transactionally executing RTM region.
|
||||
+ Copyright (C) 2021 Free Software Foundation, Inc.
|
||||
+ This file is part of the GNU C Library.
|
||||
+
|
||||
+ The GNU C Library is free software; you can redistribute it and/or
|
||||
+ modify it under the terms of the GNU Lesser General Public
|
||||
+ License as published by the Free Software Foundation; either
|
||||
+ version 2.1 of the License, or (at your option) any later version.
|
||||
+
|
||||
+ The GNU C Library is distributed in the hope that it will be useful,
|
||||
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
||||
+ Lesser General Public License for more details.
|
||||
+
|
||||
+ You should have received a copy of the GNU Lesser General Public
|
||||
+ License along with the GNU C Library; if not, see
|
||||
+ <https://www.gnu.org/licenses/>. */
|
||||
+
|
||||
+#include <tst-string-rtm.h>
|
||||
+
|
||||
+#define LOOP 3000
|
||||
+#define STRING_SIZE 1024
|
||||
+char string1[STRING_SIZE];
|
||||
+
|
||||
+__attribute__ ((noinline, noclone))
|
||||
+static int
|
||||
+prepare (void)
|
||||
+{
|
||||
+ memset (string1, 'a', STRING_SIZE - 1);
|
||||
+ string1[STRING_SIZE - 100] = 'c';
|
||||
+ char *p = strrchr (string1, 'c');
|
||||
+ if (p == &string1[STRING_SIZE - 100])
|
||||
+ return EXIT_SUCCESS;
|
||||
+ else
|
||||
+ return EXIT_FAILURE;
|
||||
+}
|
||||
+
|
||||
+__attribute__ ((noinline, noclone))
|
||||
+static int
|
||||
+function (void)
|
||||
+{
|
||||
+ char *p = strrchr (string1, 'c');
|
||||
+ if (p == &string1[STRING_SIZE - 100])
|
||||
+ return 0;
|
||||
+ else
|
||||
+ return 1;
|
||||
+}
|
||||
+
|
||||
+static int
|
||||
+do_test (void)
|
||||
+{
|
||||
+ return do_test_1 ("strrchr", LOOP, prepare, function);
|
||||
+}
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,148 +0,0 @@
|
||||
From 4e2d8f352774b56078c34648b14a2412c38384f4 Mon Sep 17 00:00:00 2001
|
||||
From: "H.J. Lu" <hjl.tools@gmail.com>
|
||||
Date: Sun, 7 Mar 2021 09:44:18 -0800
|
||||
Subject: [PATCH] x86-64: Use ZMM16-ZMM31 in AVX512 memset family functions
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
Update ifunc-memset.h/ifunc-wmemset.h to select the function optimized
|
||||
with AVX512 instructions using ZMM16-ZMM31 registers to avoid RTM abort
|
||||
with usable AVX512VL and AVX512BW since VZEROUPPER isn't needed at
|
||||
function exit.
|
||||
---
|
||||
sysdeps/x86_64/multiarch/ifunc-impl-list.c | 14 +++++++++-----
|
||||
sysdeps/x86_64/multiarch/ifunc-memset.h | 13 ++++++++-----
|
||||
sysdeps/x86_64/multiarch/ifunc-wmemset.h | 12 ++++++------
|
||||
.../multiarch/memset-avx512-unaligned-erms.S | 16 ++++++++--------
|
||||
4 files changed, 31 insertions(+), 24 deletions(-)
|
||||
|
||||
diff --git a/sysdeps/x86_64/multiarch/ifunc-impl-list.c b/sysdeps/x86_64/multiarch/ifunc-impl-list.c
|
||||
index c1efeec0..d969a156 100644
|
||||
--- a/sysdeps/x86_64/multiarch/ifunc-impl-list.c
|
||||
+++ b/sysdeps/x86_64/multiarch/ifunc-impl-list.c
|
||||
@@ -211,10 +211,12 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
|
||||
&& CPU_FEATURE_USABLE (AVX512BW)),
|
||||
__memset_chk_evex_unaligned_erms)
|
||||
IFUNC_IMPL_ADD (array, i, __memset_chk,
|
||||
- CPU_FEATURE_USABLE (AVX512F),
|
||||
+ (CPU_FEATURE_USABLE (AVX512VL)
|
||||
+ && CPU_FEATURE_USABLE (AVX512BW)),
|
||||
__memset_chk_avx512_unaligned_erms)
|
||||
IFUNC_IMPL_ADD (array, i, __memset_chk,
|
||||
- CPU_FEATURE_USABLE (AVX512F),
|
||||
+ (CPU_FEATURE_USABLE (AVX512VL)
|
||||
+ && CPU_FEATURE_USABLE (AVX512BW)),
|
||||
__memset_chk_avx512_unaligned)
|
||||
IFUNC_IMPL_ADD (array, i, __memset_chk,
|
||||
CPU_FEATURE_USABLE (AVX512F),
|
||||
@@ -252,10 +254,12 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
|
||||
&& CPU_FEATURE_USABLE (AVX512BW)),
|
||||
__memset_evex_unaligned_erms)
|
||||
IFUNC_IMPL_ADD (array, i, memset,
|
||||
- CPU_FEATURE_USABLE (AVX512F),
|
||||
+ (CPU_FEATURE_USABLE (AVX512VL)
|
||||
+ && CPU_FEATURE_USABLE (AVX512BW)),
|
||||
__memset_avx512_unaligned_erms)
|
||||
IFUNC_IMPL_ADD (array, i, memset,
|
||||
- CPU_FEATURE_USABLE (AVX512F),
|
||||
+ (CPU_FEATURE_USABLE (AVX512VL)
|
||||
+ && CPU_FEATURE_USABLE (AVX512BW)),
|
||||
__memset_avx512_unaligned)
|
||||
IFUNC_IMPL_ADD (array, i, memset,
|
||||
CPU_FEATURE_USABLE (AVX512F),
|
||||
@@ -719,7 +723,7 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
|
||||
CPU_FEATURE_USABLE (AVX512VL),
|
||||
__wmemset_evex_unaligned)
|
||||
IFUNC_IMPL_ADD (array, i, wmemset,
|
||||
- CPU_FEATURE_USABLE (AVX512F),
|
||||
+ CPU_FEATURE_USABLE (AVX512VL),
|
||||
__wmemset_avx512_unaligned))
|
||||
|
||||
#ifdef SHARED
|
||||
diff --git a/sysdeps/x86_64/multiarch/ifunc-memset.h b/sysdeps/x86_64/multiarch/ifunc-memset.h
|
||||
index 6f3375cc..19795938 100644
|
||||
--- a/sysdeps/x86_64/multiarch/ifunc-memset.h
|
||||
+++ b/sysdeps/x86_64/multiarch/ifunc-memset.h
|
||||
@@ -53,13 +53,16 @@ IFUNC_SELECTOR (void)
|
||||
if (CPU_FEATURE_USABLE_P (cpu_features, AVX512F)
|
||||
&& !CPU_FEATURES_ARCH_P (cpu_features, Prefer_No_AVX512))
|
||||
{
|
||||
- if (CPU_FEATURES_ARCH_P (cpu_features, Prefer_No_VZEROUPPER))
|
||||
- return OPTIMIZE (avx512_no_vzeroupper);
|
||||
+ if (CPU_FEATURE_USABLE_P (cpu_features, AVX512VL)
|
||||
+ && CPU_FEATURE_USABLE_P (cpu_features, AVX512BW))
|
||||
+ {
|
||||
+ if (CPU_FEATURE_USABLE_P (cpu_features, ERMS))
|
||||
+ return OPTIMIZE (avx512_unaligned_erms);
|
||||
|
||||
- if (CPU_FEATURE_USABLE_P (cpu_features, ERMS))
|
||||
- return OPTIMIZE (avx512_unaligned_erms);
|
||||
+ return OPTIMIZE (avx512_unaligned);
|
||||
+ }
|
||||
|
||||
- return OPTIMIZE (avx512_unaligned);
|
||||
+ return OPTIMIZE (avx512_no_vzeroupper);
|
||||
}
|
||||
|
||||
if (CPU_FEATURE_USABLE_P (cpu_features, AVX2))
|
||||
diff --git a/sysdeps/x86_64/multiarch/ifunc-wmemset.h b/sysdeps/x86_64/multiarch/ifunc-wmemset.h
|
||||
index bdc94c6c..98c5d406 100644
|
||||
--- a/sysdeps/x86_64/multiarch/ifunc-wmemset.h
|
||||
+++ b/sysdeps/x86_64/multiarch/ifunc-wmemset.h
|
||||
@@ -33,13 +33,13 @@ IFUNC_SELECTOR (void)
|
||||
if (CPU_FEATURE_USABLE_P (cpu_features, AVX2)
|
||||
&& CPU_FEATURES_ARCH_P (cpu_features, AVX_Fast_Unaligned_Load))
|
||||
{
|
||||
- if (CPU_FEATURE_USABLE_P (cpu_features, AVX512F)
|
||||
- && !CPU_FEATURES_ARCH_P (cpu_features, Prefer_No_AVX512)
|
||||
- && !CPU_FEATURES_ARCH_P (cpu_features, Prefer_No_VZEROUPPER))
|
||||
- return OPTIMIZE (avx512_unaligned);
|
||||
-
|
||||
if (CPU_FEATURE_USABLE_P (cpu_features, AVX512VL))
|
||||
- return OPTIMIZE (evex_unaligned);
|
||||
+ {
|
||||
+ if (!CPU_FEATURES_ARCH_P (cpu_features, Prefer_No_AVX512))
|
||||
+ return OPTIMIZE (avx512_unaligned);
|
||||
+
|
||||
+ return OPTIMIZE (evex_unaligned);
|
||||
+ }
|
||||
|
||||
if (CPU_FEATURE_USABLE_P (cpu_features, RTM))
|
||||
return OPTIMIZE (avx2_unaligned_rtm);
|
||||
diff --git a/sysdeps/x86_64/multiarch/memset-avx512-unaligned-erms.S b/sysdeps/x86_64/multiarch/memset-avx512-unaligned-erms.S
|
||||
index 0783979c..22e7b187 100644
|
||||
--- a/sysdeps/x86_64/multiarch/memset-avx512-unaligned-erms.S
|
||||
+++ b/sysdeps/x86_64/multiarch/memset-avx512-unaligned-erms.S
|
||||
@@ -1,22 +1,22 @@
|
||||
#if IS_IN (libc)
|
||||
# define VEC_SIZE 64
|
||||
-# define VEC(i) zmm##i
|
||||
+# define XMM0 xmm16
|
||||
+# define YMM0 ymm16
|
||||
+# define VEC0 zmm16
|
||||
+# define VEC(i) VEC##i
|
||||
# define VMOVU vmovdqu64
|
||||
# define VMOVA vmovdqa64
|
||||
+# define VZEROUPPER
|
||||
|
||||
# define MEMSET_VDUP_TO_VEC0_AND_SET_RETURN(d, r) \
|
||||
- vmovd d, %xmm0; \
|
||||
movq r, %rax; \
|
||||
- vpbroadcastb %xmm0, %xmm0; \
|
||||
- vpbroadcastq %xmm0, %zmm0
|
||||
+ vpbroadcastb d, %VEC0
|
||||
|
||||
# define WMEMSET_VDUP_TO_VEC0_AND_SET_RETURN(d, r) \
|
||||
- vmovd d, %xmm0; \
|
||||
movq r, %rax; \
|
||||
- vpbroadcastd %xmm0, %xmm0; \
|
||||
- vpbroadcastq %xmm0, %zmm0
|
||||
+ vpbroadcastd d, %VEC0
|
||||
|
||||
-# define SECTION(p) p##.avx512
|
||||
+# define SECTION(p) p##.evex512
|
||||
# define MEMSET_SYMBOL(p,s) p##_avx512_##s
|
||||
# define WMEMSET_SYMBOL(p,s) p##_avx512_##s
|
||||
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,230 +0,0 @@
|
||||
From b304fc201d2f6baf52ea790df8643e99772243cd Mon Sep 17 00:00:00 2001
|
||||
From: "H.J. Lu" <hjl.tools@gmail.com>
|
||||
Date: Mon, 21 Jan 2019 11:25:56 -0800
|
||||
Subject: [PATCH] x86-64 memcmp/wmemcmp: Properly handle the length parameter
|
||||
[BZ# 24097]
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
On x32, the size_t parameter may be passed in the lower 32 bits of a
|
||||
64-bit register with the non-zero upper 32 bits. The string/memory
|
||||
functions written in assembly can only use the lower 32 bits of a
|
||||
64-bit register as length or must clear the upper 32 bits before using
|
||||
the full 64-bit register for length.
|
||||
|
||||
This pach fixes memcmp/wmemcmp for x32. Tested on x86-64 and x32. On
|
||||
x86-64, libc.so is the same with and withou the fix.
|
||||
|
||||
[BZ# 24097]
|
||||
CVE-2019-6488
|
||||
* sysdeps/x86_64/multiarch/memcmp-avx2-movbe.S: Use RDX_LP for
|
||||
length. Clear the upper 32 bits of RDX register.
|
||||
* sysdeps/x86_64/multiarch/memcmp-sse4.S: Likewise.
|
||||
* sysdeps/x86_64/multiarch/memcmp-ssse3.S: Likewise.
|
||||
* sysdeps/x86_64/x32/Makefile (tests): Add tst-size_t-memcmp and
|
||||
tst-size_t-wmemcmp.
|
||||
* sysdeps/x86_64/x32/tst-size_t-memcmp.c: New file.
|
||||
* sysdeps/x86_64/x32/tst-size_t-wmemcmp.c: Likewise.
|
||||
---
|
||||
sysdeps/x86_64/multiarch/memcmp-avx2-movbe.S | 7 +-
|
||||
sysdeps/x86_64/multiarch/memcmp-sse4.S | 9 ++-
|
||||
sysdeps/x86_64/multiarch/memcmp-ssse3.S | 7 +-
|
||||
sysdeps/x86_64/x32/Makefile | 4 +-
|
||||
sysdeps/x86_64/x32/tst-size_t-memcmp.c | 76 ++++++++++++++++++++
|
||||
sysdeps/x86_64/x32/tst-size_t-wmemcmp.c | 20 ++++++
|
||||
6 files changed, 114 insertions(+), 9 deletions(-)
|
||||
create mode 100644 sysdeps/x86_64/x32/tst-size_t-memcmp.c
|
||||
create mode 100644 sysdeps/x86_64/x32/tst-size_t-wmemcmp.c
|
||||
|
||||
Conflicts:
|
||||
ChangeLog
|
||||
(removed)
|
||||
|
||||
diff --git a/sysdeps/x86_64/multiarch/memcmp-avx2-movbe.S b/sysdeps/x86_64/multiarch/memcmp-avx2-movbe.S
|
||||
index 30f764c3..e3a35b89 100644
|
||||
--- a/sysdeps/x86_64/multiarch/memcmp-avx2-movbe.S
|
||||
+++ b/sysdeps/x86_64/multiarch/memcmp-avx2-movbe.S
|
||||
@@ -58,9 +58,12 @@
|
||||
.section .text.avx,"ax",@progbits
|
||||
ENTRY (MEMCMP)
|
||||
# ifdef USE_AS_WMEMCMP
|
||||
- shl $2, %rdx
|
||||
+ shl $2, %RDX_LP
|
||||
+# elif defined __ILP32__
|
||||
+ /* Clear the upper 32 bits. */
|
||||
+ movl %edx, %edx
|
||||
# endif
|
||||
- cmpq $VEC_SIZE, %rdx
|
||||
+ cmp $VEC_SIZE, %RDX_LP
|
||||
jb L(less_vec)
|
||||
|
||||
/* From VEC to 2 * VEC. No branch when size == VEC_SIZE. */
|
||||
diff --git a/sysdeps/x86_64/multiarch/memcmp-sse4.S b/sysdeps/x86_64/multiarch/memcmp-sse4.S
|
||||
index 8e164f2c..302900f5 100644
|
||||
--- a/sysdeps/x86_64/multiarch/memcmp-sse4.S
|
||||
+++ b/sysdeps/x86_64/multiarch/memcmp-sse4.S
|
||||
@@ -42,13 +42,16 @@
|
||||
.section .text.sse4.1,"ax",@progbits
|
||||
ENTRY (MEMCMP)
|
||||
# ifdef USE_AS_WMEMCMP
|
||||
- shl $2, %rdx
|
||||
+ shl $2, %RDX_LP
|
||||
+# elif defined __ILP32__
|
||||
+ /* Clear the upper 32 bits. */
|
||||
+ mov %edx, %edx
|
||||
# endif
|
||||
pxor %xmm0, %xmm0
|
||||
- cmp $79, %rdx
|
||||
+ cmp $79, %RDX_LP
|
||||
ja L(79bytesormore)
|
||||
# ifndef USE_AS_WMEMCMP
|
||||
- cmp $1, %rdx
|
||||
+ cmp $1, %RDX_LP
|
||||
je L(firstbyte)
|
||||
# endif
|
||||
add %rdx, %rsi
|
||||
diff --git a/sysdeps/x86_64/multiarch/memcmp-ssse3.S b/sysdeps/x86_64/multiarch/memcmp-ssse3.S
|
||||
index 6f76c641..69d030fc 100644
|
||||
--- a/sysdeps/x86_64/multiarch/memcmp-ssse3.S
|
||||
+++ b/sysdeps/x86_64/multiarch/memcmp-ssse3.S
|
||||
@@ -33,9 +33,12 @@
|
||||
atom_text_section
|
||||
ENTRY (MEMCMP)
|
||||
# ifdef USE_AS_WMEMCMP
|
||||
- shl $2, %rdx
|
||||
- test %rdx, %rdx
|
||||
+ shl $2, %RDX_LP
|
||||
+ test %RDX_LP, %RDX_LP
|
||||
jz L(equal)
|
||||
+# elif defined __ILP32__
|
||||
+ /* Clear the upper 32 bits. */
|
||||
+ mov %edx, %edx
|
||||
# endif
|
||||
mov %rdx, %rcx
|
||||
mov %rdi, %rdx
|
||||
diff --git a/sysdeps/x86_64/x32/Makefile b/sysdeps/x86_64/x32/Makefile
|
||||
index 7d528889..ddec7f04 100644
|
||||
--- a/sysdeps/x86_64/x32/Makefile
|
||||
+++ b/sysdeps/x86_64/x32/Makefile
|
||||
@@ -6,9 +6,9 @@ CFLAGS-s_llround.c += -fno-builtin-lround
|
||||
endif
|
||||
|
||||
ifeq ($(subdir),string)
|
||||
-tests += tst-size_t-memchr
|
||||
+tests += tst-size_t-memchr tst-size_t-memcmp
|
||||
endif
|
||||
|
||||
ifeq ($(subdir),wcsmbs)
|
||||
-tests += tst-size_t-wmemchr
|
||||
+tests += tst-size_t-wmemchr tst-size_t-wmemcmp
|
||||
endif
|
||||
diff --git a/sysdeps/x86_64/x32/tst-size_t-memcmp.c b/sysdeps/x86_64/x32/tst-size_t-memcmp.c
|
||||
new file mode 100644
|
||||
index 00000000..9bd6fdb4
|
||||
--- /dev/null
|
||||
+++ b/sysdeps/x86_64/x32/tst-size_t-memcmp.c
|
||||
@@ -0,0 +1,76 @@
|
||||
+/* Test memcmp with size_t in the lower 32 bits of 64-bit register.
|
||||
+ Copyright (C) 2019 Free Software Foundation, Inc.
|
||||
+ This file is part of the GNU C Library.
|
||||
+
|
||||
+ The GNU C Library is free software; you can redistribute it and/or
|
||||
+ modify it under the terms of the GNU Lesser General Public
|
||||
+ License as published by the Free Software Foundation; either
|
||||
+ version 2.1 of the License, or (at your option) any later version.
|
||||
+
|
||||
+ The GNU C Library is distributed in the hope that it will be useful,
|
||||
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
||||
+ Lesser General Public License for more details.
|
||||
+
|
||||
+ You should have received a copy of the GNU Lesser General Public
|
||||
+ License along with the GNU C Library; if not, see
|
||||
+ <http://www.gnu.org/licenses/>. */
|
||||
+
|
||||
+#define TEST_MAIN
|
||||
+#ifdef WIDE
|
||||
+# define TEST_NAME "wmemcmp"
|
||||
+#else
|
||||
+# define TEST_NAME "memcmp"
|
||||
+#endif
|
||||
+
|
||||
+#include "test-size_t.h"
|
||||
+
|
||||
+#ifdef WIDE
|
||||
+# include <inttypes.h>
|
||||
+# include <wchar.h>
|
||||
+
|
||||
+# define MEMCMP wmemcmp
|
||||
+# define CHAR wchar_t
|
||||
+#else
|
||||
+# define MEMCMP memcmp
|
||||
+# define CHAR char
|
||||
+#endif
|
||||
+
|
||||
+IMPL (MEMCMP, 1)
|
||||
+
|
||||
+typedef int (*proto_t) (const CHAR *, const CHAR *, size_t);
|
||||
+
|
||||
+static int
|
||||
+__attribute__ ((noinline, noclone))
|
||||
+do_memcmp (parameter_t a, parameter_t b)
|
||||
+{
|
||||
+ return CALL (&b, a.p, b.p, a.len);
|
||||
+}
|
||||
+
|
||||
+static int
|
||||
+test_main (void)
|
||||
+{
|
||||
+ test_init ();
|
||||
+
|
||||
+ parameter_t dest = { { page_size / sizeof (CHAR) }, buf1 };
|
||||
+ parameter_t src = { { 0 }, buf2 };
|
||||
+
|
||||
+ memcpy (buf1, buf2, page_size);
|
||||
+
|
||||
+ int ret = 0;
|
||||
+ FOR_EACH_IMPL (impl, 0)
|
||||
+ {
|
||||
+ src.fn = impl->fn;
|
||||
+ int res = do_memcmp (dest, src);
|
||||
+ if (res)
|
||||
+ {
|
||||
+ error (0, 0, "Wrong result in function %s: %i != 0",
|
||||
+ impl->name, res);
|
||||
+ ret = 1;
|
||||
+ }
|
||||
+ }
|
||||
+
|
||||
+ return ret ? EXIT_FAILURE : EXIT_SUCCESS;
|
||||
+}
|
||||
+
|
||||
+#include <support/test-driver.c>
|
||||
diff --git a/sysdeps/x86_64/x32/tst-size_t-wmemcmp.c b/sysdeps/x86_64/x32/tst-size_t-wmemcmp.c
|
||||
new file mode 100644
|
||||
index 00000000..e8b5ffd0
|
||||
--- /dev/null
|
||||
+++ b/sysdeps/x86_64/x32/tst-size_t-wmemcmp.c
|
||||
@@ -0,0 +1,20 @@
|
||||
+/* Test wmemcmp with size_t in the lower 32 bits of 64-bit register.
|
||||
+ Copyright (C) 2019 Free Software Foundation, Inc.
|
||||
+ This file is part of the GNU C Library.
|
||||
+
|
||||
+ The GNU C Library is free software; you can redistribute it and/or
|
||||
+ modify it under the terms of the GNU Lesser General Public
|
||||
+ License as published by the Free Software Foundation; either
|
||||
+ version 2.1 of the License, or (at your option) any later version.
|
||||
+
|
||||
+ The GNU C Library is distributed in the hope that it will be useful,
|
||||
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
||||
+ Lesser General Public License for more details.
|
||||
+
|
||||
+ You should have received a copy of the GNU Lesser General Public
|
||||
+ License along with the GNU C Library; if not, see
|
||||
+ <http://www.gnu.org/licenses/>. */
|
||||
+
|
||||
+#define WIDE 1
|
||||
+#include "tst-size_t-memcmp.c"
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,164 +0,0 @@
|
||||
From e4fda4631017e49d4ee5a2755db34289b6860fa4 Mon Sep 17 00:00:00 2001
|
||||
From: "H.J. Lu" <hjl.tools@gmail.com>
|
||||
Date: Sun, 7 Mar 2021 09:45:23 -0800
|
||||
Subject: [PATCH] x86-64: Use ZMM16-ZMM31 in AVX512 memmove family functions
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
Update ifunc-memmove.h to select the function optimized with AVX512
|
||||
instructions using ZMM16-ZMM31 registers to avoid RTM abort with usable
|
||||
AVX512VL since VZEROUPPER isn't needed at function exit.
|
||||
---
|
||||
sysdeps/x86_64/multiarch/ifunc-impl-list.c | 24 +++++++++---------
|
||||
sysdeps/x86_64/multiarch/ifunc-memmove.h | 12 +++++----
|
||||
.../multiarch/memmove-avx512-unaligned-erms.S | 25 +++++++++++++++++--
|
||||
3 files changed, 42 insertions(+), 19 deletions(-)
|
||||
|
||||
diff --git a/sysdeps/x86_64/multiarch/ifunc-impl-list.c b/sysdeps/x86_64/multiarch/ifunc-impl-list.c
|
||||
index d969a156..fec384f6 100644
|
||||
--- a/sysdeps/x86_64/multiarch/ifunc-impl-list.c
|
||||
+++ b/sysdeps/x86_64/multiarch/ifunc-impl-list.c
|
||||
@@ -83,10 +83,10 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
|
||||
CPU_FEATURE_USABLE (AVX512F),
|
||||
__memmove_chk_avx512_no_vzeroupper)
|
||||
IFUNC_IMPL_ADD (array, i, __memmove_chk,
|
||||
- CPU_FEATURE_USABLE (AVX512F),
|
||||
+ CPU_FEATURE_USABLE (AVX512VL),
|
||||
__memmove_chk_avx512_unaligned)
|
||||
IFUNC_IMPL_ADD (array, i, __memmove_chk,
|
||||
- CPU_FEATURE_USABLE (AVX512F),
|
||||
+ CPU_FEATURE_USABLE (AVX512VL),
|
||||
__memmove_chk_avx512_unaligned_erms)
|
||||
IFUNC_IMPL_ADD (array, i, __memmove_chk,
|
||||
CPU_FEATURE_USABLE (AVX),
|
||||
@@ -148,10 +148,10 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
|
||||
CPU_FEATURE_USABLE (AVX512F),
|
||||
__memmove_avx512_no_vzeroupper)
|
||||
IFUNC_IMPL_ADD (array, i, memmove,
|
||||
- CPU_FEATURE_USABLE (AVX512F),
|
||||
+ CPU_FEATURE_USABLE (AVX512VL),
|
||||
__memmove_avx512_unaligned)
|
||||
IFUNC_IMPL_ADD (array, i, memmove,
|
||||
- CPU_FEATURE_USABLE (AVX512F),
|
||||
+ CPU_FEATURE_USABLE (AVX512VL),
|
||||
__memmove_avx512_unaligned_erms)
|
||||
IFUNC_IMPL_ADD (array, i, memmove, CPU_FEATURE_USABLE (SSSE3),
|
||||
__memmove_ssse3_back)
|
||||
@@ -733,10 +733,10 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
|
||||
CPU_FEATURE_USABLE (AVX512F),
|
||||
__memcpy_chk_avx512_no_vzeroupper)
|
||||
IFUNC_IMPL_ADD (array, i, __memcpy_chk,
|
||||
- CPU_FEATURE_USABLE (AVX512F),
|
||||
+ CPU_FEATURE_USABLE (AVX512VL),
|
||||
__memcpy_chk_avx512_unaligned)
|
||||
IFUNC_IMPL_ADD (array, i, __memcpy_chk,
|
||||
- CPU_FEATURE_USABLE (AVX512F),
|
||||
+ CPU_FEATURE_USABLE (AVX512VL),
|
||||
__memcpy_chk_avx512_unaligned_erms)
|
||||
IFUNC_IMPL_ADD (array, i, __memcpy_chk,
|
||||
CPU_FEATURE_USABLE (AVX),
|
||||
@@ -802,10 +802,10 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
|
||||
CPU_FEATURE_USABLE (AVX512F),
|
||||
__memcpy_avx512_no_vzeroupper)
|
||||
IFUNC_IMPL_ADD (array, i, memcpy,
|
||||
- CPU_FEATURE_USABLE (AVX512F),
|
||||
+ CPU_FEATURE_USABLE (AVX512VL),
|
||||
__memcpy_avx512_unaligned)
|
||||
IFUNC_IMPL_ADD (array, i, memcpy,
|
||||
- CPU_FEATURE_USABLE (AVX512F),
|
||||
+ CPU_FEATURE_USABLE (AVX512VL),
|
||||
__memcpy_avx512_unaligned_erms)
|
||||
IFUNC_IMPL_ADD (array, i, memcpy, 1, __memcpy_sse2_unaligned)
|
||||
IFUNC_IMPL_ADD (array, i, memcpy, 1,
|
||||
@@ -819,10 +819,10 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
|
||||
CPU_FEATURE_USABLE (AVX512F),
|
||||
__mempcpy_chk_avx512_no_vzeroupper)
|
||||
IFUNC_IMPL_ADD (array, i, __mempcpy_chk,
|
||||
- CPU_FEATURE_USABLE (AVX512F),
|
||||
+ CPU_FEATURE_USABLE (AVX512VL),
|
||||
__mempcpy_chk_avx512_unaligned)
|
||||
IFUNC_IMPL_ADD (array, i, __mempcpy_chk,
|
||||
- CPU_FEATURE_USABLE (AVX512F),
|
||||
+ CPU_FEATURE_USABLE (AVX512VL),
|
||||
__mempcpy_chk_avx512_unaligned_erms)
|
||||
IFUNC_IMPL_ADD (array, i, __mempcpy_chk,
|
||||
CPU_FEATURE_USABLE (AVX),
|
||||
@@ -864,10 +864,10 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
|
||||
CPU_FEATURE_USABLE (AVX512F),
|
||||
__mempcpy_avx512_no_vzeroupper)
|
||||
IFUNC_IMPL_ADD (array, i, mempcpy,
|
||||
- CPU_FEATURE_USABLE (AVX512F),
|
||||
+ CPU_FEATURE_USABLE (AVX512VL),
|
||||
__mempcpy_avx512_unaligned)
|
||||
IFUNC_IMPL_ADD (array, i, mempcpy,
|
||||
- CPU_FEATURE_USABLE (AVX512F),
|
||||
+ CPU_FEATURE_USABLE (AVX512VL),
|
||||
__mempcpy_avx512_unaligned_erms)
|
||||
IFUNC_IMPL_ADD (array, i, mempcpy,
|
||||
CPU_FEATURE_USABLE (AVX),
|
||||
diff --git a/sysdeps/x86_64/multiarch/ifunc-memmove.h b/sysdeps/x86_64/multiarch/ifunc-memmove.h
|
||||
index fa09b9fb..014e95c7 100644
|
||||
--- a/sysdeps/x86_64/multiarch/ifunc-memmove.h
|
||||
+++ b/sysdeps/x86_64/multiarch/ifunc-memmove.h
|
||||
@@ -56,13 +56,15 @@ IFUNC_SELECTOR (void)
|
||||
if (CPU_FEATURE_USABLE_P (cpu_features, AVX512F)
|
||||
&& !CPU_FEATURES_ARCH_P (cpu_features, Prefer_No_AVX512))
|
||||
{
|
||||
- if (CPU_FEATURES_ARCH_P (cpu_features, Prefer_No_VZEROUPPER))
|
||||
- return OPTIMIZE (avx512_no_vzeroupper);
|
||||
+ if (CPU_FEATURE_USABLE_P (cpu_features, AVX512VL))
|
||||
+ {
|
||||
+ if (CPU_FEATURE_USABLE_P (cpu_features, ERMS))
|
||||
+ return OPTIMIZE (avx512_unaligned_erms);
|
||||
|
||||
- if (CPU_FEATURE_USABLE_P (cpu_features, ERMS))
|
||||
- return OPTIMIZE (avx512_unaligned_erms);
|
||||
+ return OPTIMIZE (avx512_unaligned);
|
||||
+ }
|
||||
|
||||
- return OPTIMIZE (avx512_unaligned);
|
||||
+ return OPTIMIZE (avx512_no_vzeroupper);
|
||||
}
|
||||
|
||||
if (CPU_FEATURES_ARCH_P (cpu_features, AVX_Fast_Unaligned_Load))
|
||||
diff --git a/sysdeps/x86_64/multiarch/memmove-avx512-unaligned-erms.S b/sysdeps/x86_64/multiarch/memmove-avx512-unaligned-erms.S
|
||||
index aac1515c..848848ab 100644
|
||||
--- a/sysdeps/x86_64/multiarch/memmove-avx512-unaligned-erms.S
|
||||
+++ b/sysdeps/x86_64/multiarch/memmove-avx512-unaligned-erms.S
|
||||
@@ -1,11 +1,32 @@
|
||||
#if IS_IN (libc)
|
||||
# define VEC_SIZE 64
|
||||
-# define VEC(i) zmm##i
|
||||
+# define XMM0 xmm16
|
||||
+# define XMM1 xmm17
|
||||
+# define YMM0 ymm16
|
||||
+# define YMM1 ymm17
|
||||
+# define VEC0 zmm16
|
||||
+# define VEC1 zmm17
|
||||
+# define VEC2 zmm18
|
||||
+# define VEC3 zmm19
|
||||
+# define VEC4 zmm20
|
||||
+# define VEC5 zmm21
|
||||
+# define VEC6 zmm22
|
||||
+# define VEC7 zmm23
|
||||
+# define VEC8 zmm24
|
||||
+# define VEC9 zmm25
|
||||
+# define VEC10 zmm26
|
||||
+# define VEC11 zmm27
|
||||
+# define VEC12 zmm28
|
||||
+# define VEC13 zmm29
|
||||
+# define VEC14 zmm30
|
||||
+# define VEC15 zmm31
|
||||
+# define VEC(i) VEC##i
|
||||
# define VMOVNT vmovntdq
|
||||
# define VMOVU vmovdqu64
|
||||
# define VMOVA vmovdqa64
|
||||
+# define VZEROUPPER
|
||||
|
||||
-# define SECTION(p) p##.avx512
|
||||
+# define SECTION(p) p##.evex512
|
||||
# define MEMMOVE_SYMBOL(p,s) p##_avx512_##s
|
||||
|
||||
# include "memmove-vec-unaligned-erms.S"
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,71 +0,0 @@
|
||||
From 595c22ecd8e87a27fd19270ed30fdbae9ad25426 Mon Sep 17 00:00:00 2001
|
||||
From: Sunil K Pandey <skpgkp2@gmail.com>
|
||||
Date: Thu, 1 Apr 2021 15:47:04 -0700
|
||||
Subject: [PATCH] x86-64: Fix ifdef indentation in strlen-evex.S
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
Fix some indentations of ifdef in file strlen-evex.S which are off by 1
|
||||
and confusing to read.
|
||||
---
|
||||
sysdeps/x86_64/multiarch/strlen-evex.S | 16 ++++++++--------
|
||||
1 file changed, 8 insertions(+), 8 deletions(-)
|
||||
|
||||
diff --git a/sysdeps/x86_64/multiarch/strlen-evex.S b/sysdeps/x86_64/multiarch/strlen-evex.S
|
||||
index cd022509..05838190 100644
|
||||
--- a/sysdeps/x86_64/multiarch/strlen-evex.S
|
||||
+++ b/sysdeps/x86_64/multiarch/strlen-evex.S
|
||||
@@ -276,10 +276,10 @@ L(last_2x_vec):
|
||||
.p2align 4
|
||||
L(first_vec_x0_check):
|
||||
tzcntl %eax, %eax
|
||||
-# ifdef USE_AS_WCSLEN
|
||||
+# ifdef USE_AS_WCSLEN
|
||||
/* NB: Multiply wchar_t count by 4 to get the number of bytes. */
|
||||
sall $2, %eax
|
||||
-# endif
|
||||
+# endif
|
||||
/* Check the end of data. */
|
||||
cmpq %rax, %rsi
|
||||
jbe L(max)
|
||||
@@ -293,10 +293,10 @@ L(first_vec_x0_check):
|
||||
.p2align 4
|
||||
L(first_vec_x1_check):
|
||||
tzcntl %eax, %eax
|
||||
-# ifdef USE_AS_WCSLEN
|
||||
+# ifdef USE_AS_WCSLEN
|
||||
/* NB: Multiply wchar_t count by 4 to get the number of bytes. */
|
||||
sall $2, %eax
|
||||
-# endif
|
||||
+# endif
|
||||
/* Check the end of data. */
|
||||
cmpq %rax, %rsi
|
||||
jbe L(max)
|
||||
@@ -311,10 +311,10 @@ L(first_vec_x1_check):
|
||||
.p2align 4
|
||||
L(first_vec_x2_check):
|
||||
tzcntl %eax, %eax
|
||||
-# ifdef USE_AS_WCSLEN
|
||||
+# ifdef USE_AS_WCSLEN
|
||||
/* NB: Multiply wchar_t count by 4 to get the number of bytes. */
|
||||
sall $2, %eax
|
||||
-# endif
|
||||
+# endif
|
||||
/* Check the end of data. */
|
||||
cmpq %rax, %rsi
|
||||
jbe L(max)
|
||||
@@ -329,10 +329,10 @@ L(first_vec_x2_check):
|
||||
.p2align 4
|
||||
L(first_vec_x3_check):
|
||||
tzcntl %eax, %eax
|
||||
-# ifdef USE_AS_WCSLEN
|
||||
+# ifdef USE_AS_WCSLEN
|
||||
/* NB: Multiply wchar_t count by 4 to get the number of bytes. */
|
||||
sall $2, %eax
|
||||
-# endif
|
||||
+# endif
|
||||
/* Check the end of data. */
|
||||
cmpq %rax, %rsi
|
||||
jbe L(max)
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,51 +0,0 @@
|
||||
From 55bf411b451c13f0fb7ff3d3bf9a820020b45df1 Mon Sep 17 00:00:00 2001
|
||||
From: "H.J. Lu" <hjl.tools@gmail.com>
|
||||
Date: Mon, 19 Apr 2021 07:07:21 -0700
|
||||
Subject: [PATCH] x86-64: Require BMI2 for __strlen_evex and __strnlen_evex
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
Since __strlen_evex and __strnlen_evex added by
|
||||
|
||||
commit 1fd8c163a83d96ace1ff78fa6bac7aee084f6f77
|
||||
Author: H.J. Lu <hjl.tools@gmail.com>
|
||||
Date: Fri Mar 5 06:24:52 2021 -0800
|
||||
|
||||
x86-64: Add ifunc-avx2.h functions with 256-bit EVEX
|
||||
|
||||
use sarx:
|
||||
|
||||
c4 e2 6a f7 c0 sarx %edx,%eax,%eax
|
||||
|
||||
require BMI2 for __strlen_evex and __strnlen_evex in ifunc-impl-list.c.
|
||||
ifunc-avx2.h already requires BMI2 for EVEX implementation.
|
||||
---
|
||||
sysdeps/x86_64/multiarch/ifunc-impl-list.c | 6 ++++--
|
||||
1 file changed, 4 insertions(+), 2 deletions(-)
|
||||
|
||||
diff --git a/sysdeps/x86_64/multiarch/ifunc-impl-list.c b/sysdeps/x86_64/multiarch/ifunc-impl-list.c
|
||||
index fec384f6..cbfc1a5d 100644
|
||||
--- a/sysdeps/x86_64/multiarch/ifunc-impl-list.c
|
||||
+++ b/sysdeps/x86_64/multiarch/ifunc-impl-list.c
|
||||
@@ -293,7 +293,8 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
|
||||
__strlen_avx2_rtm)
|
||||
IFUNC_IMPL_ADD (array, i, strlen,
|
||||
(CPU_FEATURE_USABLE (AVX512VL)
|
||||
- && CPU_FEATURE_USABLE (AVX512BW)),
|
||||
+ && CPU_FEATURE_USABLE (AVX512BW)
|
||||
+ && CPU_FEATURE_USABLE (BMI2)),
|
||||
__strlen_evex)
|
||||
IFUNC_IMPL_ADD (array, i, strlen, 1, __strlen_sse2))
|
||||
|
||||
@@ -308,7 +309,8 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
|
||||
__strnlen_avx2_rtm)
|
||||
IFUNC_IMPL_ADD (array, i, strnlen,
|
||||
(CPU_FEATURE_USABLE (AVX512VL)
|
||||
- && CPU_FEATURE_USABLE (AVX512BW)),
|
||||
+ && CPU_FEATURE_USABLE (AVX512BW)
|
||||
+ && CPU_FEATURE_USABLE (BMI2)),
|
||||
__strnlen_evex)
|
||||
IFUNC_IMPL_ADD (array, i, strnlen, 1, __strnlen_sse2))
|
||||
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,584 +0,0 @@
|
||||
From acfd088a1963ba51cd83c78f95c0ab25ead79e04 Mon Sep 17 00:00:00 2001
|
||||
From: Noah Goldstein <goldstein.w.n@gmail.com>
|
||||
Date: Mon, 3 May 2021 03:01:58 -0400
|
||||
Subject: [PATCH] x86: Optimize memchr-avx2.S
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
No bug. This commit optimizes memchr-avx2.S. The optimizations include
|
||||
replacing some branches with cmovcc, avoiding some branches entirely
|
||||
in the less_4x_vec case, making the page cross logic less strict,
|
||||
asaving a few instructions the in loop return loop. test-memchr,
|
||||
test-rawmemchr, and test-wmemchr are all passing.
|
||||
|
||||
Signed-off-by: Noah Goldstein <goldstein.w.n@gmail.com>
|
||||
Reviewed-by: H.J. Lu <hjl.tools@gmail.com>
|
||||
---
|
||||
sysdeps/x86_64/multiarch/memchr-avx2.S | 425 ++++++++++++++-----------
|
||||
1 file changed, 247 insertions(+), 178 deletions(-)
|
||||
|
||||
diff --git a/sysdeps/x86_64/multiarch/memchr-avx2.S b/sysdeps/x86_64/multiarch/memchr-avx2.S
|
||||
index cf893e77..b377f22e 100644
|
||||
--- a/sysdeps/x86_64/multiarch/memchr-avx2.S
|
||||
+++ b/sysdeps/x86_64/multiarch/memchr-avx2.S
|
||||
@@ -26,8 +26,22 @@
|
||||
|
||||
# ifdef USE_AS_WMEMCHR
|
||||
# define VPCMPEQ vpcmpeqd
|
||||
+# define VPBROADCAST vpbroadcastd
|
||||
+# define CHAR_SIZE 4
|
||||
# else
|
||||
# define VPCMPEQ vpcmpeqb
|
||||
+# define VPBROADCAST vpbroadcastb
|
||||
+# define CHAR_SIZE 1
|
||||
+# endif
|
||||
+
|
||||
+# ifdef USE_AS_RAWMEMCHR
|
||||
+# define ERAW_PTR_REG ecx
|
||||
+# define RRAW_PTR_REG rcx
|
||||
+# define ALGN_PTR_REG rdi
|
||||
+# else
|
||||
+# define ERAW_PTR_REG edi
|
||||
+# define RRAW_PTR_REG rdi
|
||||
+# define ALGN_PTR_REG rcx
|
||||
# endif
|
||||
|
||||
# ifndef VZEROUPPER
|
||||
@@ -39,6 +53,7 @@
|
||||
# endif
|
||||
|
||||
# define VEC_SIZE 32
|
||||
+# define PAGE_SIZE 4096
|
||||
|
||||
.section SECTION(.text),"ax",@progbits
|
||||
ENTRY (MEMCHR)
|
||||
@@ -47,295 +62,349 @@ ENTRY (MEMCHR)
|
||||
test %RDX_LP, %RDX_LP
|
||||
jz L(null)
|
||||
# endif
|
||||
- movl %edi, %ecx
|
||||
- /* Broadcast CHAR to YMM0. */
|
||||
- vmovd %esi, %xmm0
|
||||
# ifdef USE_AS_WMEMCHR
|
||||
shl $2, %RDX_LP
|
||||
- vpbroadcastd %xmm0, %ymm0
|
||||
# else
|
||||
# ifdef __ILP32__
|
||||
/* Clear the upper 32 bits. */
|
||||
movl %edx, %edx
|
||||
# endif
|
||||
- vpbroadcastb %xmm0, %ymm0
|
||||
# endif
|
||||
+ /* Broadcast CHAR to YMMMATCH. */
|
||||
+ vmovd %esi, %xmm0
|
||||
+ VPBROADCAST %xmm0, %ymm0
|
||||
/* Check if we may cross page boundary with one vector load. */
|
||||
- andl $(2 * VEC_SIZE - 1), %ecx
|
||||
- cmpl $VEC_SIZE, %ecx
|
||||
- ja L(cros_page_boundary)
|
||||
+ movl %edi, %eax
|
||||
+ andl $(PAGE_SIZE - 1), %eax
|
||||
+ cmpl $(PAGE_SIZE - VEC_SIZE), %eax
|
||||
+ ja L(cross_page_boundary)
|
||||
|
||||
/* Check the first VEC_SIZE bytes. */
|
||||
- VPCMPEQ (%rdi), %ymm0, %ymm1
|
||||
+ VPCMPEQ (%rdi), %ymm0, %ymm1
|
||||
vpmovmskb %ymm1, %eax
|
||||
- testl %eax, %eax
|
||||
-
|
||||
# ifndef USE_AS_RAWMEMCHR
|
||||
- jnz L(first_vec_x0_check)
|
||||
- /* Adjust length and check the end of data. */
|
||||
- subq $VEC_SIZE, %rdx
|
||||
- jbe L(zero)
|
||||
-# else
|
||||
- jnz L(first_vec_x0)
|
||||
+ /* If length < CHAR_PER_VEC handle special. */
|
||||
+ cmpq $VEC_SIZE, %rdx
|
||||
+ jbe L(first_vec_x0)
|
||||
# endif
|
||||
-
|
||||
- /* Align data for aligned loads in the loop. */
|
||||
- addq $VEC_SIZE, %rdi
|
||||
- andl $(VEC_SIZE - 1), %ecx
|
||||
- andq $-VEC_SIZE, %rdi
|
||||
+ testl %eax, %eax
|
||||
+ jz L(aligned_more)
|
||||
+ tzcntl %eax, %eax
|
||||
+ addq %rdi, %rax
|
||||
+ VZEROUPPER_RETURN
|
||||
|
||||
# ifndef USE_AS_RAWMEMCHR
|
||||
- /* Adjust length. */
|
||||
- addq %rcx, %rdx
|
||||
+ .p2align 5
|
||||
+L(first_vec_x0):
|
||||
+ /* Check if first match was before length. */
|
||||
+ tzcntl %eax, %eax
|
||||
+ xorl %ecx, %ecx
|
||||
+ cmpl %eax, %edx
|
||||
+ leaq (%rdi, %rax), %rax
|
||||
+ cmovle %rcx, %rax
|
||||
+ VZEROUPPER_RETURN
|
||||
|
||||
- subq $(VEC_SIZE * 4), %rdx
|
||||
- jbe L(last_4x_vec_or_less)
|
||||
+L(null):
|
||||
+ xorl %eax, %eax
|
||||
+ ret
|
||||
# endif
|
||||
- jmp L(more_4x_vec)
|
||||
-
|
||||
.p2align 4
|
||||
-L(cros_page_boundary):
|
||||
- andl $(VEC_SIZE - 1), %ecx
|
||||
- andq $-VEC_SIZE, %rdi
|
||||
- VPCMPEQ (%rdi), %ymm0, %ymm1
|
||||
+L(cross_page_boundary):
|
||||
+ /* Save pointer before aligning as its original value is necessary
|
||||
+ for computer return address if byte is found or adjusting length
|
||||
+ if it is not and this is memchr. */
|
||||
+ movq %rdi, %rcx
|
||||
+ /* Align data to VEC_SIZE - 1. ALGN_PTR_REG is rcx for memchr and
|
||||
+ rdi for rawmemchr. */
|
||||
+ orq $(VEC_SIZE - 1), %ALGN_PTR_REG
|
||||
+ VPCMPEQ -(VEC_SIZE - 1)(%ALGN_PTR_REG), %ymm0, %ymm1
|
||||
vpmovmskb %ymm1, %eax
|
||||
+# ifndef USE_AS_RAWMEMCHR
|
||||
+ /* Calculate length until end of page (length checked for a
|
||||
+ match). */
|
||||
+ leaq 1(%ALGN_PTR_REG), %rsi
|
||||
+ subq %RRAW_PTR_REG, %rsi
|
||||
+# endif
|
||||
/* Remove the leading bytes. */
|
||||
- sarl %cl, %eax
|
||||
- testl %eax, %eax
|
||||
- jz L(aligned_more)
|
||||
- tzcntl %eax, %eax
|
||||
+ sarxl %ERAW_PTR_REG, %eax, %eax
|
||||
# ifndef USE_AS_RAWMEMCHR
|
||||
/* Check the end of data. */
|
||||
- cmpq %rax, %rdx
|
||||
- jbe L(zero)
|
||||
+ cmpq %rsi, %rdx
|
||||
+ jbe L(first_vec_x0)
|
||||
# endif
|
||||
- addq %rdi, %rax
|
||||
- addq %rcx, %rax
|
||||
+ testl %eax, %eax
|
||||
+ jz L(cross_page_continue)
|
||||
+ tzcntl %eax, %eax
|
||||
+ addq %RRAW_PTR_REG, %rax
|
||||
L(return_vzeroupper):
|
||||
ZERO_UPPER_VEC_REGISTERS_RETURN
|
||||
|
||||
.p2align 4
|
||||
-L(aligned_more):
|
||||
-# ifndef USE_AS_RAWMEMCHR
|
||||
- /* Calculate "rdx + rcx - VEC_SIZE" with "rdx - (VEC_SIZE - rcx)"
|
||||
- instead of "(rdx + rcx) - VEC_SIZE" to void possible addition
|
||||
- overflow. */
|
||||
- negq %rcx
|
||||
- addq $VEC_SIZE, %rcx
|
||||
+L(first_vec_x1):
|
||||
+ tzcntl %eax, %eax
|
||||
+ incq %rdi
|
||||
+ addq %rdi, %rax
|
||||
+ VZEROUPPER_RETURN
|
||||
|
||||
- /* Check the end of data. */
|
||||
- subq %rcx, %rdx
|
||||
- jbe L(zero)
|
||||
-# endif
|
||||
+ .p2align 4
|
||||
+L(first_vec_x2):
|
||||
+ tzcntl %eax, %eax
|
||||
+ addq $(VEC_SIZE + 1), %rdi
|
||||
+ addq %rdi, %rax
|
||||
+ VZEROUPPER_RETURN
|
||||
+
|
||||
+ .p2align 4
|
||||
+L(first_vec_x3):
|
||||
+ tzcntl %eax, %eax
|
||||
+ addq $(VEC_SIZE * 2 + 1), %rdi
|
||||
+ addq %rdi, %rax
|
||||
+ VZEROUPPER_RETURN
|
||||
|
||||
- addq $VEC_SIZE, %rdi
|
||||
|
||||
-# ifndef USE_AS_RAWMEMCHR
|
||||
- subq $(VEC_SIZE * 4), %rdx
|
||||
- jbe L(last_4x_vec_or_less)
|
||||
-# endif
|
||||
+ .p2align 4
|
||||
+L(first_vec_x4):
|
||||
+ tzcntl %eax, %eax
|
||||
+ addq $(VEC_SIZE * 3 + 1), %rdi
|
||||
+ addq %rdi, %rax
|
||||
+ VZEROUPPER_RETURN
|
||||
|
||||
-L(more_4x_vec):
|
||||
+ .p2align 4
|
||||
+L(aligned_more):
|
||||
/* Check the first 4 * VEC_SIZE. Only one VEC_SIZE at a time
|
||||
since data is only aligned to VEC_SIZE. */
|
||||
- VPCMPEQ (%rdi), %ymm0, %ymm1
|
||||
- vpmovmskb %ymm1, %eax
|
||||
- testl %eax, %eax
|
||||
- jnz L(first_vec_x0)
|
||||
|
||||
- VPCMPEQ VEC_SIZE(%rdi), %ymm0, %ymm1
|
||||
+# ifndef USE_AS_RAWMEMCHR
|
||||
+L(cross_page_continue):
|
||||
+ /* Align data to VEC_SIZE - 1. */
|
||||
+ xorl %ecx, %ecx
|
||||
+ subl %edi, %ecx
|
||||
+ orq $(VEC_SIZE - 1), %rdi
|
||||
+ /* esi is for adjusting length to see if near the end. */
|
||||
+ leal (VEC_SIZE * 4 + 1)(%rdi, %rcx), %esi
|
||||
+# else
|
||||
+ orq $(VEC_SIZE - 1), %rdi
|
||||
+L(cross_page_continue):
|
||||
+# endif
|
||||
+ /* Load first VEC regardless. */
|
||||
+ VPCMPEQ 1(%rdi), %ymm0, %ymm1
|
||||
vpmovmskb %ymm1, %eax
|
||||
+# ifndef USE_AS_RAWMEMCHR
|
||||
+ /* Adjust length. If near end handle specially. */
|
||||
+ subq %rsi, %rdx
|
||||
+ jbe L(last_4x_vec_or_less)
|
||||
+# endif
|
||||
testl %eax, %eax
|
||||
jnz L(first_vec_x1)
|
||||
|
||||
- VPCMPEQ (VEC_SIZE * 2)(%rdi), %ymm0, %ymm1
|
||||
+ VPCMPEQ (VEC_SIZE + 1)(%rdi), %ymm0, %ymm1
|
||||
vpmovmskb %ymm1, %eax
|
||||
testl %eax, %eax
|
||||
jnz L(first_vec_x2)
|
||||
|
||||
- VPCMPEQ (VEC_SIZE * 3)(%rdi), %ymm0, %ymm1
|
||||
+ VPCMPEQ (VEC_SIZE * 2 + 1)(%rdi), %ymm0, %ymm1
|
||||
vpmovmskb %ymm1, %eax
|
||||
testl %eax, %eax
|
||||
jnz L(first_vec_x3)
|
||||
|
||||
- addq $(VEC_SIZE * 4), %rdi
|
||||
+ VPCMPEQ (VEC_SIZE * 3 + 1)(%rdi), %ymm0, %ymm1
|
||||
+ vpmovmskb %ymm1, %eax
|
||||
+ testl %eax, %eax
|
||||
+ jnz L(first_vec_x4)
|
||||
|
||||
# ifndef USE_AS_RAWMEMCHR
|
||||
+ /* Check if at last VEC_SIZE * 4 length. */
|
||||
subq $(VEC_SIZE * 4), %rdx
|
||||
- jbe L(last_4x_vec_or_less)
|
||||
-# endif
|
||||
-
|
||||
- /* Align data to 4 * VEC_SIZE. */
|
||||
- movq %rdi, %rcx
|
||||
- andl $(4 * VEC_SIZE - 1), %ecx
|
||||
- andq $-(4 * VEC_SIZE), %rdi
|
||||
-
|
||||
-# ifndef USE_AS_RAWMEMCHR
|
||||
- /* Adjust length. */
|
||||
+ jbe L(last_4x_vec_or_less_cmpeq)
|
||||
+ /* Align data to VEC_SIZE * 4 - 1 for the loop and readjust
|
||||
+ length. */
|
||||
+ incq %rdi
|
||||
+ movl %edi, %ecx
|
||||
+ orq $(VEC_SIZE * 4 - 1), %rdi
|
||||
+ andl $(VEC_SIZE * 4 - 1), %ecx
|
||||
addq %rcx, %rdx
|
||||
+# else
|
||||
+ /* Align data to VEC_SIZE * 4 - 1 for loop. */
|
||||
+ incq %rdi
|
||||
+ orq $(VEC_SIZE * 4 - 1), %rdi
|
||||
# endif
|
||||
|
||||
+ /* Compare 4 * VEC at a time forward. */
|
||||
.p2align 4
|
||||
L(loop_4x_vec):
|
||||
- /* Compare 4 * VEC at a time forward. */
|
||||
- VPCMPEQ (%rdi), %ymm0, %ymm1
|
||||
- VPCMPEQ VEC_SIZE(%rdi), %ymm0, %ymm2
|
||||
- VPCMPEQ (VEC_SIZE * 2)(%rdi), %ymm0, %ymm3
|
||||
- VPCMPEQ (VEC_SIZE * 3)(%rdi), %ymm0, %ymm4
|
||||
-
|
||||
+ VPCMPEQ 1(%rdi), %ymm0, %ymm1
|
||||
+ VPCMPEQ (VEC_SIZE + 1)(%rdi), %ymm0, %ymm2
|
||||
+ VPCMPEQ (VEC_SIZE * 2 + 1)(%rdi), %ymm0, %ymm3
|
||||
+ VPCMPEQ (VEC_SIZE * 3 + 1)(%rdi), %ymm0, %ymm4
|
||||
vpor %ymm1, %ymm2, %ymm5
|
||||
vpor %ymm3, %ymm4, %ymm6
|
||||
vpor %ymm5, %ymm6, %ymm5
|
||||
|
||||
- vpmovmskb %ymm5, %eax
|
||||
- testl %eax, %eax
|
||||
- jnz L(4x_vec_end)
|
||||
-
|
||||
- addq $(VEC_SIZE * 4), %rdi
|
||||
-
|
||||
+ vpmovmskb %ymm5, %ecx
|
||||
# ifdef USE_AS_RAWMEMCHR
|
||||
- jmp L(loop_4x_vec)
|
||||
+ subq $-(VEC_SIZE * 4), %rdi
|
||||
+ testl %ecx, %ecx
|
||||
+ jz L(loop_4x_vec)
|
||||
# else
|
||||
- subq $(VEC_SIZE * 4), %rdx
|
||||
- ja L(loop_4x_vec)
|
||||
+ testl %ecx, %ecx
|
||||
+ jnz L(loop_4x_vec_end)
|
||||
|
||||
-L(last_4x_vec_or_less):
|
||||
- /* Less than 4 * VEC and aligned to VEC_SIZE. */
|
||||
- addl $(VEC_SIZE * 2), %edx
|
||||
- jle L(last_2x_vec)
|
||||
+ subq $-(VEC_SIZE * 4), %rdi
|
||||
|
||||
- VPCMPEQ (%rdi), %ymm0, %ymm1
|
||||
- vpmovmskb %ymm1, %eax
|
||||
- testl %eax, %eax
|
||||
- jnz L(first_vec_x0)
|
||||
+ subq $(VEC_SIZE * 4), %rdx
|
||||
+ ja L(loop_4x_vec)
|
||||
|
||||
- VPCMPEQ VEC_SIZE(%rdi), %ymm0, %ymm1
|
||||
+ /* Fall through into less than 4 remaining vectors of length case.
|
||||
+ */
|
||||
+ VPCMPEQ (VEC_SIZE * 0 + 1)(%rdi), %ymm0, %ymm1
|
||||
vpmovmskb %ymm1, %eax
|
||||
+ .p2align 4
|
||||
+L(last_4x_vec_or_less):
|
||||
+ /* Check if first VEC contained match. */
|
||||
testl %eax, %eax
|
||||
- jnz L(first_vec_x1)
|
||||
+ jnz L(first_vec_x1_check)
|
||||
|
||||
- VPCMPEQ (VEC_SIZE * 2)(%rdi), %ymm0, %ymm1
|
||||
- vpmovmskb %ymm1, %eax
|
||||
- testl %eax, %eax
|
||||
+ /* If remaining length > VEC_SIZE * 2. */
|
||||
+ addl $(VEC_SIZE * 2), %edx
|
||||
+ jg L(last_4x_vec)
|
||||
|
||||
- jnz L(first_vec_x2_check)
|
||||
- subl $VEC_SIZE, %edx
|
||||
- jle L(zero)
|
||||
+L(last_2x_vec):
|
||||
+ /* If remaining length < VEC_SIZE. */
|
||||
+ addl $VEC_SIZE, %edx
|
||||
+ jle L(zero_end)
|
||||
|
||||
- VPCMPEQ (VEC_SIZE * 3)(%rdi), %ymm0, %ymm1
|
||||
+ /* Check VEC2 and compare any match with remaining length. */
|
||||
+ VPCMPEQ (VEC_SIZE + 1)(%rdi), %ymm0, %ymm1
|
||||
vpmovmskb %ymm1, %eax
|
||||
- testl %eax, %eax
|
||||
-
|
||||
- jnz L(first_vec_x3_check)
|
||||
- xorl %eax, %eax
|
||||
+ tzcntl %eax, %eax
|
||||
+ cmpl %eax, %edx
|
||||
+ jbe L(set_zero_end)
|
||||
+ addq $(VEC_SIZE + 1), %rdi
|
||||
+ addq %rdi, %rax
|
||||
+L(zero_end):
|
||||
VZEROUPPER_RETURN
|
||||
|
||||
.p2align 4
|
||||
-L(last_2x_vec):
|
||||
- addl $(VEC_SIZE * 2), %edx
|
||||
- VPCMPEQ (%rdi), %ymm0, %ymm1
|
||||
+L(loop_4x_vec_end):
|
||||
+# endif
|
||||
+ /* rawmemchr will fall through into this if match was found in
|
||||
+ loop. */
|
||||
+
|
||||
vpmovmskb %ymm1, %eax
|
||||
testl %eax, %eax
|
||||
+ jnz L(last_vec_x1_return)
|
||||
|
||||
- jnz L(first_vec_x0_check)
|
||||
- subl $VEC_SIZE, %edx
|
||||
- jle L(zero)
|
||||
-
|
||||
- VPCMPEQ VEC_SIZE(%rdi), %ymm0, %ymm1
|
||||
- vpmovmskb %ymm1, %eax
|
||||
+ vpmovmskb %ymm2, %eax
|
||||
testl %eax, %eax
|
||||
- jnz L(first_vec_x1_check)
|
||||
- xorl %eax, %eax
|
||||
- VZEROUPPER_RETURN
|
||||
+ jnz L(last_vec_x2_return)
|
||||
|
||||
- .p2align 4
|
||||
-L(first_vec_x0_check):
|
||||
- tzcntl %eax, %eax
|
||||
- /* Check the end of data. */
|
||||
- cmpq %rax, %rdx
|
||||
- jbe L(zero)
|
||||
+ vpmovmskb %ymm3, %eax
|
||||
+ /* Combine VEC3 matches (eax) with VEC4 matches (ecx). */
|
||||
+ salq $32, %rcx
|
||||
+ orq %rcx, %rax
|
||||
+ tzcntq %rax, %rax
|
||||
+# ifdef USE_AS_RAWMEMCHR
|
||||
+ subq $(VEC_SIZE * 2 - 1), %rdi
|
||||
+# else
|
||||
+ subq $-(VEC_SIZE * 2 + 1), %rdi
|
||||
+# endif
|
||||
addq %rdi, %rax
|
||||
VZEROUPPER_RETURN
|
||||
+# ifndef USE_AS_RAWMEMCHR
|
||||
|
||||
.p2align 4
|
||||
L(first_vec_x1_check):
|
||||
tzcntl %eax, %eax
|
||||
- /* Check the end of data. */
|
||||
- cmpq %rax, %rdx
|
||||
- jbe L(zero)
|
||||
- addq $VEC_SIZE, %rax
|
||||
+ /* Adjust length. */
|
||||
+ subl $-(VEC_SIZE * 4), %edx
|
||||
+ /* Check if match within remaining length. */
|
||||
+ cmpl %eax, %edx
|
||||
+ jbe L(set_zero_end)
|
||||
+ incq %rdi
|
||||
addq %rdi, %rax
|
||||
VZEROUPPER_RETURN
|
||||
+ .p2align 4
|
||||
+L(set_zero_end):
|
||||
+ xorl %eax, %eax
|
||||
+ VZEROUPPER_RETURN
|
||||
+# endif
|
||||
|
||||
.p2align 4
|
||||
-L(first_vec_x2_check):
|
||||
+L(last_vec_x1_return):
|
||||
tzcntl %eax, %eax
|
||||
- /* Check the end of data. */
|
||||
- cmpq %rax, %rdx
|
||||
- jbe L(zero)
|
||||
- addq $(VEC_SIZE * 2), %rax
|
||||
+# ifdef USE_AS_RAWMEMCHR
|
||||
+ subq $(VEC_SIZE * 4 - 1), %rdi
|
||||
+# else
|
||||
+ incq %rdi
|
||||
+# endif
|
||||
addq %rdi, %rax
|
||||
VZEROUPPER_RETURN
|
||||
|
||||
.p2align 4
|
||||
-L(first_vec_x3_check):
|
||||
+L(last_vec_x2_return):
|
||||
tzcntl %eax, %eax
|
||||
- /* Check the end of data. */
|
||||
- cmpq %rax, %rdx
|
||||
- jbe L(zero)
|
||||
- addq $(VEC_SIZE * 3), %rax
|
||||
+# ifdef USE_AS_RAWMEMCHR
|
||||
+ subq $(VEC_SIZE * 3 - 1), %rdi
|
||||
+# else
|
||||
+ subq $-(VEC_SIZE + 1), %rdi
|
||||
+# endif
|
||||
addq %rdi, %rax
|
||||
VZEROUPPER_RETURN
|
||||
|
||||
+# ifndef USE_AS_RAWMEMCHR
|
||||
.p2align 4
|
||||
-L(zero):
|
||||
- xorl %eax, %eax
|
||||
- jmp L(return_vzeroupper)
|
||||
+L(last_4x_vec_or_less_cmpeq):
|
||||
+ VPCMPEQ (VEC_SIZE * 4 + 1)(%rdi), %ymm0, %ymm1
|
||||
+ vpmovmskb %ymm1, %eax
|
||||
+ subq $-(VEC_SIZE * 4), %rdi
|
||||
+ /* Check first VEC regardless. */
|
||||
+ testl %eax, %eax
|
||||
+ jnz L(first_vec_x1_check)
|
||||
|
||||
+ /* If remaining length <= CHAR_PER_VEC * 2. */
|
||||
+ addl $(VEC_SIZE * 2), %edx
|
||||
+ jle L(last_2x_vec)
|
||||
.p2align 4
|
||||
-L(null):
|
||||
- xorl %eax, %eax
|
||||
- ret
|
||||
-# endif
|
||||
+L(last_4x_vec):
|
||||
+ VPCMPEQ (VEC_SIZE + 1)(%rdi), %ymm0, %ymm1
|
||||
+ vpmovmskb %ymm1, %eax
|
||||
+ testl %eax, %eax
|
||||
+ jnz L(last_vec_x2_return)
|
||||
|
||||
- .p2align 4
|
||||
-L(first_vec_x0):
|
||||
- tzcntl %eax, %eax
|
||||
- addq %rdi, %rax
|
||||
- VZEROUPPER_RETURN
|
||||
+ VPCMPEQ (VEC_SIZE * 2 + 1)(%rdi), %ymm0, %ymm1
|
||||
+ vpmovmskb %ymm1, %eax
|
||||
|
||||
- .p2align 4
|
||||
-L(first_vec_x1):
|
||||
- tzcntl %eax, %eax
|
||||
- addq $VEC_SIZE, %rax
|
||||
- addq %rdi, %rax
|
||||
- VZEROUPPER_RETURN
|
||||
+ /* Create mask for possible matches within remaining length. */
|
||||
+ movq $-1, %rcx
|
||||
+ bzhiq %rdx, %rcx, %rcx
|
||||
|
||||
- .p2align 4
|
||||
-L(first_vec_x2):
|
||||
+ /* Test matches in data against length match. */
|
||||
+ andl %ecx, %eax
|
||||
+ jnz L(last_vec_x3)
|
||||
+
|
||||
+ /* if remaining length <= VEC_SIZE * 3 (Note this is after
|
||||
+ remaining length was found to be > VEC_SIZE * 2. */
|
||||
+ subl $VEC_SIZE, %edx
|
||||
+ jbe L(zero_end2)
|
||||
+
|
||||
+ VPCMPEQ (VEC_SIZE * 3 + 1)(%rdi), %ymm0, %ymm1
|
||||
+ vpmovmskb %ymm1, %eax
|
||||
+ /* Shift remaining length mask for last VEC. */
|
||||
+ shrq $32, %rcx
|
||||
+ andl %ecx, %eax
|
||||
+ jz L(zero_end2)
|
||||
tzcntl %eax, %eax
|
||||
- addq $(VEC_SIZE * 2), %rax
|
||||
+ addq $(VEC_SIZE * 3 + 1), %rdi
|
||||
addq %rdi, %rax
|
||||
+L(zero_end2):
|
||||
VZEROUPPER_RETURN
|
||||
|
||||
.p2align 4
|
||||
-L(4x_vec_end):
|
||||
- vpmovmskb %ymm1, %eax
|
||||
- testl %eax, %eax
|
||||
- jnz L(first_vec_x0)
|
||||
- vpmovmskb %ymm2, %eax
|
||||
- testl %eax, %eax
|
||||
- jnz L(first_vec_x1)
|
||||
- vpmovmskb %ymm3, %eax
|
||||
- testl %eax, %eax
|
||||
- jnz L(first_vec_x2)
|
||||
- vpmovmskb %ymm4, %eax
|
||||
- testl %eax, %eax
|
||||
-L(first_vec_x3):
|
||||
+L(last_vec_x3):
|
||||
tzcntl %eax, %eax
|
||||
- addq $(VEC_SIZE * 3), %rax
|
||||
+ subq $-(VEC_SIZE * 2 + 1), %rdi
|
||||
addq %rdi, %rax
|
||||
VZEROUPPER_RETURN
|
||||
+# endif
|
||||
|
||||
END (MEMCHR)
|
||||
#endif
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,388 +0,0 @@
|
||||
From 645a158978f9520e74074e8c14047503be4db0f0 Mon Sep 17 00:00:00 2001
|
||||
From: Noah Goldstein <goldstein.w.n@gmail.com>
|
||||
Date: Wed, 9 Jun 2021 16:25:32 -0400
|
||||
Subject: [PATCH] x86: Fix overflow bug with wmemchr-sse2 and wmemchr-avx2 [BZ
|
||||
#27974]
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
This commit fixes the bug mentioned in the previous commit.
|
||||
|
||||
The previous implementations of wmemchr in these files relied
|
||||
on n * sizeof(wchar_t) which was not guranteed by the standard.
|
||||
|
||||
The new overflow tests added in the previous commit now
|
||||
pass (As well as all the other tests).
|
||||
|
||||
Signed-off-by: Noah Goldstein <goldstein.w.n@gmail.com>
|
||||
Reviewed-by: H.J. Lu <hjl.tools@gmail.com>
|
||||
---
|
||||
sysdeps/x86_64/memchr.S | 77 +++++++++++++++++++-------
|
||||
sysdeps/x86_64/multiarch/memchr-avx2.S | 58 +++++++++++++------
|
||||
2 files changed, 98 insertions(+), 37 deletions(-)
|
||||
|
||||
diff --git a/sysdeps/x86_64/memchr.S b/sysdeps/x86_64/memchr.S
|
||||
index cb320257..24f9a0c5 100644
|
||||
--- a/sysdeps/x86_64/memchr.S
|
||||
+++ b/sysdeps/x86_64/memchr.S
|
||||
@@ -21,9 +21,11 @@
|
||||
#ifdef USE_AS_WMEMCHR
|
||||
# define MEMCHR wmemchr
|
||||
# define PCMPEQ pcmpeqd
|
||||
+# define CHAR_PER_VEC 4
|
||||
#else
|
||||
# define MEMCHR memchr
|
||||
# define PCMPEQ pcmpeqb
|
||||
+# define CHAR_PER_VEC 16
|
||||
#endif
|
||||
|
||||
/* fast SSE2 version with using pmaxub and 64 byte loop */
|
||||
@@ -33,15 +35,14 @@ ENTRY(MEMCHR)
|
||||
movd %esi, %xmm1
|
||||
mov %edi, %ecx
|
||||
|
||||
+#ifdef __ILP32__
|
||||
+ /* Clear the upper 32 bits. */
|
||||
+ movl %edx, %edx
|
||||
+#endif
|
||||
#ifdef USE_AS_WMEMCHR
|
||||
test %RDX_LP, %RDX_LP
|
||||
jz L(return_null)
|
||||
- shl $2, %RDX_LP
|
||||
#else
|
||||
-# ifdef __ILP32__
|
||||
- /* Clear the upper 32 bits. */
|
||||
- movl %edx, %edx
|
||||
-# endif
|
||||
punpcklbw %xmm1, %xmm1
|
||||
test %RDX_LP, %RDX_LP
|
||||
jz L(return_null)
|
||||
@@ -60,13 +61,16 @@ ENTRY(MEMCHR)
|
||||
test %eax, %eax
|
||||
|
||||
jnz L(matches_1)
|
||||
- sub $16, %rdx
|
||||
+ sub $CHAR_PER_VEC, %rdx
|
||||
jbe L(return_null)
|
||||
add $16, %rdi
|
||||
and $15, %ecx
|
||||
and $-16, %rdi
|
||||
+#ifdef USE_AS_WMEMCHR
|
||||
+ shr $2, %ecx
|
||||
+#endif
|
||||
add %rcx, %rdx
|
||||
- sub $64, %rdx
|
||||
+ sub $(CHAR_PER_VEC * 4), %rdx
|
||||
jbe L(exit_loop)
|
||||
jmp L(loop_prolog)
|
||||
|
||||
@@ -77,16 +81,21 @@ L(crosscache):
|
||||
movdqa (%rdi), %xmm0
|
||||
|
||||
PCMPEQ %xmm1, %xmm0
|
||||
-/* Check if there is a match. */
|
||||
+ /* Check if there is a match. */
|
||||
pmovmskb %xmm0, %eax
|
||||
-/* Remove the leading bytes. */
|
||||
+ /* Remove the leading bytes. */
|
||||
sar %cl, %eax
|
||||
test %eax, %eax
|
||||
je L(unaligned_no_match)
|
||||
-/* Check which byte is a match. */
|
||||
+ /* Check which byte is a match. */
|
||||
bsf %eax, %eax
|
||||
-
|
||||
+#ifdef USE_AS_WMEMCHR
|
||||
+ mov %eax, %esi
|
||||
+ shr $2, %esi
|
||||
+ sub %rsi, %rdx
|
||||
+#else
|
||||
sub %rax, %rdx
|
||||
+#endif
|
||||
jbe L(return_null)
|
||||
add %rdi, %rax
|
||||
add %rcx, %rax
|
||||
@@ -94,15 +103,18 @@ L(crosscache):
|
||||
|
||||
.p2align 4
|
||||
L(unaligned_no_match):
|
||||
- /* "rcx" is less than 16. Calculate "rdx + rcx - 16" by using
|
||||
+ /* "rcx" is less than 16. Calculate "rdx + rcx - 16" by using
|
||||
"rdx - (16 - rcx)" instead of "(rdx + rcx) - 16" to void
|
||||
possible addition overflow. */
|
||||
neg %rcx
|
||||
add $16, %rcx
|
||||
+#ifdef USE_AS_WMEMCHR
|
||||
+ shr $2, %ecx
|
||||
+#endif
|
||||
sub %rcx, %rdx
|
||||
jbe L(return_null)
|
||||
add $16, %rdi
|
||||
- sub $64, %rdx
|
||||
+ sub $(CHAR_PER_VEC * 4), %rdx
|
||||
jbe L(exit_loop)
|
||||
|
||||
.p2align 4
|
||||
@@ -135,7 +147,7 @@ L(loop_prolog):
|
||||
test $0x3f, %rdi
|
||||
jz L(align64_loop)
|
||||
|
||||
- sub $64, %rdx
|
||||
+ sub $(CHAR_PER_VEC * 4), %rdx
|
||||
jbe L(exit_loop)
|
||||
|
||||
movdqa (%rdi), %xmm0
|
||||
@@ -167,11 +179,14 @@ L(loop_prolog):
|
||||
mov %rdi, %rcx
|
||||
and $-64, %rdi
|
||||
and $63, %ecx
|
||||
+#ifdef USE_AS_WMEMCHR
|
||||
+ shr $2, %ecx
|
||||
+#endif
|
||||
add %rcx, %rdx
|
||||
|
||||
.p2align 4
|
||||
L(align64_loop):
|
||||
- sub $64, %rdx
|
||||
+ sub $(CHAR_PER_VEC * 4), %rdx
|
||||
jbe L(exit_loop)
|
||||
movdqa (%rdi), %xmm0
|
||||
movdqa 16(%rdi), %xmm2
|
||||
@@ -218,7 +233,7 @@ L(align64_loop):
|
||||
|
||||
.p2align 4
|
||||
L(exit_loop):
|
||||
- add $32, %edx
|
||||
+ add $(CHAR_PER_VEC * 2), %edx
|
||||
jle L(exit_loop_32)
|
||||
|
||||
movdqa (%rdi), %xmm0
|
||||
@@ -238,7 +253,7 @@ L(exit_loop):
|
||||
pmovmskb %xmm3, %eax
|
||||
test %eax, %eax
|
||||
jnz L(matches32_1)
|
||||
- sub $16, %edx
|
||||
+ sub $CHAR_PER_VEC, %edx
|
||||
jle L(return_null)
|
||||
|
||||
PCMPEQ 48(%rdi), %xmm1
|
||||
@@ -250,13 +265,13 @@ L(exit_loop):
|
||||
|
||||
.p2align 4
|
||||
L(exit_loop_32):
|
||||
- add $32, %edx
|
||||
+ add $(CHAR_PER_VEC * 2), %edx
|
||||
movdqa (%rdi), %xmm0
|
||||
PCMPEQ %xmm1, %xmm0
|
||||
pmovmskb %xmm0, %eax
|
||||
test %eax, %eax
|
||||
jnz L(matches_1)
|
||||
- sub $16, %edx
|
||||
+ sub $CHAR_PER_VEC, %edx
|
||||
jbe L(return_null)
|
||||
|
||||
PCMPEQ 16(%rdi), %xmm1
|
||||
@@ -293,7 +308,13 @@ L(matches32):
|
||||
.p2align 4
|
||||
L(matches_1):
|
||||
bsf %eax, %eax
|
||||
+#ifdef USE_AS_WMEMCHR
|
||||
+ mov %eax, %esi
|
||||
+ shr $2, %esi
|
||||
+ sub %rsi, %rdx
|
||||
+#else
|
||||
sub %rax, %rdx
|
||||
+#endif
|
||||
jbe L(return_null)
|
||||
add %rdi, %rax
|
||||
ret
|
||||
@@ -301,7 +322,13 @@ L(matches_1):
|
||||
.p2align 4
|
||||
L(matches16_1):
|
||||
bsf %eax, %eax
|
||||
+#ifdef USE_AS_WMEMCHR
|
||||
+ mov %eax, %esi
|
||||
+ shr $2, %esi
|
||||
+ sub %rsi, %rdx
|
||||
+#else
|
||||
sub %rax, %rdx
|
||||
+#endif
|
||||
jbe L(return_null)
|
||||
lea 16(%rdi, %rax), %rax
|
||||
ret
|
||||
@@ -309,7 +336,13 @@ L(matches16_1):
|
||||
.p2align 4
|
||||
L(matches32_1):
|
||||
bsf %eax, %eax
|
||||
+#ifdef USE_AS_WMEMCHR
|
||||
+ mov %eax, %esi
|
||||
+ shr $2, %esi
|
||||
+ sub %rsi, %rdx
|
||||
+#else
|
||||
sub %rax, %rdx
|
||||
+#endif
|
||||
jbe L(return_null)
|
||||
lea 32(%rdi, %rax), %rax
|
||||
ret
|
||||
@@ -317,7 +350,13 @@ L(matches32_1):
|
||||
.p2align 4
|
||||
L(matches48_1):
|
||||
bsf %eax, %eax
|
||||
+#ifdef USE_AS_WMEMCHR
|
||||
+ mov %eax, %esi
|
||||
+ shr $2, %esi
|
||||
+ sub %rsi, %rdx
|
||||
+#else
|
||||
sub %rax, %rdx
|
||||
+#endif
|
||||
jbe L(return_null)
|
||||
lea 48(%rdi, %rax), %rax
|
||||
ret
|
||||
diff --git a/sysdeps/x86_64/multiarch/memchr-avx2.S b/sysdeps/x86_64/multiarch/memchr-avx2.S
|
||||
index b377f22e..16027abb 100644
|
||||
--- a/sysdeps/x86_64/multiarch/memchr-avx2.S
|
||||
+++ b/sysdeps/x86_64/multiarch/memchr-avx2.S
|
||||
@@ -54,21 +54,19 @@
|
||||
|
||||
# define VEC_SIZE 32
|
||||
# define PAGE_SIZE 4096
|
||||
+# define CHAR_PER_VEC (VEC_SIZE / CHAR_SIZE)
|
||||
|
||||
.section SECTION(.text),"ax",@progbits
|
||||
ENTRY (MEMCHR)
|
||||
# ifndef USE_AS_RAWMEMCHR
|
||||
/* Check for zero length. */
|
||||
- test %RDX_LP, %RDX_LP
|
||||
- jz L(null)
|
||||
-# endif
|
||||
-# ifdef USE_AS_WMEMCHR
|
||||
- shl $2, %RDX_LP
|
||||
-# else
|
||||
# ifdef __ILP32__
|
||||
- /* Clear the upper 32 bits. */
|
||||
- movl %edx, %edx
|
||||
+ /* Clear upper bits. */
|
||||
+ and %RDX_LP, %RDX_LP
|
||||
+# else
|
||||
+ test %RDX_LP, %RDX_LP
|
||||
# endif
|
||||
+ jz L(null)
|
||||
# endif
|
||||
/* Broadcast CHAR to YMMMATCH. */
|
||||
vmovd %esi, %xmm0
|
||||
@@ -84,7 +82,7 @@ ENTRY (MEMCHR)
|
||||
vpmovmskb %ymm1, %eax
|
||||
# ifndef USE_AS_RAWMEMCHR
|
||||
/* If length < CHAR_PER_VEC handle special. */
|
||||
- cmpq $VEC_SIZE, %rdx
|
||||
+ cmpq $CHAR_PER_VEC, %rdx
|
||||
jbe L(first_vec_x0)
|
||||
# endif
|
||||
testl %eax, %eax
|
||||
@@ -98,6 +96,10 @@ ENTRY (MEMCHR)
|
||||
L(first_vec_x0):
|
||||
/* Check if first match was before length. */
|
||||
tzcntl %eax, %eax
|
||||
+# ifdef USE_AS_WMEMCHR
|
||||
+ /* NB: Multiply length by 4 to get byte count. */
|
||||
+ sall $2, %edx
|
||||
+# endif
|
||||
xorl %ecx, %ecx
|
||||
cmpl %eax, %edx
|
||||
leaq (%rdi, %rax), %rax
|
||||
@@ -110,12 +112,12 @@ L(null):
|
||||
# endif
|
||||
.p2align 4
|
||||
L(cross_page_boundary):
|
||||
- /* Save pointer before aligning as its original value is necessary
|
||||
- for computer return address if byte is found or adjusting length
|
||||
- if it is not and this is memchr. */
|
||||
+ /* Save pointer before aligning as its original value is
|
||||
+ necessary for computer return address if byte is found or
|
||||
+ adjusting length if it is not and this is memchr. */
|
||||
movq %rdi, %rcx
|
||||
- /* Align data to VEC_SIZE - 1. ALGN_PTR_REG is rcx for memchr and
|
||||
- rdi for rawmemchr. */
|
||||
+ /* Align data to VEC_SIZE - 1. ALGN_PTR_REG is rcx for memchr
|
||||
+ and rdi for rawmemchr. */
|
||||
orq $(VEC_SIZE - 1), %ALGN_PTR_REG
|
||||
VPCMPEQ -(VEC_SIZE - 1)(%ALGN_PTR_REG), %ymm0, %ymm1
|
||||
vpmovmskb %ymm1, %eax
|
||||
@@ -124,6 +126,10 @@ L(cross_page_boundary):
|
||||
match). */
|
||||
leaq 1(%ALGN_PTR_REG), %rsi
|
||||
subq %RRAW_PTR_REG, %rsi
|
||||
+# ifdef USE_AS_WMEMCHR
|
||||
+ /* NB: Divide bytes by 4 to get wchar_t count. */
|
||||
+ shrl $2, %esi
|
||||
+# endif
|
||||
# endif
|
||||
/* Remove the leading bytes. */
|
||||
sarxl %ERAW_PTR_REG, %eax, %eax
|
||||
@@ -181,6 +187,10 @@ L(cross_page_continue):
|
||||
orq $(VEC_SIZE - 1), %rdi
|
||||
/* esi is for adjusting length to see if near the end. */
|
||||
leal (VEC_SIZE * 4 + 1)(%rdi, %rcx), %esi
|
||||
+# ifdef USE_AS_WMEMCHR
|
||||
+ /* NB: Divide bytes by 4 to get the wchar_t count. */
|
||||
+ sarl $2, %esi
|
||||
+# endif
|
||||
# else
|
||||
orq $(VEC_SIZE - 1), %rdi
|
||||
L(cross_page_continue):
|
||||
@@ -213,7 +223,7 @@ L(cross_page_continue):
|
||||
|
||||
# ifndef USE_AS_RAWMEMCHR
|
||||
/* Check if at last VEC_SIZE * 4 length. */
|
||||
- subq $(VEC_SIZE * 4), %rdx
|
||||
+ subq $(CHAR_PER_VEC * 4), %rdx
|
||||
jbe L(last_4x_vec_or_less_cmpeq)
|
||||
/* Align data to VEC_SIZE * 4 - 1 for the loop and readjust
|
||||
length. */
|
||||
@@ -221,6 +231,10 @@ L(cross_page_continue):
|
||||
movl %edi, %ecx
|
||||
orq $(VEC_SIZE * 4 - 1), %rdi
|
||||
andl $(VEC_SIZE * 4 - 1), %ecx
|
||||
+# ifdef USE_AS_WMEMCHR
|
||||
+ /* NB: Divide bytes by 4 to get the wchar_t count. */
|
||||
+ sarl $2, %ecx
|
||||
+# endif
|
||||
addq %rcx, %rdx
|
||||
# else
|
||||
/* Align data to VEC_SIZE * 4 - 1 for loop. */
|
||||
@@ -250,15 +264,19 @@ L(loop_4x_vec):
|
||||
|
||||
subq $-(VEC_SIZE * 4), %rdi
|
||||
|
||||
- subq $(VEC_SIZE * 4), %rdx
|
||||
+ subq $(CHAR_PER_VEC * 4), %rdx
|
||||
ja L(loop_4x_vec)
|
||||
|
||||
- /* Fall through into less than 4 remaining vectors of length case.
|
||||
- */
|
||||
+ /* Fall through into less than 4 remaining vectors of length
|
||||
+ case. */
|
||||
VPCMPEQ (VEC_SIZE * 0 + 1)(%rdi), %ymm0, %ymm1
|
||||
vpmovmskb %ymm1, %eax
|
||||
.p2align 4
|
||||
L(last_4x_vec_or_less):
|
||||
+# ifdef USE_AS_WMEMCHR
|
||||
+ /* NB: Multiply length by 4 to get byte count. */
|
||||
+ sall $2, %edx
|
||||
+# endif
|
||||
/* Check if first VEC contained match. */
|
||||
testl %eax, %eax
|
||||
jnz L(first_vec_x1_check)
|
||||
@@ -355,6 +373,10 @@ L(last_vec_x2_return):
|
||||
L(last_4x_vec_or_less_cmpeq):
|
||||
VPCMPEQ (VEC_SIZE * 4 + 1)(%rdi), %ymm0, %ymm1
|
||||
vpmovmskb %ymm1, %eax
|
||||
+# ifdef USE_AS_WMEMCHR
|
||||
+ /* NB: Multiply length by 4 to get byte count. */
|
||||
+ sall $2, %edx
|
||||
+# endif
|
||||
subq $-(VEC_SIZE * 4), %rdi
|
||||
/* Check first VEC regardless. */
|
||||
testl %eax, %eax
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,767 +0,0 @@
|
||||
From aaa23c35071537e2dcf5807e956802ed215210aa Mon Sep 17 00:00:00 2001
|
||||
From: Noah Goldstein <goldstein.w.n@gmail.com>
|
||||
Date: Mon, 19 Apr 2021 19:36:07 -0400
|
||||
Subject: [PATCH] x86: Optimize strlen-avx2.S
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
No bug. This commit optimizes strlen-avx2.S. The optimizations are
|
||||
mostly small things but they add up to roughly 10-30% performance
|
||||
improvement for strlen. The results for strnlen are bit more
|
||||
ambiguous. test-strlen, test-strnlen, test-wcslen, and test-wcsnlen
|
||||
are all passing.
|
||||
|
||||
Signed-off-by: Noah Goldstein <goldstein.w.n@gmail.com>
|
||||
---
|
||||
sysdeps/x86_64/multiarch/ifunc-impl-list.c | 16 +-
|
||||
sysdeps/x86_64/multiarch/strlen-avx2.S | 532 +++++++++++++--------
|
||||
2 files changed, 334 insertions(+), 214 deletions(-)
|
||||
|
||||
diff --git a/sysdeps/x86_64/multiarch/ifunc-impl-list.c b/sysdeps/x86_64/multiarch/ifunc-impl-list.c
|
||||
index cbfc1a5d..f1a6460a 100644
|
||||
--- a/sysdeps/x86_64/multiarch/ifunc-impl-list.c
|
||||
+++ b/sysdeps/x86_64/multiarch/ifunc-impl-list.c
|
||||
@@ -285,10 +285,12 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
|
||||
/* Support sysdeps/x86_64/multiarch/strlen.c. */
|
||||
IFUNC_IMPL (i, name, strlen,
|
||||
IFUNC_IMPL_ADD (array, i, strlen,
|
||||
- CPU_FEATURE_USABLE (AVX2),
|
||||
+ (CPU_FEATURE_USABLE (AVX2)
|
||||
+ && CPU_FEATURE_USABLE (BMI2)),
|
||||
__strlen_avx2)
|
||||
IFUNC_IMPL_ADD (array, i, strlen,
|
||||
(CPU_FEATURE_USABLE (AVX2)
|
||||
+ && CPU_FEATURE_USABLE (BMI2)
|
||||
&& CPU_FEATURE_USABLE (RTM)),
|
||||
__strlen_avx2_rtm)
|
||||
IFUNC_IMPL_ADD (array, i, strlen,
|
||||
@@ -301,10 +303,12 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
|
||||
/* Support sysdeps/x86_64/multiarch/strnlen.c. */
|
||||
IFUNC_IMPL (i, name, strnlen,
|
||||
IFUNC_IMPL_ADD (array, i, strnlen,
|
||||
- CPU_FEATURE_USABLE (AVX2),
|
||||
+ (CPU_FEATURE_USABLE (AVX2)
|
||||
+ && CPU_FEATURE_USABLE (BMI2)),
|
||||
__strnlen_avx2)
|
||||
IFUNC_IMPL_ADD (array, i, strnlen,
|
||||
(CPU_FEATURE_USABLE (AVX2)
|
||||
+ && CPU_FEATURE_USABLE (BMI2)
|
||||
&& CPU_FEATURE_USABLE (RTM)),
|
||||
__strnlen_avx2_rtm)
|
||||
IFUNC_IMPL_ADD (array, i, strnlen,
|
||||
@@ -640,10 +644,12 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
|
||||
/* Support sysdeps/x86_64/multiarch/wcslen.c. */
|
||||
IFUNC_IMPL (i, name, wcslen,
|
||||
IFUNC_IMPL_ADD (array, i, wcslen,
|
||||
- CPU_FEATURE_USABLE (AVX2),
|
||||
+ (CPU_FEATURE_USABLE (AVX2)
|
||||
+ && CPU_FEATURE_USABLE (BMI2)),
|
||||
__wcslen_avx2)
|
||||
IFUNC_IMPL_ADD (array, i, wcslen,
|
||||
(CPU_FEATURE_USABLE (AVX2)
|
||||
+ && CPU_FEATURE_USABLE (BMI2)
|
||||
&& CPU_FEATURE_USABLE (RTM)),
|
||||
__wcslen_avx2_rtm)
|
||||
IFUNC_IMPL_ADD (array, i, wcslen,
|
||||
@@ -656,10 +662,12 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
|
||||
/* Support sysdeps/x86_64/multiarch/wcsnlen.c. */
|
||||
IFUNC_IMPL (i, name, wcsnlen,
|
||||
IFUNC_IMPL_ADD (array, i, wcsnlen,
|
||||
- CPU_FEATURE_USABLE (AVX2),
|
||||
+ (CPU_FEATURE_USABLE (AVX2)
|
||||
+ && CPU_FEATURE_USABLE (BMI2)),
|
||||
__wcsnlen_avx2)
|
||||
IFUNC_IMPL_ADD (array, i, wcsnlen,
|
||||
(CPU_FEATURE_USABLE (AVX2)
|
||||
+ && CPU_FEATURE_USABLE (BMI2)
|
||||
&& CPU_FEATURE_USABLE (RTM)),
|
||||
__wcsnlen_avx2_rtm)
|
||||
IFUNC_IMPL_ADD (array, i, wcsnlen,
|
||||
diff --git a/sysdeps/x86_64/multiarch/strlen-avx2.S b/sysdeps/x86_64/multiarch/strlen-avx2.S
|
||||
index 82826e10..be8a5db5 100644
|
||||
--- a/sysdeps/x86_64/multiarch/strlen-avx2.S
|
||||
+++ b/sysdeps/x86_64/multiarch/strlen-avx2.S
|
||||
@@ -27,9 +27,11 @@
|
||||
# ifdef USE_AS_WCSLEN
|
||||
# define VPCMPEQ vpcmpeqd
|
||||
# define VPMINU vpminud
|
||||
+# define CHAR_SIZE 4
|
||||
# else
|
||||
# define VPCMPEQ vpcmpeqb
|
||||
# define VPMINU vpminub
|
||||
+# define CHAR_SIZE 1
|
||||
# endif
|
||||
|
||||
# ifndef VZEROUPPER
|
||||
@@ -41,349 +43,459 @@
|
||||
# endif
|
||||
|
||||
# define VEC_SIZE 32
|
||||
+# define PAGE_SIZE 4096
|
||||
|
||||
.section SECTION(.text),"ax",@progbits
|
||||
ENTRY (STRLEN)
|
||||
# ifdef USE_AS_STRNLEN
|
||||
- /* Check for zero length. */
|
||||
+ /* Check zero length. */
|
||||
test %RSI_LP, %RSI_LP
|
||||
jz L(zero)
|
||||
+ /* Store max len in R8_LP before adjusting if using WCSLEN. */
|
||||
+ mov %RSI_LP, %R8_LP
|
||||
# ifdef USE_AS_WCSLEN
|
||||
shl $2, %RSI_LP
|
||||
# elif defined __ILP32__
|
||||
/* Clear the upper 32 bits. */
|
||||
movl %esi, %esi
|
||||
# endif
|
||||
- mov %RSI_LP, %R8_LP
|
||||
# endif
|
||||
- movl %edi, %ecx
|
||||
+ movl %edi, %eax
|
||||
movq %rdi, %rdx
|
||||
vpxor %xmm0, %xmm0, %xmm0
|
||||
-
|
||||
+ /* Clear high bits from edi. Only keeping bits relevant to page
|
||||
+ cross check. */
|
||||
+ andl $(PAGE_SIZE - 1), %eax
|
||||
/* Check if we may cross page boundary with one vector load. */
|
||||
- andl $(2 * VEC_SIZE - 1), %ecx
|
||||
- cmpl $VEC_SIZE, %ecx
|
||||
- ja L(cros_page_boundary)
|
||||
+ cmpl $(PAGE_SIZE - VEC_SIZE), %eax
|
||||
+ ja L(cross_page_boundary)
|
||||
|
||||
/* Check the first VEC_SIZE bytes. */
|
||||
- VPCMPEQ (%rdi), %ymm0, %ymm1
|
||||
- vpmovmskb %ymm1, %eax
|
||||
- testl %eax, %eax
|
||||
-
|
||||
+ VPCMPEQ (%rdi), %ymm0, %ymm1
|
||||
+ vpmovmskb %ymm1, %eax
|
||||
# ifdef USE_AS_STRNLEN
|
||||
- jnz L(first_vec_x0_check)
|
||||
- /* Adjust length and check the end of data. */
|
||||
- subq $VEC_SIZE, %rsi
|
||||
- jbe L(max)
|
||||
-# else
|
||||
- jnz L(first_vec_x0)
|
||||
+ /* If length < VEC_SIZE handle special. */
|
||||
+ cmpq $VEC_SIZE, %rsi
|
||||
+ jbe L(first_vec_x0)
|
||||
# endif
|
||||
-
|
||||
- /* Align data for aligned loads in the loop. */
|
||||
- addq $VEC_SIZE, %rdi
|
||||
- andl $(VEC_SIZE - 1), %ecx
|
||||
- andq $-VEC_SIZE, %rdi
|
||||
+ /* If empty continue to aligned_more. Otherwise return bit
|
||||
+ position of first match. */
|
||||
+ testl %eax, %eax
|
||||
+ jz L(aligned_more)
|
||||
+ tzcntl %eax, %eax
|
||||
+# ifdef USE_AS_WCSLEN
|
||||
+ shrl $2, %eax
|
||||
+# endif
|
||||
+ VZEROUPPER_RETURN
|
||||
|
||||
# ifdef USE_AS_STRNLEN
|
||||
- /* Adjust length. */
|
||||
- addq %rcx, %rsi
|
||||
+L(zero):
|
||||
+ xorl %eax, %eax
|
||||
+ ret
|
||||
|
||||
- subq $(VEC_SIZE * 4), %rsi
|
||||
- jbe L(last_4x_vec_or_less)
|
||||
+ .p2align 4
|
||||
+L(first_vec_x0):
|
||||
+ /* Set bit for max len so that tzcnt will return min of max len
|
||||
+ and position of first match. */
|
||||
+ btsq %rsi, %rax
|
||||
+ tzcntl %eax, %eax
|
||||
+# ifdef USE_AS_WCSLEN
|
||||
+ shrl $2, %eax
|
||||
+# endif
|
||||
+ VZEROUPPER_RETURN
|
||||
# endif
|
||||
- jmp L(more_4x_vec)
|
||||
|
||||
.p2align 4
|
||||
-L(cros_page_boundary):
|
||||
- andl $(VEC_SIZE - 1), %ecx
|
||||
- andq $-VEC_SIZE, %rdi
|
||||
- VPCMPEQ (%rdi), %ymm0, %ymm1
|
||||
- vpmovmskb %ymm1, %eax
|
||||
- /* Remove the leading bytes. */
|
||||
- sarl %cl, %eax
|
||||
- testl %eax, %eax
|
||||
- jz L(aligned_more)
|
||||
+L(first_vec_x1):
|
||||
tzcntl %eax, %eax
|
||||
+ /* Safe to use 32 bit instructions as these are only called for
|
||||
+ size = [1, 159]. */
|
||||
# ifdef USE_AS_STRNLEN
|
||||
- /* Check the end of data. */
|
||||
- cmpq %rax, %rsi
|
||||
- jbe L(max)
|
||||
+ /* Use ecx which was computed earlier to compute correct value.
|
||||
+ */
|
||||
+ subl $(VEC_SIZE * 4 + 1), %ecx
|
||||
+ addl %ecx, %eax
|
||||
+# else
|
||||
+ subl %edx, %edi
|
||||
+ incl %edi
|
||||
+ addl %edi, %eax
|
||||
# endif
|
||||
- addq %rdi, %rax
|
||||
- addq %rcx, %rax
|
||||
- subq %rdx, %rax
|
||||
# ifdef USE_AS_WCSLEN
|
||||
- shrq $2, %rax
|
||||
+ shrl $2, %eax
|
||||
# endif
|
||||
-L(return_vzeroupper):
|
||||
- ZERO_UPPER_VEC_REGISTERS_RETURN
|
||||
+ VZEROUPPER_RETURN
|
||||
|
||||
.p2align 4
|
||||
-L(aligned_more):
|
||||
+L(first_vec_x2):
|
||||
+ tzcntl %eax, %eax
|
||||
+ /* Safe to use 32 bit instructions as these are only called for
|
||||
+ size = [1, 159]. */
|
||||
# ifdef USE_AS_STRNLEN
|
||||
- /* "rcx" is less than VEC_SIZE. Calculate "rdx + rcx - VEC_SIZE"
|
||||
- with "rdx - (VEC_SIZE - rcx)" instead of "(rdx + rcx) - VEC_SIZE"
|
||||
- to void possible addition overflow. */
|
||||
- negq %rcx
|
||||
- addq $VEC_SIZE, %rcx
|
||||
-
|
||||
- /* Check the end of data. */
|
||||
- subq %rcx, %rsi
|
||||
- jbe L(max)
|
||||
+ /* Use ecx which was computed earlier to compute correct value.
|
||||
+ */
|
||||
+ subl $(VEC_SIZE * 3 + 1), %ecx
|
||||
+ addl %ecx, %eax
|
||||
+# else
|
||||
+ subl %edx, %edi
|
||||
+ addl $(VEC_SIZE + 1), %edi
|
||||
+ addl %edi, %eax
|
||||
# endif
|
||||
+# ifdef USE_AS_WCSLEN
|
||||
+ shrl $2, %eax
|
||||
+# endif
|
||||
+ VZEROUPPER_RETURN
|
||||
|
||||
- addq $VEC_SIZE, %rdi
|
||||
+ .p2align 4
|
||||
+L(first_vec_x3):
|
||||
+ tzcntl %eax, %eax
|
||||
+ /* Safe to use 32 bit instructions as these are only called for
|
||||
+ size = [1, 159]. */
|
||||
+# ifdef USE_AS_STRNLEN
|
||||
+ /* Use ecx which was computed earlier to compute correct value.
|
||||
+ */
|
||||
+ subl $(VEC_SIZE * 2 + 1), %ecx
|
||||
+ addl %ecx, %eax
|
||||
+# else
|
||||
+ subl %edx, %edi
|
||||
+ addl $(VEC_SIZE * 2 + 1), %edi
|
||||
+ addl %edi, %eax
|
||||
+# endif
|
||||
+# ifdef USE_AS_WCSLEN
|
||||
+ shrl $2, %eax
|
||||
+# endif
|
||||
+ VZEROUPPER_RETURN
|
||||
|
||||
+ .p2align 4
|
||||
+L(first_vec_x4):
|
||||
+ tzcntl %eax, %eax
|
||||
+ /* Safe to use 32 bit instructions as these are only called for
|
||||
+ size = [1, 159]. */
|
||||
# ifdef USE_AS_STRNLEN
|
||||
- subq $(VEC_SIZE * 4), %rsi
|
||||
- jbe L(last_4x_vec_or_less)
|
||||
+ /* Use ecx which was computed earlier to compute correct value.
|
||||
+ */
|
||||
+ subl $(VEC_SIZE + 1), %ecx
|
||||
+ addl %ecx, %eax
|
||||
+# else
|
||||
+ subl %edx, %edi
|
||||
+ addl $(VEC_SIZE * 3 + 1), %edi
|
||||
+ addl %edi, %eax
|
||||
# endif
|
||||
+# ifdef USE_AS_WCSLEN
|
||||
+ shrl $2, %eax
|
||||
+# endif
|
||||
+ VZEROUPPER_RETURN
|
||||
|
||||
-L(more_4x_vec):
|
||||
+ .p2align 5
|
||||
+L(aligned_more):
|
||||
+ /* Align data to VEC_SIZE - 1. This is the same number of
|
||||
+ instructions as using andq with -VEC_SIZE but saves 4 bytes of
|
||||
+ code on the x4 check. */
|
||||
+ orq $(VEC_SIZE - 1), %rdi
|
||||
+L(cross_page_continue):
|
||||
/* Check the first 4 * VEC_SIZE. Only one VEC_SIZE at a time
|
||||
since data is only aligned to VEC_SIZE. */
|
||||
- VPCMPEQ (%rdi), %ymm0, %ymm1
|
||||
- vpmovmskb %ymm1, %eax
|
||||
- testl %eax, %eax
|
||||
- jnz L(first_vec_x0)
|
||||
-
|
||||
- VPCMPEQ VEC_SIZE(%rdi), %ymm0, %ymm1
|
||||
- vpmovmskb %ymm1, %eax
|
||||
+# ifdef USE_AS_STRNLEN
|
||||
+ /* + 1 because rdi is aligned to VEC_SIZE - 1. + CHAR_SIZE because
|
||||
+ it simplies the logic in last_4x_vec_or_less. */
|
||||
+ leaq (VEC_SIZE * 4 + CHAR_SIZE + 1)(%rdi), %rcx
|
||||
+ subq %rdx, %rcx
|
||||
+# endif
|
||||
+ /* Load first VEC regardless. */
|
||||
+ VPCMPEQ 1(%rdi), %ymm0, %ymm1
|
||||
+# ifdef USE_AS_STRNLEN
|
||||
+ /* Adjust length. If near end handle specially. */
|
||||
+ subq %rcx, %rsi
|
||||
+ jb L(last_4x_vec_or_less)
|
||||
+# endif
|
||||
+ vpmovmskb %ymm1, %eax
|
||||
testl %eax, %eax
|
||||
jnz L(first_vec_x1)
|
||||
|
||||
- VPCMPEQ (VEC_SIZE * 2)(%rdi), %ymm0, %ymm1
|
||||
- vpmovmskb %ymm1, %eax
|
||||
+ VPCMPEQ (VEC_SIZE + 1)(%rdi), %ymm0, %ymm1
|
||||
+ vpmovmskb %ymm1, %eax
|
||||
testl %eax, %eax
|
||||
jnz L(first_vec_x2)
|
||||
|
||||
- VPCMPEQ (VEC_SIZE * 3)(%rdi), %ymm0, %ymm1
|
||||
- vpmovmskb %ymm1, %eax
|
||||
+ VPCMPEQ (VEC_SIZE * 2 + 1)(%rdi), %ymm0, %ymm1
|
||||
+ vpmovmskb %ymm1, %eax
|
||||
testl %eax, %eax
|
||||
jnz L(first_vec_x3)
|
||||
|
||||
- addq $(VEC_SIZE * 4), %rdi
|
||||
-
|
||||
-# ifdef USE_AS_STRNLEN
|
||||
- subq $(VEC_SIZE * 4), %rsi
|
||||
- jbe L(last_4x_vec_or_less)
|
||||
-# endif
|
||||
-
|
||||
- /* Align data to 4 * VEC_SIZE. */
|
||||
- movq %rdi, %rcx
|
||||
- andl $(4 * VEC_SIZE - 1), %ecx
|
||||
- andq $-(4 * VEC_SIZE), %rdi
|
||||
+ VPCMPEQ (VEC_SIZE * 3 + 1)(%rdi), %ymm0, %ymm1
|
||||
+ vpmovmskb %ymm1, %eax
|
||||
+ testl %eax, %eax
|
||||
+ jnz L(first_vec_x4)
|
||||
|
||||
+ /* Align data to VEC_SIZE * 4 - 1. */
|
||||
# ifdef USE_AS_STRNLEN
|
||||
- /* Adjust length. */
|
||||
+ /* Before adjusting length check if at last VEC_SIZE * 4. */
|
||||
+ cmpq $(VEC_SIZE * 4 - 1), %rsi
|
||||
+ jbe L(last_4x_vec_or_less_load)
|
||||
+ incq %rdi
|
||||
+ movl %edi, %ecx
|
||||
+ orq $(VEC_SIZE * 4 - 1), %rdi
|
||||
+ andl $(VEC_SIZE * 4 - 1), %ecx
|
||||
+ /* Readjust length. */
|
||||
addq %rcx, %rsi
|
||||
+# else
|
||||
+ incq %rdi
|
||||
+ orq $(VEC_SIZE * 4 - 1), %rdi
|
||||
# endif
|
||||
-
|
||||
+ /* Compare 4 * VEC at a time forward. */
|
||||
.p2align 4
|
||||
L(loop_4x_vec):
|
||||
- /* Compare 4 * VEC at a time forward. */
|
||||
- vmovdqa (%rdi), %ymm1
|
||||
- vmovdqa VEC_SIZE(%rdi), %ymm2
|
||||
- vmovdqa (VEC_SIZE * 2)(%rdi), %ymm3
|
||||
- vmovdqa (VEC_SIZE * 3)(%rdi), %ymm4
|
||||
- VPMINU %ymm1, %ymm2, %ymm5
|
||||
- VPMINU %ymm3, %ymm4, %ymm6
|
||||
- VPMINU %ymm5, %ymm6, %ymm5
|
||||
-
|
||||
- VPCMPEQ %ymm5, %ymm0, %ymm5
|
||||
- vpmovmskb %ymm5, %eax
|
||||
- testl %eax, %eax
|
||||
- jnz L(4x_vec_end)
|
||||
-
|
||||
- addq $(VEC_SIZE * 4), %rdi
|
||||
-
|
||||
-# ifndef USE_AS_STRNLEN
|
||||
- jmp L(loop_4x_vec)
|
||||
-# else
|
||||
+# ifdef USE_AS_STRNLEN
|
||||
+ /* Break if at end of length. */
|
||||
subq $(VEC_SIZE * 4), %rsi
|
||||
- ja L(loop_4x_vec)
|
||||
-
|
||||
-L(last_4x_vec_or_less):
|
||||
- /* Less than 4 * VEC and aligned to VEC_SIZE. */
|
||||
- addl $(VEC_SIZE * 2), %esi
|
||||
- jle L(last_2x_vec)
|
||||
+ jb L(last_4x_vec_or_less_cmpeq)
|
||||
+# endif
|
||||
+ /* Save some code size by microfusing VPMINU with the load. Since
|
||||
+ the matches in ymm2/ymm4 can only be returned if there where no
|
||||
+ matches in ymm1/ymm3 respectively there is no issue with overlap.
|
||||
+ */
|
||||
+ vmovdqa 1(%rdi), %ymm1
|
||||
+ VPMINU (VEC_SIZE + 1)(%rdi), %ymm1, %ymm2
|
||||
+ vmovdqa (VEC_SIZE * 2 + 1)(%rdi), %ymm3
|
||||
+ VPMINU (VEC_SIZE * 3 + 1)(%rdi), %ymm3, %ymm4
|
||||
+
|
||||
+ VPMINU %ymm2, %ymm4, %ymm5
|
||||
+ VPCMPEQ %ymm5, %ymm0, %ymm5
|
||||
+ vpmovmskb %ymm5, %ecx
|
||||
|
||||
- VPCMPEQ (%rdi), %ymm0, %ymm1
|
||||
- vpmovmskb %ymm1, %eax
|
||||
- testl %eax, %eax
|
||||
- jnz L(first_vec_x0)
|
||||
+ subq $-(VEC_SIZE * 4), %rdi
|
||||
+ testl %ecx, %ecx
|
||||
+ jz L(loop_4x_vec)
|
||||
|
||||
- VPCMPEQ VEC_SIZE(%rdi), %ymm0, %ymm1
|
||||
- vpmovmskb %ymm1, %eax
|
||||
- testl %eax, %eax
|
||||
- jnz L(first_vec_x1)
|
||||
|
||||
- VPCMPEQ (VEC_SIZE * 2)(%rdi), %ymm0, %ymm1
|
||||
- vpmovmskb %ymm1, %eax
|
||||
+ VPCMPEQ %ymm1, %ymm0, %ymm1
|
||||
+ vpmovmskb %ymm1, %eax
|
||||
+ subq %rdx, %rdi
|
||||
testl %eax, %eax
|
||||
+ jnz L(last_vec_return_x0)
|
||||
|
||||
- jnz L(first_vec_x2_check)
|
||||
- subl $VEC_SIZE, %esi
|
||||
- jle L(max)
|
||||
-
|
||||
- VPCMPEQ (VEC_SIZE * 3)(%rdi), %ymm0, %ymm1
|
||||
- vpmovmskb %ymm1, %eax
|
||||
+ VPCMPEQ %ymm2, %ymm0, %ymm2
|
||||
+ vpmovmskb %ymm2, %eax
|
||||
testl %eax, %eax
|
||||
-
|
||||
- jnz L(first_vec_x3_check)
|
||||
- movq %r8, %rax
|
||||
-# ifdef USE_AS_WCSLEN
|
||||
+ jnz L(last_vec_return_x1)
|
||||
+
|
||||
+ /* Combine last 2 VEC. */
|
||||
+ VPCMPEQ %ymm3, %ymm0, %ymm3
|
||||
+ vpmovmskb %ymm3, %eax
|
||||
+ /* rcx has combined result from all 4 VEC. It will only be used if
|
||||
+ the first 3 other VEC all did not contain a match. */
|
||||
+ salq $32, %rcx
|
||||
+ orq %rcx, %rax
|
||||
+ tzcntq %rax, %rax
|
||||
+ subq $(VEC_SIZE * 2 - 1), %rdi
|
||||
+ addq %rdi, %rax
|
||||
+# ifdef USE_AS_WCSLEN
|
||||
shrq $2, %rax
|
||||
-# endif
|
||||
+# endif
|
||||
VZEROUPPER_RETURN
|
||||
|
||||
+
|
||||
+# ifdef USE_AS_STRNLEN
|
||||
.p2align 4
|
||||
-L(last_2x_vec):
|
||||
- addl $(VEC_SIZE * 2), %esi
|
||||
- VPCMPEQ (%rdi), %ymm0, %ymm1
|
||||
- vpmovmskb %ymm1, %eax
|
||||
- testl %eax, %eax
|
||||
+L(last_4x_vec_or_less_load):
|
||||
+ /* Depending on entry adjust rdi / prepare first VEC in ymm1. */
|
||||
+ subq $-(VEC_SIZE * 4), %rdi
|
||||
+L(last_4x_vec_or_less_cmpeq):
|
||||
+ VPCMPEQ 1(%rdi), %ymm0, %ymm1
|
||||
+L(last_4x_vec_or_less):
|
||||
|
||||
- jnz L(first_vec_x0_check)
|
||||
- subl $VEC_SIZE, %esi
|
||||
- jle L(max)
|
||||
+ vpmovmskb %ymm1, %eax
|
||||
+ /* If remaining length > VEC_SIZE * 2. This works if esi is off by
|
||||
+ VEC_SIZE * 4. */
|
||||
+ testl $(VEC_SIZE * 2), %esi
|
||||
+ jnz L(last_4x_vec)
|
||||
|
||||
- VPCMPEQ VEC_SIZE(%rdi), %ymm0, %ymm1
|
||||
- vpmovmskb %ymm1, %eax
|
||||
+ /* length may have been negative or positive by an offset of
|
||||
+ VEC_SIZE * 4 depending on where this was called from. This fixes
|
||||
+ that. */
|
||||
+ andl $(VEC_SIZE * 4 - 1), %esi
|
||||
testl %eax, %eax
|
||||
- jnz L(first_vec_x1_check)
|
||||
- movq %r8, %rax
|
||||
-# ifdef USE_AS_WCSLEN
|
||||
- shrq $2, %rax
|
||||
-# endif
|
||||
- VZEROUPPER_RETURN
|
||||
+ jnz L(last_vec_x1_check)
|
||||
|
||||
- .p2align 4
|
||||
-L(first_vec_x0_check):
|
||||
+ subl $VEC_SIZE, %esi
|
||||
+ jb L(max)
|
||||
+
|
||||
+ VPCMPEQ (VEC_SIZE + 1)(%rdi), %ymm0, %ymm1
|
||||
+ vpmovmskb %ymm1, %eax
|
||||
tzcntl %eax, %eax
|
||||
/* Check the end of data. */
|
||||
- cmpq %rax, %rsi
|
||||
- jbe L(max)
|
||||
+ cmpl %eax, %esi
|
||||
+ jb L(max)
|
||||
+ subq %rdx, %rdi
|
||||
+ addl $(VEC_SIZE + 1), %eax
|
||||
addq %rdi, %rax
|
||||
- subq %rdx, %rax
|
||||
# ifdef USE_AS_WCSLEN
|
||||
shrq $2, %rax
|
||||
# endif
|
||||
VZEROUPPER_RETURN
|
||||
+# endif
|
||||
|
||||
.p2align 4
|
||||
-L(first_vec_x1_check):
|
||||
+L(last_vec_return_x0):
|
||||
tzcntl %eax, %eax
|
||||
- /* Check the end of data. */
|
||||
- cmpq %rax, %rsi
|
||||
- jbe L(max)
|
||||
- addq $VEC_SIZE, %rax
|
||||
+ subq $(VEC_SIZE * 4 - 1), %rdi
|
||||
addq %rdi, %rax
|
||||
- subq %rdx, %rax
|
||||
-# ifdef USE_AS_WCSLEN
|
||||
+# ifdef USE_AS_WCSLEN
|
||||
shrq $2, %rax
|
||||
-# endif
|
||||
+# endif
|
||||
VZEROUPPER_RETURN
|
||||
|
||||
.p2align 4
|
||||
-L(first_vec_x2_check):
|
||||
+L(last_vec_return_x1):
|
||||
tzcntl %eax, %eax
|
||||
- /* Check the end of data. */
|
||||
- cmpq %rax, %rsi
|
||||
- jbe L(max)
|
||||
- addq $(VEC_SIZE * 2), %rax
|
||||
+ subq $(VEC_SIZE * 3 - 1), %rdi
|
||||
addq %rdi, %rax
|
||||
- subq %rdx, %rax
|
||||
-# ifdef USE_AS_WCSLEN
|
||||
+# ifdef USE_AS_WCSLEN
|
||||
shrq $2, %rax
|
||||
-# endif
|
||||
+# endif
|
||||
VZEROUPPER_RETURN
|
||||
|
||||
+# ifdef USE_AS_STRNLEN
|
||||
.p2align 4
|
||||
-L(first_vec_x3_check):
|
||||
+L(last_vec_x1_check):
|
||||
+
|
||||
tzcntl %eax, %eax
|
||||
/* Check the end of data. */
|
||||
- cmpq %rax, %rsi
|
||||
- jbe L(max)
|
||||
- addq $(VEC_SIZE * 3), %rax
|
||||
+ cmpl %eax, %esi
|
||||
+ jb L(max)
|
||||
+ subq %rdx, %rdi
|
||||
+ incl %eax
|
||||
addq %rdi, %rax
|
||||
- subq %rdx, %rax
|
||||
# ifdef USE_AS_WCSLEN
|
||||
shrq $2, %rax
|
||||
# endif
|
||||
VZEROUPPER_RETURN
|
||||
|
||||
- .p2align 4
|
||||
L(max):
|
||||
movq %r8, %rax
|
||||
+ VZEROUPPER_RETURN
|
||||
+
|
||||
+ .p2align 4
|
||||
+L(last_4x_vec):
|
||||
+ /* Test first 2x VEC normally. */
|
||||
+ testl %eax, %eax
|
||||
+ jnz L(last_vec_x1)
|
||||
+
|
||||
+ VPCMPEQ (VEC_SIZE + 1)(%rdi), %ymm0, %ymm1
|
||||
+ vpmovmskb %ymm1, %eax
|
||||
+ testl %eax, %eax
|
||||
+ jnz L(last_vec_x2)
|
||||
+
|
||||
+ /* Normalize length. */
|
||||
+ andl $(VEC_SIZE * 4 - 1), %esi
|
||||
+ VPCMPEQ (VEC_SIZE * 2 + 1)(%rdi), %ymm0, %ymm1
|
||||
+ vpmovmskb %ymm1, %eax
|
||||
+ testl %eax, %eax
|
||||
+ jnz L(last_vec_x3)
|
||||
+
|
||||
+ subl $(VEC_SIZE * 3), %esi
|
||||
+ jb L(max)
|
||||
+
|
||||
+ VPCMPEQ (VEC_SIZE * 3 + 1)(%rdi), %ymm0, %ymm1
|
||||
+ vpmovmskb %ymm1, %eax
|
||||
+ tzcntl %eax, %eax
|
||||
+ /* Check the end of data. */
|
||||
+ cmpl %eax, %esi
|
||||
+ jb L(max)
|
||||
+ subq %rdx, %rdi
|
||||
+ addl $(VEC_SIZE * 3 + 1), %eax
|
||||
+ addq %rdi, %rax
|
||||
# ifdef USE_AS_WCSLEN
|
||||
shrq $2, %rax
|
||||
# endif
|
||||
VZEROUPPER_RETURN
|
||||
|
||||
- .p2align 4
|
||||
-L(zero):
|
||||
- xorl %eax, %eax
|
||||
- ret
|
||||
-# endif
|
||||
|
||||
.p2align 4
|
||||
-L(first_vec_x0):
|
||||
+L(last_vec_x1):
|
||||
+ /* essentially duplicates of first_vec_x1 but use 64 bit
|
||||
+ instructions. */
|
||||
tzcntl %eax, %eax
|
||||
+ subq %rdx, %rdi
|
||||
+ incl %eax
|
||||
addq %rdi, %rax
|
||||
- subq %rdx, %rax
|
||||
-# ifdef USE_AS_WCSLEN
|
||||
+# ifdef USE_AS_WCSLEN
|
||||
shrq $2, %rax
|
||||
-# endif
|
||||
+# endif
|
||||
VZEROUPPER_RETURN
|
||||
|
||||
.p2align 4
|
||||
-L(first_vec_x1):
|
||||
+L(last_vec_x2):
|
||||
+ /* essentially duplicates of first_vec_x1 but use 64 bit
|
||||
+ instructions. */
|
||||
tzcntl %eax, %eax
|
||||
- addq $VEC_SIZE, %rax
|
||||
+ subq %rdx, %rdi
|
||||
+ addl $(VEC_SIZE + 1), %eax
|
||||
addq %rdi, %rax
|
||||
- subq %rdx, %rax
|
||||
-# ifdef USE_AS_WCSLEN
|
||||
+# ifdef USE_AS_WCSLEN
|
||||
shrq $2, %rax
|
||||
-# endif
|
||||
+# endif
|
||||
VZEROUPPER_RETURN
|
||||
|
||||
.p2align 4
|
||||
-L(first_vec_x2):
|
||||
+L(last_vec_x3):
|
||||
tzcntl %eax, %eax
|
||||
- addq $(VEC_SIZE * 2), %rax
|
||||
+ subl $(VEC_SIZE * 2), %esi
|
||||
+ /* Check the end of data. */
|
||||
+ cmpl %eax, %esi
|
||||
+ jb L(max_end)
|
||||
+ subq %rdx, %rdi
|
||||
+ addl $(VEC_SIZE * 2 + 1), %eax
|
||||
addq %rdi, %rax
|
||||
- subq %rdx, %rax
|
||||
-# ifdef USE_AS_WCSLEN
|
||||
+# ifdef USE_AS_WCSLEN
|
||||
shrq $2, %rax
|
||||
-# endif
|
||||
+# endif
|
||||
+ VZEROUPPER_RETURN
|
||||
+L(max_end):
|
||||
+ movq %r8, %rax
|
||||
VZEROUPPER_RETURN
|
||||
+# endif
|
||||
|
||||
+ /* Cold case for crossing page with first load. */
|
||||
.p2align 4
|
||||
-L(4x_vec_end):
|
||||
- VPCMPEQ %ymm1, %ymm0, %ymm1
|
||||
- vpmovmskb %ymm1, %eax
|
||||
- testl %eax, %eax
|
||||
- jnz L(first_vec_x0)
|
||||
- VPCMPEQ %ymm2, %ymm0, %ymm2
|
||||
- vpmovmskb %ymm2, %eax
|
||||
+L(cross_page_boundary):
|
||||
+ /* Align data to VEC_SIZE - 1. */
|
||||
+ orq $(VEC_SIZE - 1), %rdi
|
||||
+ VPCMPEQ -(VEC_SIZE - 1)(%rdi), %ymm0, %ymm1
|
||||
+ vpmovmskb %ymm1, %eax
|
||||
+ /* Remove the leading bytes. sarxl only uses bits [5:0] of COUNT
|
||||
+ so no need to manually mod rdx. */
|
||||
+ sarxl %edx, %eax, %eax
|
||||
+# ifdef USE_AS_STRNLEN
|
||||
testl %eax, %eax
|
||||
- jnz L(first_vec_x1)
|
||||
- VPCMPEQ %ymm3, %ymm0, %ymm3
|
||||
- vpmovmskb %ymm3, %eax
|
||||
+ jnz L(cross_page_less_vec)
|
||||
+ leaq 1(%rdi), %rcx
|
||||
+ subq %rdx, %rcx
|
||||
+ /* Check length. */
|
||||
+ cmpq %rsi, %rcx
|
||||
+ jb L(cross_page_continue)
|
||||
+ movq %r8, %rax
|
||||
+# else
|
||||
testl %eax, %eax
|
||||
- jnz L(first_vec_x2)
|
||||
- VPCMPEQ %ymm4, %ymm0, %ymm4
|
||||
- vpmovmskb %ymm4, %eax
|
||||
-L(first_vec_x3):
|
||||
+ jz L(cross_page_continue)
|
||||
tzcntl %eax, %eax
|
||||
- addq $(VEC_SIZE * 3), %rax
|
||||
- addq %rdi, %rax
|
||||
- subq %rdx, %rax
|
||||
-# ifdef USE_AS_WCSLEN
|
||||
- shrq $2, %rax
|
||||
+# ifdef USE_AS_WCSLEN
|
||||
+ shrl $2, %eax
|
||||
+# endif
|
||||
# endif
|
||||
+L(return_vzeroupper):
|
||||
+ ZERO_UPPER_VEC_REGISTERS_RETURN
|
||||
+
|
||||
+# ifdef USE_AS_STRNLEN
|
||||
+ .p2align 4
|
||||
+L(cross_page_less_vec):
|
||||
+ tzcntl %eax, %eax
|
||||
+ cmpq %rax, %rsi
|
||||
+ cmovb %esi, %eax
|
||||
+# ifdef USE_AS_WCSLEN
|
||||
+ shrl $2, %eax
|
||||
+# endif
|
||||
VZEROUPPER_RETURN
|
||||
+# endif
|
||||
|
||||
END (STRLEN)
|
||||
#endif
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,701 +0,0 @@
|
||||
From 2a76821c3081d2c0231ecd2618f52662cb48fccd Mon Sep 17 00:00:00 2001
|
||||
From: Noah Goldstein <goldstein.w.n@gmail.com>
|
||||
Date: Mon, 3 May 2021 03:03:19 -0400
|
||||
Subject: [PATCH] x86: Optimize memchr-evex.S
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
No bug. This commit optimizes memchr-evex.S. The optimizations include
|
||||
replacing some branches with cmovcc, avoiding some branches entirely
|
||||
in the less_4x_vec case, making the page cross logic less strict,
|
||||
saving some ALU in the alignment process, and most importantly
|
||||
increasing ILP in the 4x loop. test-memchr, test-rawmemchr, and
|
||||
test-wmemchr are all passing.
|
||||
|
||||
Signed-off-by: Noah Goldstein <goldstein.w.n@gmail.com>
|
||||
Reviewed-by: H.J. Lu <hjl.tools@gmail.com>
|
||||
---
|
||||
sysdeps/x86_64/multiarch/memchr-evex.S | 547 +++++++++++++++----------
|
||||
1 file changed, 322 insertions(+), 225 deletions(-)
|
||||
|
||||
diff --git a/sysdeps/x86_64/multiarch/memchr-evex.S b/sysdeps/x86_64/multiarch/memchr-evex.S
|
||||
index 6dd5d67b..81d5cd64 100644
|
||||
--- a/sysdeps/x86_64/multiarch/memchr-evex.S
|
||||
+++ b/sysdeps/x86_64/multiarch/memchr-evex.S
|
||||
@@ -26,14 +26,28 @@
|
||||
|
||||
# ifdef USE_AS_WMEMCHR
|
||||
# define VPBROADCAST vpbroadcastd
|
||||
-# define VPCMP vpcmpd
|
||||
-# define SHIFT_REG r8d
|
||||
+# define VPMINU vpminud
|
||||
+# define VPCMP vpcmpd
|
||||
+# define VPCMPEQ vpcmpeqd
|
||||
+# define CHAR_SIZE 4
|
||||
# else
|
||||
# define VPBROADCAST vpbroadcastb
|
||||
-# define VPCMP vpcmpb
|
||||
-# define SHIFT_REG ecx
|
||||
+# define VPMINU vpminub
|
||||
+# define VPCMP vpcmpb
|
||||
+# define VPCMPEQ vpcmpeqb
|
||||
+# define CHAR_SIZE 1
|
||||
# endif
|
||||
|
||||
+# ifdef USE_AS_RAWMEMCHR
|
||||
+# define RAW_PTR_REG rcx
|
||||
+# define ALGN_PTR_REG rdi
|
||||
+# else
|
||||
+# define RAW_PTR_REG rdi
|
||||
+# define ALGN_PTR_REG rcx
|
||||
+# endif
|
||||
+
|
||||
+# define XMMZERO xmm23
|
||||
+# define YMMZERO ymm23
|
||||
# define XMMMATCH xmm16
|
||||
# define YMMMATCH ymm16
|
||||
# define YMM1 ymm17
|
||||
@@ -44,6 +58,8 @@
|
||||
# define YMM6 ymm22
|
||||
|
||||
# define VEC_SIZE 32
|
||||
+# define CHAR_PER_VEC (VEC_SIZE / CHAR_SIZE)
|
||||
+# define PAGE_SIZE 4096
|
||||
|
||||
.section .text.evex,"ax",@progbits
|
||||
ENTRY (MEMCHR)
|
||||
@@ -51,11 +67,7 @@ ENTRY (MEMCHR)
|
||||
/* Check for zero length. */
|
||||
test %RDX_LP, %RDX_LP
|
||||
jz L(zero)
|
||||
-# endif
|
||||
- movl %edi, %ecx
|
||||
-# ifdef USE_AS_WMEMCHR
|
||||
- shl $2, %RDX_LP
|
||||
-# else
|
||||
+
|
||||
# ifdef __ILP32__
|
||||
/* Clear the upper 32 bits. */
|
||||
movl %edx, %edx
|
||||
@@ -64,318 +76,403 @@ ENTRY (MEMCHR)
|
||||
/* Broadcast CHAR to YMMMATCH. */
|
||||
VPBROADCAST %esi, %YMMMATCH
|
||||
/* Check if we may cross page boundary with one vector load. */
|
||||
- andl $(2 * VEC_SIZE - 1), %ecx
|
||||
- cmpl $VEC_SIZE, %ecx
|
||||
- ja L(cros_page_boundary)
|
||||
+ movl %edi, %eax
|
||||
+ andl $(PAGE_SIZE - 1), %eax
|
||||
+ cmpl $(PAGE_SIZE - VEC_SIZE), %eax
|
||||
+ ja L(cross_page_boundary)
|
||||
|
||||
/* Check the first VEC_SIZE bytes. */
|
||||
- VPCMP $0, (%rdi), %YMMMATCH, %k1
|
||||
- kmovd %k1, %eax
|
||||
- testl %eax, %eax
|
||||
-
|
||||
+ VPCMP $0, (%rdi), %YMMMATCH, %k0
|
||||
+ kmovd %k0, %eax
|
||||
# ifndef USE_AS_RAWMEMCHR
|
||||
- jnz L(first_vec_x0_check)
|
||||
- /* Adjust length and check the end of data. */
|
||||
- subq $VEC_SIZE, %rdx
|
||||
- jbe L(zero)
|
||||
+ /* If length < CHAR_PER_VEC handle special. */
|
||||
+ cmpq $CHAR_PER_VEC, %rdx
|
||||
+ jbe L(first_vec_x0)
|
||||
+# endif
|
||||
+ testl %eax, %eax
|
||||
+ jz L(aligned_more)
|
||||
+ tzcntl %eax, %eax
|
||||
+# ifdef USE_AS_WMEMCHR
|
||||
+ /* NB: Multiply bytes by CHAR_SIZE to get the wchar_t count. */
|
||||
+ leaq (%rdi, %rax, CHAR_SIZE), %rax
|
||||
# else
|
||||
- jnz L(first_vec_x0)
|
||||
+ addq %rdi, %rax
|
||||
# endif
|
||||
-
|
||||
- /* Align data for aligned loads in the loop. */
|
||||
- addq $VEC_SIZE, %rdi
|
||||
- andl $(VEC_SIZE - 1), %ecx
|
||||
- andq $-VEC_SIZE, %rdi
|
||||
+ ret
|
||||
|
||||
# ifndef USE_AS_RAWMEMCHR
|
||||
- /* Adjust length. */
|
||||
- addq %rcx, %rdx
|
||||
-
|
||||
- subq $(VEC_SIZE * 4), %rdx
|
||||
- jbe L(last_4x_vec_or_less)
|
||||
-# endif
|
||||
- jmp L(more_4x_vec)
|
||||
+L(zero):
|
||||
+ xorl %eax, %eax
|
||||
+ ret
|
||||
|
||||
+ .p2align 5
|
||||
+L(first_vec_x0):
|
||||
+ /* Check if first match was before length. */
|
||||
+ tzcntl %eax, %eax
|
||||
+ xorl %ecx, %ecx
|
||||
+ cmpl %eax, %edx
|
||||
+ leaq (%rdi, %rax, CHAR_SIZE), %rax
|
||||
+ cmovle %rcx, %rax
|
||||
+ ret
|
||||
+# else
|
||||
+ /* NB: first_vec_x0 is 17 bytes which will leave
|
||||
+ cross_page_boundary (which is relatively cold) close enough
|
||||
+ to ideal alignment. So only realign L(cross_page_boundary) if
|
||||
+ rawmemchr. */
|
||||
.p2align 4
|
||||
-L(cros_page_boundary):
|
||||
- andl $(VEC_SIZE - 1), %ecx
|
||||
+# endif
|
||||
+L(cross_page_boundary):
|
||||
+ /* Save pointer before aligning as its original value is
|
||||
+ necessary for computer return address if byte is found or
|
||||
+ adjusting length if it is not and this is memchr. */
|
||||
+ movq %rdi, %rcx
|
||||
+ /* Align data to VEC_SIZE. ALGN_PTR_REG is rcx for memchr and rdi
|
||||
+ for rawmemchr. */
|
||||
+ andq $-VEC_SIZE, %ALGN_PTR_REG
|
||||
+ VPCMP $0, (%ALGN_PTR_REG), %YMMMATCH, %k0
|
||||
+ kmovd %k0, %r8d
|
||||
# ifdef USE_AS_WMEMCHR
|
||||
- /* NB: Divide shift count by 4 since each bit in K1 represent 4
|
||||
+ /* NB: Divide shift count by 4 since each bit in K0 represent 4
|
||||
bytes. */
|
||||
- movl %ecx, %SHIFT_REG
|
||||
- sarl $2, %SHIFT_REG
|
||||
+ sarl $2, %eax
|
||||
+# endif
|
||||
+# ifndef USE_AS_RAWMEMCHR
|
||||
+ movl $(PAGE_SIZE / CHAR_SIZE), %esi
|
||||
+ subl %eax, %esi
|
||||
# endif
|
||||
- andq $-VEC_SIZE, %rdi
|
||||
- VPCMP $0, (%rdi), %YMMMATCH, %k1
|
||||
- kmovd %k1, %eax
|
||||
- /* Remove the leading bytes. */
|
||||
- sarxl %SHIFT_REG, %eax, %eax
|
||||
- testl %eax, %eax
|
||||
- jz L(aligned_more)
|
||||
- tzcntl %eax, %eax
|
||||
# ifdef USE_AS_WMEMCHR
|
||||
- /* NB: Multiply wchar_t count by 4 to get the number of bytes. */
|
||||
- sall $2, %eax
|
||||
+ andl $(CHAR_PER_VEC - 1), %eax
|
||||
# endif
|
||||
+ /* Remove the leading bytes. */
|
||||
+ sarxl %eax, %r8d, %eax
|
||||
# ifndef USE_AS_RAWMEMCHR
|
||||
/* Check the end of data. */
|
||||
- cmpq %rax, %rdx
|
||||
- jbe L(zero)
|
||||
+ cmpq %rsi, %rdx
|
||||
+ jbe L(first_vec_x0)
|
||||
+# endif
|
||||
+ testl %eax, %eax
|
||||
+ jz L(cross_page_continue)
|
||||
+ tzcntl %eax, %eax
|
||||
+# ifdef USE_AS_WMEMCHR
|
||||
+ /* NB: Multiply bytes by CHAR_SIZE to get the wchar_t count. */
|
||||
+ leaq (%RAW_PTR_REG, %rax, CHAR_SIZE), %rax
|
||||
+# else
|
||||
+ addq %RAW_PTR_REG, %rax
|
||||
# endif
|
||||
- addq %rdi, %rax
|
||||
- addq %rcx, %rax
|
||||
ret
|
||||
|
||||
.p2align 4
|
||||
-L(aligned_more):
|
||||
-# ifndef USE_AS_RAWMEMCHR
|
||||
- /* Calculate "rdx + rcx - VEC_SIZE" with "rdx - (VEC_SIZE - rcx)"
|
||||
- instead of "(rdx + rcx) - VEC_SIZE" to void possible addition
|
||||
- overflow. */
|
||||
- negq %rcx
|
||||
- addq $VEC_SIZE, %rcx
|
||||
+L(first_vec_x1):
|
||||
+ tzcntl %eax, %eax
|
||||
+ leaq VEC_SIZE(%rdi, %rax, CHAR_SIZE), %rax
|
||||
+ ret
|
||||
|
||||
- /* Check the end of data. */
|
||||
- subq %rcx, %rdx
|
||||
- jbe L(zero)
|
||||
-# endif
|
||||
+ .p2align 4
|
||||
+L(first_vec_x2):
|
||||
+ tzcntl %eax, %eax
|
||||
+ leaq (VEC_SIZE * 2)(%rdi, %rax, CHAR_SIZE), %rax
|
||||
+ ret
|
||||
|
||||
- addq $VEC_SIZE, %rdi
|
||||
+ .p2align 4
|
||||
+L(first_vec_x3):
|
||||
+ tzcntl %eax, %eax
|
||||
+ leaq (VEC_SIZE * 3)(%rdi, %rax, CHAR_SIZE), %rax
|
||||
+ ret
|
||||
|
||||
-# ifndef USE_AS_RAWMEMCHR
|
||||
- subq $(VEC_SIZE * 4), %rdx
|
||||
- jbe L(last_4x_vec_or_less)
|
||||
-# endif
|
||||
+ .p2align 4
|
||||
+L(first_vec_x4):
|
||||
+ tzcntl %eax, %eax
|
||||
+ leaq (VEC_SIZE * 4)(%rdi, %rax, CHAR_SIZE), %rax
|
||||
+ ret
|
||||
|
||||
-L(more_4x_vec):
|
||||
+ .p2align 5
|
||||
+L(aligned_more):
|
||||
/* Check the first 4 * VEC_SIZE. Only one VEC_SIZE at a time
|
||||
since data is only aligned to VEC_SIZE. */
|
||||
- VPCMP $0, (%rdi), %YMMMATCH, %k1
|
||||
- kmovd %k1, %eax
|
||||
- testl %eax, %eax
|
||||
- jnz L(first_vec_x0)
|
||||
|
||||
- VPCMP $0, VEC_SIZE(%rdi), %YMMMATCH, %k1
|
||||
- kmovd %k1, %eax
|
||||
+# ifndef USE_AS_RAWMEMCHR
|
||||
+ /* Align data to VEC_SIZE. */
|
||||
+L(cross_page_continue):
|
||||
+ xorl %ecx, %ecx
|
||||
+ subl %edi, %ecx
|
||||
+ andq $-VEC_SIZE, %rdi
|
||||
+ /* esi is for adjusting length to see if near the end. */
|
||||
+ leal (VEC_SIZE * 5)(%rdi, %rcx), %esi
|
||||
+# ifdef USE_AS_WMEMCHR
|
||||
+ /* NB: Divide bytes by 4 to get the wchar_t count. */
|
||||
+ sarl $2, %esi
|
||||
+# endif
|
||||
+# else
|
||||
+ andq $-VEC_SIZE, %rdi
|
||||
+L(cross_page_continue):
|
||||
+# endif
|
||||
+ /* Load first VEC regardless. */
|
||||
+ VPCMP $0, (VEC_SIZE)(%rdi), %YMMMATCH, %k0
|
||||
+ kmovd %k0, %eax
|
||||
+# ifndef USE_AS_RAWMEMCHR
|
||||
+ /* Adjust length. If near end handle specially. */
|
||||
+ subq %rsi, %rdx
|
||||
+ jbe L(last_4x_vec_or_less)
|
||||
+# endif
|
||||
testl %eax, %eax
|
||||
jnz L(first_vec_x1)
|
||||
|
||||
- VPCMP $0, (VEC_SIZE * 2)(%rdi), %YMMMATCH, %k1
|
||||
- kmovd %k1, %eax
|
||||
+ VPCMP $0, (VEC_SIZE * 2)(%rdi), %YMMMATCH, %k0
|
||||
+ kmovd %k0, %eax
|
||||
testl %eax, %eax
|
||||
jnz L(first_vec_x2)
|
||||
|
||||
- VPCMP $0, (VEC_SIZE * 3)(%rdi), %YMMMATCH, %k1
|
||||
- kmovd %k1, %eax
|
||||
+ VPCMP $0, (VEC_SIZE * 3)(%rdi), %YMMMATCH, %k0
|
||||
+ kmovd %k0, %eax
|
||||
testl %eax, %eax
|
||||
jnz L(first_vec_x3)
|
||||
|
||||
- addq $(VEC_SIZE * 4), %rdi
|
||||
+ VPCMP $0, (VEC_SIZE * 4)(%rdi), %YMMMATCH, %k0
|
||||
+ kmovd %k0, %eax
|
||||
+ testl %eax, %eax
|
||||
+ jnz L(first_vec_x4)
|
||||
+
|
||||
|
||||
# ifndef USE_AS_RAWMEMCHR
|
||||
- subq $(VEC_SIZE * 4), %rdx
|
||||
- jbe L(last_4x_vec_or_less)
|
||||
-# endif
|
||||
+ /* Check if at last CHAR_PER_VEC * 4 length. */
|
||||
+ subq $(CHAR_PER_VEC * 4), %rdx
|
||||
+ jbe L(last_4x_vec_or_less_cmpeq)
|
||||
+ addq $VEC_SIZE, %rdi
|
||||
|
||||
- /* Align data to 4 * VEC_SIZE. */
|
||||
- movq %rdi, %rcx
|
||||
- andl $(4 * VEC_SIZE - 1), %ecx
|
||||
+ /* Align data to VEC_SIZE * 4 for the loop and readjust length.
|
||||
+ */
|
||||
+# ifdef USE_AS_WMEMCHR
|
||||
+ movl %edi, %ecx
|
||||
andq $-(4 * VEC_SIZE), %rdi
|
||||
-
|
||||
-# ifndef USE_AS_RAWMEMCHR
|
||||
- /* Adjust length. */
|
||||
+ andl $(VEC_SIZE * 4 - 1), %ecx
|
||||
+ /* NB: Divide bytes by 4 to get the wchar_t count. */
|
||||
+ sarl $2, %ecx
|
||||
addq %rcx, %rdx
|
||||
+# else
|
||||
+ addq %rdi, %rdx
|
||||
+ andq $-(4 * VEC_SIZE), %rdi
|
||||
+ subq %rdi, %rdx
|
||||
+# endif
|
||||
+# else
|
||||
+ addq $VEC_SIZE, %rdi
|
||||
+ andq $-(4 * VEC_SIZE), %rdi
|
||||
# endif
|
||||
|
||||
+ vpxorq %XMMZERO, %XMMZERO, %XMMZERO
|
||||
+
|
||||
+ /* Compare 4 * VEC at a time forward. */
|
||||
.p2align 4
|
||||
L(loop_4x_vec):
|
||||
- /* Compare 4 * VEC at a time forward. */
|
||||
- VPCMP $0, (%rdi), %YMMMATCH, %k1
|
||||
- VPCMP $0, VEC_SIZE(%rdi), %YMMMATCH, %k2
|
||||
- kord %k1, %k2, %k5
|
||||
- VPCMP $0, (VEC_SIZE * 2)(%rdi), %YMMMATCH, %k3
|
||||
- VPCMP $0, (VEC_SIZE * 3)(%rdi), %YMMMATCH, %k4
|
||||
-
|
||||
- kord %k3, %k4, %k6
|
||||
- kortestd %k5, %k6
|
||||
- jnz L(4x_vec_end)
|
||||
-
|
||||
- addq $(VEC_SIZE * 4), %rdi
|
||||
-
|
||||
+ /* It would be possible to save some instructions using 4x VPCMP
|
||||
+ but bottleneck on port 5 makes it not woth it. */
|
||||
+ VPCMP $4, (VEC_SIZE * 4)(%rdi), %YMMMATCH, %k1
|
||||
+ /* xor will set bytes match esi to zero. */
|
||||
+ vpxorq (VEC_SIZE * 5)(%rdi), %YMMMATCH, %YMM2
|
||||
+ vpxorq (VEC_SIZE * 6)(%rdi), %YMMMATCH, %YMM3
|
||||
+ VPCMP $0, (VEC_SIZE * 7)(%rdi), %YMMMATCH, %k3
|
||||
+ /* Reduce VEC2 / VEC3 with min and VEC1 with zero mask. */
|
||||
+ VPMINU %YMM2, %YMM3, %YMM3 {%k1} {z}
|
||||
+ VPCMP $0, %YMM3, %YMMZERO, %k2
|
||||
# ifdef USE_AS_RAWMEMCHR
|
||||
- jmp L(loop_4x_vec)
|
||||
+ subq $-(VEC_SIZE * 4), %rdi
|
||||
+ kortestd %k2, %k3
|
||||
+ jz L(loop_4x_vec)
|
||||
# else
|
||||
- subq $(VEC_SIZE * 4), %rdx
|
||||
+ kortestd %k2, %k3
|
||||
+ jnz L(loop_4x_vec_end)
|
||||
+
|
||||
+ subq $-(VEC_SIZE * 4), %rdi
|
||||
+
|
||||
+ subq $(CHAR_PER_VEC * 4), %rdx
|
||||
ja L(loop_4x_vec)
|
||||
|
||||
+ /* Fall through into less than 4 remaining vectors of length case.
|
||||
+ */
|
||||
+ VPCMP $0, (VEC_SIZE * 4)(%rdi), %YMMMATCH, %k0
|
||||
+ kmovd %k0, %eax
|
||||
+ addq $(VEC_SIZE * 3), %rdi
|
||||
+ .p2align 4
|
||||
L(last_4x_vec_or_less):
|
||||
- /* Less than 4 * VEC and aligned to VEC_SIZE. */
|
||||
- addl $(VEC_SIZE * 2), %edx
|
||||
- jle L(last_2x_vec)
|
||||
-
|
||||
- VPCMP $0, (%rdi), %YMMMATCH, %k1
|
||||
- kmovd %k1, %eax
|
||||
+ /* Check if first VEC contained match. */
|
||||
testl %eax, %eax
|
||||
- jnz L(first_vec_x0)
|
||||
+ jnz L(first_vec_x1_check)
|
||||
|
||||
- VPCMP $0, VEC_SIZE(%rdi), %YMMMATCH, %k1
|
||||
- kmovd %k1, %eax
|
||||
- testl %eax, %eax
|
||||
- jnz L(first_vec_x1)
|
||||
+ /* If remaining length > CHAR_PER_VEC * 2. */
|
||||
+ addl $(CHAR_PER_VEC * 2), %edx
|
||||
+ jg L(last_4x_vec)
|
||||
|
||||
- VPCMP $0, (VEC_SIZE * 2)(%rdi), %YMMMATCH, %k1
|
||||
- kmovd %k1, %eax
|
||||
- testl %eax, %eax
|
||||
+L(last_2x_vec):
|
||||
+ /* If remaining length < CHAR_PER_VEC. */
|
||||
+ addl $CHAR_PER_VEC, %edx
|
||||
+ jle L(zero_end)
|
||||
|
||||
- jnz L(first_vec_x2_check)
|
||||
- subl $VEC_SIZE, %edx
|
||||
- jle L(zero)
|
||||
+ /* Check VEC2 and compare any match with remaining length. */
|
||||
+ VPCMP $0, (VEC_SIZE * 2)(%rdi), %YMMMATCH, %k0
|
||||
+ kmovd %k0, %eax
|
||||
+ tzcntl %eax, %eax
|
||||
+ cmpl %eax, %edx
|
||||
+ jbe L(set_zero_end)
|
||||
+ leaq (VEC_SIZE * 2)(%rdi, %rax, CHAR_SIZE), %rax
|
||||
+L(zero_end):
|
||||
+ ret
|
||||
|
||||
- VPCMP $0, (VEC_SIZE * 3)(%rdi), %YMMMATCH, %k1
|
||||
- kmovd %k1, %eax
|
||||
- testl %eax, %eax
|
||||
|
||||
- jnz L(first_vec_x3_check)
|
||||
+ .p2align 4
|
||||
+L(first_vec_x1_check):
|
||||
+ tzcntl %eax, %eax
|
||||
+ /* Adjust length. */
|
||||
+ subl $-(CHAR_PER_VEC * 4), %edx
|
||||
+ /* Check if match within remaining length. */
|
||||
+ cmpl %eax, %edx
|
||||
+ jbe L(set_zero_end)
|
||||
+ /* NB: Multiply bytes by CHAR_SIZE to get the wchar_t count. */
|
||||
+ leaq VEC_SIZE(%rdi, %rax, CHAR_SIZE), %rax
|
||||
+ ret
|
||||
+L(set_zero_end):
|
||||
xorl %eax, %eax
|
||||
ret
|
||||
|
||||
.p2align 4
|
||||
-L(last_2x_vec):
|
||||
- addl $(VEC_SIZE * 2), %edx
|
||||
- VPCMP $0, (%rdi), %YMMMATCH, %k1
|
||||
+L(loop_4x_vec_end):
|
||||
+# endif
|
||||
+ /* rawmemchr will fall through into this if match was found in
|
||||
+ loop. */
|
||||
+
|
||||
+ /* k1 has not of matches with VEC1. */
|
||||
kmovd %k1, %eax
|
||||
- testl %eax, %eax
|
||||
+# ifdef USE_AS_WMEMCHR
|
||||
+ subl $((1 << CHAR_PER_VEC) - 1), %eax
|
||||
+# else
|
||||
+ incl %eax
|
||||
+# endif
|
||||
+ jnz L(last_vec_x1_return)
|
||||
|
||||
- jnz L(first_vec_x0_check)
|
||||
- subl $VEC_SIZE, %edx
|
||||
- jle L(zero)
|
||||
+ VPCMP $0, %YMM2, %YMMZERO, %k0
|
||||
+ kmovd %k0, %eax
|
||||
+ testl %eax, %eax
|
||||
+ jnz L(last_vec_x2_return)
|
||||
|
||||
- VPCMP $0, VEC_SIZE(%rdi), %YMMMATCH, %k1
|
||||
- kmovd %k1, %eax
|
||||
+ kmovd %k2, %eax
|
||||
testl %eax, %eax
|
||||
- jnz L(first_vec_x1_check)
|
||||
- xorl %eax, %eax
|
||||
- ret
|
||||
+ jnz L(last_vec_x3_return)
|
||||
|
||||
- .p2align 4
|
||||
-L(first_vec_x0_check):
|
||||
+ kmovd %k3, %eax
|
||||
tzcntl %eax, %eax
|
||||
-# ifdef USE_AS_WMEMCHR
|
||||
- /* NB: Multiply wchar_t count by 4 to get the number of bytes. */
|
||||
- sall $2, %eax
|
||||
+# ifdef USE_AS_RAWMEMCHR
|
||||
+ leaq (VEC_SIZE * 3)(%rdi, %rax, CHAR_SIZE), %rax
|
||||
+# else
|
||||
+ leaq (VEC_SIZE * 7)(%rdi, %rax, CHAR_SIZE), %rax
|
||||
# endif
|
||||
- /* Check the end of data. */
|
||||
- cmpq %rax, %rdx
|
||||
- jbe L(zero)
|
||||
- addq %rdi, %rax
|
||||
ret
|
||||
|
||||
.p2align 4
|
||||
-L(first_vec_x1_check):
|
||||
+L(last_vec_x1_return):
|
||||
tzcntl %eax, %eax
|
||||
-# ifdef USE_AS_WMEMCHR
|
||||
- /* NB: Multiply wchar_t count by 4 to get the number of bytes. */
|
||||
- sall $2, %eax
|
||||
-# endif
|
||||
- /* Check the end of data. */
|
||||
- cmpq %rax, %rdx
|
||||
- jbe L(zero)
|
||||
- addq $VEC_SIZE, %rax
|
||||
+# ifdef USE_AS_RAWMEMCHR
|
||||
+# ifdef USE_AS_WMEMCHR
|
||||
+ /* NB: Multiply bytes by CHAR_SIZE to get the wchar_t count. */
|
||||
+ leaq (%rdi, %rax, CHAR_SIZE), %rax
|
||||
+# else
|
||||
addq %rdi, %rax
|
||||
- ret
|
||||
-
|
||||
- .p2align 4
|
||||
-L(first_vec_x2_check):
|
||||
- tzcntl %eax, %eax
|
||||
-# ifdef USE_AS_WMEMCHR
|
||||
- /* NB: Multiply wchar_t count by 4 to get the number of bytes. */
|
||||
- sall $2, %eax
|
||||
+# endif
|
||||
+# else
|
||||
+ /* NB: Multiply bytes by CHAR_SIZE to get the wchar_t count. */
|
||||
+ leaq (VEC_SIZE * 4)(%rdi, %rax, CHAR_SIZE), %rax
|
||||
# endif
|
||||
- /* Check the end of data. */
|
||||
- cmpq %rax, %rdx
|
||||
- jbe L(zero)
|
||||
- addq $(VEC_SIZE * 2), %rax
|
||||
- addq %rdi, %rax
|
||||
ret
|
||||
|
||||
.p2align 4
|
||||
-L(first_vec_x3_check):
|
||||
+L(last_vec_x2_return):
|
||||
tzcntl %eax, %eax
|
||||
-# ifdef USE_AS_WMEMCHR
|
||||
- /* NB: Multiply wchar_t count by 4 to get the number of bytes. */
|
||||
- sall $2, %eax
|
||||
+# ifdef USE_AS_RAWMEMCHR
|
||||
+ /* NB: Multiply bytes by CHAR_SIZE to get the wchar_t count. */
|
||||
+ leaq VEC_SIZE(%rdi, %rax, CHAR_SIZE), %rax
|
||||
+# else
|
||||
+ /* NB: Multiply bytes by CHAR_SIZE to get the wchar_t count. */
|
||||
+ leaq (VEC_SIZE * 5)(%rdi, %rax, CHAR_SIZE), %rax
|
||||
# endif
|
||||
- /* Check the end of data. */
|
||||
- cmpq %rax, %rdx
|
||||
- jbe L(zero)
|
||||
- addq $(VEC_SIZE * 3), %rax
|
||||
- addq %rdi, %rax
|
||||
ret
|
||||
|
||||
.p2align 4
|
||||
-L(zero):
|
||||
- xorl %eax, %eax
|
||||
- ret
|
||||
-# endif
|
||||
-
|
||||
- .p2align 4
|
||||
-L(first_vec_x0):
|
||||
+L(last_vec_x3_return):
|
||||
tzcntl %eax, %eax
|
||||
-# ifdef USE_AS_WMEMCHR
|
||||
- /* NB: Multiply wchar_t count by 4 to get the number of bytes. */
|
||||
- leaq (%rdi, %rax, 4), %rax
|
||||
+# ifdef USE_AS_RAWMEMCHR
|
||||
+ /* NB: Multiply bytes by CHAR_SIZE to get the wchar_t count. */
|
||||
+ leaq (VEC_SIZE * 2)(%rdi, %rax, CHAR_SIZE), %rax
|
||||
# else
|
||||
- addq %rdi, %rax
|
||||
+ /* NB: Multiply bytes by CHAR_SIZE to get the wchar_t count. */
|
||||
+ leaq (VEC_SIZE * 6)(%rdi, %rax, CHAR_SIZE), %rax
|
||||
# endif
|
||||
ret
|
||||
|
||||
+
|
||||
+# ifndef USE_AS_RAWMEMCHR
|
||||
+L(last_4x_vec_or_less_cmpeq):
|
||||
+ VPCMP $0, (VEC_SIZE * 5)(%rdi), %YMMMATCH, %k0
|
||||
+ kmovd %k0, %eax
|
||||
+ subq $-(VEC_SIZE * 4), %rdi
|
||||
+ /* Check first VEC regardless. */
|
||||
+ testl %eax, %eax
|
||||
+ jnz L(first_vec_x1_check)
|
||||
+
|
||||
+ /* If remaining length <= CHAR_PER_VEC * 2. */
|
||||
+ addl $(CHAR_PER_VEC * 2), %edx
|
||||
+ jle L(last_2x_vec)
|
||||
+
|
||||
.p2align 4
|
||||
-L(first_vec_x1):
|
||||
+L(last_4x_vec):
|
||||
+ VPCMP $0, (VEC_SIZE * 2)(%rdi), %YMMMATCH, %k0
|
||||
+ kmovd %k0, %eax
|
||||
+ testl %eax, %eax
|
||||
+ jnz L(last_vec_x2)
|
||||
+
|
||||
+
|
||||
+ VPCMP $0, (VEC_SIZE * 3)(%rdi), %YMMMATCH, %k0
|
||||
+ kmovd %k0, %eax
|
||||
+ /* Create mask for possible matches within remaining length. */
|
||||
+# ifdef USE_AS_WMEMCHR
|
||||
+ movl $((1 << (CHAR_PER_VEC * 2)) - 1), %ecx
|
||||
+ bzhil %edx, %ecx, %ecx
|
||||
+# else
|
||||
+ movq $-1, %rcx
|
||||
+ bzhiq %rdx, %rcx, %rcx
|
||||
+# endif
|
||||
+ /* Test matches in data against length match. */
|
||||
+ andl %ecx, %eax
|
||||
+ jnz L(last_vec_x3)
|
||||
+
|
||||
+ /* if remaining length <= CHAR_PER_VEC * 3 (Note this is after
|
||||
+ remaining length was found to be > CHAR_PER_VEC * 2. */
|
||||
+ subl $CHAR_PER_VEC, %edx
|
||||
+ jbe L(zero_end2)
|
||||
+
|
||||
+
|
||||
+ VPCMP $0, (VEC_SIZE * 4)(%rdi), %YMMMATCH, %k0
|
||||
+ kmovd %k0, %eax
|
||||
+ /* Shift remaining length mask for last VEC. */
|
||||
+# ifdef USE_AS_WMEMCHR
|
||||
+ shrl $CHAR_PER_VEC, %ecx
|
||||
+# else
|
||||
+ shrq $CHAR_PER_VEC, %rcx
|
||||
+# endif
|
||||
+ andl %ecx, %eax
|
||||
+ jz L(zero_end2)
|
||||
tzcntl %eax, %eax
|
||||
-# ifdef USE_AS_WMEMCHR
|
||||
- /* NB: Multiply wchar_t count by 4 to get the number of bytes. */
|
||||
- leaq VEC_SIZE(%rdi, %rax, 4), %rax
|
||||
-# else
|
||||
- addq $VEC_SIZE, %rax
|
||||
- addq %rdi, %rax
|
||||
-# endif
|
||||
+ leaq (VEC_SIZE * 4)(%rdi, %rax, CHAR_SIZE), %rax
|
||||
+L(zero_end2):
|
||||
ret
|
||||
|
||||
- .p2align 4
|
||||
-L(first_vec_x2):
|
||||
+L(last_vec_x2):
|
||||
tzcntl %eax, %eax
|
||||
-# ifdef USE_AS_WMEMCHR
|
||||
- /* NB: Multiply wchar_t count by 4 to get the number of bytes. */
|
||||
- leaq (VEC_SIZE * 2)(%rdi, %rax, 4), %rax
|
||||
-# else
|
||||
- addq $(VEC_SIZE * 2), %rax
|
||||
- addq %rdi, %rax
|
||||
-# endif
|
||||
+ leaq (VEC_SIZE * 2)(%rdi, %rax, CHAR_SIZE), %rax
|
||||
ret
|
||||
|
||||
.p2align 4
|
||||
-L(4x_vec_end):
|
||||
- kmovd %k1, %eax
|
||||
- testl %eax, %eax
|
||||
- jnz L(first_vec_x0)
|
||||
- kmovd %k2, %eax
|
||||
- testl %eax, %eax
|
||||
- jnz L(first_vec_x1)
|
||||
- kmovd %k3, %eax
|
||||
- testl %eax, %eax
|
||||
- jnz L(first_vec_x2)
|
||||
- kmovd %k4, %eax
|
||||
- testl %eax, %eax
|
||||
-L(first_vec_x3):
|
||||
+L(last_vec_x3):
|
||||
tzcntl %eax, %eax
|
||||
-# ifdef USE_AS_WMEMCHR
|
||||
- /* NB: Multiply wchar_t count by 4 to get the number of bytes. */
|
||||
- leaq (VEC_SIZE * 3)(%rdi, %rax, 4), %rax
|
||||
-# else
|
||||
- addq $(VEC_SIZE * 3), %rax
|
||||
- addq %rdi, %rax
|
||||
-# endif
|
||||
+ leaq (VEC_SIZE * 3)(%rdi, %rax, CHAR_SIZE), %rax
|
||||
ret
|
||||
+# endif
|
||||
|
||||
END (MEMCHR)
|
||||
#endif
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,30 +0,0 @@
|
||||
From 6ea916adfa0ab9af6e7dc6adcf6f977dfe017835 Mon Sep 17 00:00:00 2001
|
||||
From: Alice Xu <alice.d.xu@gmail.com>
|
||||
Date: Fri, 7 May 2021 19:03:21 -0700
|
||||
Subject: [PATCH] x86-64: Fix an unknown vector operation in memchr-evex.S
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
An unknown vector operation occurred in commit 2a76821c308. Fixed it
|
||||
by using "ymm{k1}{z}" but not "ymm {k1} {z}".
|
||||
|
||||
Reviewed-by: H.J. Lu <hjl.tools@gmail.com>
|
||||
---
|
||||
sysdeps/x86_64/multiarch/memchr-evex.S | 2 +-
|
||||
1 file changed, 1 insertion(+), 1 deletion(-)
|
||||
|
||||
diff --git a/sysdeps/x86_64/multiarch/memchr-evex.S b/sysdeps/x86_64/multiarch/memchr-evex.S
|
||||
index 81d5cd64..f3fdad4f 100644
|
||||
--- a/sysdeps/x86_64/multiarch/memchr-evex.S
|
||||
+++ b/sysdeps/x86_64/multiarch/memchr-evex.S
|
||||
@@ -271,7 +271,7 @@ L(loop_4x_vec):
|
||||
vpxorq (VEC_SIZE * 6)(%rdi), %YMMMATCH, %YMM3
|
||||
VPCMP $0, (VEC_SIZE * 7)(%rdi), %YMMMATCH, %k3
|
||||
/* Reduce VEC2 / VEC3 with min and VEC1 with zero mask. */
|
||||
- VPMINU %YMM2, %YMM3, %YMM3 {%k1} {z}
|
||||
+ VPMINU %YMM2, %YMM3, %YMM3{%k1}{z}
|
||||
VPCMP $0, %YMM3, %YMMZERO, %k2
|
||||
# ifdef USE_AS_RAWMEMCHR
|
||||
subq $-(VEC_SIZE * 4), %rdi
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,566 +0,0 @@
|
||||
From a0db678071c60b6c47c468d231dd0b3694ba7a98 Mon Sep 17 00:00:00 2001
|
||||
From: "H.J. Lu" <hjl.tools@gmail.com>
|
||||
Date: Tue, 22 Jun 2021 20:42:10 -0700
|
||||
Subject: [PATCH] x86-64: Move strlen.S to multiarch/strlen-vec.S
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
Since strlen.S contains SSE2 version of strlen/strnlen and SSE4.1
|
||||
version of wcslen/wcsnlen, move strlen.S to multiarch/strlen-vec.S
|
||||
and include multiarch/strlen-vec.S from SSE2 and SSE4.1 variants.
|
||||
This also removes the unused symbols, __GI___strlen_sse2 and
|
||||
__GI___wcsnlen_sse4_1.
|
||||
---
|
||||
sysdeps/x86_64/multiarch/strlen-sse2.S | 2 +-
|
||||
sysdeps/x86_64/multiarch/strlen-vec.S | 257 ++++++++++++++++++++++
|
||||
sysdeps/x86_64/multiarch/wcsnlen-sse4_1.S | 2 +-
|
||||
sysdeps/x86_64/strlen.S | 243 +-------------------
|
||||
4 files changed, 262 insertions(+), 242 deletions(-)
|
||||
create mode 100644 sysdeps/x86_64/multiarch/strlen-vec.S
|
||||
|
||||
Conflicts:
|
||||
sysdeps/x86_64/strlen.S b/sysdeps/x86_64/strlen.S
|
||||
(Copyright dates, URL)
|
||||
|
||||
diff --git a/sysdeps/x86_64/multiarch/strlen-sse2.S b/sysdeps/x86_64/multiarch/strlen-sse2.S
|
||||
index 7bc57b8d..449c8a7f 100644
|
||||
--- a/sysdeps/x86_64/multiarch/strlen-sse2.S
|
||||
+++ b/sysdeps/x86_64/multiarch/strlen-sse2.S
|
||||
@@ -20,4 +20,4 @@
|
||||
# define strlen __strlen_sse2
|
||||
#endif
|
||||
|
||||
-#include "../strlen.S"
|
||||
+#include "strlen-vec.S"
|
||||
diff --git a/sysdeps/x86_64/multiarch/strlen-vec.S b/sysdeps/x86_64/multiarch/strlen-vec.S
|
||||
new file mode 100644
|
||||
index 00000000..8f660bb9
|
||||
--- /dev/null
|
||||
+++ b/sysdeps/x86_64/multiarch/strlen-vec.S
|
||||
@@ -0,0 +1,257 @@
|
||||
+/* SSE2 version of strlen and SSE4.1 version of wcslen.
|
||||
+ Copyright (C) 2012-2021 Free Software Foundation, Inc.
|
||||
+ This file is part of the GNU C Library.
|
||||
+
|
||||
+ The GNU C Library is free software; you can redistribute it and/or
|
||||
+ modify it under the terms of the GNU Lesser General Public
|
||||
+ License as published by the Free Software Foundation; either
|
||||
+ version 2.1 of the License, or (at your option) any later version.
|
||||
+
|
||||
+ The GNU C Library is distributed in the hope that it will be useful,
|
||||
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
||||
+ Lesser General Public License for more details.
|
||||
+
|
||||
+ You should have received a copy of the GNU Lesser General Public
|
||||
+ License along with the GNU C Library; if not, see
|
||||
+ <https://www.gnu.org/licenses/>. */
|
||||
+
|
||||
+#include <sysdep.h>
|
||||
+
|
||||
+#ifdef AS_WCSLEN
|
||||
+# define PMINU pminud
|
||||
+# define PCMPEQ pcmpeqd
|
||||
+# define SHIFT_RETURN shrq $2, %rax
|
||||
+#else
|
||||
+# define PMINU pminub
|
||||
+# define PCMPEQ pcmpeqb
|
||||
+# define SHIFT_RETURN
|
||||
+#endif
|
||||
+
|
||||
+/* Long lived register in strlen(s), strnlen(s, n) are:
|
||||
+
|
||||
+ %xmm3 - zero
|
||||
+ %rdi - s
|
||||
+ %r10 (s+n) & (~(64-1))
|
||||
+ %r11 s+n
|
||||
+*/
|
||||
+
|
||||
+
|
||||
+.text
|
||||
+ENTRY(strlen)
|
||||
+
|
||||
+/* Test 64 bytes from %rax for zero. Save result as bitmask in %rdx. */
|
||||
+#define FIND_ZERO \
|
||||
+ PCMPEQ (%rax), %xmm0; \
|
||||
+ PCMPEQ 16(%rax), %xmm1; \
|
||||
+ PCMPEQ 32(%rax), %xmm2; \
|
||||
+ PCMPEQ 48(%rax), %xmm3; \
|
||||
+ pmovmskb %xmm0, %esi; \
|
||||
+ pmovmskb %xmm1, %edx; \
|
||||
+ pmovmskb %xmm2, %r8d; \
|
||||
+ pmovmskb %xmm3, %ecx; \
|
||||
+ salq $16, %rdx; \
|
||||
+ salq $16, %rcx; \
|
||||
+ orq %rsi, %rdx; \
|
||||
+ orq %r8, %rcx; \
|
||||
+ salq $32, %rcx; \
|
||||
+ orq %rcx, %rdx;
|
||||
+
|
||||
+#ifdef AS_STRNLEN
|
||||
+/* Do not read anything when n==0. */
|
||||
+ test %RSI_LP, %RSI_LP
|
||||
+ jne L(n_nonzero)
|
||||
+ xor %rax, %rax
|
||||
+ ret
|
||||
+L(n_nonzero):
|
||||
+# ifdef AS_WCSLEN
|
||||
+ shl $2, %RSI_LP
|
||||
+# endif
|
||||
+
|
||||
+/* Initialize long lived registers. */
|
||||
+
|
||||
+ add %RDI_LP, %RSI_LP
|
||||
+ mov %RSI_LP, %R10_LP
|
||||
+ and $-64, %R10_LP
|
||||
+ mov %RSI_LP, %R11_LP
|
||||
+#endif
|
||||
+
|
||||
+ pxor %xmm0, %xmm0
|
||||
+ pxor %xmm1, %xmm1
|
||||
+ pxor %xmm2, %xmm2
|
||||
+ pxor %xmm3, %xmm3
|
||||
+ movq %rdi, %rax
|
||||
+ movq %rdi, %rcx
|
||||
+ andq $4095, %rcx
|
||||
+/* Offsets 4032-4047 will be aligned into 4032 thus fit into page. */
|
||||
+ cmpq $4047, %rcx
|
||||
+/* We cannot unify this branching as it would be ~6 cycles slower. */
|
||||
+ ja L(cross_page)
|
||||
+
|
||||
+#ifdef AS_STRNLEN
|
||||
+/* Test if end is among first 64 bytes. */
|
||||
+# define STRNLEN_PROLOG \
|
||||
+ mov %r11, %rsi; \
|
||||
+ subq %rax, %rsi; \
|
||||
+ andq $-64, %rax; \
|
||||
+ testq $-64, %rsi; \
|
||||
+ je L(strnlen_ret)
|
||||
+#else
|
||||
+# define STRNLEN_PROLOG andq $-64, %rax;
|
||||
+#endif
|
||||
+
|
||||
+/* Ignore bits in mask that come before start of string. */
|
||||
+#define PROLOG(lab) \
|
||||
+ movq %rdi, %rcx; \
|
||||
+ xorq %rax, %rcx; \
|
||||
+ STRNLEN_PROLOG; \
|
||||
+ sarq %cl, %rdx; \
|
||||
+ test %rdx, %rdx; \
|
||||
+ je L(lab); \
|
||||
+ bsfq %rdx, %rax; \
|
||||
+ SHIFT_RETURN; \
|
||||
+ ret
|
||||
+
|
||||
+#ifdef AS_STRNLEN
|
||||
+ andq $-16, %rax
|
||||
+ FIND_ZERO
|
||||
+#else
|
||||
+ /* Test first 16 bytes unaligned. */
|
||||
+ movdqu (%rax), %xmm4
|
||||
+ PCMPEQ %xmm0, %xmm4
|
||||
+ pmovmskb %xmm4, %edx
|
||||
+ test %edx, %edx
|
||||
+ je L(next48_bytes)
|
||||
+ bsf %edx, %eax /* If eax is zeroed 16bit bsf can be used. */
|
||||
+ SHIFT_RETURN
|
||||
+ ret
|
||||
+
|
||||
+L(next48_bytes):
|
||||
+/* Same as FIND_ZERO except we do not check first 16 bytes. */
|
||||
+ andq $-16, %rax
|
||||
+ PCMPEQ 16(%rax), %xmm1
|
||||
+ PCMPEQ 32(%rax), %xmm2
|
||||
+ PCMPEQ 48(%rax), %xmm3
|
||||
+ pmovmskb %xmm1, %edx
|
||||
+ pmovmskb %xmm2, %r8d
|
||||
+ pmovmskb %xmm3, %ecx
|
||||
+ salq $16, %rdx
|
||||
+ salq $16, %rcx
|
||||
+ orq %r8, %rcx
|
||||
+ salq $32, %rcx
|
||||
+ orq %rcx, %rdx
|
||||
+#endif
|
||||
+
|
||||
+ /* When no zero byte is found xmm1-3 are zero so we do not have to
|
||||
+ zero them. */
|
||||
+ PROLOG(loop)
|
||||
+
|
||||
+ .p2align 4
|
||||
+L(cross_page):
|
||||
+ andq $-64, %rax
|
||||
+ FIND_ZERO
|
||||
+ PROLOG(loop_init)
|
||||
+
|
||||
+#ifdef AS_STRNLEN
|
||||
+/* We must do this check to correctly handle strnlen (s, -1). */
|
||||
+L(strnlen_ret):
|
||||
+ bts %rsi, %rdx
|
||||
+ sarq %cl, %rdx
|
||||
+ test %rdx, %rdx
|
||||
+ je L(loop_init)
|
||||
+ bsfq %rdx, %rax
|
||||
+ SHIFT_RETURN
|
||||
+ ret
|
||||
+#endif
|
||||
+ .p2align 4
|
||||
+L(loop_init):
|
||||
+ pxor %xmm1, %xmm1
|
||||
+ pxor %xmm2, %xmm2
|
||||
+ pxor %xmm3, %xmm3
|
||||
+#ifdef AS_STRNLEN
|
||||
+ .p2align 4
|
||||
+L(loop):
|
||||
+
|
||||
+ addq $64, %rax
|
||||
+ cmpq %rax, %r10
|
||||
+ je L(exit_end)
|
||||
+
|
||||
+ movdqa (%rax), %xmm0
|
||||
+ PMINU 16(%rax), %xmm0
|
||||
+ PMINU 32(%rax), %xmm0
|
||||
+ PMINU 48(%rax), %xmm0
|
||||
+ PCMPEQ %xmm3, %xmm0
|
||||
+ pmovmskb %xmm0, %edx
|
||||
+ testl %edx, %edx
|
||||
+ jne L(exit)
|
||||
+ jmp L(loop)
|
||||
+
|
||||
+ .p2align 4
|
||||
+L(exit_end):
|
||||
+ cmp %rax, %r11
|
||||
+ je L(first) /* Do not read when end is at page boundary. */
|
||||
+ pxor %xmm0, %xmm0
|
||||
+ FIND_ZERO
|
||||
+
|
||||
+L(first):
|
||||
+ bts %r11, %rdx
|
||||
+ bsfq %rdx, %rdx
|
||||
+ addq %rdx, %rax
|
||||
+ subq %rdi, %rax
|
||||
+ SHIFT_RETURN
|
||||
+ ret
|
||||
+
|
||||
+ .p2align 4
|
||||
+L(exit):
|
||||
+ pxor %xmm0, %xmm0
|
||||
+ FIND_ZERO
|
||||
+
|
||||
+ bsfq %rdx, %rdx
|
||||
+ addq %rdx, %rax
|
||||
+ subq %rdi, %rax
|
||||
+ SHIFT_RETURN
|
||||
+ ret
|
||||
+
|
||||
+#else
|
||||
+
|
||||
+ /* Main loop. Unrolled twice to improve L2 cache performance on core2. */
|
||||
+ .p2align 4
|
||||
+L(loop):
|
||||
+
|
||||
+ movdqa 64(%rax), %xmm0
|
||||
+ PMINU 80(%rax), %xmm0
|
||||
+ PMINU 96(%rax), %xmm0
|
||||
+ PMINU 112(%rax), %xmm0
|
||||
+ PCMPEQ %xmm3, %xmm0
|
||||
+ pmovmskb %xmm0, %edx
|
||||
+ testl %edx, %edx
|
||||
+ jne L(exit64)
|
||||
+
|
||||
+ subq $-128, %rax
|
||||
+
|
||||
+ movdqa (%rax), %xmm0
|
||||
+ PMINU 16(%rax), %xmm0
|
||||
+ PMINU 32(%rax), %xmm0
|
||||
+ PMINU 48(%rax), %xmm0
|
||||
+ PCMPEQ %xmm3, %xmm0
|
||||
+ pmovmskb %xmm0, %edx
|
||||
+ testl %edx, %edx
|
||||
+ jne L(exit0)
|
||||
+ jmp L(loop)
|
||||
+
|
||||
+ .p2align 4
|
||||
+L(exit64):
|
||||
+ addq $64, %rax
|
||||
+L(exit0):
|
||||
+ pxor %xmm0, %xmm0
|
||||
+ FIND_ZERO
|
||||
+
|
||||
+ bsfq %rdx, %rdx
|
||||
+ addq %rdx, %rax
|
||||
+ subq %rdi, %rax
|
||||
+ SHIFT_RETURN
|
||||
+ ret
|
||||
+
|
||||
+#endif
|
||||
+
|
||||
+END(strlen)
|
||||
diff --git a/sysdeps/x86_64/multiarch/wcsnlen-sse4_1.S b/sysdeps/x86_64/multiarch/wcsnlen-sse4_1.S
|
||||
index a8cab0cb..5fa51fe0 100644
|
||||
--- a/sysdeps/x86_64/multiarch/wcsnlen-sse4_1.S
|
||||
+++ b/sysdeps/x86_64/multiarch/wcsnlen-sse4_1.S
|
||||
@@ -2,4 +2,4 @@
|
||||
#define AS_STRNLEN
|
||||
#define strlen __wcsnlen_sse4_1
|
||||
|
||||
-#include "../strlen.S"
|
||||
+#include "strlen-vec.S"
|
||||
diff --git a/sysdeps/x86_64/strlen.S b/sysdeps/x86_64/strlen.S
|
||||
index f845f3d4..ad047d84 100644
|
||||
--- a/sysdeps/x86_64/strlen.S
|
||||
+++ b/sysdeps/x86_64/strlen.S
|
||||
@@ -1,5 +1,5 @@
|
||||
-/* SSE2 version of strlen/wcslen.
|
||||
- Copyright (C) 2012-2018 Free Software Foundation, Inc.
|
||||
+/* SSE2 version of strlen.
|
||||
+ Copyright (C) 2021 Free Software Foundation, Inc.
|
||||
This file is part of the GNU C Library.
|
||||
|
||||
The GNU C Library is free software; you can redistribute it and/or
|
||||
@@ -16,243 +16,6 @@
|
||||
License along with the GNU C Library; if not, see
|
||||
<http://www.gnu.org/licenses/>. */
|
||||
|
||||
-#include <sysdep.h>
|
||||
+#include "multiarch/strlen-vec.S"
|
||||
|
||||
-#ifdef AS_WCSLEN
|
||||
-# define PMINU pminud
|
||||
-# define PCMPEQ pcmpeqd
|
||||
-# define SHIFT_RETURN shrq $2, %rax
|
||||
-#else
|
||||
-# define PMINU pminub
|
||||
-# define PCMPEQ pcmpeqb
|
||||
-# define SHIFT_RETURN
|
||||
-#endif
|
||||
-
|
||||
-/* Long lived register in strlen(s), strnlen(s, n) are:
|
||||
-
|
||||
- %xmm3 - zero
|
||||
- %rdi - s
|
||||
- %r10 (s+n) & (~(64-1))
|
||||
- %r11 s+n
|
||||
-*/
|
||||
-
|
||||
-
|
||||
-.text
|
||||
-ENTRY(strlen)
|
||||
-
|
||||
-/* Test 64 bytes from %rax for zero. Save result as bitmask in %rdx. */
|
||||
-#define FIND_ZERO \
|
||||
- PCMPEQ (%rax), %xmm0; \
|
||||
- PCMPEQ 16(%rax), %xmm1; \
|
||||
- PCMPEQ 32(%rax), %xmm2; \
|
||||
- PCMPEQ 48(%rax), %xmm3; \
|
||||
- pmovmskb %xmm0, %esi; \
|
||||
- pmovmskb %xmm1, %edx; \
|
||||
- pmovmskb %xmm2, %r8d; \
|
||||
- pmovmskb %xmm3, %ecx; \
|
||||
- salq $16, %rdx; \
|
||||
- salq $16, %rcx; \
|
||||
- orq %rsi, %rdx; \
|
||||
- orq %r8, %rcx; \
|
||||
- salq $32, %rcx; \
|
||||
- orq %rcx, %rdx;
|
||||
-
|
||||
-#ifdef AS_STRNLEN
|
||||
-/* Do not read anything when n==0. */
|
||||
- test %RSI_LP, %RSI_LP
|
||||
- jne L(n_nonzero)
|
||||
- xor %rax, %rax
|
||||
- ret
|
||||
-L(n_nonzero):
|
||||
-# ifdef AS_WCSLEN
|
||||
- shl $2, %RSI_LP
|
||||
-# endif
|
||||
-
|
||||
-/* Initialize long lived registers. */
|
||||
-
|
||||
- add %RDI_LP, %RSI_LP
|
||||
- mov %RSI_LP, %R10_LP
|
||||
- and $-64, %R10_LP
|
||||
- mov %RSI_LP, %R11_LP
|
||||
-#endif
|
||||
-
|
||||
- pxor %xmm0, %xmm0
|
||||
- pxor %xmm1, %xmm1
|
||||
- pxor %xmm2, %xmm2
|
||||
- pxor %xmm3, %xmm3
|
||||
- movq %rdi, %rax
|
||||
- movq %rdi, %rcx
|
||||
- andq $4095, %rcx
|
||||
-/* Offsets 4032-4047 will be aligned into 4032 thus fit into page. */
|
||||
- cmpq $4047, %rcx
|
||||
-/* We cannot unify this branching as it would be ~6 cycles slower. */
|
||||
- ja L(cross_page)
|
||||
-
|
||||
-#ifdef AS_STRNLEN
|
||||
-/* Test if end is among first 64 bytes. */
|
||||
-# define STRNLEN_PROLOG \
|
||||
- mov %r11, %rsi; \
|
||||
- subq %rax, %rsi; \
|
||||
- andq $-64, %rax; \
|
||||
- testq $-64, %rsi; \
|
||||
- je L(strnlen_ret)
|
||||
-#else
|
||||
-# define STRNLEN_PROLOG andq $-64, %rax;
|
||||
-#endif
|
||||
-
|
||||
-/* Ignore bits in mask that come before start of string. */
|
||||
-#define PROLOG(lab) \
|
||||
- movq %rdi, %rcx; \
|
||||
- xorq %rax, %rcx; \
|
||||
- STRNLEN_PROLOG; \
|
||||
- sarq %cl, %rdx; \
|
||||
- test %rdx, %rdx; \
|
||||
- je L(lab); \
|
||||
- bsfq %rdx, %rax; \
|
||||
- SHIFT_RETURN; \
|
||||
- ret
|
||||
-
|
||||
-#ifdef AS_STRNLEN
|
||||
- andq $-16, %rax
|
||||
- FIND_ZERO
|
||||
-#else
|
||||
- /* Test first 16 bytes unaligned. */
|
||||
- movdqu (%rax), %xmm4
|
||||
- PCMPEQ %xmm0, %xmm4
|
||||
- pmovmskb %xmm4, %edx
|
||||
- test %edx, %edx
|
||||
- je L(next48_bytes)
|
||||
- bsf %edx, %eax /* If eax is zeroed 16bit bsf can be used. */
|
||||
- SHIFT_RETURN
|
||||
- ret
|
||||
-
|
||||
-L(next48_bytes):
|
||||
-/* Same as FIND_ZERO except we do not check first 16 bytes. */
|
||||
- andq $-16, %rax
|
||||
- PCMPEQ 16(%rax), %xmm1
|
||||
- PCMPEQ 32(%rax), %xmm2
|
||||
- PCMPEQ 48(%rax), %xmm3
|
||||
- pmovmskb %xmm1, %edx
|
||||
- pmovmskb %xmm2, %r8d
|
||||
- pmovmskb %xmm3, %ecx
|
||||
- salq $16, %rdx
|
||||
- salq $16, %rcx
|
||||
- orq %r8, %rcx
|
||||
- salq $32, %rcx
|
||||
- orq %rcx, %rdx
|
||||
-#endif
|
||||
-
|
||||
- /* When no zero byte is found xmm1-3 are zero so we do not have to
|
||||
- zero them. */
|
||||
- PROLOG(loop)
|
||||
-
|
||||
- .p2align 4
|
||||
-L(cross_page):
|
||||
- andq $-64, %rax
|
||||
- FIND_ZERO
|
||||
- PROLOG(loop_init)
|
||||
-
|
||||
-#ifdef AS_STRNLEN
|
||||
-/* We must do this check to correctly handle strnlen (s, -1). */
|
||||
-L(strnlen_ret):
|
||||
- bts %rsi, %rdx
|
||||
- sarq %cl, %rdx
|
||||
- test %rdx, %rdx
|
||||
- je L(loop_init)
|
||||
- bsfq %rdx, %rax
|
||||
- SHIFT_RETURN
|
||||
- ret
|
||||
-#endif
|
||||
- .p2align 4
|
||||
-L(loop_init):
|
||||
- pxor %xmm1, %xmm1
|
||||
- pxor %xmm2, %xmm2
|
||||
- pxor %xmm3, %xmm3
|
||||
-#ifdef AS_STRNLEN
|
||||
- .p2align 4
|
||||
-L(loop):
|
||||
-
|
||||
- addq $64, %rax
|
||||
- cmpq %rax, %r10
|
||||
- je L(exit_end)
|
||||
-
|
||||
- movdqa (%rax), %xmm0
|
||||
- PMINU 16(%rax), %xmm0
|
||||
- PMINU 32(%rax), %xmm0
|
||||
- PMINU 48(%rax), %xmm0
|
||||
- PCMPEQ %xmm3, %xmm0
|
||||
- pmovmskb %xmm0, %edx
|
||||
- testl %edx, %edx
|
||||
- jne L(exit)
|
||||
- jmp L(loop)
|
||||
-
|
||||
- .p2align 4
|
||||
-L(exit_end):
|
||||
- cmp %rax, %r11
|
||||
- je L(first) /* Do not read when end is at page boundary. */
|
||||
- pxor %xmm0, %xmm0
|
||||
- FIND_ZERO
|
||||
-
|
||||
-L(first):
|
||||
- bts %r11, %rdx
|
||||
- bsfq %rdx, %rdx
|
||||
- addq %rdx, %rax
|
||||
- subq %rdi, %rax
|
||||
- SHIFT_RETURN
|
||||
- ret
|
||||
-
|
||||
- .p2align 4
|
||||
-L(exit):
|
||||
- pxor %xmm0, %xmm0
|
||||
- FIND_ZERO
|
||||
-
|
||||
- bsfq %rdx, %rdx
|
||||
- addq %rdx, %rax
|
||||
- subq %rdi, %rax
|
||||
- SHIFT_RETURN
|
||||
- ret
|
||||
-
|
||||
-#else
|
||||
-
|
||||
- /* Main loop. Unrolled twice to improve L2 cache performance on core2. */
|
||||
- .p2align 4
|
||||
-L(loop):
|
||||
-
|
||||
- movdqa 64(%rax), %xmm0
|
||||
- PMINU 80(%rax), %xmm0
|
||||
- PMINU 96(%rax), %xmm0
|
||||
- PMINU 112(%rax), %xmm0
|
||||
- PCMPEQ %xmm3, %xmm0
|
||||
- pmovmskb %xmm0, %edx
|
||||
- testl %edx, %edx
|
||||
- jne L(exit64)
|
||||
-
|
||||
- subq $-128, %rax
|
||||
-
|
||||
- movdqa (%rax), %xmm0
|
||||
- PMINU 16(%rax), %xmm0
|
||||
- PMINU 32(%rax), %xmm0
|
||||
- PMINU 48(%rax), %xmm0
|
||||
- PCMPEQ %xmm3, %xmm0
|
||||
- pmovmskb %xmm0, %edx
|
||||
- testl %edx, %edx
|
||||
- jne L(exit0)
|
||||
- jmp L(loop)
|
||||
-
|
||||
- .p2align 4
|
||||
-L(exit64):
|
||||
- addq $64, %rax
|
||||
-L(exit0):
|
||||
- pxor %xmm0, %xmm0
|
||||
- FIND_ZERO
|
||||
-
|
||||
- bsfq %rdx, %rdx
|
||||
- addq %rdx, %rax
|
||||
- subq %rdi, %rax
|
||||
- SHIFT_RETURN
|
||||
- ret
|
||||
-
|
||||
-#endif
|
||||
-
|
||||
-END(strlen)
|
||||
libc_hidden_builtin_def (strlen)
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,181 +0,0 @@
|
||||
From 6f573a27b6c8b4236445810a44660612323f5a73 Mon Sep 17 00:00:00 2001
|
||||
From: Noah Goldstein <goldstein.w.n@gmail.com>
|
||||
Date: Wed, 23 Jun 2021 01:19:34 -0400
|
||||
Subject: [PATCH] x86-64: Add wcslen optimize for sse4.1
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
No bug. This comment adds the ifunc / build infrastructure
|
||||
necessary for wcslen to prefer the sse4.1 implementation
|
||||
in strlen-vec.S. test-wcslen.c is passing.
|
||||
|
||||
Signed-off-by: Noah Goldstein <goldstein.w.n@gmail.com>
|
||||
Reviewed-by: H.J. Lu <hjl.tools@gmail.com>
|
||||
---
|
||||
sysdeps/x86_64/multiarch/Makefile | 4 +-
|
||||
sysdeps/x86_64/multiarch/ifunc-impl-list.c | 3 ++
|
||||
sysdeps/x86_64/multiarch/ifunc-wcslen.h | 52 ++++++++++++++++++++++
|
||||
sysdeps/x86_64/multiarch/wcslen-sse4_1.S | 4 ++
|
||||
sysdeps/x86_64/multiarch/wcslen.c | 2 +-
|
||||
sysdeps/x86_64/multiarch/wcsnlen.c | 34 +-------------
|
||||
6 files changed, 63 insertions(+), 36 deletions(-)
|
||||
create mode 100644 sysdeps/x86_64/multiarch/ifunc-wcslen.h
|
||||
create mode 100644 sysdeps/x86_64/multiarch/wcslen-sse4_1.S
|
||||
|
||||
diff --git a/sysdeps/x86_64/multiarch/Makefile b/sysdeps/x86_64/multiarch/Makefile
|
||||
index 491c7698..65fde4eb 100644
|
||||
--- a/sysdeps/x86_64/multiarch/Makefile
|
||||
+++ b/sysdeps/x86_64/multiarch/Makefile
|
||||
@@ -93,8 +93,8 @@ sysdep_routines += wmemcmp-sse4 wmemcmp-ssse3 wmemcmp-c \
|
||||
wcscpy-ssse3 wcscpy-c \
|
||||
wcschr-sse2 wcschr-avx2 \
|
||||
wcsrchr-sse2 wcsrchr-avx2 \
|
||||
- wcsnlen-sse4_1 wcsnlen-c \
|
||||
- wcslen-sse2 wcslen-avx2 wcsnlen-avx2 \
|
||||
+ wcslen-sse2 wcslen-sse4_1 wcslen-avx2 \
|
||||
+ wcsnlen-c wcsnlen-sse4_1 wcsnlen-avx2 \
|
||||
wcschr-avx2-rtm \
|
||||
wcscmp-avx2-rtm \
|
||||
wcslen-avx2-rtm \
|
||||
diff --git a/sysdeps/x86_64/multiarch/ifunc-impl-list.c b/sysdeps/x86_64/multiarch/ifunc-impl-list.c
|
||||
index f1a6460a..580913ca 100644
|
||||
--- a/sysdeps/x86_64/multiarch/ifunc-impl-list.c
|
||||
+++ b/sysdeps/x86_64/multiarch/ifunc-impl-list.c
|
||||
@@ -657,6 +657,9 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
|
||||
&& CPU_FEATURE_USABLE (AVX512BW)
|
||||
&& CPU_FEATURE_USABLE (BMI2)),
|
||||
__wcslen_evex)
|
||||
+ IFUNC_IMPL_ADD (array, i, wcsnlen,
|
||||
+ CPU_FEATURE_USABLE (SSE4_1),
|
||||
+ __wcsnlen_sse4_1)
|
||||
IFUNC_IMPL_ADD (array, i, wcslen, 1, __wcslen_sse2))
|
||||
|
||||
/* Support sysdeps/x86_64/multiarch/wcsnlen.c. */
|
||||
diff --git a/sysdeps/x86_64/multiarch/ifunc-wcslen.h b/sysdeps/x86_64/multiarch/ifunc-wcslen.h
|
||||
new file mode 100644
|
||||
index 00000000..39e33473
|
||||
--- /dev/null
|
||||
+++ b/sysdeps/x86_64/multiarch/ifunc-wcslen.h
|
||||
@@ -0,0 +1,52 @@
|
||||
+/* Common definition for ifunc selections for wcslen and wcsnlen
|
||||
+ All versions must be listed in ifunc-impl-list.c.
|
||||
+ Copyright (C) 2017-2021 Free Software Foundation, Inc.
|
||||
+ This file is part of the GNU C Library.
|
||||
+
|
||||
+ The GNU C Library is free software; you can redistribute it and/or
|
||||
+ modify it under the terms of the GNU Lesser General Public
|
||||
+ License as published by the Free Software Foundation; either
|
||||
+ version 2.1 of the License, or (at your option) any later version.
|
||||
+
|
||||
+ The GNU C Library is distributed in the hope that it will be useful,
|
||||
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
||||
+ Lesser General Public License for more details.
|
||||
+
|
||||
+ You should have received a copy of the GNU Lesser General Public
|
||||
+ License along with the GNU C Library; if not, see
|
||||
+ <https://www.gnu.org/licenses/>. */
|
||||
+
|
||||
+#include <init-arch.h>
|
||||
+
|
||||
+extern __typeof (REDIRECT_NAME) OPTIMIZE (sse2) attribute_hidden;
|
||||
+extern __typeof (REDIRECT_NAME) OPTIMIZE (sse4_1) attribute_hidden;
|
||||
+extern __typeof (REDIRECT_NAME) OPTIMIZE (avx2) attribute_hidden;
|
||||
+extern __typeof (REDIRECT_NAME) OPTIMIZE (avx2_rtm) attribute_hidden;
|
||||
+extern __typeof (REDIRECT_NAME) OPTIMIZE (evex) attribute_hidden;
|
||||
+
|
||||
+static inline void *
|
||||
+IFUNC_SELECTOR (void)
|
||||
+{
|
||||
+ const struct cpu_features* cpu_features = __get_cpu_features ();
|
||||
+
|
||||
+ if (CPU_FEATURE_USABLE_P (cpu_features, AVX2)
|
||||
+ && CPU_FEATURE_USABLE_P (cpu_features, BMI2)
|
||||
+ && CPU_FEATURES_ARCH_P (cpu_features, AVX_Fast_Unaligned_Load))
|
||||
+ {
|
||||
+ if (CPU_FEATURE_USABLE_P (cpu_features, AVX512VL)
|
||||
+ && CPU_FEATURE_USABLE_P (cpu_features, AVX512BW))
|
||||
+ return OPTIMIZE (evex);
|
||||
+
|
||||
+ if (CPU_FEATURE_USABLE_P (cpu_features, RTM))
|
||||
+ return OPTIMIZE (avx2_rtm);
|
||||
+
|
||||
+ if (!CPU_FEATURES_ARCH_P (cpu_features, Prefer_No_VZEROUPPER))
|
||||
+ return OPTIMIZE (avx2);
|
||||
+ }
|
||||
+
|
||||
+ if (CPU_FEATURE_USABLE_P (cpu_features, SSE4_1))
|
||||
+ return OPTIMIZE (sse4_1);
|
||||
+
|
||||
+ return OPTIMIZE (sse2);
|
||||
+}
|
||||
diff --git a/sysdeps/x86_64/multiarch/wcslen-sse4_1.S b/sysdeps/x86_64/multiarch/wcslen-sse4_1.S
|
||||
new file mode 100644
|
||||
index 00000000..7e62621a
|
||||
--- /dev/null
|
||||
+++ b/sysdeps/x86_64/multiarch/wcslen-sse4_1.S
|
||||
@@ -0,0 +1,4 @@
|
||||
+#define AS_WCSLEN
|
||||
+#define strlen __wcslen_sse4_1
|
||||
+
|
||||
+#include "strlen-vec.S"
|
||||
diff --git a/sysdeps/x86_64/multiarch/wcslen.c b/sysdeps/x86_64/multiarch/wcslen.c
|
||||
index 6d06e47c..3b04b75b 100644
|
||||
--- a/sysdeps/x86_64/multiarch/wcslen.c
|
||||
+++ b/sysdeps/x86_64/multiarch/wcslen.c
|
||||
@@ -24,7 +24,7 @@
|
||||
# undef __wcslen
|
||||
|
||||
# define SYMBOL_NAME wcslen
|
||||
-# include "ifunc-avx2.h"
|
||||
+# include "ifunc-wcslen.h"
|
||||
|
||||
libc_ifunc_redirected (__redirect_wcslen, __wcslen, IFUNC_SELECTOR ());
|
||||
weak_alias (__wcslen, wcslen);
|
||||
diff --git a/sysdeps/x86_64/multiarch/wcsnlen.c b/sysdeps/x86_64/multiarch/wcsnlen.c
|
||||
index 20b731ae..06736410 100644
|
||||
--- a/sysdeps/x86_64/multiarch/wcsnlen.c
|
||||
+++ b/sysdeps/x86_64/multiarch/wcsnlen.c
|
||||
@@ -24,39 +24,7 @@
|
||||
# undef __wcsnlen
|
||||
|
||||
# define SYMBOL_NAME wcsnlen
|
||||
-# include <init-arch.h>
|
||||
-
|
||||
-extern __typeof (REDIRECT_NAME) OPTIMIZE (sse2) attribute_hidden;
|
||||
-extern __typeof (REDIRECT_NAME) OPTIMIZE (sse4_1) attribute_hidden;
|
||||
-extern __typeof (REDIRECT_NAME) OPTIMIZE (avx2) attribute_hidden;
|
||||
-extern __typeof (REDIRECT_NAME) OPTIMIZE (avx2_rtm) attribute_hidden;
|
||||
-extern __typeof (REDIRECT_NAME) OPTIMIZE (evex) attribute_hidden;
|
||||
-
|
||||
-static inline void *
|
||||
-IFUNC_SELECTOR (void)
|
||||
-{
|
||||
- const struct cpu_features* cpu_features = __get_cpu_features ();
|
||||
-
|
||||
- if (CPU_FEATURE_USABLE_P (cpu_features, AVX2)
|
||||
- && CPU_FEATURES_ARCH_P (cpu_features, AVX_Fast_Unaligned_Load))
|
||||
- {
|
||||
- if (CPU_FEATURE_USABLE_P (cpu_features, AVX512VL)
|
||||
- && CPU_FEATURE_USABLE_P (cpu_features, AVX512BW)
|
||||
- && CPU_FEATURE_USABLE_P (cpu_features, BMI2))
|
||||
- return OPTIMIZE (evex);
|
||||
-
|
||||
- if (CPU_FEATURE_USABLE_P (cpu_features, RTM))
|
||||
- return OPTIMIZE (avx2_rtm);
|
||||
-
|
||||
- if (!CPU_FEATURES_ARCH_P (cpu_features, Prefer_No_VZEROUPPER))
|
||||
- return OPTIMIZE (avx2);
|
||||
- }
|
||||
-
|
||||
- if (CPU_FEATURE_USABLE_P (cpu_features, SSE4_1))
|
||||
- return OPTIMIZE (sse4_1);
|
||||
-
|
||||
- return OPTIMIZE (sse2);
|
||||
-}
|
||||
+# include "ifunc-wcslen.h"
|
||||
|
||||
libc_ifunc_redirected (__redirect_wcsnlen, __wcsnlen, IFUNC_SELECTOR ());
|
||||
weak_alias (__wcsnlen, wcsnlen);
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,396 +0,0 @@
|
||||
From 231c56760c1e2ded21ad96bbb860b1f08c556c7a Mon Sep 17 00:00:00 2001
|
||||
From: "H.J. Lu" <hjl.tools@gmail.com>
|
||||
Date: Mon, 21 Jan 2019 11:27:25 -0800
|
||||
Subject: [PATCH] x86-64 memcpy: Properly handle the length parameter [BZ#
|
||||
24097]
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
On x32, the size_t parameter may be passed in the lower 32 bits of a
|
||||
64-bit register with the non-zero upper 32 bits. The string/memory
|
||||
functions written in assembly can only use the lower 32 bits of a
|
||||
64-bit register as length or must clear the upper 32 bits before using
|
||||
the full 64-bit register for length.
|
||||
|
||||
This pach fixes memcpy for x32. Tested on x86-64 and x32. On x86-64,
|
||||
libc.so is the same with and withou the fix.
|
||||
|
||||
[BZ# 24097]
|
||||
CVE-2019-6488
|
||||
* sysdeps/x86_64/multiarch/memcpy-ssse3-back.S: Use RDX_LP for
|
||||
length. Clear the upper 32 bits of RDX register.
|
||||
* sysdeps/x86_64/multiarch/memcpy-ssse3.S: Likewise.
|
||||
* sysdeps/x86_64/multiarch/memmove-avx512-no-vzeroupper.S:
|
||||
Likewise.
|
||||
* sysdeps/x86_64/multiarch/memmove-vec-unaligned-erms.S:
|
||||
Likewise.
|
||||
* sysdeps/x86_64/x32/Makefile (tests): Add tst-size_t-memcpy.
|
||||
tst-size_t-wmemchr.
|
||||
* sysdeps/x86_64/x32/tst-size_t-memcpy.c: New file.
|
||||
---
|
||||
sysdeps/x86_64/multiarch/memcpy-ssse3-back.S | 17 ++++--
|
||||
sysdeps/x86_64/multiarch/memcpy-ssse3.S | 17 ++++--
|
||||
.../multiarch/memmove-avx512-no-vzeroupper.S | 16 +++--
|
||||
.../multiarch/memmove-vec-unaligned-erms.S | 54 +++++++++--------
|
||||
sysdeps/x86_64/x32/Makefile | 2 +-
|
||||
sysdeps/x86_64/x32/tst-size_t-memcpy.c | 58 +++++++++++++++++++
|
||||
6 files changed, 122 insertions(+), 42 deletions(-)
|
||||
create mode 100644 sysdeps/x86_64/x32/tst-size_t-memcpy.c
|
||||
|
||||
Conflicts:
|
||||
ChangeLog
|
||||
(removed)
|
||||
|
||||
diff --git a/sysdeps/x86_64/multiarch/memcpy-ssse3-back.S b/sysdeps/x86_64/multiarch/memcpy-ssse3-back.S
|
||||
index 3cd11233..568eebd3 100644
|
||||
--- a/sysdeps/x86_64/multiarch/memcpy-ssse3-back.S
|
||||
+++ b/sysdeps/x86_64/multiarch/memcpy-ssse3-back.S
|
||||
@@ -45,28 +45,33 @@
|
||||
.section .text.ssse3,"ax",@progbits
|
||||
#if !defined USE_AS_MEMPCPY && !defined USE_AS_MEMMOVE
|
||||
ENTRY (MEMPCPY_CHK)
|
||||
- cmpq %rdx, %rcx
|
||||
+ cmp %RDX_LP, %RCX_LP
|
||||
jb HIDDEN_JUMPTARGET (__chk_fail)
|
||||
END (MEMPCPY_CHK)
|
||||
|
||||
ENTRY (MEMPCPY)
|
||||
- movq %rdi, %rax
|
||||
- addq %rdx, %rax
|
||||
+ mov %RDI_LP, %RAX_LP
|
||||
+ add %RDX_LP, %RAX_LP
|
||||
jmp L(start)
|
||||
END (MEMPCPY)
|
||||
#endif
|
||||
|
||||
#if !defined USE_AS_BCOPY
|
||||
ENTRY (MEMCPY_CHK)
|
||||
- cmpq %rdx, %rcx
|
||||
+ cmp %RDX_LP, %RCX_LP
|
||||
jb HIDDEN_JUMPTARGET (__chk_fail)
|
||||
END (MEMCPY_CHK)
|
||||
#endif
|
||||
|
||||
ENTRY (MEMCPY)
|
||||
- mov %rdi, %rax
|
||||
+ mov %RDI_LP, %RAX_LP
|
||||
#ifdef USE_AS_MEMPCPY
|
||||
- add %rdx, %rax
|
||||
+ add %RDX_LP, %RAX_LP
|
||||
+#endif
|
||||
+
|
||||
+#ifdef __ILP32__
|
||||
+ /* Clear the upper 32 bits. */
|
||||
+ mov %edx, %edx
|
||||
#endif
|
||||
|
||||
#ifdef USE_AS_MEMMOVE
|
||||
diff --git a/sysdeps/x86_64/multiarch/memcpy-ssse3.S b/sysdeps/x86_64/multiarch/memcpy-ssse3.S
|
||||
index 0240bfa3..0bd5ee99 100644
|
||||
--- a/sysdeps/x86_64/multiarch/memcpy-ssse3.S
|
||||
+++ b/sysdeps/x86_64/multiarch/memcpy-ssse3.S
|
||||
@@ -45,28 +45,33 @@
|
||||
.section .text.ssse3,"ax",@progbits
|
||||
#if !defined USE_AS_MEMPCPY && !defined USE_AS_MEMMOVE
|
||||
ENTRY (MEMPCPY_CHK)
|
||||
- cmpq %rdx, %rcx
|
||||
+ cmp %RDX_LP, %RCX_LP
|
||||
jb HIDDEN_JUMPTARGET (__chk_fail)
|
||||
END (MEMPCPY_CHK)
|
||||
|
||||
ENTRY (MEMPCPY)
|
||||
- movq %rdi, %rax
|
||||
- addq %rdx, %rax
|
||||
+ mov %RDI_LP, %RAX_LP
|
||||
+ add %RDX_LP, %RAX_LP
|
||||
jmp L(start)
|
||||
END (MEMPCPY)
|
||||
#endif
|
||||
|
||||
#if !defined USE_AS_BCOPY
|
||||
ENTRY (MEMCPY_CHK)
|
||||
- cmpq %rdx, %rcx
|
||||
+ cmp %RDX_LP, %RCX_LP
|
||||
jb HIDDEN_JUMPTARGET (__chk_fail)
|
||||
END (MEMCPY_CHK)
|
||||
#endif
|
||||
|
||||
ENTRY (MEMCPY)
|
||||
- mov %rdi, %rax
|
||||
+ mov %RDI_LP, %RAX_LP
|
||||
#ifdef USE_AS_MEMPCPY
|
||||
- add %rdx, %rax
|
||||
+ add %RDX_LP, %RAX_LP
|
||||
+#endif
|
||||
+
|
||||
+#ifdef __ILP32__
|
||||
+ /* Clear the upper 32 bits. */
|
||||
+ mov %edx, %edx
|
||||
#endif
|
||||
|
||||
#ifdef USE_AS_MEMMOVE
|
||||
diff --git a/sysdeps/x86_64/multiarch/memmove-avx512-no-vzeroupper.S b/sysdeps/x86_64/multiarch/memmove-avx512-no-vzeroupper.S
|
||||
index effc3ac2..6ca2bbc9 100644
|
||||
--- a/sysdeps/x86_64/multiarch/memmove-avx512-no-vzeroupper.S
|
||||
+++ b/sysdeps/x86_64/multiarch/memmove-avx512-no-vzeroupper.S
|
||||
@@ -24,27 +24,31 @@
|
||||
|
||||
.section .text.avx512,"ax",@progbits
|
||||
ENTRY (__mempcpy_chk_avx512_no_vzeroupper)
|
||||
- cmpq %rdx, %rcx
|
||||
+ cmp %RDX_LP, %RCX_LP
|
||||
jb HIDDEN_JUMPTARGET (__chk_fail)
|
||||
END (__mempcpy_chk_avx512_no_vzeroupper)
|
||||
|
||||
ENTRY (__mempcpy_avx512_no_vzeroupper)
|
||||
- movq %rdi, %rax
|
||||
- addq %rdx, %rax
|
||||
+ mov %RDI_LP, %RAX_LP
|
||||
+ add %RDX_LP, %RAX_LP
|
||||
jmp L(start)
|
||||
END (__mempcpy_avx512_no_vzeroupper)
|
||||
|
||||
ENTRY (__memmove_chk_avx512_no_vzeroupper)
|
||||
- cmpq %rdx, %rcx
|
||||
+ cmp %RDX_LP, %RCX_LP
|
||||
jb HIDDEN_JUMPTARGET (__chk_fail)
|
||||
END (__memmove_chk_avx512_no_vzeroupper)
|
||||
|
||||
ENTRY (__memmove_avx512_no_vzeroupper)
|
||||
- mov %rdi, %rax
|
||||
+ mov %RDI_LP, %RAX_LP
|
||||
# ifdef USE_AS_MEMPCPY
|
||||
- add %rdx, %rax
|
||||
+ add %RDX_LP, %RAX_LP
|
||||
# endif
|
||||
L(start):
|
||||
+# ifdef __ILP32__
|
||||
+ /* Clear the upper 32 bits. */
|
||||
+ mov %edx, %edx
|
||||
+# endif
|
||||
lea (%rsi, %rdx), %rcx
|
||||
lea (%rdi, %rdx), %r9
|
||||
cmp $512, %rdx
|
||||
diff --git a/sysdeps/x86_64/multiarch/memmove-vec-unaligned-erms.S b/sysdeps/x86_64/multiarch/memmove-vec-unaligned-erms.S
|
||||
index c952576c..274aa1c7 100644
|
||||
--- a/sysdeps/x86_64/multiarch/memmove-vec-unaligned-erms.S
|
||||
+++ b/sysdeps/x86_64/multiarch/memmove-vec-unaligned-erms.S
|
||||
@@ -95,20 +95,20 @@
|
||||
.section SECTION(.text),"ax",@progbits
|
||||
#if defined SHARED && IS_IN (libc)
|
||||
ENTRY (MEMMOVE_CHK_SYMBOL (__mempcpy_chk, unaligned))
|
||||
- cmpq %rdx, %rcx
|
||||
+ cmp %RDX_LP, %RCX_LP
|
||||
jb HIDDEN_JUMPTARGET (__chk_fail)
|
||||
END (MEMMOVE_CHK_SYMBOL (__mempcpy_chk, unaligned))
|
||||
#endif
|
||||
|
||||
ENTRY (MEMPCPY_SYMBOL (__mempcpy, unaligned))
|
||||
- movq %rdi, %rax
|
||||
- addq %rdx, %rax
|
||||
+ mov %RDI_LP, %RAX_LP
|
||||
+ add %RDX_LP, %RAX_LP
|
||||
jmp L(start)
|
||||
END (MEMPCPY_SYMBOL (__mempcpy, unaligned))
|
||||
|
||||
#if defined SHARED && IS_IN (libc)
|
||||
ENTRY (MEMMOVE_CHK_SYMBOL (__memmove_chk, unaligned))
|
||||
- cmpq %rdx, %rcx
|
||||
+ cmp %RDX_LP, %RCX_LP
|
||||
jb HIDDEN_JUMPTARGET (__chk_fail)
|
||||
END (MEMMOVE_CHK_SYMBOL (__memmove_chk, unaligned))
|
||||
#endif
|
||||
@@ -116,9 +116,13 @@ END (MEMMOVE_CHK_SYMBOL (__memmove_chk, unaligned))
|
||||
ENTRY (MEMMOVE_SYMBOL (__memmove, unaligned))
|
||||
movq %rdi, %rax
|
||||
L(start):
|
||||
- cmpq $VEC_SIZE, %rdx
|
||||
+# ifdef __ILP32__
|
||||
+ /* Clear the upper 32 bits. */
|
||||
+ movl %edx, %edx
|
||||
+# endif
|
||||
+ cmp $VEC_SIZE, %RDX_LP
|
||||
jb L(less_vec)
|
||||
- cmpq $(VEC_SIZE * 2), %rdx
|
||||
+ cmp $(VEC_SIZE * 2), %RDX_LP
|
||||
ja L(more_2x_vec)
|
||||
#if !defined USE_MULTIARCH || !IS_IN (libc)
|
||||
L(last_2x_vec):
|
||||
@@ -138,38 +142,38 @@ END (MEMMOVE_SYMBOL (__memmove, unaligned))
|
||||
|
||||
# if VEC_SIZE == 16
|
||||
ENTRY (__mempcpy_chk_erms)
|
||||
- cmpq %rdx, %rcx
|
||||
+ cmp %RDX_LP, %RCX_LP
|
||||
jb HIDDEN_JUMPTARGET (__chk_fail)
|
||||
END (__mempcpy_chk_erms)
|
||||
|
||||
/* Only used to measure performance of REP MOVSB. */
|
||||
ENTRY (__mempcpy_erms)
|
||||
- movq %rdi, %rax
|
||||
+ mov %RDI_LP, %RAX_LP
|
||||
/* Skip zero length. */
|
||||
- testq %rdx, %rdx
|
||||
+ test %RDX_LP, %RDX_LP
|
||||
jz 2f
|
||||
- addq %rdx, %rax
|
||||
+ add %RDX_LP, %RAX_LP
|
||||
jmp L(start_movsb)
|
||||
END (__mempcpy_erms)
|
||||
|
||||
ENTRY (__memmove_chk_erms)
|
||||
- cmpq %rdx, %rcx
|
||||
+ cmp %RDX_LP, %RCX_LP
|
||||
jb HIDDEN_JUMPTARGET (__chk_fail)
|
||||
END (__memmove_chk_erms)
|
||||
|
||||
ENTRY (__memmove_erms)
|
||||
movq %rdi, %rax
|
||||
/* Skip zero length. */
|
||||
- testq %rdx, %rdx
|
||||
+ test %RDX_LP, %RDX_LP
|
||||
jz 2f
|
||||
L(start_movsb):
|
||||
- movq %rdx, %rcx
|
||||
- cmpq %rsi, %rdi
|
||||
+ mov %RDX_LP, %RCX_LP
|
||||
+ cmp %RSI_LP, %RDI_LP
|
||||
jb 1f
|
||||
/* Source == destination is less common. */
|
||||
je 2f
|
||||
- leaq (%rsi,%rcx), %rdx
|
||||
- cmpq %rdx, %rdi
|
||||
+ lea (%rsi,%rcx), %RDX_LP
|
||||
+ cmp %RDX_LP, %RDI_LP
|
||||
jb L(movsb_backward)
|
||||
1:
|
||||
rep movsb
|
||||
@@ -189,20 +193,20 @@ strong_alias (__memmove_chk_erms, __memcpy_chk_erms)
|
||||
|
||||
# ifdef SHARED
|
||||
ENTRY (MEMMOVE_CHK_SYMBOL (__mempcpy_chk, unaligned_erms))
|
||||
- cmpq %rdx, %rcx
|
||||
+ cmp %RDX_LP, %RCX_LP
|
||||
jb HIDDEN_JUMPTARGET (__chk_fail)
|
||||
END (MEMMOVE_CHK_SYMBOL (__mempcpy_chk, unaligned_erms))
|
||||
# endif
|
||||
|
||||
ENTRY (MEMMOVE_SYMBOL (__mempcpy, unaligned_erms))
|
||||
- movq %rdi, %rax
|
||||
- addq %rdx, %rax
|
||||
+ mov %RDI_LP, %RAX_LP
|
||||
+ add %RDX_LP, %RAX_LP
|
||||
jmp L(start_erms)
|
||||
END (MEMMOVE_SYMBOL (__mempcpy, unaligned_erms))
|
||||
|
||||
# ifdef SHARED
|
||||
ENTRY (MEMMOVE_CHK_SYMBOL (__memmove_chk, unaligned_erms))
|
||||
- cmpq %rdx, %rcx
|
||||
+ cmp %RDX_LP, %RCX_LP
|
||||
jb HIDDEN_JUMPTARGET (__chk_fail)
|
||||
END (MEMMOVE_CHK_SYMBOL (__memmove_chk, unaligned_erms))
|
||||
# endif
|
||||
@@ -210,9 +214,13 @@ END (MEMMOVE_CHK_SYMBOL (__memmove_chk, unaligned_erms))
|
||||
ENTRY (MEMMOVE_SYMBOL (__memmove, unaligned_erms))
|
||||
movq %rdi, %rax
|
||||
L(start_erms):
|
||||
- cmpq $VEC_SIZE, %rdx
|
||||
+# ifdef __ILP32__
|
||||
+ /* Clear the upper 32 bits. */
|
||||
+ movl %edx, %edx
|
||||
+# endif
|
||||
+ cmp $VEC_SIZE, %RDX_LP
|
||||
jb L(less_vec)
|
||||
- cmpq $(VEC_SIZE * 2), %rdx
|
||||
+ cmp $(VEC_SIZE * 2), %RDX_LP
|
||||
ja L(movsb_more_2x_vec)
|
||||
L(last_2x_vec):
|
||||
/* From VEC and to 2 * VEC. No branch when size == VEC_SIZE. */
|
||||
@@ -236,7 +244,7 @@ L(movsb):
|
||||
/* Avoid slow backward REP MOVSB. */
|
||||
jb L(more_8x_vec_backward)
|
||||
1:
|
||||
- movq %rdx, %rcx
|
||||
+ mov %RDX_LP, %RCX_LP
|
||||
rep movsb
|
||||
L(nop):
|
||||
ret
|
||||
diff --git a/sysdeps/x86_64/x32/Makefile b/sysdeps/x86_64/x32/Makefile
|
||||
index ddec7f04..2fe1e5ac 100644
|
||||
--- a/sysdeps/x86_64/x32/Makefile
|
||||
+++ b/sysdeps/x86_64/x32/Makefile
|
||||
@@ -6,7 +6,7 @@ CFLAGS-s_llround.c += -fno-builtin-lround
|
||||
endif
|
||||
|
||||
ifeq ($(subdir),string)
|
||||
-tests += tst-size_t-memchr tst-size_t-memcmp
|
||||
+tests += tst-size_t-memchr tst-size_t-memcmp tst-size_t-memcpy
|
||||
endif
|
||||
|
||||
ifeq ($(subdir),wcsmbs)
|
||||
diff --git a/sysdeps/x86_64/x32/tst-size_t-memcpy.c b/sysdeps/x86_64/x32/tst-size_t-memcpy.c
|
||||
new file mode 100644
|
||||
index 00000000..66b71e17
|
||||
--- /dev/null
|
||||
+++ b/sysdeps/x86_64/x32/tst-size_t-memcpy.c
|
||||
@@ -0,0 +1,58 @@
|
||||
+/* Test memcpy with size_t in the lower 32 bits of 64-bit register.
|
||||
+ Copyright (C) 2019 Free Software Foundation, Inc.
|
||||
+ This file is part of the GNU C Library.
|
||||
+
|
||||
+ The GNU C Library is free software; you can redistribute it and/or
|
||||
+ modify it under the terms of the GNU Lesser General Public
|
||||
+ License as published by the Free Software Foundation; either
|
||||
+ version 2.1 of the License, or (at your option) any later version.
|
||||
+
|
||||
+ The GNU C Library is distributed in the hope that it will be useful,
|
||||
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
||||
+ Lesser General Public License for more details.
|
||||
+
|
||||
+ You should have received a copy of the GNU Lesser General Public
|
||||
+ License along with the GNU C Library; if not, see
|
||||
+ <http://www.gnu.org/licenses/>. */
|
||||
+
|
||||
+#define TEST_NAME "memcpy"
|
||||
+#include "test-size_t.h"
|
||||
+
|
||||
+IMPL (memcpy, 1)
|
||||
+
|
||||
+typedef void *(*proto_t) (void *, const void *, size_t);
|
||||
+
|
||||
+static void *
|
||||
+__attribute__ ((noinline, noclone))
|
||||
+do_memcpy (parameter_t a, parameter_t b)
|
||||
+{
|
||||
+ return CALL (&b, a.p, b.p, a.len);
|
||||
+}
|
||||
+
|
||||
+static int
|
||||
+test_main (void)
|
||||
+{
|
||||
+ test_init ();
|
||||
+
|
||||
+ parameter_t dest = { { page_size }, buf1 };
|
||||
+ parameter_t src = { { 0 }, buf2 };
|
||||
+
|
||||
+ int ret = 0;
|
||||
+ FOR_EACH_IMPL (impl, 0)
|
||||
+ {
|
||||
+ src.fn = impl->fn;
|
||||
+ do_memcpy (dest, src);
|
||||
+ int res = memcmp (dest.p, src.p, dest.len);
|
||||
+ if (res)
|
||||
+ {
|
||||
+ error (0, 0, "Wrong result in function %s: %i != 0",
|
||||
+ impl->name, res);
|
||||
+ ret = 1;
|
||||
+ }
|
||||
+ }
|
||||
+
|
||||
+ return ret ? EXIT_FAILURE : EXIT_SUCCESS;
|
||||
+}
|
||||
+
|
||||
+#include <support/test-driver.c>
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,497 +0,0 @@
|
||||
From a775a7a3eb1e85b54af0b4ee5ff4dcf66772a1fb Mon Sep 17 00:00:00 2001
|
||||
From: Noah Goldstein <goldstein.w.n@gmail.com>
|
||||
Date: Wed, 23 Jun 2021 01:56:29 -0400
|
||||
Subject: [PATCH] x86: Fix overflow bug in wcsnlen-sse4_1 and wcsnlen-avx2 [BZ
|
||||
#27974]
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
This commit fixes the bug mentioned in the previous commit.
|
||||
|
||||
The previous implementations of wmemchr in these files relied
|
||||
on maxlen * sizeof(wchar_t) which was not guranteed by the standard.
|
||||
|
||||
The new overflow tests added in the previous commit now
|
||||
pass (As well as all the other tests).
|
||||
|
||||
Signed-off-by: Noah Goldstein <goldstein.w.n@gmail.com>
|
||||
Reviewed-by: H.J. Lu <hjl.tools@gmail.com>
|
||||
---
|
||||
sysdeps/x86_64/multiarch/strlen-avx2.S | 130 ++++++++++++++++++-------
|
||||
sysdeps/x86_64/multiarch/strlen-vec.S | 15 ++-
|
||||
2 files changed, 107 insertions(+), 38 deletions(-)
|
||||
|
||||
diff --git a/sysdeps/x86_64/multiarch/strlen-avx2.S b/sysdeps/x86_64/multiarch/strlen-avx2.S
|
||||
index be8a5db5..37688966 100644
|
||||
--- a/sysdeps/x86_64/multiarch/strlen-avx2.S
|
||||
+++ b/sysdeps/x86_64/multiarch/strlen-avx2.S
|
||||
@@ -44,21 +44,21 @@
|
||||
|
||||
# define VEC_SIZE 32
|
||||
# define PAGE_SIZE 4096
|
||||
+# define CHAR_PER_VEC (VEC_SIZE / CHAR_SIZE)
|
||||
|
||||
.section SECTION(.text),"ax",@progbits
|
||||
ENTRY (STRLEN)
|
||||
# ifdef USE_AS_STRNLEN
|
||||
/* Check zero length. */
|
||||
+# ifdef __ILP32__
|
||||
+ /* Clear upper bits. */
|
||||
+ and %RSI_LP, %RSI_LP
|
||||
+# else
|
||||
test %RSI_LP, %RSI_LP
|
||||
+# endif
|
||||
jz L(zero)
|
||||
/* Store max len in R8_LP before adjusting if using WCSLEN. */
|
||||
mov %RSI_LP, %R8_LP
|
||||
-# ifdef USE_AS_WCSLEN
|
||||
- shl $2, %RSI_LP
|
||||
-# elif defined __ILP32__
|
||||
- /* Clear the upper 32 bits. */
|
||||
- movl %esi, %esi
|
||||
-# endif
|
||||
# endif
|
||||
movl %edi, %eax
|
||||
movq %rdi, %rdx
|
||||
@@ -72,10 +72,10 @@ ENTRY (STRLEN)
|
||||
|
||||
/* Check the first VEC_SIZE bytes. */
|
||||
VPCMPEQ (%rdi), %ymm0, %ymm1
|
||||
- vpmovmskb %ymm1, %eax
|
||||
+ vpmovmskb %ymm1, %eax
|
||||
# ifdef USE_AS_STRNLEN
|
||||
/* If length < VEC_SIZE handle special. */
|
||||
- cmpq $VEC_SIZE, %rsi
|
||||
+ cmpq $CHAR_PER_VEC, %rsi
|
||||
jbe L(first_vec_x0)
|
||||
# endif
|
||||
/* If empty continue to aligned_more. Otherwise return bit
|
||||
@@ -84,6 +84,7 @@ ENTRY (STRLEN)
|
||||
jz L(aligned_more)
|
||||
tzcntl %eax, %eax
|
||||
# ifdef USE_AS_WCSLEN
|
||||
+ /* NB: Divide bytes by 4 to get wchar_t count. */
|
||||
shrl $2, %eax
|
||||
# endif
|
||||
VZEROUPPER_RETURN
|
||||
@@ -97,9 +98,14 @@ L(zero):
|
||||
L(first_vec_x0):
|
||||
/* Set bit for max len so that tzcnt will return min of max len
|
||||
and position of first match. */
|
||||
+# ifdef USE_AS_WCSLEN
|
||||
+ /* NB: Multiply length by 4 to get byte count. */
|
||||
+ sall $2, %esi
|
||||
+# endif
|
||||
btsq %rsi, %rax
|
||||
tzcntl %eax, %eax
|
||||
# ifdef USE_AS_WCSLEN
|
||||
+ /* NB: Divide bytes by 4 to get wchar_t count. */
|
||||
shrl $2, %eax
|
||||
# endif
|
||||
VZEROUPPER_RETURN
|
||||
@@ -113,14 +119,19 @@ L(first_vec_x1):
|
||||
# ifdef USE_AS_STRNLEN
|
||||
/* Use ecx which was computed earlier to compute correct value.
|
||||
*/
|
||||
+# ifdef USE_AS_WCSLEN
|
||||
+ leal -(VEC_SIZE * 4 + 1)(%rax, %rcx, 4), %eax
|
||||
+# else
|
||||
subl $(VEC_SIZE * 4 + 1), %ecx
|
||||
addl %ecx, %eax
|
||||
+# endif
|
||||
# else
|
||||
subl %edx, %edi
|
||||
incl %edi
|
||||
addl %edi, %eax
|
||||
# endif
|
||||
# ifdef USE_AS_WCSLEN
|
||||
+ /* NB: Divide bytes by 4 to get wchar_t count. */
|
||||
shrl $2, %eax
|
||||
# endif
|
||||
VZEROUPPER_RETURN
|
||||
@@ -133,14 +144,19 @@ L(first_vec_x2):
|
||||
# ifdef USE_AS_STRNLEN
|
||||
/* Use ecx which was computed earlier to compute correct value.
|
||||
*/
|
||||
+# ifdef USE_AS_WCSLEN
|
||||
+ leal -(VEC_SIZE * 3 + 1)(%rax, %rcx, 4), %eax
|
||||
+# else
|
||||
subl $(VEC_SIZE * 3 + 1), %ecx
|
||||
addl %ecx, %eax
|
||||
+# endif
|
||||
# else
|
||||
subl %edx, %edi
|
||||
addl $(VEC_SIZE + 1), %edi
|
||||
addl %edi, %eax
|
||||
# endif
|
||||
# ifdef USE_AS_WCSLEN
|
||||
+ /* NB: Divide bytes by 4 to get wchar_t count. */
|
||||
shrl $2, %eax
|
||||
# endif
|
||||
VZEROUPPER_RETURN
|
||||
@@ -153,14 +169,19 @@ L(first_vec_x3):
|
||||
# ifdef USE_AS_STRNLEN
|
||||
/* Use ecx which was computed earlier to compute correct value.
|
||||
*/
|
||||
+# ifdef USE_AS_WCSLEN
|
||||
+ leal -(VEC_SIZE * 2 + 1)(%rax, %rcx, 4), %eax
|
||||
+# else
|
||||
subl $(VEC_SIZE * 2 + 1), %ecx
|
||||
addl %ecx, %eax
|
||||
+# endif
|
||||
# else
|
||||
subl %edx, %edi
|
||||
addl $(VEC_SIZE * 2 + 1), %edi
|
||||
addl %edi, %eax
|
||||
# endif
|
||||
# ifdef USE_AS_WCSLEN
|
||||
+ /* NB: Divide bytes by 4 to get wchar_t count. */
|
||||
shrl $2, %eax
|
||||
# endif
|
||||
VZEROUPPER_RETURN
|
||||
@@ -173,14 +194,19 @@ L(first_vec_x4):
|
||||
# ifdef USE_AS_STRNLEN
|
||||
/* Use ecx which was computed earlier to compute correct value.
|
||||
*/
|
||||
+# ifdef USE_AS_WCSLEN
|
||||
+ leal -(VEC_SIZE * 1 + 1)(%rax, %rcx, 4), %eax
|
||||
+# else
|
||||
subl $(VEC_SIZE + 1), %ecx
|
||||
addl %ecx, %eax
|
||||
+# endif
|
||||
# else
|
||||
subl %edx, %edi
|
||||
addl $(VEC_SIZE * 3 + 1), %edi
|
||||
addl %edi, %eax
|
||||
# endif
|
||||
# ifdef USE_AS_WCSLEN
|
||||
+ /* NB: Divide bytes by 4 to get wchar_t count. */
|
||||
shrl $2, %eax
|
||||
# endif
|
||||
VZEROUPPER_RETURN
|
||||
@@ -195,10 +221,14 @@ L(cross_page_continue):
|
||||
/* Check the first 4 * VEC_SIZE. Only one VEC_SIZE at a time
|
||||
since data is only aligned to VEC_SIZE. */
|
||||
# ifdef USE_AS_STRNLEN
|
||||
- /* + 1 because rdi is aligned to VEC_SIZE - 1. + CHAR_SIZE because
|
||||
- it simplies the logic in last_4x_vec_or_less. */
|
||||
+ /* + 1 because rdi is aligned to VEC_SIZE - 1. + CHAR_SIZE
|
||||
+ because it simplies the logic in last_4x_vec_or_less. */
|
||||
leaq (VEC_SIZE * 4 + CHAR_SIZE + 1)(%rdi), %rcx
|
||||
subq %rdx, %rcx
|
||||
+# ifdef USE_AS_WCSLEN
|
||||
+ /* NB: Divide bytes by 4 to get the wchar_t count. */
|
||||
+ sarl $2, %ecx
|
||||
+# endif
|
||||
# endif
|
||||
/* Load first VEC regardless. */
|
||||
VPCMPEQ 1(%rdi), %ymm0, %ymm1
|
||||
@@ -207,34 +237,38 @@ L(cross_page_continue):
|
||||
subq %rcx, %rsi
|
||||
jb L(last_4x_vec_or_less)
|
||||
# endif
|
||||
- vpmovmskb %ymm1, %eax
|
||||
+ vpmovmskb %ymm1, %eax
|
||||
testl %eax, %eax
|
||||
jnz L(first_vec_x1)
|
||||
|
||||
VPCMPEQ (VEC_SIZE + 1)(%rdi), %ymm0, %ymm1
|
||||
- vpmovmskb %ymm1, %eax
|
||||
+ vpmovmskb %ymm1, %eax
|
||||
testl %eax, %eax
|
||||
jnz L(first_vec_x2)
|
||||
|
||||
VPCMPEQ (VEC_SIZE * 2 + 1)(%rdi), %ymm0, %ymm1
|
||||
- vpmovmskb %ymm1, %eax
|
||||
+ vpmovmskb %ymm1, %eax
|
||||
testl %eax, %eax
|
||||
jnz L(first_vec_x3)
|
||||
|
||||
VPCMPEQ (VEC_SIZE * 3 + 1)(%rdi), %ymm0, %ymm1
|
||||
- vpmovmskb %ymm1, %eax
|
||||
+ vpmovmskb %ymm1, %eax
|
||||
testl %eax, %eax
|
||||
jnz L(first_vec_x4)
|
||||
|
||||
/* Align data to VEC_SIZE * 4 - 1. */
|
||||
# ifdef USE_AS_STRNLEN
|
||||
/* Before adjusting length check if at last VEC_SIZE * 4. */
|
||||
- cmpq $(VEC_SIZE * 4 - 1), %rsi
|
||||
+ cmpq $(CHAR_PER_VEC * 4 - 1), %rsi
|
||||
jbe L(last_4x_vec_or_less_load)
|
||||
incq %rdi
|
||||
movl %edi, %ecx
|
||||
orq $(VEC_SIZE * 4 - 1), %rdi
|
||||
andl $(VEC_SIZE * 4 - 1), %ecx
|
||||
+# ifdef USE_AS_WCSLEN
|
||||
+ /* NB: Divide bytes by 4 to get the wchar_t count. */
|
||||
+ sarl $2, %ecx
|
||||
+# endif
|
||||
/* Readjust length. */
|
||||
addq %rcx, %rsi
|
||||
# else
|
||||
@@ -246,13 +280,13 @@ L(cross_page_continue):
|
||||
L(loop_4x_vec):
|
||||
# ifdef USE_AS_STRNLEN
|
||||
/* Break if at end of length. */
|
||||
- subq $(VEC_SIZE * 4), %rsi
|
||||
+ subq $(CHAR_PER_VEC * 4), %rsi
|
||||
jb L(last_4x_vec_or_less_cmpeq)
|
||||
# endif
|
||||
- /* Save some code size by microfusing VPMINU with the load. Since
|
||||
- the matches in ymm2/ymm4 can only be returned if there where no
|
||||
- matches in ymm1/ymm3 respectively there is no issue with overlap.
|
||||
- */
|
||||
+ /* Save some code size by microfusing VPMINU with the load.
|
||||
+ Since the matches in ymm2/ymm4 can only be returned if there
|
||||
+ where no matches in ymm1/ymm3 respectively there is no issue
|
||||
+ with overlap. */
|
||||
vmovdqa 1(%rdi), %ymm1
|
||||
VPMINU (VEC_SIZE + 1)(%rdi), %ymm1, %ymm2
|
||||
vmovdqa (VEC_SIZE * 2 + 1)(%rdi), %ymm3
|
||||
@@ -260,7 +294,7 @@ L(loop_4x_vec):
|
||||
|
||||
VPMINU %ymm2, %ymm4, %ymm5
|
||||
VPCMPEQ %ymm5, %ymm0, %ymm5
|
||||
- vpmovmskb %ymm5, %ecx
|
||||
+ vpmovmskb %ymm5, %ecx
|
||||
|
||||
subq $-(VEC_SIZE * 4), %rdi
|
||||
testl %ecx, %ecx
|
||||
@@ -268,27 +302,28 @@ L(loop_4x_vec):
|
||||
|
||||
|
||||
VPCMPEQ %ymm1, %ymm0, %ymm1
|
||||
- vpmovmskb %ymm1, %eax
|
||||
+ vpmovmskb %ymm1, %eax
|
||||
subq %rdx, %rdi
|
||||
testl %eax, %eax
|
||||
jnz L(last_vec_return_x0)
|
||||
|
||||
VPCMPEQ %ymm2, %ymm0, %ymm2
|
||||
- vpmovmskb %ymm2, %eax
|
||||
+ vpmovmskb %ymm2, %eax
|
||||
testl %eax, %eax
|
||||
jnz L(last_vec_return_x1)
|
||||
|
||||
/* Combine last 2 VEC. */
|
||||
VPCMPEQ %ymm3, %ymm0, %ymm3
|
||||
- vpmovmskb %ymm3, %eax
|
||||
- /* rcx has combined result from all 4 VEC. It will only be used if
|
||||
- the first 3 other VEC all did not contain a match. */
|
||||
+ vpmovmskb %ymm3, %eax
|
||||
+ /* rcx has combined result from all 4 VEC. It will only be used
|
||||
+ if the first 3 other VEC all did not contain a match. */
|
||||
salq $32, %rcx
|
||||
orq %rcx, %rax
|
||||
tzcntq %rax, %rax
|
||||
subq $(VEC_SIZE * 2 - 1), %rdi
|
||||
addq %rdi, %rax
|
||||
# ifdef USE_AS_WCSLEN
|
||||
+ /* NB: Divide bytes by 4 to get wchar_t count. */
|
||||
shrq $2, %rax
|
||||
# endif
|
||||
VZEROUPPER_RETURN
|
||||
@@ -297,15 +332,19 @@ L(loop_4x_vec):
|
||||
# ifdef USE_AS_STRNLEN
|
||||
.p2align 4
|
||||
L(last_4x_vec_or_less_load):
|
||||
- /* Depending on entry adjust rdi / prepare first VEC in ymm1. */
|
||||
+ /* Depending on entry adjust rdi / prepare first VEC in ymm1.
|
||||
+ */
|
||||
subq $-(VEC_SIZE * 4), %rdi
|
||||
L(last_4x_vec_or_less_cmpeq):
|
||||
VPCMPEQ 1(%rdi), %ymm0, %ymm1
|
||||
L(last_4x_vec_or_less):
|
||||
-
|
||||
- vpmovmskb %ymm1, %eax
|
||||
- /* If remaining length > VEC_SIZE * 2. This works if esi is off by
|
||||
- VEC_SIZE * 4. */
|
||||
+# ifdef USE_AS_WCSLEN
|
||||
+ /* NB: Multiply length by 4 to get byte count. */
|
||||
+ sall $2, %esi
|
||||
+# endif
|
||||
+ vpmovmskb %ymm1, %eax
|
||||
+ /* If remaining length > VEC_SIZE * 2. This works if esi is off
|
||||
+ by VEC_SIZE * 4. */
|
||||
testl $(VEC_SIZE * 2), %esi
|
||||
jnz L(last_4x_vec)
|
||||
|
||||
@@ -320,7 +359,7 @@ L(last_4x_vec_or_less):
|
||||
jb L(max)
|
||||
|
||||
VPCMPEQ (VEC_SIZE + 1)(%rdi), %ymm0, %ymm1
|
||||
- vpmovmskb %ymm1, %eax
|
||||
+ vpmovmskb %ymm1, %eax
|
||||
tzcntl %eax, %eax
|
||||
/* Check the end of data. */
|
||||
cmpl %eax, %esi
|
||||
@@ -329,6 +368,7 @@ L(last_4x_vec_or_less):
|
||||
addl $(VEC_SIZE + 1), %eax
|
||||
addq %rdi, %rax
|
||||
# ifdef USE_AS_WCSLEN
|
||||
+ /* NB: Divide bytes by 4 to get wchar_t count. */
|
||||
shrq $2, %rax
|
||||
# endif
|
||||
VZEROUPPER_RETURN
|
||||
@@ -340,6 +380,7 @@ L(last_vec_return_x0):
|
||||
subq $(VEC_SIZE * 4 - 1), %rdi
|
||||
addq %rdi, %rax
|
||||
# ifdef USE_AS_WCSLEN
|
||||
+ /* NB: Divide bytes by 4 to get wchar_t count. */
|
||||
shrq $2, %rax
|
||||
# endif
|
||||
VZEROUPPER_RETURN
|
||||
@@ -350,6 +391,7 @@ L(last_vec_return_x1):
|
||||
subq $(VEC_SIZE * 3 - 1), %rdi
|
||||
addq %rdi, %rax
|
||||
# ifdef USE_AS_WCSLEN
|
||||
+ /* NB: Divide bytes by 4 to get wchar_t count. */
|
||||
shrq $2, %rax
|
||||
# endif
|
||||
VZEROUPPER_RETURN
|
||||
@@ -366,6 +408,7 @@ L(last_vec_x1_check):
|
||||
incl %eax
|
||||
addq %rdi, %rax
|
||||
# ifdef USE_AS_WCSLEN
|
||||
+ /* NB: Divide bytes by 4 to get wchar_t count. */
|
||||
shrq $2, %rax
|
||||
# endif
|
||||
VZEROUPPER_RETURN
|
||||
@@ -381,14 +424,14 @@ L(last_4x_vec):
|
||||
jnz L(last_vec_x1)
|
||||
|
||||
VPCMPEQ (VEC_SIZE + 1)(%rdi), %ymm0, %ymm1
|
||||
- vpmovmskb %ymm1, %eax
|
||||
+ vpmovmskb %ymm1, %eax
|
||||
testl %eax, %eax
|
||||
jnz L(last_vec_x2)
|
||||
|
||||
/* Normalize length. */
|
||||
andl $(VEC_SIZE * 4 - 1), %esi
|
||||
VPCMPEQ (VEC_SIZE * 2 + 1)(%rdi), %ymm0, %ymm1
|
||||
- vpmovmskb %ymm1, %eax
|
||||
+ vpmovmskb %ymm1, %eax
|
||||
testl %eax, %eax
|
||||
jnz L(last_vec_x3)
|
||||
|
||||
@@ -396,7 +439,7 @@ L(last_4x_vec):
|
||||
jb L(max)
|
||||
|
||||
VPCMPEQ (VEC_SIZE * 3 + 1)(%rdi), %ymm0, %ymm1
|
||||
- vpmovmskb %ymm1, %eax
|
||||
+ vpmovmskb %ymm1, %eax
|
||||
tzcntl %eax, %eax
|
||||
/* Check the end of data. */
|
||||
cmpl %eax, %esi
|
||||
@@ -405,6 +448,7 @@ L(last_4x_vec):
|
||||
addl $(VEC_SIZE * 3 + 1), %eax
|
||||
addq %rdi, %rax
|
||||
# ifdef USE_AS_WCSLEN
|
||||
+ /* NB: Divide bytes by 4 to get wchar_t count. */
|
||||
shrq $2, %rax
|
||||
# endif
|
||||
VZEROUPPER_RETURN
|
||||
@@ -419,6 +463,7 @@ L(last_vec_x1):
|
||||
incl %eax
|
||||
addq %rdi, %rax
|
||||
# ifdef USE_AS_WCSLEN
|
||||
+ /* NB: Divide bytes by 4 to get wchar_t count. */
|
||||
shrq $2, %rax
|
||||
# endif
|
||||
VZEROUPPER_RETURN
|
||||
@@ -432,6 +477,7 @@ L(last_vec_x2):
|
||||
addl $(VEC_SIZE + 1), %eax
|
||||
addq %rdi, %rax
|
||||
# ifdef USE_AS_WCSLEN
|
||||
+ /* NB: Divide bytes by 4 to get wchar_t count. */
|
||||
shrq $2, %rax
|
||||
# endif
|
||||
VZEROUPPER_RETURN
|
||||
@@ -447,6 +493,7 @@ L(last_vec_x3):
|
||||
addl $(VEC_SIZE * 2 + 1), %eax
|
||||
addq %rdi, %rax
|
||||
# ifdef USE_AS_WCSLEN
|
||||
+ /* NB: Divide bytes by 4 to get wchar_t count. */
|
||||
shrq $2, %rax
|
||||
# endif
|
||||
VZEROUPPER_RETURN
|
||||
@@ -455,13 +502,13 @@ L(max_end):
|
||||
VZEROUPPER_RETURN
|
||||
# endif
|
||||
|
||||
- /* Cold case for crossing page with first load. */
|
||||
+ /* Cold case for crossing page with first load. */
|
||||
.p2align 4
|
||||
L(cross_page_boundary):
|
||||
/* Align data to VEC_SIZE - 1. */
|
||||
orq $(VEC_SIZE - 1), %rdi
|
||||
VPCMPEQ -(VEC_SIZE - 1)(%rdi), %ymm0, %ymm1
|
||||
- vpmovmskb %ymm1, %eax
|
||||
+ vpmovmskb %ymm1, %eax
|
||||
/* Remove the leading bytes. sarxl only uses bits [5:0] of COUNT
|
||||
so no need to manually mod rdx. */
|
||||
sarxl %edx, %eax, %eax
|
||||
@@ -470,6 +517,10 @@ L(cross_page_boundary):
|
||||
jnz L(cross_page_less_vec)
|
||||
leaq 1(%rdi), %rcx
|
||||
subq %rdx, %rcx
|
||||
+# ifdef USE_AS_WCSLEN
|
||||
+ /* NB: Divide bytes by 4 to get wchar_t count. */
|
||||
+ shrl $2, %ecx
|
||||
+# endif
|
||||
/* Check length. */
|
||||
cmpq %rsi, %rcx
|
||||
jb L(cross_page_continue)
|
||||
@@ -479,6 +530,7 @@ L(cross_page_boundary):
|
||||
jz L(cross_page_continue)
|
||||
tzcntl %eax, %eax
|
||||
# ifdef USE_AS_WCSLEN
|
||||
+ /* NB: Divide length by 4 to get wchar_t count. */
|
||||
shrl $2, %eax
|
||||
# endif
|
||||
# endif
|
||||
@@ -489,6 +541,10 @@ L(return_vzeroupper):
|
||||
.p2align 4
|
||||
L(cross_page_less_vec):
|
||||
tzcntl %eax, %eax
|
||||
+# ifdef USE_AS_WCSLEN
|
||||
+ /* NB: Multiply length by 4 to get byte count. */
|
||||
+ sall $2, %esi
|
||||
+# endif
|
||||
cmpq %rax, %rsi
|
||||
cmovb %esi, %eax
|
||||
# ifdef USE_AS_WCSLEN
|
||||
diff --git a/sysdeps/x86_64/multiarch/strlen-vec.S b/sysdeps/x86_64/multiarch/strlen-vec.S
|
||||
index 8f660bb9..439e486a 100644
|
||||
--- a/sysdeps/x86_64/multiarch/strlen-vec.S
|
||||
+++ b/sysdeps/x86_64/multiarch/strlen-vec.S
|
||||
@@ -65,12 +65,25 @@ ENTRY(strlen)
|
||||
ret
|
||||
L(n_nonzero):
|
||||
# ifdef AS_WCSLEN
|
||||
- shl $2, %RSI_LP
|
||||
+/* Check for overflow from maxlen * sizeof(wchar_t). If it would
|
||||
+ overflow the only way this program doesn't have undefined behavior
|
||||
+ is if there is a null terminator in valid memory so wcslen will
|
||||
+ suffice. */
|
||||
+ mov %RSI_LP, %R10_LP
|
||||
+ sar $62, %R10_LP
|
||||
+ test %R10_LP, %R10_LP
|
||||
+ jnz __wcslen_sse4_1
|
||||
+ sal $2, %RSI_LP
|
||||
# endif
|
||||
|
||||
+
|
||||
/* Initialize long lived registers. */
|
||||
|
||||
add %RDI_LP, %RSI_LP
|
||||
+# ifdef AS_WCSLEN
|
||||
+/* Check for overflow again from s + maxlen * sizeof(wchar_t). */
|
||||
+ jbe __wcslen_sse4_1
|
||||
+# endif
|
||||
mov %RSI_LP, %R10_LP
|
||||
and $-64, %R10_LP
|
||||
mov %RSI_LP, %R11_LP
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,745 +0,0 @@
|
||||
From 4ba65586847751372520a36757c17f114588794e Mon Sep 17 00:00:00 2001
|
||||
From: Noah Goldstein <goldstein.w.n@gmail.com>
|
||||
Date: Mon, 19 Apr 2021 19:36:06 -0400
|
||||
Subject: [PATCH] x86: Optimize strlen-evex.S
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
No bug. This commit optimizes strlen-evex.S. The
|
||||
optimizations are mostly small things but they add up to roughly
|
||||
10-30% performance improvement for strlen. The results for strnlen are
|
||||
bit more ambiguous. test-strlen, test-strnlen, test-wcslen, and
|
||||
test-wcsnlen are all passing.
|
||||
|
||||
Signed-off-by: Noah Goldstein <goldstein.w.n@gmail.com>
|
||||
---
|
||||
sysdeps/x86_64/multiarch/strlen-evex.S | 581 ++++++++++++++-----------
|
||||
1 file changed, 317 insertions(+), 264 deletions(-)
|
||||
|
||||
diff --git a/sysdeps/x86_64/multiarch/strlen-evex.S b/sysdeps/x86_64/multiarch/strlen-evex.S
|
||||
index 05838190..4bf6874b 100644
|
||||
--- a/sysdeps/x86_64/multiarch/strlen-evex.S
|
||||
+++ b/sysdeps/x86_64/multiarch/strlen-evex.S
|
||||
@@ -29,11 +29,13 @@
|
||||
# ifdef USE_AS_WCSLEN
|
||||
# define VPCMP vpcmpd
|
||||
# define VPMINU vpminud
|
||||
-# define SHIFT_REG r9d
|
||||
+# define SHIFT_REG ecx
|
||||
+# define CHAR_SIZE 4
|
||||
# else
|
||||
# define VPCMP vpcmpb
|
||||
# define VPMINU vpminub
|
||||
-# define SHIFT_REG ecx
|
||||
+# define SHIFT_REG edx
|
||||
+# define CHAR_SIZE 1
|
||||
# endif
|
||||
|
||||
# define XMMZERO xmm16
|
||||
@@ -46,132 +48,165 @@
|
||||
# define YMM6 ymm22
|
||||
|
||||
# define VEC_SIZE 32
|
||||
+# define PAGE_SIZE 4096
|
||||
+# define CHAR_PER_VEC (VEC_SIZE / CHAR_SIZE)
|
||||
|
||||
.section .text.evex,"ax",@progbits
|
||||
ENTRY (STRLEN)
|
||||
# ifdef USE_AS_STRNLEN
|
||||
- /* Check for zero length. */
|
||||
+ /* Check zero length. */
|
||||
test %RSI_LP, %RSI_LP
|
||||
jz L(zero)
|
||||
-# ifdef USE_AS_WCSLEN
|
||||
- shl $2, %RSI_LP
|
||||
-# elif defined __ILP32__
|
||||
+# ifdef __ILP32__
|
||||
/* Clear the upper 32 bits. */
|
||||
movl %esi, %esi
|
||||
# endif
|
||||
mov %RSI_LP, %R8_LP
|
||||
# endif
|
||||
- movl %edi, %ecx
|
||||
- movq %rdi, %rdx
|
||||
+ movl %edi, %eax
|
||||
vpxorq %XMMZERO, %XMMZERO, %XMMZERO
|
||||
-
|
||||
+ /* Clear high bits from edi. Only keeping bits relevant to page
|
||||
+ cross check. */
|
||||
+ andl $(PAGE_SIZE - 1), %eax
|
||||
/* Check if we may cross page boundary with one vector load. */
|
||||
- andl $(2 * VEC_SIZE - 1), %ecx
|
||||
- cmpl $VEC_SIZE, %ecx
|
||||
- ja L(cros_page_boundary)
|
||||
+ cmpl $(PAGE_SIZE - VEC_SIZE), %eax
|
||||
+ ja L(cross_page_boundary)
|
||||
|
||||
/* Check the first VEC_SIZE bytes. Each bit in K0 represents a
|
||||
null byte. */
|
||||
VPCMP $0, (%rdi), %YMMZERO, %k0
|
||||
kmovd %k0, %eax
|
||||
- testl %eax, %eax
|
||||
-
|
||||
# ifdef USE_AS_STRNLEN
|
||||
- jnz L(first_vec_x0_check)
|
||||
- /* Adjust length and check the end of data. */
|
||||
- subq $VEC_SIZE, %rsi
|
||||
- jbe L(max)
|
||||
-# else
|
||||
- jnz L(first_vec_x0)
|
||||
+ /* If length < CHAR_PER_VEC handle special. */
|
||||
+ cmpq $CHAR_PER_VEC, %rsi
|
||||
+ jbe L(first_vec_x0)
|
||||
# endif
|
||||
-
|
||||
- /* Align data for aligned loads in the loop. */
|
||||
- addq $VEC_SIZE, %rdi
|
||||
- andl $(VEC_SIZE - 1), %ecx
|
||||
- andq $-VEC_SIZE, %rdi
|
||||
-
|
||||
+ testl %eax, %eax
|
||||
+ jz L(aligned_more)
|
||||
+ tzcntl %eax, %eax
|
||||
+ ret
|
||||
# ifdef USE_AS_STRNLEN
|
||||
- /* Adjust length. */
|
||||
- addq %rcx, %rsi
|
||||
+L(zero):
|
||||
+ xorl %eax, %eax
|
||||
+ ret
|
||||
|
||||
- subq $(VEC_SIZE * 4), %rsi
|
||||
- jbe L(last_4x_vec_or_less)
|
||||
+ .p2align 4
|
||||
+L(first_vec_x0):
|
||||
+ /* Set bit for max len so that tzcnt will return min of max len
|
||||
+ and position of first match. */
|
||||
+ btsq %rsi, %rax
|
||||
+ tzcntl %eax, %eax
|
||||
+ ret
|
||||
# endif
|
||||
- jmp L(more_4x_vec)
|
||||
|
||||
.p2align 4
|
||||
-L(cros_page_boundary):
|
||||
- andl $(VEC_SIZE - 1), %ecx
|
||||
- andq $-VEC_SIZE, %rdi
|
||||
-
|
||||
-# ifdef USE_AS_WCSLEN
|
||||
- /* NB: Divide shift count by 4 since each bit in K0 represent 4
|
||||
- bytes. */
|
||||
- movl %ecx, %SHIFT_REG
|
||||
- sarl $2, %SHIFT_REG
|
||||
+L(first_vec_x1):
|
||||
+ tzcntl %eax, %eax
|
||||
+ /* Safe to use 32 bit instructions as these are only called for
|
||||
+ size = [1, 159]. */
|
||||
+# ifdef USE_AS_STRNLEN
|
||||
+ /* Use ecx which was computed earlier to compute correct value.
|
||||
+ */
|
||||
+ leal -(CHAR_PER_VEC * 4 + 1)(%rcx, %rax), %eax
|
||||
+# else
|
||||
+ subl %edx, %edi
|
||||
+# ifdef USE_AS_WCSLEN
|
||||
+ /* NB: Divide bytes by 4 to get the wchar_t count. */
|
||||
+ sarl $2, %edi
|
||||
+# endif
|
||||
+ leal CHAR_PER_VEC(%rdi, %rax), %eax
|
||||
# endif
|
||||
- VPCMP $0, (%rdi), %YMMZERO, %k0
|
||||
- kmovd %k0, %eax
|
||||
+ ret
|
||||
|
||||
- /* Remove the leading bytes. */
|
||||
- sarxl %SHIFT_REG, %eax, %eax
|
||||
- testl %eax, %eax
|
||||
- jz L(aligned_more)
|
||||
+ .p2align 4
|
||||
+L(first_vec_x2):
|
||||
tzcntl %eax, %eax
|
||||
-# ifdef USE_AS_WCSLEN
|
||||
- /* NB: Multiply wchar_t count by 4 to get the number of bytes. */
|
||||
- sall $2, %eax
|
||||
-# endif
|
||||
+ /* Safe to use 32 bit instructions as these are only called for
|
||||
+ size = [1, 159]. */
|
||||
# ifdef USE_AS_STRNLEN
|
||||
- /* Check the end of data. */
|
||||
- cmpq %rax, %rsi
|
||||
- jbe L(max)
|
||||
-# endif
|
||||
- addq %rdi, %rax
|
||||
- addq %rcx, %rax
|
||||
- subq %rdx, %rax
|
||||
-# ifdef USE_AS_WCSLEN
|
||||
- shrq $2, %rax
|
||||
+ /* Use ecx which was computed earlier to compute correct value.
|
||||
+ */
|
||||
+ leal -(CHAR_PER_VEC * 3 + 1)(%rcx, %rax), %eax
|
||||
+# else
|
||||
+ subl %edx, %edi
|
||||
+# ifdef USE_AS_WCSLEN
|
||||
+ /* NB: Divide bytes by 4 to get the wchar_t count. */
|
||||
+ sarl $2, %edi
|
||||
+# endif
|
||||
+ leal (CHAR_PER_VEC * 2)(%rdi, %rax), %eax
|
||||
# endif
|
||||
ret
|
||||
|
||||
.p2align 4
|
||||
-L(aligned_more):
|
||||
+L(first_vec_x3):
|
||||
+ tzcntl %eax, %eax
|
||||
+ /* Safe to use 32 bit instructions as these are only called for
|
||||
+ size = [1, 159]. */
|
||||
# ifdef USE_AS_STRNLEN
|
||||
- /* "rcx" is less than VEC_SIZE. Calculate "rdx + rcx - VEC_SIZE"
|
||||
- with "rdx - (VEC_SIZE - rcx)" instead of "(rdx + rcx) - VEC_SIZE"
|
||||
- to void possible addition overflow. */
|
||||
- negq %rcx
|
||||
- addq $VEC_SIZE, %rcx
|
||||
-
|
||||
- /* Check the end of data. */
|
||||
- subq %rcx, %rsi
|
||||
- jbe L(max)
|
||||
+ /* Use ecx which was computed earlier to compute correct value.
|
||||
+ */
|
||||
+ leal -(CHAR_PER_VEC * 2 + 1)(%rcx, %rax), %eax
|
||||
+# else
|
||||
+ subl %edx, %edi
|
||||
+# ifdef USE_AS_WCSLEN
|
||||
+ /* NB: Divide bytes by 4 to get the wchar_t count. */
|
||||
+ sarl $2, %edi
|
||||
+# endif
|
||||
+ leal (CHAR_PER_VEC * 3)(%rdi, %rax), %eax
|
||||
# endif
|
||||
+ ret
|
||||
|
||||
- addq $VEC_SIZE, %rdi
|
||||
-
|
||||
+ .p2align 4
|
||||
+L(first_vec_x4):
|
||||
+ tzcntl %eax, %eax
|
||||
+ /* Safe to use 32 bit instructions as these are only called for
|
||||
+ size = [1, 159]. */
|
||||
# ifdef USE_AS_STRNLEN
|
||||
- subq $(VEC_SIZE * 4), %rsi
|
||||
- jbe L(last_4x_vec_or_less)
|
||||
+ /* Use ecx which was computed earlier to compute correct value.
|
||||
+ */
|
||||
+ leal -(CHAR_PER_VEC + 1)(%rcx, %rax), %eax
|
||||
+# else
|
||||
+ subl %edx, %edi
|
||||
+# ifdef USE_AS_WCSLEN
|
||||
+ /* NB: Divide bytes by 4 to get the wchar_t count. */
|
||||
+ sarl $2, %edi
|
||||
+# endif
|
||||
+ leal (CHAR_PER_VEC * 4)(%rdi, %rax), %eax
|
||||
# endif
|
||||
+ ret
|
||||
|
||||
-L(more_4x_vec):
|
||||
+ .p2align 5
|
||||
+L(aligned_more):
|
||||
+ movq %rdi, %rdx
|
||||
+ /* Align data to VEC_SIZE. */
|
||||
+ andq $-(VEC_SIZE), %rdi
|
||||
+L(cross_page_continue):
|
||||
/* Check the first 4 * VEC_SIZE. Only one VEC_SIZE at a time
|
||||
since data is only aligned to VEC_SIZE. */
|
||||
- VPCMP $0, (%rdi), %YMMZERO, %k0
|
||||
- kmovd %k0, %eax
|
||||
- testl %eax, %eax
|
||||
- jnz L(first_vec_x0)
|
||||
-
|
||||
+# ifdef USE_AS_STRNLEN
|
||||
+ /* + CHAR_SIZE because it simplies the logic in
|
||||
+ last_4x_vec_or_less. */
|
||||
+ leaq (VEC_SIZE * 5 + CHAR_SIZE)(%rdi), %rcx
|
||||
+ subq %rdx, %rcx
|
||||
+# ifdef USE_AS_WCSLEN
|
||||
+ /* NB: Divide bytes by 4 to get the wchar_t count. */
|
||||
+ sarl $2, %ecx
|
||||
+# endif
|
||||
+# endif
|
||||
+ /* Load first VEC regardless. */
|
||||
VPCMP $0, VEC_SIZE(%rdi), %YMMZERO, %k0
|
||||
+# ifdef USE_AS_STRNLEN
|
||||
+ /* Adjust length. If near end handle specially. */
|
||||
+ subq %rcx, %rsi
|
||||
+ jb L(last_4x_vec_or_less)
|
||||
+# endif
|
||||
kmovd %k0, %eax
|
||||
testl %eax, %eax
|
||||
jnz L(first_vec_x1)
|
||||
|
||||
VPCMP $0, (VEC_SIZE * 2)(%rdi), %YMMZERO, %k0
|
||||
kmovd %k0, %eax
|
||||
- testl %eax, %eax
|
||||
+ test %eax, %eax
|
||||
jnz L(first_vec_x2)
|
||||
|
||||
VPCMP $0, (VEC_SIZE * 3)(%rdi), %YMMZERO, %k0
|
||||
@@ -179,258 +214,276 @@ L(more_4x_vec):
|
||||
testl %eax, %eax
|
||||
jnz L(first_vec_x3)
|
||||
|
||||
- addq $(VEC_SIZE * 4), %rdi
|
||||
-
|
||||
-# ifdef USE_AS_STRNLEN
|
||||
- subq $(VEC_SIZE * 4), %rsi
|
||||
- jbe L(last_4x_vec_or_less)
|
||||
-# endif
|
||||
-
|
||||
- /* Align data to 4 * VEC_SIZE. */
|
||||
- movq %rdi, %rcx
|
||||
- andl $(4 * VEC_SIZE - 1), %ecx
|
||||
- andq $-(4 * VEC_SIZE), %rdi
|
||||
+ VPCMP $0, (VEC_SIZE * 4)(%rdi), %YMMZERO, %k0
|
||||
+ kmovd %k0, %eax
|
||||
+ testl %eax, %eax
|
||||
+ jnz L(first_vec_x4)
|
||||
|
||||
+ addq $VEC_SIZE, %rdi
|
||||
# ifdef USE_AS_STRNLEN
|
||||
- /* Adjust length. */
|
||||
+ /* Check if at last VEC_SIZE * 4 length. */
|
||||
+ cmpq $(CHAR_PER_VEC * 4 - 1), %rsi
|
||||
+ jbe L(last_4x_vec_or_less_load)
|
||||
+ movl %edi, %ecx
|
||||
+ andl $(VEC_SIZE * 4 - 1), %ecx
|
||||
+# ifdef USE_AS_WCSLEN
|
||||
+ /* NB: Divide bytes by 4 to get the wchar_t count. */
|
||||
+ sarl $2, %ecx
|
||||
+# endif
|
||||
+ /* Readjust length. */
|
||||
addq %rcx, %rsi
|
||||
# endif
|
||||
+ /* Align data to VEC_SIZE * 4. */
|
||||
+ andq $-(VEC_SIZE * 4), %rdi
|
||||
|
||||
+ /* Compare 4 * VEC at a time forward. */
|
||||
.p2align 4
|
||||
L(loop_4x_vec):
|
||||
- /* Compare 4 * VEC at a time forward. */
|
||||
- VMOVA (%rdi), %YMM1
|
||||
- VMOVA VEC_SIZE(%rdi), %YMM2
|
||||
- VMOVA (VEC_SIZE * 2)(%rdi), %YMM3
|
||||
- VMOVA (VEC_SIZE * 3)(%rdi), %YMM4
|
||||
-
|
||||
- VPMINU %YMM1, %YMM2, %YMM5
|
||||
- VPMINU %YMM3, %YMM4, %YMM6
|
||||
+ /* Load first VEC regardless. */
|
||||
+ VMOVA (VEC_SIZE * 4)(%rdi), %YMM1
|
||||
+# ifdef USE_AS_STRNLEN
|
||||
+ /* Break if at end of length. */
|
||||
+ subq $(CHAR_PER_VEC * 4), %rsi
|
||||
+ jb L(last_4x_vec_or_less_cmpeq)
|
||||
+# endif
|
||||
+ /* Save some code size by microfusing VPMINU with the load. Since
|
||||
+ the matches in ymm2/ymm4 can only be returned if there where no
|
||||
+ matches in ymm1/ymm3 respectively there is no issue with overlap.
|
||||
+ */
|
||||
+ VPMINU (VEC_SIZE * 5)(%rdi), %YMM1, %YMM2
|
||||
+ VMOVA (VEC_SIZE * 6)(%rdi), %YMM3
|
||||
+ VPMINU (VEC_SIZE * 7)(%rdi), %YMM3, %YMM4
|
||||
+
|
||||
+ VPCMP $0, %YMM2, %YMMZERO, %k0
|
||||
+ VPCMP $0, %YMM4, %YMMZERO, %k1
|
||||
+ subq $-(VEC_SIZE * 4), %rdi
|
||||
+ kortestd %k0, %k1
|
||||
+ jz L(loop_4x_vec)
|
||||
+
|
||||
+ /* Check if end was in first half. */
|
||||
+ kmovd %k0, %eax
|
||||
+ subq %rdx, %rdi
|
||||
+# ifdef USE_AS_WCSLEN
|
||||
+ shrq $2, %rdi
|
||||
+# endif
|
||||
+ testl %eax, %eax
|
||||
+ jz L(second_vec_return)
|
||||
|
||||
- VPMINU %YMM5, %YMM6, %YMM5
|
||||
- VPCMP $0, %YMM5, %YMMZERO, %k0
|
||||
- ktestd %k0, %k0
|
||||
- jnz L(4x_vec_end)
|
||||
+ VPCMP $0, %YMM1, %YMMZERO, %k2
|
||||
+ kmovd %k2, %edx
|
||||
+ /* Combine VEC1 matches (edx) with VEC2 matches (eax). */
|
||||
+# ifdef USE_AS_WCSLEN
|
||||
+ sall $CHAR_PER_VEC, %eax
|
||||
+ orl %edx, %eax
|
||||
+ tzcntl %eax, %eax
|
||||
+# else
|
||||
+ salq $CHAR_PER_VEC, %rax
|
||||
+ orq %rdx, %rax
|
||||
+ tzcntq %rax, %rax
|
||||
+# endif
|
||||
+ addq %rdi, %rax
|
||||
+ ret
|
||||
|
||||
- addq $(VEC_SIZE * 4), %rdi
|
||||
|
||||
-# ifndef USE_AS_STRNLEN
|
||||
- jmp L(loop_4x_vec)
|
||||
-# else
|
||||
- subq $(VEC_SIZE * 4), %rsi
|
||||
- ja L(loop_4x_vec)
|
||||
+# ifdef USE_AS_STRNLEN
|
||||
|
||||
+L(last_4x_vec_or_less_load):
|
||||
+ /* Depending on entry adjust rdi / prepare first VEC in YMM1. */
|
||||
+ VMOVA (VEC_SIZE * 4)(%rdi), %YMM1
|
||||
+L(last_4x_vec_or_less_cmpeq):
|
||||
+ VPCMP $0, %YMM1, %YMMZERO, %k0
|
||||
+ addq $(VEC_SIZE * 3), %rdi
|
||||
L(last_4x_vec_or_less):
|
||||
- /* Less than 4 * VEC and aligned to VEC_SIZE. */
|
||||
- addl $(VEC_SIZE * 2), %esi
|
||||
- jle L(last_2x_vec)
|
||||
-
|
||||
- VPCMP $0, (%rdi), %YMMZERO, %k0
|
||||
kmovd %k0, %eax
|
||||
+ /* If remaining length > VEC_SIZE * 2. This works if esi is off by
|
||||
+ VEC_SIZE * 4. */
|
||||
+ testl $(CHAR_PER_VEC * 2), %esi
|
||||
+ jnz L(last_4x_vec)
|
||||
+
|
||||
+ /* length may have been negative or positive by an offset of
|
||||
+ CHAR_PER_VEC * 4 depending on where this was called from. This
|
||||
+ fixes that. */
|
||||
+ andl $(CHAR_PER_VEC * 4 - 1), %esi
|
||||
testl %eax, %eax
|
||||
- jnz L(first_vec_x0)
|
||||
+ jnz L(last_vec_x1_check)
|
||||
|
||||
- VPCMP $0, VEC_SIZE(%rdi), %YMMZERO, %k0
|
||||
- kmovd %k0, %eax
|
||||
- testl %eax, %eax
|
||||
- jnz L(first_vec_x1)
|
||||
+ /* Check the end of data. */
|
||||
+ subl $CHAR_PER_VEC, %esi
|
||||
+ jb L(max)
|
||||
|
||||
VPCMP $0, (VEC_SIZE * 2)(%rdi), %YMMZERO, %k0
|
||||
kmovd %k0, %eax
|
||||
- testl %eax, %eax
|
||||
- jnz L(first_vec_x2_check)
|
||||
- subl $VEC_SIZE, %esi
|
||||
- jle L(max)
|
||||
+ tzcntl %eax, %eax
|
||||
+ /* Check the end of data. */
|
||||
+ cmpl %eax, %esi
|
||||
+ jb L(max)
|
||||
|
||||
- VPCMP $0, (VEC_SIZE * 3)(%rdi), %YMMZERO, %k0
|
||||
- kmovd %k0, %eax
|
||||
- testl %eax, %eax
|
||||
- jnz L(first_vec_x3_check)
|
||||
+ subq %rdx, %rdi
|
||||
+# ifdef USE_AS_WCSLEN
|
||||
+ /* NB: Divide bytes by 4 to get the wchar_t count. */
|
||||
+ sarq $2, %rdi
|
||||
+# endif
|
||||
+ leaq (CHAR_PER_VEC * 2)(%rdi, %rax), %rax
|
||||
+ ret
|
||||
+L(max):
|
||||
movq %r8, %rax
|
||||
+ ret
|
||||
+# endif
|
||||
+
|
||||
+ /* Placed here in strnlen so that the jcc L(last_4x_vec_or_less)
|
||||
+ in the 4x VEC loop can use 2 byte encoding. */
|
||||
+ .p2align 4
|
||||
+L(second_vec_return):
|
||||
+ VPCMP $0, %YMM3, %YMMZERO, %k0
|
||||
+ /* Combine YMM3 matches (k0) with YMM4 matches (k1). */
|
||||
+# ifdef USE_AS_WCSLEN
|
||||
+ kunpckbw %k0, %k1, %k0
|
||||
+ kmovd %k0, %eax
|
||||
+ tzcntl %eax, %eax
|
||||
+# else
|
||||
+ kunpckdq %k0, %k1, %k0
|
||||
+ kmovq %k0, %rax
|
||||
+ tzcntq %rax, %rax
|
||||
+# endif
|
||||
+ leaq (CHAR_PER_VEC * 2)(%rdi, %rax), %rax
|
||||
+ ret
|
||||
+
|
||||
+
|
||||
+# ifdef USE_AS_STRNLEN
|
||||
+L(last_vec_x1_check):
|
||||
+ tzcntl %eax, %eax
|
||||
+ /* Check the end of data. */
|
||||
+ cmpl %eax, %esi
|
||||
+ jb L(max)
|
||||
+ subq %rdx, %rdi
|
||||
# ifdef USE_AS_WCSLEN
|
||||
- shrq $2, %rax
|
||||
+ /* NB: Divide bytes by 4 to get the wchar_t count. */
|
||||
+ sarq $2, %rdi
|
||||
# endif
|
||||
+ leaq (CHAR_PER_VEC)(%rdi, %rax), %rax
|
||||
ret
|
||||
|
||||
.p2align 4
|
||||
-L(last_2x_vec):
|
||||
- addl $(VEC_SIZE * 2), %esi
|
||||
+L(last_4x_vec):
|
||||
+ /* Test first 2x VEC normally. */
|
||||
+ testl %eax, %eax
|
||||
+ jnz L(last_vec_x1)
|
||||
|
||||
- VPCMP $0, (%rdi), %YMMZERO, %k0
|
||||
+ VPCMP $0, (VEC_SIZE * 2)(%rdi), %YMMZERO, %k0
|
||||
kmovd %k0, %eax
|
||||
testl %eax, %eax
|
||||
- jnz L(first_vec_x0_check)
|
||||
- subl $VEC_SIZE, %esi
|
||||
- jle L(max)
|
||||
+ jnz L(last_vec_x2)
|
||||
|
||||
- VPCMP $0, VEC_SIZE(%rdi), %YMMZERO, %k0
|
||||
+ /* Normalize length. */
|
||||
+ andl $(CHAR_PER_VEC * 4 - 1), %esi
|
||||
+ VPCMP $0, (VEC_SIZE * 3)(%rdi), %YMMZERO, %k0
|
||||
kmovd %k0, %eax
|
||||
testl %eax, %eax
|
||||
- jnz L(first_vec_x1_check)
|
||||
- movq %r8, %rax
|
||||
-# ifdef USE_AS_WCSLEN
|
||||
- shrq $2, %rax
|
||||
-# endif
|
||||
- ret
|
||||
+ jnz L(last_vec_x3)
|
||||
|
||||
- .p2align 4
|
||||
-L(first_vec_x0_check):
|
||||
+ /* Check the end of data. */
|
||||
+ subl $(CHAR_PER_VEC * 3), %esi
|
||||
+ jb L(max)
|
||||
+
|
||||
+ VPCMP $0, (VEC_SIZE * 4)(%rdi), %YMMZERO, %k0
|
||||
+ kmovd %k0, %eax
|
||||
tzcntl %eax, %eax
|
||||
-# ifdef USE_AS_WCSLEN
|
||||
- /* NB: Multiply wchar_t count by 4 to get the number of bytes. */
|
||||
- sall $2, %eax
|
||||
-# endif
|
||||
/* Check the end of data. */
|
||||
- cmpq %rax, %rsi
|
||||
- jbe L(max)
|
||||
- addq %rdi, %rax
|
||||
- subq %rdx, %rax
|
||||
+ cmpl %eax, %esi
|
||||
+ jb L(max_end)
|
||||
+
|
||||
+ subq %rdx, %rdi
|
||||
# ifdef USE_AS_WCSLEN
|
||||
- shrq $2, %rax
|
||||
+ /* NB: Divide bytes by 4 to get the wchar_t count. */
|
||||
+ sarq $2, %rdi
|
||||
# endif
|
||||
+ leaq (CHAR_PER_VEC * 4)(%rdi, %rax), %rax
|
||||
ret
|
||||
|
||||
.p2align 4
|
||||
-L(first_vec_x1_check):
|
||||
+L(last_vec_x1):
|
||||
tzcntl %eax, %eax
|
||||
+ subq %rdx, %rdi
|
||||
# ifdef USE_AS_WCSLEN
|
||||
- /* NB: Multiply wchar_t count by 4 to get the number of bytes. */
|
||||
- sall $2, %eax
|
||||
-# endif
|
||||
- /* Check the end of data. */
|
||||
- cmpq %rax, %rsi
|
||||
- jbe L(max)
|
||||
- addq $VEC_SIZE, %rax
|
||||
- addq %rdi, %rax
|
||||
- subq %rdx, %rax
|
||||
-# ifdef USE_AS_WCSLEN
|
||||
- shrq $2, %rax
|
||||
+ /* NB: Divide bytes by 4 to get the wchar_t count. */
|
||||
+ sarq $2, %rdi
|
||||
# endif
|
||||
+ leaq (CHAR_PER_VEC)(%rdi, %rax), %rax
|
||||
ret
|
||||
|
||||
.p2align 4
|
||||
-L(first_vec_x2_check):
|
||||
+L(last_vec_x2):
|
||||
tzcntl %eax, %eax
|
||||
+ subq %rdx, %rdi
|
||||
# ifdef USE_AS_WCSLEN
|
||||
- /* NB: Multiply wchar_t count by 4 to get the number of bytes. */
|
||||
- sall $2, %eax
|
||||
-# endif
|
||||
- /* Check the end of data. */
|
||||
- cmpq %rax, %rsi
|
||||
- jbe L(max)
|
||||
- addq $(VEC_SIZE * 2), %rax
|
||||
- addq %rdi, %rax
|
||||
- subq %rdx, %rax
|
||||
-# ifdef USE_AS_WCSLEN
|
||||
- shrq $2, %rax
|
||||
+ /* NB: Divide bytes by 4 to get the wchar_t count. */
|
||||
+ sarq $2, %rdi
|
||||
# endif
|
||||
+ leaq (CHAR_PER_VEC * 2)(%rdi, %rax), %rax
|
||||
ret
|
||||
|
||||
.p2align 4
|
||||
-L(first_vec_x3_check):
|
||||
+L(last_vec_x3):
|
||||
tzcntl %eax, %eax
|
||||
-# ifdef USE_AS_WCSLEN
|
||||
- /* NB: Multiply wchar_t count by 4 to get the number of bytes. */
|
||||
- sall $2, %eax
|
||||
-# endif
|
||||
+ subl $(CHAR_PER_VEC * 2), %esi
|
||||
/* Check the end of data. */
|
||||
- cmpq %rax, %rsi
|
||||
- jbe L(max)
|
||||
- addq $(VEC_SIZE * 3), %rax
|
||||
- addq %rdi, %rax
|
||||
- subq %rdx, %rax
|
||||
+ cmpl %eax, %esi
|
||||
+ jb L(max_end)
|
||||
+ subq %rdx, %rdi
|
||||
# ifdef USE_AS_WCSLEN
|
||||
- shrq $2, %rax
|
||||
+ /* NB: Divide bytes by 4 to get the wchar_t count. */
|
||||
+ sarq $2, %rdi
|
||||
# endif
|
||||
+ leaq (CHAR_PER_VEC * 3)(%rdi, %rax), %rax
|
||||
ret
|
||||
-
|
||||
- .p2align 4
|
||||
-L(max):
|
||||
+L(max_end):
|
||||
movq %r8, %rax
|
||||
-# ifdef USE_AS_WCSLEN
|
||||
- shrq $2, %rax
|
||||
-# endif
|
||||
- ret
|
||||
-
|
||||
- .p2align 4
|
||||
-L(zero):
|
||||
- xorl %eax, %eax
|
||||
ret
|
||||
# endif
|
||||
|
||||
+ /* Cold case for crossing page with first load. */
|
||||
.p2align 4
|
||||
-L(first_vec_x0):
|
||||
- tzcntl %eax, %eax
|
||||
-# ifdef USE_AS_WCSLEN
|
||||
- /* NB: Multiply wchar_t count by 4 to get the number of bytes. */
|
||||
- sall $2, %eax
|
||||
-# endif
|
||||
- addq %rdi, %rax
|
||||
- subq %rdx, %rax
|
||||
+L(cross_page_boundary):
|
||||
+ movq %rdi, %rdx
|
||||
+ /* Align data to VEC_SIZE. */
|
||||
+ andq $-VEC_SIZE, %rdi
|
||||
+ VPCMP $0, (%rdi), %YMMZERO, %k0
|
||||
+ kmovd %k0, %eax
|
||||
+ /* Remove the leading bytes. */
|
||||
# ifdef USE_AS_WCSLEN
|
||||
- shrq $2, %rax
|
||||
+ /* NB: Divide shift count by 4 since each bit in K0 represent 4
|
||||
+ bytes. */
|
||||
+ movl %edx, %ecx
|
||||
+ shrl $2, %ecx
|
||||
+ andl $(CHAR_PER_VEC - 1), %ecx
|
||||
# endif
|
||||
- ret
|
||||
-
|
||||
- .p2align 4
|
||||
-L(first_vec_x1):
|
||||
+ /* SHIFT_REG is ecx for USE_AS_WCSLEN and edx otherwise. */
|
||||
+ sarxl %SHIFT_REG, %eax, %eax
|
||||
+ testl %eax, %eax
|
||||
+# ifndef USE_AS_STRNLEN
|
||||
+ jz L(cross_page_continue)
|
||||
tzcntl %eax, %eax
|
||||
-# ifdef USE_AS_WCSLEN
|
||||
- /* NB: Multiply wchar_t count by 4 to get the number of bytes. */
|
||||
- sall $2, %eax
|
||||
-# endif
|
||||
- addq $VEC_SIZE, %rax
|
||||
- addq %rdi, %rax
|
||||
- subq %rdx, %rax
|
||||
-# ifdef USE_AS_WCSLEN
|
||||
- shrq $2, %rax
|
||||
-# endif
|
||||
ret
|
||||
-
|
||||
- .p2align 4
|
||||
-L(first_vec_x2):
|
||||
- tzcntl %eax, %eax
|
||||
-# ifdef USE_AS_WCSLEN
|
||||
- /* NB: Multiply wchar_t count by 4 to get the number of bytes. */
|
||||
- sall $2, %eax
|
||||
-# endif
|
||||
- addq $(VEC_SIZE * 2), %rax
|
||||
- addq %rdi, %rax
|
||||
- subq %rdx, %rax
|
||||
-# ifdef USE_AS_WCSLEN
|
||||
- shrq $2, %rax
|
||||
-# endif
|
||||
+# else
|
||||
+ jnz L(cross_page_less_vec)
|
||||
+# ifndef USE_AS_WCSLEN
|
||||
+ movl %edx, %ecx
|
||||
+ andl $(CHAR_PER_VEC - 1), %ecx
|
||||
+# endif
|
||||
+ movl $CHAR_PER_VEC, %eax
|
||||
+ subl %ecx, %eax
|
||||
+ /* Check the end of data. */
|
||||
+ cmpq %rax, %rsi
|
||||
+ ja L(cross_page_continue)
|
||||
+ movl %esi, %eax
|
||||
ret
|
||||
-
|
||||
- .p2align 4
|
||||
-L(4x_vec_end):
|
||||
- VPCMP $0, %YMM1, %YMMZERO, %k0
|
||||
- kmovd %k0, %eax
|
||||
- testl %eax, %eax
|
||||
- jnz L(first_vec_x0)
|
||||
- VPCMP $0, %YMM2, %YMMZERO, %k1
|
||||
- kmovd %k1, %eax
|
||||
- testl %eax, %eax
|
||||
- jnz L(first_vec_x1)
|
||||
- VPCMP $0, %YMM3, %YMMZERO, %k2
|
||||
- kmovd %k2, %eax
|
||||
- testl %eax, %eax
|
||||
- jnz L(first_vec_x2)
|
||||
- VPCMP $0, %YMM4, %YMMZERO, %k3
|
||||
- kmovd %k3, %eax
|
||||
-L(first_vec_x3):
|
||||
+L(cross_page_less_vec):
|
||||
tzcntl %eax, %eax
|
||||
-# ifdef USE_AS_WCSLEN
|
||||
- /* NB: Multiply wchar_t count by 4 to get the number of bytes. */
|
||||
- sall $2, %eax
|
||||
-# endif
|
||||
- addq $(VEC_SIZE * 3), %rax
|
||||
- addq %rdi, %rax
|
||||
- subq %rdx, %rax
|
||||
-# ifdef USE_AS_WCSLEN
|
||||
- shrq $2, %rax
|
||||
-# endif
|
||||
+ /* Select min of length and position of first null. */
|
||||
+ cmpq %rax, %rsi
|
||||
+ cmovb %esi, %eax
|
||||
ret
|
||||
+# endif
|
||||
|
||||
END (STRLEN)
|
||||
#endif
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,158 +0,0 @@
|
||||
From ea8e465a6b8d0f26c72bcbe453a854de3abf68ec Mon Sep 17 00:00:00 2001
|
||||
From: "H.J. Lu" <hjl.tools@gmail.com>
|
||||
Date: Wed, 30 Jun 2021 10:47:06 -0700
|
||||
Subject: [PATCH] x86: Check RTM_ALWAYS_ABORT for RTM [BZ #28033]
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
From
|
||||
|
||||
https://www.intel.com/content/www/us/en/support/articles/000059422/processors.html
|
||||
|
||||
* Intel TSX will be disabled by default.
|
||||
* The processor will force abort all Restricted Transactional Memory (RTM)
|
||||
transactions by default.
|
||||
* A new CPUID bit CPUID.07H.0H.EDX[11](RTM_ALWAYS_ABORT) will be enumerated,
|
||||
which is set to indicate to updated software that the loaded microcode is
|
||||
forcing RTM abort.
|
||||
* On processors that enumerate support for RTM, the CPUID enumeration bits
|
||||
for Intel TSX (CPUID.07H.0H.EBX[11] and CPUID.07H.0H.EBX[4]) continue to
|
||||
be set by default after microcode update.
|
||||
* Workloads that were benefited from Intel TSX might experience a change
|
||||
in performance.
|
||||
* System software may use a new bit in Model-Specific Register (MSR) 0x10F
|
||||
TSX_FORCE_ABORT[TSX_CPUID_CLEAR] functionality to clear the Hardware Lock
|
||||
Elision (HLE) and RTM bits to indicate to software that Intel TSX is
|
||||
disabled.
|
||||
|
||||
1. Add RTM_ALWAYS_ABORT to CPUID features.
|
||||
2. Set RTM usable only if RTM_ALWAYS_ABORT isn't set. This skips the
|
||||
string/tst-memchr-rtm etc. testcases on the affected processors, which
|
||||
always fail after a microcde update.
|
||||
3. Check RTM feature, instead of usability, against /proc/cpuinfo.
|
||||
|
||||
This fixes BZ #28033.
|
||||
---
|
||||
manual/platform.texi | 3 +++
|
||||
sysdeps/x86/cpu-features.c | 5 ++++-
|
||||
sysdeps/x86/sys/platform/x86.h | 6 +++---
|
||||
sysdeps/x86/tst-cpu-features-supports.c | 2 +-
|
||||
sysdeps/x86/tst-get-cpu-features.c | 2 ++
|
||||
5 files changed, 13 insertions(+), 5 deletions(-)
|
||||
|
||||
Conflicts:
|
||||
sysdeps/x86/bits/platform/x86.h
|
||||
(doesn't exist)
|
||||
sysdeps/x86/bits/platform/x86.h
|
||||
(account for lack of upstream renames)
|
||||
|
||||
diff --git a/manual/platform.texi b/manual/platform.texi
|
||||
index 8fec2933..b7e8aef7 100644
|
||||
--- a/manual/platform.texi
|
||||
+++ b/manual/platform.texi
|
||||
@@ -510,6 +510,9 @@ capability.
|
||||
@item
|
||||
@code{RTM} -- RTM instruction extensions.
|
||||
|
||||
+@item
|
||||
+@code{RTM_ALWAYS_ABORT} -- Transactions always abort, making RTM unusable.
|
||||
+
|
||||
@item
|
||||
@code{SDBG} -- IA32_DEBUG_INTERFACE MSR for silicon debug.
|
||||
|
||||
diff --git a/sysdeps/x86/cpu-features.c b/sysdeps/x86/cpu-features.c
|
||||
index 3610ee5c..4889f062 100644
|
||||
--- a/sysdeps/x86/cpu-features.c
|
||||
+++ b/sysdeps/x86/cpu-features.c
|
||||
@@ -74,7 +74,6 @@ update_usable (struct cpu_features *cpu_features)
|
||||
CPU_FEATURE_SET_USABLE (cpu_features, HLE);
|
||||
CPU_FEATURE_SET_USABLE (cpu_features, BMI2);
|
||||
CPU_FEATURE_SET_USABLE (cpu_features, ERMS);
|
||||
- CPU_FEATURE_SET_USABLE (cpu_features, RTM);
|
||||
CPU_FEATURE_SET_USABLE (cpu_features, RDSEED);
|
||||
CPU_FEATURE_SET_USABLE (cpu_features, ADX);
|
||||
CPU_FEATURE_SET_USABLE (cpu_features, CLFLUSHOPT);
|
||||
@@ -90,6 +89,7 @@ update_usable (struct cpu_features *cpu_features)
|
||||
CPU_FEATURE_SET_USABLE (cpu_features, MOVDIRI);
|
||||
CPU_FEATURE_SET_USABLE (cpu_features, MOVDIR64B);
|
||||
CPU_FEATURE_SET_USABLE (cpu_features, FSRM);
|
||||
+ CPU_FEATURE_SET_USABLE (cpu_features, RTM_ALWAYS_ABORT);
|
||||
CPU_FEATURE_SET_USABLE (cpu_features, SERIALIZE);
|
||||
CPU_FEATURE_SET_USABLE (cpu_features, TSXLDTRK);
|
||||
CPU_FEATURE_SET_USABLE (cpu_features, LAHF64_SAHF64);
|
||||
@@ -779,6 +779,9 @@ no_cpuid:
|
||||
GLRO(dl_platform) = "i586";
|
||||
#endif
|
||||
|
||||
+ if (!CPU_FEATURES_CPU_P (cpu_features, RTM_ALWAYS_ABORT))
|
||||
+ CPU_FEATURE_SET_USABLE (cpu_features, RTM);
|
||||
+
|
||||
#if CET_ENABLED
|
||||
# if HAVE_TUNABLES
|
||||
TUNABLE_GET (x86_ibt, tunable_val_t *,
|
||||
diff --git a/sysdeps/x86/sys/platform/x86.h b/sysdeps/x86/sys/platform/x86.h
|
||||
index e5cc7c68..7a434926 100644
|
||||
--- a/sysdeps/x86/sys/platform/x86.h
|
||||
+++ b/sysdeps/x86/sys/platform/x86.h
|
||||
@@ -247,7 +247,7 @@ extern const struct cpu_features *__x86_get_cpu_features (unsigned int)
|
||||
#define bit_cpu_AVX512_VP2INTERSECT (1u << 8)
|
||||
#define bit_cpu_INDEX_7_EDX_9 (1u << 9)
|
||||
#define bit_cpu_MD_CLEAR (1u << 10)
|
||||
-#define bit_cpu_INDEX_7_EDX_11 (1u << 11)
|
||||
+#define bit_cpu_RTM_ALWAYS_ABORT (1u << 11)
|
||||
#define bit_cpu_INDEX_7_EDX_12 (1u << 12)
|
||||
#define bit_cpu_INDEX_7_EDX_13 (1u << 13)
|
||||
#define bit_cpu_SERIALIZE (1u << 14)
|
||||
@@ -471,7 +471,7 @@ extern const struct cpu_features *__x86_get_cpu_features (unsigned int)
|
||||
#define index_cpu_AVX512_VP2INTERSECT COMMON_CPUID_INDEX_7
|
||||
#define index_cpu_INDEX_7_EDX_9 COMMON_CPUID_INDEX_7
|
||||
#define index_cpu_MD_CLEAR COMMON_CPUID_INDEX_7
|
||||
-#define index_cpu_INDEX_7_EDX_11 COMMON_CPUID_INDEX_7
|
||||
+#define index_cpu_RTM_ALWAYS_ABORT COMMON_CPUID_INDEX_7
|
||||
#define index_cpu_INDEX_7_EDX_12 COMMON_CPUID_INDEX_7
|
||||
#define index_cpu_INDEX_7_EDX_13 COMMON_CPUID_INDEX_7
|
||||
#define index_cpu_SERIALIZE COMMON_CPUID_INDEX_7
|
||||
@@ -695,7 +695,7 @@ extern const struct cpu_features *__x86_get_cpu_features (unsigned int)
|
||||
#define reg_AVX512_VP2INTERSECT edx
|
||||
#define reg_INDEX_7_EDX_9 edx
|
||||
#define reg_MD_CLEAR edx
|
||||
-#define reg_INDEX_7_EDX_11 edx
|
||||
+#define reg_RTM_ALWAYS_ABORT edx
|
||||
#define reg_INDEX_7_EDX_12 edx
|
||||
#define reg_INDEX_7_EDX_13 edx
|
||||
#define reg_SERIALIZE edx
|
||||
diff --git a/sysdeps/x86/tst-cpu-features-supports.c b/sysdeps/x86/tst-cpu-features-supports.c
|
||||
index 287cf01f..8100a319 100644
|
||||
--- a/sysdeps/x86/tst-cpu-features-supports.c
|
||||
+++ b/sysdeps/x86/tst-cpu-features-supports.c
|
||||
@@ -152,7 +152,7 @@ do_test (int argc, char **argv)
|
||||
fails += CHECK_SUPPORTS (rdpid, RDPID);
|
||||
fails += CHECK_SUPPORTS (rdrnd, RDRAND);
|
||||
fails += CHECK_SUPPORTS (rdseed, RDSEED);
|
||||
- fails += CHECK_SUPPORTS (rtm, RTM);
|
||||
+ fails += CHECK_CPU_SUPPORTS (rtm, RTM);
|
||||
fails += CHECK_SUPPORTS (serialize, SERIALIZE);
|
||||
fails += CHECK_SUPPORTS (sha, SHA);
|
||||
fails += CHECK_CPU_SUPPORTS (shstk, SHSTK);
|
||||
diff --git a/sysdeps/x86/tst-get-cpu-features.c b/sysdeps/x86/tst-get-cpu-features.c
|
||||
index 2763deb6..0717e5d8 100644
|
||||
--- a/sysdeps/x86/tst-get-cpu-features.c
|
||||
+++ b/sysdeps/x86/tst-get-cpu-features.c
|
||||
@@ -183,6 +183,7 @@ do_test (void)
|
||||
CHECK_CPU_FEATURE (UINTR);
|
||||
CHECK_CPU_FEATURE (AVX512_VP2INTERSECT);
|
||||
CHECK_CPU_FEATURE (MD_CLEAR);
|
||||
+ CHECK_CPU_FEATURE (RTM_ALWAYS_ABORT);
|
||||
CHECK_CPU_FEATURE (SERIALIZE);
|
||||
CHECK_CPU_FEATURE (HYBRID);
|
||||
CHECK_CPU_FEATURE (TSXLDTRK);
|
||||
@@ -344,6 +345,7 @@ do_test (void)
|
||||
CHECK_CPU_FEATURE_USABLE (FSRM);
|
||||
CHECK_CPU_FEATURE_USABLE (AVX512_VP2INTERSECT);
|
||||
CHECK_CPU_FEATURE_USABLE (MD_CLEAR);
|
||||
+ CHECK_CPU_FEATURE_USABLE (RTM_ALWAYS_ABORT);
|
||||
CHECK_CPU_FEATURE_USABLE (SERIALIZE);
|
||||
CHECK_CPU_FEATURE_USABLE (HYBRID);
|
||||
CHECK_CPU_FEATURE_USABLE (TSXLDTRK);
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,51 +0,0 @@
|
||||
From 0679442defedf7e52a94264975880ab8674736b2 Mon Sep 17 00:00:00 2001
|
||||
From: Noah Goldstein <goldstein.w.n@gmail.com>
|
||||
Date: Thu, 8 Jul 2021 16:13:19 -0400
|
||||
Subject: [PATCH] x86: Remove wcsnlen-sse4_1 from wcslen ifunc-impl-list [BZ
|
||||
#28064]
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
The following commit
|
||||
|
||||
commit 6f573a27b6c8b4236445810a44660612323f5a73
|
||||
Author: Noah Goldstein <goldstein.w.n@gmail.com>
|
||||
Date: Wed Jun 23 01:19:34 2021 -0400
|
||||
|
||||
x86-64: Add wcslen optimize for sse4.1
|
||||
|
||||
Added wcsnlen-sse4.1 to the wcslen ifunc implementation list and did
|
||||
not add wcslen-sse4.1 to wcslen ifunc implementation list. This commit
|
||||
fixes that by removing wcsnlen-sse4.1 from the wcslen ifunc
|
||||
implementation list and adding wcslen-sse4.1 to the ifunc
|
||||
implementation list.
|
||||
|
||||
Testing:
|
||||
test-wcslen.c, test-rsi-wcslen.c, and test-rsi-strlen.c are passing as
|
||||
well as all other tests in wcsmbs and string.
|
||||
|
||||
Signed-off-by: Noah Goldstein <goldstein.w.n@gmail.com>
|
||||
Reviewed-by: H.J. Lu <hjl.tools@gmail.com>
|
||||
Reviewed-by: H.J. Lu <hjl.tools@gmail.com>
|
||||
---
|
||||
sysdeps/x86_64/multiarch/ifunc-impl-list.c | 4 ++--
|
||||
1 file changed, 2 insertions(+), 2 deletions(-)
|
||||
|
||||
diff --git a/sysdeps/x86_64/multiarch/ifunc-impl-list.c b/sysdeps/x86_64/multiarch/ifunc-impl-list.c
|
||||
index 580913ca..695cdba6 100644
|
||||
--- a/sysdeps/x86_64/multiarch/ifunc-impl-list.c
|
||||
+++ b/sysdeps/x86_64/multiarch/ifunc-impl-list.c
|
||||
@@ -657,9 +657,9 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
|
||||
&& CPU_FEATURE_USABLE (AVX512BW)
|
||||
&& CPU_FEATURE_USABLE (BMI2)),
|
||||
__wcslen_evex)
|
||||
- IFUNC_IMPL_ADD (array, i, wcsnlen,
|
||||
+ IFUNC_IMPL_ADD (array, i, wcslen,
|
||||
CPU_FEATURE_USABLE (SSE4_1),
|
||||
- __wcsnlen_sse4_1)
|
||||
+ __wcslen_sse4_1)
|
||||
IFUNC_IMPL_ADD (array, i, wcslen, 1, __wcslen_sse2))
|
||||
|
||||
/* Support sysdeps/x86_64/multiarch/wcsnlen.c. */
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,51 +0,0 @@
|
||||
From 55c7bcc71b84123d5d4bd2814366a6b05fcf8ebd Mon Sep 17 00:00:00 2001
|
||||
From: "H.J. Lu" <hjl.tools@gmail.com>
|
||||
Date: Sat, 9 May 2020 12:04:23 -0700
|
||||
Subject: [PATCH] x86-64: Use RDX_LP on __x86_shared_non_temporal_threshold [BZ
|
||||
#25966]
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
Since __x86_shared_non_temporal_threshold is defined as
|
||||
|
||||
long int __x86_shared_non_temporal_threshold;
|
||||
|
||||
and long int is 4 bytes for x32, use RDX_LP to compare against
|
||||
__x86_shared_non_temporal_threshold in assembly code.
|
||||
---
|
||||
sysdeps/x86_64/multiarch/memmove-vec-unaligned-erms.S | 6 +++---
|
||||
1 file changed, 3 insertions(+), 3 deletions(-)
|
||||
|
||||
diff --git a/sysdeps/x86_64/multiarch/memmove-vec-unaligned-erms.S b/sysdeps/x86_64/multiarch/memmove-vec-unaligned-erms.S
|
||||
index 71f5954d..673b73aa 100644
|
||||
--- a/sysdeps/x86_64/multiarch/memmove-vec-unaligned-erms.S
|
||||
+++ b/sysdeps/x86_64/multiarch/memmove-vec-unaligned-erms.S
|
||||
@@ -245,7 +245,7 @@ L(return):
|
||||
#endif
|
||||
|
||||
L(movsb):
|
||||
- cmpq __x86_shared_non_temporal_threshold(%rip), %rdx
|
||||
+ cmp __x86_shared_non_temporal_threshold(%rip), %RDX_LP
|
||||
jae L(more_8x_vec)
|
||||
cmpq %rsi, %rdi
|
||||
jb 1f
|
||||
@@ -397,7 +397,7 @@ L(more_8x_vec):
|
||||
addq %r8, %rdx
|
||||
#if (defined USE_MULTIARCH || VEC_SIZE == 16) && IS_IN (libc)
|
||||
/* Check non-temporal store threshold. */
|
||||
- cmpq __x86_shared_non_temporal_threshold(%rip), %rdx
|
||||
+ cmp __x86_shared_non_temporal_threshold(%rip), %RDX_LP
|
||||
ja L(large_forward)
|
||||
#endif
|
||||
L(loop_4x_vec_forward):
|
||||
@@ -448,7 +448,7 @@ L(more_8x_vec_backward):
|
||||
subq %r8, %rdx
|
||||
#if (defined USE_MULTIARCH || VEC_SIZE == 16) && IS_IN (libc)
|
||||
/* Check non-temporal store threshold. */
|
||||
- cmpq __x86_shared_non_temporal_threshold(%rip), %rdx
|
||||
+ cmp __x86_shared_non_temporal_threshold(%rip), %RDX_LP
|
||||
ja L(large_backward)
|
||||
#endif
|
||||
L(loop_4x_vec_backward):
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,44 +0,0 @@
|
||||
From a35a59036ebae3efcdf5e8167610e0656fca9770 Mon Sep 17 00:00:00 2001
|
||||
From: "H.J. Lu" <hjl.tools@gmail.com>
|
||||
Date: Thu, 11 Jun 2020 12:41:18 -0700
|
||||
Subject: [PATCH] x86_64: Use %xmmN with vpxor to clear a vector register
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
Since "vpxor %xmmN, %xmmN, %xmmN" clears the whole vector register, use
|
||||
%xmmN, instead of %ymmN, with vpxor to clear a vector register.
|
||||
---
|
||||
sysdeps/x86_64/multiarch/strcmp-avx2.S | 4 ++--
|
||||
sysdeps/x86_64/multiarch/strrchr-avx2.S | 2 +-
|
||||
2 files changed, 3 insertions(+), 3 deletions(-)
|
||||
|
||||
diff --git a/sysdeps/x86_64/multiarch/strcmp-avx2.S b/sysdeps/x86_64/multiarch/strcmp-avx2.S
|
||||
index 433ae047..70d8499b 100644
|
||||
--- a/sysdeps/x86_64/multiarch/strcmp-avx2.S
|
||||
+++ b/sysdeps/x86_64/multiarch/strcmp-avx2.S
|
||||
@@ -105,8 +105,8 @@ ENTRY (STRCMP)
|
||||
# endif
|
||||
movl %edi, %eax
|
||||
xorl %edx, %edx
|
||||
- /* Make %ymm7 all zeros in this function. */
|
||||
- vpxor %ymm7, %ymm7, %ymm7
|
||||
+ /* Make %xmm7 (%ymm7) all zeros in this function. */
|
||||
+ vpxor %xmm7, %xmm7, %xmm7
|
||||
orl %esi, %eax
|
||||
andl $(PAGE_SIZE - 1), %eax
|
||||
cmpl $(PAGE_SIZE - (VEC_SIZE * 4)), %eax
|
||||
diff --git a/sysdeps/x86_64/multiarch/strrchr-avx2.S b/sysdeps/x86_64/multiarch/strrchr-avx2.S
|
||||
index 9f22a15e..c949410b 100644
|
||||
--- a/sysdeps/x86_64/multiarch/strrchr-avx2.S
|
||||
+++ b/sysdeps/x86_64/multiarch/strrchr-avx2.S
|
||||
@@ -48,7 +48,7 @@ ENTRY (STRRCHR)
|
||||
movl %edi, %ecx
|
||||
/* Broadcast CHAR to YMM4. */
|
||||
VPBROADCAST %xmm4, %ymm4
|
||||
- vpxor %ymm0, %ymm0, %ymm0
|
||||
+ vpxor %xmm0, %xmm0, %xmm0
|
||||
|
||||
/* Check if we may cross page boundary with one vector load. */
|
||||
andl $(2 * VEC_SIZE - 1), %ecx
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,359 +0,0 @@
|
||||
From 1f745ecc2109890886b161d4791e1406fdfc29b8 Mon Sep 17 00:00:00 2001
|
||||
From: noah <goldstein.w.n@gmail.com>
|
||||
Date: Wed, 3 Feb 2021 00:38:59 -0500
|
||||
Subject: [PATCH] x86-64: Refactor and improve performance of strchr-avx2.S
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
No bug. Just seemed the performance could be improved a bit. Observed
|
||||
and expected behavior are unchanged. Optimized body of main
|
||||
loop. Updated page cross logic and optimized accordingly. Made a few
|
||||
minor instruction selection modifications. No regressions in test
|
||||
suite. Both test-strchrnul and test-strchr passed.
|
||||
---
|
||||
sysdeps/x86_64/multiarch/strchr-avx2.S | 225 ++++++++++++-------------
|
||||
sysdeps/x86_64/multiarch/strchr.c | 4 +-
|
||||
2 files changed, 114 insertions(+), 115 deletions(-)
|
||||
|
||||
Conflicts:
|
||||
sysdeps/x86_64/multiarch/strchr.c
|
||||
(account for missing upstream macros)
|
||||
|
||||
diff --git a/sysdeps/x86_64/multiarch/strchr-avx2.S b/sysdeps/x86_64/multiarch/strchr-avx2.S
|
||||
index da7d2620..919d256c 100644
|
||||
--- a/sysdeps/x86_64/multiarch/strchr-avx2.S
|
||||
+++ b/sysdeps/x86_64/multiarch/strchr-avx2.S
|
||||
@@ -27,10 +27,12 @@
|
||||
# ifdef USE_AS_WCSCHR
|
||||
# define VPBROADCAST vpbroadcastd
|
||||
# define VPCMPEQ vpcmpeqd
|
||||
+# define VPMINU vpminud
|
||||
# define CHAR_REG esi
|
||||
# else
|
||||
# define VPBROADCAST vpbroadcastb
|
||||
# define VPCMPEQ vpcmpeqb
|
||||
+# define VPMINU vpminub
|
||||
# define CHAR_REG sil
|
||||
# endif
|
||||
|
||||
@@ -43,71 +45,54 @@
|
||||
# endif
|
||||
|
||||
# define VEC_SIZE 32
|
||||
+# define PAGE_SIZE 4096
|
||||
|
||||
.section SECTION(.text),"ax",@progbits
|
||||
ENTRY (STRCHR)
|
||||
movl %edi, %ecx
|
||||
- /* Broadcast CHAR to YMM0. */
|
||||
+# ifndef USE_AS_STRCHRNUL
|
||||
+ xorl %edx, %edx
|
||||
+# endif
|
||||
+
|
||||
+ /* Broadcast CHAR to YMM0. */
|
||||
vmovd %esi, %xmm0
|
||||
vpxor %xmm9, %xmm9, %xmm9
|
||||
VPBROADCAST %xmm0, %ymm0
|
||||
- /* Check if we may cross page boundary with one vector load. */
|
||||
- andl $(2 * VEC_SIZE - 1), %ecx
|
||||
- cmpl $VEC_SIZE, %ecx
|
||||
- ja L(cros_page_boundary)
|
||||
|
||||
- /* Check the first VEC_SIZE bytes. Search for both CHAR and the
|
||||
- null byte. */
|
||||
- vmovdqu (%rdi), %ymm8
|
||||
- VPCMPEQ %ymm8, %ymm0, %ymm1
|
||||
- VPCMPEQ %ymm8, %ymm9, %ymm2
|
||||
- vpor %ymm1, %ymm2, %ymm1
|
||||
- vpmovmskb %ymm1, %eax
|
||||
- testl %eax, %eax
|
||||
- jnz L(first_vec_x0)
|
||||
+ /* Check if we cross page boundary with one vector load. */
|
||||
+ andl $(PAGE_SIZE - 1), %ecx
|
||||
+ cmpl $(PAGE_SIZE - VEC_SIZE), %ecx
|
||||
+ ja L(cross_page_boundary)
|
||||
|
||||
- /* Align data for aligned loads in the loop. */
|
||||
- addq $VEC_SIZE, %rdi
|
||||
- andl $(VEC_SIZE - 1), %ecx
|
||||
- andq $-VEC_SIZE, %rdi
|
||||
-
|
||||
- jmp L(more_4x_vec)
|
||||
-
|
||||
- .p2align 4
|
||||
-L(cros_page_boundary):
|
||||
- andl $(VEC_SIZE - 1), %ecx
|
||||
- andq $-VEC_SIZE, %rdi
|
||||
+ /* Check the first VEC_SIZE bytes. Search for both CHAR and the
|
||||
+ null byte. */
|
||||
vmovdqu (%rdi), %ymm8
|
||||
VPCMPEQ %ymm8, %ymm0, %ymm1
|
||||
VPCMPEQ %ymm8, %ymm9, %ymm2
|
||||
vpor %ymm1, %ymm2, %ymm1
|
||||
vpmovmskb %ymm1, %eax
|
||||
- /* Remove the leading bytes. */
|
||||
- sarl %cl, %eax
|
||||
testl %eax, %eax
|
||||
- jz L(aligned_more)
|
||||
- /* Found CHAR or the null byte. */
|
||||
+ jz L(more_vecs)
|
||||
tzcntl %eax, %eax
|
||||
- addq %rcx, %rax
|
||||
-# ifdef USE_AS_STRCHRNUL
|
||||
+ /* Found CHAR or the null byte. */
|
||||
addq %rdi, %rax
|
||||
-# else
|
||||
- xorl %edx, %edx
|
||||
- leaq (%rdi, %rax), %rax
|
||||
- cmp (%rax), %CHAR_REG
|
||||
+# ifndef USE_AS_STRCHRNUL
|
||||
+ cmp (%rax), %CHAR_REG
|
||||
cmovne %rdx, %rax
|
||||
# endif
|
||||
L(return_vzeroupper):
|
||||
ZERO_UPPER_VEC_REGISTERS_RETURN
|
||||
|
||||
.p2align 4
|
||||
+L(more_vecs):
|
||||
+ /* Align data for aligned loads in the loop. */
|
||||
+ andq $-VEC_SIZE, %rdi
|
||||
L(aligned_more):
|
||||
- addq $VEC_SIZE, %rdi
|
||||
|
||||
-L(more_4x_vec):
|
||||
- /* Check the first 4 * VEC_SIZE. Only one VEC_SIZE at a time
|
||||
- since data is only aligned to VEC_SIZE. */
|
||||
- vmovdqa (%rdi), %ymm8
|
||||
+ /* Check the next 4 * VEC_SIZE. Only one VEC_SIZE at a time
|
||||
+ since data is only aligned to VEC_SIZE. */
|
||||
+ vmovdqa VEC_SIZE(%rdi), %ymm8
|
||||
+ addq $VEC_SIZE, %rdi
|
||||
VPCMPEQ %ymm8, %ymm0, %ymm1
|
||||
VPCMPEQ %ymm8, %ymm9, %ymm2
|
||||
vpor %ymm1, %ymm2, %ymm1
|
||||
@@ -137,61 +122,24 @@ L(more_4x_vec):
|
||||
vpor %ymm1, %ymm2, %ymm1
|
||||
vpmovmskb %ymm1, %eax
|
||||
testl %eax, %eax
|
||||
- jnz L(first_vec_x3)
|
||||
-
|
||||
- addq $(VEC_SIZE * 4), %rdi
|
||||
-
|
||||
- /* Align data to 4 * VEC_SIZE. */
|
||||
- movq %rdi, %rcx
|
||||
- andl $(4 * VEC_SIZE - 1), %ecx
|
||||
- andq $-(4 * VEC_SIZE), %rdi
|
||||
-
|
||||
- .p2align 4
|
||||
-L(loop_4x_vec):
|
||||
- /* Compare 4 * VEC at a time forward. */
|
||||
- vmovdqa (%rdi), %ymm5
|
||||
- vmovdqa VEC_SIZE(%rdi), %ymm6
|
||||
- vmovdqa (VEC_SIZE * 2)(%rdi), %ymm7
|
||||
- vmovdqa (VEC_SIZE * 3)(%rdi), %ymm8
|
||||
-
|
||||
- VPCMPEQ %ymm5, %ymm0, %ymm1
|
||||
- VPCMPEQ %ymm6, %ymm0, %ymm2
|
||||
- VPCMPEQ %ymm7, %ymm0, %ymm3
|
||||
- VPCMPEQ %ymm8, %ymm0, %ymm4
|
||||
-
|
||||
- VPCMPEQ %ymm5, %ymm9, %ymm5
|
||||
- VPCMPEQ %ymm6, %ymm9, %ymm6
|
||||
- VPCMPEQ %ymm7, %ymm9, %ymm7
|
||||
- VPCMPEQ %ymm8, %ymm9, %ymm8
|
||||
-
|
||||
- vpor %ymm1, %ymm5, %ymm1
|
||||
- vpor %ymm2, %ymm6, %ymm2
|
||||
- vpor %ymm3, %ymm7, %ymm3
|
||||
- vpor %ymm4, %ymm8, %ymm4
|
||||
-
|
||||
- vpor %ymm1, %ymm2, %ymm5
|
||||
- vpor %ymm3, %ymm4, %ymm6
|
||||
-
|
||||
- vpor %ymm5, %ymm6, %ymm5
|
||||
-
|
||||
- vpmovmskb %ymm5, %eax
|
||||
- testl %eax, %eax
|
||||
- jnz L(4x_vec_end)
|
||||
-
|
||||
- addq $(VEC_SIZE * 4), %rdi
|
||||
+ jz L(prep_loop_4x)
|
||||
|
||||
- jmp L(loop_4x_vec)
|
||||
+ tzcntl %eax, %eax
|
||||
+ leaq (VEC_SIZE * 3)(%rdi, %rax), %rax
|
||||
+# ifndef USE_AS_STRCHRNUL
|
||||
+ cmp (%rax), %CHAR_REG
|
||||
+ cmovne %rdx, %rax
|
||||
+# endif
|
||||
+ VZEROUPPER
|
||||
+ ret
|
||||
|
||||
.p2align 4
|
||||
L(first_vec_x0):
|
||||
- /* Found CHAR or the null byte. */
|
||||
tzcntl %eax, %eax
|
||||
-# ifdef USE_AS_STRCHRNUL
|
||||
+ /* Found CHAR or the null byte. */
|
||||
addq %rdi, %rax
|
||||
-# else
|
||||
- xorl %edx, %edx
|
||||
- leaq (%rdi, %rax), %rax
|
||||
- cmp (%rax), %CHAR_REG
|
||||
+# ifndef USE_AS_STRCHRNUL
|
||||
+ cmp (%rax), %CHAR_REG
|
||||
cmovne %rdx, %rax
|
||||
# endif
|
||||
VZEROUPPER_RETURN
|
||||
@@ -199,13 +147,9 @@ L(first_vec_x0):
|
||||
.p2align 4
|
||||
L(first_vec_x1):
|
||||
tzcntl %eax, %eax
|
||||
-# ifdef USE_AS_STRCHRNUL
|
||||
- addq $VEC_SIZE, %rax
|
||||
- addq %rdi, %rax
|
||||
-# else
|
||||
- xorl %edx, %edx
|
||||
leaq VEC_SIZE(%rdi, %rax), %rax
|
||||
- cmp (%rax), %CHAR_REG
|
||||
+# ifndef USE_AS_STRCHRNUL
|
||||
+ cmp (%rax), %CHAR_REG
|
||||
cmovne %rdx, %rax
|
||||
# endif
|
||||
VZEROUPPER_RETURN
|
||||
@@ -213,42 +157,97 @@ L(first_vec_x1):
|
||||
.p2align 4
|
||||
L(first_vec_x2):
|
||||
tzcntl %eax, %eax
|
||||
-# ifdef USE_AS_STRCHRNUL
|
||||
- addq $(VEC_SIZE * 2), %rax
|
||||
- addq %rdi, %rax
|
||||
-# else
|
||||
- xorl %edx, %edx
|
||||
+ /* Found CHAR or the null byte. */
|
||||
leaq (VEC_SIZE * 2)(%rdi, %rax), %rax
|
||||
- cmp (%rax), %CHAR_REG
|
||||
+# ifndef USE_AS_STRCHRNUL
|
||||
+ cmp (%rax), %CHAR_REG
|
||||
cmovne %rdx, %rax
|
||||
# endif
|
||||
VZEROUPPER_RETURN
|
||||
|
||||
+L(prep_loop_4x):
|
||||
+ /* Align data to 4 * VEC_SIZE. */
|
||||
+ andq $-(VEC_SIZE * 4), %rdi
|
||||
+
|
||||
.p2align 4
|
||||
-L(4x_vec_end):
|
||||
+L(loop_4x_vec):
|
||||
+ /* Compare 4 * VEC at a time forward. */
|
||||
+ vmovdqa (VEC_SIZE * 4)(%rdi), %ymm5
|
||||
+ vmovdqa (VEC_SIZE * 5)(%rdi), %ymm6
|
||||
+ vmovdqa (VEC_SIZE * 6)(%rdi), %ymm7
|
||||
+ vmovdqa (VEC_SIZE * 7)(%rdi), %ymm8
|
||||
+
|
||||
+ /* Leaves only CHARS matching esi as 0. */
|
||||
+ vpxor %ymm5, %ymm0, %ymm1
|
||||
+ vpxor %ymm6, %ymm0, %ymm2
|
||||
+ vpxor %ymm7, %ymm0, %ymm3
|
||||
+ vpxor %ymm8, %ymm0, %ymm4
|
||||
+
|
||||
+ VPMINU %ymm1, %ymm5, %ymm1
|
||||
+ VPMINU %ymm2, %ymm6, %ymm2
|
||||
+ VPMINU %ymm3, %ymm7, %ymm3
|
||||
+ VPMINU %ymm4, %ymm8, %ymm4
|
||||
+
|
||||
+ VPMINU %ymm1, %ymm2, %ymm5
|
||||
+ VPMINU %ymm3, %ymm4, %ymm6
|
||||
+
|
||||
+ VPMINU %ymm5, %ymm6, %ymm5
|
||||
+
|
||||
+ VPCMPEQ %ymm5, %ymm9, %ymm5
|
||||
+ vpmovmskb %ymm5, %eax
|
||||
+
|
||||
+ addq $(VEC_SIZE * 4), %rdi
|
||||
+ testl %eax, %eax
|
||||
+ jz L(loop_4x_vec)
|
||||
+
|
||||
+ VPCMPEQ %ymm1, %ymm9, %ymm1
|
||||
vpmovmskb %ymm1, %eax
|
||||
testl %eax, %eax
|
||||
jnz L(first_vec_x0)
|
||||
+
|
||||
+ VPCMPEQ %ymm2, %ymm9, %ymm2
|
||||
vpmovmskb %ymm2, %eax
|
||||
testl %eax, %eax
|
||||
jnz L(first_vec_x1)
|
||||
- vpmovmskb %ymm3, %eax
|
||||
- testl %eax, %eax
|
||||
- jnz L(first_vec_x2)
|
||||
+
|
||||
+ VPCMPEQ %ymm3, %ymm9, %ymm3
|
||||
+ VPCMPEQ %ymm4, %ymm9, %ymm4
|
||||
+ vpmovmskb %ymm3, %ecx
|
||||
vpmovmskb %ymm4, %eax
|
||||
+ salq $32, %rax
|
||||
+ orq %rcx, %rax
|
||||
+ tzcntq %rax, %rax
|
||||
+ leaq (VEC_SIZE * 2)(%rdi, %rax), %rax
|
||||
+# ifndef USE_AS_STRCHRNUL
|
||||
+ cmp (%rax), %CHAR_REG
|
||||
+ cmovne %rdx, %rax
|
||||
+# endif
|
||||
+ VZEROUPPER
|
||||
+ ret
|
||||
+
|
||||
+ /* Cold case for crossing page with first load. */
|
||||
+ .p2align 4
|
||||
+L(cross_page_boundary):
|
||||
+ andq $-VEC_SIZE, %rdi
|
||||
+ andl $(VEC_SIZE - 1), %ecx
|
||||
+
|
||||
+ vmovdqa (%rdi), %ymm8
|
||||
+ VPCMPEQ %ymm8, %ymm0, %ymm1
|
||||
+ VPCMPEQ %ymm8, %ymm9, %ymm2
|
||||
+ vpor %ymm1, %ymm2, %ymm1
|
||||
+ vpmovmskb %ymm1, %eax
|
||||
+ /* Remove the leading bits. */
|
||||
+ sarxl %ecx, %eax, %eax
|
||||
testl %eax, %eax
|
||||
-L(first_vec_x3):
|
||||
+ jz L(aligned_more)
|
||||
tzcntl %eax, %eax
|
||||
-# ifdef USE_AS_STRCHRNUL
|
||||
- addq $(VEC_SIZE * 3), %rax
|
||||
+ addq %rcx, %rdi
|
||||
addq %rdi, %rax
|
||||
-# else
|
||||
- xorl %edx, %edx
|
||||
- leaq (VEC_SIZE * 3)(%rdi, %rax), %rax
|
||||
- cmp (%rax), %CHAR_REG
|
||||
+# ifndef USE_AS_STRCHRNUL
|
||||
+ cmp (%rax), %CHAR_REG
|
||||
cmovne %rdx, %rax
|
||||
# endif
|
||||
VZEROUPPER_RETURN
|
||||
|
||||
END (STRCHR)
|
||||
-#endif
|
||||
+# endif
|
||||
diff --git a/sysdeps/x86_64/multiarch/strchr.c b/sysdeps/x86_64/multiarch/strchr.c
|
||||
index 7e582f02..5225bd4f 100644
|
||||
--- a/sysdeps/x86_64/multiarch/strchr.c
|
||||
+++ b/sysdeps/x86_64/multiarch/strchr.c
|
||||
@@ -38,11 +38,11 @@ IFUNC_SELECTOR (void)
|
||||
const struct cpu_features* cpu_features = __get_cpu_features ();
|
||||
|
||||
if (CPU_FEATURE_USABLE_P (cpu_features, AVX2)
|
||||
+ && CPU_FEATURE_USABLE_P (cpu_features, BMI2)
|
||||
&& CPU_FEATURES_ARCH_P (cpu_features, AVX_Fast_Unaligned_Load))
|
||||
{
|
||||
if (CPU_FEATURE_USABLE_P (cpu_features, AVX512VL)
|
||||
- && CPU_FEATURE_USABLE_P (cpu_features, AVX512BW)
|
||||
- && CPU_FEATURE_USABLE_P (cpu_features, BMI2))
|
||||
+ && CPU_FEATURE_USABLE_P (cpu_features, AVX512BW))
|
||||
return OPTIMIZE (evex);
|
||||
|
||||
if (CPU_FEATURE_USABLE_P (cpu_features, RTM))
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,67 +0,0 @@
|
||||
From 3ec5d83d2a237d39e7fd6ef7a0bc8ac4c171a4a5 Mon Sep 17 00:00:00 2001
|
||||
From: "H.J. Lu" <hjl.tools@gmail.com>
|
||||
Date: Sat, 25 Jan 2020 14:19:40 -0800
|
||||
Subject: [PATCH] x86-64: Avoid rep movsb with short distance [BZ #27130]
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
When copying with "rep movsb", if the distance between source and
|
||||
destination is N*4GB + [1..63] with N >= 0, performance may be very
|
||||
slow. This patch updates memmove-vec-unaligned-erms.S for AVX and
|
||||
AVX512 versions with the distance in RCX:
|
||||
|
||||
cmpl $63, %ecx
|
||||
// Don't use "rep movsb" if ECX <= 63
|
||||
jbe L(Don't use rep movsb")
|
||||
Use "rep movsb"
|
||||
|
||||
Benchtests data with bench-memcpy, bench-memcpy-large, bench-memcpy-random
|
||||
and bench-memcpy-walk on Skylake, Ice Lake and Tiger Lake show that its
|
||||
performance impact is within noise range as "rep movsb" is only used for
|
||||
data size >= 4KB.
|
||||
---
|
||||
.../multiarch/memmove-vec-unaligned-erms.S | 21 +++++++++++++++++++
|
||||
1 file changed, 21 insertions(+)
|
||||
|
||||
diff --git a/sysdeps/x86_64/multiarch/memmove-vec-unaligned-erms.S b/sysdeps/x86_64/multiarch/memmove-vec-unaligned-erms.S
|
||||
index 673b73aa..c475fed4 100644
|
||||
--- a/sysdeps/x86_64/multiarch/memmove-vec-unaligned-erms.S
|
||||
+++ b/sysdeps/x86_64/multiarch/memmove-vec-unaligned-erms.S
|
||||
@@ -64,6 +64,13 @@
|
||||
# endif
|
||||
#endif
|
||||
|
||||
+/* Avoid short distance rep movsb only with non-SSE vector. */
|
||||
+#ifndef AVOID_SHORT_DISTANCE_REP_MOVSB
|
||||
+# define AVOID_SHORT_DISTANCE_REP_MOVSB (VEC_SIZE > 16)
|
||||
+#else
|
||||
+# define AVOID_SHORT_DISTANCE_REP_MOVSB 0
|
||||
+#endif
|
||||
+
|
||||
#ifndef PREFETCH
|
||||
# define PREFETCH(addr) prefetcht0 addr
|
||||
#endif
|
||||
@@ -255,7 +262,21 @@ L(movsb):
|
||||
cmpq %r9, %rdi
|
||||
/* Avoid slow backward REP MOVSB. */
|
||||
jb L(more_8x_vec_backward)
|
||||
+# if AVOID_SHORT_DISTANCE_REP_MOVSB
|
||||
+ movq %rdi, %rcx
|
||||
+ subq %rsi, %rcx
|
||||
+ jmp 2f
|
||||
+# endif
|
||||
1:
|
||||
+# if AVOID_SHORT_DISTANCE_REP_MOVSB
|
||||
+ movq %rsi, %rcx
|
||||
+ subq %rdi, %rcx
|
||||
+2:
|
||||
+/* Avoid "rep movsb" if RCX, the distance between source and destination,
|
||||
+ is N*4GB + [1..63] with N >= 0. */
|
||||
+ cmpl $63, %ecx
|
||||
+ jbe L(more_2x_vec) /* Avoid "rep movsb" if ECX <= 63. */
|
||||
+# endif
|
||||
mov %RDX_LP, %RCX_LP
|
||||
rep movsb
|
||||
L(nop):
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,449 +0,0 @@
|
||||
From 1a8605b6cd257e8a74e29b5b71c057211f5fb847 Mon Sep 17 00:00:00 2001
|
||||
From: noah <goldstein.w.n@gmail.com>
|
||||
Date: Sat, 3 Apr 2021 04:12:15 -0400
|
||||
Subject: [PATCH] x86: Update large memcpy case in memmove-vec-unaligned-erms.S
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
No Bug. This commit updates the large memcpy case (no overlap). The
|
||||
update is to perform memcpy on either 2 or 4 contiguous pages at
|
||||
once. This 1) helps to alleviate the affects of false memory aliasing
|
||||
when destination and source have a close 4k alignment and 2) In most
|
||||
cases and for most DRAM units is a modestly more efficient access
|
||||
pattern. These changes are a clear performance improvement for
|
||||
VEC_SIZE =16/32, though more ambiguous for VEC_SIZE=64. test-memcpy,
|
||||
test-memccpy, test-mempcpy, test-memmove, and tst-memmove-overflow all
|
||||
pass.
|
||||
|
||||
Signed-off-by: Noah Goldstein <goldstein.w.n@gmail.com>
|
||||
---
|
||||
.../multiarch/memmove-vec-unaligned-erms.S | 338 ++++++++++++++----
|
||||
1 file changed, 265 insertions(+), 73 deletions(-)
|
||||
|
||||
Conflicts:
|
||||
sysdeps/x86_64/multiarch/memmove-vec-unaligned-erms.S
|
||||
(different number of sections)
|
||||
|
||||
diff --git a/sysdeps/x86_64/multiarch/memmove-vec-unaligned-erms.S b/sysdeps/x86_64/multiarch/memmove-vec-unaligned-erms.S
|
||||
index c475fed4..3e2dd6bc 100644
|
||||
--- a/sysdeps/x86_64/multiarch/memmove-vec-unaligned-erms.S
|
||||
+++ b/sysdeps/x86_64/multiarch/memmove-vec-unaligned-erms.S
|
||||
@@ -32,7 +32,16 @@
|
||||
overlapping addresses.
|
||||
6. If size >= __x86_shared_non_temporal_threshold and there is no
|
||||
overlap between destination and source, use non-temporal store
|
||||
- instead of aligned store. */
|
||||
+ instead of aligned store copying from either 2 or 4 pages at
|
||||
+ once.
|
||||
+ 8. For point 7) if size < 16 * __x86_shared_non_temporal_threshold
|
||||
+ and source and destination do not page alias, copy from 2 pages
|
||||
+ at once using non-temporal stores. Page aliasing in this case is
|
||||
+ considered true if destination's page alignment - sources' page
|
||||
+ alignment is less than 8 * VEC_SIZE.
|
||||
+ 9. If size >= 16 * __x86_shared_non_temporal_threshold or source
|
||||
+ and destination do page alias copy from 4 pages at once using
|
||||
+ non-temporal stores. */
|
||||
|
||||
#include <sysdep.h>
|
||||
|
||||
@@ -64,6 +73,34 @@
|
||||
# endif
|
||||
#endif
|
||||
|
||||
+#ifndef PAGE_SIZE
|
||||
+# define PAGE_SIZE 4096
|
||||
+#endif
|
||||
+
|
||||
+#if PAGE_SIZE != 4096
|
||||
+# error Unsupported PAGE_SIZE
|
||||
+#endif
|
||||
+
|
||||
+#ifndef LOG_PAGE_SIZE
|
||||
+# define LOG_PAGE_SIZE 12
|
||||
+#endif
|
||||
+
|
||||
+#if PAGE_SIZE != (1 << LOG_PAGE_SIZE)
|
||||
+# error Invalid LOG_PAGE_SIZE
|
||||
+#endif
|
||||
+
|
||||
+/* Byte per page for large_memcpy inner loop. */
|
||||
+#if VEC_SIZE == 64
|
||||
+# define LARGE_LOAD_SIZE (VEC_SIZE * 2)
|
||||
+#else
|
||||
+# define LARGE_LOAD_SIZE (VEC_SIZE * 4)
|
||||
+#endif
|
||||
+
|
||||
+/* Amount to shift rdx by to compare for memcpy_large_4x. */
|
||||
+#ifndef LOG_4X_MEMCPY_THRESH
|
||||
+# define LOG_4X_MEMCPY_THRESH 4
|
||||
+#endif
|
||||
+
|
||||
/* Avoid short distance rep movsb only with non-SSE vector. */
|
||||
#ifndef AVOID_SHORT_DISTANCE_REP_MOVSB
|
||||
# define AVOID_SHORT_DISTANCE_REP_MOVSB (VEC_SIZE > 16)
|
||||
@@ -103,6 +140,28 @@
|
||||
# error Unsupported PREFETCH_SIZE!
|
||||
#endif
|
||||
|
||||
+#if LARGE_LOAD_SIZE == (VEC_SIZE * 2)
|
||||
+# define LOAD_ONE_SET(base, offset, vec0, vec1, ...) \
|
||||
+ VMOVU (offset)base, vec0; \
|
||||
+ VMOVU ((offset) + VEC_SIZE)base, vec1;
|
||||
+# define STORE_ONE_SET(base, offset, vec0, vec1, ...) \
|
||||
+ VMOVNT vec0, (offset)base; \
|
||||
+ VMOVNT vec1, ((offset) + VEC_SIZE)base;
|
||||
+#elif LARGE_LOAD_SIZE == (VEC_SIZE * 4)
|
||||
+# define LOAD_ONE_SET(base, offset, vec0, vec1, vec2, vec3) \
|
||||
+ VMOVU (offset)base, vec0; \
|
||||
+ VMOVU ((offset) + VEC_SIZE)base, vec1; \
|
||||
+ VMOVU ((offset) + VEC_SIZE * 2)base, vec2; \
|
||||
+ VMOVU ((offset) + VEC_SIZE * 3)base, vec3;
|
||||
+# define STORE_ONE_SET(base, offset, vec0, vec1, vec2, vec3) \
|
||||
+ VMOVNT vec0, (offset)base; \
|
||||
+ VMOVNT vec1, ((offset) + VEC_SIZE)base; \
|
||||
+ VMOVNT vec2, ((offset) + VEC_SIZE * 2)base; \
|
||||
+ VMOVNT vec3, ((offset) + VEC_SIZE * 3)base;
|
||||
+#else
|
||||
+# error Invalid LARGE_LOAD_SIZE
|
||||
+#endif
|
||||
+
|
||||
#ifndef SECTION
|
||||
# error SECTION is not defined!
|
||||
#endif
|
||||
@@ -390,6 +449,15 @@ L(last_4x_vec):
|
||||
VZEROUPPER_RETURN
|
||||
|
||||
L(more_8x_vec):
|
||||
+ /* Check if non-temporal move candidate. */
|
||||
+#if (defined USE_MULTIARCH || VEC_SIZE == 16) && IS_IN (libc)
|
||||
+ /* Check non-temporal store threshold. */
|
||||
+ cmp __x86_shared_non_temporal_threshold(%rip), %RDX_LP
|
||||
+ ja L(large_memcpy_2x)
|
||||
+#endif
|
||||
+ /* Entry if rdx is greater than non-temporal threshold but there
|
||||
+ is overlap. */
|
||||
+L(more_8x_vec_check):
|
||||
cmpq %rsi, %rdi
|
||||
ja L(more_8x_vec_backward)
|
||||
/* Source == destination is less common. */
|
||||
@@ -416,24 +484,21 @@ L(more_8x_vec):
|
||||
subq %r8, %rdi
|
||||
/* Adjust length. */
|
||||
addq %r8, %rdx
|
||||
-#if (defined USE_MULTIARCH || VEC_SIZE == 16) && IS_IN (libc)
|
||||
- /* Check non-temporal store threshold. */
|
||||
- cmp __x86_shared_non_temporal_threshold(%rip), %RDX_LP
|
||||
- ja L(large_forward)
|
||||
-#endif
|
||||
+
|
||||
+ .p2align 4
|
||||
L(loop_4x_vec_forward):
|
||||
/* Copy 4 * VEC a time forward. */
|
||||
VMOVU (%rsi), %VEC(0)
|
||||
VMOVU VEC_SIZE(%rsi), %VEC(1)
|
||||
VMOVU (VEC_SIZE * 2)(%rsi), %VEC(2)
|
||||
VMOVU (VEC_SIZE * 3)(%rsi), %VEC(3)
|
||||
- addq $(VEC_SIZE * 4), %rsi
|
||||
- subq $(VEC_SIZE * 4), %rdx
|
||||
+ subq $-(VEC_SIZE * 4), %rsi
|
||||
+ addq $-(VEC_SIZE * 4), %rdx
|
||||
VMOVA %VEC(0), (%rdi)
|
||||
VMOVA %VEC(1), VEC_SIZE(%rdi)
|
||||
VMOVA %VEC(2), (VEC_SIZE * 2)(%rdi)
|
||||
VMOVA %VEC(3), (VEC_SIZE * 3)(%rdi)
|
||||
- addq $(VEC_SIZE * 4), %rdi
|
||||
+ subq $-(VEC_SIZE * 4), %rdi
|
||||
cmpq $(VEC_SIZE * 4), %rdx
|
||||
ja L(loop_4x_vec_forward)
|
||||
/* Store the last 4 * VEC. */
|
||||
@@ -467,24 +532,21 @@ L(more_8x_vec_backward):
|
||||
subq %r8, %r9
|
||||
/* Adjust length. */
|
||||
subq %r8, %rdx
|
||||
-#if (defined USE_MULTIARCH || VEC_SIZE == 16) && IS_IN (libc)
|
||||
- /* Check non-temporal store threshold. */
|
||||
- cmp __x86_shared_non_temporal_threshold(%rip), %RDX_LP
|
||||
- ja L(large_backward)
|
||||
-#endif
|
||||
+
|
||||
+ .p2align 4
|
||||
L(loop_4x_vec_backward):
|
||||
/* Copy 4 * VEC a time backward. */
|
||||
VMOVU (%rcx), %VEC(0)
|
||||
VMOVU -VEC_SIZE(%rcx), %VEC(1)
|
||||
VMOVU -(VEC_SIZE * 2)(%rcx), %VEC(2)
|
||||
VMOVU -(VEC_SIZE * 3)(%rcx), %VEC(3)
|
||||
- subq $(VEC_SIZE * 4), %rcx
|
||||
- subq $(VEC_SIZE * 4), %rdx
|
||||
+ addq $-(VEC_SIZE * 4), %rcx
|
||||
+ addq $-(VEC_SIZE * 4), %rdx
|
||||
VMOVA %VEC(0), (%r9)
|
||||
VMOVA %VEC(1), -VEC_SIZE(%r9)
|
||||
VMOVA %VEC(2), -(VEC_SIZE * 2)(%r9)
|
||||
VMOVA %VEC(3), -(VEC_SIZE * 3)(%r9)
|
||||
- subq $(VEC_SIZE * 4), %r9
|
||||
+ addq $-(VEC_SIZE * 4), %r9
|
||||
cmpq $(VEC_SIZE * 4), %rdx
|
||||
ja L(loop_4x_vec_backward)
|
||||
/* Store the first 4 * VEC. */
|
||||
@@ -497,72 +559,202 @@ L(loop_4x_vec_backward):
|
||||
VZEROUPPER_RETURN
|
||||
|
||||
#if (defined USE_MULTIARCH || VEC_SIZE == 16) && IS_IN (libc)
|
||||
-L(large_forward):
|
||||
+ .p2align 4
|
||||
+L(large_memcpy_2x):
|
||||
+ /* Compute absolute value of difference between source and
|
||||
+ destination. */
|
||||
+ movq %rdi, %r9
|
||||
+ subq %rsi, %r9
|
||||
+ movq %r9, %r8
|
||||
+ leaq -1(%r9), %rcx
|
||||
+ sarq $63, %r8
|
||||
+ xorq %r8, %r9
|
||||
+ subq %r8, %r9
|
||||
/* Don't use non-temporal store if there is overlap between
|
||||
- destination and source since destination may be in cache
|
||||
- when source is loaded. */
|
||||
- leaq (%rdi, %rdx), %r10
|
||||
- cmpq %r10, %rsi
|
||||
- jb L(loop_4x_vec_forward)
|
||||
-L(loop_large_forward):
|
||||
+ destination and source since destination may be in cache when
|
||||
+ source is loaded. */
|
||||
+ cmpq %r9, %rdx
|
||||
+ ja L(more_8x_vec_check)
|
||||
+
|
||||
+ /* Cache align destination. First store the first 64 bytes then
|
||||
+ adjust alignments. */
|
||||
+ VMOVU (%rsi), %VEC(8)
|
||||
+#if VEC_SIZE < 64
|
||||
+ VMOVU VEC_SIZE(%rsi), %VEC(9)
|
||||
+#if VEC_SIZE < 32
|
||||
+ VMOVU (VEC_SIZE * 2)(%rsi), %VEC(10)
|
||||
+ VMOVU (VEC_SIZE * 3)(%rsi), %VEC(11)
|
||||
+#endif
|
||||
+#endif
|
||||
+ VMOVU %VEC(8), (%rdi)
|
||||
+#if VEC_SIZE < 64
|
||||
+ VMOVU %VEC(9), VEC_SIZE(%rdi)
|
||||
+#if VEC_SIZE < 32
|
||||
+ VMOVU %VEC(10), (VEC_SIZE * 2)(%rdi)
|
||||
+ VMOVU %VEC(11), (VEC_SIZE * 3)(%rdi)
|
||||
+#endif
|
||||
+#endif
|
||||
+ /* Adjust source, destination, and size. */
|
||||
+ movq %rdi, %r8
|
||||
+ andq $63, %r8
|
||||
+ /* Get the negative of offset for alignment. */
|
||||
+ subq $64, %r8
|
||||
+ /* Adjust source. */
|
||||
+ subq %r8, %rsi
|
||||
+ /* Adjust destination which should be aligned now. */
|
||||
+ subq %r8, %rdi
|
||||
+ /* Adjust length. */
|
||||
+ addq %r8, %rdx
|
||||
+
|
||||
+ /* Test if source and destination addresses will alias. If they do
|
||||
+ the larger pipeline in large_memcpy_4x alleviated the
|
||||
+ performance drop. */
|
||||
+ testl $(PAGE_SIZE - VEC_SIZE * 8), %ecx
|
||||
+ jz L(large_memcpy_4x)
|
||||
+
|
||||
+ movq %rdx, %r10
|
||||
+ shrq $LOG_4X_MEMCPY_THRESH, %r10
|
||||
+ cmp __x86_shared_non_temporal_threshold(%rip), %r10
|
||||
+ jae L(large_memcpy_4x)
|
||||
+
|
||||
+ /* edx will store remainder size for copying tail. */
|
||||
+ andl $(PAGE_SIZE * 2 - 1), %edx
|
||||
+ /* r10 stores outer loop counter. */
|
||||
+ shrq $((LOG_PAGE_SIZE + 1) - LOG_4X_MEMCPY_THRESH), %r10
|
||||
+ /* Copy 4x VEC at a time from 2 pages. */
|
||||
+ .p2align 4
|
||||
+L(loop_large_memcpy_2x_outer):
|
||||
+ /* ecx stores inner loop counter. */
|
||||
+ movl $(PAGE_SIZE / LARGE_LOAD_SIZE), %ecx
|
||||
+L(loop_large_memcpy_2x_inner):
|
||||
+ PREFETCH_ONE_SET(1, (%rsi), PREFETCHED_LOAD_SIZE)
|
||||
+ PREFETCH_ONE_SET(1, (%rsi), PREFETCHED_LOAD_SIZE * 2)
|
||||
+ PREFETCH_ONE_SET(1, (%rsi), PAGE_SIZE + PREFETCHED_LOAD_SIZE)
|
||||
+ PREFETCH_ONE_SET(1, (%rsi), PAGE_SIZE + PREFETCHED_LOAD_SIZE * 2)
|
||||
+ /* Load vectors from rsi. */
|
||||
+ LOAD_ONE_SET((%rsi), 0, %VEC(0), %VEC(1), %VEC(2), %VEC(3))
|
||||
+ LOAD_ONE_SET((%rsi), PAGE_SIZE, %VEC(4), %VEC(5), %VEC(6), %VEC(7))
|
||||
+ subq $-LARGE_LOAD_SIZE, %rsi
|
||||
+ /* Non-temporal store vectors to rdi. */
|
||||
+ STORE_ONE_SET((%rdi), 0, %VEC(0), %VEC(1), %VEC(2), %VEC(3))
|
||||
+ STORE_ONE_SET((%rdi), PAGE_SIZE, %VEC(4), %VEC(5), %VEC(6), %VEC(7))
|
||||
+ subq $-LARGE_LOAD_SIZE, %rdi
|
||||
+ decl %ecx
|
||||
+ jnz L(loop_large_memcpy_2x_inner)
|
||||
+ addq $PAGE_SIZE, %rdi
|
||||
+ addq $PAGE_SIZE, %rsi
|
||||
+ decq %r10
|
||||
+ jne L(loop_large_memcpy_2x_outer)
|
||||
+ sfence
|
||||
+
|
||||
+ /* Check if only last 4 loads are needed. */
|
||||
+ cmpl $(VEC_SIZE * 4), %edx
|
||||
+ jbe L(large_memcpy_2x_end)
|
||||
+
|
||||
+ /* Handle the last 2 * PAGE_SIZE bytes. */
|
||||
+L(loop_large_memcpy_2x_tail):
|
||||
/* Copy 4 * VEC a time forward with non-temporal stores. */
|
||||
- PREFETCH_ONE_SET (1, (%rsi), PREFETCHED_LOAD_SIZE * 2)
|
||||
- PREFETCH_ONE_SET (1, (%rsi), PREFETCHED_LOAD_SIZE * 3)
|
||||
+ PREFETCH_ONE_SET (1, (%rsi), PREFETCHED_LOAD_SIZE)
|
||||
+ PREFETCH_ONE_SET (1, (%rdi), PREFETCHED_LOAD_SIZE)
|
||||
VMOVU (%rsi), %VEC(0)
|
||||
VMOVU VEC_SIZE(%rsi), %VEC(1)
|
||||
VMOVU (VEC_SIZE * 2)(%rsi), %VEC(2)
|
||||
VMOVU (VEC_SIZE * 3)(%rsi), %VEC(3)
|
||||
- addq $PREFETCHED_LOAD_SIZE, %rsi
|
||||
- subq $PREFETCHED_LOAD_SIZE, %rdx
|
||||
- VMOVNT %VEC(0), (%rdi)
|
||||
- VMOVNT %VEC(1), VEC_SIZE(%rdi)
|
||||
- VMOVNT %VEC(2), (VEC_SIZE * 2)(%rdi)
|
||||
- VMOVNT %VEC(3), (VEC_SIZE * 3)(%rdi)
|
||||
- addq $PREFETCHED_LOAD_SIZE, %rdi
|
||||
- cmpq $PREFETCHED_LOAD_SIZE, %rdx
|
||||
- ja L(loop_large_forward)
|
||||
- sfence
|
||||
+ subq $-(VEC_SIZE * 4), %rsi
|
||||
+ addl $-(VEC_SIZE * 4), %edx
|
||||
+ VMOVA %VEC(0), (%rdi)
|
||||
+ VMOVA %VEC(1), VEC_SIZE(%rdi)
|
||||
+ VMOVA %VEC(2), (VEC_SIZE * 2)(%rdi)
|
||||
+ VMOVA %VEC(3), (VEC_SIZE * 3)(%rdi)
|
||||
+ subq $-(VEC_SIZE * 4), %rdi
|
||||
+ cmpl $(VEC_SIZE * 4), %edx
|
||||
+ ja L(loop_large_memcpy_2x_tail)
|
||||
+
|
||||
+L(large_memcpy_2x_end):
|
||||
/* Store the last 4 * VEC. */
|
||||
- VMOVU %VEC(5), (%rcx)
|
||||
- VMOVU %VEC(6), -VEC_SIZE(%rcx)
|
||||
- VMOVU %VEC(7), -(VEC_SIZE * 2)(%rcx)
|
||||
- VMOVU %VEC(8), -(VEC_SIZE * 3)(%rcx)
|
||||
- /* Store the first VEC. */
|
||||
- VMOVU %VEC(4), (%r11)
|
||||
+ VMOVU -(VEC_SIZE * 4)(%rsi, %rdx), %VEC(0)
|
||||
+ VMOVU -(VEC_SIZE * 3)(%rsi, %rdx), %VEC(1)
|
||||
+ VMOVU -(VEC_SIZE * 2)(%rsi, %rdx), %VEC(2)
|
||||
+ VMOVU -VEC_SIZE(%rsi, %rdx), %VEC(3)
|
||||
+
|
||||
+ VMOVU %VEC(0), -(VEC_SIZE * 4)(%rdi, %rdx)
|
||||
+ VMOVU %VEC(1), -(VEC_SIZE * 3)(%rdi, %rdx)
|
||||
+ VMOVU %VEC(2), -(VEC_SIZE * 2)(%rdi, %rdx)
|
||||
+ VMOVU %VEC(3), -VEC_SIZE(%rdi, %rdx)
|
||||
VZEROUPPER_RETURN
|
||||
|
||||
-L(large_backward):
|
||||
- /* Don't use non-temporal store if there is overlap between
|
||||
- destination and source since destination may be in cache
|
||||
- when source is loaded. */
|
||||
- leaq (%rcx, %rdx), %r10
|
||||
- cmpq %r10, %r9
|
||||
- jb L(loop_4x_vec_backward)
|
||||
-L(loop_large_backward):
|
||||
- /* Copy 4 * VEC a time backward with non-temporal stores. */
|
||||
- PREFETCH_ONE_SET (-1, (%rcx), -PREFETCHED_LOAD_SIZE * 2)
|
||||
- PREFETCH_ONE_SET (-1, (%rcx), -PREFETCHED_LOAD_SIZE * 3)
|
||||
- VMOVU (%rcx), %VEC(0)
|
||||
- VMOVU -VEC_SIZE(%rcx), %VEC(1)
|
||||
- VMOVU -(VEC_SIZE * 2)(%rcx), %VEC(2)
|
||||
- VMOVU -(VEC_SIZE * 3)(%rcx), %VEC(3)
|
||||
- subq $PREFETCHED_LOAD_SIZE, %rcx
|
||||
- subq $PREFETCHED_LOAD_SIZE, %rdx
|
||||
- VMOVNT %VEC(0), (%r9)
|
||||
- VMOVNT %VEC(1), -VEC_SIZE(%r9)
|
||||
- VMOVNT %VEC(2), -(VEC_SIZE * 2)(%r9)
|
||||
- VMOVNT %VEC(3), -(VEC_SIZE * 3)(%r9)
|
||||
- subq $PREFETCHED_LOAD_SIZE, %r9
|
||||
- cmpq $PREFETCHED_LOAD_SIZE, %rdx
|
||||
- ja L(loop_large_backward)
|
||||
+ .p2align 4
|
||||
+L(large_memcpy_4x):
|
||||
+ movq %rdx, %r10
|
||||
+ /* edx will store remainder size for copying tail. */
|
||||
+ andl $(PAGE_SIZE * 4 - 1), %edx
|
||||
+ /* r10 stores outer loop counter. */
|
||||
+ shrq $(LOG_PAGE_SIZE + 2), %r10
|
||||
+ /* Copy 4x VEC at a time from 4 pages. */
|
||||
+ .p2align 4
|
||||
+L(loop_large_memcpy_4x_outer):
|
||||
+ /* ecx stores inner loop counter. */
|
||||
+ movl $(PAGE_SIZE / LARGE_LOAD_SIZE), %ecx
|
||||
+L(loop_large_memcpy_4x_inner):
|
||||
+ /* Only one prefetch set per page as doing 4 pages give more time
|
||||
+ for prefetcher to keep up. */
|
||||
+ PREFETCH_ONE_SET(1, (%rsi), PREFETCHED_LOAD_SIZE)
|
||||
+ PREFETCH_ONE_SET(1, (%rsi), PAGE_SIZE + PREFETCHED_LOAD_SIZE)
|
||||
+ PREFETCH_ONE_SET(1, (%rsi), PAGE_SIZE * 2 + PREFETCHED_LOAD_SIZE)
|
||||
+ PREFETCH_ONE_SET(1, (%rsi), PAGE_SIZE * 3 + PREFETCHED_LOAD_SIZE)
|
||||
+ /* Load vectors from rsi. */
|
||||
+ LOAD_ONE_SET((%rsi), 0, %VEC(0), %VEC(1), %VEC(2), %VEC(3))
|
||||
+ LOAD_ONE_SET((%rsi), PAGE_SIZE, %VEC(4), %VEC(5), %VEC(6), %VEC(7))
|
||||
+ LOAD_ONE_SET((%rsi), PAGE_SIZE * 2, %VEC(8), %VEC(9), %VEC(10), %VEC(11))
|
||||
+ LOAD_ONE_SET((%rsi), PAGE_SIZE * 3, %VEC(12), %VEC(13), %VEC(14), %VEC(15))
|
||||
+ subq $-LARGE_LOAD_SIZE, %rsi
|
||||
+ /* Non-temporal store vectors to rdi. */
|
||||
+ STORE_ONE_SET((%rdi), 0, %VEC(0), %VEC(1), %VEC(2), %VEC(3))
|
||||
+ STORE_ONE_SET((%rdi), PAGE_SIZE, %VEC(4), %VEC(5), %VEC(6), %VEC(7))
|
||||
+ STORE_ONE_SET((%rdi), PAGE_SIZE * 2, %VEC(8), %VEC(9), %VEC(10), %VEC(11))
|
||||
+ STORE_ONE_SET((%rdi), PAGE_SIZE * 3, %VEC(12), %VEC(13), %VEC(14), %VEC(15))
|
||||
+ subq $-LARGE_LOAD_SIZE, %rdi
|
||||
+ decl %ecx
|
||||
+ jnz L(loop_large_memcpy_4x_inner)
|
||||
+ addq $(PAGE_SIZE * 3), %rdi
|
||||
+ addq $(PAGE_SIZE * 3), %rsi
|
||||
+ decq %r10
|
||||
+ jne L(loop_large_memcpy_4x_outer)
|
||||
sfence
|
||||
- /* Store the first 4 * VEC. */
|
||||
- VMOVU %VEC(4), (%rdi)
|
||||
- VMOVU %VEC(5), VEC_SIZE(%rdi)
|
||||
- VMOVU %VEC(6), (VEC_SIZE * 2)(%rdi)
|
||||
- VMOVU %VEC(7), (VEC_SIZE * 3)(%rdi)
|
||||
- /* Store the last VEC. */
|
||||
- VMOVU %VEC(8), (%r11)
|
||||
+ /* Check if only last 4 loads are needed. */
|
||||
+ cmpl $(VEC_SIZE * 4), %edx
|
||||
+ jbe L(large_memcpy_4x_end)
|
||||
+
|
||||
+ /* Handle the last 4 * PAGE_SIZE bytes. */
|
||||
+L(loop_large_memcpy_4x_tail):
|
||||
+ /* Copy 4 * VEC a time forward with non-temporal stores. */
|
||||
+ PREFETCH_ONE_SET (1, (%rsi), PREFETCHED_LOAD_SIZE)
|
||||
+ PREFETCH_ONE_SET (1, (%rdi), PREFETCHED_LOAD_SIZE)
|
||||
+ VMOVU (%rsi), %VEC(0)
|
||||
+ VMOVU VEC_SIZE(%rsi), %VEC(1)
|
||||
+ VMOVU (VEC_SIZE * 2)(%rsi), %VEC(2)
|
||||
+ VMOVU (VEC_SIZE * 3)(%rsi), %VEC(3)
|
||||
+ subq $-(VEC_SIZE * 4), %rsi
|
||||
+ addl $-(VEC_SIZE * 4), %edx
|
||||
+ VMOVA %VEC(0), (%rdi)
|
||||
+ VMOVA %VEC(1), VEC_SIZE(%rdi)
|
||||
+ VMOVA %VEC(2), (VEC_SIZE * 2)(%rdi)
|
||||
+ VMOVA %VEC(3), (VEC_SIZE * 3)(%rdi)
|
||||
+ subq $-(VEC_SIZE * 4), %rdi
|
||||
+ cmpl $(VEC_SIZE * 4), %edx
|
||||
+ ja L(loop_large_memcpy_4x_tail)
|
||||
+
|
||||
+L(large_memcpy_4x_end):
|
||||
+ /* Store the last 4 * VEC. */
|
||||
+ VMOVU -(VEC_SIZE * 4)(%rsi, %rdx), %VEC(0)
|
||||
+ VMOVU -(VEC_SIZE * 3)(%rsi, %rdx), %VEC(1)
|
||||
+ VMOVU -(VEC_SIZE * 2)(%rsi, %rdx), %VEC(2)
|
||||
+ VMOVU -VEC_SIZE(%rsi, %rdx), %VEC(3)
|
||||
+
|
||||
+ VMOVU %VEC(0), -(VEC_SIZE * 4)(%rdi, %rdx)
|
||||
+ VMOVU %VEC(1), -(VEC_SIZE * 3)(%rdi, %rdx)
|
||||
+ VMOVU %VEC(2), -(VEC_SIZE * 2)(%rdi, %rdx)
|
||||
+ VMOVU %VEC(3), -VEC_SIZE(%rdi, %rdx)
|
||||
VZEROUPPER_RETURN
|
||||
#endif
|
||||
END (MEMMOVE_SYMBOL (__memmove, unaligned_erms))
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,151 +0,0 @@
|
||||
From ecd8b842cf37ea112e59cd9085ff1f1b6e208ae0 Mon Sep 17 00:00:00 2001
|
||||
From: "H.J. Lu" <hjl.tools@gmail.com>
|
||||
Date: Mon, 21 Jan 2019 11:29:58 -0800
|
||||
Subject: [PATCH] x86-64 memrchr: Properly handle the length parameter [BZ#
|
||||
24097]
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
On x32, the size_t parameter may be passed in the lower 32 bits of a
|
||||
64-bit register with the non-zero upper 32 bits. The string/memory
|
||||
functions written in assembly can only use the lower 32 bits of a
|
||||
64-bit register as length or must clear the upper 32 bits before using
|
||||
the full 64-bit register for length.
|
||||
|
||||
This pach fixes memrchr for x32. Tested on x86-64 and x32. On x86-64,
|
||||
libc.so is the same with and withou the fix.
|
||||
|
||||
[BZ# 24097]
|
||||
CVE-2019-6488
|
||||
* sysdeps/x86_64/memrchr.S: Use RDX_LP for length.
|
||||
* sysdeps/x86_64/multiarch/memrchr-avx2.S: Likewise.
|
||||
* sysdeps/x86_64/x32/Makefile (tests): Add tst-size_t-memrchr.
|
||||
* sysdeps/x86_64/x32/tst-size_t-memrchr.c: New file.
|
||||
---
|
||||
sysdeps/x86_64/memrchr.S | 4 +-
|
||||
sysdeps/x86_64/multiarch/memrchr-avx2.S | 4 +-
|
||||
sysdeps/x86_64/x32/Makefile | 3 +-
|
||||
sysdeps/x86_64/x32/tst-size_t-memrchr.c | 57 +++++++++++++++++++++++++
|
||||
4 files changed, 63 insertions(+), 5 deletions(-)
|
||||
create mode 100644 sysdeps/x86_64/x32/tst-size_t-memrchr.c
|
||||
|
||||
Conflicts:
|
||||
ChangeLog
|
||||
(removed)
|
||||
|
||||
diff --git a/sysdeps/x86_64/memrchr.S b/sysdeps/x86_64/memrchr.S
|
||||
index b8e3fa1d..dc82f8f7 100644
|
||||
--- a/sysdeps/x86_64/memrchr.S
|
||||
+++ b/sysdeps/x86_64/memrchr.S
|
||||
@@ -24,13 +24,13 @@
|
||||
ENTRY (__memrchr)
|
||||
movd %esi, %xmm1
|
||||
|
||||
- sub $16, %rdx
|
||||
+ sub $16, %RDX_LP
|
||||
jbe L(length_less16)
|
||||
|
||||
punpcklbw %xmm1, %xmm1
|
||||
punpcklbw %xmm1, %xmm1
|
||||
|
||||
- add %rdx, %rdi
|
||||
+ add %RDX_LP, %RDI_LP
|
||||
pshufd $0, %xmm1, %xmm1
|
||||
|
||||
movdqu (%rdi), %xmm0
|
||||
diff --git a/sysdeps/x86_64/multiarch/memrchr-avx2.S b/sysdeps/x86_64/multiarch/memrchr-avx2.S
|
||||
index b41a58bc..ce488dd9 100644
|
||||
--- a/sysdeps/x86_64/multiarch/memrchr-avx2.S
|
||||
+++ b/sysdeps/x86_64/multiarch/memrchr-avx2.S
|
||||
@@ -32,10 +32,10 @@ ENTRY (__memrchr_avx2)
|
||||
vmovd %esi, %xmm0
|
||||
vpbroadcastb %xmm0, %ymm0
|
||||
|
||||
- subq $VEC_SIZE, %rdx
|
||||
+ sub $VEC_SIZE, %RDX_LP
|
||||
jbe L(last_vec_or_less)
|
||||
|
||||
- addq %rdx, %rdi
|
||||
+ add %RDX_LP, %RDI_LP
|
||||
|
||||
/* Check the last VEC_SIZE bytes. */
|
||||
vpcmpeqb (%rdi), %ymm0, %ymm1
|
||||
diff --git a/sysdeps/x86_64/x32/Makefile b/sysdeps/x86_64/x32/Makefile
|
||||
index 2fe1e5ac..e99dbd7c 100644
|
||||
--- a/sysdeps/x86_64/x32/Makefile
|
||||
+++ b/sysdeps/x86_64/x32/Makefile
|
||||
@@ -6,7 +6,8 @@ CFLAGS-s_llround.c += -fno-builtin-lround
|
||||
endif
|
||||
|
||||
ifeq ($(subdir),string)
|
||||
-tests += tst-size_t-memchr tst-size_t-memcmp tst-size_t-memcpy
|
||||
+tests += tst-size_t-memchr tst-size_t-memcmp tst-size_t-memcpy \
|
||||
+ tst-size_t-memrchr
|
||||
endif
|
||||
|
||||
ifeq ($(subdir),wcsmbs)
|
||||
diff --git a/sysdeps/x86_64/x32/tst-size_t-memrchr.c b/sysdeps/x86_64/x32/tst-size_t-memrchr.c
|
||||
new file mode 100644
|
||||
index 00000000..c83699c0
|
||||
--- /dev/null
|
||||
+++ b/sysdeps/x86_64/x32/tst-size_t-memrchr.c
|
||||
@@ -0,0 +1,57 @@
|
||||
+/* Test memrchr with size_t in the lower 32 bits of 64-bit register.
|
||||
+ Copyright (C) 2019 Free Software Foundation, Inc.
|
||||
+ This file is part of the GNU C Library.
|
||||
+
|
||||
+ The GNU C Library is free software; you can redistribute it and/or
|
||||
+ modify it under the terms of the GNU Lesser General Public
|
||||
+ License as published by the Free Software Foundation; either
|
||||
+ version 2.1 of the License, or (at your option) any later version.
|
||||
+
|
||||
+ The GNU C Library is distributed in the hope that it will be useful,
|
||||
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
||||
+ Lesser General Public License for more details.
|
||||
+
|
||||
+ You should have received a copy of the GNU Lesser General Public
|
||||
+ License along with the GNU C Library; if not, see
|
||||
+ <http://www.gnu.org/licenses/>. */
|
||||
+
|
||||
+#define TEST_NAME "memrchr"
|
||||
+#include "test-size_t.h"
|
||||
+
|
||||
+IMPL (memchr, 1)
|
||||
+
|
||||
+typedef void * (*proto_t) (const void *, int, size_t);
|
||||
+
|
||||
+static void *
|
||||
+__attribute__ ((noinline, noclone))
|
||||
+do_memrchr (parameter_t a, parameter_t b)
|
||||
+{
|
||||
+ return CALL (&b, a.p, (uintptr_t) b.p, a.len);
|
||||
+}
|
||||
+
|
||||
+static int
|
||||
+test_main (void)
|
||||
+{
|
||||
+ test_init ();
|
||||
+
|
||||
+ parameter_t src = { { page_size }, buf2 };
|
||||
+ parameter_t c = { { 0 }, (void *) (uintptr_t) 0x12 };
|
||||
+
|
||||
+ int ret = 0;
|
||||
+ FOR_EACH_IMPL (impl, 0)
|
||||
+ {
|
||||
+ c.fn = impl->fn;
|
||||
+ void * res = do_memrchr (src, c);
|
||||
+ if (res)
|
||||
+ {
|
||||
+ error (0, 0, "Wrong result in function %s: %p != NULL",
|
||||
+ impl->name, res);
|
||||
+ ret = 1;
|
||||
+ }
|
||||
+ }
|
||||
+
|
||||
+ return ret ? EXIT_FAILURE : EXIT_SUCCESS;
|
||||
+}
|
||||
+
|
||||
+#include <support/test-driver.c>
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,92 +0,0 @@
|
||||
From 83c5b368226c34a2f0a5287df40fc290b2b34359 Mon Sep 17 00:00:00 2001
|
||||
From: "H.J. Lu" <hjl.tools@gmail.com>
|
||||
Date: Mon, 19 Apr 2021 10:45:07 -0700
|
||||
Subject: [PATCH] x86-64: Require BMI2 for strchr-avx2.S
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
Since strchr-avx2.S updated by
|
||||
|
||||
commit 1f745ecc2109890886b161d4791e1406fdfc29b8
|
||||
Author: noah <goldstein.w.n@gmail.com>
|
||||
Date: Wed Feb 3 00:38:59 2021 -0500
|
||||
|
||||
x86-64: Refactor and improve performance of strchr-avx2.S
|
||||
|
||||
uses sarx:
|
||||
|
||||
c4 e2 72 f7 c0 sarx %ecx,%eax,%eax
|
||||
|
||||
for strchr-avx2 family functions, require BMI2 in ifunc-impl-list.c and
|
||||
ifunc-avx2.h.
|
||||
---
|
||||
sysdeps/x86_64/multiarch/ifunc-avx2.h | 4 ++--
|
||||
sysdeps/x86_64/multiarch/ifunc-impl-list.c | 12 +++++++++---
|
||||
2 files changed, 11 insertions(+), 5 deletions(-)
|
||||
|
||||
diff --git a/sysdeps/x86_64/multiarch/ifunc-avx2.h b/sysdeps/x86_64/multiarch/ifunc-avx2.h
|
||||
index e0f30e61..ef72b73f 100644
|
||||
--- a/sysdeps/x86_64/multiarch/ifunc-avx2.h
|
||||
+++ b/sysdeps/x86_64/multiarch/ifunc-avx2.h
|
||||
@@ -30,11 +30,11 @@ IFUNC_SELECTOR (void)
|
||||
const struct cpu_features* cpu_features = __get_cpu_features ();
|
||||
|
||||
if (CPU_FEATURE_USABLE_P (cpu_features, AVX2)
|
||||
+ && CPU_FEATURE_USABLE_P (cpu_features, BMI2)
|
||||
&& CPU_FEATURES_ARCH_P (cpu_features, AVX_Fast_Unaligned_Load))
|
||||
{
|
||||
if (CPU_FEATURE_USABLE_P (cpu_features, AVX512VL)
|
||||
- && CPU_FEATURE_USABLE_P (cpu_features, AVX512BW)
|
||||
- && CPU_FEATURE_USABLE_P (cpu_features, BMI2))
|
||||
+ && CPU_FEATURE_USABLE_P (cpu_features, AVX512BW))
|
||||
return OPTIMIZE (evex);
|
||||
|
||||
if (CPU_FEATURE_USABLE_P (cpu_features, RTM))
|
||||
diff --git a/sysdeps/x86_64/multiarch/ifunc-impl-list.c b/sysdeps/x86_64/multiarch/ifunc-impl-list.c
|
||||
index 695cdba6..85b8863a 100644
|
||||
--- a/sysdeps/x86_64/multiarch/ifunc-impl-list.c
|
||||
+++ b/sysdeps/x86_64/multiarch/ifunc-impl-list.c
|
||||
@@ -400,10 +400,12 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
|
||||
/* Support sysdeps/x86_64/multiarch/strchr.c. */
|
||||
IFUNC_IMPL (i, name, strchr,
|
||||
IFUNC_IMPL_ADD (array, i, strchr,
|
||||
- CPU_FEATURE_USABLE (AVX2),
|
||||
+ (CPU_FEATURE_USABLE (AVX2)
|
||||
+ && CPU_FEATURE_USABLE (BMI2)),
|
||||
__strchr_avx2)
|
||||
IFUNC_IMPL_ADD (array, i, strchr,
|
||||
(CPU_FEATURE_USABLE (AVX2)
|
||||
+ && CPU_FEATURE_USABLE (BMI2)
|
||||
&& CPU_FEATURE_USABLE (RTM)),
|
||||
__strchr_avx2_rtm)
|
||||
IFUNC_IMPL_ADD (array, i, strchr,
|
||||
@@ -417,10 +419,12 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
|
||||
/* Support sysdeps/x86_64/multiarch/strchrnul.c. */
|
||||
IFUNC_IMPL (i, name, strchrnul,
|
||||
IFUNC_IMPL_ADD (array, i, strchrnul,
|
||||
- CPU_FEATURE_USABLE (AVX2),
|
||||
+ (CPU_FEATURE_USABLE (AVX2)
|
||||
+ && CPU_FEATURE_USABLE (BMI2)),
|
||||
__strchrnul_avx2)
|
||||
IFUNC_IMPL_ADD (array, i, strchrnul,
|
||||
(CPU_FEATURE_USABLE (AVX2)
|
||||
+ && CPU_FEATURE_USABLE (BMI2)
|
||||
&& CPU_FEATURE_USABLE (RTM)),
|
||||
__strchrnul_avx2_rtm)
|
||||
IFUNC_IMPL_ADD (array, i, strchrnul,
|
||||
@@ -574,10 +578,12 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
|
||||
/* Support sysdeps/x86_64/multiarch/wcschr.c. */
|
||||
IFUNC_IMPL (i, name, wcschr,
|
||||
IFUNC_IMPL_ADD (array, i, wcschr,
|
||||
- CPU_FEATURE_USABLE (AVX2),
|
||||
+ (CPU_FEATURE_USABLE (AVX2)
|
||||
+ && CPU_FEATURE_USABLE (BMI2)),
|
||||
__wcschr_avx2)
|
||||
IFUNC_IMPL_ADD (array, i, wcschr,
|
||||
(CPU_FEATURE_USABLE (AVX2)
|
||||
+ && CPU_FEATURE_USABLE (BMI2)
|
||||
&& CPU_FEATURE_USABLE (RTM)),
|
||||
__wcschr_avx2_rtm)
|
||||
IFUNC_IMPL_ADD (array, i, wcschr,
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,265 +0,0 @@
|
||||
From f53790272ce7bdc5ecd14b45f65d0464d2a61a3a Mon Sep 17 00:00:00 2001
|
||||
From: Noah Goldstein <goldstein.w.n@gmail.com>
|
||||
Date: Mon, 19 Apr 2021 17:48:10 -0400
|
||||
Subject: [PATCH] x86: Optimize less_vec evex and avx512
|
||||
memset-vec-unaligned-erms.S
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
No bug. This commit adds optimized cased for less_vec memset case that
|
||||
uses the avx512vl/avx512bw mask store avoiding the excessive
|
||||
branches. test-memset and test-wmemset are passing.
|
||||
|
||||
Signed-off-by: Noah Goldstein <goldstein.w.n@gmail.com>
|
||||
---
|
||||
sysdeps/x86_64/multiarch/ifunc-impl-list.c | 40 ++++++++++-----
|
||||
sysdeps/x86_64/multiarch/ifunc-memset.h | 6 ++-
|
||||
.../multiarch/memset-avx512-unaligned-erms.S | 2 +-
|
||||
.../multiarch/memset-evex-unaligned-erms.S | 2 +-
|
||||
.../multiarch/memset-vec-unaligned-erms.S | 51 +++++++++++++++----
|
||||
5 files changed, 74 insertions(+), 27 deletions(-)
|
||||
|
||||
diff --git a/sysdeps/x86_64/multiarch/ifunc-impl-list.c b/sysdeps/x86_64/multiarch/ifunc-impl-list.c
|
||||
index 85b8863a..d59d65f8 100644
|
||||
--- a/sysdeps/x86_64/multiarch/ifunc-impl-list.c
|
||||
+++ b/sysdeps/x86_64/multiarch/ifunc-impl-list.c
|
||||
@@ -204,19 +204,23 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
|
||||
__memset_chk_avx2_unaligned_erms_rtm)
|
||||
IFUNC_IMPL_ADD (array, i, __memset_chk,
|
||||
(CPU_FEATURE_USABLE (AVX512VL)
|
||||
- && CPU_FEATURE_USABLE (AVX512BW)),
|
||||
+ && CPU_FEATURE_USABLE (AVX512BW)
|
||||
+ && CPU_FEATURE_USABLE (BMI2)),
|
||||
__memset_chk_evex_unaligned)
|
||||
IFUNC_IMPL_ADD (array, i, __memset_chk,
|
||||
(CPU_FEATURE_USABLE (AVX512VL)
|
||||
- && CPU_FEATURE_USABLE (AVX512BW)),
|
||||
+ && CPU_FEATURE_USABLE (AVX512BW)
|
||||
+ && CPU_FEATURE_USABLE (BMI2)),
|
||||
__memset_chk_evex_unaligned_erms)
|
||||
IFUNC_IMPL_ADD (array, i, __memset_chk,
|
||||
(CPU_FEATURE_USABLE (AVX512VL)
|
||||
- && CPU_FEATURE_USABLE (AVX512BW)),
|
||||
+ && CPU_FEATURE_USABLE (AVX512BW)
|
||||
+ && CPU_FEATURE_USABLE (BMI2)),
|
||||
__memset_chk_avx512_unaligned_erms)
|
||||
IFUNC_IMPL_ADD (array, i, __memset_chk,
|
||||
(CPU_FEATURE_USABLE (AVX512VL)
|
||||
- && CPU_FEATURE_USABLE (AVX512BW)),
|
||||
+ && CPU_FEATURE_USABLE (AVX512BW)
|
||||
+ && CPU_FEATURE_USABLE (BMI2)),
|
||||
__memset_chk_avx512_unaligned)
|
||||
IFUNC_IMPL_ADD (array, i, __memset_chk,
|
||||
CPU_FEATURE_USABLE (AVX512F),
|
||||
@@ -247,19 +251,23 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
|
||||
__memset_avx2_unaligned_erms_rtm)
|
||||
IFUNC_IMPL_ADD (array, i, memset,
|
||||
(CPU_FEATURE_USABLE (AVX512VL)
|
||||
- && CPU_FEATURE_USABLE (AVX512BW)),
|
||||
+ && CPU_FEATURE_USABLE (AVX512BW)
|
||||
+ && CPU_FEATURE_USABLE (BMI2)),
|
||||
__memset_evex_unaligned)
|
||||
IFUNC_IMPL_ADD (array, i, memset,
|
||||
(CPU_FEATURE_USABLE (AVX512VL)
|
||||
- && CPU_FEATURE_USABLE (AVX512BW)),
|
||||
+ && CPU_FEATURE_USABLE (AVX512BW)
|
||||
+ && CPU_FEATURE_USABLE (BMI2)),
|
||||
__memset_evex_unaligned_erms)
|
||||
IFUNC_IMPL_ADD (array, i, memset,
|
||||
(CPU_FEATURE_USABLE (AVX512VL)
|
||||
- && CPU_FEATURE_USABLE (AVX512BW)),
|
||||
+ && CPU_FEATURE_USABLE (AVX512BW)
|
||||
+ && CPU_FEATURE_USABLE (BMI2)),
|
||||
__memset_avx512_unaligned_erms)
|
||||
IFUNC_IMPL_ADD (array, i, memset,
|
||||
(CPU_FEATURE_USABLE (AVX512VL)
|
||||
- && CPU_FEATURE_USABLE (AVX512BW)),
|
||||
+ && CPU_FEATURE_USABLE (AVX512BW)
|
||||
+ && CPU_FEATURE_USABLE (BMI2)),
|
||||
__memset_avx512_unaligned)
|
||||
IFUNC_IMPL_ADD (array, i, memset,
|
||||
CPU_FEATURE_USABLE (AVX512F),
|
||||
@@ -739,10 +747,14 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
|
||||
&& CPU_FEATURE_USABLE (RTM)),
|
||||
__wmemset_avx2_unaligned_rtm)
|
||||
IFUNC_IMPL_ADD (array, i, wmemset,
|
||||
- CPU_FEATURE_USABLE (AVX512VL),
|
||||
+ (CPU_FEATURE_USABLE (AVX512VL)
|
||||
+ && CPU_FEATURE_USABLE (AVX512BW)
|
||||
+ && CPU_FEATURE_USABLE (BMI2)),
|
||||
__wmemset_evex_unaligned)
|
||||
IFUNC_IMPL_ADD (array, i, wmemset,
|
||||
- CPU_FEATURE_USABLE (AVX512VL),
|
||||
+ (CPU_FEATURE_USABLE (AVX512VL)
|
||||
+ && CPU_FEATURE_USABLE (AVX512BW)
|
||||
+ && CPU_FEATURE_USABLE (BMI2)),
|
||||
__wmemset_avx512_unaligned))
|
||||
|
||||
#ifdef SHARED
|
||||
@@ -946,10 +958,14 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
|
||||
CPU_FEATURE_USABLE (AVX2),
|
||||
__wmemset_chk_avx2_unaligned)
|
||||
IFUNC_IMPL_ADD (array, i, __wmemset_chk,
|
||||
- CPU_FEATURE_USABLE (AVX512VL),
|
||||
+ (CPU_FEATURE_USABLE (AVX512VL)
|
||||
+ && CPU_FEATURE_USABLE (AVX512BW)
|
||||
+ && CPU_FEATURE_USABLE (BMI2)),
|
||||
__wmemset_chk_evex_unaligned)
|
||||
IFUNC_IMPL_ADD (array, i, __wmemset_chk,
|
||||
- CPU_FEATURE_USABLE (AVX512F),
|
||||
+ (CPU_FEATURE_USABLE (AVX512VL)
|
||||
+ && CPU_FEATURE_USABLE (AVX512BW)
|
||||
+ && CPU_FEATURE_USABLE (BMI2)),
|
||||
__wmemset_chk_avx512_unaligned))
|
||||
#endif
|
||||
|
||||
diff --git a/sysdeps/x86_64/multiarch/ifunc-memset.h b/sysdeps/x86_64/multiarch/ifunc-memset.h
|
||||
index 19795938..100e3707 100644
|
||||
--- a/sysdeps/x86_64/multiarch/ifunc-memset.h
|
||||
+++ b/sysdeps/x86_64/multiarch/ifunc-memset.h
|
||||
@@ -54,7 +54,8 @@ IFUNC_SELECTOR (void)
|
||||
&& !CPU_FEATURES_ARCH_P (cpu_features, Prefer_No_AVX512))
|
||||
{
|
||||
if (CPU_FEATURE_USABLE_P (cpu_features, AVX512VL)
|
||||
- && CPU_FEATURE_USABLE_P (cpu_features, AVX512BW))
|
||||
+ && CPU_FEATURE_USABLE_P (cpu_features, AVX512BW)
|
||||
+ && CPU_FEATURE_USABLE_P (cpu_features, BMI2))
|
||||
{
|
||||
if (CPU_FEATURE_USABLE_P (cpu_features, ERMS))
|
||||
return OPTIMIZE (avx512_unaligned_erms);
|
||||
@@ -68,7 +69,8 @@ IFUNC_SELECTOR (void)
|
||||
if (CPU_FEATURE_USABLE_P (cpu_features, AVX2))
|
||||
{
|
||||
if (CPU_FEATURE_USABLE_P (cpu_features, AVX512VL)
|
||||
- && CPU_FEATURE_USABLE_P (cpu_features, AVX512BW))
|
||||
+ && CPU_FEATURE_USABLE_P (cpu_features, AVX512BW)
|
||||
+ && CPU_FEATURE_USABLE_P (cpu_features, BMI2))
|
||||
{
|
||||
if (CPU_FEATURE_USABLE_P (cpu_features, ERMS))
|
||||
return OPTIMIZE (evex_unaligned_erms);
|
||||
diff --git a/sysdeps/x86_64/multiarch/memset-avx512-unaligned-erms.S b/sysdeps/x86_64/multiarch/memset-avx512-unaligned-erms.S
|
||||
index 22e7b187..8ad842fc 100644
|
||||
--- a/sysdeps/x86_64/multiarch/memset-avx512-unaligned-erms.S
|
||||
+++ b/sysdeps/x86_64/multiarch/memset-avx512-unaligned-erms.S
|
||||
@@ -19,6 +19,6 @@
|
||||
# define SECTION(p) p##.evex512
|
||||
# define MEMSET_SYMBOL(p,s) p##_avx512_##s
|
||||
# define WMEMSET_SYMBOL(p,s) p##_avx512_##s
|
||||
-
|
||||
+# define USE_LESS_VEC_MASK_STORE 1
|
||||
# include "memset-vec-unaligned-erms.S"
|
||||
#endif
|
||||
diff --git a/sysdeps/x86_64/multiarch/memset-evex-unaligned-erms.S b/sysdeps/x86_64/multiarch/memset-evex-unaligned-erms.S
|
||||
index ae0a4d6e..640f0929 100644
|
||||
--- a/sysdeps/x86_64/multiarch/memset-evex-unaligned-erms.S
|
||||
+++ b/sysdeps/x86_64/multiarch/memset-evex-unaligned-erms.S
|
||||
@@ -19,6 +19,6 @@
|
||||
# define SECTION(p) p##.evex
|
||||
# define MEMSET_SYMBOL(p,s) p##_evex_##s
|
||||
# define WMEMSET_SYMBOL(p,s) p##_evex_##s
|
||||
-
|
||||
+# define USE_LESS_VEC_MASK_STORE 1
|
||||
# include "memset-vec-unaligned-erms.S"
|
||||
#endif
|
||||
diff --git a/sysdeps/x86_64/multiarch/memset-vec-unaligned-erms.S b/sysdeps/x86_64/multiarch/memset-vec-unaligned-erms.S
|
||||
index bae5cba4..f877ac9d 100644
|
||||
--- a/sysdeps/x86_64/multiarch/memset-vec-unaligned-erms.S
|
||||
+++ b/sysdeps/x86_64/multiarch/memset-vec-unaligned-erms.S
|
||||
@@ -63,6 +63,8 @@
|
||||
# endif
|
||||
#endif
|
||||
|
||||
+#define PAGE_SIZE 4096
|
||||
+
|
||||
#ifndef SECTION
|
||||
# error SECTION is not defined!
|
||||
#endif
|
||||
@@ -213,11 +215,38 @@ L(loop):
|
||||
cmpq %rcx, %rdx
|
||||
jne L(loop)
|
||||
VZEROUPPER_SHORT_RETURN
|
||||
+
|
||||
+ .p2align 4
|
||||
L(less_vec):
|
||||
/* Less than 1 VEC. */
|
||||
# if VEC_SIZE != 16 && VEC_SIZE != 32 && VEC_SIZE != 64
|
||||
# error Unsupported VEC_SIZE!
|
||||
# endif
|
||||
+# ifdef USE_LESS_VEC_MASK_STORE
|
||||
+ /* Clear high bits from edi. Only keeping bits relevant to page
|
||||
+ cross check. Note that we are using rax which is set in
|
||||
+ MEMSET_VDUP_TO_VEC0_AND_SET_RETURN as ptr from here on out.
|
||||
+ */
|
||||
+ andl $(PAGE_SIZE - 1), %edi
|
||||
+ /* Check if VEC_SIZE store cross page. Mask stores suffer serious
|
||||
+ performance degradation when it has to fault supress. */
|
||||
+ cmpl $(PAGE_SIZE - VEC_SIZE), %edi
|
||||
+ ja L(cross_page)
|
||||
+# if VEC_SIZE > 32
|
||||
+ movq $-1, %rcx
|
||||
+ bzhiq %rdx, %rcx, %rcx
|
||||
+ kmovq %rcx, %k1
|
||||
+# else
|
||||
+ movl $-1, %ecx
|
||||
+ bzhil %edx, %ecx, %ecx
|
||||
+ kmovd %ecx, %k1
|
||||
+# endif
|
||||
+ vmovdqu8 %VEC(0), (%rax) {%k1}
|
||||
+ VZEROUPPER_RETURN
|
||||
+
|
||||
+ .p2align 4
|
||||
+L(cross_page):
|
||||
+# endif
|
||||
# if VEC_SIZE > 32
|
||||
cmpb $32, %dl
|
||||
jae L(between_32_63)
|
||||
@@ -234,36 +263,36 @@ L(less_vec):
|
||||
cmpb $1, %dl
|
||||
ja L(between_2_3)
|
||||
jb 1f
|
||||
- movb %cl, (%rdi)
|
||||
+ movb %cl, (%rax)
|
||||
1:
|
||||
VZEROUPPER_RETURN
|
||||
# if VEC_SIZE > 32
|
||||
/* From 32 to 63. No branch when size == 32. */
|
||||
L(between_32_63):
|
||||
- VMOVU %YMM0, -32(%rdi,%rdx)
|
||||
- VMOVU %YMM0, (%rdi)
|
||||
+ VMOVU %YMM0, -32(%rax,%rdx)
|
||||
+ VMOVU %YMM0, (%rax)
|
||||
VZEROUPPER_RETURN
|
||||
# endif
|
||||
# if VEC_SIZE > 16
|
||||
/* From 16 to 31. No branch when size == 16. */
|
||||
L(between_16_31):
|
||||
- VMOVU %XMM0, -16(%rdi,%rdx)
|
||||
- VMOVU %XMM0, (%rdi)
|
||||
+ VMOVU %XMM0, -16(%rax,%rdx)
|
||||
+ VMOVU %XMM0, (%rax)
|
||||
VZEROUPPER_RETURN
|
||||
# endif
|
||||
/* From 8 to 15. No branch when size == 8. */
|
||||
L(between_8_15):
|
||||
- movq %rcx, -8(%rdi,%rdx)
|
||||
- movq %rcx, (%rdi)
|
||||
+ movq %rcx, -8(%rax,%rdx)
|
||||
+ movq %rcx, (%rax)
|
||||
VZEROUPPER_RETURN
|
||||
L(between_4_7):
|
||||
/* From 4 to 7. No branch when size == 4. */
|
||||
- movl %ecx, -4(%rdi,%rdx)
|
||||
- movl %ecx, (%rdi)
|
||||
+ movl %ecx, -4(%rax,%rdx)
|
||||
+ movl %ecx, (%rax)
|
||||
VZEROUPPER_RETURN
|
||||
L(between_2_3):
|
||||
/* From 2 to 3. No branch when size == 2. */
|
||||
- movw %cx, -2(%rdi,%rdx)
|
||||
- movw %cx, (%rdi)
|
||||
+ movw %cx, -2(%rax,%rdx)
|
||||
+ movw %cx, (%rax)
|
||||
VZEROUPPER_RETURN
|
||||
END (MEMSET_SYMBOL (__memset, unaligned_erms))
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,396 +0,0 @@
|
||||
From ccabe7971f508709d034b63b8672f6f751a3d356 Mon Sep 17 00:00:00 2001
|
||||
From: Noah Goldstein <goldstein.w.n@gmail.com>
|
||||
Date: Fri, 23 Apr 2021 15:56:24 -0400
|
||||
Subject: [PATCH] x86: Optimize strchr-avx2.S
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
No bug. This commit optimizes strchr-avx2.S. The optimizations are all
|
||||
small things such as save an ALU in the alignment process, saving a
|
||||
few instructions in the loop return, saving some bytes in the main
|
||||
loop, and increasing the ILP in the return cases. test-strchr,
|
||||
test-strchrnul, test-wcschr, and test-wcschrnul are all passing.
|
||||
|
||||
Signed-off-by: Noah Goldstein <goldstein.w.n@gmail.com>
|
||||
---
|
||||
sysdeps/x86_64/multiarch/strchr-avx2.S | 290 +++++++++++++++----------
|
||||
1 file changed, 170 insertions(+), 120 deletions(-)
|
||||
|
||||
Conflics:
|
||||
sysdeps/x86_64/multiarch/strchr-avx2.S
|
||||
(rearranged to account for branch changes)
|
||||
|
||||
diff --git a/sysdeps/x86_64/multiarch/strchr-avx2.S b/sysdeps/x86_64/multiarch/strchr-avx2.S
|
||||
index 919d256c..5884726b 100644
|
||||
--- a/sysdeps/x86_64/multiarch/strchr-avx2.S
|
||||
+++ b/sysdeps/x86_64/multiarch/strchr-avx2.S
|
||||
@@ -49,133 +49,144 @@
|
||||
|
||||
.section SECTION(.text),"ax",@progbits
|
||||
ENTRY (STRCHR)
|
||||
- movl %edi, %ecx
|
||||
-# ifndef USE_AS_STRCHRNUL
|
||||
- xorl %edx, %edx
|
||||
-# endif
|
||||
-
|
||||
/* Broadcast CHAR to YMM0. */
|
||||
vmovd %esi, %xmm0
|
||||
+ movl %edi, %eax
|
||||
+ andl $(PAGE_SIZE - 1), %eax
|
||||
+ VPBROADCAST %xmm0, %ymm0
|
||||
vpxor %xmm9, %xmm9, %xmm9
|
||||
- VPBROADCAST %xmm0, %ymm0
|
||||
|
||||
/* Check if we cross page boundary with one vector load. */
|
||||
- andl $(PAGE_SIZE - 1), %ecx
|
||||
- cmpl $(PAGE_SIZE - VEC_SIZE), %ecx
|
||||
- ja L(cross_page_boundary)
|
||||
+ cmpl $(PAGE_SIZE - VEC_SIZE), %eax
|
||||
+ ja L(cross_page_boundary)
|
||||
|
||||
/* Check the first VEC_SIZE bytes. Search for both CHAR and the
|
||||
null byte. */
|
||||
vmovdqu (%rdi), %ymm8
|
||||
- VPCMPEQ %ymm8, %ymm0, %ymm1
|
||||
- VPCMPEQ %ymm8, %ymm9, %ymm2
|
||||
+ VPCMPEQ %ymm8, %ymm0, %ymm1
|
||||
+ VPCMPEQ %ymm8, %ymm9, %ymm2
|
||||
vpor %ymm1, %ymm2, %ymm1
|
||||
vpmovmskb %ymm1, %eax
|
||||
testl %eax, %eax
|
||||
- jz L(more_vecs)
|
||||
+ jz L(aligned_more)
|
||||
tzcntl %eax, %eax
|
||||
+# ifndef USE_AS_STRCHRNUL
|
||||
/* Found CHAR or the null byte. */
|
||||
+ cmp (%rdi, %rax), %CHAR_REG
|
||||
+ jne L(zero)
|
||||
+# endif
|
||||
addq %rdi, %rax
|
||||
+ VZEROUPPER_RETURN
|
||||
+
|
||||
+ /* .p2align 5 helps keep performance more consistent if ENTRY()
|
||||
+ alignment % 32 was either 16 or 0. As well this makes the
|
||||
+ alignment % 32 of the loop_4x_vec fixed which makes tuning it
|
||||
+ easier. */
|
||||
+ .p2align 5
|
||||
+L(first_vec_x4):
|
||||
+ tzcntl %eax, %eax
|
||||
+ addq $(VEC_SIZE * 3 + 1), %rdi
|
||||
# ifndef USE_AS_STRCHRNUL
|
||||
- cmp (%rax), %CHAR_REG
|
||||
- cmovne %rdx, %rax
|
||||
+ /* Found CHAR or the null byte. */
|
||||
+ cmp (%rdi, %rax), %CHAR_REG
|
||||
+ jne L(zero)
|
||||
# endif
|
||||
-L(return_vzeroupper):
|
||||
- ZERO_UPPER_VEC_REGISTERS_RETURN
|
||||
-
|
||||
- .p2align 4
|
||||
-L(more_vecs):
|
||||
- /* Align data for aligned loads in the loop. */
|
||||
- andq $-VEC_SIZE, %rdi
|
||||
-L(aligned_more):
|
||||
-
|
||||
- /* Check the next 4 * VEC_SIZE. Only one VEC_SIZE at a time
|
||||
- since data is only aligned to VEC_SIZE. */
|
||||
- vmovdqa VEC_SIZE(%rdi), %ymm8
|
||||
- addq $VEC_SIZE, %rdi
|
||||
- VPCMPEQ %ymm8, %ymm0, %ymm1
|
||||
- VPCMPEQ %ymm8, %ymm9, %ymm2
|
||||
- vpor %ymm1, %ymm2, %ymm1
|
||||
- vpmovmskb %ymm1, %eax
|
||||
- testl %eax, %eax
|
||||
- jnz L(first_vec_x0)
|
||||
-
|
||||
- vmovdqa VEC_SIZE(%rdi), %ymm8
|
||||
- VPCMPEQ %ymm8, %ymm0, %ymm1
|
||||
- VPCMPEQ %ymm8, %ymm9, %ymm2
|
||||
- vpor %ymm1, %ymm2, %ymm1
|
||||
- vpmovmskb %ymm1, %eax
|
||||
- testl %eax, %eax
|
||||
- jnz L(first_vec_x1)
|
||||
-
|
||||
- vmovdqa (VEC_SIZE * 2)(%rdi), %ymm8
|
||||
- VPCMPEQ %ymm8, %ymm0, %ymm1
|
||||
- VPCMPEQ %ymm8, %ymm9, %ymm2
|
||||
- vpor %ymm1, %ymm2, %ymm1
|
||||
- vpmovmskb %ymm1, %eax
|
||||
- testl %eax, %eax
|
||||
- jnz L(first_vec_x2)
|
||||
-
|
||||
- vmovdqa (VEC_SIZE * 3)(%rdi), %ymm8
|
||||
- VPCMPEQ %ymm8, %ymm0, %ymm1
|
||||
- VPCMPEQ %ymm8, %ymm9, %ymm2
|
||||
- vpor %ymm1, %ymm2, %ymm1
|
||||
- vpmovmskb %ymm1, %eax
|
||||
- testl %eax, %eax
|
||||
- jz L(prep_loop_4x)
|
||||
+ addq %rdi, %rax
|
||||
+ VZEROUPPER_RETURN
|
||||
|
||||
- tzcntl %eax, %eax
|
||||
- leaq (VEC_SIZE * 3)(%rdi, %rax), %rax
|
||||
# ifndef USE_AS_STRCHRNUL
|
||||
- cmp (%rax), %CHAR_REG
|
||||
- cmovne %rdx, %rax
|
||||
+L(zero):
|
||||
+ xorl %eax, %eax
|
||||
+ VZEROUPPER_RETURN
|
||||
# endif
|
||||
- VZEROUPPER
|
||||
- ret
|
||||
+
|
||||
|
||||
.p2align 4
|
||||
-L(first_vec_x0):
|
||||
+L(first_vec_x1):
|
||||
tzcntl %eax, %eax
|
||||
- /* Found CHAR or the null byte. */
|
||||
- addq %rdi, %rax
|
||||
+ incq %rdi
|
||||
# ifndef USE_AS_STRCHRNUL
|
||||
- cmp (%rax), %CHAR_REG
|
||||
- cmovne %rdx, %rax
|
||||
+ /* Found CHAR or the null byte. */
|
||||
+ cmp (%rdi, %rax), %CHAR_REG
|
||||
+ jne L(zero)
|
||||
# endif
|
||||
+ addq %rdi, %rax
|
||||
VZEROUPPER_RETURN
|
||||
|
||||
.p2align 4
|
||||
-L(first_vec_x1):
|
||||
+L(first_vec_x2):
|
||||
tzcntl %eax, %eax
|
||||
- leaq VEC_SIZE(%rdi, %rax), %rax
|
||||
+ addq $(VEC_SIZE + 1), %rdi
|
||||
# ifndef USE_AS_STRCHRNUL
|
||||
- cmp (%rax), %CHAR_REG
|
||||
- cmovne %rdx, %rax
|
||||
+ /* Found CHAR or the null byte. */
|
||||
+ cmp (%rdi, %rax), %CHAR_REG
|
||||
+ jne L(zero)
|
||||
# endif
|
||||
+ addq %rdi, %rax
|
||||
VZEROUPPER_RETURN
|
||||
|
||||
.p2align 4
|
||||
-L(first_vec_x2):
|
||||
+L(first_vec_x3):
|
||||
tzcntl %eax, %eax
|
||||
- /* Found CHAR or the null byte. */
|
||||
- leaq (VEC_SIZE * 2)(%rdi, %rax), %rax
|
||||
+ addq $(VEC_SIZE * 2 + 1), %rdi
|
||||
# ifndef USE_AS_STRCHRNUL
|
||||
- cmp (%rax), %CHAR_REG
|
||||
- cmovne %rdx, %rax
|
||||
+ /* Found CHAR or the null byte. */
|
||||
+ cmp (%rdi, %rax), %CHAR_REG
|
||||
+ jne L(zero)
|
||||
# endif
|
||||
+ addq %rdi, %rax
|
||||
VZEROUPPER_RETURN
|
||||
|
||||
-L(prep_loop_4x):
|
||||
- /* Align data to 4 * VEC_SIZE. */
|
||||
- andq $-(VEC_SIZE * 4), %rdi
|
||||
+ .p2align 4
|
||||
+L(aligned_more):
|
||||
+ /* Align data to VEC_SIZE - 1. This is the same number of
|
||||
+ instructions as using andq -VEC_SIZE but saves 4 bytes of code
|
||||
+ on x4 check. */
|
||||
+ orq $(VEC_SIZE - 1), %rdi
|
||||
+L(cross_page_continue):
|
||||
+ /* Check the next 4 * VEC_SIZE. Only one VEC_SIZE at a time
|
||||
+ since data is only aligned to VEC_SIZE. */
|
||||
+ vmovdqa 1(%rdi), %ymm8
|
||||
+ VPCMPEQ %ymm8, %ymm0, %ymm1
|
||||
+ VPCMPEQ %ymm8, %ymm9, %ymm2
|
||||
+ vpor %ymm1, %ymm2, %ymm1
|
||||
+ vpmovmskb %ymm1, %eax
|
||||
+ testl %eax, %eax
|
||||
+ jnz L(first_vec_x1)
|
||||
+
|
||||
+ vmovdqa (VEC_SIZE + 1)(%rdi), %ymm8
|
||||
+ VPCMPEQ %ymm8, %ymm0, %ymm1
|
||||
+ VPCMPEQ %ymm8, %ymm9, %ymm2
|
||||
+ vpor %ymm1, %ymm2, %ymm1
|
||||
+ vpmovmskb %ymm1, %eax
|
||||
+ testl %eax, %eax
|
||||
+ jnz L(first_vec_x2)
|
||||
+
|
||||
+ vmovdqa (VEC_SIZE * 2 + 1)(%rdi), %ymm8
|
||||
+ VPCMPEQ %ymm8, %ymm0, %ymm1
|
||||
+ VPCMPEQ %ymm8, %ymm9, %ymm2
|
||||
+ vpor %ymm1, %ymm2, %ymm1
|
||||
+ vpmovmskb %ymm1, %eax
|
||||
+ testl %eax, %eax
|
||||
+ jnz L(first_vec_x3)
|
||||
|
||||
+ vmovdqa (VEC_SIZE * 3 + 1)(%rdi), %ymm8
|
||||
+ VPCMPEQ %ymm8, %ymm0, %ymm1
|
||||
+ VPCMPEQ %ymm8, %ymm9, %ymm2
|
||||
+ vpor %ymm1, %ymm2, %ymm1
|
||||
+ vpmovmskb %ymm1, %eax
|
||||
+ testl %eax, %eax
|
||||
+ jnz L(first_vec_x4)
|
||||
+ /* Align data to VEC_SIZE * 4 - 1. */
|
||||
+ addq $(VEC_SIZE * 4 + 1), %rdi
|
||||
+ andq $-(VEC_SIZE * 4), %rdi
|
||||
.p2align 4
|
||||
L(loop_4x_vec):
|
||||
/* Compare 4 * VEC at a time forward. */
|
||||
- vmovdqa (VEC_SIZE * 4)(%rdi), %ymm5
|
||||
- vmovdqa (VEC_SIZE * 5)(%rdi), %ymm6
|
||||
- vmovdqa (VEC_SIZE * 6)(%rdi), %ymm7
|
||||
- vmovdqa (VEC_SIZE * 7)(%rdi), %ymm8
|
||||
+ vmovdqa (%rdi), %ymm5
|
||||
+ vmovdqa (VEC_SIZE)(%rdi), %ymm6
|
||||
+ vmovdqa (VEC_SIZE * 2)(%rdi), %ymm7
|
||||
+ vmovdqa (VEC_SIZE * 3)(%rdi), %ymm8
|
||||
|
||||
/* Leaves only CHARS matching esi as 0. */
|
||||
vpxor %ymm5, %ymm0, %ymm1
|
||||
@@ -191,63 +202,102 @@ L(loop_4x_vec):
|
||||
VPMINU %ymm1, %ymm2, %ymm5
|
||||
VPMINU %ymm3, %ymm4, %ymm6
|
||||
|
||||
- VPMINU %ymm5, %ymm6, %ymm5
|
||||
+ VPMINU %ymm5, %ymm6, %ymm6
|
||||
|
||||
- VPCMPEQ %ymm5, %ymm9, %ymm5
|
||||
- vpmovmskb %ymm5, %eax
|
||||
+ VPCMPEQ %ymm6, %ymm9, %ymm6
|
||||
+ vpmovmskb %ymm6, %ecx
|
||||
+ subq $-(VEC_SIZE * 4), %rdi
|
||||
+ testl %ecx, %ecx
|
||||
+ jz L(loop_4x_vec)
|
||||
|
||||
- addq $(VEC_SIZE * 4), %rdi
|
||||
- testl %eax, %eax
|
||||
- jz L(loop_4x_vec)
|
||||
|
||||
- VPCMPEQ %ymm1, %ymm9, %ymm1
|
||||
+ VPCMPEQ %ymm1, %ymm9, %ymm1
|
||||
vpmovmskb %ymm1, %eax
|
||||
testl %eax, %eax
|
||||
- jnz L(first_vec_x0)
|
||||
+ jnz L(last_vec_x0)
|
||||
+
|
||||
|
||||
- VPCMPEQ %ymm2, %ymm9, %ymm2
|
||||
+ VPCMPEQ %ymm5, %ymm9, %ymm2
|
||||
vpmovmskb %ymm2, %eax
|
||||
testl %eax, %eax
|
||||
- jnz L(first_vec_x1)
|
||||
+ jnz L(last_vec_x1)
|
||||
+
|
||||
+ VPCMPEQ %ymm3, %ymm9, %ymm3
|
||||
+ vpmovmskb %ymm3, %eax
|
||||
+ /* rcx has combined result from all 4 VEC. It will only be used
|
||||
+ if the first 3 other VEC all did not contain a match. */
|
||||
+ salq $32, %rcx
|
||||
+ orq %rcx, %rax
|
||||
+ tzcntq %rax, %rax
|
||||
+ subq $(VEC_SIZE * 2), %rdi
|
||||
+# ifndef USE_AS_STRCHRNUL
|
||||
+ /* Found CHAR or the null byte. */
|
||||
+ cmp (%rdi, %rax), %CHAR_REG
|
||||
+ jne L(zero_end)
|
||||
+# endif
|
||||
+ addq %rdi, %rax
|
||||
+ VZEROUPPER_RETURN
|
||||
+
|
||||
+
|
||||
+ .p2align 4
|
||||
+L(last_vec_x0):
|
||||
+ tzcntl %eax, %eax
|
||||
+ addq $-(VEC_SIZE * 4), %rdi
|
||||
+# ifndef USE_AS_STRCHRNUL
|
||||
+ /* Found CHAR or the null byte. */
|
||||
+ cmp (%rdi, %rax), %CHAR_REG
|
||||
+ jne L(zero_end)
|
||||
+# endif
|
||||
+ addq %rdi, %rax
|
||||
+ VZEROUPPER_RETURN
|
||||
|
||||
- VPCMPEQ %ymm3, %ymm9, %ymm3
|
||||
- VPCMPEQ %ymm4, %ymm9, %ymm4
|
||||
- vpmovmskb %ymm3, %ecx
|
||||
- vpmovmskb %ymm4, %eax
|
||||
- salq $32, %rax
|
||||
- orq %rcx, %rax
|
||||
- tzcntq %rax, %rax
|
||||
- leaq (VEC_SIZE * 2)(%rdi, %rax), %rax
|
||||
# ifndef USE_AS_STRCHRNUL
|
||||
- cmp (%rax), %CHAR_REG
|
||||
- cmovne %rdx, %rax
|
||||
+L(zero_end):
|
||||
+ xorl %eax, %eax
|
||||
+ VZEROUPPER_RETURN
|
||||
# endif
|
||||
- VZEROUPPER
|
||||
- ret
|
||||
+
|
||||
+ .p2align 4
|
||||
+L(last_vec_x1):
|
||||
+ tzcntl %eax, %eax
|
||||
+ subq $(VEC_SIZE * 3), %rdi
|
||||
+# ifndef USE_AS_STRCHRNUL
|
||||
+ /* Found CHAR or the null byte. */
|
||||
+ cmp (%rdi, %rax), %CHAR_REG
|
||||
+ jne L(zero_end)
|
||||
+# endif
|
||||
+ addq %rdi, %rax
|
||||
+ VZEROUPPER_RETURN
|
||||
+
|
||||
|
||||
/* Cold case for crossing page with first load. */
|
||||
.p2align 4
|
||||
L(cross_page_boundary):
|
||||
- andq $-VEC_SIZE, %rdi
|
||||
- andl $(VEC_SIZE - 1), %ecx
|
||||
-
|
||||
- vmovdqa (%rdi), %ymm8
|
||||
- VPCMPEQ %ymm8, %ymm0, %ymm1
|
||||
- VPCMPEQ %ymm8, %ymm9, %ymm2
|
||||
+ movq %rdi, %rdx
|
||||
+ /* Align rdi to VEC_SIZE - 1. */
|
||||
+ orq $(VEC_SIZE - 1), %rdi
|
||||
+ vmovdqa -(VEC_SIZE - 1)(%rdi), %ymm8
|
||||
+ VPCMPEQ %ymm8, %ymm0, %ymm1
|
||||
+ VPCMPEQ %ymm8, %ymm9, %ymm2
|
||||
vpor %ymm1, %ymm2, %ymm1
|
||||
vpmovmskb %ymm1, %eax
|
||||
- /* Remove the leading bits. */
|
||||
- sarxl %ecx, %eax, %eax
|
||||
+ /* Remove the leading bytes. sarxl only uses bits [5:0] of COUNT
|
||||
+ so no need to manually mod edx. */
|
||||
+ sarxl %edx, %eax, %eax
|
||||
testl %eax, %eax
|
||||
- jz L(aligned_more)
|
||||
+ jz L(cross_page_continue)
|
||||
tzcntl %eax, %eax
|
||||
- addq %rcx, %rdi
|
||||
- addq %rdi, %rax
|
||||
# ifndef USE_AS_STRCHRNUL
|
||||
- cmp (%rax), %CHAR_REG
|
||||
- cmovne %rdx, %rax
|
||||
+ xorl %ecx, %ecx
|
||||
+ /* Found CHAR or the null byte. */
|
||||
+ cmp (%rdx, %rax), %CHAR_REG
|
||||
+ leaq (%rdx, %rax), %rax
|
||||
+ cmovne %rcx, %rax
|
||||
+# else
|
||||
+ addq %rdx, %rax
|
||||
# endif
|
||||
- VZEROUPPER_RETURN
|
||||
+L(return_vzeroupper):
|
||||
+ ZERO_UPPER_VEC_REGISTERS_RETURN
|
||||
|
||||
END (STRCHR)
|
||||
# endif
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,532 +0,0 @@
|
||||
From 7f3e7c262cab4e2401e4331a6ef29c428de02044 Mon Sep 17 00:00:00 2001
|
||||
From: Noah Goldstein <goldstein.w.n@gmail.com>
|
||||
Date: Fri, 23 Apr 2021 15:56:25 -0400
|
||||
Subject: [PATCH] x86: Optimize strchr-evex.S
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
No bug. This commit optimizes strchr-evex.S. The optimizations are
|
||||
mostly small things such as save an ALU in the alignment process,
|
||||
saving a few instructions in the loop return. The one significant
|
||||
change is saving 2 instructions in the 4x loop. test-strchr,
|
||||
test-strchrnul, test-wcschr, and test-wcschrnul are all passing.
|
||||
|
||||
Signed-off-by: Noah Goldstein <goldstein.w.n@gmail.com>
|
||||
---
|
||||
sysdeps/x86_64/multiarch/strchr-evex.S | 392 ++++++++++++++-----------
|
||||
1 file changed, 218 insertions(+), 174 deletions(-)
|
||||
|
||||
diff --git a/sysdeps/x86_64/multiarch/strchr-evex.S b/sysdeps/x86_64/multiarch/strchr-evex.S
|
||||
index ddc86a70..7f9d4ee4 100644
|
||||
--- a/sysdeps/x86_64/multiarch/strchr-evex.S
|
||||
+++ b/sysdeps/x86_64/multiarch/strchr-evex.S
|
||||
@@ -32,13 +32,15 @@
|
||||
# define VPCMP vpcmpd
|
||||
# define VPMINU vpminud
|
||||
# define CHAR_REG esi
|
||||
-# define SHIFT_REG r8d
|
||||
+# define SHIFT_REG ecx
|
||||
+# define CHAR_SIZE 4
|
||||
# else
|
||||
# define VPBROADCAST vpbroadcastb
|
||||
# define VPCMP vpcmpb
|
||||
# define VPMINU vpminub
|
||||
# define CHAR_REG sil
|
||||
-# define SHIFT_REG ecx
|
||||
+# define SHIFT_REG edx
|
||||
+# define CHAR_SIZE 1
|
||||
# endif
|
||||
|
||||
# define XMMZERO xmm16
|
||||
@@ -56,23 +58,20 @@
|
||||
|
||||
# define VEC_SIZE 32
|
||||
# define PAGE_SIZE 4096
|
||||
+# define CHAR_PER_VEC (VEC_SIZE / CHAR_SIZE)
|
||||
|
||||
.section .text.evex,"ax",@progbits
|
||||
ENTRY (STRCHR)
|
||||
- movl %edi, %ecx
|
||||
-# ifndef USE_AS_STRCHRNUL
|
||||
- xorl %edx, %edx
|
||||
-# endif
|
||||
-
|
||||
/* Broadcast CHAR to YMM0. */
|
||||
- VPBROADCAST %esi, %YMM0
|
||||
-
|
||||
+ VPBROADCAST %esi, %YMM0
|
||||
+ movl %edi, %eax
|
||||
+ andl $(PAGE_SIZE - 1), %eax
|
||||
vpxorq %XMMZERO, %XMMZERO, %XMMZERO
|
||||
|
||||
- /* Check if we cross page boundary with one vector load. */
|
||||
- andl $(PAGE_SIZE - 1), %ecx
|
||||
- cmpl $(PAGE_SIZE - VEC_SIZE), %ecx
|
||||
- ja L(cross_page_boundary)
|
||||
+ /* Check if we cross page boundary with one vector load.
|
||||
+ Otherwise it is safe to use an unaligned load. */
|
||||
+ cmpl $(PAGE_SIZE - VEC_SIZE), %eax
|
||||
+ ja L(cross_page_boundary)
|
||||
|
||||
/* Check the first VEC_SIZE bytes. Search for both CHAR and the
|
||||
null bytes. */
|
||||
@@ -83,251 +82,296 @@ ENTRY (STRCHR)
|
||||
VPMINU %YMM2, %YMM1, %YMM2
|
||||
/* Each bit in K0 represents a CHAR or a null byte in YMM1. */
|
||||
VPCMP $0, %YMMZERO, %YMM2, %k0
|
||||
- ktestd %k0, %k0
|
||||
- jz L(more_vecs)
|
||||
kmovd %k0, %eax
|
||||
+ testl %eax, %eax
|
||||
+ jz L(aligned_more)
|
||||
tzcntl %eax, %eax
|
||||
- /* Found CHAR or the null byte. */
|
||||
# ifdef USE_AS_WCSCHR
|
||||
- /* NB: Multiply wchar_t count by 4 to get the number of bytes. */
|
||||
- leaq (%rdi, %rax, 4), %rax
|
||||
+ /* NB: Multiply wchar_t count by 4 to get the number of bytes.
|
||||
+ */
|
||||
+ leaq (%rdi, %rax, CHAR_SIZE), %rax
|
||||
# else
|
||||
addq %rdi, %rax
|
||||
# endif
|
||||
# ifndef USE_AS_STRCHRNUL
|
||||
- cmp (%rax), %CHAR_REG
|
||||
- cmovne %rdx, %rax
|
||||
+ /* Found CHAR or the null byte. */
|
||||
+ cmp (%rax), %CHAR_REG
|
||||
+ jne L(zero)
|
||||
# endif
|
||||
ret
|
||||
|
||||
- .p2align 4
|
||||
-L(more_vecs):
|
||||
- /* Align data for aligned loads in the loop. */
|
||||
- andq $-VEC_SIZE, %rdi
|
||||
-L(aligned_more):
|
||||
-
|
||||
- /* Check the next 4 * VEC_SIZE. Only one VEC_SIZE at a time
|
||||
- since data is only aligned to VEC_SIZE. */
|
||||
- VMOVA VEC_SIZE(%rdi), %YMM1
|
||||
- addq $VEC_SIZE, %rdi
|
||||
-
|
||||
- /* Leaves only CHARS matching esi as 0. */
|
||||
- vpxorq %YMM1, %YMM0, %YMM2
|
||||
- VPMINU %YMM2, %YMM1, %YMM2
|
||||
- /* Each bit in K0 represents a CHAR or a null byte in YMM1. */
|
||||
- VPCMP $0, %YMMZERO, %YMM2, %k0
|
||||
- kmovd %k0, %eax
|
||||
- testl %eax, %eax
|
||||
- jnz L(first_vec_x0)
|
||||
-
|
||||
- VMOVA VEC_SIZE(%rdi), %YMM1
|
||||
- /* Leaves only CHARS matching esi as 0. */
|
||||
- vpxorq %YMM1, %YMM0, %YMM2
|
||||
- VPMINU %YMM2, %YMM1, %YMM2
|
||||
- /* Each bit in K0 represents a CHAR or a null byte in YMM1. */
|
||||
- VPCMP $0, %YMMZERO, %YMM2, %k0
|
||||
- kmovd %k0, %eax
|
||||
- testl %eax, %eax
|
||||
- jnz L(first_vec_x1)
|
||||
-
|
||||
- VMOVA (VEC_SIZE * 2)(%rdi), %YMM1
|
||||
- /* Leaves only CHARS matching esi as 0. */
|
||||
- vpxorq %YMM1, %YMM0, %YMM2
|
||||
- VPMINU %YMM2, %YMM1, %YMM2
|
||||
- /* Each bit in K0 represents a CHAR or a null byte in YMM1. */
|
||||
- VPCMP $0, %YMMZERO, %YMM2, %k0
|
||||
- kmovd %k0, %eax
|
||||
- testl %eax, %eax
|
||||
- jnz L(first_vec_x2)
|
||||
-
|
||||
- VMOVA (VEC_SIZE * 3)(%rdi), %YMM1
|
||||
- /* Leaves only CHARS matching esi as 0. */
|
||||
- vpxorq %YMM1, %YMM0, %YMM2
|
||||
- VPMINU %YMM2, %YMM1, %YMM2
|
||||
- /* Each bit in K0 represents a CHAR or a null byte in YMM1. */
|
||||
- VPCMP $0, %YMMZERO, %YMM2, %k0
|
||||
- ktestd %k0, %k0
|
||||
- jz L(prep_loop_4x)
|
||||
-
|
||||
- kmovd %k0, %eax
|
||||
+ /* .p2align 5 helps keep performance more consistent if ENTRY()
|
||||
+ alignment % 32 was either 16 or 0. As well this makes the
|
||||
+ alignment % 32 of the loop_4x_vec fixed which makes tuning it
|
||||
+ easier. */
|
||||
+ .p2align 5
|
||||
+L(first_vec_x3):
|
||||
tzcntl %eax, %eax
|
||||
+# ifndef USE_AS_STRCHRNUL
|
||||
/* Found CHAR or the null byte. */
|
||||
-# ifdef USE_AS_WCSCHR
|
||||
- /* NB: Multiply wchar_t count by 4 to get the number of bytes. */
|
||||
- leaq (VEC_SIZE * 3)(%rdi, %rax, 4), %rax
|
||||
-# else
|
||||
- leaq (VEC_SIZE * 3)(%rdi, %rax), %rax
|
||||
+ cmp (VEC_SIZE * 3)(%rdi, %rax, CHAR_SIZE), %CHAR_REG
|
||||
+ jne L(zero)
|
||||
# endif
|
||||
+ /* NB: Multiply sizeof char type (1 or 4) to get the number of
|
||||
+ bytes. */
|
||||
+ leaq (VEC_SIZE * 3)(%rdi, %rax, CHAR_SIZE), %rax
|
||||
+ ret
|
||||
+
|
||||
# ifndef USE_AS_STRCHRNUL
|
||||
- cmp (%rax), %CHAR_REG
|
||||
- cmovne %rdx, %rax
|
||||
-# endif
|
||||
+L(zero):
|
||||
+ xorl %eax, %eax
|
||||
ret
|
||||
+# endif
|
||||
|
||||
.p2align 4
|
||||
-L(first_vec_x0):
|
||||
+L(first_vec_x4):
|
||||
+# ifndef USE_AS_STRCHRNUL
|
||||
+ /* Check to see if first match was CHAR (k0) or null (k1). */
|
||||
+ kmovd %k0, %eax
|
||||
tzcntl %eax, %eax
|
||||
- /* Found CHAR or the null byte. */
|
||||
-# ifdef USE_AS_WCSCHR
|
||||
- /* NB: Multiply wchar_t count by 4 to get the number of bytes. */
|
||||
- leaq (%rdi, %rax, 4), %rax
|
||||
+ kmovd %k1, %ecx
|
||||
+ /* bzhil will not be 0 if first match was null. */
|
||||
+ bzhil %eax, %ecx, %ecx
|
||||
+ jne L(zero)
|
||||
# else
|
||||
- addq %rdi, %rax
|
||||
-# endif
|
||||
-# ifndef USE_AS_STRCHRNUL
|
||||
- cmp (%rax), %CHAR_REG
|
||||
- cmovne %rdx, %rax
|
||||
+ /* Combine CHAR and null matches. */
|
||||
+ kord %k0, %k1, %k0
|
||||
+ kmovd %k0, %eax
|
||||
+ tzcntl %eax, %eax
|
||||
# endif
|
||||
+ /* NB: Multiply sizeof char type (1 or 4) to get the number of
|
||||
+ bytes. */
|
||||
+ leaq (VEC_SIZE * 4)(%rdi, %rax, CHAR_SIZE), %rax
|
||||
ret
|
||||
|
||||
.p2align 4
|
||||
L(first_vec_x1):
|
||||
tzcntl %eax, %eax
|
||||
- /* Found CHAR or the null byte. */
|
||||
-# ifdef USE_AS_WCSCHR
|
||||
- /* NB: Multiply wchar_t count by 4 to get the number of bytes. */
|
||||
- leaq VEC_SIZE(%rdi, %rax, 4), %rax
|
||||
-# else
|
||||
- leaq VEC_SIZE(%rdi, %rax), %rax
|
||||
-# endif
|
||||
# ifndef USE_AS_STRCHRNUL
|
||||
- cmp (%rax), %CHAR_REG
|
||||
- cmovne %rdx, %rax
|
||||
+ /* Found CHAR or the null byte. */
|
||||
+ cmp (VEC_SIZE)(%rdi, %rax, CHAR_SIZE), %CHAR_REG
|
||||
+ jne L(zero)
|
||||
+
|
||||
# endif
|
||||
+ /* NB: Multiply sizeof char type (1 or 4) to get the number of
|
||||
+ bytes. */
|
||||
+ leaq (VEC_SIZE)(%rdi, %rax, CHAR_SIZE), %rax
|
||||
ret
|
||||
|
||||
.p2align 4
|
||||
L(first_vec_x2):
|
||||
+# ifndef USE_AS_STRCHRNUL
|
||||
+ /* Check to see if first match was CHAR (k0) or null (k1). */
|
||||
+ kmovd %k0, %eax
|
||||
tzcntl %eax, %eax
|
||||
- /* Found CHAR or the null byte. */
|
||||
-# ifdef USE_AS_WCSCHR
|
||||
- /* NB: Multiply wchar_t count by 4 to get the number of bytes. */
|
||||
- leaq (VEC_SIZE * 2)(%rdi, %rax, 4), %rax
|
||||
+ kmovd %k1, %ecx
|
||||
+ /* bzhil will not be 0 if first match was null. */
|
||||
+ bzhil %eax, %ecx, %ecx
|
||||
+ jne L(zero)
|
||||
# else
|
||||
- leaq (VEC_SIZE * 2)(%rdi, %rax), %rax
|
||||
-# endif
|
||||
-# ifndef USE_AS_STRCHRNUL
|
||||
- cmp (%rax), %CHAR_REG
|
||||
- cmovne %rdx, %rax
|
||||
+ /* Combine CHAR and null matches. */
|
||||
+ kord %k0, %k1, %k0
|
||||
+ kmovd %k0, %eax
|
||||
+ tzcntl %eax, %eax
|
||||
# endif
|
||||
+ /* NB: Multiply sizeof char type (1 or 4) to get the number of
|
||||
+ bytes. */
|
||||
+ leaq (VEC_SIZE * 2)(%rdi, %rax, CHAR_SIZE), %rax
|
||||
ret
|
||||
|
||||
-L(prep_loop_4x):
|
||||
- /* Align data to 4 * VEC_SIZE. */
|
||||
+ .p2align 4
|
||||
+L(aligned_more):
|
||||
+ /* Align data to VEC_SIZE. */
|
||||
+ andq $-VEC_SIZE, %rdi
|
||||
+L(cross_page_continue):
|
||||
+ /* Check the next 4 * VEC_SIZE. Only one VEC_SIZE at a time since
|
||||
+ data is only aligned to VEC_SIZE. Use two alternating methods
|
||||
+ for checking VEC to balance latency and port contention. */
|
||||
+
|
||||
+ /* This method has higher latency but has better port
|
||||
+ distribution. */
|
||||
+ VMOVA (VEC_SIZE)(%rdi), %YMM1
|
||||
+ /* Leaves only CHARS matching esi as 0. */
|
||||
+ vpxorq %YMM1, %YMM0, %YMM2
|
||||
+ VPMINU %YMM2, %YMM1, %YMM2
|
||||
+ /* Each bit in K0 represents a CHAR or a null byte in YMM1. */
|
||||
+ VPCMP $0, %YMMZERO, %YMM2, %k0
|
||||
+ kmovd %k0, %eax
|
||||
+ testl %eax, %eax
|
||||
+ jnz L(first_vec_x1)
|
||||
+
|
||||
+ /* This method has higher latency but has better port
|
||||
+ distribution. */
|
||||
+ VMOVA (VEC_SIZE * 2)(%rdi), %YMM1
|
||||
+ /* Each bit in K0 represents a CHAR in YMM1. */
|
||||
+ VPCMP $0, %YMM1, %YMM0, %k0
|
||||
+ /* Each bit in K1 represents a CHAR in YMM1. */
|
||||
+ VPCMP $0, %YMM1, %YMMZERO, %k1
|
||||
+ kortestd %k0, %k1
|
||||
+ jnz L(first_vec_x2)
|
||||
+
|
||||
+ VMOVA (VEC_SIZE * 3)(%rdi), %YMM1
|
||||
+ /* Leaves only CHARS matching esi as 0. */
|
||||
+ vpxorq %YMM1, %YMM0, %YMM2
|
||||
+ VPMINU %YMM2, %YMM1, %YMM2
|
||||
+ /* Each bit in K0 represents a CHAR or a null byte in YMM1. */
|
||||
+ VPCMP $0, %YMMZERO, %YMM2, %k0
|
||||
+ kmovd %k0, %eax
|
||||
+ testl %eax, %eax
|
||||
+ jnz L(first_vec_x3)
|
||||
+
|
||||
+ VMOVA (VEC_SIZE * 4)(%rdi), %YMM1
|
||||
+ /* Each bit in K0 represents a CHAR in YMM1. */
|
||||
+ VPCMP $0, %YMM1, %YMM0, %k0
|
||||
+ /* Each bit in K1 represents a CHAR in YMM1. */
|
||||
+ VPCMP $0, %YMM1, %YMMZERO, %k1
|
||||
+ kortestd %k0, %k1
|
||||
+ jnz L(first_vec_x4)
|
||||
+
|
||||
+ /* Align data to VEC_SIZE * 4 for the loop. */
|
||||
+ addq $VEC_SIZE, %rdi
|
||||
andq $-(VEC_SIZE * 4), %rdi
|
||||
|
||||
.p2align 4
|
||||
L(loop_4x_vec):
|
||||
- /* Compare 4 * VEC at a time forward. */
|
||||
+ /* Check 4x VEC at a time. No penalty to imm32 offset with evex
|
||||
+ encoding. */
|
||||
VMOVA (VEC_SIZE * 4)(%rdi), %YMM1
|
||||
VMOVA (VEC_SIZE * 5)(%rdi), %YMM2
|
||||
VMOVA (VEC_SIZE * 6)(%rdi), %YMM3
|
||||
VMOVA (VEC_SIZE * 7)(%rdi), %YMM4
|
||||
|
||||
- /* Leaves only CHARS matching esi as 0. */
|
||||
+ /* For YMM1 and YMM3 use xor to set the CHARs matching esi to
|
||||
+ zero. */
|
||||
vpxorq %YMM1, %YMM0, %YMM5
|
||||
- vpxorq %YMM2, %YMM0, %YMM6
|
||||
+ /* For YMM2 and YMM4 cmp not equals to CHAR and store result in
|
||||
+ k register. Its possible to save either 1 or 2 instructions
|
||||
+ using cmp no equals method for either YMM1 or YMM1 and YMM3
|
||||
+ respectively but bottleneck on p5 makes it not worth it. */
|
||||
+ VPCMP $4, %YMM0, %YMM2, %k2
|
||||
vpxorq %YMM3, %YMM0, %YMM7
|
||||
- vpxorq %YMM4, %YMM0, %YMM8
|
||||
-
|
||||
- VPMINU %YMM5, %YMM1, %YMM5
|
||||
- VPMINU %YMM6, %YMM2, %YMM6
|
||||
- VPMINU %YMM7, %YMM3, %YMM7
|
||||
- VPMINU %YMM8, %YMM4, %YMM8
|
||||
-
|
||||
- VPMINU %YMM5, %YMM6, %YMM1
|
||||
- VPMINU %YMM7, %YMM8, %YMM2
|
||||
-
|
||||
- VPMINU %YMM1, %YMM2, %YMM1
|
||||
-
|
||||
- /* Each bit in K0 represents a CHAR or a null byte. */
|
||||
- VPCMP $0, %YMMZERO, %YMM1, %k0
|
||||
-
|
||||
- addq $(VEC_SIZE * 4), %rdi
|
||||
-
|
||||
- ktestd %k0, %k0
|
||||
+ VPCMP $4, %YMM0, %YMM4, %k4
|
||||
+
|
||||
+ /* Use min to select all zeros from either xor or end of string).
|
||||
+ */
|
||||
+ VPMINU %YMM1, %YMM5, %YMM1
|
||||
+ VPMINU %YMM3, %YMM7, %YMM3
|
||||
+
|
||||
+ /* Use min + zeromask to select for zeros. Since k2 and k4 will
|
||||
+ have 0 as positions that matched with CHAR which will set
|
||||
+ zero in the corresponding destination bytes in YMM2 / YMM4.
|
||||
+ */
|
||||
+ VPMINU %YMM1, %YMM2, %YMM2{%k2}{z}
|
||||
+ VPMINU %YMM3, %YMM4, %YMM4
|
||||
+ VPMINU %YMM2, %YMM4, %YMM4{%k4}{z}
|
||||
+
|
||||
+ VPCMP $0, %YMMZERO, %YMM4, %k1
|
||||
+ kmovd %k1, %ecx
|
||||
+ subq $-(VEC_SIZE * 4), %rdi
|
||||
+ testl %ecx, %ecx
|
||||
jz L(loop_4x_vec)
|
||||
|
||||
- /* Each bit in K0 represents a CHAR or a null byte in YMM1. */
|
||||
- VPCMP $0, %YMMZERO, %YMM5, %k0
|
||||
+ VPCMP $0, %YMMZERO, %YMM1, %k0
|
||||
kmovd %k0, %eax
|
||||
testl %eax, %eax
|
||||
- jnz L(first_vec_x0)
|
||||
+ jnz L(last_vec_x1)
|
||||
|
||||
- /* Each bit in K1 represents a CHAR or a null byte in YMM2. */
|
||||
- VPCMP $0, %YMMZERO, %YMM6, %k1
|
||||
- kmovd %k1, %eax
|
||||
+ VPCMP $0, %YMMZERO, %YMM2, %k0
|
||||
+ kmovd %k0, %eax
|
||||
testl %eax, %eax
|
||||
- jnz L(first_vec_x1)
|
||||
-
|
||||
- /* Each bit in K2 represents a CHAR or a null byte in YMM3. */
|
||||
- VPCMP $0, %YMMZERO, %YMM7, %k2
|
||||
- /* Each bit in K3 represents a CHAR or a null byte in YMM4. */
|
||||
- VPCMP $0, %YMMZERO, %YMM8, %k3
|
||||
+ jnz L(last_vec_x2)
|
||||
|
||||
+ VPCMP $0, %YMMZERO, %YMM3, %k0
|
||||
+ kmovd %k0, %eax
|
||||
+ /* Combine YMM3 matches (eax) with YMM4 matches (ecx). */
|
||||
# ifdef USE_AS_WCSCHR
|
||||
- /* NB: Each bit in K2/K3 represents 4-byte element. */
|
||||
- kshiftlw $8, %k3, %k1
|
||||
+ sall $8, %ecx
|
||||
+ orl %ecx, %eax
|
||||
+ tzcntl %eax, %eax
|
||||
# else
|
||||
- kshiftlq $32, %k3, %k1
|
||||
+ salq $32, %rcx
|
||||
+ orq %rcx, %rax
|
||||
+ tzcntq %rax, %rax
|
||||
# endif
|
||||
+# ifndef USE_AS_STRCHRNUL
|
||||
+ /* Check if match was CHAR or null. */
|
||||
+ cmp (VEC_SIZE * 2)(%rdi, %rax, CHAR_SIZE), %CHAR_REG
|
||||
+ jne L(zero_end)
|
||||
+# endif
|
||||
+ /* NB: Multiply sizeof char type (1 or 4) to get the number of
|
||||
+ bytes. */
|
||||
+ leaq (VEC_SIZE * 2)(%rdi, %rax, CHAR_SIZE), %rax
|
||||
+ ret
|
||||
|
||||
- /* Each bit in K1 represents a NULL or a mismatch. */
|
||||
- korq %k1, %k2, %k1
|
||||
- kmovq %k1, %rax
|
||||
+# ifndef USE_AS_STRCHRNUL
|
||||
+L(zero_end):
|
||||
+ xorl %eax, %eax
|
||||
+ ret
|
||||
+# endif
|
||||
|
||||
- tzcntq %rax, %rax
|
||||
-# ifdef USE_AS_WCSCHR
|
||||
- /* NB: Multiply wchar_t count by 4 to get the number of bytes. */
|
||||
- leaq (VEC_SIZE * 2)(%rdi, %rax, 4), %rax
|
||||
-# else
|
||||
- leaq (VEC_SIZE * 2)(%rdi, %rax), %rax
|
||||
+ .p2align 4
|
||||
+L(last_vec_x1):
|
||||
+ tzcntl %eax, %eax
|
||||
+# ifndef USE_AS_STRCHRNUL
|
||||
+ /* Check if match was null. */
|
||||
+ cmp (%rdi, %rax, CHAR_SIZE), %CHAR_REG
|
||||
+ jne L(zero_end)
|
||||
# endif
|
||||
+ /* NB: Multiply sizeof char type (1 or 4) to get the number of
|
||||
+ bytes. */
|
||||
+ leaq (%rdi, %rax, CHAR_SIZE), %rax
|
||||
+ ret
|
||||
+
|
||||
+ .p2align 4
|
||||
+L(last_vec_x2):
|
||||
+ tzcntl %eax, %eax
|
||||
# ifndef USE_AS_STRCHRNUL
|
||||
- cmp (%rax), %CHAR_REG
|
||||
- cmovne %rdx, %rax
|
||||
+ /* Check if match was null. */
|
||||
+ cmp (VEC_SIZE)(%rdi, %rax, CHAR_SIZE), %CHAR_REG
|
||||
+ jne L(zero_end)
|
||||
# endif
|
||||
+ /* NB: Multiply sizeof char type (1 or 4) to get the number of
|
||||
+ bytes. */
|
||||
+ leaq (VEC_SIZE)(%rdi, %rax, CHAR_SIZE), %rax
|
||||
ret
|
||||
|
||||
/* Cold case for crossing page with first load. */
|
||||
.p2align 4
|
||||
L(cross_page_boundary):
|
||||
+ movq %rdi, %rdx
|
||||
+ /* Align rdi. */
|
||||
andq $-VEC_SIZE, %rdi
|
||||
- andl $(VEC_SIZE - 1), %ecx
|
||||
-
|
||||
VMOVA (%rdi), %YMM1
|
||||
-
|
||||
/* Leaves only CHARS matching esi as 0. */
|
||||
vpxorq %YMM1, %YMM0, %YMM2
|
||||
VPMINU %YMM2, %YMM1, %YMM2
|
||||
/* Each bit in K0 represents a CHAR or a null byte in YMM1. */
|
||||
VPCMP $0, %YMMZERO, %YMM2, %k0
|
||||
kmovd %k0, %eax
|
||||
- testl %eax, %eax
|
||||
-
|
||||
+ /* Remove the leading bits. */
|
||||
# ifdef USE_AS_WCSCHR
|
||||
+ movl %edx, %SHIFT_REG
|
||||
/* NB: Divide shift count by 4 since each bit in K1 represent 4
|
||||
bytes. */
|
||||
- movl %ecx, %SHIFT_REG
|
||||
- sarl $2, %SHIFT_REG
|
||||
+ sarl $2, %SHIFT_REG
|
||||
+ andl $(CHAR_PER_VEC - 1), %SHIFT_REG
|
||||
# endif
|
||||
-
|
||||
- /* Remove the leading bits. */
|
||||
sarxl %SHIFT_REG, %eax, %eax
|
||||
+ /* If eax is zero continue. */
|
||||
testl %eax, %eax
|
||||
-
|
||||
- jz L(aligned_more)
|
||||
+ jz L(cross_page_continue)
|
||||
tzcntl %eax, %eax
|
||||
- addq %rcx, %rdi
|
||||
+# ifndef USE_AS_STRCHRNUL
|
||||
+ /* Check to see if match was CHAR or null. */
|
||||
+ cmp (%rdx, %rax, CHAR_SIZE), %CHAR_REG
|
||||
+ jne L(zero_end)
|
||||
+# endif
|
||||
# ifdef USE_AS_WCSCHR
|
||||
- /* NB: Multiply wchar_t count by 4 to get the number of bytes. */
|
||||
- leaq (%rdi, %rax, 4), %rax
|
||||
+ /* NB: Multiply wchar_t count by 4 to get the number of
|
||||
+ bytes. */
|
||||
+ leaq (%rdx, %rax, CHAR_SIZE), %rax
|
||||
# else
|
||||
- addq %rdi, %rax
|
||||
-# endif
|
||||
-# ifndef USE_AS_STRCHRNUL
|
||||
- cmp (%rax), %CHAR_REG
|
||||
- cmovne %rdx, %rax
|
||||
+ addq %rdx, %rax
|
||||
# endif
|
||||
ret
|
||||
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,536 +0,0 @@
|
||||
From 104c7b1967c3e78435c6f7eab5e225a7eddf9c6e Mon Sep 17 00:00:00 2001
|
||||
From: Noah Goldstein <goldstein.w.n@gmail.com>
|
||||
Date: Tue, 4 May 2021 19:02:40 -0400
|
||||
Subject: [PATCH] x86: Add EVEX optimized memchr family not safe for RTM
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
No bug.
|
||||
|
||||
This commit adds a new implementation for EVEX memchr that is not safe
|
||||
for RTM because it uses vzeroupper. The benefit is that by using
|
||||
ymm0-ymm15 it can use vpcmpeq and vpternlogd in the 4x loop which is
|
||||
faster than the RTM safe version which cannot use vpcmpeq because
|
||||
there is no EVEX encoding for the instruction. All parts of the
|
||||
implementation aside from the 4x loop are the same for the two
|
||||
versions and the optimization is only relevant for large sizes.
|
||||
|
||||
Tigerlake:
|
||||
size , algn , Pos , Cur T , New T , Win , Dif
|
||||
512 , 6 , 192 , 9.2 , 9.04 , no-RTM , 0.16
|
||||
512 , 7 , 224 , 9.19 , 8.98 , no-RTM , 0.21
|
||||
2048 , 0 , 256 , 10.74 , 10.54 , no-RTM , 0.2
|
||||
2048 , 0 , 512 , 14.81 , 14.87 , RTM , 0.06
|
||||
2048 , 0 , 1024 , 22.97 , 22.57 , no-RTM , 0.4
|
||||
2048 , 0 , 2048 , 37.49 , 34.51 , no-RTM , 2.98 <--
|
||||
|
||||
Icelake:
|
||||
size , algn , Pos , Cur T , New T , Win , Dif
|
||||
512 , 6 , 192 , 7.6 , 7.3 , no-RTM , 0.3
|
||||
512 , 7 , 224 , 7.63 , 7.27 , no-RTM , 0.36
|
||||
2048 , 0 , 256 , 8.48 , 8.38 , no-RTM , 0.1
|
||||
2048 , 0 , 512 , 11.57 , 11.42 , no-RTM , 0.15
|
||||
2048 , 0 , 1024 , 17.92 , 17.38 , no-RTM , 0.54
|
||||
2048 , 0 , 2048 , 30.37 , 27.34 , no-RTM , 3.03 <--
|
||||
|
||||
test-memchr, test-wmemchr, and test-rawmemchr are all passing.
|
||||
|
||||
Signed-off-by: Noah Goldstein <goldstein.w.n@gmail.com>
|
||||
Reviewed-by: H.J. Lu <hjl.tools@gmail.com>
|
||||
---
|
||||
sysdeps/x86_64/multiarch/Makefile | 7 +-
|
||||
sysdeps/x86_64/multiarch/ifunc-evex.h | 55 ++++++
|
||||
sysdeps/x86_64/multiarch/ifunc-impl-list.c | 15 ++
|
||||
sysdeps/x86_64/multiarch/memchr-evex-rtm.S | 8 +
|
||||
sysdeps/x86_64/multiarch/memchr-evex.S | 161 ++++++++++++++----
|
||||
sysdeps/x86_64/multiarch/memchr.c | 2 +-
|
||||
sysdeps/x86_64/multiarch/rawmemchr-evex-rtm.S | 3 +
|
||||
sysdeps/x86_64/multiarch/rawmemchr.c | 2 +-
|
||||
sysdeps/x86_64/multiarch/wmemchr-evex-rtm.S | 3 +
|
||||
sysdeps/x86_64/multiarch/wmemchr.c | 2 +-
|
||||
10 files changed, 217 insertions(+), 41 deletions(-)
|
||||
create mode 100644 sysdeps/x86_64/multiarch/ifunc-evex.h
|
||||
create mode 100644 sysdeps/x86_64/multiarch/memchr-evex-rtm.S
|
||||
create mode 100644 sysdeps/x86_64/multiarch/rawmemchr-evex-rtm.S
|
||||
create mode 100644 sysdeps/x86_64/multiarch/wmemchr-evex-rtm.S
|
||||
|
||||
diff --git a/sysdeps/x86_64/multiarch/Makefile b/sysdeps/x86_64/multiarch/Makefile
|
||||
index 65fde4eb..26be4095 100644
|
||||
--- a/sysdeps/x86_64/multiarch/Makefile
|
||||
+++ b/sysdeps/x86_64/multiarch/Makefile
|
||||
@@ -77,7 +77,9 @@ sysdep_routines += strncat-c stpncpy-c strncpy-c \
|
||||
strncmp-evex \
|
||||
strncpy-evex \
|
||||
strnlen-evex \
|
||||
- strrchr-evex
|
||||
+ strrchr-evex \
|
||||
+ memchr-evex-rtm \
|
||||
+ rawmemchr-evex-rtm
|
||||
CFLAGS-varshift.c += -msse4
|
||||
CFLAGS-strcspn-c.c += -msse4
|
||||
CFLAGS-strpbrk-c.c += -msse4
|
||||
@@ -110,7 +112,8 @@ sysdep_routines += wmemcmp-sse4 wmemcmp-ssse3 wmemcmp-c \
|
||||
wcsnlen-evex \
|
||||
wcsrchr-evex \
|
||||
wmemchr-evex \
|
||||
- wmemcmp-evex-movbe
|
||||
+ wmemcmp-evex-movbe \
|
||||
+ wmemchr-evex-rtm
|
||||
endif
|
||||
|
||||
ifeq ($(subdir),debug)
|
||||
diff --git a/sysdeps/x86_64/multiarch/ifunc-evex.h b/sysdeps/x86_64/multiarch/ifunc-evex.h
|
||||
new file mode 100644
|
||||
index 00000000..fc391edb
|
||||
--- /dev/null
|
||||
+++ b/sysdeps/x86_64/multiarch/ifunc-evex.h
|
||||
@@ -0,0 +1,55 @@
|
||||
+/* Common definition for ifunc selection optimized with EVEX.
|
||||
+ All versions must be listed in ifunc-impl-list.c.
|
||||
+ Copyright (C) 2017-2021 Free Software Foundation, Inc.
|
||||
+ This file is part of the GNU C Library.
|
||||
+
|
||||
+ The GNU C Library is free software; you can redistribute it and/or
|
||||
+ modify it under the terms of the GNU Lesser General Public
|
||||
+ License as published by the Free Software Foundation; either
|
||||
+ version 2.1 of the License, or (at your option) any later version.
|
||||
+
|
||||
+ The GNU C Library is distributed in the hope that it will be useful,
|
||||
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
||||
+ Lesser General Public License for more details.
|
||||
+
|
||||
+ You should have received a copy of the GNU Lesser General Public
|
||||
+ License along with the GNU C Library; if not, see
|
||||
+ <https://www.gnu.org/licenses/>. */
|
||||
+
|
||||
+#include <init-arch.h>
|
||||
+
|
||||
+extern __typeof (REDIRECT_NAME) OPTIMIZE (sse2) attribute_hidden;
|
||||
+extern __typeof (REDIRECT_NAME) OPTIMIZE (avx2) attribute_hidden;
|
||||
+extern __typeof (REDIRECT_NAME) OPTIMIZE (avx2_rtm) attribute_hidden;
|
||||
+extern __typeof (REDIRECT_NAME) OPTIMIZE (evex) attribute_hidden;
|
||||
+extern __typeof (REDIRECT_NAME) OPTIMIZE (evex_rtm) attribute_hidden;
|
||||
+
|
||||
+
|
||||
+static inline void *
|
||||
+IFUNC_SELECTOR (void)
|
||||
+{
|
||||
+ const struct cpu_features* cpu_features = __get_cpu_features ();
|
||||
+
|
||||
+ if (CPU_FEATURE_USABLE_P (cpu_features, AVX2)
|
||||
+ && CPU_FEATURE_USABLE_P (cpu_features, BMI2)
|
||||
+ && CPU_FEATURES_ARCH_P (cpu_features, AVX_Fast_Unaligned_Load))
|
||||
+ {
|
||||
+ if (CPU_FEATURE_USABLE_P (cpu_features, AVX512VL)
|
||||
+ && CPU_FEATURE_USABLE_P (cpu_features, AVX512BW))
|
||||
+ {
|
||||
+ if (CPU_FEATURE_USABLE_P (cpu_features, RTM))
|
||||
+ return OPTIMIZE (evex_rtm);
|
||||
+
|
||||
+ return OPTIMIZE (evex);
|
||||
+ }
|
||||
+
|
||||
+ if (CPU_FEATURE_USABLE_P (cpu_features, RTM))
|
||||
+ return OPTIMIZE (avx2_rtm);
|
||||
+
|
||||
+ if (!CPU_FEATURES_ARCH_P (cpu_features, Prefer_No_VZEROUPPER))
|
||||
+ return OPTIMIZE (avx2);
|
||||
+ }
|
||||
+
|
||||
+ return OPTIMIZE (sse2);
|
||||
+}
|
||||
diff --git a/sysdeps/x86_64/multiarch/ifunc-impl-list.c b/sysdeps/x86_64/multiarch/ifunc-impl-list.c
|
||||
index d59d65f8..ac097e8d 100644
|
||||
--- a/sysdeps/x86_64/multiarch/ifunc-impl-list.c
|
||||
+++ b/sysdeps/x86_64/multiarch/ifunc-impl-list.c
|
||||
@@ -52,6 +52,11 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
|
||||
&& CPU_FEATURE_USABLE (AVX512BW)
|
||||
&& CPU_FEATURE_USABLE (BMI2)),
|
||||
__memchr_evex)
|
||||
+ IFUNC_IMPL_ADD (array, i, memchr,
|
||||
+ (CPU_FEATURE_USABLE (AVX512VL)
|
||||
+ && CPU_FEATURE_USABLE (AVX512BW)
|
||||
+ && CPU_FEATURE_USABLE (BMI2)),
|
||||
+ __memchr_evex_rtm)
|
||||
IFUNC_IMPL_ADD (array, i, memchr, 1, __memchr_sse2))
|
||||
|
||||
/* Support sysdeps/x86_64/multiarch/memcmp.c. */
|
||||
@@ -288,6 +293,11 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
|
||||
&& CPU_FEATURE_USABLE (AVX512BW)
|
||||
&& CPU_FEATURE_USABLE (BMI2)),
|
||||
__rawmemchr_evex)
|
||||
+ IFUNC_IMPL_ADD (array, i, rawmemchr,
|
||||
+ (CPU_FEATURE_USABLE (AVX512VL)
|
||||
+ && CPU_FEATURE_USABLE (AVX512BW)
|
||||
+ && CPU_FEATURE_USABLE (BMI2)),
|
||||
+ __rawmemchr_evex_rtm)
|
||||
IFUNC_IMPL_ADD (array, i, rawmemchr, 1, __rawmemchr_sse2))
|
||||
|
||||
/* Support sysdeps/x86_64/multiarch/strlen.c. */
|
||||
@@ -711,6 +721,11 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
|
||||
&& CPU_FEATURE_USABLE (AVX512BW)
|
||||
&& CPU_FEATURE_USABLE (BMI2)),
|
||||
__wmemchr_evex)
|
||||
+ IFUNC_IMPL_ADD (array, i, wmemchr,
|
||||
+ (CPU_FEATURE_USABLE (AVX512VL)
|
||||
+ && CPU_FEATURE_USABLE (AVX512BW)
|
||||
+ && CPU_FEATURE_USABLE (BMI2)),
|
||||
+ __wmemchr_evex_rtm)
|
||||
IFUNC_IMPL_ADD (array, i, wmemchr, 1, __wmemchr_sse2))
|
||||
|
||||
/* Support sysdeps/x86_64/multiarch/wmemcmp.c. */
|
||||
diff --git a/sysdeps/x86_64/multiarch/memchr-evex-rtm.S b/sysdeps/x86_64/multiarch/memchr-evex-rtm.S
|
||||
new file mode 100644
|
||||
index 00000000..19871882
|
||||
--- /dev/null
|
||||
+++ b/sysdeps/x86_64/multiarch/memchr-evex-rtm.S
|
||||
@@ -0,0 +1,8 @@
|
||||
+#ifndef MEMCHR
|
||||
+# define MEMCHR __memchr_evex_rtm
|
||||
+#endif
|
||||
+
|
||||
+#define USE_IN_RTM 1
|
||||
+#define SECTION(p) p##.evex.rtm
|
||||
+
|
||||
+#include "memchr-evex.S"
|
||||
diff --git a/sysdeps/x86_64/multiarch/memchr-evex.S b/sysdeps/x86_64/multiarch/memchr-evex.S
|
||||
index f3fdad4f..4d0ed6d1 100644
|
||||
--- a/sysdeps/x86_64/multiarch/memchr-evex.S
|
||||
+++ b/sysdeps/x86_64/multiarch/memchr-evex.S
|
||||
@@ -38,10 +38,32 @@
|
||||
# define CHAR_SIZE 1
|
||||
# endif
|
||||
|
||||
+ /* In the 4x loop the RTM and non-RTM versions have data pointer
|
||||
+ off by VEC_SIZE * 4 with RTM version being VEC_SIZE * 4 greater.
|
||||
+ This is represented by BASE_OFFSET. As well because the RTM
|
||||
+ version uses vpcmp which stores a bit per element compared where
|
||||
+ the non-RTM version uses vpcmpeq which stores a bit per byte
|
||||
+ compared RET_SCALE of CHAR_SIZE is only relevant for the RTM
|
||||
+ version. */
|
||||
+# ifdef USE_IN_RTM
|
||||
+# define VZEROUPPER
|
||||
+# define BASE_OFFSET (VEC_SIZE * 4)
|
||||
+# define RET_SCALE CHAR_SIZE
|
||||
+# else
|
||||
+# define VZEROUPPER vzeroupper
|
||||
+# define BASE_OFFSET 0
|
||||
+# define RET_SCALE 1
|
||||
+# endif
|
||||
+
|
||||
+ /* In the return from 4x loop memchr and rawmemchr versions have
|
||||
+ data pointers off by VEC_SIZE * 4 with memchr version being
|
||||
+ VEC_SIZE * 4 greater. */
|
||||
# ifdef USE_AS_RAWMEMCHR
|
||||
+# define RET_OFFSET (BASE_OFFSET - (VEC_SIZE * 4))
|
||||
# define RAW_PTR_REG rcx
|
||||
# define ALGN_PTR_REG rdi
|
||||
# else
|
||||
+# define RET_OFFSET BASE_OFFSET
|
||||
# define RAW_PTR_REG rdi
|
||||
# define ALGN_PTR_REG rcx
|
||||
# endif
|
||||
@@ -57,11 +79,15 @@
|
||||
# define YMM5 ymm21
|
||||
# define YMM6 ymm22
|
||||
|
||||
+# ifndef SECTION
|
||||
+# define SECTION(p) p##.evex
|
||||
+# endif
|
||||
+
|
||||
# define VEC_SIZE 32
|
||||
# define CHAR_PER_VEC (VEC_SIZE / CHAR_SIZE)
|
||||
# define PAGE_SIZE 4096
|
||||
|
||||
- .section .text.evex,"ax",@progbits
|
||||
+ .section SECTION(.text),"ax",@progbits
|
||||
ENTRY (MEMCHR)
|
||||
# ifndef USE_AS_RAWMEMCHR
|
||||
/* Check for zero length. */
|
||||
@@ -237,14 +263,15 @@ L(cross_page_continue):
|
||||
/* Check if at last CHAR_PER_VEC * 4 length. */
|
||||
subq $(CHAR_PER_VEC * 4), %rdx
|
||||
jbe L(last_4x_vec_or_less_cmpeq)
|
||||
- addq $VEC_SIZE, %rdi
|
||||
+ /* +VEC_SIZE if USE_IN_RTM otherwise +VEC_SIZE * 5. */
|
||||
+ addq $(VEC_SIZE + (VEC_SIZE * 4 - BASE_OFFSET)), %rdi
|
||||
|
||||
/* Align data to VEC_SIZE * 4 for the loop and readjust length.
|
||||
*/
|
||||
# ifdef USE_AS_WMEMCHR
|
||||
movl %edi, %ecx
|
||||
andq $-(4 * VEC_SIZE), %rdi
|
||||
- andl $(VEC_SIZE * 4 - 1), %ecx
|
||||
+ subl %edi, %ecx
|
||||
/* NB: Divide bytes by 4 to get the wchar_t count. */
|
||||
sarl $2, %ecx
|
||||
addq %rcx, %rdx
|
||||
@@ -254,15 +281,28 @@ L(cross_page_continue):
|
||||
subq %rdi, %rdx
|
||||
# endif
|
||||
# else
|
||||
- addq $VEC_SIZE, %rdi
|
||||
+ addq $(VEC_SIZE + (VEC_SIZE * 4 - BASE_OFFSET)), %rdi
|
||||
andq $-(4 * VEC_SIZE), %rdi
|
||||
# endif
|
||||
-
|
||||
+# ifdef USE_IN_RTM
|
||||
vpxorq %XMMZERO, %XMMZERO, %XMMZERO
|
||||
+# else
|
||||
+ /* copy ymmmatch to ymm0 so we can use vpcmpeq which is not
|
||||
+ encodable with EVEX registers (ymm16-ymm31). */
|
||||
+ vmovdqa64 %YMMMATCH, %ymm0
|
||||
+# endif
|
||||
|
||||
/* Compare 4 * VEC at a time forward. */
|
||||
.p2align 4
|
||||
L(loop_4x_vec):
|
||||
+ /* Two versions of the loop. One that does not require
|
||||
+ vzeroupper by not using ymm0-ymm15 and another does that require
|
||||
+ vzeroupper because it uses ymm0-ymm15. The reason why ymm0-ymm15
|
||||
+ is used at all is because there is no EVEX encoding vpcmpeq and
|
||||
+ with vpcmpeq this loop can be performed more efficiently. The
|
||||
+ non-vzeroupper version is safe for RTM while the vzeroupper
|
||||
+ version should be prefered if RTM are not supported. */
|
||||
+# ifdef USE_IN_RTM
|
||||
/* It would be possible to save some instructions using 4x VPCMP
|
||||
but bottleneck on port 5 makes it not woth it. */
|
||||
VPCMP $4, (VEC_SIZE * 4)(%rdi), %YMMMATCH, %k1
|
||||
@@ -273,12 +313,55 @@ L(loop_4x_vec):
|
||||
/* Reduce VEC2 / VEC3 with min and VEC1 with zero mask. */
|
||||
VPMINU %YMM2, %YMM3, %YMM3{%k1}{z}
|
||||
VPCMP $0, %YMM3, %YMMZERO, %k2
|
||||
+# else
|
||||
+ /* Since vptern can only take 3x vectors fastest to do 1 vec
|
||||
+ seperately with EVEX vpcmp. */
|
||||
+# ifdef USE_AS_WMEMCHR
|
||||
+ /* vptern can only accept masks for epi32/epi64 so can only save
|
||||
+ instruction using not equals mask on vptern with wmemchr. */
|
||||
+ VPCMP $4, (%rdi), %YMMMATCH, %k1
|
||||
+# else
|
||||
+ VPCMP $0, (%rdi), %YMMMATCH, %k1
|
||||
+# endif
|
||||
+ /* Compare 3x with vpcmpeq and or them all together with vptern.
|
||||
+ */
|
||||
+ VPCMPEQ VEC_SIZE(%rdi), %ymm0, %ymm2
|
||||
+ VPCMPEQ (VEC_SIZE * 2)(%rdi), %ymm0, %ymm3
|
||||
+ VPCMPEQ (VEC_SIZE * 3)(%rdi), %ymm0, %ymm4
|
||||
+# ifdef USE_AS_WMEMCHR
|
||||
+ /* This takes the not of or between ymm2, ymm3, ymm4 as well as
|
||||
+ combines result from VEC0 with zero mask. */
|
||||
+ vpternlogd $1, %ymm2, %ymm3, %ymm4{%k1}{z}
|
||||
+ vpmovmskb %ymm4, %ecx
|
||||
+# else
|
||||
+ /* 254 is mask for oring ymm2, ymm3, ymm4 into ymm4. */
|
||||
+ vpternlogd $254, %ymm2, %ymm3, %ymm4
|
||||
+ vpmovmskb %ymm4, %ecx
|
||||
+ kmovd %k1, %eax
|
||||
+# endif
|
||||
+# endif
|
||||
+
|
||||
# ifdef USE_AS_RAWMEMCHR
|
||||
subq $-(VEC_SIZE * 4), %rdi
|
||||
+# endif
|
||||
+# ifdef USE_IN_RTM
|
||||
kortestd %k2, %k3
|
||||
+# else
|
||||
+# ifdef USE_AS_WMEMCHR
|
||||
+ /* ecx contains not of matches. All 1s means no matches. incl will
|
||||
+ overflow and set zeroflag if that is the case. */
|
||||
+ incl %ecx
|
||||
+# else
|
||||
+ /* If either VEC1 (eax) or VEC2-VEC4 (ecx) are not zero. Adding
|
||||
+ to ecx is not an issue because if eax is non-zero it will be
|
||||
+ used for returning the match. If it is zero the add does
|
||||
+ nothing. */
|
||||
+ addq %rax, %rcx
|
||||
+# endif
|
||||
+# endif
|
||||
+# ifdef USE_AS_RAWMEMCHR
|
||||
jz L(loop_4x_vec)
|
||||
# else
|
||||
- kortestd %k2, %k3
|
||||
jnz L(loop_4x_vec_end)
|
||||
|
||||
subq $-(VEC_SIZE * 4), %rdi
|
||||
@@ -288,10 +371,11 @@ L(loop_4x_vec):
|
||||
|
||||
/* Fall through into less than 4 remaining vectors of length case.
|
||||
*/
|
||||
- VPCMP $0, (VEC_SIZE * 4)(%rdi), %YMMMATCH, %k0
|
||||
+ VPCMP $0, BASE_OFFSET(%rdi), %YMMMATCH, %k0
|
||||
+ addq $(BASE_OFFSET - VEC_SIZE), %rdi
|
||||
kmovd %k0, %eax
|
||||
- addq $(VEC_SIZE * 3), %rdi
|
||||
- .p2align 4
|
||||
+ VZEROUPPER
|
||||
+
|
||||
L(last_4x_vec_or_less):
|
||||
/* Check if first VEC contained match. */
|
||||
testl %eax, %eax
|
||||
@@ -338,73 +422,78 @@ L(loop_4x_vec_end):
|
||||
/* rawmemchr will fall through into this if match was found in
|
||||
loop. */
|
||||
|
||||
+# if defined USE_IN_RTM || defined USE_AS_WMEMCHR
|
||||
/* k1 has not of matches with VEC1. */
|
||||
kmovd %k1, %eax
|
||||
-# ifdef USE_AS_WMEMCHR
|
||||
+# ifdef USE_AS_WMEMCHR
|
||||
subl $((1 << CHAR_PER_VEC) - 1), %eax
|
||||
-# else
|
||||
+# else
|
||||
incl %eax
|
||||
+# endif
|
||||
+# else
|
||||
+ /* eax already has matches for VEC1. */
|
||||
+ testl %eax, %eax
|
||||
# endif
|
||||
jnz L(last_vec_x1_return)
|
||||
|
||||
+# ifdef USE_IN_RTM
|
||||
VPCMP $0, %YMM2, %YMMZERO, %k0
|
||||
kmovd %k0, %eax
|
||||
+# else
|
||||
+ vpmovmskb %ymm2, %eax
|
||||
+# endif
|
||||
testl %eax, %eax
|
||||
jnz L(last_vec_x2_return)
|
||||
|
||||
+# ifdef USE_IN_RTM
|
||||
kmovd %k2, %eax
|
||||
testl %eax, %eax
|
||||
jnz L(last_vec_x3_return)
|
||||
|
||||
kmovd %k3, %eax
|
||||
tzcntl %eax, %eax
|
||||
-# ifdef USE_AS_RAWMEMCHR
|
||||
- leaq (VEC_SIZE * 3)(%rdi, %rax, CHAR_SIZE), %rax
|
||||
+ leaq (VEC_SIZE * 3 + RET_OFFSET)(%rdi, %rax, CHAR_SIZE), %rax
|
||||
# else
|
||||
- leaq (VEC_SIZE * 7)(%rdi, %rax, CHAR_SIZE), %rax
|
||||
+ vpmovmskb %ymm3, %eax
|
||||
+ /* Combine matches in VEC3 (eax) with matches in VEC4 (ecx). */
|
||||
+ salq $VEC_SIZE, %rcx
|
||||
+ orq %rcx, %rax
|
||||
+ tzcntq %rax, %rax
|
||||
+ leaq (VEC_SIZE * 2 + RET_OFFSET)(%rdi, %rax), %rax
|
||||
+ VZEROUPPER
|
||||
# endif
|
||||
ret
|
||||
|
||||
.p2align 4
|
||||
L(last_vec_x1_return):
|
||||
tzcntl %eax, %eax
|
||||
-# ifdef USE_AS_RAWMEMCHR
|
||||
-# ifdef USE_AS_WMEMCHR
|
||||
+# if defined USE_AS_WMEMCHR || RET_OFFSET != 0
|
||||
/* NB: Multiply bytes by CHAR_SIZE to get the wchar_t count. */
|
||||
- leaq (%rdi, %rax, CHAR_SIZE), %rax
|
||||
-# else
|
||||
- addq %rdi, %rax
|
||||
-# endif
|
||||
+ leaq RET_OFFSET(%rdi, %rax, CHAR_SIZE), %rax
|
||||
# else
|
||||
- /* NB: Multiply bytes by CHAR_SIZE to get the wchar_t count. */
|
||||
- leaq (VEC_SIZE * 4)(%rdi, %rax, CHAR_SIZE), %rax
|
||||
+ addq %rdi, %rax
|
||||
# endif
|
||||
+ VZEROUPPER
|
||||
ret
|
||||
|
||||
.p2align 4
|
||||
L(last_vec_x2_return):
|
||||
tzcntl %eax, %eax
|
||||
-# ifdef USE_AS_RAWMEMCHR
|
||||
- /* NB: Multiply bytes by CHAR_SIZE to get the wchar_t count. */
|
||||
- leaq VEC_SIZE(%rdi, %rax, CHAR_SIZE), %rax
|
||||
-# else
|
||||
- /* NB: Multiply bytes by CHAR_SIZE to get the wchar_t count. */
|
||||
- leaq (VEC_SIZE * 5)(%rdi, %rax, CHAR_SIZE), %rax
|
||||
-# endif
|
||||
+ /* NB: Multiply bytes by RET_SCALE to get the wchar_t count
|
||||
+ if relevant (RET_SCALE = CHAR_SIZE if USE_AS_WMEMCHAR and
|
||||
+ USE_IN_RTM are both defined. Otherwise RET_SCALE = 1. */
|
||||
+ leaq (VEC_SIZE + RET_OFFSET)(%rdi, %rax, RET_SCALE), %rax
|
||||
+ VZEROUPPER
|
||||
ret
|
||||
|
||||
+# ifdef USE_IN_RTM
|
||||
.p2align 4
|
||||
L(last_vec_x3_return):
|
||||
tzcntl %eax, %eax
|
||||
-# ifdef USE_AS_RAWMEMCHR
|
||||
- /* NB: Multiply bytes by CHAR_SIZE to get the wchar_t count. */
|
||||
- leaq (VEC_SIZE * 2)(%rdi, %rax, CHAR_SIZE), %rax
|
||||
-# else
|
||||
/* NB: Multiply bytes by CHAR_SIZE to get the wchar_t count. */
|
||||
- leaq (VEC_SIZE * 6)(%rdi, %rax, CHAR_SIZE), %rax
|
||||
-# endif
|
||||
+ leaq (VEC_SIZE * 2 + RET_OFFSET)(%rdi, %rax, CHAR_SIZE), %rax
|
||||
ret
|
||||
-
|
||||
+# endif
|
||||
|
||||
# ifndef USE_AS_RAWMEMCHR
|
||||
L(last_4x_vec_or_less_cmpeq):
|
||||
diff --git a/sysdeps/x86_64/multiarch/memchr.c b/sysdeps/x86_64/multiarch/memchr.c
|
||||
index 016f5784..f28aea77 100644
|
||||
--- a/sysdeps/x86_64/multiarch/memchr.c
|
||||
+++ b/sysdeps/x86_64/multiarch/memchr.c
|
||||
@@ -24,7 +24,7 @@
|
||||
# undef memchr
|
||||
|
||||
# define SYMBOL_NAME memchr
|
||||
-# include "ifunc-avx2.h"
|
||||
+# include "ifunc-evex.h"
|
||||
|
||||
libc_ifunc_redirected (__redirect_memchr, memchr, IFUNC_SELECTOR ());
|
||||
strong_alias (memchr, __memchr)
|
||||
diff --git a/sysdeps/x86_64/multiarch/rawmemchr-evex-rtm.S b/sysdeps/x86_64/multiarch/rawmemchr-evex-rtm.S
|
||||
new file mode 100644
|
||||
index 00000000..deda1ca3
|
||||
--- /dev/null
|
||||
+++ b/sysdeps/x86_64/multiarch/rawmemchr-evex-rtm.S
|
||||
@@ -0,0 +1,3 @@
|
||||
+#define MEMCHR __rawmemchr_evex_rtm
|
||||
+#define USE_AS_RAWMEMCHR 1
|
||||
+#include "memchr-evex-rtm.S"
|
||||
diff --git a/sysdeps/x86_64/multiarch/rawmemchr.c b/sysdeps/x86_64/multiarch/rawmemchr.c
|
||||
index 8a0bc313..1f764f35 100644
|
||||
--- a/sysdeps/x86_64/multiarch/rawmemchr.c
|
||||
+++ b/sysdeps/x86_64/multiarch/rawmemchr.c
|
||||
@@ -26,7 +26,7 @@
|
||||
# undef __rawmemchr
|
||||
|
||||
# define SYMBOL_NAME rawmemchr
|
||||
-# include "ifunc-avx2.h"
|
||||
+# include "ifunc-evex.h"
|
||||
|
||||
libc_ifunc_redirected (__redirect_rawmemchr, __rawmemchr,
|
||||
IFUNC_SELECTOR ());
|
||||
diff --git a/sysdeps/x86_64/multiarch/wmemchr-evex-rtm.S b/sysdeps/x86_64/multiarch/wmemchr-evex-rtm.S
|
||||
new file mode 100644
|
||||
index 00000000..a346cd35
|
||||
--- /dev/null
|
||||
+++ b/sysdeps/x86_64/multiarch/wmemchr-evex-rtm.S
|
||||
@@ -0,0 +1,3 @@
|
||||
+#define MEMCHR __wmemchr_evex_rtm
|
||||
+#define USE_AS_WMEMCHR 1
|
||||
+#include "memchr-evex-rtm.S"
|
||||
diff --git a/sysdeps/x86_64/multiarch/wmemchr.c b/sysdeps/x86_64/multiarch/wmemchr.c
|
||||
index 6d833702..f9c91915 100644
|
||||
--- a/sysdeps/x86_64/multiarch/wmemchr.c
|
||||
+++ b/sysdeps/x86_64/multiarch/wmemchr.c
|
||||
@@ -26,7 +26,7 @@
|
||||
# undef __wmemchr
|
||||
|
||||
# define SYMBOL_NAME wmemchr
|
||||
-# include "ifunc-avx2.h"
|
||||
+# include "ifunc-evex.h"
|
||||
|
||||
libc_ifunc_redirected (__redirect_wmemchr, __wmemchr, IFUNC_SELECTOR ());
|
||||
weak_alias (__wmemchr, wmemchr)
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,873 +0,0 @@
|
||||
From 16d12015c57701b08d7bbed6ec536641bcafb428 Mon Sep 17 00:00:00 2001
|
||||
From: Noah Goldstein <goldstein.w.n@gmail.com>
|
||||
Date: Mon, 17 May 2021 13:56:52 -0400
|
||||
Subject: [PATCH] x86: Optimize memcmp-avx2-movbe.S
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
No bug. This commit optimizes memcmp-avx2.S. The optimizations include
|
||||
adding a new vec compare path for small sizes, reorganizing the entry
|
||||
control flow, and removing some unnecissary ALU instructions from the
|
||||
main loop. test-memcmp and test-wmemcmp are both passing.
|
||||
|
||||
Signed-off-by: Noah Goldstein <goldstein.w.n@gmail.com>
|
||||
Reviewed-by: H.J. Lu <hjl.tools@gmail.com>
|
||||
---
|
||||
sysdeps/x86_64/multiarch/ifunc-impl-list.c | 6 +
|
||||
sysdeps/x86_64/multiarch/ifunc-memcmp.h | 1 +
|
||||
sysdeps/x86_64/multiarch/memcmp-avx2-movbe.S | 676 +++++++++++--------
|
||||
3 files changed, 402 insertions(+), 281 deletions(-)
|
||||
|
||||
diff --git a/sysdeps/x86_64/multiarch/ifunc-impl-list.c b/sysdeps/x86_64/multiarch/ifunc-impl-list.c
|
||||
index ac097e8d..8be0d78a 100644
|
||||
--- a/sysdeps/x86_64/multiarch/ifunc-impl-list.c
|
||||
+++ b/sysdeps/x86_64/multiarch/ifunc-impl-list.c
|
||||
@@ -63,16 +63,19 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
|
||||
IFUNC_IMPL (i, name, memcmp,
|
||||
IFUNC_IMPL_ADD (array, i, memcmp,
|
||||
(CPU_FEATURE_USABLE (AVX2)
|
||||
+ && CPU_FEATURE_USABLE (BMI2)
|
||||
&& CPU_FEATURE_USABLE (MOVBE)),
|
||||
__memcmp_avx2_movbe)
|
||||
IFUNC_IMPL_ADD (array, i, memcmp,
|
||||
(CPU_FEATURE_USABLE (AVX2)
|
||||
+ && CPU_FEATURE_USABLE (BMI2)
|
||||
&& CPU_FEATURE_USABLE (MOVBE)
|
||||
&& CPU_FEATURE_USABLE (RTM)),
|
||||
__memcmp_avx2_movbe_rtm)
|
||||
IFUNC_IMPL_ADD (array, i, memcmp,
|
||||
(CPU_FEATURE_USABLE (AVX512VL)
|
||||
&& CPU_FEATURE_USABLE (AVX512BW)
|
||||
+ && CPU_FEATURE_USABLE (BMI2)
|
||||
&& CPU_FEATURE_USABLE (MOVBE)),
|
||||
__memcmp_evex_movbe)
|
||||
IFUNC_IMPL_ADD (array, i, memcmp, CPU_FEATURE_USABLE (SSE4_1),
|
||||
@@ -732,16 +735,19 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
|
||||
IFUNC_IMPL (i, name, wmemcmp,
|
||||
IFUNC_IMPL_ADD (array, i, wmemcmp,
|
||||
(CPU_FEATURE_USABLE (AVX2)
|
||||
+ && CPU_FEATURE_USABLE (BMI2)
|
||||
&& CPU_FEATURE_USABLE (MOVBE)),
|
||||
__wmemcmp_avx2_movbe)
|
||||
IFUNC_IMPL_ADD (array, i, wmemcmp,
|
||||
(CPU_FEATURE_USABLE (AVX2)
|
||||
+ && CPU_FEATURE_USABLE (BMI2)
|
||||
&& CPU_FEATURE_USABLE (MOVBE)
|
||||
&& CPU_FEATURE_USABLE (RTM)),
|
||||
__wmemcmp_avx2_movbe_rtm)
|
||||
IFUNC_IMPL_ADD (array, i, wmemcmp,
|
||||
(CPU_FEATURE_USABLE (AVX512VL)
|
||||
&& CPU_FEATURE_USABLE (AVX512BW)
|
||||
+ && CPU_FEATURE_USABLE (BMI2)
|
||||
&& CPU_FEATURE_USABLE (MOVBE)),
|
||||
__wmemcmp_evex_movbe)
|
||||
IFUNC_IMPL_ADD (array, i, wmemcmp, CPU_FEATURE_USABLE (SSE4_1),
|
||||
diff --git a/sysdeps/x86_64/multiarch/ifunc-memcmp.h b/sysdeps/x86_64/multiarch/ifunc-memcmp.h
|
||||
index 8043c635..690dffe8 100644
|
||||
--- a/sysdeps/x86_64/multiarch/ifunc-memcmp.h
|
||||
+++ b/sysdeps/x86_64/multiarch/ifunc-memcmp.h
|
||||
@@ -33,6 +33,7 @@ IFUNC_SELECTOR (void)
|
||||
|
||||
if (CPU_FEATURE_USABLE_P (cpu_features, AVX2)
|
||||
&& CPU_FEATURE_USABLE_P (cpu_features, MOVBE)
|
||||
+ && CPU_FEATURE_USABLE_P (cpu_features, BMI2)
|
||||
&& CPU_FEATURES_ARCH_P (cpu_features, AVX_Fast_Unaligned_Load))
|
||||
{
|
||||
if (CPU_FEATURE_USABLE_P (cpu_features, AVX512VL)
|
||||
diff --git a/sysdeps/x86_64/multiarch/memcmp-avx2-movbe.S b/sysdeps/x86_64/multiarch/memcmp-avx2-movbe.S
|
||||
index 9d5c9c72..16fc673e 100644
|
||||
--- a/sysdeps/x86_64/multiarch/memcmp-avx2-movbe.S
|
||||
+++ b/sysdeps/x86_64/multiarch/memcmp-avx2-movbe.S
|
||||
@@ -19,17 +19,23 @@
|
||||
#if IS_IN (libc)
|
||||
|
||||
/* memcmp/wmemcmp is implemented as:
|
||||
- 1. For size from 2 to 7 bytes, load as big endian with movbe and bswap
|
||||
- to avoid branches.
|
||||
- 2. Use overlapping compare to avoid branch.
|
||||
- 3. Use vector compare when size >= 4 bytes for memcmp or size >= 8
|
||||
- bytes for wmemcmp.
|
||||
- 4. If size is 8 * VEC_SIZE or less, unroll the loop.
|
||||
- 5. Compare 4 * VEC_SIZE at a time with the aligned first memory
|
||||
+ 1. Use ymm vector compares when possible. The only case where
|
||||
+ vector compares is not possible for when size < VEC_SIZE
|
||||
+ and loading from either s1 or s2 would cause a page cross.
|
||||
+ 2. For size from 2 to 7 bytes on page cross, load as big endian
|
||||
+ with movbe and bswap to avoid branches.
|
||||
+ 3. Use xmm vector compare when size >= 4 bytes for memcmp or
|
||||
+ size >= 8 bytes for wmemcmp.
|
||||
+ 4. Optimistically compare up to first 4 * VEC_SIZE one at a
|
||||
+ to check for early mismatches. Only do this if its guranteed the
|
||||
+ work is not wasted.
|
||||
+ 5. If size is 8 * VEC_SIZE or less, unroll the loop.
|
||||
+ 6. Compare 4 * VEC_SIZE at a time with the aligned first memory
|
||||
area.
|
||||
- 6. Use 2 vector compares when size is 2 * VEC_SIZE or less.
|
||||
- 7. Use 4 vector compares when size is 4 * VEC_SIZE or less.
|
||||
- 8. Use 8 vector compares when size is 8 * VEC_SIZE or less. */
|
||||
+ 7. Use 2 vector compares when size is 2 * VEC_SIZE or less.
|
||||
+ 8. Use 4 vector compares when size is 4 * VEC_SIZE or less.
|
||||
+ 9. Use 8 vector compares when size is 8 * VEC_SIZE or less. */
|
||||
+
|
||||
|
||||
# include <sysdep.h>
|
||||
|
||||
@@ -38,8 +44,10 @@
|
||||
# endif
|
||||
|
||||
# ifdef USE_AS_WMEMCMP
|
||||
+# define CHAR_SIZE 4
|
||||
# define VPCMPEQ vpcmpeqd
|
||||
# else
|
||||
+# define CHAR_SIZE 1
|
||||
# define VPCMPEQ vpcmpeqb
|
||||
# endif
|
||||
|
||||
@@ -52,7 +60,7 @@
|
||||
# endif
|
||||
|
||||
# define VEC_SIZE 32
|
||||
-# define VEC_MASK ((1 << VEC_SIZE) - 1)
|
||||
+# define PAGE_SIZE 4096
|
||||
|
||||
/* Warning!
|
||||
wmemcmp has to use SIGNED comparison for elements.
|
||||
@@ -71,136 +79,359 @@ ENTRY (MEMCMP)
|
||||
jb L(less_vec)
|
||||
|
||||
/* From VEC to 2 * VEC. No branch when size == VEC_SIZE. */
|
||||
- vmovdqu (%rsi), %ymm2
|
||||
- VPCMPEQ (%rdi), %ymm2, %ymm2
|
||||
- vpmovmskb %ymm2, %eax
|
||||
- subl $VEC_MASK, %eax
|
||||
- jnz L(first_vec)
|
||||
+ vmovdqu (%rsi), %ymm1
|
||||
+ VPCMPEQ (%rdi), %ymm1, %ymm1
|
||||
+ vpmovmskb %ymm1, %eax
|
||||
+ /* NB: eax must be destination register if going to
|
||||
+ L(return_vec_[0,2]). For L(return_vec_3 destination register
|
||||
+ must be ecx. */
|
||||
+ incl %eax
|
||||
+ jnz L(return_vec_0)
|
||||
|
||||
cmpq $(VEC_SIZE * 2), %rdx
|
||||
- jbe L(last_vec)
|
||||
-
|
||||
- VPCMPEQ %ymm0, %ymm0, %ymm0
|
||||
- /* More than 2 * VEC. */
|
||||
- cmpq $(VEC_SIZE * 8), %rdx
|
||||
- ja L(more_8x_vec)
|
||||
- cmpq $(VEC_SIZE * 4), %rdx
|
||||
- jb L(last_4x_vec)
|
||||
-
|
||||
- /* From 4 * VEC to 8 * VEC, inclusively. */
|
||||
- vmovdqu (%rsi), %ymm1
|
||||
- VPCMPEQ (%rdi), %ymm1, %ymm1
|
||||
+ jbe L(last_1x_vec)
|
||||
|
||||
+ /* Check second VEC no matter what. */
|
||||
vmovdqu VEC_SIZE(%rsi), %ymm2
|
||||
- VPCMPEQ VEC_SIZE(%rdi), %ymm2, %ymm2
|
||||
+ VPCMPEQ VEC_SIZE(%rdi), %ymm2, %ymm2
|
||||
+ vpmovmskb %ymm2, %eax
|
||||
+ /* If all 4 VEC where equal eax will be all 1s so incl will
|
||||
+ overflow and set zero flag. */
|
||||
+ incl %eax
|
||||
+ jnz L(return_vec_1)
|
||||
|
||||
- vmovdqu (VEC_SIZE * 2)(%rsi), %ymm3
|
||||
- VPCMPEQ (VEC_SIZE * 2)(%rdi), %ymm3, %ymm3
|
||||
+ /* Less than 4 * VEC. */
|
||||
+ cmpq $(VEC_SIZE * 4), %rdx
|
||||
+ jbe L(last_2x_vec)
|
||||
|
||||
+ /* Check third and fourth VEC no matter what. */
|
||||
+ vmovdqu (VEC_SIZE * 2)(%rsi), %ymm3
|
||||
+ VPCMPEQ (VEC_SIZE * 2)(%rdi), %ymm3, %ymm3
|
||||
+ vpmovmskb %ymm3, %eax
|
||||
+ incl %eax
|
||||
+ jnz L(return_vec_2)
|
||||
vmovdqu (VEC_SIZE * 3)(%rsi), %ymm4
|
||||
- VPCMPEQ (VEC_SIZE * 3)(%rdi), %ymm4, %ymm4
|
||||
+ VPCMPEQ (VEC_SIZE * 3)(%rdi), %ymm4, %ymm4
|
||||
+ vpmovmskb %ymm4, %ecx
|
||||
+ incl %ecx
|
||||
+ jnz L(return_vec_3)
|
||||
|
||||
- vpand %ymm1, %ymm2, %ymm5
|
||||
- vpand %ymm3, %ymm4, %ymm6
|
||||
- vpand %ymm5, %ymm6, %ymm5
|
||||
+ /* Go to 4x VEC loop. */
|
||||
+ cmpq $(VEC_SIZE * 8), %rdx
|
||||
+ ja L(more_8x_vec)
|
||||
|
||||
- vptest %ymm0, %ymm5
|
||||
- jnc L(4x_vec_end)
|
||||
+ /* Handle remainder of size = 4 * VEC + 1 to 8 * VEC without any
|
||||
+ branches. */
|
||||
|
||||
+ /* Load first two VEC from s2 before adjusting addresses. */
|
||||
+ vmovdqu -(VEC_SIZE * 4)(%rsi, %rdx), %ymm1
|
||||
+ vmovdqu -(VEC_SIZE * 3)(%rsi, %rdx), %ymm2
|
||||
leaq -(4 * VEC_SIZE)(%rdi, %rdx), %rdi
|
||||
leaq -(4 * VEC_SIZE)(%rsi, %rdx), %rsi
|
||||
- vmovdqu (%rsi), %ymm1
|
||||
- VPCMPEQ (%rdi), %ymm1, %ymm1
|
||||
|
||||
- vmovdqu VEC_SIZE(%rsi), %ymm2
|
||||
- VPCMPEQ VEC_SIZE(%rdi), %ymm2, %ymm2
|
||||
- vpand %ymm2, %ymm1, %ymm5
|
||||
+ /* Wait to load from s1 until addressed adjust due to
|
||||
+ unlamination of microfusion with complex address mode. */
|
||||
+ VPCMPEQ (%rdi), %ymm1, %ymm1
|
||||
+ VPCMPEQ (VEC_SIZE)(%rdi), %ymm2, %ymm2
|
||||
|
||||
vmovdqu (VEC_SIZE * 2)(%rsi), %ymm3
|
||||
- VPCMPEQ (VEC_SIZE * 2)(%rdi), %ymm3, %ymm3
|
||||
- vpand %ymm3, %ymm5, %ymm5
|
||||
-
|
||||
+ VPCMPEQ (VEC_SIZE * 2)(%rdi), %ymm3, %ymm3
|
||||
vmovdqu (VEC_SIZE * 3)(%rsi), %ymm4
|
||||
- VPCMPEQ (VEC_SIZE * 3)(%rdi), %ymm4, %ymm4
|
||||
- vpand %ymm4, %ymm5, %ymm5
|
||||
+ VPCMPEQ (VEC_SIZE * 3)(%rdi), %ymm4, %ymm4
|
||||
|
||||
- vptest %ymm0, %ymm5
|
||||
- jnc L(4x_vec_end)
|
||||
- xorl %eax, %eax
|
||||
+ /* Reduce VEC0 - VEC4. */
|
||||
+ vpand %ymm1, %ymm2, %ymm5
|
||||
+ vpand %ymm3, %ymm4, %ymm6
|
||||
+ vpand %ymm5, %ymm6, %ymm7
|
||||
+ vpmovmskb %ymm7, %ecx
|
||||
+ incl %ecx
|
||||
+ jnz L(return_vec_0_1_2_3)
|
||||
+ /* NB: eax must be zero to reach here. */
|
||||
+ VZEROUPPER_RETURN
|
||||
+
|
||||
+ .p2align 4
|
||||
+L(return_vec_0):
|
||||
+ tzcntl %eax, %eax
|
||||
+# ifdef USE_AS_WMEMCMP
|
||||
+ movl (%rdi, %rax), %ecx
|
||||
+ xorl %edx, %edx
|
||||
+ cmpl (%rsi, %rax), %ecx
|
||||
+ /* NB: no partial register stall here because xorl zero idiom
|
||||
+ above. */
|
||||
+ setg %dl
|
||||
+ leal -1(%rdx, %rdx), %eax
|
||||
+# else
|
||||
+ movzbl (%rsi, %rax), %ecx
|
||||
+ movzbl (%rdi, %rax), %eax
|
||||
+ subl %ecx, %eax
|
||||
+# endif
|
||||
L(return_vzeroupper):
|
||||
ZERO_UPPER_VEC_REGISTERS_RETURN
|
||||
|
||||
.p2align 4
|
||||
-L(last_2x_vec):
|
||||
- /* From VEC to 2 * VEC. No branch when size == VEC_SIZE. */
|
||||
- vmovdqu (%rsi), %ymm2
|
||||
- VPCMPEQ (%rdi), %ymm2, %ymm2
|
||||
- vpmovmskb %ymm2, %eax
|
||||
- subl $VEC_MASK, %eax
|
||||
- jnz L(first_vec)
|
||||
+L(return_vec_1):
|
||||
+ tzcntl %eax, %eax
|
||||
+# ifdef USE_AS_WMEMCMP
|
||||
+ movl VEC_SIZE(%rdi, %rax), %ecx
|
||||
+ xorl %edx, %edx
|
||||
+ cmpl VEC_SIZE(%rsi, %rax), %ecx
|
||||
+ setg %dl
|
||||
+ leal -1(%rdx, %rdx), %eax
|
||||
+# else
|
||||
+ movzbl VEC_SIZE(%rsi, %rax), %ecx
|
||||
+ movzbl VEC_SIZE(%rdi, %rax), %eax
|
||||
+ subl %ecx, %eax
|
||||
+# endif
|
||||
+ VZEROUPPER_RETURN
|
||||
+
|
||||
+ .p2align 4
|
||||
+L(return_vec_2):
|
||||
+ tzcntl %eax, %eax
|
||||
+# ifdef USE_AS_WMEMCMP
|
||||
+ movl (VEC_SIZE * 2)(%rdi, %rax), %ecx
|
||||
+ xorl %edx, %edx
|
||||
+ cmpl (VEC_SIZE * 2)(%rsi, %rax), %ecx
|
||||
+ setg %dl
|
||||
+ leal -1(%rdx, %rdx), %eax
|
||||
+# else
|
||||
+ movzbl (VEC_SIZE * 2)(%rsi, %rax), %ecx
|
||||
+ movzbl (VEC_SIZE * 2)(%rdi, %rax), %eax
|
||||
+ subl %ecx, %eax
|
||||
+# endif
|
||||
+ VZEROUPPER_RETURN
|
||||
+
|
||||
+ /* NB: p2align 5 here to ensure 4x loop is 32 byte aligned. */
|
||||
+ .p2align 5
|
||||
+L(8x_return_vec_0_1_2_3):
|
||||
+ /* Returning from L(more_8x_vec) requires restoring rsi. */
|
||||
+ addq %rdi, %rsi
|
||||
+L(return_vec_0_1_2_3):
|
||||
+ vpmovmskb %ymm1, %eax
|
||||
+ incl %eax
|
||||
+ jnz L(return_vec_0)
|
||||
|
||||
-L(last_vec):
|
||||
- /* Use overlapping loads to avoid branches. */
|
||||
- leaq -VEC_SIZE(%rdi, %rdx), %rdi
|
||||
- leaq -VEC_SIZE(%rsi, %rdx), %rsi
|
||||
- vmovdqu (%rsi), %ymm2
|
||||
- VPCMPEQ (%rdi), %ymm2, %ymm2
|
||||
vpmovmskb %ymm2, %eax
|
||||
- subl $VEC_MASK, %eax
|
||||
- jnz L(first_vec)
|
||||
+ incl %eax
|
||||
+ jnz L(return_vec_1)
|
||||
+
|
||||
+ vpmovmskb %ymm3, %eax
|
||||
+ incl %eax
|
||||
+ jnz L(return_vec_2)
|
||||
+L(return_vec_3):
|
||||
+ tzcntl %ecx, %ecx
|
||||
+# ifdef USE_AS_WMEMCMP
|
||||
+ movl (VEC_SIZE * 3)(%rdi, %rcx), %eax
|
||||
+ xorl %edx, %edx
|
||||
+ cmpl (VEC_SIZE * 3)(%rsi, %rcx), %eax
|
||||
+ setg %dl
|
||||
+ leal -1(%rdx, %rdx), %eax
|
||||
+# else
|
||||
+ movzbl (VEC_SIZE * 3)(%rdi, %rcx), %eax
|
||||
+ movzbl (VEC_SIZE * 3)(%rsi, %rcx), %ecx
|
||||
+ subl %ecx, %eax
|
||||
+# endif
|
||||
+ VZEROUPPER_RETURN
|
||||
+
|
||||
+ .p2align 4
|
||||
+L(more_8x_vec):
|
||||
+ /* Set end of s1 in rdx. */
|
||||
+ leaq -(VEC_SIZE * 4)(%rdi, %rdx), %rdx
|
||||
+ /* rsi stores s2 - s1. This allows loop to only update one
|
||||
+ pointer. */
|
||||
+ subq %rdi, %rsi
|
||||
+ /* Align s1 pointer. */
|
||||
+ andq $-VEC_SIZE, %rdi
|
||||
+ /* Adjust because first 4x vec where check already. */
|
||||
+ subq $-(VEC_SIZE * 4), %rdi
|
||||
+ .p2align 4
|
||||
+L(loop_4x_vec):
|
||||
+ /* rsi has s2 - s1 so get correct address by adding s1 (in rdi).
|
||||
+ */
|
||||
+ vmovdqu (%rsi, %rdi), %ymm1
|
||||
+ VPCMPEQ (%rdi), %ymm1, %ymm1
|
||||
+
|
||||
+ vmovdqu VEC_SIZE(%rsi, %rdi), %ymm2
|
||||
+ VPCMPEQ VEC_SIZE(%rdi), %ymm2, %ymm2
|
||||
+
|
||||
+ vmovdqu (VEC_SIZE * 2)(%rsi, %rdi), %ymm3
|
||||
+ VPCMPEQ (VEC_SIZE * 2)(%rdi), %ymm3, %ymm3
|
||||
+
|
||||
+ vmovdqu (VEC_SIZE * 3)(%rsi, %rdi), %ymm4
|
||||
+ VPCMPEQ (VEC_SIZE * 3)(%rdi), %ymm4, %ymm4
|
||||
+
|
||||
+ vpand %ymm1, %ymm2, %ymm5
|
||||
+ vpand %ymm3, %ymm4, %ymm6
|
||||
+ vpand %ymm5, %ymm6, %ymm7
|
||||
+ vpmovmskb %ymm7, %ecx
|
||||
+ incl %ecx
|
||||
+ jnz L(8x_return_vec_0_1_2_3)
|
||||
+ subq $-(VEC_SIZE * 4), %rdi
|
||||
+ /* Check if s1 pointer at end. */
|
||||
+ cmpq %rdx, %rdi
|
||||
+ jb L(loop_4x_vec)
|
||||
+
|
||||
+ subq %rdx, %rdi
|
||||
+ /* rdi has 4 * VEC_SIZE - remaining length. */
|
||||
+ cmpl $(VEC_SIZE * 3), %edi
|
||||
+ jae L(8x_last_1x_vec)
|
||||
+ /* Load regardless of branch. */
|
||||
+ vmovdqu (VEC_SIZE * 2)(%rsi, %rdx), %ymm3
|
||||
+ cmpl $(VEC_SIZE * 2), %edi
|
||||
+ jae L(8x_last_2x_vec)
|
||||
+
|
||||
+ /* Check last 4 VEC. */
|
||||
+ vmovdqu (%rsi, %rdx), %ymm1
|
||||
+ VPCMPEQ (%rdx), %ymm1, %ymm1
|
||||
+
|
||||
+ vmovdqu VEC_SIZE(%rsi, %rdx), %ymm2
|
||||
+ VPCMPEQ VEC_SIZE(%rdx), %ymm2, %ymm2
|
||||
+
|
||||
+ VPCMPEQ (VEC_SIZE * 2)(%rdx), %ymm3, %ymm3
|
||||
+
|
||||
+ vmovdqu (VEC_SIZE * 3)(%rsi, %rdx), %ymm4
|
||||
+ VPCMPEQ (VEC_SIZE * 3)(%rdx), %ymm4, %ymm4
|
||||
+
|
||||
+ vpand %ymm1, %ymm2, %ymm5
|
||||
+ vpand %ymm3, %ymm4, %ymm6
|
||||
+ vpand %ymm5, %ymm6, %ymm7
|
||||
+ vpmovmskb %ymm7, %ecx
|
||||
+ /* Restore s1 pointer to rdi. */
|
||||
+ movq %rdx, %rdi
|
||||
+ incl %ecx
|
||||
+ jnz L(8x_return_vec_0_1_2_3)
|
||||
+ /* NB: eax must be zero to reach here. */
|
||||
+ VZEROUPPER_RETURN
|
||||
+
|
||||
+ /* Only entry is from L(more_8x_vec). */
|
||||
+ .p2align 4
|
||||
+L(8x_last_2x_vec):
|
||||
+ /* Check second to last VEC. rdx store end pointer of s1 and
|
||||
+ ymm3 has already been loaded with second to last VEC from s2.
|
||||
+ */
|
||||
+ VPCMPEQ (VEC_SIZE * 2)(%rdx), %ymm3, %ymm3
|
||||
+ vpmovmskb %ymm3, %eax
|
||||
+ incl %eax
|
||||
+ jnz L(8x_return_vec_2)
|
||||
+ /* Check last VEC. */
|
||||
+ .p2align 4
|
||||
+L(8x_last_1x_vec):
|
||||
+ vmovdqu (VEC_SIZE * 3)(%rsi, %rdx), %ymm4
|
||||
+ VPCMPEQ (VEC_SIZE * 3)(%rdx), %ymm4, %ymm4
|
||||
+ vpmovmskb %ymm4, %eax
|
||||
+ incl %eax
|
||||
+ jnz L(8x_return_vec_3)
|
||||
VZEROUPPER_RETURN
|
||||
|
||||
.p2align 4
|
||||
-L(first_vec):
|
||||
- /* A byte or int32 is different within 16 or 32 bytes. */
|
||||
- tzcntl %eax, %ecx
|
||||
+L(last_2x_vec):
|
||||
+ /* Check second to last VEC. */
|
||||
+ vmovdqu -(VEC_SIZE * 2)(%rsi, %rdx), %ymm1
|
||||
+ VPCMPEQ -(VEC_SIZE * 2)(%rdi, %rdx), %ymm1, %ymm1
|
||||
+ vpmovmskb %ymm1, %eax
|
||||
+ incl %eax
|
||||
+ jnz L(return_vec_1_end)
|
||||
+ /* Check last VEC. */
|
||||
+L(last_1x_vec):
|
||||
+ vmovdqu -(VEC_SIZE * 1)(%rsi, %rdx), %ymm1
|
||||
+ VPCMPEQ -(VEC_SIZE * 1)(%rdi, %rdx), %ymm1, %ymm1
|
||||
+ vpmovmskb %ymm1, %eax
|
||||
+ incl %eax
|
||||
+ jnz L(return_vec_0_end)
|
||||
+ VZEROUPPER_RETURN
|
||||
+
|
||||
+ .p2align 4
|
||||
+L(8x_return_vec_2):
|
||||
+ subq $VEC_SIZE, %rdx
|
||||
+L(8x_return_vec_3):
|
||||
+ tzcntl %eax, %eax
|
||||
+ addq %rdx, %rax
|
||||
# ifdef USE_AS_WMEMCMP
|
||||
- xorl %eax, %eax
|
||||
- movl (%rdi, %rcx), %edx
|
||||
- cmpl (%rsi, %rcx), %edx
|
||||
-L(wmemcmp_return):
|
||||
- setl %al
|
||||
- negl %eax
|
||||
- orl $1, %eax
|
||||
+ movl (VEC_SIZE * 3)(%rax), %ecx
|
||||
+ xorl %edx, %edx
|
||||
+ cmpl (VEC_SIZE * 3)(%rsi, %rax), %ecx
|
||||
+ setg %dl
|
||||
+ leal -1(%rdx, %rdx), %eax
|
||||
# else
|
||||
- movzbl (%rdi, %rcx), %eax
|
||||
- movzbl (%rsi, %rcx), %edx
|
||||
- sub %edx, %eax
|
||||
+ movzbl (VEC_SIZE * 3)(%rsi, %rax), %ecx
|
||||
+ movzbl (VEC_SIZE * 3)(%rax), %eax
|
||||
+ subl %ecx, %eax
|
||||
# endif
|
||||
VZEROUPPER_RETURN
|
||||
|
||||
-# ifdef USE_AS_WMEMCMP
|
||||
.p2align 4
|
||||
-L(4):
|
||||
- xorl %eax, %eax
|
||||
- movl (%rdi), %edx
|
||||
- cmpl (%rsi), %edx
|
||||
- jne L(wmemcmp_return)
|
||||
- ret
|
||||
+L(return_vec_1_end):
|
||||
+ tzcntl %eax, %eax
|
||||
+ addl %edx, %eax
|
||||
+# ifdef USE_AS_WMEMCMP
|
||||
+ movl -(VEC_SIZE * 2)(%rdi, %rax), %ecx
|
||||
+ xorl %edx, %edx
|
||||
+ cmpl -(VEC_SIZE * 2)(%rsi, %rax), %ecx
|
||||
+ setg %dl
|
||||
+ leal -1(%rdx, %rdx), %eax
|
||||
# else
|
||||
+ movzbl -(VEC_SIZE * 2)(%rsi, %rax), %ecx
|
||||
+ movzbl -(VEC_SIZE * 2)(%rdi, %rax), %eax
|
||||
+ subl %ecx, %eax
|
||||
+# endif
|
||||
+ VZEROUPPER_RETURN
|
||||
+
|
||||
.p2align 4
|
||||
-L(between_4_7):
|
||||
- /* Load as big endian with overlapping movbe to avoid branches. */
|
||||
- movbe (%rdi), %eax
|
||||
- movbe (%rsi), %ecx
|
||||
- shlq $32, %rax
|
||||
- shlq $32, %rcx
|
||||
- movbe -4(%rdi, %rdx), %edi
|
||||
- movbe -4(%rsi, %rdx), %esi
|
||||
- orq %rdi, %rax
|
||||
- orq %rsi, %rcx
|
||||
- subq %rcx, %rax
|
||||
- je L(exit)
|
||||
- sbbl %eax, %eax
|
||||
- orl $1, %eax
|
||||
- ret
|
||||
+L(return_vec_0_end):
|
||||
+ tzcntl %eax, %eax
|
||||
+ addl %edx, %eax
|
||||
+# ifdef USE_AS_WMEMCMP
|
||||
+ movl -VEC_SIZE(%rdi, %rax), %ecx
|
||||
+ xorl %edx, %edx
|
||||
+ cmpl -VEC_SIZE(%rsi, %rax), %ecx
|
||||
+ setg %dl
|
||||
+ leal -1(%rdx, %rdx), %eax
|
||||
+# else
|
||||
+ movzbl -VEC_SIZE(%rsi, %rax), %ecx
|
||||
+ movzbl -VEC_SIZE(%rdi, %rax), %eax
|
||||
+ subl %ecx, %eax
|
||||
+# endif
|
||||
+ VZEROUPPER_RETURN
|
||||
|
||||
.p2align 4
|
||||
-L(exit):
|
||||
- ret
|
||||
+L(less_vec):
|
||||
+ /* Check if one or less CHAR. This is necessary for size = 0 but
|
||||
+ is also faster for size = CHAR_SIZE. */
|
||||
+ cmpl $CHAR_SIZE, %edx
|
||||
+ jbe L(one_or_less)
|
||||
+
|
||||
+ /* Check if loading one VEC from either s1 or s2 could cause a
|
||||
+ page cross. This can have false positives but is by far the
|
||||
+ fastest method. */
|
||||
+ movl %edi, %eax
|
||||
+ orl %esi, %eax
|
||||
+ andl $(PAGE_SIZE - 1), %eax
|
||||
+ cmpl $(PAGE_SIZE - VEC_SIZE), %eax
|
||||
+ jg L(page_cross_less_vec)
|
||||
+
|
||||
+ /* No page cross possible. */
|
||||
+ vmovdqu (%rsi), %ymm2
|
||||
+ VPCMPEQ (%rdi), %ymm2, %ymm2
|
||||
+ vpmovmskb %ymm2, %eax
|
||||
+ incl %eax
|
||||
+ /* Result will be zero if s1 and s2 match. Otherwise first set
|
||||
+ bit will be first mismatch. */
|
||||
+ bzhil %edx, %eax, %edx
|
||||
+ jnz L(return_vec_0)
|
||||
+ xorl %eax, %eax
|
||||
+ VZEROUPPER_RETURN
|
||||
|
||||
.p2align 4
|
||||
-L(between_2_3):
|
||||
+L(page_cross_less_vec):
|
||||
+ /* if USE_AS_WMEMCMP it can only be 0, 4, 8, 12, 16, 20, 24, 28
|
||||
+ bytes. */
|
||||
+ cmpl $16, %edx
|
||||
+ jae L(between_16_31)
|
||||
+# ifndef USE_AS_WMEMCMP
|
||||
+ cmpl $8, %edx
|
||||
+ jae L(between_8_15)
|
||||
+ cmpl $4, %edx
|
||||
+ jae L(between_4_7)
|
||||
+
|
||||
/* Load as big endian to avoid branches. */
|
||||
movzwl (%rdi), %eax
|
||||
movzwl (%rsi), %ecx
|
||||
@@ -208,223 +439,106 @@ L(between_2_3):
|
||||
shll $8, %ecx
|
||||
bswap %eax
|
||||
bswap %ecx
|
||||
- movb -1(%rdi, %rdx), %al
|
||||
- movb -1(%rsi, %rdx), %cl
|
||||
+ movzbl -1(%rdi, %rdx), %edi
|
||||
+ movzbl -1(%rsi, %rdx), %esi
|
||||
+ orl %edi, %eax
|
||||
+ orl %esi, %ecx
|
||||
/* Subtraction is okay because the upper 8 bits are zero. */
|
||||
subl %ecx, %eax
|
||||
+ /* No ymm register was touched. */
|
||||
ret
|
||||
|
||||
.p2align 4
|
||||
-L(1):
|
||||
- movzbl (%rdi), %eax
|
||||
+L(one_or_less):
|
||||
+ jb L(zero)
|
||||
movzbl (%rsi), %ecx
|
||||
+ movzbl (%rdi), %eax
|
||||
subl %ecx, %eax
|
||||
- ret
|
||||
-# endif
|
||||
-
|
||||
- .p2align 4
|
||||
-L(zero):
|
||||
- xorl %eax, %eax
|
||||
+ /* No ymm register was touched. */
|
||||
ret
|
||||
|
||||
.p2align 4
|
||||
-L(less_vec):
|
||||
-# ifdef USE_AS_WMEMCMP
|
||||
- /* It can only be 0, 4, 8, 12, 16, 20, 24, 28 bytes. */
|
||||
- cmpb $4, %dl
|
||||
- je L(4)
|
||||
- jb L(zero)
|
||||
-# else
|
||||
- cmpb $1, %dl
|
||||
- je L(1)
|
||||
- jb L(zero)
|
||||
- cmpb $4, %dl
|
||||
- jb L(between_2_3)
|
||||
- cmpb $8, %dl
|
||||
- jb L(between_4_7)
|
||||
+L(between_8_15):
|
||||
# endif
|
||||
- cmpb $16, %dl
|
||||
- jae L(between_16_31)
|
||||
- /* It is between 8 and 15 bytes. */
|
||||
+ /* If USE_AS_WMEMCMP fall through into 8-15 byte case. */
|
||||
vmovq (%rdi), %xmm1
|
||||
vmovq (%rsi), %xmm2
|
||||
- VPCMPEQ %xmm1, %xmm2, %xmm2
|
||||
+ VPCMPEQ %xmm1, %xmm2, %xmm2
|
||||
vpmovmskb %xmm2, %eax
|
||||
- subl $0xffff, %eax
|
||||
- jnz L(first_vec)
|
||||
+ subl $0xffff, %eax
|
||||
+ jnz L(return_vec_0)
|
||||
/* Use overlapping loads to avoid branches. */
|
||||
leaq -8(%rdi, %rdx), %rdi
|
||||
leaq -8(%rsi, %rdx), %rsi
|
||||
vmovq (%rdi), %xmm1
|
||||
vmovq (%rsi), %xmm2
|
||||
- VPCMPEQ %xmm1, %xmm2, %xmm2
|
||||
+ VPCMPEQ %xmm1, %xmm2, %xmm2
|
||||
vpmovmskb %xmm2, %eax
|
||||
- subl $0xffff, %eax
|
||||
- jnz L(first_vec)
|
||||
+ subl $0xffff, %eax
|
||||
+ jnz L(return_vec_0)
|
||||
+ /* No ymm register was touched. */
|
||||
+ ret
|
||||
+
|
||||
+ .p2align 4
|
||||
+L(zero):
|
||||
+ xorl %eax, %eax
|
||||
ret
|
||||
|
||||
.p2align 4
|
||||
L(between_16_31):
|
||||
/* From 16 to 31 bytes. No branch when size == 16. */
|
||||
vmovdqu (%rsi), %xmm2
|
||||
- VPCMPEQ (%rdi), %xmm2, %xmm2
|
||||
+ VPCMPEQ (%rdi), %xmm2, %xmm2
|
||||
vpmovmskb %xmm2, %eax
|
||||
- subl $0xffff, %eax
|
||||
- jnz L(first_vec)
|
||||
+ subl $0xffff, %eax
|
||||
+ jnz L(return_vec_0)
|
||||
|
||||
/* Use overlapping loads to avoid branches. */
|
||||
+
|
||||
+ vmovdqu -16(%rsi, %rdx), %xmm2
|
||||
leaq -16(%rdi, %rdx), %rdi
|
||||
leaq -16(%rsi, %rdx), %rsi
|
||||
- vmovdqu (%rsi), %xmm2
|
||||
- VPCMPEQ (%rdi), %xmm2, %xmm2
|
||||
+ VPCMPEQ (%rdi), %xmm2, %xmm2
|
||||
vpmovmskb %xmm2, %eax
|
||||
- subl $0xffff, %eax
|
||||
- jnz L(first_vec)
|
||||
+ subl $0xffff, %eax
|
||||
+ jnz L(return_vec_0)
|
||||
+ /* No ymm register was touched. */
|
||||
ret
|
||||
|
||||
- .p2align 4
|
||||
-L(more_8x_vec):
|
||||
- /* More than 8 * VEC. Check the first VEC. */
|
||||
- vmovdqu (%rsi), %ymm2
|
||||
- VPCMPEQ (%rdi), %ymm2, %ymm2
|
||||
- vpmovmskb %ymm2, %eax
|
||||
- subl $VEC_MASK, %eax
|
||||
- jnz L(first_vec)
|
||||
-
|
||||
- /* Align the first memory area for aligned loads in the loop.
|
||||
- Compute how much the first memory area is misaligned. */
|
||||
- movq %rdi, %rcx
|
||||
- andl $(VEC_SIZE - 1), %ecx
|
||||
- /* Get the negative of offset for alignment. */
|
||||
- subq $VEC_SIZE, %rcx
|
||||
- /* Adjust the second memory area. */
|
||||
- subq %rcx, %rsi
|
||||
- /* Adjust the first memory area which should be aligned now. */
|
||||
- subq %rcx, %rdi
|
||||
- /* Adjust length. */
|
||||
- addq %rcx, %rdx
|
||||
-
|
||||
-L(loop_4x_vec):
|
||||
- /* Compare 4 * VEC at a time forward. */
|
||||
- vmovdqu (%rsi), %ymm1
|
||||
- VPCMPEQ (%rdi), %ymm1, %ymm1
|
||||
-
|
||||
- vmovdqu VEC_SIZE(%rsi), %ymm2
|
||||
- VPCMPEQ VEC_SIZE(%rdi), %ymm2, %ymm2
|
||||
- vpand %ymm2, %ymm1, %ymm5
|
||||
-
|
||||
- vmovdqu (VEC_SIZE * 2)(%rsi), %ymm3
|
||||
- VPCMPEQ (VEC_SIZE * 2)(%rdi), %ymm3, %ymm3
|
||||
- vpand %ymm3, %ymm5, %ymm5
|
||||
-
|
||||
- vmovdqu (VEC_SIZE * 3)(%rsi), %ymm4
|
||||
- VPCMPEQ (VEC_SIZE * 3)(%rdi), %ymm4, %ymm4
|
||||
- vpand %ymm4, %ymm5, %ymm5
|
||||
-
|
||||
- vptest %ymm0, %ymm5
|
||||
- jnc L(4x_vec_end)
|
||||
-
|
||||
- addq $(VEC_SIZE * 4), %rdi
|
||||
- addq $(VEC_SIZE * 4), %rsi
|
||||
-
|
||||
- subq $(VEC_SIZE * 4), %rdx
|
||||
- cmpq $(VEC_SIZE * 4), %rdx
|
||||
- jae L(loop_4x_vec)
|
||||
-
|
||||
- /* Less than 4 * VEC. */
|
||||
- cmpq $VEC_SIZE, %rdx
|
||||
- jbe L(last_vec)
|
||||
- cmpq $(VEC_SIZE * 2), %rdx
|
||||
- jbe L(last_2x_vec)
|
||||
-
|
||||
-L(last_4x_vec):
|
||||
- /* From 2 * VEC to 4 * VEC. */
|
||||
- vmovdqu (%rsi), %ymm2
|
||||
- VPCMPEQ (%rdi), %ymm2, %ymm2
|
||||
- vpmovmskb %ymm2, %eax
|
||||
- subl $VEC_MASK, %eax
|
||||
- jnz L(first_vec)
|
||||
-
|
||||
- addq $VEC_SIZE, %rdi
|
||||
- addq $VEC_SIZE, %rsi
|
||||
- vmovdqu (%rsi), %ymm2
|
||||
- VPCMPEQ (%rdi), %ymm2, %ymm2
|
||||
- vpmovmskb %ymm2, %eax
|
||||
- subl $VEC_MASK, %eax
|
||||
- jnz L(first_vec)
|
||||
-
|
||||
- /* Use overlapping loads to avoid branches. */
|
||||
- leaq -(3 * VEC_SIZE)(%rdi, %rdx), %rdi
|
||||
- leaq -(3 * VEC_SIZE)(%rsi, %rdx), %rsi
|
||||
- vmovdqu (%rsi), %ymm2
|
||||
- VPCMPEQ (%rdi), %ymm2, %ymm2
|
||||
- vpmovmskb %ymm2, %eax
|
||||
- subl $VEC_MASK, %eax
|
||||
- jnz L(first_vec)
|
||||
-
|
||||
- addq $VEC_SIZE, %rdi
|
||||
- addq $VEC_SIZE, %rsi
|
||||
- vmovdqu (%rsi), %ymm2
|
||||
- VPCMPEQ (%rdi), %ymm2, %ymm2
|
||||
- vpmovmskb %ymm2, %eax
|
||||
- subl $VEC_MASK, %eax
|
||||
- jnz L(first_vec)
|
||||
- VZEROUPPER_RETURN
|
||||
-
|
||||
- .p2align 4
|
||||
-L(4x_vec_end):
|
||||
- vpmovmskb %ymm1, %eax
|
||||
- subl $VEC_MASK, %eax
|
||||
- jnz L(first_vec)
|
||||
- vpmovmskb %ymm2, %eax
|
||||
- subl $VEC_MASK, %eax
|
||||
- jnz L(first_vec_x1)
|
||||
- vpmovmskb %ymm3, %eax
|
||||
- subl $VEC_MASK, %eax
|
||||
- jnz L(first_vec_x2)
|
||||
- vpmovmskb %ymm4, %eax
|
||||
- subl $VEC_MASK, %eax
|
||||
- tzcntl %eax, %ecx
|
||||
# ifdef USE_AS_WMEMCMP
|
||||
- xorl %eax, %eax
|
||||
- movl (VEC_SIZE * 3)(%rdi, %rcx), %edx
|
||||
- cmpl (VEC_SIZE * 3)(%rsi, %rcx), %edx
|
||||
- jmp L(wmemcmp_return)
|
||||
-# else
|
||||
- movzbl (VEC_SIZE * 3)(%rdi, %rcx), %eax
|
||||
- movzbl (VEC_SIZE * 3)(%rsi, %rcx), %edx
|
||||
- sub %edx, %eax
|
||||
-# endif
|
||||
- VZEROUPPER_RETURN
|
||||
-
|
||||
.p2align 4
|
||||
-L(first_vec_x1):
|
||||
- tzcntl %eax, %ecx
|
||||
-# ifdef USE_AS_WMEMCMP
|
||||
- xorl %eax, %eax
|
||||
- movl VEC_SIZE(%rdi, %rcx), %edx
|
||||
- cmpl VEC_SIZE(%rsi, %rcx), %edx
|
||||
- jmp L(wmemcmp_return)
|
||||
+L(one_or_less):
|
||||
+ jb L(zero)
|
||||
+ movl (%rdi), %ecx
|
||||
+ xorl %edx, %edx
|
||||
+ cmpl (%rsi), %ecx
|
||||
+ je L(zero)
|
||||
+ setg %dl
|
||||
+ leal -1(%rdx, %rdx), %eax
|
||||
+ /* No ymm register was touched. */
|
||||
+ ret
|
||||
# else
|
||||
- movzbl VEC_SIZE(%rdi, %rcx), %eax
|
||||
- movzbl VEC_SIZE(%rsi, %rcx), %edx
|
||||
- sub %edx, %eax
|
||||
-# endif
|
||||
- VZEROUPPER_RETURN
|
||||
|
||||
.p2align 4
|
||||
-L(first_vec_x2):
|
||||
- tzcntl %eax, %ecx
|
||||
-# ifdef USE_AS_WMEMCMP
|
||||
- xorl %eax, %eax
|
||||
- movl (VEC_SIZE * 2)(%rdi, %rcx), %edx
|
||||
- cmpl (VEC_SIZE * 2)(%rsi, %rcx), %edx
|
||||
- jmp L(wmemcmp_return)
|
||||
-# else
|
||||
- movzbl (VEC_SIZE * 2)(%rdi, %rcx), %eax
|
||||
- movzbl (VEC_SIZE * 2)(%rsi, %rcx), %edx
|
||||
- sub %edx, %eax
|
||||
+L(between_4_7):
|
||||
+ /* Load as big endian with overlapping movbe to avoid branches.
|
||||
+ */
|
||||
+ movbe (%rdi), %eax
|
||||
+ movbe (%rsi), %ecx
|
||||
+ shlq $32, %rax
|
||||
+ shlq $32, %rcx
|
||||
+ movbe -4(%rdi, %rdx), %edi
|
||||
+ movbe -4(%rsi, %rdx), %esi
|
||||
+ orq %rdi, %rax
|
||||
+ orq %rsi, %rcx
|
||||
+ subq %rcx, %rax
|
||||
+ jz L(zero_4_7)
|
||||
+ sbbl %eax, %eax
|
||||
+ orl $1, %eax
|
||||
+L(zero_4_7):
|
||||
+ /* No ymm register was touched. */
|
||||
+ ret
|
||||
# endif
|
||||
- VZEROUPPER_RETURN
|
||||
+
|
||||
END (MEMCMP)
|
||||
#endif
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,851 +0,0 @@
|
||||
From 4ad473e97acdc5f6d811755b67c09f2128a644ce Mon Sep 17 00:00:00 2001
|
||||
From: Noah Goldstein <goldstein.w.n@gmail.com>
|
||||
Date: Mon, 17 May 2021 13:57:24 -0400
|
||||
Subject: [PATCH] x86: Optimize memcmp-evex-movbe.S
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
No bug. This commit optimizes memcmp-evex.S. The optimizations include
|
||||
adding a new vec compare path for small sizes, reorganizing the entry
|
||||
control flow, removing some unnecissary ALU instructions from the main
|
||||
loop, and most importantly replacing the heavy use of vpcmp + kand
|
||||
logic with vpxor + vptern. test-memcmp and test-wmemcmp are both
|
||||
passing.
|
||||
|
||||
Signed-off-by: Noah Goldstein <goldstein.w.n@gmail.com>
|
||||
Reviewed-by: H.J. Lu <hjl.tools@gmail.com>
|
||||
---
|
||||
sysdeps/x86_64/multiarch/memcmp-evex-movbe.S | 710 +++++++++++--------
|
||||
1 file changed, 408 insertions(+), 302 deletions(-)
|
||||
|
||||
diff --git a/sysdeps/x86_64/multiarch/memcmp-evex-movbe.S b/sysdeps/x86_64/multiarch/memcmp-evex-movbe.S
|
||||
index 9c093972..654dc7ac 100644
|
||||
--- a/sysdeps/x86_64/multiarch/memcmp-evex-movbe.S
|
||||
+++ b/sysdeps/x86_64/multiarch/memcmp-evex-movbe.S
|
||||
@@ -19,17 +19,22 @@
|
||||
#if IS_IN (libc)
|
||||
|
||||
/* memcmp/wmemcmp is implemented as:
|
||||
- 1. For size from 2 to 7 bytes, load as big endian with movbe and bswap
|
||||
- to avoid branches.
|
||||
- 2. Use overlapping compare to avoid branch.
|
||||
- 3. Use vector compare when size >= 4 bytes for memcmp or size >= 8
|
||||
- bytes for wmemcmp.
|
||||
- 4. If size is 8 * VEC_SIZE or less, unroll the loop.
|
||||
- 5. Compare 4 * VEC_SIZE at a time with the aligned first memory
|
||||
+ 1. Use ymm vector compares when possible. The only case where
|
||||
+ vector compares is not possible for when size < CHAR_PER_VEC
|
||||
+ and loading from either s1 or s2 would cause a page cross.
|
||||
+ 2. For size from 2 to 7 bytes on page cross, load as big endian
|
||||
+ with movbe and bswap to avoid branches.
|
||||
+ 3. Use xmm vector compare when size >= 4 bytes for memcmp or
|
||||
+ size >= 8 bytes for wmemcmp.
|
||||
+ 4. Optimistically compare up to first 4 * CHAR_PER_VEC one at a
|
||||
+ to check for early mismatches. Only do this if its guranteed the
|
||||
+ work is not wasted.
|
||||
+ 5. If size is 8 * VEC_SIZE or less, unroll the loop.
|
||||
+ 6. Compare 4 * VEC_SIZE at a time with the aligned first memory
|
||||
area.
|
||||
- 6. Use 2 vector compares when size is 2 * VEC_SIZE or less.
|
||||
- 7. Use 4 vector compares when size is 4 * VEC_SIZE or less.
|
||||
- 8. Use 8 vector compares when size is 8 * VEC_SIZE or less. */
|
||||
+ 7. Use 2 vector compares when size is 2 * CHAR_PER_VEC or less.
|
||||
+ 8. Use 4 vector compares when size is 4 * CHAR_PER_VEC or less.
|
||||
+ 9. Use 8 vector compares when size is 8 * CHAR_PER_VEC or less. */
|
||||
|
||||
# include <sysdep.h>
|
||||
|
||||
@@ -40,11 +45,21 @@
|
||||
# define VMOVU vmovdqu64
|
||||
|
||||
# ifdef USE_AS_WMEMCMP
|
||||
-# define VPCMPEQ vpcmpeqd
|
||||
+# define CHAR_SIZE 4
|
||||
+# define VPCMP vpcmpd
|
||||
# else
|
||||
-# define VPCMPEQ vpcmpeqb
|
||||
+# define CHAR_SIZE 1
|
||||
+# define VPCMP vpcmpub
|
||||
# endif
|
||||
|
||||
+# define VEC_SIZE 32
|
||||
+# define PAGE_SIZE 4096
|
||||
+# define CHAR_PER_VEC (VEC_SIZE / CHAR_SIZE)
|
||||
+
|
||||
+# define XMM0 xmm16
|
||||
+# define XMM1 xmm17
|
||||
+# define XMM2 xmm18
|
||||
+# define YMM0 ymm16
|
||||
# define XMM1 xmm17
|
||||
# define XMM2 xmm18
|
||||
# define YMM1 ymm17
|
||||
@@ -54,15 +69,6 @@
|
||||
# define YMM5 ymm21
|
||||
# define YMM6 ymm22
|
||||
|
||||
-# define VEC_SIZE 32
|
||||
-# ifdef USE_AS_WMEMCMP
|
||||
-# define VEC_MASK 0xff
|
||||
-# define XMM_MASK 0xf
|
||||
-# else
|
||||
-# define VEC_MASK 0xffffffff
|
||||
-# define XMM_MASK 0xffff
|
||||
-# endif
|
||||
-
|
||||
/* Warning!
|
||||
wmemcmp has to use SIGNED comparison for elements.
|
||||
memcmp has to use UNSIGNED comparison for elemnts.
|
||||
@@ -70,145 +76,370 @@
|
||||
|
||||
.section .text.evex,"ax",@progbits
|
||||
ENTRY (MEMCMP)
|
||||
-# ifdef USE_AS_WMEMCMP
|
||||
- shl $2, %RDX_LP
|
||||
-# elif defined __ILP32__
|
||||
+# ifdef __ILP32__
|
||||
/* Clear the upper 32 bits. */
|
||||
movl %edx, %edx
|
||||
# endif
|
||||
- cmp $VEC_SIZE, %RDX_LP
|
||||
+ cmp $CHAR_PER_VEC, %RDX_LP
|
||||
jb L(less_vec)
|
||||
|
||||
/* From VEC to 2 * VEC. No branch when size == VEC_SIZE. */
|
||||
- VMOVU (%rsi), %YMM2
|
||||
- VPCMPEQ (%rdi), %YMM2, %k1
|
||||
+ VMOVU (%rsi), %YMM1
|
||||
+ /* Use compare not equals to directly check for mismatch. */
|
||||
+ VPCMP $4, (%rdi), %YMM1, %k1
|
||||
kmovd %k1, %eax
|
||||
- subl $VEC_MASK, %eax
|
||||
- jnz L(first_vec)
|
||||
-
|
||||
- cmpq $(VEC_SIZE * 2), %rdx
|
||||
- jbe L(last_vec)
|
||||
-
|
||||
- /* More than 2 * VEC. */
|
||||
- cmpq $(VEC_SIZE * 8), %rdx
|
||||
- ja L(more_8x_vec)
|
||||
- cmpq $(VEC_SIZE * 4), %rdx
|
||||
- jb L(last_4x_vec)
|
||||
+ /* NB: eax must be destination register if going to
|
||||
+ L(return_vec_[0,2]). For L(return_vec_3 destination register
|
||||
+ must be ecx. */
|
||||
+ testl %eax, %eax
|
||||
+ jnz L(return_vec_0)
|
||||
|
||||
- /* From 4 * VEC to 8 * VEC, inclusively. */
|
||||
- VMOVU (%rsi), %YMM1
|
||||
- VPCMPEQ (%rdi), %YMM1, %k1
|
||||
+ cmpq $(CHAR_PER_VEC * 2), %rdx
|
||||
+ jbe L(last_1x_vec)
|
||||
|
||||
+ /* Check second VEC no matter what. */
|
||||
VMOVU VEC_SIZE(%rsi), %YMM2
|
||||
- VPCMPEQ VEC_SIZE(%rdi), %YMM2, %k2
|
||||
+ VPCMP $4, VEC_SIZE(%rdi), %YMM2, %k1
|
||||
+ kmovd %k1, %eax
|
||||
+ testl %eax, %eax
|
||||
+ jnz L(return_vec_1)
|
||||
+
|
||||
+ /* Less than 4 * VEC. */
|
||||
+ cmpq $(CHAR_PER_VEC * 4), %rdx
|
||||
+ jbe L(last_2x_vec)
|
||||
|
||||
+ /* Check third and fourth VEC no matter what. */
|
||||
VMOVU (VEC_SIZE * 2)(%rsi), %YMM3
|
||||
- VPCMPEQ (VEC_SIZE * 2)(%rdi), %YMM3, %k3
|
||||
+ VPCMP $4, (VEC_SIZE * 2)(%rdi), %YMM3, %k1
|
||||
+ kmovd %k1, %eax
|
||||
+ testl %eax, %eax
|
||||
+ jnz L(return_vec_2)
|
||||
|
||||
VMOVU (VEC_SIZE * 3)(%rsi), %YMM4
|
||||
- VPCMPEQ (VEC_SIZE * 3)(%rdi), %YMM4, %k4
|
||||
+ VPCMP $4, (VEC_SIZE * 3)(%rdi), %YMM4, %k1
|
||||
+ kmovd %k1, %ecx
|
||||
+ testl %ecx, %ecx
|
||||
+ jnz L(return_vec_3)
|
||||
|
||||
- kandd %k1, %k2, %k5
|
||||
- kandd %k3, %k4, %k6
|
||||
- kandd %k5, %k6, %k6
|
||||
+ /* Zero YMM0. 4x VEC reduction is done with vpxor + vtern so
|
||||
+ compare with zero to get a mask is needed. */
|
||||
+ vpxorq %XMM0, %XMM0, %XMM0
|
||||
|
||||
- kmovd %k6, %eax
|
||||
- cmpl $VEC_MASK, %eax
|
||||
- jne L(4x_vec_end)
|
||||
+ /* Go to 4x VEC loop. */
|
||||
+ cmpq $(CHAR_PER_VEC * 8), %rdx
|
||||
+ ja L(more_8x_vec)
|
||||
|
||||
- leaq -(4 * VEC_SIZE)(%rdi, %rdx), %rdi
|
||||
- leaq -(4 * VEC_SIZE)(%rsi, %rdx), %rsi
|
||||
- VMOVU (%rsi), %YMM1
|
||||
- VPCMPEQ (%rdi), %YMM1, %k1
|
||||
+ /* Handle remainder of size = 4 * VEC + 1 to 8 * VEC without any
|
||||
+ branches. */
|
||||
|
||||
- VMOVU VEC_SIZE(%rsi), %YMM2
|
||||
- VPCMPEQ VEC_SIZE(%rdi), %YMM2, %k2
|
||||
- kandd %k1, %k2, %k5
|
||||
+ /* Load first two VEC from s2 before adjusting addresses. */
|
||||
+ VMOVU -(VEC_SIZE * 4)(%rsi, %rdx, CHAR_SIZE), %YMM1
|
||||
+ VMOVU -(VEC_SIZE * 3)(%rsi, %rdx, CHAR_SIZE), %YMM2
|
||||
+ leaq -(4 * VEC_SIZE)(%rdi, %rdx, CHAR_SIZE), %rdi
|
||||
+ leaq -(4 * VEC_SIZE)(%rsi, %rdx, CHAR_SIZE), %rsi
|
||||
+
|
||||
+ /* Wait to load from s1 until addressed adjust due to
|
||||
+ unlamination of microfusion with complex address mode. */
|
||||
+
|
||||
+ /* vpxor will be all 0s if s1 and s2 are equal. Otherwise it
|
||||
+ will have some 1s. */
|
||||
+ vpxorq (%rdi), %YMM1, %YMM1
|
||||
+ vpxorq (VEC_SIZE)(%rdi), %YMM2, %YMM2
|
||||
|
||||
VMOVU (VEC_SIZE * 2)(%rsi), %YMM3
|
||||
- VPCMPEQ (VEC_SIZE * 2)(%rdi), %YMM3, %k3
|
||||
- kandd %k3, %k5, %k5
|
||||
+ vpxorq (VEC_SIZE * 2)(%rdi), %YMM3, %YMM3
|
||||
+ /* Or together YMM1, YMM2, and YMM3 into YMM3. */
|
||||
+ vpternlogd $0xfe, %YMM1, %YMM2, %YMM3
|
||||
|
||||
VMOVU (VEC_SIZE * 3)(%rsi), %YMM4
|
||||
- VPCMPEQ (VEC_SIZE * 3)(%rdi), %YMM4, %k4
|
||||
- kandd %k4, %k5, %k5
|
||||
+ /* Ternary logic to xor (VEC_SIZE * 3)(%rdi) with YMM4 while
|
||||
+ oring with YMM3. Result is stored in YMM4. */
|
||||
+ vpternlogd $0xde, (VEC_SIZE * 3)(%rdi), %YMM3, %YMM4
|
||||
+ /* Compare YMM4 with 0. If any 1s s1 and s2 don't match. */
|
||||
+ VPCMP $4, %YMM4, %YMM0, %k1
|
||||
+ kmovd %k1, %ecx
|
||||
+ testl %ecx, %ecx
|
||||
+ jnz L(return_vec_0_1_2_3)
|
||||
+ /* NB: eax must be zero to reach here. */
|
||||
+ ret
|
||||
|
||||
- kmovd %k5, %eax
|
||||
- cmpl $VEC_MASK, %eax
|
||||
- jne L(4x_vec_end)
|
||||
- xorl %eax, %eax
|
||||
+ /* NB: aligning 32 here allows for the rest of the jump targets
|
||||
+ to be tuned for 32 byte alignment. Most important this ensures
|
||||
+ the L(more_8x_vec) loop is 32 byte aligned. */
|
||||
+ .p2align 5
|
||||
+L(less_vec):
|
||||
+ /* Check if one or less CHAR. This is necessary for size = 0 but
|
||||
+ is also faster for size = CHAR_SIZE. */
|
||||
+ cmpl $1, %edx
|
||||
+ jbe L(one_or_less)
|
||||
+
|
||||
+ /* Check if loading one VEC from either s1 or s2 could cause a
|
||||
+ page cross. This can have false positives but is by far the
|
||||
+ fastest method. */
|
||||
+ movl %edi, %eax
|
||||
+ orl %esi, %eax
|
||||
+ andl $(PAGE_SIZE - 1), %eax
|
||||
+ cmpl $(PAGE_SIZE - VEC_SIZE), %eax
|
||||
+ jg L(page_cross_less_vec)
|
||||
+
|
||||
+ /* No page cross possible. */
|
||||
+ VMOVU (%rsi), %YMM2
|
||||
+ VPCMP $4, (%rdi), %YMM2, %k1
|
||||
+ kmovd %k1, %eax
|
||||
+ /* Create mask in ecx for potentially in bound matches. */
|
||||
+ bzhil %edx, %eax, %eax
|
||||
+ jnz L(return_vec_0)
|
||||
ret
|
||||
|
||||
.p2align 4
|
||||
-L(last_2x_vec):
|
||||
- /* From VEC to 2 * VEC. No branch when size == VEC_SIZE. */
|
||||
- VMOVU (%rsi), %YMM2
|
||||
- VPCMPEQ (%rdi), %YMM2, %k2
|
||||
- kmovd %k2, %eax
|
||||
- subl $VEC_MASK, %eax
|
||||
- jnz L(first_vec)
|
||||
+L(return_vec_0):
|
||||
+ tzcntl %eax, %eax
|
||||
+# ifdef USE_AS_WMEMCMP
|
||||
+ movl (%rdi, %rax, CHAR_SIZE), %ecx
|
||||
+ xorl %edx, %edx
|
||||
+ cmpl (%rsi, %rax, CHAR_SIZE), %ecx
|
||||
+ /* NB: no partial register stall here because xorl zero idiom
|
||||
+ above. */
|
||||
+ setg %dl
|
||||
+ leal -1(%rdx, %rdx), %eax
|
||||
+# else
|
||||
+ movzbl (%rsi, %rax), %ecx
|
||||
+ movzbl (%rdi, %rax), %eax
|
||||
+ subl %ecx, %eax
|
||||
+# endif
|
||||
+ ret
|
||||
|
||||
-L(last_vec):
|
||||
- /* Use overlapping loads to avoid branches. */
|
||||
- leaq -VEC_SIZE(%rdi, %rdx), %rdi
|
||||
- leaq -VEC_SIZE(%rsi, %rdx), %rsi
|
||||
- VMOVU (%rsi), %YMM2
|
||||
- VPCMPEQ (%rdi), %YMM2, %k2
|
||||
- kmovd %k2, %eax
|
||||
- subl $VEC_MASK, %eax
|
||||
- jnz L(first_vec)
|
||||
+ /* NB: No p2align necessary. Alignment % 16 is naturally 1
|
||||
+ which is good enough for a target not in a loop. */
|
||||
+L(return_vec_1):
|
||||
+ tzcntl %eax, %eax
|
||||
+# ifdef USE_AS_WMEMCMP
|
||||
+ movl VEC_SIZE(%rdi, %rax, CHAR_SIZE), %ecx
|
||||
+ xorl %edx, %edx
|
||||
+ cmpl VEC_SIZE(%rsi, %rax, CHAR_SIZE), %ecx
|
||||
+ setg %dl
|
||||
+ leal -1(%rdx, %rdx), %eax
|
||||
+# else
|
||||
+ movzbl VEC_SIZE(%rsi, %rax), %ecx
|
||||
+ movzbl VEC_SIZE(%rdi, %rax), %eax
|
||||
+ subl %ecx, %eax
|
||||
+# endif
|
||||
ret
|
||||
|
||||
- .p2align 4
|
||||
-L(first_vec):
|
||||
- /* A byte or int32 is different within 16 or 32 bytes. */
|
||||
- tzcntl %eax, %ecx
|
||||
+ /* NB: No p2align necessary. Alignment % 16 is naturally 2
|
||||
+ which is good enough for a target not in a loop. */
|
||||
+L(return_vec_2):
|
||||
+ tzcntl %eax, %eax
|
||||
# ifdef USE_AS_WMEMCMP
|
||||
- xorl %eax, %eax
|
||||
- movl (%rdi, %rcx, 4), %edx
|
||||
- cmpl (%rsi, %rcx, 4), %edx
|
||||
-L(wmemcmp_return):
|
||||
- setl %al
|
||||
- negl %eax
|
||||
- orl $1, %eax
|
||||
+ movl (VEC_SIZE * 2)(%rdi, %rax, CHAR_SIZE), %ecx
|
||||
+ xorl %edx, %edx
|
||||
+ cmpl (VEC_SIZE * 2)(%rsi, %rax, CHAR_SIZE), %ecx
|
||||
+ setg %dl
|
||||
+ leal -1(%rdx, %rdx), %eax
|
||||
# else
|
||||
- movzbl (%rdi, %rcx), %eax
|
||||
- movzbl (%rsi, %rcx), %edx
|
||||
- sub %edx, %eax
|
||||
+ movzbl (VEC_SIZE * 2)(%rsi, %rax), %ecx
|
||||
+ movzbl (VEC_SIZE * 2)(%rdi, %rax), %eax
|
||||
+ subl %ecx, %eax
|
||||
# endif
|
||||
ret
|
||||
|
||||
+ .p2align 4
|
||||
+L(8x_return_vec_0_1_2_3):
|
||||
+ /* Returning from L(more_8x_vec) requires restoring rsi. */
|
||||
+ addq %rdi, %rsi
|
||||
+L(return_vec_0_1_2_3):
|
||||
+ VPCMP $4, %YMM1, %YMM0, %k0
|
||||
+ kmovd %k0, %eax
|
||||
+ testl %eax, %eax
|
||||
+ jnz L(return_vec_0)
|
||||
+
|
||||
+ VPCMP $4, %YMM2, %YMM0, %k0
|
||||
+ kmovd %k0, %eax
|
||||
+ testl %eax, %eax
|
||||
+ jnz L(return_vec_1)
|
||||
+
|
||||
+ VPCMP $4, %YMM3, %YMM0, %k0
|
||||
+ kmovd %k0, %eax
|
||||
+ testl %eax, %eax
|
||||
+ jnz L(return_vec_2)
|
||||
+L(return_vec_3):
|
||||
+ tzcntl %ecx, %ecx
|
||||
# ifdef USE_AS_WMEMCMP
|
||||
+ movl (VEC_SIZE * 3)(%rdi, %rcx, CHAR_SIZE), %eax
|
||||
+ xorl %edx, %edx
|
||||
+ cmpl (VEC_SIZE * 3)(%rsi, %rcx, CHAR_SIZE), %eax
|
||||
+ setg %dl
|
||||
+ leal -1(%rdx, %rdx), %eax
|
||||
+# else
|
||||
+ movzbl (VEC_SIZE * 3)(%rdi, %rcx), %eax
|
||||
+ movzbl (VEC_SIZE * 3)(%rsi, %rcx), %ecx
|
||||
+ subl %ecx, %eax
|
||||
+# endif
|
||||
+ ret
|
||||
+
|
||||
.p2align 4
|
||||
-L(4):
|
||||
- xorl %eax, %eax
|
||||
- movl (%rdi), %edx
|
||||
- cmpl (%rsi), %edx
|
||||
- jne L(wmemcmp_return)
|
||||
+L(more_8x_vec):
|
||||
+ /* Set end of s1 in rdx. */
|
||||
+ leaq -(VEC_SIZE * 4)(%rdi, %rdx, CHAR_SIZE), %rdx
|
||||
+ /* rsi stores s2 - s1. This allows loop to only update one
|
||||
+ pointer. */
|
||||
+ subq %rdi, %rsi
|
||||
+ /* Align s1 pointer. */
|
||||
+ andq $-VEC_SIZE, %rdi
|
||||
+ /* Adjust because first 4x vec where check already. */
|
||||
+ subq $-(VEC_SIZE * 4), %rdi
|
||||
+ .p2align 4
|
||||
+L(loop_4x_vec):
|
||||
+ VMOVU (%rsi, %rdi), %YMM1
|
||||
+ vpxorq (%rdi), %YMM1, %YMM1
|
||||
+
|
||||
+ VMOVU VEC_SIZE(%rsi, %rdi), %YMM2
|
||||
+ vpxorq VEC_SIZE(%rdi), %YMM2, %YMM2
|
||||
+
|
||||
+ VMOVU (VEC_SIZE * 2)(%rsi, %rdi), %YMM3
|
||||
+ vpxorq (VEC_SIZE * 2)(%rdi), %YMM3, %YMM3
|
||||
+ vpternlogd $0xfe, %YMM1, %YMM2, %YMM3
|
||||
+
|
||||
+ VMOVU (VEC_SIZE * 3)(%rsi, %rdi), %YMM4
|
||||
+ vpternlogd $0xde, (VEC_SIZE * 3)(%rdi), %YMM3, %YMM4
|
||||
+ VPCMP $4, %YMM4, %YMM0, %k1
|
||||
+ kmovd %k1, %ecx
|
||||
+ testl %ecx, %ecx
|
||||
+ jnz L(8x_return_vec_0_1_2_3)
|
||||
+ subq $-(VEC_SIZE * 4), %rdi
|
||||
+ cmpq %rdx, %rdi
|
||||
+ jb L(loop_4x_vec)
|
||||
+
|
||||
+ subq %rdx, %rdi
|
||||
+ /* rdi has 4 * VEC_SIZE - remaining length. */
|
||||
+ cmpl $(VEC_SIZE * 3), %edi
|
||||
+ jae L(8x_last_1x_vec)
|
||||
+ /* Load regardless of branch. */
|
||||
+ VMOVU (VEC_SIZE * 2)(%rsi, %rdx), %YMM3
|
||||
+ cmpl $(VEC_SIZE * 2), %edi
|
||||
+ jae L(8x_last_2x_vec)
|
||||
+
|
||||
+ VMOVU (%rsi, %rdx), %YMM1
|
||||
+ vpxorq (%rdx), %YMM1, %YMM1
|
||||
+
|
||||
+ VMOVU VEC_SIZE(%rsi, %rdx), %YMM2
|
||||
+ vpxorq VEC_SIZE(%rdx), %YMM2, %YMM2
|
||||
+
|
||||
+ vpxorq (VEC_SIZE * 2)(%rdx), %YMM3, %YMM3
|
||||
+ vpternlogd $0xfe, %YMM1, %YMM2, %YMM3
|
||||
+
|
||||
+ VMOVU (VEC_SIZE * 3)(%rsi, %rdx), %YMM4
|
||||
+ vpternlogd $0xde, (VEC_SIZE * 3)(%rdx), %YMM3, %YMM4
|
||||
+ VPCMP $4, %YMM4, %YMM0, %k1
|
||||
+ kmovd %k1, %ecx
|
||||
+ /* Restore s1 pointer to rdi. */
|
||||
+ movq %rdx, %rdi
|
||||
+ testl %ecx, %ecx
|
||||
+ jnz L(8x_return_vec_0_1_2_3)
|
||||
+ /* NB: eax must be zero to reach here. */
|
||||
+ ret
|
||||
+
|
||||
+ /* Only entry is from L(more_8x_vec). */
|
||||
+ .p2align 4
|
||||
+L(8x_last_2x_vec):
|
||||
+ VPCMP $4, (VEC_SIZE * 2)(%rdx), %YMM3, %k1
|
||||
+ kmovd %k1, %eax
|
||||
+ testl %eax, %eax
|
||||
+ jnz L(8x_return_vec_2)
|
||||
+ /* Naturally aligned to 16 bytes. */
|
||||
+L(8x_last_1x_vec):
|
||||
+ VMOVU (VEC_SIZE * 3)(%rsi, %rdx), %YMM1
|
||||
+ VPCMP $4, (VEC_SIZE * 3)(%rdx), %YMM1, %k1
|
||||
+ kmovd %k1, %eax
|
||||
+ testl %eax, %eax
|
||||
+ jnz L(8x_return_vec_3)
|
||||
+ ret
|
||||
+
|
||||
+ .p2align 4
|
||||
+L(last_2x_vec):
|
||||
+ /* Check second to last VEC. */
|
||||
+ VMOVU -(VEC_SIZE * 2)(%rsi, %rdx, CHAR_SIZE), %YMM1
|
||||
+ VPCMP $4, -(VEC_SIZE * 2)(%rdi, %rdx, CHAR_SIZE), %YMM1, %k1
|
||||
+ kmovd %k1, %eax
|
||||
+ testl %eax, %eax
|
||||
+ jnz L(return_vec_1_end)
|
||||
+
|
||||
+ /* Check last VEC. */
|
||||
+ .p2align 4
|
||||
+L(last_1x_vec):
|
||||
+ VMOVU -(VEC_SIZE * 1)(%rsi, %rdx, CHAR_SIZE), %YMM1
|
||||
+ VPCMP $4, -(VEC_SIZE * 1)(%rdi, %rdx, CHAR_SIZE), %YMM1, %k1
|
||||
+ kmovd %k1, %eax
|
||||
+ testl %eax, %eax
|
||||
+ jnz L(return_vec_0_end)
|
||||
ret
|
||||
+
|
||||
+ .p2align 4
|
||||
+L(8x_return_vec_2):
|
||||
+ subq $VEC_SIZE, %rdx
|
||||
+L(8x_return_vec_3):
|
||||
+ tzcntl %eax, %eax
|
||||
+# ifdef USE_AS_WMEMCMP
|
||||
+ leaq (%rdx, %rax, CHAR_SIZE), %rax
|
||||
+ movl (VEC_SIZE * 3)(%rax), %ecx
|
||||
+ xorl %edx, %edx
|
||||
+ cmpl (VEC_SIZE * 3)(%rsi, %rax), %ecx
|
||||
+ setg %dl
|
||||
+ leal -1(%rdx, %rdx), %eax
|
||||
# else
|
||||
+ addq %rdx, %rax
|
||||
+ movzbl (VEC_SIZE * 3)(%rsi, %rax), %ecx
|
||||
+ movzbl (VEC_SIZE * 3)(%rax), %eax
|
||||
+ subl %ecx, %eax
|
||||
+# endif
|
||||
+ ret
|
||||
+
|
||||
.p2align 4
|
||||
-L(between_4_7):
|
||||
- /* Load as big endian with overlapping movbe to avoid branches. */
|
||||
- movbe (%rdi), %eax
|
||||
- movbe (%rsi), %ecx
|
||||
- shlq $32, %rax
|
||||
- shlq $32, %rcx
|
||||
- movbe -4(%rdi, %rdx), %edi
|
||||
- movbe -4(%rsi, %rdx), %esi
|
||||
- orq %rdi, %rax
|
||||
- orq %rsi, %rcx
|
||||
- subq %rcx, %rax
|
||||
- je L(exit)
|
||||
- sbbl %eax, %eax
|
||||
- orl $1, %eax
|
||||
+L(return_vec_0_end):
|
||||
+ tzcntl %eax, %eax
|
||||
+ addl %edx, %eax
|
||||
+# ifdef USE_AS_WMEMCMP
|
||||
+ movl -VEC_SIZE(%rdi, %rax, CHAR_SIZE), %ecx
|
||||
+ xorl %edx, %edx
|
||||
+ cmpl -VEC_SIZE(%rsi, %rax, CHAR_SIZE), %ecx
|
||||
+ setg %dl
|
||||
+ leal -1(%rdx, %rdx), %eax
|
||||
+# else
|
||||
+ movzbl -VEC_SIZE(%rsi, %rax), %ecx
|
||||
+ movzbl -VEC_SIZE(%rdi, %rax), %eax
|
||||
+ subl %ecx, %eax
|
||||
+# endif
|
||||
ret
|
||||
|
||||
.p2align 4
|
||||
-L(exit):
|
||||
+L(return_vec_1_end):
|
||||
+ tzcntl %eax, %eax
|
||||
+ addl %edx, %eax
|
||||
+# ifdef USE_AS_WMEMCMP
|
||||
+ movl -(VEC_SIZE * 2)(%rdi, %rax, CHAR_SIZE), %ecx
|
||||
+ xorl %edx, %edx
|
||||
+ cmpl -(VEC_SIZE * 2)(%rsi, %rax, CHAR_SIZE), %ecx
|
||||
+ setg %dl
|
||||
+ leal -1(%rdx, %rdx), %eax
|
||||
+# else
|
||||
+ movzbl -(VEC_SIZE * 2)(%rsi, %rax), %ecx
|
||||
+ movzbl -(VEC_SIZE * 2)(%rdi, %rax), %eax
|
||||
+ subl %ecx, %eax
|
||||
+# endif
|
||||
ret
|
||||
|
||||
+
|
||||
.p2align 4
|
||||
+L(page_cross_less_vec):
|
||||
+ /* if USE_AS_WMEMCMP it can only be 0, 4, 8, 12, 16, 20, 24, 28
|
||||
+ bytes. */
|
||||
+ cmpl $(16 / CHAR_SIZE), %edx
|
||||
+ jae L(between_16_31)
|
||||
+# ifndef USE_AS_WMEMCMP
|
||||
+ cmpl $8, %edx
|
||||
+ jae L(between_8_15)
|
||||
+ cmpl $4, %edx
|
||||
+ jae L(between_4_7)
|
||||
L(between_2_3):
|
||||
/* Load as big endian to avoid branches. */
|
||||
movzwl (%rdi), %eax
|
||||
@@ -217,224 +448,99 @@ L(between_2_3):
|
||||
shll $8, %ecx
|
||||
bswap %eax
|
||||
bswap %ecx
|
||||
- movb -1(%rdi, %rdx), %al
|
||||
- movb -1(%rsi, %rdx), %cl
|
||||
+ movzbl -1(%rdi, %rdx), %edi
|
||||
+ movzbl -1(%rsi, %rdx), %esi
|
||||
+ orl %edi, %eax
|
||||
+ orl %esi, %ecx
|
||||
/* Subtraction is okay because the upper 8 bits are zero. */
|
||||
subl %ecx, %eax
|
||||
ret
|
||||
-
|
||||
.p2align 4
|
||||
-L(1):
|
||||
- movzbl (%rdi), %eax
|
||||
+L(one_or_less):
|
||||
+ jb L(zero)
|
||||
movzbl (%rsi), %ecx
|
||||
+ movzbl (%rdi), %eax
|
||||
subl %ecx, %eax
|
||||
ret
|
||||
-# endif
|
||||
-
|
||||
- .p2align 4
|
||||
-L(zero):
|
||||
- xorl %eax, %eax
|
||||
- ret
|
||||
|
||||
.p2align 4
|
||||
-L(less_vec):
|
||||
-# ifdef USE_AS_WMEMCMP
|
||||
- /* It can only be 0, 4, 8, 12, 16, 20, 24, 28 bytes. */
|
||||
- cmpb $4, %dl
|
||||
- je L(4)
|
||||
- jb L(zero)
|
||||
-# else
|
||||
- cmpb $1, %dl
|
||||
- je L(1)
|
||||
- jb L(zero)
|
||||
- cmpb $4, %dl
|
||||
- jb L(between_2_3)
|
||||
- cmpb $8, %dl
|
||||
- jb L(between_4_7)
|
||||
+L(between_8_15):
|
||||
# endif
|
||||
- cmpb $16, %dl
|
||||
- jae L(between_16_31)
|
||||
- /* It is between 8 and 15 bytes. */
|
||||
+ /* If USE_AS_WMEMCMP fall through into 8-15 byte case. */
|
||||
vmovq (%rdi), %XMM1
|
||||
vmovq (%rsi), %XMM2
|
||||
- VPCMPEQ %XMM1, %XMM2, %k2
|
||||
- kmovw %k2, %eax
|
||||
- subl $XMM_MASK, %eax
|
||||
- jnz L(first_vec)
|
||||
+ VPCMP $4, %XMM1, %XMM2, %k1
|
||||
+ kmovd %k1, %eax
|
||||
+ testl %eax, %eax
|
||||
+ jnz L(return_vec_0)
|
||||
/* Use overlapping loads to avoid branches. */
|
||||
- leaq -8(%rdi, %rdx), %rdi
|
||||
- leaq -8(%rsi, %rdx), %rsi
|
||||
+ leaq -8(%rdi, %rdx, CHAR_SIZE), %rdi
|
||||
+ leaq -8(%rsi, %rdx, CHAR_SIZE), %rsi
|
||||
vmovq (%rdi), %XMM1
|
||||
vmovq (%rsi), %XMM2
|
||||
- VPCMPEQ %XMM1, %XMM2, %k2
|
||||
- kmovw %k2, %eax
|
||||
- subl $XMM_MASK, %eax
|
||||
- jnz L(first_vec)
|
||||
+ VPCMP $4, %XMM1, %XMM2, %k1
|
||||
+ kmovd %k1, %eax
|
||||
+ testl %eax, %eax
|
||||
+ jnz L(return_vec_0)
|
||||
ret
|
||||
|
||||
.p2align 4
|
||||
-L(between_16_31):
|
||||
- /* From 16 to 31 bytes. No branch when size == 16. */
|
||||
- VMOVU (%rsi), %XMM2
|
||||
- VPCMPEQ (%rdi), %XMM2, %k2
|
||||
- kmovw %k2, %eax
|
||||
- subl $XMM_MASK, %eax
|
||||
- jnz L(first_vec)
|
||||
-
|
||||
- /* Use overlapping loads to avoid branches. */
|
||||
- leaq -16(%rdi, %rdx), %rdi
|
||||
- leaq -16(%rsi, %rdx), %rsi
|
||||
- VMOVU (%rsi), %XMM2
|
||||
- VPCMPEQ (%rdi), %XMM2, %k2
|
||||
- kmovw %k2, %eax
|
||||
- subl $XMM_MASK, %eax
|
||||
- jnz L(first_vec)
|
||||
+L(zero):
|
||||
+ xorl %eax, %eax
|
||||
ret
|
||||
|
||||
.p2align 4
|
||||
-L(more_8x_vec):
|
||||
- /* More than 8 * VEC. Check the first VEC. */
|
||||
- VMOVU (%rsi), %YMM2
|
||||
- VPCMPEQ (%rdi), %YMM2, %k2
|
||||
- kmovd %k2, %eax
|
||||
- subl $VEC_MASK, %eax
|
||||
- jnz L(first_vec)
|
||||
-
|
||||
- /* Align the first memory area for aligned loads in the loop.
|
||||
- Compute how much the first memory area is misaligned. */
|
||||
- movq %rdi, %rcx
|
||||
- andl $(VEC_SIZE - 1), %ecx
|
||||
- /* Get the negative of offset for alignment. */
|
||||
- subq $VEC_SIZE, %rcx
|
||||
- /* Adjust the second memory area. */
|
||||
- subq %rcx, %rsi
|
||||
- /* Adjust the first memory area which should be aligned now. */
|
||||
- subq %rcx, %rdi
|
||||
- /* Adjust length. */
|
||||
- addq %rcx, %rdx
|
||||
-
|
||||
-L(loop_4x_vec):
|
||||
- /* Compare 4 * VEC at a time forward. */
|
||||
- VMOVU (%rsi), %YMM1
|
||||
- VPCMPEQ (%rdi), %YMM1, %k1
|
||||
-
|
||||
- VMOVU VEC_SIZE(%rsi), %YMM2
|
||||
- VPCMPEQ VEC_SIZE(%rdi), %YMM2, %k2
|
||||
- kandd %k2, %k1, %k5
|
||||
-
|
||||
- VMOVU (VEC_SIZE * 2)(%rsi), %YMM3
|
||||
- VPCMPEQ (VEC_SIZE * 2)(%rdi), %YMM3, %k3
|
||||
- kandd %k3, %k5, %k5
|
||||
-
|
||||
- VMOVU (VEC_SIZE * 3)(%rsi), %YMM4
|
||||
- VPCMPEQ (VEC_SIZE * 3)(%rdi), %YMM4, %k4
|
||||
- kandd %k4, %k5, %k5
|
||||
-
|
||||
- kmovd %k5, %eax
|
||||
- cmpl $VEC_MASK, %eax
|
||||
- jne L(4x_vec_end)
|
||||
-
|
||||
- addq $(VEC_SIZE * 4), %rdi
|
||||
- addq $(VEC_SIZE * 4), %rsi
|
||||
-
|
||||
- subq $(VEC_SIZE * 4), %rdx
|
||||
- cmpq $(VEC_SIZE * 4), %rdx
|
||||
- jae L(loop_4x_vec)
|
||||
-
|
||||
- /* Less than 4 * VEC. */
|
||||
- cmpq $VEC_SIZE, %rdx
|
||||
- jbe L(last_vec)
|
||||
- cmpq $(VEC_SIZE * 2), %rdx
|
||||
- jbe L(last_2x_vec)
|
||||
-
|
||||
-L(last_4x_vec):
|
||||
- /* From 2 * VEC to 4 * VEC. */
|
||||
- VMOVU (%rsi), %YMM2
|
||||
- VPCMPEQ (%rdi), %YMM2, %k2
|
||||
- kmovd %k2, %eax
|
||||
- subl $VEC_MASK, %eax
|
||||
- jnz L(first_vec)
|
||||
-
|
||||
- addq $VEC_SIZE, %rdi
|
||||
- addq $VEC_SIZE, %rsi
|
||||
- VMOVU (%rsi), %YMM2
|
||||
- VPCMPEQ (%rdi), %YMM2, %k2
|
||||
- kmovd %k2, %eax
|
||||
- subl $VEC_MASK, %eax
|
||||
- jnz L(first_vec)
|
||||
+L(between_16_31):
|
||||
+ /* From 16 to 31 bytes. No branch when size == 16. */
|
||||
+ VMOVU (%rsi), %XMM2
|
||||
+ VPCMP $4, (%rdi), %XMM2, %k1
|
||||
+ kmovd %k1, %eax
|
||||
+ testl %eax, %eax
|
||||
+ jnz L(return_vec_0)
|
||||
|
||||
/* Use overlapping loads to avoid branches. */
|
||||
- leaq -(3 * VEC_SIZE)(%rdi, %rdx), %rdi
|
||||
- leaq -(3 * VEC_SIZE)(%rsi, %rdx), %rsi
|
||||
- VMOVU (%rsi), %YMM2
|
||||
- VPCMPEQ (%rdi), %YMM2, %k2
|
||||
- kmovd %k2, %eax
|
||||
- subl $VEC_MASK, %eax
|
||||
- jnz L(first_vec)
|
||||
|
||||
- addq $VEC_SIZE, %rdi
|
||||
- addq $VEC_SIZE, %rsi
|
||||
- VMOVU (%rsi), %YMM2
|
||||
- VPCMPEQ (%rdi), %YMM2, %k2
|
||||
- kmovd %k2, %eax
|
||||
- subl $VEC_MASK, %eax
|
||||
- jnz L(first_vec)
|
||||
- ret
|
||||
-
|
||||
- .p2align 4
|
||||
-L(4x_vec_end):
|
||||
+ VMOVU -16(%rsi, %rdx, CHAR_SIZE), %XMM2
|
||||
+ leaq -16(%rdi, %rdx, CHAR_SIZE), %rdi
|
||||
+ leaq -16(%rsi, %rdx, CHAR_SIZE), %rsi
|
||||
+ VPCMP $4, (%rdi), %XMM2, %k1
|
||||
kmovd %k1, %eax
|
||||
- subl $VEC_MASK, %eax
|
||||
- jnz L(first_vec)
|
||||
- kmovd %k2, %eax
|
||||
- subl $VEC_MASK, %eax
|
||||
- jnz L(first_vec_x1)
|
||||
- kmovd %k3, %eax
|
||||
- subl $VEC_MASK, %eax
|
||||
- jnz L(first_vec_x2)
|
||||
- kmovd %k4, %eax
|
||||
- subl $VEC_MASK, %eax
|
||||
- tzcntl %eax, %ecx
|
||||
-# ifdef USE_AS_WMEMCMP
|
||||
- xorl %eax, %eax
|
||||
- movl (VEC_SIZE * 3)(%rdi, %rcx, 4), %edx
|
||||
- cmpl (VEC_SIZE * 3)(%rsi, %rcx, 4), %edx
|
||||
- jmp L(wmemcmp_return)
|
||||
-# else
|
||||
- movzbl (VEC_SIZE * 3)(%rdi, %rcx), %eax
|
||||
- movzbl (VEC_SIZE * 3)(%rsi, %rcx), %edx
|
||||
- sub %edx, %eax
|
||||
-# endif
|
||||
+ testl %eax, %eax
|
||||
+ jnz L(return_vec_0)
|
||||
ret
|
||||
|
||||
- .p2align 4
|
||||
-L(first_vec_x1):
|
||||
- tzcntl %eax, %ecx
|
||||
# ifdef USE_AS_WMEMCMP
|
||||
- xorl %eax, %eax
|
||||
- movl VEC_SIZE(%rdi, %rcx, 4), %edx
|
||||
- cmpl VEC_SIZE(%rsi, %rcx, 4), %edx
|
||||
- jmp L(wmemcmp_return)
|
||||
-# else
|
||||
- movzbl VEC_SIZE(%rdi, %rcx), %eax
|
||||
- movzbl VEC_SIZE(%rsi, %rcx), %edx
|
||||
- sub %edx, %eax
|
||||
-# endif
|
||||
+ .p2align 4
|
||||
+L(one_or_less):
|
||||
+ jb L(zero)
|
||||
+ movl (%rdi), %ecx
|
||||
+ xorl %edx, %edx
|
||||
+ cmpl (%rsi), %ecx
|
||||
+ je L(zero)
|
||||
+ setg %dl
|
||||
+ leal -1(%rdx, %rdx), %eax
|
||||
ret
|
||||
+# else
|
||||
|
||||
.p2align 4
|
||||
-L(first_vec_x2):
|
||||
- tzcntl %eax, %ecx
|
||||
-# ifdef USE_AS_WMEMCMP
|
||||
- xorl %eax, %eax
|
||||
- movl (VEC_SIZE * 2)(%rdi, %rcx, 4), %edx
|
||||
- cmpl (VEC_SIZE * 2)(%rsi, %rcx, 4), %edx
|
||||
- jmp L(wmemcmp_return)
|
||||
-# else
|
||||
- movzbl (VEC_SIZE * 2)(%rdi, %rcx), %eax
|
||||
- movzbl (VEC_SIZE * 2)(%rsi, %rcx), %edx
|
||||
- sub %edx, %eax
|
||||
-# endif
|
||||
+L(between_4_7):
|
||||
+ /* Load as big endian with overlapping movbe to avoid branches.
|
||||
+ */
|
||||
+ movbe (%rdi), %eax
|
||||
+ movbe (%rsi), %ecx
|
||||
+ shlq $32, %rax
|
||||
+ shlq $32, %rcx
|
||||
+ movbe -4(%rdi, %rdx), %edi
|
||||
+ movbe -4(%rsi, %rdx), %esi
|
||||
+ orq %rdi, %rax
|
||||
+ orq %rsi, %rcx
|
||||
+ subq %rcx, %rax
|
||||
+ jz L(zero_4_7)
|
||||
+ sbbl %eax, %eax
|
||||
+ orl $1, %eax
|
||||
+L(zero_4_7):
|
||||
ret
|
||||
+# endif
|
||||
+
|
||||
END (MEMCMP)
|
||||
#endif
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,104 +0,0 @@
|
||||
From 6abf27980a947f9b6e514d6b33b83059d39566ae Mon Sep 17 00:00:00 2001
|
||||
From: Noah Goldstein <goldstein.w.n@gmail.com>
|
||||
Date: Thu, 20 May 2021 13:13:51 -0400
|
||||
Subject: [PATCH] x86: Improve memset-vec-unaligned-erms.S
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
No bug. This commit makes a few small improvements to
|
||||
memset-vec-unaligned-erms.S. The changes are 1) only aligning to 64
|
||||
instead of 128. Either alignment will perform equally well in a loop
|
||||
and 128 just increases the odds of having to do an extra iteration
|
||||
which can be significant overhead for small values. 2) Align some
|
||||
targets and the loop. 3) Remove an ALU from the alignment process. 4)
|
||||
Reorder the last 4x VEC so that they are stored after the loop. 5)
|
||||
Move the condition for leq 8x VEC to before the alignment
|
||||
process. test-memset and test-wmemset are both passing.
|
||||
|
||||
Signed-off-by: Noah Goldstein <goldstein.w.n@gmail.com>
|
||||
Reviewed-by: H.J. Lu <hjl.tools@gmail.com>
|
||||
---
|
||||
.../multiarch/memset-vec-unaligned-erms.S | 50 +++++++++++--------
|
||||
1 file changed, 28 insertions(+), 22 deletions(-)
|
||||
|
||||
diff --git a/sysdeps/x86_64/multiarch/memset-vec-unaligned-erms.S b/sysdeps/x86_64/multiarch/memset-vec-unaligned-erms.S
|
||||
index f877ac9d..909c33f6 100644
|
||||
--- a/sysdeps/x86_64/multiarch/memset-vec-unaligned-erms.S
|
||||
+++ b/sysdeps/x86_64/multiarch/memset-vec-unaligned-erms.S
|
||||
@@ -173,17 +173,22 @@ ENTRY (MEMSET_SYMBOL (__memset, unaligned_erms))
|
||||
VMOVU %VEC(0), (%rdi)
|
||||
VZEROUPPER_RETURN
|
||||
|
||||
+ .p2align 4
|
||||
L(stosb_more_2x_vec):
|
||||
cmp __x86_rep_stosb_threshold(%rip), %RDX_LP
|
||||
ja L(stosb)
|
||||
+#else
|
||||
+ .p2align 4
|
||||
#endif
|
||||
L(more_2x_vec):
|
||||
- cmpq $(VEC_SIZE * 4), %rdx
|
||||
- ja L(loop_start)
|
||||
+ /* Stores to first 2x VEC before cmp as any path forward will
|
||||
+ require it. */
|
||||
VMOVU %VEC(0), (%rdi)
|
||||
VMOVU %VEC(0), VEC_SIZE(%rdi)
|
||||
- VMOVU %VEC(0), -VEC_SIZE(%rdi,%rdx)
|
||||
+ cmpq $(VEC_SIZE * 4), %rdx
|
||||
+ ja L(loop_start)
|
||||
VMOVU %VEC(0), -(VEC_SIZE * 2)(%rdi,%rdx)
|
||||
+ VMOVU %VEC(0), -VEC_SIZE(%rdi,%rdx)
|
||||
L(return):
|
||||
#if VEC_SIZE > 16
|
||||
ZERO_UPPER_VEC_REGISTERS_RETURN
|
||||
@@ -192,28 +197,29 @@ L(return):
|
||||
#endif
|
||||
|
||||
L(loop_start):
|
||||
- leaq (VEC_SIZE * 4)(%rdi), %rcx
|
||||
- VMOVU %VEC(0), (%rdi)
|
||||
- andq $-(VEC_SIZE * 4), %rcx
|
||||
- VMOVU %VEC(0), -VEC_SIZE(%rdi,%rdx)
|
||||
- VMOVU %VEC(0), VEC_SIZE(%rdi)
|
||||
- VMOVU %VEC(0), -(VEC_SIZE * 2)(%rdi,%rdx)
|
||||
VMOVU %VEC(0), (VEC_SIZE * 2)(%rdi)
|
||||
- VMOVU %VEC(0), -(VEC_SIZE * 3)(%rdi,%rdx)
|
||||
VMOVU %VEC(0), (VEC_SIZE * 3)(%rdi)
|
||||
- VMOVU %VEC(0), -(VEC_SIZE * 4)(%rdi,%rdx)
|
||||
- addq %rdi, %rdx
|
||||
- andq $-(VEC_SIZE * 4), %rdx
|
||||
- cmpq %rdx, %rcx
|
||||
- je L(return)
|
||||
+ cmpq $(VEC_SIZE * 8), %rdx
|
||||
+ jbe L(loop_end)
|
||||
+ andq $-(VEC_SIZE * 2), %rdi
|
||||
+ subq $-(VEC_SIZE * 4), %rdi
|
||||
+ leaq -(VEC_SIZE * 4)(%rax, %rdx), %rcx
|
||||
+ .p2align 4
|
||||
L(loop):
|
||||
- VMOVA %VEC(0), (%rcx)
|
||||
- VMOVA %VEC(0), VEC_SIZE(%rcx)
|
||||
- VMOVA %VEC(0), (VEC_SIZE * 2)(%rcx)
|
||||
- VMOVA %VEC(0), (VEC_SIZE * 3)(%rcx)
|
||||
- addq $(VEC_SIZE * 4), %rcx
|
||||
- cmpq %rcx, %rdx
|
||||
- jne L(loop)
|
||||
+ VMOVA %VEC(0), (%rdi)
|
||||
+ VMOVA %VEC(0), VEC_SIZE(%rdi)
|
||||
+ VMOVA %VEC(0), (VEC_SIZE * 2)(%rdi)
|
||||
+ VMOVA %VEC(0), (VEC_SIZE * 3)(%rdi)
|
||||
+ subq $-(VEC_SIZE * 4), %rdi
|
||||
+ cmpq %rcx, %rdi
|
||||
+ jb L(loop)
|
||||
+L(loop_end):
|
||||
+ /* NB: rax is set as ptr in MEMSET_VDUP_TO_VEC0_AND_SET_RETURN.
|
||||
+ rdx as length is also unchanged. */
|
||||
+ VMOVU %VEC(0), -(VEC_SIZE * 4)(%rax, %rdx)
|
||||
+ VMOVU %VEC(0), -(VEC_SIZE * 3)(%rax, %rdx)
|
||||
+ VMOVU %VEC(0), -(VEC_SIZE * 2)(%rax, %rdx)
|
||||
+ VMOVU %VEC(0), -VEC_SIZE(%rax, %rdx)
|
||||
VZEROUPPER_SHORT_RETURN
|
||||
|
||||
.p2align 4
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,84 +0,0 @@
|
||||
From 1b992204f68af851e905c16016756fd4421e1934 Mon Sep 17 00:00:00 2001
|
||||
From: Noah Goldstein <goldstein.w.n@gmail.com>
|
||||
Date: Sun, 23 May 2021 19:43:24 -0400
|
||||
Subject: [PATCH] x86: Improve memmove-vec-unaligned-erms.S
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
This patch changes the condition for copy 4x VEC so that if length is
|
||||
exactly equal to 4 * VEC_SIZE it will use the 4x VEC case instead of
|
||||
8x VEC case.
|
||||
|
||||
Results For Skylake memcpy-avx2-erms
|
||||
size, al1 , al2 , Cur T , New T , Win , New / Cur
|
||||
128 , 0 , 0 , 9.137 , 6.873 , New , 75.22
|
||||
128 , 7 , 0 , 12.933 , 7.732 , New , 59.79
|
||||
128 , 0 , 7 , 11.852 , 6.76 , New , 57.04
|
||||
128 , 7 , 7 , 12.587 , 6.808 , New , 54.09
|
||||
|
||||
Results For Icelake memcpy-evex-erms
|
||||
size, al1 , al2 , Cur T , New T , Win , New / Cur
|
||||
128 , 0 , 0 , 9.963 , 5.416 , New , 54.36
|
||||
128 , 7 , 0 , 16.467 , 8.061 , New , 48.95
|
||||
128 , 0 , 7 , 14.388 , 7.644 , New , 53.13
|
||||
128 , 7 , 7 , 14.546 , 7.642 , New , 52.54
|
||||
|
||||
Results For Tigerlake memcpy-evex-erms
|
||||
size, al1 , al2 , Cur T , New T , Win , New / Cur
|
||||
128 , 0 , 0 , 8.979 , 4.95 , New , 55.13
|
||||
128 , 7 , 0 , 14.245 , 7.122 , New , 50.0
|
||||
128 , 0 , 7 , 12.668 , 6.675 , New , 52.69
|
||||
128 , 7 , 7 , 13.042 , 6.802 , New , 52.15
|
||||
|
||||
Results For Skylake memmove-avx2-erms
|
||||
size, al1 , al2 , Cur T , New T , Win , New / Cur
|
||||
128 , 0 , 32 , 6.181 , 5.691 , New , 92.07
|
||||
128 , 32 , 0 , 6.165 , 5.752 , New , 93.3
|
||||
128 , 0 , 7 , 13.923 , 9.37 , New , 67.3
|
||||
128 , 7 , 0 , 12.049 , 10.182 , New , 84.5
|
||||
|
||||
Results For Icelake memmove-evex-erms
|
||||
size, al1 , al2 , Cur T , New T , Win , New / Cur
|
||||
128 , 0 , 32 , 5.479 , 4.889 , New , 89.23
|
||||
128 , 32 , 0 , 5.127 , 4.911 , New , 95.79
|
||||
128 , 0 , 7 , 18.885 , 13.547 , New , 71.73
|
||||
128 , 7 , 0 , 15.565 , 14.436 , New , 92.75
|
||||
|
||||
Results For Tigerlake memmove-evex-erms
|
||||
size, al1 , al2 , Cur T , New T , Win , New / Cur
|
||||
128 , 0 , 32 , 5.275 , 4.815 , New , 91.28
|
||||
128 , 32 , 0 , 5.376 , 4.565 , New , 84.91
|
||||
128 , 0 , 7 , 19.426 , 14.273 , New , 73.47
|
||||
128 , 7 , 0 , 15.924 , 14.951 , New , 93.89
|
||||
|
||||
Signed-off-by: Noah Goldstein <goldstein.w.n@gmail.com>
|
||||
---
|
||||
sysdeps/x86_64/multiarch/memmove-vec-unaligned-erms.S | 6 +++---
|
||||
1 file changed, 3 insertions(+), 3 deletions(-)
|
||||
|
||||
diff --git a/sysdeps/x86_64/multiarch/memmove-vec-unaligned-erms.S b/sysdeps/x86_64/multiarch/memmove-vec-unaligned-erms.S
|
||||
index 3e2dd6bc..572cef04 100644
|
||||
--- a/sysdeps/x86_64/multiarch/memmove-vec-unaligned-erms.S
|
||||
+++ b/sysdeps/x86_64/multiarch/memmove-vec-unaligned-erms.S
|
||||
@@ -417,8 +417,8 @@ L(more_2x_vec):
|
||||
cmpq $(VEC_SIZE * 8), %rdx
|
||||
ja L(more_8x_vec)
|
||||
cmpq $(VEC_SIZE * 4), %rdx
|
||||
- jb L(last_4x_vec)
|
||||
- /* Copy from 4 * VEC to 8 * VEC, inclusively. */
|
||||
+ jbe L(last_4x_vec)
|
||||
+ /* Copy from 4 * VEC + 1 to 8 * VEC, inclusively. */
|
||||
VMOVU (%rsi), %VEC(0)
|
||||
VMOVU VEC_SIZE(%rsi), %VEC(1)
|
||||
VMOVU (VEC_SIZE * 2)(%rsi), %VEC(2)
|
||||
@@ -437,7 +437,7 @@ L(more_2x_vec):
|
||||
VMOVU %VEC(7), -(VEC_SIZE * 4)(%rdi,%rdx)
|
||||
VZEROUPPER_RETURN
|
||||
L(last_4x_vec):
|
||||
- /* Copy from 2 * VEC to 4 * VEC. */
|
||||
+ /* Copy from 2 * VEC + 1 to 4 * VEC, inclusively. */
|
||||
VMOVU (%rsi), %VEC(0)
|
||||
VMOVU VEC_SIZE(%rsi), %VEC(1)
|
||||
VMOVU -VEC_SIZE(%rsi,%rdx), %VEC(2)
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,55 +0,0 @@
|
||||
From 08cbcd4dbc686bb38ec3093aff2f919fbff5ec17 Mon Sep 17 00:00:00 2001
|
||||
From: Noah Goldstein <goldstein.w.n@gmail.com>
|
||||
Date: Wed, 23 Jun 2021 19:19:34 -0400
|
||||
Subject: [PATCH] x86: Remove unnecessary overflow check from wcsnlen-sse4_1.S
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
No bug. The way wcsnlen will check if near the end of maxlen
|
||||
is the following macro:
|
||||
|
||||
mov %r11, %rsi; \
|
||||
subq %rax, %rsi; \
|
||||
andq $-64, %rax; \
|
||||
testq $-64, %rsi; \
|
||||
je L(strnlen_ret)
|
||||
|
||||
Which words independently of s + maxlen overflowing. So the
|
||||
second overflow check is unnecissary for correctness and
|
||||
just extra overhead in the common no overflow case.
|
||||
|
||||
test-strlen.c, test-wcslen.c, test-strnlen.c and test-wcsnlen.c are
|
||||
all passing
|
||||
|
||||
Signed-off-by: Noah Goldstein <goldstein.w.n@gmail.com>
|
||||
Reviewed-by: H.J. Lu <hjl.tools@gmail.com>
|
||||
---
|
||||
sysdeps/x86_64/multiarch/strlen-vec.S | 7 -------
|
||||
1 file changed, 7 deletions(-)
|
||||
|
||||
diff --git a/sysdeps/x86_64/multiarch/strlen-vec.S b/sysdeps/x86_64/multiarch/strlen-vec.S
|
||||
index 439e486a..b7657282 100644
|
||||
--- a/sysdeps/x86_64/multiarch/strlen-vec.S
|
||||
+++ b/sysdeps/x86_64/multiarch/strlen-vec.S
|
||||
@@ -71,19 +71,12 @@ L(n_nonzero):
|
||||
suffice. */
|
||||
mov %RSI_LP, %R10_LP
|
||||
sar $62, %R10_LP
|
||||
- test %R10_LP, %R10_LP
|
||||
jnz __wcslen_sse4_1
|
||||
sal $2, %RSI_LP
|
||||
# endif
|
||||
|
||||
-
|
||||
/* Initialize long lived registers. */
|
||||
-
|
||||
add %RDI_LP, %RSI_LP
|
||||
-# ifdef AS_WCSLEN
|
||||
-/* Check for overflow again from s + maxlen * sizeof(wchar_t). */
|
||||
- jbe __wcslen_sse4_1
|
||||
-# endif
|
||||
mov %RSI_LP, %R10_LP
|
||||
and $-64, %R10_LP
|
||||
mov %RSI_LP, %R11_LP
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,290 +0,0 @@
|
||||
From 82d0b4a4d76db554eb6757acb790fcea30b19965 Mon Sep 17 00:00:00 2001
|
||||
From: "H.J. Lu" <hjl.tools@gmail.com>
|
||||
Date: Mon, 21 Jan 2019 11:32:24 -0800
|
||||
Subject: [PATCH] x86-64 memset/wmemset: Properly handle the length parameter
|
||||
[BZ# 24097]
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
On x32, the size_t parameter may be passed in the lower 32 bits of a
|
||||
64-bit register with the non-zero upper 32 bits. The string/memory
|
||||
functions written in assembly can only use the lower 32 bits of a
|
||||
64-bit register as length or must clear the upper 32 bits before using
|
||||
the full 64-bit register for length.
|
||||
|
||||
This pach fixes memset/wmemset for x32. Tested on x86-64 and x32. On
|
||||
x86-64, libc.so is the same with and withou the fix.
|
||||
|
||||
[BZ# 24097]
|
||||
CVE-2019-6488
|
||||
* sysdeps/x86_64/multiarch/memset-avx512-no-vzeroupper.S: Use
|
||||
RDX_LP for length. Clear the upper 32 bits of RDX register.
|
||||
* sysdeps/x86_64/multiarch/memset-vec-unaligned-erms.S: Likewise.
|
||||
* sysdeps/x86_64/x32/Makefile (tests): Add tst-size_t-wmemset.
|
||||
* sysdeps/x86_64/x32/tst-size_t-memset.c: New file.
|
||||
* sysdeps/x86_64/x32/tst-size_t-wmemset.c: Likewise.
|
||||
---
|
||||
.../multiarch/memset-avx512-no-vzeroupper.S | 6 +-
|
||||
.../multiarch/memset-vec-unaligned-erms.S | 34 +++++----
|
||||
sysdeps/x86_64/x32/Makefile | 4 +-
|
||||
sysdeps/x86_64/x32/tst-size_t-memset.c | 73 +++++++++++++++++++
|
||||
sysdeps/x86_64/x32/tst-size_t-wmemset.c | 20 +++++
|
||||
5 files changed, 121 insertions(+), 16 deletions(-)
|
||||
create mode 100644 sysdeps/x86_64/x32/tst-size_t-memset.c
|
||||
create mode 100644 sysdeps/x86_64/x32/tst-size_t-wmemset.c
|
||||
|
||||
Conflicts:
|
||||
ChangeLog
|
||||
(removed)
|
||||
|
||||
diff --git a/sysdeps/x86_64/multiarch/memset-avx512-no-vzeroupper.S b/sysdeps/x86_64/multiarch/memset-avx512-no-vzeroupper.S
|
||||
index 689cc119..99e25519 100644
|
||||
--- a/sysdeps/x86_64/multiarch/memset-avx512-no-vzeroupper.S
|
||||
+++ b/sysdeps/x86_64/multiarch/memset-avx512-no-vzeroupper.S
|
||||
@@ -29,12 +29,16 @@
|
||||
.section .text.avx512,"ax",@progbits
|
||||
#if defined PIC
|
||||
ENTRY (MEMSET_CHK)
|
||||
- cmpq %rdx, %rcx
|
||||
+ cmp %RDX_LP, %RCX_LP
|
||||
jb HIDDEN_JUMPTARGET (__chk_fail)
|
||||
END (MEMSET_CHK)
|
||||
#endif
|
||||
|
||||
ENTRY (MEMSET)
|
||||
+# ifdef __ILP32__
|
||||
+ /* Clear the upper 32 bits. */
|
||||
+ mov %edx, %edx
|
||||
+# endif
|
||||
vpxor %xmm0, %xmm0, %xmm0
|
||||
vmovd %esi, %xmm1
|
||||
lea (%rdi, %rdx), %rsi
|
||||
diff --git a/sysdeps/x86_64/multiarch/memset-vec-unaligned-erms.S b/sysdeps/x86_64/multiarch/memset-vec-unaligned-erms.S
|
||||
index 270a1d49..9a0fd818 100644
|
||||
--- a/sysdeps/x86_64/multiarch/memset-vec-unaligned-erms.S
|
||||
+++ b/sysdeps/x86_64/multiarch/memset-vec-unaligned-erms.S
|
||||
@@ -65,8 +65,8 @@
|
||||
.section SECTION(.text),"ax",@progbits
|
||||
#if VEC_SIZE == 16 && IS_IN (libc)
|
||||
ENTRY (__bzero)
|
||||
- movq %rdi, %rax /* Set return value. */
|
||||
- movq %rsi, %rdx /* Set n. */
|
||||
+ mov %RDI_LP, %RAX_LP /* Set return value. */
|
||||
+ mov %RSI_LP, %RDX_LP /* Set n. */
|
||||
pxor %xmm0, %xmm0
|
||||
jmp L(entry_from_bzero)
|
||||
END (__bzero)
|
||||
@@ -76,13 +76,13 @@ weak_alias (__bzero, bzero)
|
||||
#if IS_IN (libc)
|
||||
# if defined SHARED
|
||||
ENTRY_CHK (WMEMSET_CHK_SYMBOL (__wmemset_chk, unaligned))
|
||||
- cmpq %rdx, %rcx
|
||||
+ cmp %RDX_LP, %RCX_LP
|
||||
jb HIDDEN_JUMPTARGET (__chk_fail)
|
||||
END_CHK (WMEMSET_CHK_SYMBOL (__wmemset_chk, unaligned))
|
||||
# endif
|
||||
|
||||
ENTRY (WMEMSET_SYMBOL (__wmemset, unaligned))
|
||||
- shlq $2, %rdx
|
||||
+ shl $2, %RDX_LP
|
||||
WMEMSET_VDUP_TO_VEC0_AND_SET_RETURN (%esi, %rdi)
|
||||
jmp L(entry_from_bzero)
|
||||
END (WMEMSET_SYMBOL (__wmemset, unaligned))
|
||||
@@ -90,13 +90,17 @@ END (WMEMSET_SYMBOL (__wmemset, unaligned))
|
||||
|
||||
#if defined SHARED && IS_IN (libc)
|
||||
ENTRY_CHK (MEMSET_CHK_SYMBOL (__memset_chk, unaligned))
|
||||
- cmpq %rdx, %rcx
|
||||
+ cmp %RDX_LP, %RCX_LP
|
||||
jb HIDDEN_JUMPTARGET (__chk_fail)
|
||||
END_CHK (MEMSET_CHK_SYMBOL (__memset_chk, unaligned))
|
||||
#endif
|
||||
|
||||
ENTRY (MEMSET_SYMBOL (__memset, unaligned))
|
||||
MEMSET_VDUP_TO_VEC0_AND_SET_RETURN (%esi, %rdi)
|
||||
+# ifdef __ILP32__
|
||||
+ /* Clear the upper 32 bits. */
|
||||
+ mov %edx, %edx
|
||||
+# endif
|
||||
L(entry_from_bzero):
|
||||
cmpq $VEC_SIZE, %rdx
|
||||
jb L(less_vec)
|
||||
@@ -112,14 +116,14 @@ END (MEMSET_SYMBOL (__memset, unaligned))
|
||||
|
||||
# if VEC_SIZE == 16
|
||||
ENTRY (__memset_chk_erms)
|
||||
- cmpq %rdx, %rcx
|
||||
+ cmp %RDX_LP, %RCX_LP
|
||||
jb HIDDEN_JUMPTARGET (__chk_fail)
|
||||
END (__memset_chk_erms)
|
||||
|
||||
/* Only used to measure performance of REP STOSB. */
|
||||
ENTRY (__memset_erms)
|
||||
/* Skip zero length. */
|
||||
- testq %rdx, %rdx
|
||||
+ test %RDX_LP, %RDX_LP
|
||||
jnz L(stosb)
|
||||
movq %rdi, %rax
|
||||
ret
|
||||
@@ -131,11 +135,11 @@ ENTRY (MEMSET_SYMBOL (__memset, erms))
|
||||
L(stosb):
|
||||
/* Issue vzeroupper before rep stosb. */
|
||||
VZEROUPPER
|
||||
- movq %rdx, %rcx
|
||||
+ mov %RDX_LP, %RCX_LP
|
||||
movzbl %sil, %eax
|
||||
- movq %rdi, %rdx
|
||||
+ mov %RDI_LP, %RDX_LP
|
||||
rep stosb
|
||||
- movq %rdx, %rax
|
||||
+ mov %RDX_LP, %RAX_LP
|
||||
ret
|
||||
# if VEC_SIZE == 16
|
||||
END (__memset_erms)
|
||||
@@ -145,16 +149,20 @@ END (MEMSET_SYMBOL (__memset, erms))
|
||||
|
||||
# if defined SHARED && IS_IN (libc)
|
||||
ENTRY_CHK (MEMSET_CHK_SYMBOL (__memset_chk, unaligned_erms))
|
||||
- cmpq %rdx, %rcx
|
||||
+ cmp %RDX_LP, %RCX_LP
|
||||
jb HIDDEN_JUMPTARGET (__chk_fail)
|
||||
END_CHK (MEMSET_CHK_SYMBOL (__memset_chk, unaligned_erms))
|
||||
# endif
|
||||
|
||||
ENTRY (MEMSET_SYMBOL (__memset, unaligned_erms))
|
||||
MEMSET_VDUP_TO_VEC0_AND_SET_RETURN (%esi, %rdi)
|
||||
- cmpq $VEC_SIZE, %rdx
|
||||
+# ifdef __ILP32__
|
||||
+ /* Clear the upper 32 bits. */
|
||||
+ mov %edx, %edx
|
||||
+# endif
|
||||
+ cmp $VEC_SIZE, %RDX_LP
|
||||
jb L(less_vec)
|
||||
- cmpq $(VEC_SIZE * 2), %rdx
|
||||
+ cmp $(VEC_SIZE * 2), %RDX_LP
|
||||
ja L(stosb_more_2x_vec)
|
||||
/* From VEC and to 2 * VEC. No branch when size == VEC_SIZE. */
|
||||
VMOVU %VEC(0), -VEC_SIZE(%rdi,%rdx)
|
||||
diff --git a/sysdeps/x86_64/x32/Makefile b/sysdeps/x86_64/x32/Makefile
|
||||
index e99dbd7c..98bd9ae9 100644
|
||||
--- a/sysdeps/x86_64/x32/Makefile
|
||||
+++ b/sysdeps/x86_64/x32/Makefile
|
||||
@@ -7,9 +7,9 @@ endif
|
||||
|
||||
ifeq ($(subdir),string)
|
||||
tests += tst-size_t-memchr tst-size_t-memcmp tst-size_t-memcpy \
|
||||
- tst-size_t-memrchr
|
||||
+ tst-size_t-memrchr tst-size_t-memset
|
||||
endif
|
||||
|
||||
ifeq ($(subdir),wcsmbs)
|
||||
-tests += tst-size_t-wmemchr tst-size_t-wmemcmp
|
||||
+tests += tst-size_t-wmemchr tst-size_t-wmemcmp tst-size_t-wmemset
|
||||
endif
|
||||
diff --git a/sysdeps/x86_64/x32/tst-size_t-memset.c b/sysdeps/x86_64/x32/tst-size_t-memset.c
|
||||
new file mode 100644
|
||||
index 00000000..2c367af6
|
||||
--- /dev/null
|
||||
+++ b/sysdeps/x86_64/x32/tst-size_t-memset.c
|
||||
@@ -0,0 +1,73 @@
|
||||
+/* Test memset with size_t in the lower 32 bits of 64-bit register.
|
||||
+ Copyright (C) 2019 Free Software Foundation, Inc.
|
||||
+ This file is part of the GNU C Library.
|
||||
+
|
||||
+ The GNU C Library is free software; you can redistribute it and/or
|
||||
+ modify it under the terms of the GNU Lesser General Public
|
||||
+ License as published by the Free Software Foundation; either
|
||||
+ version 2.1 of the License, or (at your option) any later version.
|
||||
+
|
||||
+ The GNU C Library is distributed in the hope that it will be useful,
|
||||
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
||||
+ Lesser General Public License for more details.
|
||||
+
|
||||
+ You should have received a copy of the GNU Lesser General Public
|
||||
+ License along with the GNU C Library; if not, see
|
||||
+ <http://www.gnu.org/licenses/>. */
|
||||
+
|
||||
+#ifdef WIDE
|
||||
+# define TEST_NAME "wmemset"
|
||||
+#else
|
||||
+# define TEST_NAME "memset"
|
||||
+#endif /* WIDE */
|
||||
+
|
||||
+#include "test-size_t.h"
|
||||
+
|
||||
+#ifdef WIDE
|
||||
+# include <wchar.h>
|
||||
+# define MEMSET wmemset
|
||||
+# define CHAR wchar_t
|
||||
+#else
|
||||
+# define MEMSET memset
|
||||
+# define CHAR char
|
||||
+#endif /* WIDE */
|
||||
+
|
||||
+IMPL (MEMSET, 1)
|
||||
+
|
||||
+typedef CHAR *(*proto_t) (CHAR *, int, size_t);
|
||||
+
|
||||
+static void *
|
||||
+__attribute__ ((noinline, noclone))
|
||||
+do_memset (parameter_t a, parameter_t b)
|
||||
+{
|
||||
+ return CALL (&b, a.p, (uintptr_t) b.p, a.len);
|
||||
+}
|
||||
+
|
||||
+static int
|
||||
+test_main (void)
|
||||
+{
|
||||
+ test_init ();
|
||||
+
|
||||
+ CHAR ch = 0x23;
|
||||
+ parameter_t src = { { page_size / sizeof (CHAR) }, buf2 };
|
||||
+ parameter_t c = { { 0 }, (void *) (uintptr_t) ch };
|
||||
+
|
||||
+ int ret = 0;
|
||||
+ FOR_EACH_IMPL (impl, 0)
|
||||
+ {
|
||||
+ c.fn = impl->fn;
|
||||
+ CHAR *p = (CHAR *) do_memset (src, c);
|
||||
+ size_t i;
|
||||
+ for (i = 0; i < src.len; i++)
|
||||
+ if (p[i] != ch)
|
||||
+ {
|
||||
+ error (0, 0, "Wrong result in function %s", impl->name);
|
||||
+ ret = 1;
|
||||
+ }
|
||||
+ }
|
||||
+
|
||||
+ return ret ? EXIT_FAILURE : EXIT_SUCCESS;
|
||||
+}
|
||||
+
|
||||
+#include <support/test-driver.c>
|
||||
diff --git a/sysdeps/x86_64/x32/tst-size_t-wmemset.c b/sysdeps/x86_64/x32/tst-size_t-wmemset.c
|
||||
new file mode 100644
|
||||
index 00000000..955eb488
|
||||
--- /dev/null
|
||||
+++ b/sysdeps/x86_64/x32/tst-size_t-wmemset.c
|
||||
@@ -0,0 +1,20 @@
|
||||
+/* Test wmemset with size_t in the lower 32 bits of 64-bit register.
|
||||
+ Copyright (C) 2019 Free Software Foundation, Inc.
|
||||
+ This file is part of the GNU C Library.
|
||||
+
|
||||
+ The GNU C Library is free software; you can redistribute it and/or
|
||||
+ modify it under the terms of the GNU Lesser General Public
|
||||
+ License as published by the Free Software Foundation; either
|
||||
+ version 2.1 of the License, or (at your option) any later version.
|
||||
+
|
||||
+ The GNU C Library is distributed in the hope that it will be useful,
|
||||
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
||||
+ Lesser General Public License for more details.
|
||||
+
|
||||
+ You should have received a copy of the GNU Lesser General Public
|
||||
+ License along with the GNU C Library; if not, see
|
||||
+ <http://www.gnu.org/licenses/>. */
|
||||
+
|
||||
+#define WIDE 1
|
||||
+#include "tst-size_t-memset.c"
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,43 +0,0 @@
|
||||
From 447954a206837b5f153869cfeeeab44631c3fac9 Mon Sep 17 00:00:00 2001
|
||||
Author: Shen-Ta Hsieh <ibmibmibm.tw@gmail.com> 2021-05-23 21:43:10
|
||||
Committer: H.J. Lu <hjl.tools@gmail.com> 2021-06-27 10:56:57
|
||||
Parent: 2c16cb88a6e5ace0fb7cedca86860ea7bde522a7 (Linux: Move timer helper routines from librt to libc)
|
||||
Child: 1683249d17e14827b6579529742eb895027dfa84 (x86_64: roundeven with sse4.1 support)
|
||||
Branches: master, remotes/origin/master and many more (41)
|
||||
Follows: glibc-2.33.9000
|
||||
Precedes: glibc-2.34
|
||||
|
||||
math: redirect roundeven function
|
||||
|
||||
This patch redirect roundeven function for futhermore changes.
|
||||
|
||||
Signed-off-by: Shen-Ta Hsieh <ibmibmibm.tw@gmail.com>
|
||||
Reviewed-by: H.J. Lu <hjl.tools@gmail.com>
|
||||
|
||||
Conflicts:
|
||||
*
|
||||
(rewritten for older branch)
|
||||
|
||||
diff --git a/sysdeps/ieee754/dbl-64/wordsize-64/s_roundeven.c b/sysdeps/ieee754/dbl-64/wordsize-64/s_roundeven.c
|
||||
index 7bbbb2dc..8728d0f2 100644
|
||||
--- a/sysdeps/ieee754/dbl-64/wordsize-64/s_roundeven.c
|
||||
+++ b/sysdeps/ieee754/dbl-64/wordsize-64/s_roundeven.c
|
||||
@@ -17,6 +17,7 @@
|
||||
License along with the GNU C Library; if not, see
|
||||
<http://www.gnu.org/licenses/>. */
|
||||
|
||||
+#define NO_MATH_REDIRECT
|
||||
#include <math.h>
|
||||
#include <math_private.h>
|
||||
#include <libm-alias-double.h>
|
||||
@@ -67,5 +68,6 @@ __roundeven (double x)
|
||||
INSERT_WORDS64 (x, ix);
|
||||
return x;
|
||||
}
|
||||
-hidden_def (__roundeven)
|
||||
+#ifndef __roundeven
|
||||
libm_alias_double (__roundeven, roundeven)
|
||||
+#endif
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,118 +0,0 @@
|
||||
From 447954a206837b5f153869cfeeeab44631c3fac9 Mon Sep 17 00:00:00 2001
|
||||
From: Shen-Ta Hsieh <ibmibmibm.tw@gmail.com>
|
||||
Date: Mon, 24 May 2021 09:43:10 +0800
|
||||
Subject: [PATCH] math: redirect roundeven function
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
This patch redirect roundeven function for futhermore changes.
|
||||
|
||||
Signed-off-by: Shen-Ta Hsieh <ibmibmibm.tw@gmail.com>
|
||||
Reviewed-by: H.J. Lu <hjl.tools@gmail.com>
|
||||
---
|
||||
include/math.h | 3 ++-
|
||||
sysdeps/ieee754/dbl-64/s_roundeven.c | 4 +++-
|
||||
sysdeps/ieee754/float128/s_roundevenf128.c | 1 +
|
||||
sysdeps/ieee754/flt-32/s_roundevenf.c | 3 +++
|
||||
sysdeps/ieee754/ldbl-128/s_roundevenl.c | 1 +
|
||||
sysdeps/ieee754/ldbl-96/s_roundevenl.c | 1 +
|
||||
6 files changed, 11 insertions(+), 2 deletions(-)
|
||||
|
||||
Conflicts:
|
||||
include/math.h
|
||||
(missing MATH_REDIRECT macros)
|
||||
|
||||
diff --git a/include/math.h b/include/math.h
|
||||
index e21d34b8..1f9f9a54 100644
|
||||
--- a/include/math.h
|
||||
+++ b/include/math.h
|
||||
@@ -38,7 +38,6 @@ libm_hidden_proto (__issignaling)
|
||||
libm_hidden_proto (__issignalingf)
|
||||
libm_hidden_proto (__exp)
|
||||
libm_hidden_proto (__expf)
|
||||
-libm_hidden_proto (__roundeven)
|
||||
|
||||
# ifndef __NO_LONG_DOUBLE_MATH
|
||||
libm_hidden_proto (__fpclassifyl)
|
||||
@@ -56,6 +55,8 @@ libm_hidden_proto (__expm1f128)
|
||||
|
||||
# if !(defined __FINITE_MATH_ONLY__ && __FINITE_MATH_ONLY__ > 0)
|
||||
# ifndef NO_MATH_REDIRECT
|
||||
+float (roundevenf) (float) asm ("__roundevenf");
|
||||
+double (roundeven) (double) asm ("__roundeven");
|
||||
/* Declare sqrt for use within GLIBC. Compilers typically inline sqrt as a
|
||||
single instruction. Use an asm to avoid use of PLTs if it doesn't. */
|
||||
float (sqrtf) (float) asm ("__ieee754_sqrtf");
|
||||
diff --git a/sysdeps/ieee754/dbl-64/s_roundeven.c b/sysdeps/ieee754/dbl-64/s_roundeven.c
|
||||
index 1438e81d..61962184 100644
|
||||
--- a/sysdeps/ieee754/dbl-64/s_roundeven.c
|
||||
+++ b/sysdeps/ieee754/dbl-64/s_roundeven.c
|
||||
@@ -17,6 +17,7 @@
|
||||
License along with the GNU C Library; if not, see
|
||||
<http://www.gnu.org/licenses/>. */
|
||||
|
||||
+#define NO_MATH_REDIRECT
|
||||
#include <math.h>
|
||||
#include <math_private.h>
|
||||
#include <libm-alias-double.h>
|
||||
@@ -101,5 +102,6 @@ __roundeven (double x)
|
||||
INSERT_WORDS (x, hx, lx);
|
||||
return x;
|
||||
}
|
||||
-hidden_def (__roundeven)
|
||||
+#ifndef __roundeven
|
||||
libm_alias_double (__roundeven, roundeven)
|
||||
+#endif
|
||||
diff --git a/sysdeps/ieee754/float128/s_roundevenf128.c b/sysdeps/ieee754/float128/s_roundevenf128.c
|
||||
index 5a9b3f39..e0faf727 100644
|
||||
--- a/sysdeps/ieee754/float128/s_roundevenf128.c
|
||||
+++ b/sysdeps/ieee754/float128/s_roundevenf128.c
|
||||
@@ -1,2 +1,3 @@
|
||||
+#define NO_MATH_REDIRECT
|
||||
#include <float128_private.h>
|
||||
#include "../ldbl-128/s_roundevenl.c"
|
||||
diff --git a/sysdeps/ieee754/flt-32/s_roundevenf.c b/sysdeps/ieee754/flt-32/s_roundevenf.c
|
||||
index 90f991d5..a661875e 100644
|
||||
--- a/sysdeps/ieee754/flt-32/s_roundevenf.c
|
||||
+++ b/sysdeps/ieee754/flt-32/s_roundevenf.c
|
||||
@@ -17,6 +17,7 @@
|
||||
License along with the GNU C Library; if not, see
|
||||
<http://www.gnu.org/licenses/>. */
|
||||
|
||||
+#define NO_MATH_REDIRECT
|
||||
#include <math.h>
|
||||
#include <math_private.h>
|
||||
#include <libm-alias-float.h>
|
||||
@@ -67,4 +68,6 @@ __roundevenf (float x)
|
||||
SET_FLOAT_WORD (x, ix);
|
||||
return x;
|
||||
}
|
||||
+#ifndef __roundevenf
|
||||
libm_alias_float (__roundeven, roundeven)
|
||||
+#endif
|
||||
diff --git a/sysdeps/ieee754/ldbl-128/s_roundevenl.c b/sysdeps/ieee754/ldbl-128/s_roundevenl.c
|
||||
index 5fc59af4..b9375b6c 100644
|
||||
--- a/sysdeps/ieee754/ldbl-128/s_roundevenl.c
|
||||
+++ b/sysdeps/ieee754/ldbl-128/s_roundevenl.c
|
||||
@@ -17,6 +17,7 @@
|
||||
License along with the GNU C Library; if not, see
|
||||
<http://www.gnu.org/licenses/>. */
|
||||
|
||||
+#define NO_MATH_REDIRECT
|
||||
#include <math.h>
|
||||
#include <math_private.h>
|
||||
#include <libm-alias-ldouble.h>
|
||||
diff --git a/sysdeps/ieee754/ldbl-96/s_roundevenl.c b/sysdeps/ieee754/ldbl-96/s_roundevenl.c
|
||||
index be2e4fa4..65031ab7 100644
|
||||
--- a/sysdeps/ieee754/ldbl-96/s_roundevenl.c
|
||||
+++ b/sysdeps/ieee754/ldbl-96/s_roundevenl.c
|
||||
@@ -17,6 +17,7 @@
|
||||
License along with the GNU C Library; if not, see
|
||||
<http://www.gnu.org/licenses/>. */
|
||||
|
||||
+#define NO_MATH_REDIRECT
|
||||
#include <math.h>
|
||||
#include <math_private.h>
|
||||
#include <libm-alias-ldouble.h>
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,242 +0,0 @@
|
||||
From 1683249d17e14827b6579529742eb895027dfa84 Mon Sep 17 00:00:00 2001
|
||||
From: Shen-Ta Hsieh <ibmibmibm.tw@gmail.com>
|
||||
Date: Mon, 24 May 2021 09:43:11 +0800
|
||||
Subject: [PATCH] x86_64: roundeven with sse4.1 support
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
This patch adds support for the sse4.1 hardware floating point
|
||||
roundeven.
|
||||
|
||||
Here is some benchmark results on my systems:
|
||||
|
||||
=AMD Ryzen 9 3900X 12-Core Processor=
|
||||
|
||||
* benchmark result before this commit
|
||||
| | roundeven | roundevenf |
|
||||
|------------|--------------|--------------|
|
||||
| duration | 3.75587e+09 | 3.75114e+09 |
|
||||
| iterations | 3.93053e+08 | 4.35402e+08 |
|
||||
| max | 52.592 | 58.71 |
|
||||
| min | 7.98 | 7.22 |
|
||||
| mean | 9.55563 | 8.61535 |
|
||||
|
||||
* benchmark result after this commit
|
||||
| | roundeven | roundevenf |
|
||||
|------------|---------------|--------------|
|
||||
| duration | 3.73815e+09 | 3.73738e+09 |
|
||||
| iterations | 5.82692e+08 | 5.91498e+08 |
|
||||
| max | 56.468 | 51.642 |
|
||||
| min | 6.27 | 6.156 |
|
||||
| mean | 6.41532 | 6.3185 |
|
||||
|
||||
=Intel(R) Pentium(R) CPU D1508 @ 2.20GHz=
|
||||
|
||||
* benchmark result before this commit
|
||||
| | roundeven | roundevenf |
|
||||
|------------|--------------|--------------|
|
||||
| duration | 2.18208e+09 | 2.18258e+09 |
|
||||
| iterations | 2.39932e+08 | 2.46924e+08 |
|
||||
| max | 96.378 | 98.035 |
|
||||
| min | 6.776 | 5.94 |
|
||||
| mean | 9.09456 | 8.83907 |
|
||||
|
||||
* benchmark result after this commit
|
||||
| | roundeven | roundevenf |
|
||||
|------------|--------------|--------------|
|
||||
| duration | 2.17415e+09 | 2.17005e+09 |
|
||||
| iterations | 3.56193e+08 | 4.09824e+08 |
|
||||
| max | 51.693 | 97.192 |
|
||||
| min | 5.926 | 5.093 |
|
||||
| mean | 6.10385 | 5.29507 |
|
||||
|
||||
Signed-off-by: Shen-Ta Hsieh <ibmibmibm.tw@gmail.com>
|
||||
Reviewed-by: H.J. Lu <hjl.tools@gmail.com>
|
||||
---
|
||||
sysdeps/x86_64/fpu/multiarch/Makefile | 5 +--
|
||||
sysdeps/x86_64/fpu/multiarch/s_roundeven-c.c | 2 ++
|
||||
.../x86_64/fpu/multiarch/s_roundeven-sse4_1.S | 24 ++++++++++++++
|
||||
sysdeps/x86_64/fpu/multiarch/s_roundeven.c | 31 +++++++++++++++++++
|
||||
sysdeps/x86_64/fpu/multiarch/s_roundevenf-c.c | 3 ++
|
||||
.../fpu/multiarch/s_roundevenf-sse4_1.S | 24 ++++++++++++++
|
||||
sysdeps/x86_64/fpu/multiarch/s_roundevenf.c | 31 +++++++++++++++++++
|
||||
7 files changed, 118 insertions(+), 2 deletions(-)
|
||||
create mode 100644 sysdeps/x86_64/fpu/multiarch/s_roundeven-c.c
|
||||
create mode 100644 sysdeps/x86_64/fpu/multiarch/s_roundeven-sse4_1.S
|
||||
create mode 100644 sysdeps/x86_64/fpu/multiarch/s_roundeven.c
|
||||
create mode 100644 sysdeps/x86_64/fpu/multiarch/s_roundevenf-c.c
|
||||
create mode 100644 sysdeps/x86_64/fpu/multiarch/s_roundevenf-sse4_1.S
|
||||
create mode 100644 sysdeps/x86_64/fpu/multiarch/s_roundevenf.c
|
||||
|
||||
diff --git a/sysdeps/x86_64/fpu/multiarch/Makefile b/sysdeps/x86_64/fpu/multiarch/Makefile
|
||||
index 9f387248..6ddd1c01 100644
|
||||
--- a/sysdeps/x86_64/fpu/multiarch/Makefile
|
||||
+++ b/sysdeps/x86_64/fpu/multiarch/Makefile
|
||||
@@ -1,11 +1,12 @@
|
||||
ifeq ($(subdir),math)
|
||||
libm-sysdep_routines += s_floor-c s_ceil-c s_floorf-c s_ceilf-c \
|
||||
s_rint-c s_rintf-c s_nearbyint-c s_nearbyintf-c \
|
||||
- s_trunc-c s_truncf-c
|
||||
+ s_roundeven-c s_roundevenf-c s_trunc-c s_truncf-c
|
||||
|
||||
libm-sysdep_routines += s_ceil-sse4_1 s_ceilf-sse4_1 s_floor-sse4_1 \
|
||||
s_floorf-sse4_1 s_nearbyint-sse4_1 \
|
||||
- s_nearbyintf-sse4_1 s_rint-sse4_1 s_rintf-sse4_1 \
|
||||
+ s_nearbyintf-sse4_1 s_roundeven-sse4_1 \
|
||||
+ s_roundevenf-sse4_1 s_rint-sse4_1 s_rintf-sse4_1 \
|
||||
s_trunc-sse4_1 s_truncf-sse4_1
|
||||
|
||||
libm-sysdep_routines += e_exp-fma e_log-fma e_pow-fma s_atan-fma \
|
||||
diff --git a/sysdeps/x86_64/fpu/multiarch/s_roundeven-c.c b/sysdeps/x86_64/fpu/multiarch/s_roundeven-c.c
|
||||
new file mode 100644
|
||||
index 00000000..c7be43cb
|
||||
--- /dev/null
|
||||
+++ b/sysdeps/x86_64/fpu/multiarch/s_roundeven-c.c
|
||||
@@ -0,0 +1,2 @@
|
||||
+#define __roundeven __roundeven_c
|
||||
+#include <sysdeps/ieee754/dbl-64/s_roundeven.c>
|
||||
diff --git a/sysdeps/x86_64/fpu/multiarch/s_roundeven-sse4_1.S b/sysdeps/x86_64/fpu/multiarch/s_roundeven-sse4_1.S
|
||||
new file mode 100644
|
||||
index 00000000..6ae8f6b1
|
||||
--- /dev/null
|
||||
+++ b/sysdeps/x86_64/fpu/multiarch/s_roundeven-sse4_1.S
|
||||
@@ -0,0 +1,24 @@
|
||||
+/* Copyright (C) 2021 Free Software Foundation, Inc.
|
||||
+ This file is part of the GNU C Library.
|
||||
+
|
||||
+ The GNU C Library is free software; you can redistribute it and/or
|
||||
+ modify it under the terms of the GNU Lesser General Public
|
||||
+ License as published by the Free Software Foundation; either
|
||||
+ version 2.1 of the License, or (at your option) any later version.
|
||||
+
|
||||
+ The GNU C Library is distributed in the hope that it will be useful,
|
||||
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
||||
+ Lesser General Public License for more details.
|
||||
+
|
||||
+ You should have received a copy of the GNU Lesser General Public
|
||||
+ License along with the GNU C Library; if not, see
|
||||
+ <https://www.gnu.org/licenses/>. */
|
||||
+
|
||||
+#include <sysdep.h>
|
||||
+
|
||||
+ .section .text.sse4.1,"ax",@progbits
|
||||
+ENTRY(__roundeven_sse41)
|
||||
+ roundsd $8, %xmm0, %xmm0
|
||||
+ ret
|
||||
+END(__roundeven_sse41)
|
||||
diff --git a/sysdeps/x86_64/fpu/multiarch/s_roundeven.c b/sysdeps/x86_64/fpu/multiarch/s_roundeven.c
|
||||
new file mode 100644
|
||||
index 00000000..d92eda65
|
||||
--- /dev/null
|
||||
+++ b/sysdeps/x86_64/fpu/multiarch/s_roundeven.c
|
||||
@@ -0,0 +1,31 @@
|
||||
+/* Multiple versions of __roundeven.
|
||||
+ Copyright (C) 2021 Free Software Foundation, Inc.
|
||||
+ This file is part of the GNU C Library.
|
||||
+
|
||||
+ The GNU C Library is free software; you can redistribute it and/or
|
||||
+ modify it under the terms of the GNU Lesser General Public
|
||||
+ License as published by the Free Software Foundation; either
|
||||
+ version 2.1 of the License, or (at your option) any later version.
|
||||
+
|
||||
+ The GNU C Library is distributed in the hope that it will be useful,
|
||||
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
||||
+ Lesser General Public License for more details.
|
||||
+
|
||||
+ You should have received a copy of the GNU Lesser General Public
|
||||
+ License along with the GNU C Library; if not, see
|
||||
+ <https://www.gnu.org/licenses/>. */
|
||||
+
|
||||
+#include <libm-alias-double.h>
|
||||
+
|
||||
+#define roundeven __redirect_roundeven
|
||||
+#define __roundeven __redirect___roundeven
|
||||
+#include <math.h>
|
||||
+#undef roundeven
|
||||
+#undef __roundeven
|
||||
+
|
||||
+#define SYMBOL_NAME roundeven
|
||||
+#include "ifunc-sse4_1.h"
|
||||
+
|
||||
+libc_ifunc_redirected (__redirect_roundeven, __roundeven, IFUNC_SELECTOR ());
|
||||
+libm_alias_double (__roundeven, roundeven)
|
||||
diff --git a/sysdeps/x86_64/fpu/multiarch/s_roundevenf-c.c b/sysdeps/x86_64/fpu/multiarch/s_roundevenf-c.c
|
||||
new file mode 100644
|
||||
index 00000000..72a6e7d1
|
||||
--- /dev/null
|
||||
+++ b/sysdeps/x86_64/fpu/multiarch/s_roundevenf-c.c
|
||||
@@ -0,0 +1,3 @@
|
||||
+#undef __roundevenf
|
||||
+#define __roundevenf __roundevenf_c
|
||||
+#include <sysdeps/ieee754/flt-32/s_roundevenf.c>
|
||||
diff --git a/sysdeps/x86_64/fpu/multiarch/s_roundevenf-sse4_1.S b/sysdeps/x86_64/fpu/multiarch/s_roundevenf-sse4_1.S
|
||||
new file mode 100644
|
||||
index 00000000..a76e1080
|
||||
--- /dev/null
|
||||
+++ b/sysdeps/x86_64/fpu/multiarch/s_roundevenf-sse4_1.S
|
||||
@@ -0,0 +1,24 @@
|
||||
+/* Copyright (C) 2021 Free Software Foundation, Inc.
|
||||
+ This file is part of the GNU C Library.
|
||||
+
|
||||
+ The GNU C Library is free software; you can redistribute it and/or
|
||||
+ modify it under the terms of the GNU Lesser General Public
|
||||
+ License as published by the Free Software Foundation; either
|
||||
+ version 2.1 of the License, or (at your option) any later version.
|
||||
+
|
||||
+ The GNU C Library is distributed in the hope that it will be useful,
|
||||
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
||||
+ Lesser General Public License for more details.
|
||||
+
|
||||
+ You should have received a copy of the GNU Lesser General Public
|
||||
+ License along with the GNU C Library; if not, see
|
||||
+ <https://www.gnu.org/licenses/>. */
|
||||
+
|
||||
+#include <sysdep.h>
|
||||
+
|
||||
+ .section .text.sse4.1,"ax",@progbits
|
||||
+ENTRY(__roundevenf_sse41)
|
||||
+ roundss $8, %xmm0, %xmm0
|
||||
+ ret
|
||||
+END(__roundevenf_sse41)
|
||||
diff --git a/sysdeps/x86_64/fpu/multiarch/s_roundevenf.c b/sysdeps/x86_64/fpu/multiarch/s_roundevenf.c
|
||||
new file mode 100644
|
||||
index 00000000..2ee196e6
|
||||
--- /dev/null
|
||||
+++ b/sysdeps/x86_64/fpu/multiarch/s_roundevenf.c
|
||||
@@ -0,0 +1,31 @@
|
||||
+/* Multiple versions of __roundevenf.
|
||||
+ Copyright (C) 2021 Free Software Foundation, Inc.
|
||||
+ This file is part of the GNU C Library.
|
||||
+
|
||||
+ The GNU C Library is free software; you can redistribute it and/or
|
||||
+ modify it under the terms of the GNU Lesser General Public
|
||||
+ License as published by the Free Software Foundation; either
|
||||
+ version 2.1 of the License, or (at your option) any later version.
|
||||
+
|
||||
+ The GNU C Library is distributed in the hope that it will be useful,
|
||||
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
||||
+ Lesser General Public License for more details.
|
||||
+
|
||||
+ You should have received a copy of the GNU Lesser General Public
|
||||
+ License along with the GNU C Library; if not, see
|
||||
+ <https://www.gnu.org/licenses/>. */
|
||||
+
|
||||
+#include <libm-alias-float.h>
|
||||
+
|
||||
+#define roundevenf __redirect_roundevenf
|
||||
+#define __roundevenf __redirect___roundevenf
|
||||
+#include <math.h>
|
||||
+#undef roundevenf
|
||||
+#undef __roundevenf
|
||||
+
|
||||
+#define SYMBOL_NAME roundevenf
|
||||
+#include "ifunc-sse4_1.h"
|
||||
+
|
||||
+libc_ifunc_redirected (__redirect_roundevenf, __roundevenf, IFUNC_SELECTOR ());
|
||||
+libm_alias_float (__roundeven, roundeven)
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,41 +0,0 @@
|
||||
From 7e08db3359c86c94918feb33a1182cd0ff3bb10b Mon Sep 17 00:00:00 2001
|
||||
From: Noah Goldstein <goldstein.w.n@gmail.com>
|
||||
Date: Sun, 9 Jan 2022 16:02:28 -0600
|
||||
Subject: [PATCH] x86: Fix __wcsncmp_evex in strcmp-evex.S [BZ# 28755]
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
Fixes [BZ# 28755] for wcsncmp by redirecting length >= 2^56 to
|
||||
__wcscmp_evex. For x86_64 this covers the entire address range so any
|
||||
length larger could not possibly be used to bound `s1` or `s2`.
|
||||
|
||||
test-strcmp, test-strncmp, test-wcscmp, and test-wcsncmp all pass.
|
||||
|
||||
Signed-off-by: Noah Goldstein <goldstein.w.n@gmail.com>
|
||||
---
|
||||
sysdeps/x86_64/multiarch/strcmp-evex.S | 10 ++++++++++
|
||||
1 file changed, 10 insertions(+)
|
||||
|
||||
diff --git a/sysdeps/x86_64/multiarch/strcmp-evex.S b/sysdeps/x86_64/multiarch/strcmp-evex.S
|
||||
index 459eeed0..d5aa6daa 100644
|
||||
--- a/sysdeps/x86_64/multiarch/strcmp-evex.S
|
||||
+++ b/sysdeps/x86_64/multiarch/strcmp-evex.S
|
||||
@@ -97,6 +97,16 @@ ENTRY (STRCMP)
|
||||
je L(char0)
|
||||
jb L(zero)
|
||||
# ifdef USE_AS_WCSCMP
|
||||
+# ifndef __ILP32__
|
||||
+ movq %rdx, %rcx
|
||||
+ /* Check if length could overflow when multiplied by
|
||||
+ sizeof(wchar_t). Checking top 8 bits will cover all potential
|
||||
+ overflow cases as well as redirect cases where its impossible to
|
||||
+ length to bound a valid memory region. In these cases just use
|
||||
+ 'wcscmp'. */
|
||||
+ shrq $56, %rcx
|
||||
+ jnz __wcscmp_evex
|
||||
+# endif
|
||||
/* Convert units: from wide to byte char. */
|
||||
shl $2, %RDX_LP
|
||||
# endif
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,45 +0,0 @@
|
||||
From bad852b61b79503fcb3c5fc379c70f768df3e1fb Mon Sep 17 00:00:00 2001
|
||||
From: Noah Goldstein <goldstein.w.n@gmail.com>
|
||||
Date: Sat, 23 Oct 2021 01:26:47 -0400
|
||||
Subject: [PATCH] x86: Replace sse2 instructions with avx in
|
||||
memcmp-evex-movbe.S
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
This commit replaces two usages of SSE2 'movups' with AVX 'vmovdqu'.
|
||||
|
||||
it could potentially be dangerous to use SSE2 if this function is ever
|
||||
called without using 'vzeroupper' beforehand. While compilers appear
|
||||
to use 'vzeroupper' before function calls if AVX2 has been used, using
|
||||
SSE2 here is more brittle. Since it is not absolutely necessary it
|
||||
should be avoided.
|
||||
|
||||
It costs 2-extra bytes but the extra bytes should only eat into
|
||||
alignment padding.
|
||||
Reviewed-by: H.J. Lu <hjl.tools@gmail.com>
|
||||
---
|
||||
sysdeps/x86_64/multiarch/memcmp-evex-movbe.S | 4 ++--
|
||||
1 file changed, 2 insertions(+), 2 deletions(-)
|
||||
|
||||
diff --git a/sysdeps/x86_64/multiarch/memcmp-evex-movbe.S b/sysdeps/x86_64/multiarch/memcmp-evex-movbe.S
|
||||
index 2761b54f..640f6757 100644
|
||||
--- a/sysdeps/x86_64/multiarch/memcmp-evex-movbe.S
|
||||
+++ b/sysdeps/x86_64/multiarch/memcmp-evex-movbe.S
|
||||
@@ -561,13 +561,13 @@ L(between_16_31):
|
||||
/* From 16 to 31 bytes. No branch when size == 16. */
|
||||
|
||||
/* Use movups to save code size. */
|
||||
- movups (%rsi), %xmm2
|
||||
+ vmovdqu (%rsi), %xmm2
|
||||
VPCMP $4, (%rdi), %xmm2, %k1
|
||||
kmovd %k1, %eax
|
||||
testl %eax, %eax
|
||||
jnz L(return_vec_0_lv)
|
||||
/* Use overlapping loads to avoid branches. */
|
||||
- movups -16(%rsi, %rdx, CHAR_SIZE), %xmm2
|
||||
+ vmovdqu -16(%rsi, %rdx, CHAR_SIZE), %xmm2
|
||||
VPCMP $4, -16(%rdi, %rdx, CHAR_SIZE), %xmm2, %k1
|
||||
addl $(CHAR_PER_VEC - (16 / CHAR_SIZE)), %edx
|
||||
kmovd %k1, %eax
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,300 +0,0 @@
|
||||
From ee915088a0231cd421054dbd8abab7aadf331153 Mon Sep 17 00:00:00 2001
|
||||
From: "H.J. Lu" <hjl.tools@gmail.com>
|
||||
Date: Mon, 21 Jan 2019 11:33:52 -0800
|
||||
Subject: [PATCH] x86-64 strncmp family: Properly handle the length parameter
|
||||
[BZ# 24097]
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
On x32, the size_t parameter may be passed in the lower 32 bits of a
|
||||
64-bit register with the non-zero upper 32 bits. The string/memory
|
||||
functions written in assembly can only use the lower 32 bits of a
|
||||
64-bit register as length or must clear the upper 32 bits before using
|
||||
the full 64-bit register for length.
|
||||
|
||||
This pach fixes the strncmp family for x32. Tested on x86-64 and x32.
|
||||
On x86-64, libc.so is the same with and withou the fix.
|
||||
|
||||
[BZ# 24097]
|
||||
CVE-2019-6488
|
||||
* sysdeps/x86_64/multiarch/strcmp-avx2.S: Use RDX_LP for length.
|
||||
* sysdeps/x86_64/multiarch/strcmp-sse42.S: Likewise.
|
||||
* sysdeps/x86_64/strcmp.S: Likewise.
|
||||
* sysdeps/x86_64/x32/Makefile (tests): Add tst-size_t-strncasecmp,
|
||||
tst-size_t-strncmp and tst-size_t-wcsncmp.
|
||||
* sysdeps/x86_64/x32/tst-size_t-strncasecmp.c: New file.
|
||||
* sysdeps/x86_64/x32/tst-size_t-strncmp.c: Likewise.
|
||||
* sysdeps/x86_64/x32/tst-size_t-wcsncmp.c: Likewise.
|
||||
---
|
||||
sysdeps/x86_64/multiarch/strcmp-avx2.S | 6 +-
|
||||
sysdeps/x86_64/multiarch/strcmp-sse42.S | 6 +-
|
||||
sysdeps/x86_64/strcmp.S | 6 +-
|
||||
sysdeps/x86_64/x32/Makefile | 6 +-
|
||||
sysdeps/x86_64/x32/tst-size_t-strncasecmp.c | 59 ++++++++++++++++
|
||||
sysdeps/x86_64/x32/tst-size_t-strncmp.c | 78 +++++++++++++++++++++
|
||||
sysdeps/x86_64/x32/tst-size_t-wcsncmp.c | 20 ++++++
|
||||
7 files changed, 170 insertions(+), 11 deletions(-)
|
||||
create mode 100644 sysdeps/x86_64/x32/tst-size_t-strncasecmp.c
|
||||
create mode 100644 sysdeps/x86_64/x32/tst-size_t-strncmp.c
|
||||
create mode 100644 sysdeps/x86_64/x32/tst-size_t-wcsncmp.c
|
||||
|
||||
Conflicts:
|
||||
ChangeLog
|
||||
(removed)
|
||||
|
||||
diff --git a/sysdeps/x86_64/multiarch/strcmp-avx2.S b/sysdeps/x86_64/multiarch/strcmp-avx2.S
|
||||
index 327e3d87..156c1949 100644
|
||||
--- a/sysdeps/x86_64/multiarch/strcmp-avx2.S
|
||||
+++ b/sysdeps/x86_64/multiarch/strcmp-avx2.S
|
||||
@@ -79,15 +79,15 @@
|
||||
ENTRY (STRCMP)
|
||||
# ifdef USE_AS_STRNCMP
|
||||
/* Check for simple cases (0 or 1) in offset. */
|
||||
- cmp $1, %rdx
|
||||
+ cmp $1, %RDX_LP
|
||||
je L(char0)
|
||||
jb L(zero)
|
||||
# ifdef USE_AS_WCSCMP
|
||||
/* Convert units: from wide to byte char. */
|
||||
- shl $2, %rdx
|
||||
+ shl $2, %RDX_LP
|
||||
# endif
|
||||
/* Register %r11 tracks the maximum offset. */
|
||||
- movq %rdx, %r11
|
||||
+ mov %RDX_LP, %R11_LP
|
||||
# endif
|
||||
movl %edi, %eax
|
||||
xorl %edx, %edx
|
||||
diff --git a/sysdeps/x86_64/multiarch/strcmp-sse42.S b/sysdeps/x86_64/multiarch/strcmp-sse42.S
|
||||
index d3c07bd2..a1ebea46 100644
|
||||
--- a/sysdeps/x86_64/multiarch/strcmp-sse42.S
|
||||
+++ b/sysdeps/x86_64/multiarch/strcmp-sse42.S
|
||||
@@ -156,11 +156,11 @@ STRCMP_SSE42:
|
||||
#endif
|
||||
|
||||
#if defined USE_AS_STRNCMP || defined USE_AS_STRNCASECMP_L
|
||||
- test %rdx, %rdx
|
||||
+ test %RDX_LP, %RDX_LP
|
||||
je LABEL(strcmp_exitz)
|
||||
- cmp $1, %rdx
|
||||
+ cmp $1, %RDX_LP
|
||||
je LABEL(Byte0)
|
||||
- mov %rdx, %r11
|
||||
+ mov %RDX_LP, %R11_LP
|
||||
#endif
|
||||
mov %esi, %ecx
|
||||
mov %edi, %eax
|
||||
diff --git a/sysdeps/x86_64/strcmp.S b/sysdeps/x86_64/strcmp.S
|
||||
index e16945b9..f47c8ad4 100644
|
||||
--- a/sysdeps/x86_64/strcmp.S
|
||||
+++ b/sysdeps/x86_64/strcmp.S
|
||||
@@ -135,11 +135,11 @@ ENTRY (STRCMP)
|
||||
* This implementation uses SSE to compare up to 16 bytes at a time.
|
||||
*/
|
||||
#if defined USE_AS_STRNCMP || defined USE_AS_STRNCASECMP_L
|
||||
- test %rdx, %rdx
|
||||
+ test %RDX_LP, %RDX_LP
|
||||
je LABEL(strcmp_exitz)
|
||||
- cmp $1, %rdx
|
||||
+ cmp $1, %RDX_LP
|
||||
je LABEL(Byte0)
|
||||
- mov %rdx, %r11
|
||||
+ mov %RDX_LP, %R11_LP
|
||||
#endif
|
||||
mov %esi, %ecx
|
||||
mov %edi, %eax
|
||||
diff --git a/sysdeps/x86_64/x32/Makefile b/sysdeps/x86_64/x32/Makefile
|
||||
index 98bd9ae9..db302839 100644
|
||||
--- a/sysdeps/x86_64/x32/Makefile
|
||||
+++ b/sysdeps/x86_64/x32/Makefile
|
||||
@@ -7,9 +7,11 @@ endif
|
||||
|
||||
ifeq ($(subdir),string)
|
||||
tests += tst-size_t-memchr tst-size_t-memcmp tst-size_t-memcpy \
|
||||
- tst-size_t-memrchr tst-size_t-memset
|
||||
+ tst-size_t-memrchr tst-size_t-memset tst-size_t-strncasecmp \
|
||||
+ tst-size_t-strncmp
|
||||
endif
|
||||
|
||||
ifeq ($(subdir),wcsmbs)
|
||||
-tests += tst-size_t-wmemchr tst-size_t-wmemcmp tst-size_t-wmemset
|
||||
+tests += tst-size_t-wmemchr tst-size_t-wmemcmp tst-size_t-wmemset \
|
||||
+ tst-size_t-wcsncmp
|
||||
endif
|
||||
diff --git a/sysdeps/x86_64/x32/tst-size_t-strncasecmp.c b/sysdeps/x86_64/x32/tst-size_t-strncasecmp.c
|
||||
new file mode 100644
|
||||
index 00000000..86233593
|
||||
--- /dev/null
|
||||
+++ b/sysdeps/x86_64/x32/tst-size_t-strncasecmp.c
|
||||
@@ -0,0 +1,59 @@
|
||||
+/* Test strncaecmp with size_t in the lower 32 bits of 64-bit register.
|
||||
+ Copyright (C) 2019 Free Software Foundation, Inc.
|
||||
+ This file is part of the GNU C Library.
|
||||
+
|
||||
+ The GNU C Library is free software; you can redistribute it and/or
|
||||
+ modify it under the terms of the GNU Lesser General Public
|
||||
+ License as published by the Free Software Foundation; either
|
||||
+ version 2.1 of the License, or (at your option) any later version.
|
||||
+
|
||||
+ The GNU C Library is distributed in the hope that it will be useful,
|
||||
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
||||
+ Lesser General Public License for more details.
|
||||
+
|
||||
+ You should have received a copy of the GNU Lesser General Public
|
||||
+ License along with the GNU C Library; if not, see
|
||||
+ <http://www.gnu.org/licenses/>. */
|
||||
+
|
||||
+#define TEST_NAME "strncasecmp"
|
||||
+#include "test-size_t.h"
|
||||
+
|
||||
+IMPL (strncasecmp, 1)
|
||||
+
|
||||
+typedef int (*proto_t) (const char *, const char *, size_t);
|
||||
+
|
||||
+static int
|
||||
+__attribute__ ((noinline, noclone))
|
||||
+do_strncasecmp (parameter_t a, parameter_t b)
|
||||
+{
|
||||
+ return CALL (&b, a.p, b.p, a.len);
|
||||
+}
|
||||
+
|
||||
+static int
|
||||
+test_main (void)
|
||||
+{
|
||||
+ test_init ();
|
||||
+
|
||||
+ parameter_t dest = { { page_size }, buf1 };
|
||||
+ parameter_t src = { { 0 }, buf2 };
|
||||
+
|
||||
+ strncpy ((char *) buf1, (const char *) buf2, page_size);
|
||||
+
|
||||
+ int ret = 0;
|
||||
+ FOR_EACH_IMPL (impl, 0)
|
||||
+ {
|
||||
+ src.fn = impl->fn;
|
||||
+ int res = do_strncasecmp (dest, src);
|
||||
+ if (res)
|
||||
+ {
|
||||
+ error (0, 0, "Wrong result in function %s: %i != 0",
|
||||
+ impl->name, res);
|
||||
+ ret = 1;
|
||||
+ }
|
||||
+ }
|
||||
+
|
||||
+ return ret ? EXIT_FAILURE : EXIT_SUCCESS;
|
||||
+}
|
||||
+
|
||||
+#include <support/test-driver.c>
|
||||
diff --git a/sysdeps/x86_64/x32/tst-size_t-strncmp.c b/sysdeps/x86_64/x32/tst-size_t-strncmp.c
|
||||
new file mode 100644
|
||||
index 00000000..54e6bd83
|
||||
--- /dev/null
|
||||
+++ b/sysdeps/x86_64/x32/tst-size_t-strncmp.c
|
||||
@@ -0,0 +1,78 @@
|
||||
+/* Test strncmp with size_t in the lower 32 bits of 64-bit register.
|
||||
+ Copyright (C) 2019 Free Software Foundation, Inc.
|
||||
+ This file is part of the GNU C Library.
|
||||
+
|
||||
+ The GNU C Library is free software; you can redistribute it and/or
|
||||
+ modify it under the terms of the GNU Lesser General Public
|
||||
+ License as published by the Free Software Foundation; either
|
||||
+ version 2.1 of the License, or (at your option) any later version.
|
||||
+
|
||||
+ The GNU C Library is distributed in the hope that it will be useful,
|
||||
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
||||
+ Lesser General Public License for more details.
|
||||
+
|
||||
+ You should have received a copy of the GNU Lesser General Public
|
||||
+ License along with the GNU C Library; if not, see
|
||||
+ <http://www.gnu.org/licenses/>. */
|
||||
+
|
||||
+#ifdef WIDE
|
||||
+# define TEST_NAME "wcsncmp"
|
||||
+#else
|
||||
+# define TEST_NAME "strncmp"
|
||||
+#endif
|
||||
+
|
||||
+#include "test-size_t.h"
|
||||
+
|
||||
+#ifdef WIDE
|
||||
+# include <wchar.h>
|
||||
+
|
||||
+# define STRNCMP wcsncmp
|
||||
+# define STRNCPY wcsncpy
|
||||
+# define CHAR wchar_t
|
||||
+#else
|
||||
+# define STRNCMP strncmp
|
||||
+# define STRNCPY strncpy
|
||||
+# define CHAR char
|
||||
+#endif
|
||||
+
|
||||
+IMPL (STRNCMP, 1)
|
||||
+
|
||||
+typedef int (*proto_t) (const CHAR *, const CHAR *, size_t);
|
||||
+
|
||||
+
|
||||
+static int
|
||||
+__attribute__ ((noinline, noclone))
|
||||
+do_strncmp (parameter_t a, parameter_t b)
|
||||
+{
|
||||
+ return CALL (&b, a.p, b.p, a.len);
|
||||
+}
|
||||
+
|
||||
+static int
|
||||
+test_main (void)
|
||||
+{
|
||||
+ test_init ();
|
||||
+
|
||||
+ size_t size = page_size / sizeof (CHAR);
|
||||
+ parameter_t dest = { { size }, buf1 };
|
||||
+ parameter_t src = { { 0 }, buf2 };
|
||||
+
|
||||
+ STRNCPY ((CHAR *) buf1, (const CHAR *) buf2, size);
|
||||
+
|
||||
+ int ret = 0;
|
||||
+ FOR_EACH_IMPL (impl, 0)
|
||||
+ {
|
||||
+ src.fn = impl->fn;
|
||||
+ int res = do_strncmp (dest, src);
|
||||
+ if (res)
|
||||
+ {
|
||||
+ error (0, 0, "Wrong result in function %s: %i != 0",
|
||||
+ impl->name, res);
|
||||
+ ret = 1;
|
||||
+ }
|
||||
+ }
|
||||
+
|
||||
+ return ret ? EXIT_FAILURE : EXIT_SUCCESS;
|
||||
+}
|
||||
+
|
||||
+#include <support/test-driver.c>
|
||||
diff --git a/sysdeps/x86_64/x32/tst-size_t-wcsncmp.c b/sysdeps/x86_64/x32/tst-size_t-wcsncmp.c
|
||||
new file mode 100644
|
||||
index 00000000..4829647c
|
||||
--- /dev/null
|
||||
+++ b/sysdeps/x86_64/x32/tst-size_t-wcsncmp.c
|
||||
@@ -0,0 +1,20 @@
|
||||
+/* Test wcsncmp with size_t in the lower 32 bits of 64-bit register.
|
||||
+ Copyright (C) 2019 Free Software Foundation, Inc.
|
||||
+ This file is part of the GNU C Library.
|
||||
+
|
||||
+ The GNU C Library is free software; you can redistribute it and/or
|
||||
+ modify it under the terms of the GNU Lesser General Public
|
||||
+ License as published by the Free Software Foundation; either
|
||||
+ version 2.1 of the License, or (at your option) any later version.
|
||||
+
|
||||
+ The GNU C Library is distributed in the hope that it will be useful,
|
||||
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
||||
+ Lesser General Public License for more details.
|
||||
+
|
||||
+ You should have received a copy of the GNU Lesser General Public
|
||||
+ License along with the GNU C Library; if not, see
|
||||
+ <http://www.gnu.org/licenses/>. */
|
||||
+
|
||||
+#define WIDE 1
|
||||
+#include "tst-size_t-strncmp.c"
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,56 +0,0 @@
|
||||
From cf2c57526ba4b57e6863ad4db8a868e2678adce8 Mon Sep 17 00:00:00 2001
|
||||
From: "H.J. Lu" <hjl.tools@gmail.com>
|
||||
Date: Fri, 30 Apr 2021 05:58:59 -0700
|
||||
Subject: [PATCH] x86: Set rep_movsb_threshold to 2112 on processors with FSRM
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
The glibc memcpy benchmark on Intel Core i7-1065G7 (Ice Lake) showed
|
||||
that REP MOVSB became faster after 2112 bytes:
|
||||
|
||||
Vector Move REP MOVSB
|
||||
length=2112, align1=0, align2=0: 24.20 24.40
|
||||
length=2112, align1=1, align2=0: 26.07 23.13
|
||||
length=2112, align1=0, align2=1: 27.18 28.13
|
||||
length=2112, align1=1, align2=1: 26.23 25.16
|
||||
length=2176, align1=0, align2=0: 23.18 22.52
|
||||
length=2176, align1=2, align2=0: 25.45 22.52
|
||||
length=2176, align1=0, align2=2: 27.14 27.82
|
||||
length=2176, align1=2, align2=2: 22.73 25.56
|
||||
length=2240, align1=0, align2=0: 24.62 24.25
|
||||
length=2240, align1=3, align2=0: 29.77 27.15
|
||||
length=2240, align1=0, align2=3: 35.55 29.93
|
||||
length=2240, align1=3, align2=3: 34.49 25.15
|
||||
length=2304, align1=0, align2=0: 34.75 26.64
|
||||
length=2304, align1=4, align2=0: 32.09 22.63
|
||||
length=2304, align1=0, align2=4: 28.43 31.24
|
||||
|
||||
Use REP MOVSB for data size > 2112 bytes in memcpy on processors with
|
||||
fast short REP MOVSB (FSRM).
|
||||
|
||||
* sysdeps/x86/dl-cacheinfo.h (dl_init_cacheinfo): Set
|
||||
rep_movsb_threshold to 2112 on processors with fast short REP
|
||||
MOVSB (FSRM).
|
||||
---
|
||||
sysdeps/x86/cacheinfo.h | 6 ++++++
|
||||
1 file changed, 6 insertions(+)
|
||||
|
||||
diff --git a/sysdeps/x86/cacheinfo.h b/sysdeps/x86/cacheinfo.h
|
||||
index f72f634a..cc3941d3 100644
|
||||
--- a/sysdeps/x86/cacheinfo.h
|
||||
+++ b/sysdeps/x86/cacheinfo.h
|
||||
@@ -430,6 +430,12 @@ init_cacheinfo (void)
|
||||
rep_movsb_threshold = 2048 * (16 / 16);
|
||||
minimum_rep_movsb_threshold = 16 * 8;
|
||||
}
|
||||
+
|
||||
+ /* NB: The default REP MOVSB threshold is 2112 on processors with fast
|
||||
+ short REP MOVSB (FSRM). */
|
||||
+ if (CPU_FEATURE_USABLE_P (cpu_features, FSRM))
|
||||
+ rep_movsb_threshold = 2112;
|
||||
+
|
||||
if (cpu_features->rep_movsb_threshold > minimum_rep_movsb_threshold)
|
||||
__x86_rep_movsb_threshold = cpu_features->rep_movsb_threshold;
|
||||
else
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,51 +0,0 @@
|
||||
From d672a98a1af106bd68deb15576710cd61363f7a6 Mon Sep 17 00:00:00 2001
|
||||
From: "H.J. Lu" <hjl.tools@gmail.com>
|
||||
Date: Tue, 2 Nov 2021 18:33:07 -0700
|
||||
Subject: [PATCH] Add LLL_MUTEX_READ_LOCK [BZ #28537]
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
CAS instruction is expensive. From the x86 CPU's point of view, getting
|
||||
a cache line for writing is more expensive than reading. See Appendix
|
||||
A.2 Spinlock in:
|
||||
|
||||
https://www.intel.com/content/dam/www/public/us/en/documents/white-papers/xeon-lock-scaling-analysis-paper.pdf
|
||||
|
||||
The full compare and swap will grab the cache line exclusive and cause
|
||||
excessive cache line bouncing.
|
||||
|
||||
Add LLL_MUTEX_READ_LOCK to do an atomic load and skip CAS in spinlock
|
||||
loop if compare may fail to reduce cache line bouncing on contended locks.
|
||||
|
||||
Reviewed-by: Szabolcs Nagy <szabolcs.nagy@arm.com>
|
||||
---
|
||||
nptl/pthread_mutex_lock.c | 7 +++++++
|
||||
1 file changed, 7 insertions(+)
|
||||
|
||||
diff --git a/nptl/pthread_mutex_lock.c b/nptl/pthread_mutex_lock.c
|
||||
index 60ada70d..eb4d8baa 100644
|
||||
--- a/nptl/pthread_mutex_lock.c
|
||||
+++ b/nptl/pthread_mutex_lock.c
|
||||
@@ -56,6 +56,11 @@
|
||||
#define FORCE_ELISION(m, s)
|
||||
#endif
|
||||
|
||||
+#ifndef LLL_MUTEX_READ_LOCK
|
||||
+# define LLL_MUTEX_READ_LOCK(mutex) \
|
||||
+ atomic_load_relaxed (&(mutex)->__data.__lock)
|
||||
+#endif
|
||||
+
|
||||
static int __pthread_mutex_lock_full (pthread_mutex_t *mutex)
|
||||
__attribute_noinline__;
|
||||
|
||||
@@ -136,6 +141,8 @@ __pthread_mutex_lock (pthread_mutex_t *mutex)
|
||||
break;
|
||||
}
|
||||
atomic_spin_nop ();
|
||||
+ if (LLL_MUTEX_READ_LOCK (mutex) != 0)
|
||||
+ continue;
|
||||
}
|
||||
while (LLL_MUTEX_TRYLOCK (mutex) != 0);
|
||||
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,71 +0,0 @@
|
||||
From 120ac6d238825452e8024e2f627da33b2508dfd3 Mon Sep 17 00:00:00 2001
|
||||
From: "H.J. Lu" <hjl.tools@gmail.com>
|
||||
Date: Fri, 12 Nov 2021 11:47:42 -0800
|
||||
Subject: [PATCH] Move assignment out of the CAS condition
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
Update
|
||||
|
||||
commit 49302b8fdf9103b6fc0a398678668a22fa19574c
|
||||
Author: H.J. Lu <hjl.tools@gmail.com>
|
||||
Date: Thu Nov 11 06:54:01 2021 -0800
|
||||
|
||||
Avoid extra load with CAS in __pthread_mutex_clocklock_common [BZ #28537]
|
||||
|
||||
Replace boolean CAS with value CAS to avoid the extra load.
|
||||
|
||||
and
|
||||
|
||||
commit 0b82747dc48d5bf0871bdc6da8cb6eec1256355f
|
||||
Author: H.J. Lu <hjl.tools@gmail.com>
|
||||
Date: Thu Nov 11 06:31:51 2021 -0800
|
||||
|
||||
Avoid extra load with CAS in __pthread_mutex_lock_full [BZ #28537]
|
||||
|
||||
Replace boolean CAS with value CAS to avoid the extra load.
|
||||
|
||||
by moving assignment out of the CAS condition.
|
||||
---
|
||||
nptl/pthread_mutex_lock.c | 7 +++----
|
||||
nptl/pthread_mutex_timedlock.c | 7 +++----
|
||||
2 files changed, 6 insertions(+), 8 deletions(-)
|
||||
|
||||
diff --git a/nptl/pthread_mutex_lock.c b/nptl/pthread_mutex_lock.c
|
||||
index eb4d8baa..a633d95e 100644
|
||||
--- a/nptl/pthread_mutex_lock.c
|
||||
+++ b/nptl/pthread_mutex_lock.c
|
||||
@@ -299,10 +299,9 @@ __pthread_mutex_lock_full (pthread_mutex_t *mutex)
|
||||
meantime. */
|
||||
if ((oldval & FUTEX_WAITERS) == 0)
|
||||
{
|
||||
- int val;
|
||||
- if ((val = atomic_compare_and_exchange_val_acq
|
||||
- (&mutex->__data.__lock, oldval | FUTEX_WAITERS,
|
||||
- oldval)) != oldval)
|
||||
+ int val = atomic_compare_and_exchange_val_acq
|
||||
+ (&mutex->__data.__lock, oldval | FUTEX_WAITERS, oldval);
|
||||
+ if (val != oldval)
|
||||
{
|
||||
oldval = val;
|
||||
continue;
|
||||
diff --git a/nptl/pthread_mutex_timedlock.c b/nptl/pthread_mutex_timedlock.c
|
||||
index c4627ef6..a76c30b7 100644
|
||||
--- a/nptl/pthread_mutex_timedlock.c
|
||||
+++ b/nptl/pthread_mutex_timedlock.c
|
||||
@@ -269,10 +269,9 @@ __pthread_mutex_timedlock (pthread_mutex_t *mutex,
|
||||
meantime. */
|
||||
if ((oldval & FUTEX_WAITERS) == 0)
|
||||
{
|
||||
- int val;
|
||||
- if ((val = atomic_compare_and_exchange_val_acq
|
||||
- (&mutex->__data.__lock, oldval | FUTEX_WAITERS,
|
||||
- oldval)) != oldval)
|
||||
+ int val = atomic_compare_and_exchange_val_acq
|
||||
+ (&mutex->__data.__lock, oldval | FUTEX_WAITERS, oldval);
|
||||
+ if (val != oldval)
|
||||
{
|
||||
oldval = val;
|
||||
continue;
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,60 +0,0 @@
|
||||
From 4df1fa6ddc8925a75f3da644d5da3bb16eb33f02 Mon Sep 17 00:00:00 2001
|
||||
From: Noah Goldstein <goldstein.w.n@gmail.com>
|
||||
Date: Fri, 3 Dec 2021 15:29:25 -0800
|
||||
Subject: [PATCH] x86-64: Use notl in EVEX strcmp [BZ #28646]
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
Must use notl %edi here as lower bits are for CHAR comparisons
|
||||
potentially out of range thus can be 0 without indicating mismatch.
|
||||
This fixes BZ #28646.
|
||||
|
||||
Co-Authored-By: H.J. Lu <hjl.tools@gmail.com>
|
||||
---
|
||||
sysdeps/x86_64/multiarch/strcmp-evex.S | 14 ++++++++------
|
||||
1 file changed, 8 insertions(+), 6 deletions(-)
|
||||
|
||||
Conflicts:
|
||||
string/test-strcmp.c
|
||||
(new check omitted)
|
||||
|
||||
diff --git a/sysdeps/x86_64/multiarch/strcmp-evex.S b/sysdeps/x86_64/multiarch/strcmp-evex.S
|
||||
index 82f12ac8..6f5c4bf9 100644
|
||||
--- a/sysdeps/x86_64/multiarch/strcmp-evex.S
|
||||
+++ b/sysdeps/x86_64/multiarch/strcmp-evex.S
|
||||
@@ -656,12 +656,13 @@ L(loop_cross_page):
|
||||
in YMM3 and 32 bytes at VEC_SIZE(%rdx, %r10). */
|
||||
VPCMP $0, VEC_SIZE(%rdx, %r10), %YMM3, %k3{%k4}
|
||||
kmovd %k3, %edi
|
||||
+ /* Must use notl %edi here as lower bits are for CHAR
|
||||
+ comparisons potentially out of range thus can be 0 without
|
||||
+ indicating mismatch. */
|
||||
+ notl %edi
|
||||
# ifdef USE_AS_WCSCMP
|
||||
/* Don't use subl since it is the upper 8 bits of EDI below. */
|
||||
- notl %edi
|
||||
andl $0xff, %edi
|
||||
-# else
|
||||
- incl %edi
|
||||
# endif
|
||||
|
||||
# ifdef USE_AS_WCSCMP
|
||||
@@ -743,12 +744,13 @@ L(loop_cross_page_2_vec):
|
||||
in YMM1 and 32 bytes at (VEC_SIZE * 3)(%rdx, %r10). */
|
||||
VPCMP $0, (VEC_SIZE * 3)(%rdx, %r10), %YMM1, %k3{%k4}
|
||||
kmovd %k3, %edi
|
||||
+ /* Must use notl %edi here as lower bits are for CHAR
|
||||
+ comparisons potentially out of range thus can be 0 without
|
||||
+ indicating mismatch. */
|
||||
+ notl %edi
|
||||
# ifdef USE_AS_WCSCMP
|
||||
/* Don't use subl since it is the upper 8 bits of EDI below. */
|
||||
- notl %edi
|
||||
andl $0xff, %edi
|
||||
-# else
|
||||
- incl %edi
|
||||
# endif
|
||||
|
||||
# ifdef USE_AS_WCSCMP
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,153 +0,0 @@
|
||||
From c7c54f65b080affb87a1513dee449c8ad6143c8b Mon Sep 17 00:00:00 2001
|
||||
From: "H.J. Lu" <hjl.tools@gmail.com>
|
||||
Date: Mon, 21 Jan 2019 11:35:18 -0800
|
||||
Subject: [PATCH] x86-64 strncpy: Properly handle the length parameter [BZ#
|
||||
24097]
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
On x32, the size_t parameter may be passed in the lower 32 bits of a
|
||||
64-bit register with the non-zero upper 32 bits. The string/memory
|
||||
functions written in assembly can only use the lower 32 bits of a
|
||||
64-bit register as length or must clear the upper 32 bits before using
|
||||
the full 64-bit register for length.
|
||||
|
||||
This pach fixes strncpy for x32. Tested on x86-64 and x32. On x86-64,
|
||||
libc.so is the same with and withou the fix.
|
||||
|
||||
[BZ# 24097]
|
||||
CVE-2019-6488
|
||||
* sysdeps/x86_64/multiarch/strcpy-avx2.S: Use RDX_LP for length.
|
||||
* sysdeps/x86_64/multiarch/strcpy-sse2-unaligned.S: Likewise.
|
||||
* sysdeps/x86_64/multiarch/strcpy-ssse3.S: Likewise.
|
||||
* sysdeps/x86_64/x32/Makefile (tests): Add tst-size_t-strncpy.
|
||||
* sysdeps/x86_64/x32/tst-size_t-strncpy.c: New file.
|
||||
---
|
||||
.../x86_64/multiarch/strcpy-sse2-unaligned.S | 4 +-
|
||||
sysdeps/x86_64/multiarch/strcpy-ssse3.S | 6 +-
|
||||
sysdeps/x86_64/x32/Makefile | 2 +-
|
||||
sysdeps/x86_64/x32/tst-size_t-strncpy.c | 58 +++++++++++++++++++
|
||||
4 files changed, 64 insertions(+), 6 deletions(-)
|
||||
create mode 100644 sysdeps/x86_64/x32/tst-size_t-strncpy.c
|
||||
|
||||
Conflicts:
|
||||
ChangeLog
|
||||
(removed)
|
||||
sysdeps/x86_64/multiarch/strcpy-avx2.S
|
||||
(skipped, only needed for x32 arch)
|
||||
|
||||
diff --git a/sysdeps/x86_64/multiarch/strcpy-sse2-unaligned.S b/sysdeps/x86_64/multiarch/strcpy-sse2-unaligned.S
|
||||
index 72bf7e85..50aca22d 100644
|
||||
--- a/sysdeps/x86_64/multiarch/strcpy-sse2-unaligned.S
|
||||
+++ b/sysdeps/x86_64/multiarch/strcpy-sse2-unaligned.S
|
||||
@@ -40,8 +40,8 @@
|
||||
.text
|
||||
ENTRY (STRCPY)
|
||||
# ifdef USE_AS_STRNCPY
|
||||
- mov %rdx, %r8
|
||||
- test %r8, %r8
|
||||
+ mov %RDX_LP, %R8_LP
|
||||
+ test %R8_LP, %R8_LP
|
||||
jz L(ExitZero)
|
||||
# endif
|
||||
mov %rsi, %rcx
|
||||
diff --git a/sysdeps/x86_64/multiarch/strcpy-ssse3.S b/sysdeps/x86_64/multiarch/strcpy-ssse3.S
|
||||
index 9858d0c4..0a62814a 100644
|
||||
--- a/sysdeps/x86_64/multiarch/strcpy-ssse3.S
|
||||
+++ b/sysdeps/x86_64/multiarch/strcpy-ssse3.S
|
||||
@@ -31,13 +31,13 @@ ENTRY (STRCPY)
|
||||
|
||||
mov %rsi, %rcx
|
||||
# ifdef USE_AS_STRNCPY
|
||||
- mov %rdx, %r8
|
||||
+ mov %RDX_LP, %R8_LP
|
||||
# endif
|
||||
mov %rdi, %rdx
|
||||
# ifdef USE_AS_STRNCPY
|
||||
- test %r8, %r8
|
||||
+ test %R8_LP, %R8_LP
|
||||
jz L(Exit0)
|
||||
- cmp $8, %r8
|
||||
+ cmp $8, %R8_LP
|
||||
jbe L(StrncpyExit8Bytes)
|
||||
# endif
|
||||
cmpb $0, (%rcx)
|
||||
diff --git a/sysdeps/x86_64/x32/Makefile b/sysdeps/x86_64/x32/Makefile
|
||||
index db302839..2a9e20a9 100644
|
||||
--- a/sysdeps/x86_64/x32/Makefile
|
||||
+++ b/sysdeps/x86_64/x32/Makefile
|
||||
@@ -8,7 +8,7 @@ endif
|
||||
ifeq ($(subdir),string)
|
||||
tests += tst-size_t-memchr tst-size_t-memcmp tst-size_t-memcpy \
|
||||
tst-size_t-memrchr tst-size_t-memset tst-size_t-strncasecmp \
|
||||
- tst-size_t-strncmp
|
||||
+ tst-size_t-strncmp tst-size_t-strncpy
|
||||
endif
|
||||
|
||||
ifeq ($(subdir),wcsmbs)
|
||||
diff --git a/sysdeps/x86_64/x32/tst-size_t-strncpy.c b/sysdeps/x86_64/x32/tst-size_t-strncpy.c
|
||||
new file mode 100644
|
||||
index 00000000..4dec71e6
|
||||
--- /dev/null
|
||||
+++ b/sysdeps/x86_64/x32/tst-size_t-strncpy.c
|
||||
@@ -0,0 +1,58 @@
|
||||
+/* Test strncpy with size_t in the lower 32 bits of 64-bit register.
|
||||
+ Copyright (C) 2019 Free Software Foundation, Inc.
|
||||
+ This file is part of the GNU C Library.
|
||||
+
|
||||
+ The GNU C Library is free software; you can redistribute it and/or
|
||||
+ modify it under the terms of the GNU Lesser General Public
|
||||
+ License as published by the Free Software Foundation; either
|
||||
+ version 2.1 of the License, or (at your option) any later version.
|
||||
+
|
||||
+ The GNU C Library is distributed in the hope that it will be useful,
|
||||
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
||||
+ Lesser General Public License for more details.
|
||||
+
|
||||
+ You should have received a copy of the GNU Lesser General Public
|
||||
+ License along with the GNU C Library; if not, see
|
||||
+ <http://www.gnu.org/licenses/>. */
|
||||
+
|
||||
+#define TEST_NAME "strncpy"
|
||||
+#include "test-size_t.h"
|
||||
+
|
||||
+IMPL (strncpy, 1)
|
||||
+
|
||||
+typedef char *(*proto_t) (char *, const char*, size_t);
|
||||
+
|
||||
+static void *
|
||||
+__attribute__ ((noinline, noclone))
|
||||
+do_strncpy (parameter_t a, parameter_t b)
|
||||
+{
|
||||
+ return CALL (&b, a.p, b.p, a.len);
|
||||
+}
|
||||
+
|
||||
+static int
|
||||
+test_main (void)
|
||||
+{
|
||||
+ test_init ();
|
||||
+
|
||||
+ parameter_t dest = { { page_size }, buf1 };
|
||||
+ parameter_t src = { { 0 }, buf2 };
|
||||
+
|
||||
+ int ret = 0;
|
||||
+ FOR_EACH_IMPL (impl, 0)
|
||||
+ {
|
||||
+ src.fn = impl->fn;
|
||||
+ do_strncpy (dest, src);
|
||||
+ int res = strncmp (dest.p, src.p, dest.len);
|
||||
+ if (res)
|
||||
+ {
|
||||
+ error (0, 0, "Wrong result in function %s: %i != 0",
|
||||
+ impl->name, res);
|
||||
+ ret = 1;
|
||||
+ }
|
||||
+ }
|
||||
+
|
||||
+ return ret ? EXIT_FAILURE : EXIT_SUCCESS;
|
||||
+}
|
||||
+
|
||||
+#include <support/test-driver.c>
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,43 +0,0 @@
|
||||
From 6b8dbbd03ac88f169b65b5c7d7278576a11d2e44 Mon Sep 17 00:00:00 2001
|
||||
From: Jangwoong Kim <6812skiii@gmail.com>
|
||||
Date: Tue, 14 Dec 2021 21:30:51 +0900
|
||||
Subject: [PATCH] nptl: Effectively skip CAS in spinlock loop
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
The commit:
|
||||
"Add LLL_MUTEX_READ_LOCK [BZ #28537]"
|
||||
SHA1: d672a98a1af106bd68deb15576710cd61363f7a6
|
||||
|
||||
introduced LLL_MUTEX_READ_LOCK, to skip CAS in spinlock loop
|
||||
if atomic load fails. But, "continue" inside of do-while loop
|
||||
does not skip the evaluation of escape expression, thus CAS
|
||||
is not skipped.
|
||||
|
||||
Replace do-while with while and skip LLL_MUTEX_TRYLOCK if
|
||||
LLL_MUTEX_READ_LOCK fails.
|
||||
|
||||
Reviewed-by: H.J. Lu <hjl.tools@gmail.com>
|
||||
---
|
||||
nptl/pthread_mutex_lock.c | 5 ++---
|
||||
1 file changed, 2 insertions(+), 3 deletions(-)
|
||||
|
||||
diff --git a/nptl/pthread_mutex_lock.c b/nptl/pthread_mutex_lock.c
|
||||
index a633d95e..d96a9933 100644
|
||||
--- a/nptl/pthread_mutex_lock.c
|
||||
+++ b/nptl/pthread_mutex_lock.c
|
||||
@@ -141,10 +141,9 @@ __pthread_mutex_lock (pthread_mutex_t *mutex)
|
||||
break;
|
||||
}
|
||||
atomic_spin_nop ();
|
||||
- if (LLL_MUTEX_READ_LOCK (mutex) != 0)
|
||||
- continue;
|
||||
}
|
||||
- while (LLL_MUTEX_TRYLOCK (mutex) != 0);
|
||||
+ while (LLL_MUTEX_READ_LOCK (mutex) != 0
|
||||
+ || LLL_MUTEX_TRYLOCK (mutex) != 0);
|
||||
|
||||
mutex->__data.__spins += (cnt - mutex->__data.__spins) / 8;
|
||||
}
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,37 +0,0 @@
|
||||
From b98d0bbf747f39770e0caba7e984ce9f8f900330 Mon Sep 17 00:00:00 2001
|
||||
From: Noah Goldstein <goldstein.w.n@gmail.com>
|
||||
Date: Fri, 18 Feb 2022 17:00:25 -0600
|
||||
Subject: [PATCH] x86: Fix TEST_NAME to make it a string in tst-strncmp-rtm.c
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
Previously TEST_NAME was passing a function pointer. This didn't fail
|
||||
because of the -Wno-error flag (to allow for overflow sizes passed
|
||||
to strncmp/wcsncmp)
|
||||
|
||||
Reviewed-by: H.J. Lu <hjl.tools@gmail.com>
|
||||
---
|
||||
sysdeps/x86/tst-strncmp-rtm.c | 4 ++--
|
||||
1 file changed, 2 insertions(+), 2 deletions(-)
|
||||
|
||||
diff --git a/sysdeps/x86/tst-strncmp-rtm.c b/sysdeps/x86/tst-strncmp-rtm.c
|
||||
index 4e9f094f..aef9866c 100644
|
||||
--- a/sysdeps/x86/tst-strncmp-rtm.c
|
||||
+++ b/sysdeps/x86/tst-strncmp-rtm.c
|
||||
@@ -23,12 +23,12 @@
|
||||
# define CHAR wchar_t
|
||||
# define MEMSET wmemset
|
||||
# define STRNCMP wcsncmp
|
||||
-# define TEST_NAME wcsncmp
|
||||
+# define TEST_NAME "wcsncmp"
|
||||
#else /* !WIDE */
|
||||
# define CHAR char
|
||||
# define MEMSET memset
|
||||
# define STRNCMP strncmp
|
||||
-# define TEST_NAME strncmp
|
||||
+# define TEST_NAME "strncmp"
|
||||
#endif /* !WIDE */
|
||||
|
||||
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,33 +0,0 @@
|
||||
From c15efd011cea3d8f0494269eb539583215a1feed Mon Sep 17 00:00:00 2001
|
||||
From: "H.J. Lu" <hjl.tools@gmail.com>
|
||||
Date: Fri, 4 Feb 2022 11:09:10 -0800
|
||||
Subject: [PATCH] x86-64: Fix strcmp-avx2.S
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
Change "movl %edx, %rdx" to "movl %edx, %edx" in:
|
||||
|
||||
commit b77b06e0e296f1a2276c27a67e1d44f2cfa38d45
|
||||
Author: Noah Goldstein <goldstein.w.n@gmail.com>
|
||||
Date: Mon Jan 10 15:35:38 2022 -0600
|
||||
|
||||
x86: Optimize strcmp-avx2.S
|
||||
---
|
||||
sysdeps/x86_64/multiarch/strcmp-avx2.S | 2 +-
|
||||
1 file changed, 1 insertion(+), 1 deletion(-)
|
||||
|
||||
diff --git a/sysdeps/x86_64/multiarch/strcmp-avx2.S b/sysdeps/x86_64/multiarch/strcmp-avx2.S
|
||||
index 554ffe4c..04675aa4 100644
|
||||
--- a/sysdeps/x86_64/multiarch/strcmp-avx2.S
|
||||
+++ b/sysdeps/x86_64/multiarch/strcmp-avx2.S
|
||||
@@ -106,7 +106,7 @@ ENTRY(STRCMP)
|
||||
# ifdef USE_AS_STRNCMP
|
||||
# ifdef __ILP32__
|
||||
/* Clear the upper 32 bits. */
|
||||
- movl %edx, %rdx
|
||||
+ movl %edx, %edx
|
||||
# endif
|
||||
cmp $1, %RDX_LP
|
||||
/* Signed comparison intentional. We use this branch to also
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,33 +0,0 @@
|
||||
From 0e0199a9e02ebe42e2b36958964d63f03573c382 Mon Sep 17 00:00:00 2001
|
||||
From: "H.J. Lu" <hjl.tools@gmail.com>
|
||||
Date: Fri, 4 Feb 2022 11:11:08 -0800
|
||||
Subject: [PATCH] x86-64: Fix strcmp-evex.S
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
Change "movl %edx, %rdx" to "movl %edx, %edx" in:
|
||||
|
||||
commit 8418eb3ff4b781d31c4ed5dc6c0bd7356bc45db9
|
||||
Author: Noah Goldstein <goldstein.w.n@gmail.com>
|
||||
Date: Mon Jan 10 15:35:39 2022 -0600
|
||||
|
||||
x86: Optimize strcmp-evex.S
|
||||
---
|
||||
sysdeps/x86_64/multiarch/strcmp-evex.S | 2 +-
|
||||
1 file changed, 1 insertion(+), 1 deletion(-)
|
||||
|
||||
diff --git a/sysdeps/x86_64/multiarch/strcmp-evex.S b/sysdeps/x86_64/multiarch/strcmp-evex.S
|
||||
index 99d8409a..ed56af8e 100644
|
||||
--- a/sysdeps/x86_64/multiarch/strcmp-evex.S
|
||||
+++ b/sysdeps/x86_64/multiarch/strcmp-evex.S
|
||||
@@ -116,7 +116,7 @@ ENTRY(STRCMP)
|
||||
# ifdef USE_AS_STRNCMP
|
||||
# ifdef __ILP32__
|
||||
/* Clear the upper 32 bits. */
|
||||
- movl %edx, %rdx
|
||||
+ movl %edx, %edx
|
||||
# endif
|
||||
cmp $1, %RDX_LP
|
||||
/* Signed comparison intentional. We use this branch to also
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,40 +0,0 @@
|
||||
From 1b0c60f95bbe2eded80b2bb5be75c0e45b11cde1 Mon Sep 17 00:00:00 2001
|
||||
From: Noah Goldstein <goldstein.w.n@gmail.com>
|
||||
Date: Mon, 7 Feb 2022 00:32:23 -0600
|
||||
Subject: [PATCH] x86: Remove SSSE3 instruction for broadcast in memset.S (SSE2
|
||||
Only)
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
commit b62ace2740a106222e124cc86956448fa07abf4d
|
||||
Author: Noah Goldstein <goldstein.w.n@gmail.com>
|
||||
Date: Sun Feb 6 00:54:18 2022 -0600
|
||||
|
||||
x86: Improve vec generation in memset-vec-unaligned-erms.S
|
||||
|
||||
Revert usage of 'pshufb' in broadcast logic as it is an SSSE3
|
||||
instruction and memset.S is restricted to only SSE2 instructions.
|
||||
---
|
||||
sysdeps/x86_64/memset.S | 7 ++++---
|
||||
1 file changed, 4 insertions(+), 3 deletions(-)
|
||||
|
||||
diff --git a/sysdeps/x86_64/memset.S b/sysdeps/x86_64/memset.S
|
||||
index 27debd2b..4cb4aa71 100644
|
||||
--- a/sysdeps/x86_64/memset.S
|
||||
+++ b/sysdeps/x86_64/memset.S
|
||||
@@ -30,9 +30,10 @@
|
||||
|
||||
# define MEMSET_SET_VEC0_AND_SET_RETURN(d, r) \
|
||||
movd d, %xmm0; \
|
||||
- pxor %xmm1, %xmm1; \
|
||||
- pshufb %xmm1, %xmm0; \
|
||||
- movq r, %rax
|
||||
+ movq r, %rax; \
|
||||
+ punpcklbw %xmm0, %xmm0; \
|
||||
+ punpcklwd %xmm0, %xmm0; \
|
||||
+ pshufd $0, %xmm0, %xmm0
|
||||
|
||||
# define WMEMSET_SET_VEC0_AND_SET_RETURN(d, r) \
|
||||
movd d, %xmm0; \
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,218 +0,0 @@
|
||||
From 5165de69c0908e28a380cbd4bb054e55ea4abc95 Mon Sep 17 00:00:00 2001
|
||||
From: "H.J. Lu" <hjl.tools@gmail.com>
|
||||
Date: Mon, 21 Jan 2019 11:36:36 -0800
|
||||
Subject: [PATCH] x86-64 strnlen/wcsnlen: Properly handle the length parameter
|
||||
[BZ# 24097]
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
On x32, the size_t parameter may be passed in the lower 32 bits of a
|
||||
64-bit register with the non-zero upper 32 bits. The string/memory
|
||||
functions written in assembly can only use the lower 32 bits of a
|
||||
64-bit register as length or must clear the upper 32 bits before using
|
||||
the full 64-bit register for length.
|
||||
|
||||
This pach fixes strnlen/wcsnlen for x32. Tested on x86-64 and x32. On
|
||||
x86-64, libc.so is the same with and withou the fix.
|
||||
|
||||
[BZ# 24097]
|
||||
CVE-2019-6488
|
||||
* sysdeps/x86_64/multiarch/strlen-avx2.S: Use RSI_LP for length.
|
||||
Clear the upper 32 bits of RSI register.
|
||||
* sysdeps/x86_64/strlen.S: Use RSI_LP for length.
|
||||
* sysdeps/x86_64/x32/Makefile (tests): Add tst-size_t-strnlen
|
||||
and tst-size_t-wcsnlen.
|
||||
* sysdeps/x86_64/x32/tst-size_t-strnlen.c: New file.
|
||||
* sysdeps/x86_64/x32/tst-size_t-wcsnlen.c: Likewise.
|
||||
---
|
||||
sysdeps/x86_64/multiarch/strlen-avx2.S | 9 ++--
|
||||
sysdeps/x86_64/strlen.S | 12 ++---
|
||||
sysdeps/x86_64/x32/Makefile | 4 +-
|
||||
sysdeps/x86_64/x32/tst-size_t-strnlen.c | 72 +++++++++++++++++++++++++
|
||||
sysdeps/x86_64/x32/tst-size_t-wcsnlen.c | 20 +++++++
|
||||
5 files changed, 106 insertions(+), 11 deletions(-)
|
||||
create mode 100644 sysdeps/x86_64/x32/tst-size_t-strnlen.c
|
||||
create mode 100644 sysdeps/x86_64/x32/tst-size_t-wcsnlen.c
|
||||
|
||||
Conflicts:
|
||||
ChangeLog
|
||||
(removed)
|
||||
|
||||
diff --git a/sysdeps/x86_64/multiarch/strlen-avx2.S b/sysdeps/x86_64/multiarch/strlen-avx2.S
|
||||
index fb2418cd..645e0446 100644
|
||||
--- a/sysdeps/x86_64/multiarch/strlen-avx2.S
|
||||
+++ b/sysdeps/x86_64/multiarch/strlen-avx2.S
|
||||
@@ -42,12 +42,15 @@
|
||||
ENTRY (STRLEN)
|
||||
# ifdef USE_AS_STRNLEN
|
||||
/* Check for zero length. */
|
||||
- testq %rsi, %rsi
|
||||
+ test %RSI_LP, %RSI_LP
|
||||
jz L(zero)
|
||||
# ifdef USE_AS_WCSLEN
|
||||
- shl $2, %rsi
|
||||
+ shl $2, %RSI_LP
|
||||
+# elif defined __ILP32__
|
||||
+ /* Clear the upper 32 bits. */
|
||||
+ movl %esi, %esi
|
||||
# endif
|
||||
- movq %rsi, %r8
|
||||
+ mov %RSI_LP, %R8_LP
|
||||
# endif
|
||||
movl %edi, %ecx
|
||||
movq %rdi, %rdx
|
||||
diff --git a/sysdeps/x86_64/strlen.S b/sysdeps/x86_64/strlen.S
|
||||
index 01cb5fa8..f845f3d4 100644
|
||||
--- a/sysdeps/x86_64/strlen.S
|
||||
+++ b/sysdeps/x86_64/strlen.S
|
||||
@@ -59,21 +59,21 @@ ENTRY(strlen)
|
||||
|
||||
#ifdef AS_STRNLEN
|
||||
/* Do not read anything when n==0. */
|
||||
- test %rsi, %rsi
|
||||
+ test %RSI_LP, %RSI_LP
|
||||
jne L(n_nonzero)
|
||||
xor %rax, %rax
|
||||
ret
|
||||
L(n_nonzero):
|
||||
# ifdef AS_WCSLEN
|
||||
- shlq $2, %rsi
|
||||
+ shl $2, %RSI_LP
|
||||
# endif
|
||||
|
||||
/* Initialize long lived registers. */
|
||||
|
||||
- add %rdi, %rsi
|
||||
- mov %rsi, %r10
|
||||
- and $-64, %r10
|
||||
- mov %rsi, %r11
|
||||
+ add %RDI_LP, %RSI_LP
|
||||
+ mov %RSI_LP, %R10_LP
|
||||
+ and $-64, %R10_LP
|
||||
+ mov %RSI_LP, %R11_LP
|
||||
#endif
|
||||
|
||||
pxor %xmm0, %xmm0
|
||||
diff --git a/sysdeps/x86_64/x32/Makefile b/sysdeps/x86_64/x32/Makefile
|
||||
index 2a9e20a9..1557724b 100644
|
||||
--- a/sysdeps/x86_64/x32/Makefile
|
||||
+++ b/sysdeps/x86_64/x32/Makefile
|
||||
@@ -8,10 +8,10 @@ endif
|
||||
ifeq ($(subdir),string)
|
||||
tests += tst-size_t-memchr tst-size_t-memcmp tst-size_t-memcpy \
|
||||
tst-size_t-memrchr tst-size_t-memset tst-size_t-strncasecmp \
|
||||
- tst-size_t-strncmp tst-size_t-strncpy
|
||||
+ tst-size_t-strncmp tst-size_t-strncpy tst-size_t-strnlen
|
||||
endif
|
||||
|
||||
ifeq ($(subdir),wcsmbs)
|
||||
tests += tst-size_t-wmemchr tst-size_t-wmemcmp tst-size_t-wmemset \
|
||||
- tst-size_t-wcsncmp
|
||||
+ tst-size_t-wcsncmp tst-size_t-wcsnlen
|
||||
endif
|
||||
diff --git a/sysdeps/x86_64/x32/tst-size_t-strnlen.c b/sysdeps/x86_64/x32/tst-size_t-strnlen.c
|
||||
new file mode 100644
|
||||
index 00000000..690a4a8a
|
||||
--- /dev/null
|
||||
+++ b/sysdeps/x86_64/x32/tst-size_t-strnlen.c
|
||||
@@ -0,0 +1,72 @@
|
||||
+/* Test strnlen with size_t in the lower 32 bits of 64-bit register.
|
||||
+ Copyright (C) 2019 Free Software Foundation, Inc.
|
||||
+ This file is part of the GNU C Library.
|
||||
+
|
||||
+ The GNU C Library is free software; you can redistribute it and/or
|
||||
+ modify it under the terms of the GNU Lesser General Public
|
||||
+ License as published by the Free Software Foundation; either
|
||||
+ version 2.1 of the License, or (at your option) any later version.
|
||||
+
|
||||
+ The GNU C Library is distributed in the hope that it will be useful,
|
||||
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
||||
+ Lesser General Public License for more details.
|
||||
+
|
||||
+ You should have received a copy of the GNU Lesser General Public
|
||||
+ License along with the GNU C Library; if not, see
|
||||
+ <http://www.gnu.org/licenses/>. */
|
||||
+
|
||||
+#ifdef WIDE
|
||||
+# define TEST_NAME "wcsnlen"
|
||||
+#else
|
||||
+# define TEST_NAME "strnlen"
|
||||
+#endif /* WIDE */
|
||||
+
|
||||
+#include "test-size_t.h"
|
||||
+
|
||||
+#ifdef WIDE
|
||||
+# include <wchar.h>
|
||||
+# define STRNLEN wcsnlen
|
||||
+# define CHAR wchar_t
|
||||
+#else
|
||||
+# define STRNLEN strnlen
|
||||
+# define CHAR char
|
||||
+#endif /* WIDE */
|
||||
+
|
||||
+IMPL (STRNLEN, 1)
|
||||
+
|
||||
+typedef size_t (*proto_t) (const CHAR *, size_t);
|
||||
+
|
||||
+static size_t
|
||||
+__attribute__ ((noinline, noclone))
|
||||
+do_strnlen (parameter_t a, parameter_t b)
|
||||
+{
|
||||
+ return CALL (&a, a.p, b.len);
|
||||
+}
|
||||
+
|
||||
+static int
|
||||
+test_main (void)
|
||||
+{
|
||||
+ test_init ();
|
||||
+
|
||||
+ size_t size = page_size / sizeof (CHAR);
|
||||
+ parameter_t src = { { 0 }, buf2 };
|
||||
+ parameter_t c = { { size }, (void *) (uintptr_t) 'a' };
|
||||
+
|
||||
+ int ret = 0;
|
||||
+ FOR_EACH_IMPL (impl, 0)
|
||||
+ {
|
||||
+ src.fn = impl->fn;
|
||||
+ size_t res = do_strnlen (src, c);
|
||||
+ if (res != size)
|
||||
+ {
|
||||
+ error (0, 0, "Wrong result in function %s: 0x%x != 0x%x",
|
||||
+ impl->name, res, size);
|
||||
+ ret = 1;
|
||||
+ }
|
||||
+ }
|
||||
+
|
||||
+ return ret ? EXIT_FAILURE : EXIT_SUCCESS;
|
||||
+}
|
||||
+
|
||||
+#include <support/test-driver.c>
|
||||
diff --git a/sysdeps/x86_64/x32/tst-size_t-wcsnlen.c b/sysdeps/x86_64/x32/tst-size_t-wcsnlen.c
|
||||
new file mode 100644
|
||||
index 00000000..093b4bbe
|
||||
--- /dev/null
|
||||
+++ b/sysdeps/x86_64/x32/tst-size_t-wcsnlen.c
|
||||
@@ -0,0 +1,20 @@
|
||||
+/* Test wcsnlen with size_t in the lower 32 bits of 64-bit register.
|
||||
+ Copyright (C) 2019 Free Software Foundation, Inc.
|
||||
+ This file is part of the GNU C Library.
|
||||
+
|
||||
+ The GNU C Library is free software; you can redistribute it and/or
|
||||
+ modify it under the terms of the GNU Lesser General Public
|
||||
+ License as published by the Free Software Foundation; either
|
||||
+ version 2.1 of the License, or (at your option) any later version.
|
||||
+
|
||||
+ The GNU C Library is distributed in the hope that it will be useful,
|
||||
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
||||
+ Lesser General Public License for more details.
|
||||
+
|
||||
+ You should have received a copy of the GNU Lesser General Public
|
||||
+ License along with the GNU C Library; if not, see
|
||||
+ <http://www.gnu.org/licenses/>. */
|
||||
+
|
||||
+#define WIDE 1
|
||||
+#include "tst-size_t-strnlen.c"
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,33 +0,0 @@
|
||||
From 7912236f4a597deb092650ca79f33504ddb4af28 Mon Sep 17 00:00:00 2001
|
||||
From: Noah Goldstein <goldstein.w.n@gmail.com>
|
||||
Date: Sat, 12 Feb 2022 00:45:00 -0600
|
||||
Subject: [PATCH] x86: Set .text section in memset-vec-unaligned-erms
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
commit 3d9f171bfb5325bd5f427e9fc386453358c6e840
|
||||
Author: H.J. Lu <hjl.tools@gmail.com>
|
||||
Date: Mon Feb 7 05:55:15 2022 -0800
|
||||
|
||||
x86-64: Optimize bzero
|
||||
|
||||
Remove setting the .text section for the code. This commit
|
||||
adds that back.
|
||||
---
|
||||
sysdeps/x86_64/multiarch/memset-vec-unaligned-erms.S | 1 +
|
||||
1 file changed, 1 insertion(+)
|
||||
|
||||
diff --git a/sysdeps/x86_64/multiarch/memset-vec-unaligned-erms.S b/sysdeps/x86_64/multiarch/memset-vec-unaligned-erms.S
|
||||
index 06f5f5d7..4fb475c0 100644
|
||||
--- a/sysdeps/x86_64/multiarch/memset-vec-unaligned-erms.S
|
||||
+++ b/sysdeps/x86_64/multiarch/memset-vec-unaligned-erms.S
|
||||
@@ -114,6 +114,7 @@
|
||||
# error SECTION is not defined!
|
||||
#endif
|
||||
|
||||
+ .section SECTION(.text), "ax", @progbits
|
||||
#if IS_IN (libc)
|
||||
# if defined SHARED
|
||||
ENTRY_CHK (WMEMSET_CHK_SYMBOL (__wmemset_chk, unaligned))
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,36 +0,0 @@
|
||||
From 0fb8800029d230b3711bf722b2a47db92d0e273f Mon Sep 17 00:00:00 2001
|
||||
From: "H.J. Lu" <hjl.tools@gmail.com>
|
||||
Date: Thu, 10 Feb 2022 11:52:50 -0800
|
||||
Subject: [PATCH] x86-64: Remove bzero weak alias in SS2 memset
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
commit 3d9f171bfb5325bd5f427e9fc386453358c6e840
|
||||
Author: H.J. Lu <hjl.tools@gmail.com>
|
||||
Date: Mon Feb 7 05:55:15 2022 -0800
|
||||
|
||||
x86-64: Optimize bzero
|
||||
|
||||
added the optimized bzero. Remove bzero weak alias in SS2 memset to
|
||||
avoid undefined __bzero in memset-sse2-unaligned-erms.
|
||||
---
|
||||
sysdeps/x86_64/multiarch/memset-sse2-unaligned-erms.S | 4 +---
|
||||
1 file changed, 1 insertion(+), 3 deletions(-)
|
||||
|
||||
diff --git a/sysdeps/x86_64/multiarch/memset-sse2-unaligned-erms.S b/sysdeps/x86_64/multiarch/memset-sse2-unaligned-erms.S
|
||||
index 8f579ad6..af51362b 100644
|
||||
--- a/sysdeps/x86_64/multiarch/memset-sse2-unaligned-erms.S
|
||||
+++ b/sysdeps/x86_64/multiarch/memset-sse2-unaligned-erms.S
|
||||
@@ -31,9 +31,7 @@
|
||||
# endif
|
||||
|
||||
# undef weak_alias
|
||||
-# define weak_alias(original, alias) \
|
||||
- .weak bzero; bzero = __bzero
|
||||
-
|
||||
+# define weak_alias(original, alias)
|
||||
# undef strong_alias
|
||||
# define strong_alias(ignored1, ignored2)
|
||||
#endif
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,29 +0,0 @@
|
||||
From bf92893a14ebc161b08b28acc24fa06ae6be19cb Mon Sep 17 00:00:00 2001
|
||||
From: Adhemerval Zanella <adhemerval.zanella@linaro.org>
|
||||
Date: Thu, 10 Feb 2022 11:23:24 -0300
|
||||
Subject: [PATCH] x86_64: Remove bcopy optimizations
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
The symbols is not present in current POSIX specification and compiler
|
||||
already generates memmove call.
|
||||
---
|
||||
sysdeps/x86_64/multiarch/bcopy.S | 7 -------
|
||||
1 file changed, 7 deletions(-)
|
||||
delete mode 100644 sysdeps/x86_64/multiarch/bcopy.S
|
||||
|
||||
diff --git a/sysdeps/x86_64/multiarch/bcopy.S b/sysdeps/x86_64/multiarch/bcopy.S
|
||||
deleted file mode 100644
|
||||
index 639f02bd..00000000
|
||||
--- a/sysdeps/x86_64/multiarch/bcopy.S
|
||||
+++ /dev/null
|
||||
@@ -1,7 +0,0 @@
|
||||
-#include <sysdep.h>
|
||||
-
|
||||
- .text
|
||||
-ENTRY(bcopy)
|
||||
- xchg %rdi, %rsi
|
||||
- jmp __libc_memmove /* Branch to IFUNC memmove. */
|
||||
-END(bcopy)
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,206 +0,0 @@
|
||||
From 3f635fb43389b54f682fc9ed2acc0b2aaf4a923d Mon Sep 17 00:00:00 2001
|
||||
From: "H.J. Lu" <hjl.tools@gmail.com>
|
||||
Date: Mon, 4 Feb 2019 06:31:01 -0800
|
||||
Subject: [PATCH] x86-64 memcmp: Use unsigned Jcc instructions on size [BZ
|
||||
#24155]
|
||||
Content-type: text/plain; charset=UTF-8
|
||||
|
||||
Since the size argument is unsigned. we should use unsigned Jcc
|
||||
instructions, instead of signed, to check size.
|
||||
|
||||
Tested on x86-64 and x32, with and without --disable-multi-arch.
|
||||
|
||||
[BZ #24155]
|
||||
CVE-2019-7309
|
||||
* NEWS: Updated for CVE-2019-7309.
|
||||
* sysdeps/x86_64/memcmp.S: Use RDX_LP for size. Clear the
|
||||
upper 32 bits of RDX register for x32. Use unsigned Jcc
|
||||
instructions, instead of signed.
|
||||
* sysdeps/x86_64/x32/Makefile (tests): Add tst-size_t-memcmp-2.
|
||||
* sysdeps/x86_64/x32/tst-size_t-memcmp-2.c: New test.
|
||||
---
|
||||
sysdeps/x86_64/memcmp.S | 20 +++---
|
||||
sysdeps/x86_64/x32/Makefile | 3 +-
|
||||
sysdeps/x86_64/x32/tst-size_t-memcmp-2.c | 79 ++++++++++++++++++++++++
|
||||
3 files changed, 93 insertions(+), 9 deletions(-)
|
||||
create mode 100644 sysdeps/x86_64/x32/tst-size_t-memcmp-2.c
|
||||
|
||||
Conflics:
|
||||
ChangeLog
|
||||
(removed)
|
||||
NEWS
|
||||
(removed)
|
||||
|
||||
diff --git a/sysdeps/x86_64/memcmp.S b/sysdeps/x86_64/memcmp.S
|
||||
index bcb4a2e8..45918d37 100644
|
||||
--- a/sysdeps/x86_64/memcmp.S
|
||||
+++ b/sysdeps/x86_64/memcmp.S
|
||||
@@ -21,14 +21,18 @@
|
||||
|
||||
.text
|
||||
ENTRY (memcmp)
|
||||
- test %rdx, %rdx
|
||||
+#ifdef __ILP32__
|
||||
+ /* Clear the upper 32 bits. */
|
||||
+ movl %edx, %edx
|
||||
+#endif
|
||||
+ test %RDX_LP, %RDX_LP
|
||||
jz L(finz)
|
||||
cmpq $1, %rdx
|
||||
- jle L(finr1b)
|
||||
+ jbe L(finr1b)
|
||||
subq %rdi, %rsi
|
||||
movq %rdx, %r10
|
||||
cmpq $32, %r10
|
||||
- jge L(gt32)
|
||||
+ jae L(gt32)
|
||||
/* Handle small chunks and last block of less than 32 bytes. */
|
||||
L(small):
|
||||
testq $1, %r10
|
||||
@@ -156,7 +160,7 @@ L(A32):
|
||||
movq %r11, %r10
|
||||
andq $-32, %r10
|
||||
cmpq %r10, %rdi
|
||||
- jge L(mt16)
|
||||
+ jae L(mt16)
|
||||
/* Pre-unroll to be ready for unrolled 64B loop. */
|
||||
testq $32, %rdi
|
||||
jz L(A64)
|
||||
@@ -178,7 +182,7 @@ L(A64):
|
||||
movq %r11, %r10
|
||||
andq $-64, %r10
|
||||
cmpq %r10, %rdi
|
||||
- jge L(mt32)
|
||||
+ jae L(mt32)
|
||||
|
||||
L(A64main):
|
||||
movdqu (%rdi,%rsi), %xmm0
|
||||
@@ -216,7 +220,7 @@ L(mt32):
|
||||
movq %r11, %r10
|
||||
andq $-32, %r10
|
||||
cmpq %r10, %rdi
|
||||
- jge L(mt16)
|
||||
+ jae L(mt16)
|
||||
|
||||
L(A32main):
|
||||
movdqu (%rdi,%rsi), %xmm0
|
||||
@@ -254,7 +258,7 @@ L(ATR):
|
||||
movq %r11, %r10
|
||||
andq $-32, %r10
|
||||
cmpq %r10, %rdi
|
||||
- jge L(mt16)
|
||||
+ jae L(mt16)
|
||||
testq $16, %rdi
|
||||
jz L(ATR32)
|
||||
|
||||
@@ -325,7 +329,7 @@ L(ATR64main):
|
||||
movq %r11, %r10
|
||||
andq $-32, %r10
|
||||
cmpq %r10, %rdi
|
||||
- jge L(mt16)
|
||||
+ jae L(mt16)
|
||||
|
||||
L(ATR32res):
|
||||
movdqa (%rdi,%rsi), %xmm0
|
||||
diff --git a/sysdeps/x86_64/x32/Makefile b/sysdeps/x86_64/x32/Makefile
|
||||
index 1557724b..87489565 100644
|
||||
--- a/sysdeps/x86_64/x32/Makefile
|
||||
+++ b/sysdeps/x86_64/x32/Makefile
|
||||
@@ -8,7 +8,8 @@ endif
|
||||
ifeq ($(subdir),string)
|
||||
tests += tst-size_t-memchr tst-size_t-memcmp tst-size_t-memcpy \
|
||||
tst-size_t-memrchr tst-size_t-memset tst-size_t-strncasecmp \
|
||||
- tst-size_t-strncmp tst-size_t-strncpy tst-size_t-strnlen
|
||||
+ tst-size_t-strncmp tst-size_t-strncpy tst-size_t-strnlen \
|
||||
+ tst-size_t-memcmp-2
|
||||
endif
|
||||
|
||||
ifeq ($(subdir),wcsmbs)
|
||||
diff --git a/sysdeps/x86_64/x32/tst-size_t-memcmp-2.c b/sysdeps/x86_64/x32/tst-size_t-memcmp-2.c
|
||||
new file mode 100644
|
||||
index 00000000..d8ae1a08
|
||||
--- /dev/null
|
||||
+++ b/sysdeps/x86_64/x32/tst-size_t-memcmp-2.c
|
||||
@@ -0,0 +1,79 @@
|
||||
+/* Test memcmp with size_t in the lower 32 bits of 64-bit register.
|
||||
+ Copyright (C) 2019 Free Software Foundation, Inc.
|
||||
+ This file is part of the GNU C Library.
|
||||
+
|
||||
+ The GNU C Library is free software; you can redistribute it and/or
|
||||
+ modify it under the terms of the GNU Lesser General Public
|
||||
+ License as published by the Free Software Foundation; either
|
||||
+ version 2.1 of the License, or (at your option) any later version.
|
||||
+
|
||||
+ The GNU C Library is distributed in the hope that it will be useful,
|
||||
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
||||
+ Lesser General Public License for more details.
|
||||
+
|
||||
+ You should have received a copy of the GNU Lesser General Public
|
||||
+ License along with the GNU C Library; if not, see
|
||||
+ <http://www.gnu.org/licenses/>. */
|
||||
+
|
||||
+#define TEST_MAIN
|
||||
+#ifdef WIDE
|
||||
+# define TEST_NAME "wmemcmp"
|
||||
+#else
|
||||
+# define TEST_NAME "memcmp"
|
||||
+#endif
|
||||
+
|
||||
+#include "test-size_t.h"
|
||||
+
|
||||
+#ifdef WIDE
|
||||
+# include <inttypes.h>
|
||||
+# include <wchar.h>
|
||||
+
|
||||
+# define MEMCMP wmemcmp
|
||||
+# define CHAR wchar_t
|
||||
+#else
|
||||
+# define MEMCMP memcmp
|
||||
+# define CHAR char
|
||||
+#endif
|
||||
+
|
||||
+IMPL (MEMCMP, 1)
|
||||
+
|
||||
+typedef int (*proto_t) (const CHAR *, const CHAR *, size_t);
|
||||
+
|
||||
+static int
|
||||
+__attribute__ ((noinline, noclone))
|
||||
+do_memcmp (parameter_t a, parameter_t b)
|
||||
+{
|
||||
+ return CALL (&b, a.p, b.p, a.len);
|
||||
+}
|
||||
+
|
||||
+static int
|
||||
+test_main (void)
|
||||
+{
|
||||
+ test_init ();
|
||||
+
|
||||
+ parameter_t dest = { { page_size / sizeof (CHAR) }, buf1 };
|
||||
+ parameter_t src = { { 0 }, buf2 };
|
||||
+
|
||||
+ memcpy (buf1, buf2, page_size);
|
||||
+
|
||||
+ CHAR *p = (CHAR *) buf1;
|
||||
+ p[page_size / sizeof (CHAR) - 1] = (CHAR) 1;
|
||||
+
|
||||
+ int ret = 0;
|
||||
+ FOR_EACH_IMPL (impl, 0)
|
||||
+ {
|
||||
+ src.fn = impl->fn;
|
||||
+ int res = do_memcmp (dest, src);
|
||||
+ if (res >= 0)
|
||||
+ {
|
||||
+ error (0, 0, "Wrong result in function %s: %i >= 0",
|
||||
+ impl->name, res);
|
||||
+ ret = 1;
|
||||
+ }
|
||||
+ }
|
||||
+
|
||||
+ return ret ? EXIT_FAILURE : EXIT_SUCCESS;
|
||||
+}
|
||||
+
|
||||
+#include <support/test-driver.c>
|
||||
--
|
||||
GitLab
|
||||
|
||||
@ -1,31 +0,0 @@
|
||||
Based on the following commit, adjusted for glibc-2.28 in RHEL-8:
|
||||
|
||||
commit 5eabdb6a6ac1599d23dd5966a37417215950245f
|
||||
Author: Andreas Schwab <schwab@suse.de>
|
||||
Date: Wed Dec 6 14:48:22 2023 +0100
|
||||
|
||||
getaddrinfo: translate ENOMEM to EAI_MEMORY (bug 31163)
|
||||
|
||||
When __resolv_context_get returns NULL due to out of memory, translate it
|
||||
to a return value of EAI_MEMORY.
|
||||
|
||||
diff --git a/sysdeps/posix/getaddrinfo.c b/sysdeps/posix/getaddrinfo.c
|
||||
index 46046504a6858f2e..d0708f3e84e20025 100644
|
||||
--- a/sysdeps/posix/getaddrinfo.c
|
||||
+++ b/sysdeps/posix/getaddrinfo.c
|
||||
@@ -777,7 +777,14 @@ gaih_inet (const char *name, const struct gaih_service *service,
|
||||
res_ctx = __resolv_context_get ();
|
||||
res_enable_inet6 = __resolv_context_disable_inet6 (res_ctx);
|
||||
if (res_ctx == NULL)
|
||||
- no_more = 1;
|
||||
+ {
|
||||
+ if (errno == ENOMEM)
|
||||
+ {
|
||||
+ result = -EAI_MEMORY;
|
||||
+ goto free_and_return;
|
||||
+ }
|
||||
+ no_more = 1;
|
||||
+ }
|
||||
|
||||
while (!no_more)
|
||||
{
|
||||
@ -1,112 +0,0 @@
|
||||
This downstream-only patch compensates for the missing backport of
|
||||
commit 2d651eb9265d1366d7b9e881bfddd46db9c1ecc4 ("x86: Move
|
||||
x86 processor cache info to cpu_features"). Without it,
|
||||
ld.so --list-diagnostics prints values that have not been properly
|
||||
initalized from CPUID data.
|
||||
|
||||
diff --git a/sysdeps/x86/cacheinfo.h b/sysdeps/x86/cacheinfo.h
|
||||
index 10ebadd819d9efff..d8421fab83ab08ac 100644
|
||||
--- a/sysdeps/x86/cacheinfo.h
|
||||
+++ b/sysdeps/x86/cacheinfo.h
|
||||
@@ -19,31 +19,42 @@
|
||||
#include <assert.h>
|
||||
#include <unistd.h>
|
||||
|
||||
+/* When building ld.so, do not export any of the variables. They are
|
||||
+ only used for diagnostics and are not initialized during regular
|
||||
+ operation. */
|
||||
+#if IS_IN (rtld)
|
||||
+# define CACHEINFO_VARIABLE(name, initializer) \
|
||||
+ static long int name = initializer
|
||||
+#else
|
||||
+# define CACHEINFO_VARIABLE(name, initializer) \
|
||||
+ long int name attribute_hidden = initializer
|
||||
+#endif
|
||||
+
|
||||
/* Data cache size for use in memory and string routines, typically
|
||||
L1 size, rounded to multiple of 256 bytes. */
|
||||
-long int __x86_data_cache_size_half attribute_hidden = 32 * 1024 / 2;
|
||||
-long int __x86_data_cache_size attribute_hidden = 32 * 1024;
|
||||
+CACHEINFO_VARIABLE (__x86_data_cache_size_half, 32 * 1024 / 2);
|
||||
+CACHEINFO_VARIABLE (__x86_data_cache_size, 32 * 1024);
|
||||
/* Similar to __x86_data_cache_size_half, but not rounded. */
|
||||
-long int __x86_raw_data_cache_size_half attribute_hidden = 32 * 1024 / 2;
|
||||
+CACHEINFO_VARIABLE (__x86_raw_data_cache_size_half, 32 * 1024 / 2);
|
||||
/* Similar to __x86_data_cache_size, but not rounded. */
|
||||
-long int __x86_raw_data_cache_size attribute_hidden = 32 * 1024;
|
||||
+CACHEINFO_VARIABLE (__x86_raw_data_cache_size, 32 * 1024);
|
||||
/* Shared cache size for use in memory and string routines, typically
|
||||
L2 or L3 size, rounded to multiple of 256 bytes. */
|
||||
-long int __x86_shared_cache_size_half attribute_hidden = 1024 * 1024 / 2;
|
||||
-long int __x86_shared_cache_size attribute_hidden = 1024 * 1024;
|
||||
+CACHEINFO_VARIABLE (__x86_shared_cache_size_half, 1024 * 1024 / 2);
|
||||
+CACHEINFO_VARIABLE (__x86_shared_cache_size, 1024 * 1024);
|
||||
/* Similar to __x86_shared_cache_size_half, but not rounded. */
|
||||
-long int __x86_raw_shared_cache_size_half attribute_hidden = 1024 * 1024 / 2;
|
||||
+CACHEINFO_VARIABLE (__x86_raw_shared_cache_size_half, 1024 * 1024 / 2);
|
||||
/* Similar to __x86_shared_cache_size, but not rounded. */
|
||||
-long int __x86_raw_shared_cache_size attribute_hidden = 1024 * 1024;
|
||||
+CACHEINFO_VARIABLE (__x86_raw_shared_cache_size, 1024 * 1024);
|
||||
|
||||
/* Threshold to use non temporal store. */
|
||||
-long int __x86_shared_non_temporal_threshold attribute_hidden;
|
||||
+CACHEINFO_VARIABLE (__x86_shared_non_temporal_threshold, 0);
|
||||
|
||||
/* Threshold to use Enhanced REP MOVSB. */
|
||||
-long int __x86_rep_movsb_threshold attribute_hidden = 2048;
|
||||
+CACHEINFO_VARIABLE (__x86_rep_movsb_threshold, 2048);
|
||||
|
||||
/* Threshold to use Enhanced REP STOSB. */
|
||||
-long int __x86_rep_stosb_threshold attribute_hidden = 2048;
|
||||
+CACHEINFO_VARIABLE (__x86_rep_stosb_threshold, 2048);
|
||||
|
||||
static void
|
||||
get_common_cache_info (long int *shared_ptr, long int * shared_per_thread_ptr, unsigned int *threads_ptr,
|
||||
diff --git a/sysdeps/x86/dl-diagnostics-cpu.c b/sysdeps/x86/dl-diagnostics-cpu.c
|
||||
index 0ba286a828b69937..9215604ecf22344c 100644
|
||||
--- a/sysdeps/x86/dl-diagnostics-cpu.c
|
||||
+++ b/sysdeps/x86/dl-diagnostics-cpu.c
|
||||
@@ -19,6 +19,13 @@
|
||||
#include <dl-diagnostics.h>
|
||||
#include <ldsodefs.h>
|
||||
|
||||
+#include <assert.h>
|
||||
+#include <unistd.h>
|
||||
+#include <cpu-features.h>
|
||||
+#include <cpuid.h>
|
||||
+#include <dl-cacheinfo.h>
|
||||
+#include <cacheinfo.h>
|
||||
+
|
||||
static void
|
||||
print_cpu_features_value (const char *label, uint64_t value)
|
||||
{
|
||||
@@ -81,19 +88,21 @@ _dl_diagnostics_cpu (void)
|
||||
#include "cpu-features-preferred_feature_index_1.def"
|
||||
#undef BIT
|
||||
|
||||
+ /* The cache information variables are only used for diagnostics and
|
||||
+ are not initialized during startup. The values used at run time
|
||||
+ are only in libc.so.6. */
|
||||
+ init_cacheinfo ();
|
||||
+
|
||||
print_cpu_features_value ("xsave_state_size",
|
||||
cpu_features->xsave_state_size);
|
||||
print_cpu_features_value ("xsave_state_full_size",
|
||||
cpu_features->xsave_state_full_size);
|
||||
- print_cpu_features_value ("data_cache_size", cpu_features->data_cache_size);
|
||||
- print_cpu_features_value ("shared_cache_size",
|
||||
- cpu_features->shared_cache_size);
|
||||
+ print_cpu_features_value ("data_cache_size", __x86_data_cache_size);
|
||||
+ print_cpu_features_value ("shared_cache_size", __x86_shared_cache_size);
|
||||
print_cpu_features_value ("non_temporal_threshold",
|
||||
- cpu_features->non_temporal_threshold);
|
||||
- print_cpu_features_value ("rep_movsb_threshold",
|
||||
- cpu_features->rep_movsb_threshold);
|
||||
- print_cpu_features_value ("rep_stosb_threshold",
|
||||
- cpu_features->rep_stosb_threshold);
|
||||
+ __x86_shared_non_temporal_threshold);
|
||||
+ print_cpu_features_value ("rep_movsb_threshold", __x86_rep_movsb_threshold);
|
||||
+ print_cpu_features_value ("rep_stosb_threshold", __x86_rep_stosb_threshold);
|
||||
_Static_assert (offsetof (struct cpu_features, rep_stosb_threshold)
|
||||
+ sizeof (cpu_features->rep_stosb_threshold)
|
||||
== sizeof (*cpu_features),
|
||||
@ -1,22 +0,0 @@
|
||||
Work around in the test case, the fact that RHEL-8 NSS modules
|
||||
infrastructure incorrectly allows merging in the hosts database. This
|
||||
is a RHEL-8 only fix.
|
||||
|
||||
diff --git a/nss/tst-nss-gai-actions.c b/nss/tst-nss-gai-actions.c
|
||||
index efca6cd1837a172a..c35e752896eceb2a 100644
|
||||
--- a/nss/tst-nss-gai-actions.c
|
||||
+++ b/nss/tst-nss-gai-actions.c
|
||||
@@ -87,6 +87,13 @@ do_one_test (int action, int family, bool canon)
|
||||
case ACTION_MERGE:
|
||||
if (ret == 0)
|
||||
{
|
||||
+ if (hints.ai_flags == 0 && hints.ai_family == AF_INET)
|
||||
+ {
|
||||
+ printf ("***** RHEL-8 limitation: "
|
||||
+ "NSS modules infrastructure incorrectly allows MERGE\n");
|
||||
+ return;
|
||||
+ }
|
||||
+
|
||||
char *formatted = support_format_addrinfo (ai, ret);
|
||||
|
||||
printf ("merge unexpectedly succeeded:\n %s\n", formatted);
|
||||
@ -1,97 +0,0 @@
|
||||
commit 2ac579f9c25388a7734948d77b03e4dd10f35334
|
||||
Author: DJ Delorie <dj@redhat.com>
|
||||
Date: Mon Sep 30 16:04:52 2019 -0400
|
||||
|
||||
Add run-one-test convenience target and makefile help text
|
||||
|
||||
Adds "make test" for re-running just one test. Also adds
|
||||
"make help" for help with our Makefile targets, and adds a
|
||||
mini-help when you just run "make".
|
||||
|
||||
Reviewed-by: Carlos O'Donell <carlos@redhat.com>
|
||||
|
||||
diff --git a/Makefile b/Makefile
|
||||
index 6d73241bbc811c13..6518f62ee0676b0d 100644
|
||||
--- a/Makefile
|
||||
+++ b/Makefile
|
||||
@@ -26,8 +26,17 @@ include Makeconfig
|
||||
|
||||
|
||||
# This is the default target; it makes everything except the tests.
|
||||
-.PHONY: all
|
||||
-all: lib others
|
||||
+.PHONY: all help minihelp
|
||||
+all: minihelp lib others
|
||||
+
|
||||
+help:
|
||||
+ @sed '0,/^help-starts-here$$/d' Makefile.help
|
||||
+
|
||||
+minihelp:
|
||||
+ @echo
|
||||
+ @echo type \"make help\" for help with common glibc makefile targets
|
||||
+ @echo
|
||||
+
|
||||
|
||||
ifneq ($(AUTOCONF),no)
|
||||
|
||||
@@ -503,3 +512,12 @@ FORCE:
|
||||
|
||||
iconvdata/% localedata/% po/%: FORCE
|
||||
$(MAKE) $(PARALLELMFLAGS) -C $(@D) $(@F)
|
||||
+
|
||||
+# Convenience target to rerun one test, from the top of the build tree
|
||||
+# Example: make test t=wcsmbs/test-wcsnlen
|
||||
+.PHONY: test
|
||||
+test :
|
||||
+ @-rm -f $(objpfx)$t.out
|
||||
+ $(MAKE) subdir=$(dir $t) -C $(dir $t) ..=../ $(objpfx)$t.out
|
||||
+ @cat $(objpfx)$t.test-result
|
||||
+ @cat $(objpfx)$t.out
|
||||
diff --git a/Makefile.help b/Makefile.help
|
||||
new file mode 100644
|
||||
index 0000000000000000..3b043bce013cc2b4
|
||||
--- /dev/null
|
||||
+++ b/Makefile.help
|
||||
@@ -0,0 +1,42 @@
|
||||
+# Copyright (C) 2019 Free Software Foundation, Inc.
|
||||
+# This file is part of the GNU C Library.
|
||||
+
|
||||
+# The GNU C Library is free software; you can redistribute it and/or
|
||||
+# modify it under the terms of the GNU Lesser General Public
|
||||
+# License as published by the Free Software Foundation; either
|
||||
+# version 2.1 of the License, or (at your option) any later version.
|
||||
+
|
||||
+# The GNU C Library is distributed in the hope that it will be useful,
|
||||
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
||||
+# Lesser General Public License for more details.
|
||||
+
|
||||
+# You should have received a copy of the GNU Lesser General Public
|
||||
+# License along with the GNU C Library; if not, see
|
||||
+# <https://www.gnu.org/licenses/>.
|
||||
+
|
||||
+This is the file that gets printed when the user runs "make help",
|
||||
+starting just after the "help-starts-here" line.
|
||||
+
|
||||
+help-starts-here
|
||||
+
|
||||
+all
|
||||
+ The usual default; builds everything but doesn't run the
|
||||
+ tests.
|
||||
+
|
||||
+check (or tests)
|
||||
+ Runs the standard set of tests.
|
||||
+
|
||||
+test
|
||||
+ Runs one test. Use like this:
|
||||
+ make test t=wcsmbs/test-wcsnlen
|
||||
+ Note that this will rebuild the test if needed, but will not
|
||||
+ rebuild what "make all" would have rebuilt.
|
||||
+
|
||||
+--
|
||||
+Other useful hints:
|
||||
+
|
||||
+builddir$ rm testroot.pristine/install.stamp
|
||||
+ Forces the testroot to be reinstalled the next time you run
|
||||
+ the testsuite (or just rm -rf testroot.pristine)
|
||||
+
|
||||
@ -1,34 +0,0 @@
|
||||
commit 56e098118a31753a9f755948bb5a47bc7111e214
|
||||
Author: Andreas Schwab <schwab@suse.de>
|
||||
Date: Thu Aug 15 12:14:35 2019 +0200
|
||||
|
||||
Update i386 libm-test-ulps
|
||||
|
||||
Conflicts: ChangeLog removed
|
||||
|
||||
diff --git a/sysdeps/i386/fpu/libm-test-ulps b/sysdeps/i386/fpu/libm-test-ulps
|
||||
index e83bae71b4..2232296fe0 100644
|
||||
--- a/sysdeps/i386/fpu/libm-test-ulps
|
||||
+++ b/sysdeps/i386/fpu/libm-test-ulps
|
||||
@@ -1158,8 +1158,8 @@ float128: 4
|
||||
idouble: 4
|
||||
ifloat: 5
|
||||
ifloat128: 4
|
||||
-ildouble: 7
|
||||
-ldouble: 7
|
||||
+ildouble: 8
|
||||
+ldouble: 8
|
||||
|
||||
Function: Imaginary part of "clog10_upward":
|
||||
double: 2
|
||||
@@ -2222,8 +2222,8 @@ float128: 8
|
||||
idouble: 5
|
||||
ifloat: 5
|
||||
ifloat128: 8
|
||||
-ildouble: 5
|
||||
-ldouble: 5
|
||||
+ildouble: 6
|
||||
+ldouble: 6
|
||||
|
||||
Function: "log":
|
||||
double: 1
|
||||
@ -1,26 +0,0 @@
|
||||
Author: Patsy Griffin <patsy@redhat.com>
|
||||
|
||||
i386: update ulps
|
||||
|
||||
This change fixes 3 test failures:
|
||||
math/test-ildouble-lgamma
|
||||
math/test-ldouble-finite-lgamma
|
||||
math/test-ldouble-lgamma
|
||||
|
||||
This is a downstream only patch as upstream removed entries for
|
||||
i{float,double,ldouble} by commit: 1c15464ca05f36db5c582856d3770d5e8bde9d61.
|
||||
The ldouble change is already upstream.
|
||||
|
||||
--- a/sysdeps/i386/fpu/libm-test-ulps 2024-08-06 15:51:18.182808710 -0400
|
||||
+++ b/sysdeps/i386/fpu/libm-test-ulps 2024-08-06 18:01:50.579719841 -0400
|
||||
@@ -2030,8 +2030,8 @@ double: 5
|
||||
float: 5
|
||||
idouble: 5
|
||||
ifloat: 5
|
||||
-ildouble: 5
|
||||
-ldouble: 5
|
||||
+ildouble: 6
|
||||
+ldouble: 6
|
||||
|
||||
Function: "hypot":
|
||||
double: 1
|
||||
@ -1,25 +0,0 @@
|
||||
Author: Florian Weimer <fweimer@redhat.com>
|
||||
Date: Wed Jul 4 16:16:57 2018 +0200
|
||||
|
||||
Makeconfig (ASFLAGS): Always append required assembler flags.
|
||||
|
||||
Submitted upstream here:
|
||||
|
||||
https://sourceware.org/ml/libc-alpha/2018-07/msg00077.html
|
||||
|
||||
Otherwise, we lose essential flags such as -Wa,--noexecstack due to
|
||||
the way += works in make due to the ASFLAGS command line override.
|
||||
|
||||
diff --git a/Makeconfig b/Makeconfig
|
||||
index b0b27f0113ac18b8..92e76d6200bbcd5b 100644
|
||||
--- a/Makeconfig
|
||||
+++ b/Makeconfig
|
||||
@@ -1047,7 +1047,7 @@ endif
|
||||
ifndef ASFLAGS
|
||||
ASFLAGS := $(filter -g% -fdebug-prefix-map=%,$(CFLAGS))
|
||||
endif
|
||||
-ASFLAGS += -Werror=undef $(ASFLAGS-config) $(asflags-cpu)
|
||||
+override ASFLAGS += -Werror=undef $(ASFLAGS-config) $(asflags-cpu)
|
||||
|
||||
ifndef BUILD_CC
|
||||
BUILD_CC = $(CC)
|
||||
@ -1,286 +0,0 @@
|
||||
Short description: Add C.UTF-8 support.
|
||||
Author(s): Fedora glibc team <glibc@lists.fedoraproject.org>
|
||||
Origin: PATCH
|
||||
Upstream status: not-submitted
|
||||
|
||||
This patch needs to upstream as part of Carlos O'Donell
|
||||
<carlos@redhat.com>'s work on enabling upstream C.UTF-8 support. This
|
||||
work is currently blocked on cleaning up the test results to prove that
|
||||
full code-point sorting is working as intended.
|
||||
|
||||
Note that this patch does not provide full code-point sorting as
|
||||
expected.
|
||||
|
||||
This patch needs to upstream as soon as possible since it would be nice
|
||||
to have this in F29 and fixed.
|
||||
|
||||
From 2eda7b462b415105f5a05c1323372d4e39d46439 Mon Sep 17 00:00:00 2001
|
||||
From: Mike FABIAN <mfabian@redhat.com>
|
||||
Date: Mon, 10 Aug 2015 15:58:12 +0200
|
||||
Subject: [PATCH] Add a C.UTF-8 locale
|
||||
|
||||
---
|
||||
localedata/SUPPORTED | 1 +
|
||||
localedata/locales/C | 238 +++++++++++++++++++++++++++++++++++++++++++++++++++
|
||||
2 files changed, 239 insertions(+)
|
||||
create mode 100644 localedata/locales/C
|
||||
|
||||
diff --git a/localedata/SUPPORTED b/localedata/SUPPORTED
|
||||
index 8ca023e..2a78391 100644
|
||||
--- a/localedata/SUPPORTED
|
||||
+++ b/localedata/SUPPORTED
|
||||
@@ -1,6 +1,7 @@
|
||||
# This file names the currently supported and somewhat tested locales.
|
||||
# If you have any additions please file a glibc bug report.
|
||||
SUPPORTED-LOCALES=\
|
||||
+C.UTF-8/UTF-8 \
|
||||
aa_DJ.UTF-8/UTF-8 \
|
||||
aa_DJ/ISO-8859-1 \
|
||||
aa_ER/UTF-8 \
|
||||
diff --git a/localedata/locales/C b/localedata/locales/C
|
||||
new file mode 100644
|
||||
index 0000000..fdf460e
|
||||
--- /dev/null
|
||||
+++ b/localedata/locales/C
|
||||
@@ -0,0 +1,238 @@
|
||||
+escape_char /
|
||||
+comment_char %
|
||||
+% Locale for C locale in UTF-8
|
||||
+
|
||||
+LC_IDENTIFICATION
|
||||
+title "C locale"
|
||||
+source ""
|
||||
+address ""
|
||||
+contact ""
|
||||
+email "mfabian@redhat.com"
|
||||
+tel ""
|
||||
+fax ""
|
||||
+language "C"
|
||||
+territory ""
|
||||
+revision "1.0"
|
||||
+date "2015-08-10"
|
||||
+%
|
||||
+category "i18n:2012";LC_IDENTIFICATION
|
||||
+category "i18n:2012";LC_CTYPE
|
||||
+category "i18n:2012";LC_COLLATE
|
||||
+category "i18n:2012";LC_TIME
|
||||
+category "i18n:2012";LC_NUMERIC
|
||||
+category "i18n:2012";LC_MONETARY
|
||||
+category "i18n:2012";LC_MESSAGES
|
||||
+category "i18n:2012";LC_PAPER
|
||||
+category "i18n:2012";LC_NAME
|
||||
+category "i18n:2012";LC_ADDRESS
|
||||
+category "i18n:2012";LC_TELEPHONE
|
||||
+category "i18n:2012";LC_MEASUREMENT
|
||||
+END LC_IDENTIFICATION
|
||||
+
|
||||
+LC_CTYPE
|
||||
+copy "i18n"
|
||||
+
|
||||
+translit_start
|
||||
+include "translit_combining";""
|
||||
+translit_end
|
||||
+
|
||||
+END LC_CTYPE
|
||||
+
|
||||
+LC_COLLATE
|
||||
+order_start forward
|
||||
+<U0000>
|
||||
+..
|
||||
+<UFFFF>
|
||||
+<U10000>
|
||||
+..
|
||||
+<U1FFFF>
|
||||
+<U20000>
|
||||
+..
|
||||
+<U2FFFF>
|
||||
+<UE0000>
|
||||
+..
|
||||
+<UEFFFF>
|
||||
+<UF0000>
|
||||
+..
|
||||
+<UFFFFF>
|
||||
+<U100000>
|
||||
+..
|
||||
+<U10FFFF>
|
||||
+UNDEFINED
|
||||
+order_end
|
||||
+END LC_COLLATE
|
||||
+
|
||||
+LC_MONETARY
|
||||
+% This is the 14652 i18n fdcc-set definition for
|
||||
+% the LC_MONETARY category
|
||||
+% (except for the int_curr_symbol and currency_symbol, they are empty in
|
||||
+% the 14652 i18n fdcc-set definition and also empty in
|
||||
+% glibc/locale/C-monetary.c. But localedef complains in that case).
|
||||
+%
|
||||
+% Using "USD" for int_curr_symbol. But maybe "XXX" would be better?
|
||||
+% XXX is "No currency" (https://en.wikipedia.org/wiki/ISO_4217)
|
||||
+int_curr_symbol "<U0055><U0053><U0044><U0020>"
|
||||
+% Using "$" for currency_symbol. But maybe <U00A4> would be better?
|
||||
+% U+00A4 is the "generic currency symbol"
|
||||
+% (https://en.wikipedia.org/wiki/Currency_sign_%28typography%29)
|
||||
+currency_symbol "<U0024>"
|
||||
+mon_decimal_point "<U002E>"
|
||||
+mon_thousands_sep ""
|
||||
+mon_grouping -1
|
||||
+positive_sign ""
|
||||
+negative_sign "<U002D>"
|
||||
+int_frac_digits -1
|
||||
+frac_digits -1
|
||||
+p_cs_precedes -1
|
||||
+int_p_sep_by_space -1
|
||||
+p_sep_by_space -1
|
||||
+n_cs_precedes -1
|
||||
+int_n_sep_by_space -1
|
||||
+n_sep_by_space -1
|
||||
+p_sign_posn -1
|
||||
+n_sign_posn -1
|
||||
+%
|
||||
+END LC_MONETARY
|
||||
+
|
||||
+LC_NUMERIC
|
||||
+% This is the POSIX Locale definition for
|
||||
+% the LC_NUMERIC category.
|
||||
+%
|
||||
+decimal_point "<U002E>"
|
||||
+thousands_sep ""
|
||||
+grouping -1
|
||||
+END LC_NUMERIC
|
||||
+
|
||||
+LC_TIME
|
||||
+% This is the POSIX Locale definition for
|
||||
+% the LC_TIME category.
|
||||
+%
|
||||
+% Abbreviated weekday names (%a)
|
||||
+abday "<U0053><U0075><U006E>";"<U004D><U006F><U006E>";/
|
||||
+ "<U0054><U0075><U0065>";"<U0057><U0065><U0064>";/
|
||||
+ "<U0054><U0068><U0075>";"<U0046><U0072><U0069>";/
|
||||
+ "<U0053><U0061><U0074>"
|
||||
+
|
||||
+% Full weekday names (%A)
|
||||
+day "<U0053><U0075><U006E><U0064><U0061><U0079>";/
|
||||
+ "<U004D><U006F><U006E><U0064><U0061><U0079>";/
|
||||
+ "<U0054><U0075><U0065><U0073><U0064><U0061><U0079>";/
|
||||
+ "<U0057><U0065><U0064><U006E><U0065><U0073><U0064><U0061><U0079>";/
|
||||
+ "<U0054><U0068><U0075><U0072><U0073><U0064><U0061><U0079>";/
|
||||
+ "<U0046><U0072><U0069><U0064><U0061><U0079>";/
|
||||
+ "<U0053><U0061><U0074><U0075><U0072><U0064><U0061><U0079>"
|
||||
+
|
||||
+% Abbreviated month names (%b)
|
||||
+abmon "<U004A><U0061><U006E>";"<U0046><U0065><U0062>";/
|
||||
+ "<U004D><U0061><U0072>";"<U0041><U0070><U0072>";/
|
||||
+ "<U004D><U0061><U0079>";"<U004A><U0075><U006E>";/
|
||||
+ "<U004A><U0075><U006C>";"<U0041><U0075><U0067>";/
|
||||
+ "<U0053><U0065><U0070>";"<U004F><U0063><U0074>";/
|
||||
+ "<U004E><U006F><U0076>";"<U0044><U0065><U0063>"
|
||||
+
|
||||
+% Full month names (%B)
|
||||
+mon "<U004A><U0061><U006E><U0075><U0061><U0072><U0079>";/
|
||||
+ "<U0046><U0065><U0062><U0072><U0075><U0061><U0072><U0079>";/
|
||||
+ "<U004D><U0061><U0072><U0063><U0068>";/
|
||||
+ "<U0041><U0070><U0072><U0069><U006C>";/
|
||||
+ "<U004D><U0061><U0079>";/
|
||||
+ "<U004A><U0075><U006E><U0065>";/
|
||||
+ "<U004A><U0075><U006C><U0079>";/
|
||||
+ "<U0041><U0075><U0067><U0075><U0073><U0074>";/
|
||||
+ "<U0053><U0065><U0070><U0074><U0065><U006D><U0062><U0065><U0072>";/
|
||||
+ "<U004F><U0063><U0074><U006F><U0062><U0065><U0072>";/
|
||||
+ "<U004E><U006F><U0076><U0065><U006D><U0062><U0065><U0072>";/
|
||||
+ "<U0044><U0065><U0063><U0065><U006D><U0062><U0065><U0072>"
|
||||
+
|
||||
+% Week description, consists of three fields:
|
||||
+% 1. Number of days in a week.
|
||||
+% 2. Gregorian date that is a first weekday (19971130 for Sunday, 19971201 for Monday).
|
||||
+% 3. The weekday number to be contained in the first week of the year.
|
||||
+%
|
||||
+% ISO 8601 conforming applications should use the values 7, 19971201 (a
|
||||
+% Monday), and 4 (Thursday), respectively.
|
||||
+week 7;19971201;4
|
||||
+first_weekday 1
|
||||
+first_workday 1
|
||||
+
|
||||
+% Appropriate date and time representation (%c)
|
||||
+% "%a %b %e %H:%M:%S %Y"
|
||||
+d_t_fmt "<U0025><U0061><U0020><U0025><U0062><U0020><U0025><U0065><U0020><U0025><U0048><U003A><U0025><U004D><U003A><U0025><U0053><U0020><U0025><U0059>"
|
||||
+
|
||||
+% Appropriate date representation (%x)
|
||||
+% "%m/%d/%y"
|
||||
+d_fmt "<U0025><U006D><U002F><U0025><U0064><U002F><U0025><U0079>"
|
||||
+
|
||||
+% Appropriate time representation (%X)
|
||||
+% "%H:%M:%S"
|
||||
+t_fmt "<U0025><U0048><U003A><U0025><U004D><U003A><U0025><U0053>"
|
||||
+
|
||||
+% Appropriate AM/PM time representation (%r)
|
||||
+% "%I:%M:%S %p"
|
||||
+t_fmt_ampm "<U0025><U0049><U003A><U0025><U004D><U003A><U0025><U0053><U0020><U0025><U0070>"
|
||||
+
|
||||
+% Equivalent of AM/PM (%p) "AM"/"PM"
|
||||
+%
|
||||
+am_pm "<U0041><U004D>";"<U0050><U004D>"
|
||||
+
|
||||
+% Appropriate date representation (date(1)) "%a %b %e %H:%M:%S %Z %Y"
|
||||
+date_fmt "<U0025><U0061><U0020><U0025><U0062><U0020><U0025><U0065><U0020><U0025><U0048><U003A><U0025><U004D><U003A><U0025><U0053><U0020><U0025><U005A><U0020><U0025><U0059>"
|
||||
+END LC_TIME
|
||||
+
|
||||
+LC_MESSAGES
|
||||
+% This is the POSIX Locale definition for
|
||||
+% the LC_NUMERIC category.
|
||||
+%
|
||||
+yesexpr "<U005E><U005B><U0079><U0059><U005D>"
|
||||
+noexpr "<U005E><U005B><U006E><U004E><U005D>"
|
||||
+yesstr "<U0059><U0065><U0073>"
|
||||
+nostr "<U004E><U006F>"
|
||||
+END LC_MESSAGES
|
||||
+
|
||||
+LC_PAPER
|
||||
+% This is the ISO/IEC 14652 "i18n" definition for
|
||||
+% the LC_PAPER category.
|
||||
+% (A4 paper, this is also used in the built in C/POSIX
|
||||
+% locale in glibc/locale/C-paper.c)
|
||||
+height 297
|
||||
+width 210
|
||||
+END LC_PAPER
|
||||
+
|
||||
+LC_NAME
|
||||
+% This is the ISO/IEC 14652 "i18n" definition for
|
||||
+% the LC_NAME category.
|
||||
+% "%p%t%g%t%m%t%f"
|
||||
+% (also used in the built in C/POSIX locale in glibc/locale/C-name.c)
|
||||
+name_fmt "<U0025><U0070><U0025><U0074><U0025><U0067><U0025><U0074>/
|
||||
+<U0025><U006D><U0025><U0074><U0025><U0066>"
|
||||
+END LC_NAME
|
||||
+
|
||||
+LC_ADDRESS
|
||||
+% This is the ISO/IEC 14652 "i18n" definition for
|
||||
+% the LC_ADDRESS category.
|
||||
+% "%a%N%f%N%d%N%b%N%s %h %e %r%N%C-%z %T%N%c%N"
|
||||
+% (also used in the built in C/POSIX locale in glibc/locale/C-address.c)
|
||||
+postal_fmt "<U0025><U0061><U0025><U004E><U0025><U0066><U0025><U004E>/
|
||||
+<U0025><U0064><U0025><U004E><U0025><U0062><U0025><U004E><U0025><U0073>/
|
||||
+<U0020><U0025><U0068><U0020><U0025><U0065><U0020><U0025><U0072><U0025>/
|
||||
+<U004E><U0025><U0043><U002D><U0025><U007A><U0020><U0025><U0054><U0025>/
|
||||
+<U004E><U0025><U0063><U0025><U004E>"
|
||||
+END LC_ADDRESS
|
||||
+
|
||||
+LC_TELEPHONE
|
||||
+% This is the ISO/IEC 14652 "i18n" definition for
|
||||
+% the LC_TELEPHONE category.
|
||||
+% "+%c %a %l"
|
||||
+tel_int_fmt "<U002B><U0025><U0063><U0020><U0025><U0061><U0020><U0025>/
|
||||
+<U006C>"
|
||||
+% (also used in the built in C/POSIX locale in glibc/locale/C-telephone.c)
|
||||
+END LC_TELEPHONE
|
||||
+
|
||||
+LC_MEASUREMENT
|
||||
+% This is the ISO/IEC 14652 "i18n" definition for
|
||||
+% the LC_MEASUREMENT category.
|
||||
+% (same as in the built in C/POSIX locale in glibc/locale/C-measurement.c)
|
||||
+%metric
|
||||
+measurement 1
|
||||
+END LC_MEASUREMENT
|
||||
+
|
||||
--
|
||||
2.4.3
|
||||
|
||||
@ -1,15 +0,0 @@
|
||||
Short description: Adjust CS_PATH return value.
|
||||
Author(s): Fedora glibc team <glibc@lists.fedoraproject.org>
|
||||
Origin: PATCH
|
||||
Upstream status: not-needed
|
||||
|
||||
In Fedora we should return only /usr/bin because /bin is just a symlink
|
||||
to /usr/bin after MoveToUsr transition (which glibc has not really
|
||||
completed).
|
||||
|
||||
diff -pruN a/sysdeps/unix/confstr.h b/sysdeps/unix/confstr.h
|
||||
--- a/sysdeps/unix/confstr.h 2012-12-25 08:32:13.000000000 +0530
|
||||
+++ b/sysdeps/unix/confstr.h 2014-09-05 20:02:55.698275219 +0530
|
||||
@@ -1 +1 @@
|
||||
-#define CS_PATH "/bin:/usr/bin"
|
||||
+#define CS_PATH "/usr/bin"
|
||||
@ -1,91 +0,0 @@
|
||||
Short description: Cleanup use of _dl_starting_up.
|
||||
Author(s): Fedora glibc team <glibc@lists.fedoraproject.org>
|
||||
Origin: PATCH
|
||||
Upstream status: https://sourceware.org/ml/libc-alpha/2014-02/msg00589.html
|
||||
|
||||
Upstream discussions:
|
||||
https://sourceware.org/ml/libc-alpha/2014-02/msg00580.html
|
||||
|
||||
Based on the following commit:
|
||||
~~~
|
||||
From 16552c01a66633c9e412984d9d92616bd4e5303c Mon Sep 17 00:00:00 2001
|
||||
From: Andreas Schwab <schwab@redhat.com>
|
||||
Date: Fri, 11 Jun 2010 11:04:11 +0200
|
||||
Subject: [PATCH] Properly set __libc_multiple_libcs
|
||||
|
||||
* elf/rtld.c (_dl_starting_up): Always define.
|
||||
(dl_main): Always set _dl_starting_up.
|
||||
* elf/dl-support.c (_dl_starting_up): Always define.
|
||||
* elf/dl-init.c (_dl_init): Always clear _dl_starting_up.
|
||||
|
||||
---
|
||||
ChangeLog | 7 +++++++
|
||||
elf/dl-init.c | 4 ----
|
||||
elf/dl-support.c | 2 --
|
||||
elf/rtld.c | 4 ----
|
||||
4 files changed, 7 insertions(+), 10 deletions(-)
|
||||
~~~
|
||||
|
||||
This patch needs to go upstream to get cleaned up, but has always involed
|
||||
analysis of the GNU/Hurd parts of the change and that stalled out, but
|
||||
perhaps with build-many-glibcs we can now test these changes more easily.
|
||||
|
||||
Index: b/elf/dl-init.c
|
||||
===================================================================
|
||||
--- a/elf/dl-init.c
|
||||
+++ b/elf/dl-init.c
|
||||
@@ -119,8 +119,6 @@ _dl_init (struct link_map *main_map, int
|
||||
while (i-- > 0)
|
||||
call_init (main_map->l_initfini[i], argc, argv, env);
|
||||
|
||||
-#ifndef HAVE_INLINED_SYSCALLS
|
||||
/* Finished starting up. */
|
||||
_dl_starting_up = 0;
|
||||
-#endif
|
||||
}
|
||||
Index: b/elf/dl-support.c
|
||||
===================================================================
|
||||
--- a/elf/dl-support.c
|
||||
+++ b/elf/dl-support.c
|
||||
@@ -117,10 +117,8 @@ struct r_scope_elem _dl_initial_searchli
|
||||
.r_nlist = 1,
|
||||
};
|
||||
|
||||
-#ifndef HAVE_INLINED_SYSCALLS
|
||||
/* Nonzero during startup. */
|
||||
int _dl_starting_up = 1;
|
||||
-#endif
|
||||
|
||||
/* Random data provided by the kernel. */
|
||||
void *_dl_random;
|
||||
Index: b/elf/rtld.c
|
||||
===================================================================
|
||||
--- a/elf/rtld.c
|
||||
+++ b/elf/rtld.c
|
||||
@@ -214,7 +214,6 @@ audit_list_iter_next (struct audit_list_
|
||||
return iter->previous->name;
|
||||
}
|
||||
|
||||
-#ifndef HAVE_INLINED_SYSCALLS
|
||||
/* Set nonzero during loading and initialization of executable and
|
||||
libraries, cleared before the executable's entry point runs. This
|
||||
must not be initialized to nonzero, because the unused dynamic
|
||||
@@ -224,7 +223,6 @@ audit_list_iter_next (struct audit_list_
|
||||
never be called. */
|
||||
int _dl_starting_up = 0;
|
||||
rtld_hidden_def (_dl_starting_up)
|
||||
-#endif
|
||||
|
||||
/* This is the structure which defines all variables global to ld.so
|
||||
(except those which cannot be added for some reason). */
|
||||
@@ -898,10 +896,8 @@ dl_main (const ElfW(Phdr) *phdr,
|
||||
/* Process the environment variable which control the behaviour. */
|
||||
process_envvars (&mode);
|
||||
|
||||
-#ifndef HAVE_INLINED_SYSCALLS
|
||||
/* Set up a flag which tells we are just starting. */
|
||||
_dl_starting_up = 1;
|
||||
-#endif
|
||||
|
||||
if (*user_entry == (ElfW(Addr)) ENTRY_POINT)
|
||||
{
|
||||
@ -1,21 +0,0 @@
|
||||
Short description: Fedora-specific glibc install locale changes.
|
||||
Author(s): Fedora glibc team <glibc@lists.fedoraproject.org>
|
||||
Origin: PATCH
|
||||
Upstream status: not-needed
|
||||
|
||||
The Fedora glibc build and install does not need the normal install
|
||||
behaviour which updates the locale archive. The Fedora install phase
|
||||
in the spec file of the rpm will handle this manually.
|
||||
|
||||
diff --git a/localedata/Makefile b/localedata/Makefile
|
||||
index a5f3c92d58954dfc..56719c7c714aa0f1 100644
|
||||
--- a/localedata/Makefile
|
||||
+++ b/localedata/Makefile
|
||||
@@ -218,6 +218,7 @@ $(INSTALL-SUPPORTED-LOCALES): install-locales-dir
|
||||
echo -n '...'; \
|
||||
input=`echo $$locale | sed 's/\([^.]*\)[^@]*\(.*\)/\1\2/'`; \
|
||||
$(LOCALEDEF) $$flags --alias-file=../intl/locale.alias \
|
||||
+ --no-archive \
|
||||
-i locales/$$input -f charmaps/$$charset \
|
||||
$(addprefix --prefix=,$(install_root)) $$locale \
|
||||
&& echo ' done'; \
|
||||
@ -1,46 +0,0 @@
|
||||
Short description: Allow access to internal locale archive functions.
|
||||
Author(s): Fedora glibc team <glibc@lists.fedoraproject.org>
|
||||
Origin: PATCH
|
||||
Upstream status: not-needed
|
||||
|
||||
This is a part of commit glibc-2.3.3-1492-ga891c7b,
|
||||
needed for fedora/build-locale-archive.c only.
|
||||
|
||||
2007-04-16 Jakub Jelinek <jakub@redhat.com>
|
||||
|
||||
* locale/programs/locarchive.c (add_alias, insert_name): Remove static.
|
||||
|
||||
diff -Nrup a/locale/programs/locarchive.c b/locale/programs/locarchive.c
|
||||
--- a/locale/programs/locarchive.c 2012-06-05 07:42:49.000000000 -0600
|
||||
+++ b/locale/programs/locarchive.c 2012-06-07 12:15:21.585319540 -0600
|
||||
@@ -252,9 +252,9 @@ oldlocrecentcmp (const void *a, const vo
|
||||
/* forward decls for below */
|
||||
static uint32_t add_locale (struct locarhandle *ah, const char *name,
|
||||
locale_data_t data, bool replace);
|
||||
-static void add_alias (struct locarhandle *ah, const char *alias,
|
||||
- bool replace, const char *oldname,
|
||||
- uint32_t *locrec_offset_p);
|
||||
+void add_alias (struct locarhandle *ah, const char *alias,
|
||||
+ bool replace, const char *oldname,
|
||||
+ uint32_t *locrec_offset_p);
|
||||
|
||||
|
||||
static bool
|
||||
@@ -635,7 +635,7 @@ close_archive (struct locarhandle *ah)
|
||||
#include "../../intl/explodename.c"
|
||||
#include "../../intl/l10nflist.c"
|
||||
|
||||
-static struct namehashent *
|
||||
+struct namehashent *
|
||||
insert_name (struct locarhandle *ah,
|
||||
const char *name, size_t name_len, bool replace)
|
||||
{
|
||||
@@ -693,7 +693,7 @@ insert_name (struct locarhandle *ah,
|
||||
return &namehashtab[idx];
|
||||
}
|
||||
|
||||
-static void
|
||||
+void
|
||||
add_alias (struct locarhandle *ah, const char *alias, bool replace,
|
||||
const char *oldname, uint32_t *locrec_offset_p)
|
||||
{
|
||||
@ -1,31 +0,0 @@
|
||||
Short description: Fedora-specific enabling batch read in NSS.
|
||||
Author(s): Fedora glibc team <glibc@lists.fedoraproject.org>
|
||||
Origin: PATCH
|
||||
Bug-RHEL: #188246
|
||||
Upstream status: not-submitted
|
||||
|
||||
Enable batch read in NSS. It's not clear if this is always a win or
|
||||
just a win for NIS+, this needs to be analyzed and sent upstream or
|
||||
removed.
|
||||
|
||||
From baba5d9461d4e8a581ac26fe4412ad783ffc73e7 Mon Sep 17 00:00:00 2001
|
||||
From: Jakub Jelinek <jakub@redhat.com>
|
||||
Date: Mon, 1 May 2006 08:02:53 +0000
|
||||
Subject: [PATCH] Enable SETENT_BATCH_READ nis/nss option by default
|
||||
|
||||
* Mon May 1 2006 Jakub Jelinek <jakub@redhat.com> 2.4.90-4
|
||||
- SETENT_BATCH_READ /etc/default/nss option for speeding up
|
||||
some usages of NIS+ (#188246)
|
||||
|
||||
diff --git a/nis/nss b/nis/nss
|
||||
--- a/nis/nss
|
||||
+++ b/nis/nss
|
||||
@@ -25,7 +25,7 @@
|
||||
# memory with every getXXent() call. Otherwise each getXXent() call
|
||||
# might result into a network communication with the server to get
|
||||
# the next entry.
|
||||
-#SETENT_BATCH_READ=TRUE
|
||||
+SETENT_BATCH_READ=TRUE
|
||||
#
|
||||
# ADJUNCT_AS_SHADOW
|
||||
# If set to TRUE, the passwd routines in the NIS NSS module will not
|
||||
@ -1,38 +0,0 @@
|
||||
Short description: Do not define _XOPEN_STREAMS.
|
||||
Author(s): Fedora glibc team <glibc@lists.fedoraproject.org>
|
||||
Origin: PATCH
|
||||
Bug-Fedora: #436349
|
||||
Upstream status: not-submitted
|
||||
|
||||
This patch should go upstream. Not defining _XOPEN_STREAMS is the
|
||||
same as setting it to -1 for POSIX conformance. The headers setting
|
||||
needs to be reviewed indepedently.
|
||||
|
||||
This is part of commit glibc-2.3.3-1564-gd0b6ac6
|
||||
|
||||
* Fri Mar 14 2008 Jakub Jelinek <jakub@redhat.com> 2.7.90-11
|
||||
- remove <stropts.h>, define _XOPEN_STREAMS -1 (#436349)
|
||||
|
||||
diff -Nrup a/nptl/sysdeps/unix/sysv/linux/bits/posix_opt.h b/nptl/sysdeps/unix/sysv/linux/bits/posix_opt.h
|
||||
--- a/sysdeps/unix/sysv/linux/bits/posix_opt.h 2012-06-05 07:42:49.000000000 -0600
|
||||
+++ b/sysdeps/unix/sysv/linux/bits/posix_opt.h 2012-06-07 12:15:21.817318674 -0600
|
||||
@@ -188,4 +188,7 @@
|
||||
/* Typed memory objects are not available. */
|
||||
#define _POSIX_TYPED_MEMORY_OBJECTS -1
|
||||
|
||||
+/* Streams are not available. */
|
||||
+#define _XOPEN_STREAMS -1
|
||||
+
|
||||
#endif /* bits/posix_opt.h */
|
||||
diff -Nrup a/streams/Makefile b/streams/Makefile
|
||||
--- a/streams/Makefile 2012-06-05 07:42:49.000000000 -0600
|
||||
+++ b/streams/Makefile 2012-06-07 12:15:21.824318649 -0600
|
||||
@@ -20,7 +20,7 @@
|
||||
|
||||
include ../Makeconfig
|
||||
|
||||
-headers = stropts.h sys/stropts.h bits/stropts.h bits/xtitypes.h
|
||||
+#headers = stropts.h sys/stropts.h bits/stropts.h bits/xtitypes.h
|
||||
routines = isastream getmsg getpmsg putmsg putpmsg fattach fdetach
|
||||
|
||||
include ../Rules
|
||||
@ -1,35 +0,0 @@
|
||||
Downstream-specific patch to link DSO sorting tests with -ldl
|
||||
if needed. Upstream does not need this because <dlfcn.h> interfaces
|
||||
are part of libc.
|
||||
|
||||
diff --git a/scripts/dso-ordering-test.py b/scripts/dso-ordering-test.py
|
||||
index 43b5ec4d920ad6a3..ae85e0f4a6ae5b3e 100644
|
||||
--- a/scripts/dso-ordering-test.py
|
||||
+++ b/scripts/dso-ordering-test.py
|
||||
@@ -657,6 +657,8 @@ def process_testcase(t):
|
||||
% (test_name + "-" + dep + ".FAKE.so",
|
||||
("$(objpfx)" + test_subdir + "/"
|
||||
+ test_name + "-" + dep + ".so")))
|
||||
+ makefile.write(
|
||||
+ "LDLIBS-%s += -Wl,--as-needed -ldl -Wl,--no-as-needed\n" % dso)
|
||||
rule = ("$(objpfx)" + test_subdir + "/"
|
||||
+ test_name + "-" + dep + ".FAKE.os: "
|
||||
"$(objpfx)" + test_srcdir
|
||||
@@ -685,6 +687,8 @@ def process_testcase(t):
|
||||
+ test_descr.soname_map[o] + ".so")
|
||||
ldflags += (" -Wl,-soname=" + soname)
|
||||
makefile.write("LDFLAGS-%s = %s\n" % (dso, ldflags))
|
||||
+ makefile.write(
|
||||
+ "LDLIBS-%s += -Wl,--as-needed -ldl -Wl,--no-as-needed\n" % dso)
|
||||
if o in test_descr.callrefs:
|
||||
makefile.write("%s-no-z-defs = yes\n" % (dso))
|
||||
|
||||
@@ -702,6 +706,8 @@ def process_testcase(t):
|
||||
+ test_descr.soname_map['#'] + ".so")
|
||||
ldflags += (" -Wl,-soname=" + soname)
|
||||
makefile.write("LDFLAGS-%s = %s\n" % (test_name, ldflags))
|
||||
+ makefile.write(
|
||||
+ "LDLIBS-%s += -Wl,--as-needed -ldl -Wl,--no-as-needed\n" % test_name)
|
||||
rule = ("$(objpfx)" + test_subdir + "/" + test_name + ".o: "
|
||||
"$(objpfx)" + test_srcdir + test_name + ".c\n"
|
||||
"\t$(compile.c) $(OUTPUT_OPTION)\n")
|
||||
@ -1,357 +0,0 @@
|
||||
commit 3a0588ae48fb35384a6bd33f9b66403badfa1262
|
||||
Author: Adhemerval Zanella <adhemerval.zanella@linaro.org>
|
||||
Date: Tue Feb 8 15:22:49 2022 -0300
|
||||
|
||||
elf: Fix DFS sorting algorithm for LD_TRACE_LOADED_OBJECTS with missing libraries (BZ #28868)
|
||||
|
||||
On _dl_map_object the underlying file is not opened in trace mode
|
||||
(in other cases where the underlying file can't be opened,
|
||||
_dl_map_object quits with an error). If there any missing libraries
|
||||
being processed, they will not be considered on final nlist size
|
||||
passed on _dl_sort_maps later in the function. And it is then used by
|
||||
_dl_sort_maps_dfs on the stack allocated working maps:
|
||||
|
||||
222 /* Array to hold RPO sorting results, before we copy back to maps[]. */
|
||||
223 struct link_map *rpo[nmaps];
|
||||
224
|
||||
225 /* The 'head' position during each DFS iteration. Note that we start at
|
||||
226 one past the last element due to first-decrement-then-store (see the
|
||||
227 bottom of above dfs_traversal() routine). */
|
||||
228 struct link_map **rpo_head = &rpo[nmaps];
|
||||
|
||||
However while transversing the 'l_initfini' on dfs_traversal it will
|
||||
still consider the l_faked maps and thus update rpo more times than the
|
||||
allocated working 'rpo', overflowing the stack object.
|
||||
|
||||
As suggested in bugzilla, one option would be to avoid sorting the maps
|
||||
for trace mode. However I think ignoring l_faked object does make
|
||||
sense (there is one less constraint to call the sorting function), it
|
||||
allows a slight less stack usage for trace, and it is slight simpler
|
||||
solution.
|
||||
|
||||
The tests does trigger the stack overflow, however I tried to make
|
||||
it more generic to check different scenarios or missing objects.
|
||||
|
||||
Checked on x86_64-linux-gnu.
|
||||
|
||||
Reviewed-by: Siddhesh Poyarekar <siddhesh@sourceware.org>
|
||||
|
||||
Conflicts:
|
||||
elf/Makefile
|
||||
(differences in backported tests)
|
||||
|
||||
diff --git a/elf/Makefile b/elf/Makefile
|
||||
index 22a8060f7d3bb1a1..634c3113227d64a6 100644
|
||||
--- a/elf/Makefile
|
||||
+++ b/elf/Makefile
|
||||
@@ -584,6 +584,11 @@ modules-names = \
|
||||
libmarkermod5-3 \
|
||||
libmarkermod5-4 \
|
||||
libmarkermod5-5 \
|
||||
+ libtracemod1-1 \
|
||||
+ libtracemod2-1 \
|
||||
+ libtracemod3-1 \
|
||||
+ libtracemod4-1 \
|
||||
+ libtracemod5-1 \
|
||||
ltglobmod1 \
|
||||
ltglobmod2 \
|
||||
neededobj1 \
|
||||
@@ -983,6 +988,11 @@ tests-special += \
|
||||
$(objpfx)tst-initorder2-cmp.out \
|
||||
$(objpfx)tst-unused-dep-cmp.out \
|
||||
$(objpfx)tst-unused-dep.out \
|
||||
+ $(objpfx)tst-trace1.out \
|
||||
+ $(objpfx)tst-trace2.out \
|
||||
+ $(objpfx)tst-trace3.out \
|
||||
+ $(objpfx)tst-trace4.out \
|
||||
+ $(objpfx)tst-trace5.out \
|
||||
# tests-special
|
||||
endif
|
||||
|
||||
@@ -2619,6 +2629,51 @@ $(objpfx)tst-rtld-run-static.out: $(objpfx)/ldconfig
|
||||
|
||||
$(objpfx)tst-dlmopen-gethostbyname: $(libdl)
|
||||
$(objpfx)tst-dlmopen-gethostbyname.out: $(objpfx)tst-dlmopen-gethostbyname-mod.so
|
||||
+
|
||||
+LDFLAGS-libtracemod1-1.so += -Wl,-soname,libtracemod1.so
|
||||
+LDFLAGS-libtracemod2-1.so += -Wl,-soname,libtracemod2.so
|
||||
+LDFLAGS-libtracemod3-1.so += -Wl,-soname,libtracemod3.so
|
||||
+LDFLAGS-libtracemod4-1.so += -Wl,-soname,libtracemod4.so
|
||||
+LDFLAGS-libtracemod5-1.so += -Wl,-soname,libtracemod5.so
|
||||
+
|
||||
+$(objpfx)libtracemod1-1.so: $(objpfx)libtracemod2-1.so \
|
||||
+ $(objpfx)libtracemod3-1.so
|
||||
+$(objpfx)libtracemod2-1.so: $(objpfx)libtracemod4-1.so \
|
||||
+ $(objpfx)libtracemod5-1.so
|
||||
+
|
||||
+define libtracemod-x
|
||||
+$(objpfx)libtracemod$(1)/libtracemod$(1).so: $(objpfx)libtracemod$(1)-1.so
|
||||
+ $$(make-target-directory)
|
||||
+ cp $$< $$@
|
||||
+endef
|
||||
+libtracemod-suffixes = 1 2 3 4 5
|
||||
+$(foreach i,$(libtracemod-suffixes), $(eval $(call libtracemod-x,$(i))))
|
||||
+
|
||||
+define tst-trace-skeleton
|
||||
+$(objpfx)tst-trace$(1).out: $(objpfx)libtracemod1/libtracemod1.so \
|
||||
+ $(objpfx)libtracemod2/libtracemod2.so \
|
||||
+ $(objpfx)libtracemod3/libtracemod3.so \
|
||||
+ $(objpfx)libtracemod4/libtracemod4.so \
|
||||
+ $(objpfx)libtracemod5/libtracemod5.so \
|
||||
+ $(..)scripts/tst-ld-trace.py \
|
||||
+ tst-trace$(1).exp
|
||||
+ ${ $(PYTHON) $(..)scripts/tst-ld-trace.py \
|
||||
+ "$(test-wrapper-env) $(elf-objpfx)$(rtld-installed-name) \
|
||||
+ --library-path $(common-objpfx):$(strip $(2)) \
|
||||
+ $(objpfx)libtracemod1/libtracemod1.so" tst-trace$(1).exp \
|
||||
+ } > $$@; $$(evaluate-test)
|
||||
+endef
|
||||
+
|
||||
+$(eval $(call tst-trace-skeleton,1,))
|
||||
+$(eval $(call tst-trace-skeleton,2,\
|
||||
+ $(objpfx)libtracemod2))
|
||||
+$(eval $(call tst-trace-skeleton,3,\
|
||||
+ $(objpfx)libtracemod2:$(objpfx)libtracemod3))
|
||||
+$(eval $(call tst-trace-skeleton,4,\
|
||||
+ $(objpfx)libtracemod2:$(objpfx)libtracemod3:$(objpfx)libtracemod4))
|
||||
+$(eval $(call tst-trace-skeleton,5,\
|
||||
+ $(objpfx)libtracemod2:$(objpfx)libtracemod3:$(objpfx)libtracemod4:$(objpfx)libtracemod5))
|
||||
+
|
||||
$(objpfx)tst-audit-tlsdesc: $(objpfx)tst-audit-tlsdesc-mod1.so \
|
||||
$(objpfx)tst-audit-tlsdesc-mod2.so \
|
||||
$(shared-thread-library)
|
||||
diff --git a/elf/dl-deps.c b/elf/dl-deps.c
|
||||
index 9365d54c8e03e5f4..9ff589c8562b2dd1 100644
|
||||
--- a/elf/dl-deps.c
|
||||
+++ b/elf/dl-deps.c
|
||||
@@ -489,6 +489,8 @@ _dl_map_object_deps (struct link_map *map,
|
||||
|
||||
for (nlist = 0, runp = known; runp; runp = runp->next)
|
||||
{
|
||||
+ /* _dl_sort_maps ignores l_faked object, so it is safe to not consider
|
||||
+ them for nlist. */
|
||||
if (__builtin_expect (trace_mode, 0) && runp->map->l_faked)
|
||||
/* This can happen when we trace the loading. */
|
||||
--map->l_searchlist.r_nlist;
|
||||
diff --git a/elf/dl-sort-maps.c b/elf/dl-sort-maps.c
|
||||
index 398a08f28c4d9ff1..99354dc08a010dd3 100644
|
||||
--- a/elf/dl-sort-maps.c
|
||||
+++ b/elf/dl-sort-maps.c
|
||||
@@ -140,7 +140,9 @@ static void
|
||||
dfs_traversal (struct link_map ***rpo, struct link_map *map,
|
||||
bool *do_reldeps)
|
||||
{
|
||||
- if (map->l_visited)
|
||||
+ /* _dl_map_object_deps ignores l_faked objects when calculating the
|
||||
+ number of maps before calling _dl_sort_maps, ignore them as well. */
|
||||
+ if (map->l_visited || map->l_faked)
|
||||
return;
|
||||
|
||||
map->l_visited = 1;
|
||||
diff --git a/elf/libtracemod1-1.c b/elf/libtracemod1-1.c
|
||||
new file mode 100644
|
||||
index 0000000000000000..7c89c9a5a40b9668
|
||||
--- /dev/null
|
||||
+++ b/elf/libtracemod1-1.c
|
||||
@@ -0,0 +1 @@
|
||||
+/* Empty */
|
||||
diff --git a/elf/libtracemod2-1.c b/elf/libtracemod2-1.c
|
||||
new file mode 100644
|
||||
index 0000000000000000..7c89c9a5a40b9668
|
||||
--- /dev/null
|
||||
+++ b/elf/libtracemod2-1.c
|
||||
@@ -0,0 +1 @@
|
||||
+/* Empty */
|
||||
diff --git a/elf/libtracemod3-1.c b/elf/libtracemod3-1.c
|
||||
new file mode 100644
|
||||
index 0000000000000000..7c89c9a5a40b9668
|
||||
--- /dev/null
|
||||
+++ b/elf/libtracemod3-1.c
|
||||
@@ -0,0 +1 @@
|
||||
+/* Empty */
|
||||
diff --git a/elf/libtracemod4-1.c b/elf/libtracemod4-1.c
|
||||
new file mode 100644
|
||||
index 0000000000000000..7c89c9a5a40b9668
|
||||
--- /dev/null
|
||||
+++ b/elf/libtracemod4-1.c
|
||||
@@ -0,0 +1 @@
|
||||
+/* Empty */
|
||||
diff --git a/elf/libtracemod5-1.c b/elf/libtracemod5-1.c
|
||||
new file mode 100644
|
||||
index 0000000000000000..7c89c9a5a40b9668
|
||||
--- /dev/null
|
||||
+++ b/elf/libtracemod5-1.c
|
||||
@@ -0,0 +1 @@
|
||||
+/* Empty */
|
||||
diff --git a/elf/tst-trace1.exp b/elf/tst-trace1.exp
|
||||
new file mode 100644
|
||||
index 0000000000000000..4a6f5211a68fe2c8
|
||||
--- /dev/null
|
||||
+++ b/elf/tst-trace1.exp
|
||||
@@ -0,0 +1,4 @@
|
||||
+ld 1
|
||||
+libc 1
|
||||
+libtracemod2.so 0
|
||||
+libtracemod3.so 0
|
||||
diff --git a/elf/tst-trace2.exp b/elf/tst-trace2.exp
|
||||
new file mode 100644
|
||||
index 0000000000000000..e13506e2eb9aeca2
|
||||
--- /dev/null
|
||||
+++ b/elf/tst-trace2.exp
|
||||
@@ -0,0 +1,6 @@
|
||||
+ld 1
|
||||
+libc 1
|
||||
+libtracemod2.so 1
|
||||
+libtracemod3.so 0
|
||||
+libtracemod4.so 0
|
||||
+libtracemod5.so 0
|
||||
diff --git a/elf/tst-trace3.exp b/elf/tst-trace3.exp
|
||||
new file mode 100644
|
||||
index 0000000000000000..e574549d12a53d72
|
||||
--- /dev/null
|
||||
+++ b/elf/tst-trace3.exp
|
||||
@@ -0,0 +1,6 @@
|
||||
+ld 1
|
||||
+libc 1
|
||||
+libtracemod2.so 1
|
||||
+libtracemod3.so 1
|
||||
+libtracemod4.so 0
|
||||
+libtracemod5.so 0
|
||||
diff --git a/elf/tst-trace4.exp b/elf/tst-trace4.exp
|
||||
new file mode 100644
|
||||
index 0000000000000000..31ca97b35bde0009
|
||||
--- /dev/null
|
||||
+++ b/elf/tst-trace4.exp
|
||||
@@ -0,0 +1,6 @@
|
||||
+ld 1
|
||||
+libc 1
|
||||
+libtracemod2.so 1
|
||||
+libtracemod3.so 1
|
||||
+libtracemod4.so 1
|
||||
+libtracemod5.so 0
|
||||
diff --git a/elf/tst-trace5.exp b/elf/tst-trace5.exp
|
||||
new file mode 100644
|
||||
index 0000000000000000..5d7d95372656396f
|
||||
--- /dev/null
|
||||
+++ b/elf/tst-trace5.exp
|
||||
@@ -0,0 +1,6 @@
|
||||
+ld 1
|
||||
+libc 1
|
||||
+libtracemod2.so 1
|
||||
+libtracemod3.so 1
|
||||
+libtracemod4.so 1
|
||||
+libtracemod5.so 1
|
||||
diff --git a/scripts/tst-ld-trace.py b/scripts/tst-ld-trace.py
|
||||
new file mode 100755
|
||||
index 0000000000000000..f5a402800377f44b
|
||||
--- /dev/null
|
||||
+++ b/scripts/tst-ld-trace.py
|
||||
@@ -0,0 +1,108 @@
|
||||
+#!/usr/bin/python3
|
||||
+# Dump the output of LD_TRACE_LOADED_OBJECTS in architecture neutral format.
|
||||
+# Copyright (C) 2022 Free Software Foundation, Inc.
|
||||
+# Copyright The GNU Toolchain Authors.
|
||||
+# This file is part of the GNU C Library.
|
||||
+#
|
||||
+# The GNU C Library is free software; you can redistribute it and/or
|
||||
+# modify it under the terms of the GNU Lesser General Public
|
||||
+# License as published by the Free Software Foundation; either
|
||||
+# version 2.1 of the License, or (at your option) any later version.
|
||||
+#
|
||||
+# The GNU C Library is distributed in the hope that it will be useful,
|
||||
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
||||
+# Lesser General Public License for more details.
|
||||
+#
|
||||
+# You should have received a copy of the GNU Lesser General Public
|
||||
+# License along with the GNU C Library; if not, see
|
||||
+# <https://www.gnu.org/licenses/>.
|
||||
+
|
||||
+import argparse
|
||||
+import os
|
||||
+import subprocess
|
||||
+import sys
|
||||
+
|
||||
+try:
|
||||
+ subprocess.run
|
||||
+except:
|
||||
+ class _CompletedProcess:
|
||||
+ def __init__(self, args, returncode, stdout=None, stderr=None):
|
||||
+ self.args = args
|
||||
+ self.returncode = returncode
|
||||
+ self.stdout = stdout
|
||||
+ self.stderr = stderr
|
||||
+
|
||||
+ def _run(*popenargs, input=None, timeout=None, check=False, **kwargs):
|
||||
+ assert(timeout is None)
|
||||
+ with subprocess.Popen(*popenargs, **kwargs) as process:
|
||||
+ try:
|
||||
+ stdout, stderr = process.communicate(input)
|
||||
+ except:
|
||||
+ process.kill()
|
||||
+ process.wait()
|
||||
+ raise
|
||||
+ returncode = process.poll()
|
||||
+ if check and returncode:
|
||||
+ raise subprocess.CalledProcessError(returncode, popenargs)
|
||||
+ return _CompletedProcess(popenargs, returncode, stdout, stderr)
|
||||
+
|
||||
+ subprocess.run = _run
|
||||
+
|
||||
+def is_vdso(lib):
|
||||
+ return lib.startswith('linux-gate') or lib.startswith('linux-vdso')
|
||||
+
|
||||
+
|
||||
+def parse_trace(cmd, fref):
|
||||
+ new_env = os.environ.copy()
|
||||
+ new_env['LD_TRACE_LOADED_OBJECTS'] = '1'
|
||||
+ trace_out = subprocess.run(cmd, stdout=subprocess.PIPE, check=True,
|
||||
+ universal_newlines=True, env=new_env).stdout
|
||||
+ trace = []
|
||||
+ for line in trace_out.splitlines():
|
||||
+ line = line.strip()
|
||||
+ if is_vdso(line):
|
||||
+ continue
|
||||
+ fields = line.split('=>' if '=>' in line else ' ')
|
||||
+ lib = os.path.basename(fields[0].strip())
|
||||
+ if lib.startswith('ld'):
|
||||
+ lib = 'ld'
|
||||
+ elif lib.startswith('libc'):
|
||||
+ lib = 'libc'
|
||||
+ found = 1 if fields[1].strip() != 'not found' else 0
|
||||
+ trace += ['{} {}'.format(lib, found)]
|
||||
+ trace = sorted(trace)
|
||||
+
|
||||
+ reference = sorted(line.replace('\n','') for line in fref.readlines())
|
||||
+
|
||||
+ ret = 0 if trace == reference else 1
|
||||
+ if ret != 0:
|
||||
+ for i in reference:
|
||||
+ if i not in trace:
|
||||
+ print("Only in {}: {}".format(fref.name, i))
|
||||
+ for i in trace:
|
||||
+ if i not in reference:
|
||||
+ print("Only in trace: {}".format(i))
|
||||
+
|
||||
+ sys.exit(ret)
|
||||
+
|
||||
+
|
||||
+def get_parser():
|
||||
+ parser = argparse.ArgumentParser(description=__doc__)
|
||||
+ parser.add_argument('command',
|
||||
+ help='comand to run')
|
||||
+ parser.add_argument('reference',
|
||||
+ help='reference file to compare')
|
||||
+ return parser
|
||||
+
|
||||
+
|
||||
+def main(argv):
|
||||
+ parser = get_parser()
|
||||
+ opts = parser.parse_args(argv)
|
||||
+ with open(opts.reference, 'r') as fref:
|
||||
+ # Remove the initial 'env' command.
|
||||
+ parse_trace(opts.command.split()[1:], fref)
|
||||
+
|
||||
+
|
||||
+if __name__ == '__main__':
|
||||
+ main(sys.argv[1:])
|
||||
@ -1,36 +0,0 @@
|
||||
commit a2211c76c3b994099fd58a06d6072d7495d699cd
|
||||
Author: Florian Weimer <fweimer@redhat.com>
|
||||
Date: Fri Mar 18 18:18:35 2022 +0100
|
||||
|
||||
scripts/dso-ordering-test.py: Fix C&P error in * callrefs processing
|
||||
|
||||
The elf/dso-sort-tests-src subdirectory is not changed by this commit,
|
||||
so it seems that the cut-and-paste error was not material.
|
||||
|
||||
Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
|
||||
|
||||
diff --git a/scripts/dso-ordering-test.py b/scripts/dso-ordering-test.py
|
||||
index bde0406be9da14fc..ee476c810c76f1b0 100644
|
||||
--- a/scripts/dso-ordering-test.py
|
||||
+++ b/scripts/dso-ordering-test.py
|
||||
@@ -551,17 +551,17 @@ def process_testcase(t):
|
||||
if obj in t.deps:
|
||||
deps = t.deps[obj]
|
||||
if '*' in deps:
|
||||
- t.deps[obj].remove('*')
|
||||
+ deps.remove('*')
|
||||
t.add_deps([obj], non_dep_tgt_objs)
|
||||
if obj in t.callrefs:
|
||||
deps = t.callrefs[obj]
|
||||
if '*' in deps:
|
||||
- t.deps[obj].remove('*')
|
||||
+ deps.remove('*')
|
||||
t.add_callrefs([obj], non_dep_tgt_objs)
|
||||
if "#" in t.deps:
|
||||
deps = t.deps["#"]
|
||||
if '*' in deps:
|
||||
- t.deps["#"].remove('*')
|
||||
+ deps.remove('*')
|
||||
t.add_deps(["#"], non_dep_tgt_objs)
|
||||
|
||||
# If no main program was specified in dependency description, make a
|
||||
@ -1,49 +0,0 @@
|
||||
Backport of this Fedora Rawhide commit but split out into a distinct
|
||||
patch.
|
||||
|
||||
commit 72195d44855ab96875f117acb75c37f98dcb26a9
|
||||
Author: Carlos O'Donell <carlos@redhat.com>
|
||||
Date: Thu Jun 6 23:58:21 2019 -0400
|
||||
|
||||
locale: Fix C.UTF-8 ranges.
|
||||
|
||||
The ellipsis range support only allows <Uxxxx> or <Uxxxxxxxx> as
|
||||
valid unicode code points, otherwise it treats it as a symbol and
|
||||
since we don't define the symbol the entire range is unused.
|
||||
|
||||
diff --git a/localedata/locales/C b/localedata/locales/C
|
||||
index b2c2d1dc417cde69..30d9563213b8cb0f 100644
|
||||
--- a/localedata/locales/C
|
||||
+++ b/localedata/locales/C
|
||||
@@ -43,21 +43,21 @@ order_start forward
|
||||
<U0000>
|
||||
..
|
||||
<UFFFF>
|
||||
-<U10000>
|
||||
+<U00010000>
|
||||
..
|
||||
-<U1FFFF>
|
||||
-<U20000>
|
||||
+<U0001FFFF>
|
||||
+<U00020000>
|
||||
..
|
||||
-<U2FFFF>
|
||||
-<UE0000>
|
||||
+<U0002FFFF>
|
||||
+<U000E0000>
|
||||
..
|
||||
-<UEFFFF>
|
||||
-<UF0000>
|
||||
+<U000EFFFF>
|
||||
+<U000F0000>
|
||||
..
|
||||
-<UFFFFF>
|
||||
-<U100000>
|
||||
+<U000FFFFF>
|
||||
+<U00100000>
|
||||
..
|
||||
-<U10FFFF>
|
||||
+<U0010FFFF>
|
||||
UNDEFINED
|
||||
order_end
|
||||
END LC_COLLATE
|
||||
@ -1,185 +0,0 @@
|
||||
commit 96cd0558bcd69481ccc42e1b392f0c0b36fce2b0
|
||||
Author: Florian Weimer <fweimer@redhat.com>
|
||||
Date: Wed Nov 28 19:59:45 2018 +0100
|
||||
|
||||
support: Add signal support to support_capture_subprocess_check
|
||||
|
||||
Signal zero does not terminate a process, so it is safe to use negative
|
||||
values for signal numbers.
|
||||
|
||||
Adjust libio/tst-vtables-common.c to use this new functionality,
|
||||
instead of determining the termination status for a signal indirectly.
|
||||
|
||||
diff --git a/libio/tst-vtables-common.c b/libio/tst-vtables-common.c
|
||||
index 5e3101206919fa1b..85e246cd1131f8e8 100644
|
||||
--- a/libio/tst-vtables-common.c
|
||||
+++ b/libio/tst-vtables-common.c
|
||||
@@ -380,21 +380,6 @@ without_compatibility_fflush (void *closure)
|
||||
_exit (1);
|
||||
}
|
||||
|
||||
-/* Exit status after abnormal termination. */
|
||||
-static int termination_status;
|
||||
-
|
||||
-static void
|
||||
-init_termination_status (void)
|
||||
-{
|
||||
- pid_t pid = xfork ();
|
||||
- if (pid == 0)
|
||||
- abort ();
|
||||
- xwaitpid (pid, &termination_status, 0);
|
||||
-
|
||||
- TEST_VERIFY (WIFSIGNALED (termination_status));
|
||||
- TEST_COMPARE (WTERMSIG (termination_status), SIGABRT);
|
||||
-}
|
||||
-
|
||||
static void
|
||||
check_for_termination (const char *name, void (*callback) (void *))
|
||||
{
|
||||
@@ -404,7 +389,7 @@ check_for_termination (const char *name, void (*callback) (void *))
|
||||
shared->calls = 0;
|
||||
struct support_capture_subprocess proc
|
||||
= support_capture_subprocess (callback, NULL);
|
||||
- support_capture_subprocess_check (&proc, name, termination_status,
|
||||
+ support_capture_subprocess_check (&proc, name, -SIGABRT,
|
||||
sc_allow_stderr);
|
||||
const char *message
|
||||
= "Fatal error: glibc detected an invalid stdio handle\n";
|
||||
@@ -491,7 +476,6 @@ run_tests (bool initially_disabled)
|
||||
|
||||
shared = support_shared_allocate (sizeof (*shared));
|
||||
shared->initially_disabled = initially_disabled;
|
||||
- init_termination_status ();
|
||||
|
||||
if (initially_disabled)
|
||||
{
|
||||
diff --git a/support/capture_subprocess.h b/support/capture_subprocess.h
|
||||
index d5eac84d09ae325f..2d2384e73df0d2d0 100644
|
||||
--- a/support/capture_subprocess.h
|
||||
+++ b/support/capture_subprocess.h
|
||||
@@ -55,13 +55,16 @@ enum support_capture_allow
|
||||
sc_allow_stderr = 0x04,
|
||||
};
|
||||
|
||||
-/* Check that the subprocess exited with STATUS and that only the
|
||||
- allowed outputs happened. ALLOWED is a combination of
|
||||
- support_capture_allow flags. Report errors under the CONTEXT
|
||||
- message. */
|
||||
+/* Check that the subprocess exited and that only the allowed outputs
|
||||
+ happened. If STATUS_OR_SIGNAL is nonnegative, it is the expected
|
||||
+ (decoded) exit status of the process, as returned by WEXITSTATUS.
|
||||
+ If STATUS_OR_SIGNAL is negative, -STATUS_OR_SIGNAL is the expected
|
||||
+ termination signal, as returned by WTERMSIG. ALLOWED is a
|
||||
+ combination of support_capture_allow flags. Report errors under
|
||||
+ the CONTEXT message. */
|
||||
void support_capture_subprocess_check (struct support_capture_subprocess *,
|
||||
- const char *context, int status,
|
||||
- int allowed)
|
||||
+ const char *context,
|
||||
+ int status_or_signal, int allowed)
|
||||
__attribute__ ((nonnull (1, 2)));
|
||||
|
||||
#endif /* SUPPORT_CAPTURE_SUBPROCESS_H */
|
||||
diff --git a/support/support_capture_subprocess_check.c b/support/support_capture_subprocess_check.c
|
||||
index ff5ee89fb02599ae..8b4c352c96227b78 100644
|
||||
--- a/support/support_capture_subprocess_check.c
|
||||
+++ b/support/support_capture_subprocess_check.c
|
||||
@@ -20,6 +20,7 @@
|
||||
#include <stdio.h>
|
||||
#include <support/capture_subprocess.h>
|
||||
#include <support/check.h>
|
||||
+#include <sys/wait.h>
|
||||
|
||||
static void
|
||||
print_context (const char *context, bool *failed)
|
||||
@@ -31,9 +32,22 @@ print_context (const char *context, bool *failed)
|
||||
printf ("error: subprocess failed: %s\n", context);
|
||||
}
|
||||
|
||||
+static void
|
||||
+print_actual_status (struct support_capture_subprocess *proc)
|
||||
+{
|
||||
+ if (WIFEXITED (proc->status))
|
||||
+ printf ("error: actual exit status: %d [0x%x]\n",
|
||||
+ WEXITSTATUS (proc->status), proc->status);
|
||||
+ else if (WIFSIGNALED (proc->status))
|
||||
+ printf ("error: actual termination signal: %d [0x%x]\n",
|
||||
+ WTERMSIG (proc->status), proc->status);
|
||||
+ else
|
||||
+ printf ("error: actual undecoded exit status: [0x%x]\n", proc->status);
|
||||
+}
|
||||
+
|
||||
void
|
||||
support_capture_subprocess_check (struct support_capture_subprocess *proc,
|
||||
- const char *context, int status,
|
||||
+ const char *context, int status_or_signal,
|
||||
int allowed)
|
||||
{
|
||||
TEST_VERIFY ((allowed & sc_allow_none)
|
||||
@@ -44,11 +58,28 @@ support_capture_subprocess_check (struct support_capture_subprocess *proc,
|
||||
|| (allowed & sc_allow_stderr))));
|
||||
|
||||
bool failed = false;
|
||||
- if (proc->status != status)
|
||||
+ if (status_or_signal >= 0)
|
||||
{
|
||||
- print_context (context, &failed);
|
||||
- printf ("error: expected exit status: %d\n", status);
|
||||
- printf ("error: actual exit status: %d\n", proc->status);
|
||||
+ /* Expect regular termination. */
|
||||
+ if (!(WIFEXITED (proc->status)
|
||||
+ && WEXITSTATUS (proc->status) == status_or_signal))
|
||||
+ {
|
||||
+ print_context (context, &failed);
|
||||
+ printf ("error: expected exit status: %d\n", status_or_signal);
|
||||
+ print_actual_status (proc);
|
||||
+ }
|
||||
+ }
|
||||
+ else
|
||||
+ {
|
||||
+ /* status_or_signal < 0. Expect termination by signal. */
|
||||
+ if (!(WIFSIGNALED (proc->status)
|
||||
+ && WTERMSIG (proc->status) == -status_or_signal))
|
||||
+ {
|
||||
+ print_context (context, &failed);
|
||||
+ printf ("error: expected termination signal: %d\n",
|
||||
+ -status_or_signal);
|
||||
+ print_actual_status (proc);
|
||||
+ }
|
||||
}
|
||||
if (!(allowed & sc_allow_stdout) && proc->out.length != 0)
|
||||
{
|
||||
diff --git a/support/tst-support_capture_subprocess.c b/support/tst-support_capture_subprocess.c
|
||||
index 63b6699622f97fcc..99570879eedd65b1 100644
|
||||
--- a/support/tst-support_capture_subprocess.c
|
||||
+++ b/support/tst-support_capture_subprocess.c
|
||||
@@ -285,15 +285,29 @@ do_multiple_tests (enum test_type type)
|
||||
|
||||
check_stream ("stdout", &result.out, test.out);
|
||||
check_stream ("stderr", &result.err, test.err);
|
||||
+
|
||||
+ /* Allowed output for support_capture_subprocess_check. */
|
||||
+ int check_allow = 0;
|
||||
+ if (lengths[length_idx_stdout] > 0)
|
||||
+ check_allow |= sc_allow_stdout;
|
||||
+ if (lengths[length_idx_stderr] > 0)
|
||||
+ check_allow |= sc_allow_stderr;
|
||||
+ if (check_allow == 0)
|
||||
+ check_allow = sc_allow_none;
|
||||
+
|
||||
if (test.signal != 0)
|
||||
{
|
||||
TEST_VERIFY (WIFSIGNALED (result.status));
|
||||
TEST_VERIFY (WTERMSIG (result.status) == test.signal);
|
||||
+ support_capture_subprocess_check (&result, "signal",
|
||||
+ -SIGTERM, check_allow);
|
||||
}
|
||||
else
|
||||
{
|
||||
TEST_VERIFY (WIFEXITED (result.status));
|
||||
TEST_VERIFY (WEXITSTATUS (result.status) == test.status);
|
||||
+ support_capture_subprocess_check (&result, "exit",
|
||||
+ test.status, check_allow);
|
||||
}
|
||||
support_capture_subprocess_free (&result);
|
||||
free (test.out);
|
||||
@ -1,42 +0,0 @@
|
||||
commit e37c2cf299b61ce18f62852f6c5624c27829b610
|
||||
Author: Florian Weimer <fweimer@redhat.com>
|
||||
Date: Thu Oct 31 18:48:43 2019 +0100
|
||||
|
||||
Move _dl_open_check to its original place in dl_open_worker
|
||||
|
||||
This reverts the non-test change from commit d0093c5cefb7f7a4143f
|
||||
("Call _dl_open_check after relocation [BZ #24259]"), given that
|
||||
the underlying bug has been fixed properly in commit 61b74477fa7f63
|
||||
("Remove all loaded objects if dlopen fails, ignoring NODELETE
|
||||
[BZ #20839]").
|
||||
|
||||
Tested on x86-64-linux-gnu, with and without --enable-cet.
|
||||
|
||||
Change-Id: I995a6cfb89f25d2b0cf5e606428c2a93eb48fc33
|
||||
|
||||
diff --git a/elf/dl-open.c b/elf/dl-open.c
|
||||
index 25838b073ac1edaf..e13968d4d7c4c83f 100644
|
||||
--- a/elf/dl-open.c
|
||||
+++ b/elf/dl-open.c
|
||||
@@ -619,6 +619,8 @@ dl_open_worker (void *a)
|
||||
_dl_debug_state ();
|
||||
LIBC_PROBE (map_complete, 3, args->nsid, r, new);
|
||||
|
||||
+ _dl_open_check (new);
|
||||
+
|
||||
/* Print scope information. */
|
||||
if (__glibc_unlikely (GLRO(dl_debug_mask) & DL_DEBUG_SCOPES))
|
||||
_dl_show_scope (new, 0);
|
||||
@@ -699,12 +701,6 @@ dl_open_worker (void *a)
|
||||
_dl_relocate_object (l, l->l_scope, reloc_mode, 0);
|
||||
}
|
||||
|
||||
- /* NB: Workaround for [BZ #20839] which doesn't remove the NODELETE
|
||||
- object when _dl_open_check throws an exception. Move it after
|
||||
- relocation to avoid leaving the NODELETE object mapped without
|
||||
- relocation. */
|
||||
- _dl_open_check (new);
|
||||
-
|
||||
/* This only performs the memory allocations. The actual update of
|
||||
the scopes happens below, after failure is impossible. */
|
||||
resize_scopes (new);
|
||||
@ -1,27 +0,0 @@
|
||||
commit 61a7c9df71ee4e6f94b56c20f0d37c6e17d5f284
|
||||
Author: Florian Weimer <fweimer@redhat.com>
|
||||
Date: Mon Dec 2 14:53:16 2019 +0100
|
||||
|
||||
elf/tst-dlopenfail: Disable --no-as-needed for tst-dlopenfailmod1.so
|
||||
|
||||
Otherwise, the shared object dependency which triggers the load
|
||||
failure is dropped, invalidating the test.
|
||||
|
||||
diff --git a/elf/Makefile b/elf/Makefile
|
||||
index bf7c41f38be42184..467e810e784bb96d 100644
|
||||
--- a/elf/Makefile
|
||||
+++ b/elf/Makefile
|
||||
@@ -1543,8 +1543,11 @@ LDFLAGS-tst-finilazyfailmod.so = \
|
||||
$(objpfx)tst-dlopenfail: $(libdl)
|
||||
$(objpfx)tst-dlopenfail.out: \
|
||||
$(objpfx)tst-dlopenfailmod1.so $(objpfx)tst-dlopenfailmod2.so
|
||||
-# Order matters here. tst-dlopenfaillinkmod.so's soname ensures
|
||||
-# a run-time loader failure.
|
||||
+# Order matters here. tst-dlopenfaillinkmod.so's soname ensures a
|
||||
+# run-time loader failure. --as-needed breaks this test because
|
||||
+# nothing actually references tst-dlopenfailmod2.so (with its soname
|
||||
+# tst-dlopenfail-missingmod.so).
|
||||
+LDFLAGS-tst-dlopenfailmod1.so = -Wl,--no-as-needed
|
||||
$(objpfx)tst-dlopenfailmod1.so: \
|
||||
$(shared-thread-library) $(objpfx)tst-dlopenfaillinkmod.so
|
||||
LDFLAGS-tst-dlopenfaillinkmod.so = -Wl,-soname,tst-dlopenfail-missingmod.so
|
||||
File diff suppressed because it is too large
Load Diff
@ -1,328 +0,0 @@
|
||||
commit f8ed116aa574435c6e28260f21963233682d3b57
|
||||
Author: Florian Weimer <fweimer@redhat.com>
|
||||
Date: Fri Dec 13 10:18:46 2019 +0100
|
||||
|
||||
dlopen: Rework handling of pending NODELETE status
|
||||
|
||||
Commit a2e8aa0d9ea648068d8be52dd7b15f1b6a008e23 ("Block signals during
|
||||
the initial part of dlopen") was deemed necessary because of
|
||||
read-modify-write operations like the one in add_dependency in
|
||||
elf/dl-lookup.c. In the old code, we check for any kind of NODELETE
|
||||
status and bail out:
|
||||
|
||||
/* Redo the NODELETE check, as when dl_load_lock wasn't held
|
||||
yet this could have changed. */
|
||||
if (map->l_nodelete != link_map_nodelete_inactive)
|
||||
goto out;
|
||||
|
||||
And then set pending status (during relocation):
|
||||
|
||||
if (flags & DL_LOOKUP_FOR_RELOCATE)
|
||||
map->l_nodelete = link_map_nodelete_pending;
|
||||
else
|
||||
map->l_nodelete = link_map_nodelete_active;
|
||||
|
||||
If a signal arrives during relocation and the signal handler, through
|
||||
lazy binding, adds a global scope dependency on the same map, it will
|
||||
set map->l_nodelete to link_map_nodelete_active. This will be
|
||||
overwritten with link_map_nodelete_pending by the dlopen relocation
|
||||
code.
|
||||
|
||||
To avoid such problems in relation to the l_nodelete member, this
|
||||
commit introduces two flags for active NODELETE status (irrevocable)
|
||||
and pending NODELETE status (revocable until activate_nodelete is
|
||||
invoked). As a result, NODELETE processing in dlopen does not
|
||||
introduce further reasons why lazy binding from signal handlers
|
||||
is unsafe during dlopen, and a subsequent commit can remove signal
|
||||
blocking from dlopen.
|
||||
|
||||
This does not address pre-existing issues (unrelated to the NODELETE
|
||||
changes) which make lazy binding in a signal handler during dlopen
|
||||
unsafe, such as the use of malloc in both cases.
|
||||
|
||||
Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
|
||||
Reviewed-by: Carlos O'Donell <carlos@redhat.com>
|
||||
|
||||
diff --git a/elf/dl-close.c b/elf/dl-close.c
|
||||
index 243a028c443173c1..fa7f3e8174576e46 100644
|
||||
--- a/elf/dl-close.c
|
||||
+++ b/elf/dl-close.c
|
||||
@@ -197,7 +197,7 @@ _dl_close_worker (struct link_map *map, bool force)
|
||||
/* Check whether this object is still used. */
|
||||
if (l->l_type == lt_loaded
|
||||
&& l->l_direct_opencount == 0
|
||||
- && l->l_nodelete != link_map_nodelete_active
|
||||
+ && !l->l_nodelete_active
|
||||
/* See CONCURRENCY NOTES in cxa_thread_atexit_impl.c to know why
|
||||
acquire is sufficient and correct. */
|
||||
&& atomic_load_acquire (&l->l_tls_dtor_count) == 0
|
||||
@@ -279,8 +279,7 @@ _dl_close_worker (struct link_map *map, bool force)
|
||||
|
||||
if (!used[i])
|
||||
{
|
||||
- assert (imap->l_type == lt_loaded
|
||||
- && imap->l_nodelete != link_map_nodelete_active);
|
||||
+ assert (imap->l_type == lt_loaded && !imap->l_nodelete_active);
|
||||
|
||||
/* Call its termination function. Do not do it for
|
||||
half-cooked objects. Temporarily disable exception
|
||||
@@ -820,7 +819,7 @@ _dl_close (void *_map)
|
||||
before we took the lock. There is no way to detect this (see below)
|
||||
so we proceed assuming this isn't the case. First see whether we
|
||||
can remove the object at all. */
|
||||
- if (__glibc_unlikely (map->l_nodelete == link_map_nodelete_active))
|
||||
+ if (__glibc_unlikely (map->l_nodelete_active))
|
||||
{
|
||||
/* Nope. Do nothing. */
|
||||
__rtld_lock_unlock_recursive (GL(dl_load_lock));
|
||||
diff --git a/elf/dl-lookup.c b/elf/dl-lookup.c
|
||||
index 35a3f96a6296294a..01724a54f8840f9f 100644
|
||||
--- a/elf/dl-lookup.c
|
||||
+++ b/elf/dl-lookup.c
|
||||
@@ -187,6 +187,28 @@ enter_unique_sym (struct unique_sym *table, size_t size,
|
||||
table[idx].map = map;
|
||||
}
|
||||
|
||||
+/* Mark MAP as NODELETE according to the lookup mode in FLAGS. During
|
||||
+ initial relocation, NODELETE state is pending only. */
|
||||
+static void
|
||||
+mark_nodelete (struct link_map *map, int flags)
|
||||
+{
|
||||
+ if (flags & DL_LOOKUP_FOR_RELOCATE)
|
||||
+ map->l_nodelete_pending = true;
|
||||
+ else
|
||||
+ map->l_nodelete_active = true;
|
||||
+}
|
||||
+
|
||||
+/* Return true if MAP is marked as NODELETE according to the lookup
|
||||
+ mode in FLAGS> */
|
||||
+static bool
|
||||
+is_nodelete (struct link_map *map, int flags)
|
||||
+{
|
||||
+ /* Non-pending NODELETE always counts. Pending NODELETE only counts
|
||||
+ during initial relocation processing. */
|
||||
+ return map->l_nodelete_active
|
||||
+ || ((flags & DL_LOOKUP_FOR_RELOCATE) && map->l_nodelete_pending);
|
||||
+}
|
||||
+
|
||||
/* Utility function for do_lookup_x. Lookup an STB_GNU_UNIQUE symbol
|
||||
in the unique symbol table, creating a new entry if necessary.
|
||||
Return the matching symbol in RESULT. */
|
||||
@@ -311,8 +333,7 @@ do_lookup_unique (const char *undef_name, uint_fast32_t new_hash,
|
||||
enter_unique_sym (entries, size,
|
||||
new_hash, strtab + sym->st_name, sym, map);
|
||||
|
||||
- if (map->l_type == lt_loaded
|
||||
- && map->l_nodelete == link_map_nodelete_inactive)
|
||||
+ if (map->l_type == lt_loaded && !is_nodelete (map, flags))
|
||||
{
|
||||
/* Make sure we don't unload this object by
|
||||
setting the appropriate flag. */
|
||||
@@ -320,10 +341,7 @@ do_lookup_unique (const char *undef_name, uint_fast32_t new_hash,
|
||||
_dl_debug_printf ("\
|
||||
marking %s [%lu] as NODELETE due to unique symbol\n",
|
||||
map->l_name, map->l_ns);
|
||||
- if (flags & DL_LOOKUP_FOR_RELOCATE)
|
||||
- map->l_nodelete = link_map_nodelete_pending;
|
||||
- else
|
||||
- map->l_nodelete = link_map_nodelete_active;
|
||||
+ mark_nodelete (map, flags);
|
||||
}
|
||||
}
|
||||
++tab->n_elements;
|
||||
@@ -586,7 +604,7 @@ add_dependency (struct link_map *undef_map, struct link_map *map, int flags)
|
||||
dependencies may pick an dependency which can be dlclose'd, but
|
||||
such IFUNC resolvers are undefined anyway. */
|
||||
assert (map->l_type == lt_loaded);
|
||||
- if (map->l_nodelete != link_map_nodelete_inactive)
|
||||
+ if (is_nodelete (map, flags))
|
||||
return 0;
|
||||
|
||||
struct link_map_reldeps *l_reldeps
|
||||
@@ -694,17 +712,16 @@ add_dependency (struct link_map *undef_map, struct link_map *map, int flags)
|
||||
|
||||
/* Redo the NODELETE check, as when dl_load_lock wasn't held
|
||||
yet this could have changed. */
|
||||
- if (map->l_nodelete != link_map_nodelete_inactive)
|
||||
+ if (is_nodelete (map, flags))
|
||||
goto out;
|
||||
|
||||
/* If the object with the undefined reference cannot be removed ever
|
||||
just make sure the same is true for the object which contains the
|
||||
definition. */
|
||||
- if (undef_map->l_type != lt_loaded
|
||||
- || (undef_map->l_nodelete != link_map_nodelete_inactive))
|
||||
+ if (undef_map->l_type != lt_loaded || is_nodelete (map, flags))
|
||||
{
|
||||
if (__glibc_unlikely (GLRO (dl_debug_mask) & DL_DEBUG_BINDINGS)
|
||||
- && map->l_nodelete == link_map_nodelete_inactive)
|
||||
+ && !is_nodelete (map, flags))
|
||||
{
|
||||
if (undef_map->l_name[0] == '\0')
|
||||
_dl_debug_printf ("\
|
||||
@@ -716,11 +733,7 @@ marking %s [%lu] as NODELETE due to reference to %s [%lu]\n",
|
||||
map->l_name, map->l_ns,
|
||||
undef_map->l_name, undef_map->l_ns);
|
||||
}
|
||||
-
|
||||
- if (flags & DL_LOOKUP_FOR_RELOCATE)
|
||||
- map->l_nodelete = link_map_nodelete_pending;
|
||||
- else
|
||||
- map->l_nodelete = link_map_nodelete_active;
|
||||
+ mark_nodelete (map, flags);
|
||||
goto out;
|
||||
}
|
||||
|
||||
@@ -746,17 +759,14 @@ marking %s [%lu] as NODELETE due to reference to %s [%lu]\n",
|
||||
cannot be unloaded. This is semantically the correct
|
||||
behavior. */
|
||||
if (__glibc_unlikely (GLRO (dl_debug_mask) & DL_DEBUG_BINDINGS)
|
||||
- && map->l_nodelete == link_map_nodelete_inactive)
|
||||
+ && !is_nodelete (map, flags))
|
||||
_dl_debug_printf ("\
|
||||
marking %s [%lu] as NODELETE due to memory allocation failure\n",
|
||||
map->l_name, map->l_ns);
|
||||
- if (flags & DL_LOOKUP_FOR_RELOCATE)
|
||||
- /* In case of non-lazy binding, we could actually
|
||||
- report the memory allocation error, but for now, we
|
||||
- use the conservative approximation as well. */
|
||||
- map->l_nodelete = link_map_nodelete_pending;
|
||||
- else
|
||||
- map->l_nodelete = link_map_nodelete_active;
|
||||
+ /* In case of non-lazy binding, we could actually report
|
||||
+ the memory allocation error, but for now, we use the
|
||||
+ conservative approximation as well. */
|
||||
+ mark_nodelete (map, flags);
|
||||
goto out;
|
||||
}
|
||||
else
|
||||
diff --git a/elf/dl-open.c b/elf/dl-open.c
|
||||
index c7ed85b7ee99a296..a382bfae8aa3a2f8 100644
|
||||
--- a/elf/dl-open.c
|
||||
+++ b/elf/dl-open.c
|
||||
@@ -440,13 +440,17 @@ activate_nodelete (struct link_map *new)
|
||||
NODELETE status for objects outside the local scope. */
|
||||
for (struct link_map *l = GL (dl_ns)[new->l_ns]._ns_loaded; l != NULL;
|
||||
l = l->l_next)
|
||||
- if (l->l_nodelete == link_map_nodelete_pending)
|
||||
+ if (l->l_nodelete_pending)
|
||||
{
|
||||
if (__glibc_unlikely (GLRO (dl_debug_mask) & DL_DEBUG_FILES))
|
||||
_dl_debug_printf ("activating NODELETE for %s [%lu]\n",
|
||||
l->l_name, l->l_ns);
|
||||
|
||||
- l->l_nodelete = link_map_nodelete_active;
|
||||
+ l->l_nodelete_active = true;
|
||||
+
|
||||
+ /* This is just a debugging aid, to indicate that
|
||||
+ activate_nodelete has run for this map. */
|
||||
+ l->l_nodelete_pending = false;
|
||||
}
|
||||
}
|
||||
|
||||
@@ -549,10 +553,10 @@ dl_open_worker (void *a)
|
||||
if (__glibc_unlikely (mode & RTLD_NODELETE))
|
||||
{
|
||||
if (__glibc_unlikely (GLRO (dl_debug_mask) & DL_DEBUG_FILES)
|
||||
- && new->l_nodelete == link_map_nodelete_inactive)
|
||||
+ && !new->l_nodelete_active)
|
||||
_dl_debug_printf ("marking %s [%lu] as NODELETE\n",
|
||||
new->l_name, new->l_ns);
|
||||
- new->l_nodelete = link_map_nodelete_active;
|
||||
+ new->l_nodelete_active = true;
|
||||
}
|
||||
|
||||
/* Finalize the addition to the global scope. */
|
||||
@@ -568,7 +572,7 @@ dl_open_worker (void *a)
|
||||
/* Schedule NODELETE marking for the directly loaded object if
|
||||
requested. */
|
||||
if (__glibc_unlikely (mode & RTLD_NODELETE))
|
||||
- new->l_nodelete = link_map_nodelete_pending;
|
||||
+ new->l_nodelete_pending = true;
|
||||
|
||||
/* Load that object's dependencies. */
|
||||
_dl_map_object_deps (new, NULL, 0, 0,
|
||||
@@ -680,7 +684,7 @@ dl_open_worker (void *a)
|
||||
_dl_start_profile ();
|
||||
|
||||
/* Prevent unloading the object. */
|
||||
- GL(dl_profile_map)->l_nodelete = link_map_nodelete_active;
|
||||
+ GL(dl_profile_map)->l_nodelete_active = true;
|
||||
}
|
||||
}
|
||||
else
|
||||
@@ -879,9 +883,9 @@ no more namespaces available for dlmopen()"));
|
||||
happens inside dl_open_worker. */
|
||||
__libc_signal_restore_set (&args.original_signal_mask);
|
||||
|
||||
- /* All link_map_nodelete_pending objects should have been
|
||||
- deleted at this point, which is why it is not necessary
|
||||
- to reset the flag here. */
|
||||
+ /* All l_nodelete_pending objects should have been deleted
|
||||
+ at this point, which is why it is not necessary to reset
|
||||
+ the flag here. */
|
||||
}
|
||||
else
|
||||
__libc_signal_restore_set (&args.original_signal_mask);
|
||||
diff --git a/elf/get-dynamic-info.h b/elf/get-dynamic-info.h
|
||||
index ea286abaea0128d1..78ba7e76db9706cc 100644
|
||||
--- a/elf/get-dynamic-info.h
|
||||
+++ b/elf/get-dynamic-info.h
|
||||
@@ -164,7 +164,7 @@ elf_get_dynamic_info (struct link_map *l, ElfW(Dyn) *temp)
|
||||
{
|
||||
l->l_flags_1 = info[VERSYMIDX (DT_FLAGS_1)]->d_un.d_val;
|
||||
if (l->l_flags_1 & DF_1_NODELETE)
|
||||
- l->l_nodelete = link_map_nodelete_pending;
|
||||
+ l->l_nodelete_pending = true;
|
||||
|
||||
/* Only DT_1_SUPPORTED_MASK bits are supported, and we would like
|
||||
to assert this, but we can't. Users have been setting
|
||||
diff --git a/include/link.h b/include/link.h
|
||||
index a277b77cad6b52b1..e90fa79a0b332087 100644
|
||||
--- a/include/link.h
|
||||
+++ b/include/link.h
|
||||
@@ -79,22 +79,6 @@ struct r_search_path_struct
|
||||
int malloced;
|
||||
};
|
||||
|
||||
-/* Type used by the l_nodelete member. */
|
||||
-enum link_map_nodelete
|
||||
-{
|
||||
- /* This link map can be deallocated. */
|
||||
- link_map_nodelete_inactive = 0, /* Zero-initialized in _dl_new_object. */
|
||||
-
|
||||
- /* This link map cannot be deallocated. */
|
||||
- link_map_nodelete_active,
|
||||
-
|
||||
- /* This link map cannot be deallocated after dlopen has succeded.
|
||||
- dlopen turns this into link_map_nodelete_active. dlclose treats
|
||||
- this intermediate state as link_map_nodelete_active. */
|
||||
- link_map_nodelete_pending,
|
||||
-};
|
||||
-
|
||||
-
|
||||
/* Structure describing a loaded shared object. The `l_next' and `l_prev'
|
||||
members form a chain of all the shared objects loaded at startup.
|
||||
|
||||
@@ -218,10 +202,17 @@ struct link_map
|
||||
freed, ie. not allocated with
|
||||
the dummy malloc in ld.so. */
|
||||
|
||||
- /* Actually of type enum link_map_nodelete. Separate byte due to
|
||||
- a read in add_dependency in elf/dl-lookup.c outside the loader
|
||||
- lock. Only valid for l_type == lt_loaded. */
|
||||
- unsigned char l_nodelete;
|
||||
+ /* NODELETE status of the map. Only valid for maps of type
|
||||
+ lt_loaded. Lazy binding sets l_nodelete_active directly,
|
||||
+ potentially from signal handlers. Initial loading of an
|
||||
+ DF_1_NODELETE object set l_nodelete_pending. Relocation may
|
||||
+ set l_nodelete_pending as well. l_nodelete_pending maps are
|
||||
+ promoted to l_nodelete_active status in the final stages of
|
||||
+ dlopen, prior to calling ELF constructors. dlclose only
|
||||
+ refuses to unload l_nodelete_active maps, the pending status is
|
||||
+ ignored. */
|
||||
+ bool l_nodelete_active;
|
||||
+ bool l_nodelete_pending;
|
||||
|
||||
#include <link_map.h>
|
||||
|
||||
@ -1,134 +0,0 @@
|
||||
commit f7649d5780aa4682393b9daedd653e4d9c12784c
|
||||
Author: Florian Weimer <fweimer@redhat.com>
|
||||
Date: Fri Dec 13 10:23:10 2019 +0100
|
||||
|
||||
dlopen: Do not block signals
|
||||
|
||||
Blocking signals causes issues with certain anti-malware solutions
|
||||
which rely on an unblocked SIGSYS signal for system calls they
|
||||
intercept.
|
||||
|
||||
This reverts commit a2e8aa0d9ea648068d8be52dd7b15f1b6a008e23
|
||||
("Block signals during the initial part of dlopen") and adds
|
||||
comments related to async signal safety to active_nodelete and
|
||||
its caller.
|
||||
|
||||
Note that this does not make lazy binding async-signal-safe with regards
|
||||
to dlopen. It merely avoids introducing new async-signal-safety hazards
|
||||
as part of the NODELETE changes.
|
||||
|
||||
Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
|
||||
Reviewed-by: Carlos O'Donell <carlos@redhat.com>
|
||||
|
||||
diff --git a/elf/dl-open.c b/elf/dl-open.c
|
||||
index a382bfae8aa3a2f8..d834b89754d2b073 100644
|
||||
--- a/elf/dl-open.c
|
||||
+++ b/elf/dl-open.c
|
||||
@@ -34,7 +34,6 @@
|
||||
#include <atomic.h>
|
||||
#include <libc-internal.h>
|
||||
#include <array_length.h>
|
||||
-#include <internal-signals.h>
|
||||
|
||||
#include <dl-dst.h>
|
||||
#include <dl-prop.h>
|
||||
@@ -53,10 +52,6 @@ struct dl_open_args
|
||||
/* Namespace ID. */
|
||||
Lmid_t nsid;
|
||||
|
||||
- /* Original signal mask. Used for unblocking signal handlers before
|
||||
- running ELF constructors. */
|
||||
- sigset_t original_signal_mask;
|
||||
-
|
||||
/* Original value of _ns_global_scope_pending_adds. Set by
|
||||
dl_open_worker. Only valid if nsid is a real namespace
|
||||
(non-negative). */
|
||||
@@ -446,6 +441,9 @@ activate_nodelete (struct link_map *new)
|
||||
_dl_debug_printf ("activating NODELETE for %s [%lu]\n",
|
||||
l->l_name, l->l_ns);
|
||||
|
||||
+ /* The flag can already be true at this point, e.g. a signal
|
||||
+ handler may have triggered lazy binding and set NODELETE
|
||||
+ status immediately. */
|
||||
l->l_nodelete_active = true;
|
||||
|
||||
/* This is just a debugging aid, to indicate that
|
||||
@@ -520,16 +518,12 @@ dl_open_worker (void *a)
|
||||
if (new == NULL)
|
||||
{
|
||||
assert (mode & RTLD_NOLOAD);
|
||||
- __libc_signal_restore_set (&args->original_signal_mask);
|
||||
return;
|
||||
}
|
||||
|
||||
if (__glibc_unlikely (mode & __RTLD_SPROF))
|
||||
- {
|
||||
- /* This happens only if we load a DSO for 'sprof'. */
|
||||
- __libc_signal_restore_set (&args->original_signal_mask);
|
||||
- return;
|
||||
- }
|
||||
+ /* This happens only if we load a DSO for 'sprof'. */
|
||||
+ return;
|
||||
|
||||
/* This object is directly loaded. */
|
||||
++new->l_direct_opencount;
|
||||
@@ -565,7 +559,6 @@ dl_open_worker (void *a)
|
||||
|
||||
assert (_dl_debug_initialize (0, args->nsid)->r_state == RT_CONSISTENT);
|
||||
|
||||
- __libc_signal_restore_set (&args->original_signal_mask);
|
||||
return;
|
||||
}
|
||||
|
||||
@@ -709,6 +702,12 @@ dl_open_worker (void *a)
|
||||
All memory allocations for new objects must have happened
|
||||
before. */
|
||||
|
||||
+ /* Finalize the NODELETE status first. This comes before
|
||||
+ update_scopes, so that lazy binding will not see pending NODELETE
|
||||
+ state for newly loaded objects. There is a compiler barrier in
|
||||
+ update_scopes which ensures that the changes from
|
||||
+ activate_nodelete are visible before new objects show up in the
|
||||
+ local scope. */
|
||||
activate_nodelete (new);
|
||||
|
||||
/* Second stage after resize_scopes: Actually perform the scope
|
||||
@@ -742,10 +741,6 @@ dl_open_worker (void *a)
|
||||
if (mode & RTLD_GLOBAL)
|
||||
add_to_global_resize (new);
|
||||
|
||||
- /* Unblock signals. Data structures are now consistent, and
|
||||
- application code may run. */
|
||||
- __libc_signal_restore_set (&args->original_signal_mask);
|
||||
-
|
||||
/* Run the initializer functions of new objects. Temporarily
|
||||
disable the exception handler, so that lazy binding failures are
|
||||
fatal. */
|
||||
@@ -835,10 +830,6 @@ no more namespaces available for dlmopen()"));
|
||||
args.argv = argv;
|
||||
args.env = env;
|
||||
|
||||
- /* Recursive lazy binding during manipulation of the dynamic loader
|
||||
- structures may result in incorrect behavior. */
|
||||
- __libc_signal_block_all (&args.original_signal_mask);
|
||||
-
|
||||
struct dl_exception exception;
|
||||
int errcode = _dl_catch_exception (&exception, dl_open_worker, &args);
|
||||
|
||||
@@ -879,16 +870,10 @@ no more namespaces available for dlmopen()"));
|
||||
|
||||
_dl_close_worker (args.map, true);
|
||||
|
||||
- /* Restore the signal mask. In the success case, this
|
||||
- happens inside dl_open_worker. */
|
||||
- __libc_signal_restore_set (&args.original_signal_mask);
|
||||
-
|
||||
/* All l_nodelete_pending objects should have been deleted
|
||||
at this point, which is why it is not necessary to reset
|
||||
the flag here. */
|
||||
}
|
||||
- else
|
||||
- __libc_signal_restore_set (&args.original_signal_mask);
|
||||
|
||||
assert (_dl_debug_initialize (0, args.nsid)->r_state == RT_CONSISTENT);
|
||||
|
||||
Some files were not shown because too many files have changed in this diff Show More
Loading…
Reference in New Issue
Block a user