valgrind/valgrind-3.17.0-s390-z15.patch
Mark Wielaard 4a7f098214 3.17.0-6 - z15 support
Add valgrind-3.17.0-s390-prep.patch
Add valgrind-3.17.0-s390-z15.patch
Add valgrind-3.17.0-s390-z13-vec-fix.patch

Resolves: #1920584
Valgrind support of IBM Z hardware z15
2021-06-24 18:53:27 +02:00

2414 lines
90 KiB
Diff

From 3fbde55a5696c9273084ee2c44daca752e407597 Mon Sep 17 00:00:00 2001
From: Andreas Arnez <arnez@linux.ibm.com>
Date: Tue, 26 Jan 2021 15:06:47 +0100
Subject: [PATCH 01/13] s390x: Misc-insn-3, bitwise logical 3-way instructions
Add support for the instructions NCRK, NCGRK, NNRK, NNGRK, NORK, NOGRK,
NXRK, NXGRK, OCRK, and OCGRK. Introduce a common helper and use it for
the existing instructions NRK, NGRK, XRK, XGRK, ORK, and OGRK as well.
---
VEX/priv/guest_s390_toIR.c | 154 ++++++++++++++++++++++++++-----------
1 file changed, 109 insertions(+), 45 deletions(-)
diff --git a/VEX/priv/guest_s390_toIR.c b/VEX/priv/guest_s390_toIR.c
index a73dcfb14..f8afd5b96 100644
--- a/VEX/priv/guest_s390_toIR.c
+++ b/VEX/priv/guest_s390_toIR.c
@@ -5022,8 +5022,12 @@ s390_irgen_NGR(UChar r1, UChar r2)
return "ngr";
}
+/* Helper for bitwise logical instructions with two 32-bit input operands and a
+ 32-bit output operand. `inv3' and `inv' indicate whether to invert (build
+ bitwise complement of) operand 3 or the result, respectively. */
static const HChar *
-s390_irgen_NRK(UChar r3, UChar r1, UChar r2)
+s390_irgen_logicalK32(UChar r3, UChar r1, UChar r2,
+ const HChar *mnem, IROp op, Bool inv3, Bool inv)
{
IRTemp op2 = newTemp(Ity_I32);
IRTemp op3 = newTemp(Ity_I32);
@@ -5031,15 +5035,19 @@ s390_irgen_NRK(UChar r3, UChar r1, UChar r2)
assign(op2, get_gpr_w1(r2));
assign(op3, get_gpr_w1(r3));
- assign(result, binop(Iop_And32, mkexpr(op2), mkexpr(op3)));
+ IRExpr* tmp = binop(op, mkexpr(op2),
+ inv3 ? unop(Iop_Not32, mkexpr(op3)) : mkexpr(op3));
+ assign(result, inv ? unop(Iop_Not32, tmp) : tmp);
s390_cc_thunk_putZ(S390_CC_OP_BITWISE, result);
put_gpr_w1(r1, mkexpr(result));
- return "nrk";
+ return mnem;
}
+/* Same as s390_irgen_logicalK32, but for 64-bit operands. */
static const HChar *
-s390_irgen_NGRK(UChar r3, UChar r1, UChar r2)
+s390_irgen_logicalK64(UChar r3, UChar r1, UChar r2,
+ const HChar *mnem, IROp op, Bool inv3, Bool inv)
{
IRTemp op2 = newTemp(Ity_I64);
IRTemp op3 = newTemp(Ity_I64);
@@ -5047,11 +5055,49 @@ s390_irgen_NGRK(UChar r3, UChar r1, UChar r2)
assign(op2, get_gpr_dw0(r2));
assign(op3, get_gpr_dw0(r3));
- assign(result, binop(Iop_And64, mkexpr(op2), mkexpr(op3)));
+ IRExpr* tmp = binop(op, mkexpr(op2),
+ inv3 ? unop(Iop_Not64, mkexpr(op3)) : mkexpr(op3));
+ assign(result, inv ? unop(Iop_Not64, tmp) : tmp);
s390_cc_thunk_putZ(S390_CC_OP_BITWISE, result);
put_gpr_dw0(r1, mkexpr(result));
- return "ngrk";
+ return mnem;
+}
+
+static const HChar *
+s390_irgen_NRK(UChar r3, UChar r1, UChar r2)
+{
+ return s390_irgen_logicalK32(r3, r1, r2, "nrk", Iop_And32, False, False);
+}
+
+static const HChar *
+s390_irgen_NGRK(UChar r3, UChar r1, UChar r2)
+{
+ return s390_irgen_logicalK64(r3, r1, r2, "ngrk", Iop_And64, False, False);
+}
+
+static const HChar *
+s390_irgen_NCRK(UChar r3, UChar r1, UChar r2)
+{
+ return s390_irgen_logicalK32(r3, r1, r2, "ncrk", Iop_And32, True, False);
+}
+
+static const HChar *
+s390_irgen_NCGRK(UChar r3, UChar r1, UChar r2)
+{
+ return s390_irgen_logicalK64(r3, r1, r2, "ncgrk", Iop_And64, True, False);
+}
+
+static const HChar *
+s390_irgen_NNRK(UChar r3, UChar r1, UChar r2)
+{
+ return s390_irgen_logicalK32(r3, r1, r2, "nnrk", Iop_And32, False, True);
+}
+
+static const HChar *
+s390_irgen_NNGRK(UChar r3, UChar r1, UChar r2)
+{
+ return s390_irgen_logicalK64(r3, r1, r2, "nngrk", Iop_And64, False, True);
}
static const HChar *
@@ -7071,33 +7117,25 @@ s390_irgen_XGR(UChar r1, UChar r2)
static const HChar *
s390_irgen_XRK(UChar r3, UChar r1, UChar r2)
{
- IRTemp op2 = newTemp(Ity_I32);
- IRTemp op3 = newTemp(Ity_I32);
- IRTemp result = newTemp(Ity_I32);
-
- assign(op2, get_gpr_w1(r2));
- assign(op3, get_gpr_w1(r3));
- assign(result, binop(Iop_Xor32, mkexpr(op2), mkexpr(op3)));
- s390_cc_thunk_putZ(S390_CC_OP_BITWISE, result);
- put_gpr_w1(r1, mkexpr(result));
-
- return "xrk";
+ return s390_irgen_logicalK32(r3, r1, r2, "xrk", Iop_Xor32, False, False);
}
static const HChar *
s390_irgen_XGRK(UChar r3, UChar r1, UChar r2)
{
- IRTemp op2 = newTemp(Ity_I64);
- IRTemp op3 = newTemp(Ity_I64);
- IRTemp result = newTemp(Ity_I64);
+ return s390_irgen_logicalK64(r3, r1, r2, "xgrk", Iop_Xor64, False, False);
+}
- assign(op2, get_gpr_dw0(r2));
- assign(op3, get_gpr_dw0(r3));
- assign(result, binop(Iop_Xor64, mkexpr(op2), mkexpr(op3)));
- s390_cc_thunk_putZ(S390_CC_OP_BITWISE, result);
- put_gpr_dw0(r1, mkexpr(result));
+static const HChar *
+s390_irgen_NXRK(UChar r3, UChar r1, UChar r2)
+{
+ return s390_irgen_logicalK32(r3, r1, r2, "nxrk", Iop_Xor32, False, True);
+}
- return "xgrk";
+static const HChar *
+s390_irgen_NXGRK(UChar r3, UChar r1, UChar r2)
+{
+ return s390_irgen_logicalK64(r3, r1, r2, "nxgrk", Iop_Xor64, False, True);
}
static const HChar *
@@ -8920,33 +8958,37 @@ s390_irgen_OGR(UChar r1, UChar r2)
static const HChar *
s390_irgen_ORK(UChar r3, UChar r1, UChar r2)
{
- IRTemp op2 = newTemp(Ity_I32);
- IRTemp op3 = newTemp(Ity_I32);
- IRTemp result = newTemp(Ity_I32);
+ return s390_irgen_logicalK32(r3, r1, r2, "ork", Iop_Or32, False, False);
+}
- assign(op2, get_gpr_w1(r2));
- assign(op3, get_gpr_w1(r3));
- assign(result, binop(Iop_Or32, mkexpr(op2), mkexpr(op3)));
- s390_cc_thunk_putZ(S390_CC_OP_BITWISE, result);
- put_gpr_w1(r1, mkexpr(result));
+static const HChar *
+s390_irgen_OGRK(UChar r3, UChar r1, UChar r2)
+{
+ return s390_irgen_logicalK64(r3, r1, r2, "ogrk", Iop_Or64, False, False);
+}
- return "ork";
+static const HChar *
+s390_irgen_OCRK(UChar r3, UChar r1, UChar r2)
+{
+ return s390_irgen_logicalK32(r3, r1, r2, "ocrk", Iop_Or32, True, False);
}
static const HChar *
-s390_irgen_OGRK(UChar r3, UChar r1, UChar r2)
+s390_irgen_OCGRK(UChar r3, UChar r1, UChar r2)
{
- IRTemp op2 = newTemp(Ity_I64);
- IRTemp op3 = newTemp(Ity_I64);
- IRTemp result = newTemp(Ity_I64);
+ return s390_irgen_logicalK64(r3, r1, r2, "ocgrk", Iop_Or64, True, False);
+}
- assign(op2, get_gpr_dw0(r2));
- assign(op3, get_gpr_dw0(r3));
- assign(result, binop(Iop_Or64, mkexpr(op2), mkexpr(op3)));
- s390_cc_thunk_putZ(S390_CC_OP_BITWISE, result);
- put_gpr_dw0(r1, mkexpr(result));
+static const HChar *
+s390_irgen_NORK(UChar r3, UChar r1, UChar r2)
+{
+ return s390_irgen_logicalK32(r3, r1, r2, "nork", Iop_Or32, False, True);
+}
- return "ogrk";
+static const HChar *
+s390_irgen_NOGRK(UChar r3, UChar r1, UChar r2)
+{
+ return s390_irgen_logicalK64(r3, r1, r2, "nogrk", Iop_Or64, False, True);
}
static const HChar *
@@ -20031,12 +20073,28 @@ s390_decode_4byte_and_irgen(const UChar *bytes)
case 0xb961: s390_format_RRF_U0RR(s390_irgen_CLGRT, RRF2_m3(ovl),
RRF2_r1(ovl), RRF2_r2(ovl),
S390_XMNM_CAB); goto ok;
+ case 0xb964: s390_format_RRF_R0RR2(s390_irgen_NNGRK, RRF4_r3(ovl),
+ RRF4_r1(ovl), RRF4_r2(ovl)); goto ok;
+ case 0xb965: s390_format_RRF_R0RR2(s390_irgen_OCGRK, RRF4_r3(ovl),
+ RRF4_r1(ovl), RRF4_r2(ovl)); goto ok;
+ case 0xb966: s390_format_RRF_R0RR2(s390_irgen_NOGRK, RRF4_r3(ovl),
+ RRF4_r1(ovl), RRF4_r2(ovl)); goto ok;
+ case 0xb967: s390_format_RRF_R0RR2(s390_irgen_NXGRK, RRF4_r3(ovl),
+ RRF4_r1(ovl), RRF4_r2(ovl)); goto ok;
case 0xb972: s390_format_RRF_U0RR(s390_irgen_CRT, RRF2_m3(ovl),
RRF2_r1(ovl), RRF2_r2(ovl),
S390_XMNM_CAB); goto ok;
case 0xb973: s390_format_RRF_U0RR(s390_irgen_CLRT, RRF2_m3(ovl),
RRF2_r1(ovl), RRF2_r2(ovl),
S390_XMNM_CAB); goto ok;
+ case 0xb974: s390_format_RRF_R0RR2(s390_irgen_NNRK, RRF4_r3(ovl),
+ RRF4_r1(ovl), RRF4_r2(ovl)); goto ok;
+ case 0xb975: s390_format_RRF_R0RR2(s390_irgen_OCRK, RRF4_r3(ovl),
+ RRF4_r1(ovl), RRF4_r2(ovl)); goto ok;
+ case 0xb976: s390_format_RRF_R0RR2(s390_irgen_NORK, RRF4_r3(ovl),
+ RRF4_r1(ovl), RRF4_r2(ovl)); goto ok;
+ case 0xb977: s390_format_RRF_R0RR2(s390_irgen_NXRK, RRF4_r3(ovl),
+ RRF4_r1(ovl), RRF4_r2(ovl)); goto ok;
case 0xb980: s390_format_RRE_RR(s390_irgen_NGR, RRE_r1(ovl),
RRE_r2(ovl)); goto ok;
case 0xb981: s390_format_RRE_RR(s390_irgen_OGR, RRE_r1(ovl),
@@ -20148,6 +20206,9 @@ s390_decode_4byte_and_irgen(const UChar *bytes)
case 0xb9e4: s390_format_RRF_R0RR2(s390_irgen_NGRK, RRF4_r3(ovl),
RRF4_r1(ovl), RRF4_r2(ovl));
goto ok;
+ case 0xb9e5: s390_format_RRF_R0RR2(s390_irgen_NCGRK, RRF4_r3(ovl),
+ RRF4_r1(ovl), RRF4_r2(ovl));
+ goto ok;
case 0xb9e6: s390_format_RRF_R0RR2(s390_irgen_OGRK, RRF4_r3(ovl),
RRF4_r1(ovl), RRF4_r2(ovl));
goto ok;
@@ -20178,6 +20239,9 @@ s390_decode_4byte_and_irgen(const UChar *bytes)
case 0xb9f4: s390_format_RRF_R0RR2(s390_irgen_NRK, RRF4_r3(ovl),
RRF4_r1(ovl), RRF4_r2(ovl));
goto ok;
+ case 0xb9f5: s390_format_RRF_R0RR2(s390_irgen_NCRK, RRF4_r3(ovl),
+ RRF4_r1(ovl), RRF4_r2(ovl));
+ goto ok;
case 0xb9f6: s390_format_RRF_R0RR2(s390_irgen_ORK, RRF4_r3(ovl),
RRF4_r1(ovl), RRF4_r2(ovl));
goto ok;
--
2.23.0
From 748421b31ab6b15cc849bd6b9588ad759b807324 Mon Sep 17 00:00:00 2001
From: Andreas Arnez <arnez@linux.ibm.com>
Date: Wed, 27 Jan 2021 18:11:06 +0100
Subject: [PATCH 02/13] s390x: Misc-insn-3, "select" instructions
Add support for the instructions SELR, SELGR, and SELFHR.
---
VEX/priv/guest_s390_toIR.c | 43 ++++++++++++++++++++++++++++++++++++++
1 file changed, 43 insertions(+)
diff --git a/VEX/priv/guest_s390_toIR.c b/VEX/priv/guest_s390_toIR.c
index f8afd5b96..41265631b 100644
--- a/VEX/priv/guest_s390_toIR.c
+++ b/VEX/priv/guest_s390_toIR.c
@@ -3113,6 +3113,16 @@ s390_format_RRF_FUFF2(const HChar *(*irgen)(UChar, UChar, UChar, UChar),
s390_disasm(ENC5(MNM, FPR, FPR, FPR, UINT), mnm, r1, r2, r3, m4);
}
+static void
+s390_format_RRF_RURR(const HChar *(*irgen)(UChar, UChar, UChar, UChar),
+ UChar r3, UChar m4, UChar r1, UChar r2)
+{
+ const HChar *mnm = irgen(r3, m4, r1, r2);
+
+ if (UNLIKELY(vex_traceflags & VEX_TRACE_FE))
+ s390_disasm(ENC5(MNM, GPR, GPR, GPR, UINT), mnm, r1, r3, r2, m4);
+}
+
static void
s390_format_RRF_R0RR2(const HChar *(*irgen)(UChar r3, UChar r1, UChar r2),
UChar r3, UChar r1, UChar r2)
@@ -19254,6 +19264,30 @@ s390_irgen_VBPERM(UChar v1, UChar v2, UChar v3)
return "vbperm";
}
+static const HChar *
+s390_irgen_SELR(UChar r3, UChar m4, UChar r1, UChar r2)
+{
+ IRExpr* cond = binop(Iop_CmpNE32, s390_call_calculate_cond(m4), mkU32(0));
+ put_gpr_w1(r1, mkite(cond, get_gpr_w1(r2), get_gpr_w1(r3)));
+ return "selr";
+}
+
+static const HChar *
+s390_irgen_SELGR(UChar r3, UChar m4, UChar r1, UChar r2)
+{
+ IRExpr* cond = binop(Iop_CmpNE32, s390_call_calculate_cond(m4), mkU32(0));
+ put_gpr_dw0(r1, mkite(cond, get_gpr_dw0(r2), get_gpr_dw0(r3)));
+ return "selgr";
+}
+
+static const HChar *
+s390_irgen_SELFHR(UChar r3, UChar m4, UChar r1, UChar r2)
+{
+ IRExpr* cond = binop(Iop_CmpNE32, s390_call_calculate_cond(m4), mkU32(0));
+ put_gpr_w0(r1, mkite(cond, get_gpr_w0(r2), get_gpr_w0(r3)));
+ return "selfhr";
+}
+
/* New insns are added here.
If an insn is contingent on a facility being installed also
check whether the list of supported facilities in function
@@ -20163,6 +20197,9 @@ s390_decode_4byte_and_irgen(const UChar *bytes)
case 0xb9bd: /* TRTRE */ goto unimplemented;
case 0xb9be: /* SRSTU */ goto unimplemented;
case 0xb9bf: /* TRTE */ goto unimplemented;
+ case 0xb9c0: s390_format_RRF_RURR(s390_irgen_SELFHR, RRF4_r3(ovl),
+ RRF4_m4(ovl), RRF4_r1(ovl),
+ RRF4_r2(ovl)); goto ok;
case 0xb9c8: s390_format_RRF_R0RR2(s390_irgen_AHHHR, RRF4_r3(ovl),
RRF4_r1(ovl), RRF4_r2(ovl));
goto ok;
@@ -20203,6 +20240,9 @@ s390_decode_4byte_and_irgen(const UChar *bytes)
case 0xb9e2: s390_format_RRF_U0RR(s390_irgen_LOCGR, RRF3_r3(ovl),
RRF3_r1(ovl), RRF3_r2(ovl),
S390_XMNM_LOCGR); goto ok;
+ case 0xb9e3: s390_format_RRF_RURR(s390_irgen_SELGR, RRF4_r3(ovl),
+ RRF4_m4(ovl), RRF4_r1(ovl),
+ RRF4_r2(ovl)); goto ok;
case 0xb9e4: s390_format_RRF_R0RR2(s390_irgen_NGRK, RRF4_r3(ovl),
RRF4_r1(ovl), RRF4_r2(ovl));
goto ok;
@@ -20233,6 +20273,9 @@ s390_decode_4byte_and_irgen(const UChar *bytes)
case 0xb9ed: s390_format_RRF_R0RR2(s390_irgen_MSGRKC, RRF4_r3(ovl),
RRF4_r1(ovl), RRF4_r2(ovl));
goto ok;
+ case 0xb9f0: s390_format_RRF_RURR(s390_irgen_SELR, RRF4_r3(ovl),
+ RRF4_m4(ovl), RRF4_r1(ovl),
+ RRF4_r2(ovl)); goto ok;
case 0xb9f2: s390_format_RRF_U0RR(s390_irgen_LOCR, RRF3_r3(ovl),
RRF3_r1(ovl), RRF3_r2(ovl),
S390_XMNM_LOCR); goto ok;
--
2.23.0
From 31cbd583e858f47a86ada087d21a6abc13ba04f2 Mon Sep 17 00:00:00 2001
From: Andreas Arnez <arnez@linux.ibm.com>
Date: Thu, 28 Jan 2021 19:47:00 +0100
Subject: [PATCH 03/13] s390x: Misc-insn-3, new POPCNT variant
Add support for the new POPCNT variant that has bit 0 of the M3 field set
and yields the total number of one bits in its 64-bit operand.
---
VEX/priv/guest_s390_toIR.c | 44 ++++++++++++++++++++++++++------------
1 file changed, 30 insertions(+), 14 deletions(-)
diff --git a/VEX/priv/guest_s390_toIR.c b/VEX/priv/guest_s390_toIR.c
index 41265631b..ca9e6dc03 100644
--- a/VEX/priv/guest_s390_toIR.c
+++ b/VEX/priv/guest_s390_toIR.c
@@ -3073,6 +3073,20 @@ s390_format_RRF_U0RR(const HChar *(*irgen)(UChar m3, UChar r1, UChar r2),
s390_disasm(ENC3(XMNM, GPR, GPR), xmnm_kind, m3, r1, r2);
}
+static void
+s390_format_RRFa_U0RR(const HChar *(*irgen)(UChar m3, UChar r1, UChar r2),
+ UChar m3, UChar r1, UChar r2)
+{
+ const HChar *mnm = irgen(m3, r1, r2);
+
+ if (UNLIKELY(vex_traceflags & VEX_TRACE_FE)) {
+ if (m3 != 0)
+ s390_disasm(ENC4(MNM, GPR, GPR, UINT), mnm, r1, r2, m3);
+ else
+ s390_disasm(ENC3(MNM, GPR, GPR), mnm, r1, r2);
+ }
+}
+
static void
s390_format_RRF_F0FF2(const HChar *(*irgen)(UChar, UChar, UChar),
UChar r3, UChar r1, UChar r2)
@@ -15112,30 +15126,32 @@ s390_irgen_FLOGR(UChar r1, UChar r2)
}
static const HChar *
-s390_irgen_POPCNT(UChar r1, UChar r2)
+s390_irgen_POPCNT(UChar m3, UChar r1, UChar r2)
{
- Int i;
+ s390_insn_assert("popcnt", (m3 & 7) == 0);
+
+ static const ULong masks[] = {
+ 0x5555555555555555, 0x3333333333333333, 0x0F0F0F0F0F0F0F0F,
+ 0x00FF00FF00FF00FF, 0x0000FFFF0000FFFF, 0x00000000FFFFFFFF,
+ };
+ Int i, n;
IRTemp val = newTemp(Ity_I64);
- IRTemp mask[3];
assign(val, get_gpr_dw0(r2));
- for (i = 0; i < 3; i++) {
- mask[i] = newTemp(Ity_I64);
- }
- assign(mask[0], mkU64(0x5555555555555555ULL));
- assign(mask[1], mkU64(0x3333333333333333ULL));
- assign(mask[2], mkU64(0x0F0F0F0F0F0F0F0FULL));
- for (i = 0; i < 3; i++) {
+ n = (m3 & 8) ? 6 : 3;
+ for (i = 0; i < n; i++) {
+ IRTemp mask = newTemp(Ity_I64);
IRTemp tmp = newTemp(Ity_I64);
+ assign (mask, mkU64(masks[i]));
assign(tmp,
binop(Iop_Add64,
binop(Iop_And64,
mkexpr(val),
- mkexpr(mask[i])),
+ mkexpr(mask)),
binop(Iop_And64,
binop(Iop_Shr64, mkexpr(val), mkU8(1 << i)),
- mkexpr(mask[i]))));
+ mkexpr(mask))));
val = tmp;
}
s390_cc_thunk_putZ(S390_CC_OP_BITWISE, val);
@@ -20235,8 +20251,8 @@ s390_decode_4byte_and_irgen(const UChar *bytes)
case 0xb9e0: s390_format_RRF_U0RR(s390_irgen_LOCFHR, RRF3_r3(ovl),
RRF3_r1(ovl), RRF3_r2(ovl),
S390_XMNM_LOCFHR); goto ok;
- case 0xb9e1: s390_format_RRE_RR(s390_irgen_POPCNT, RRE_r1(ovl),
- RRE_r2(ovl)); goto ok;
+ case 0xb9e1: s390_format_RRFa_U0RR(s390_irgen_POPCNT, RRF3_r3(ovl),
+ RRF3_r1(ovl), RRF3_r2(ovl)); goto ok;
case 0xb9e2: s390_format_RRF_U0RR(s390_irgen_LOCGR, RRF3_r3(ovl),
RRF3_r1(ovl), RRF3_r2(ovl),
S390_XMNM_LOCGR); goto ok;
--
2.23.0
From 64352d57f93711ce76fd481558dcf6d65e26b19f Mon Sep 17 00:00:00 2001
From: Andreas Arnez <arnez@linux.ibm.com>
Date: Fri, 29 Jan 2021 20:13:05 +0100
Subject: [PATCH 04/13] s390x: Misc-insn-3, MVCRL
Add support for the "move right to left" instruction MVCRL.
---
VEX/priv/guest_s390_toIR.c | 47 ++++++++++++++++++++++++++++++++++++++
1 file changed, 47 insertions(+)
diff --git a/VEX/priv/guest_s390_toIR.c b/VEX/priv/guest_s390_toIR.c
index ca9e6dc03..9f7d98f8c 100644
--- a/VEX/priv/guest_s390_toIR.c
+++ b/VEX/priv/guest_s390_toIR.c
@@ -3562,6 +3562,25 @@ s390_format_SS_L0RDRD(const HChar *(*irgen)(UChar, IRTemp, IRTemp),
s390_disasm(ENC3(MNM, UDLB, UDXB), mnm, d1, l, b1, d2, 0, b2);
}
+static void
+s390_format_SSE_RDRD(const HChar *(*irgen)(IRTemp, IRTemp),
+ UChar b1, UShort d1, UChar b2, UShort d2)
+{
+ const HChar *mnm;
+ IRTemp op1addr = newTemp(Ity_I64);
+ IRTemp op2addr = newTemp(Ity_I64);
+
+ assign(op1addr, binop(Iop_Add64, mkU64(d1), b1 != 0 ? get_gpr_dw0(b1) :
+ mkU64(0)));
+ assign(op2addr, binop(Iop_Add64, mkU64(d2), b2 != 0 ? get_gpr_dw0(b2) :
+ mkU64(0)));
+
+ mnm = irgen(op1addr, op2addr);
+
+ if (UNLIKELY(vex_traceflags & VEX_TRACE_FE))
+ s390_disasm(ENC2(UDXB, UDXB), mnm, d1, 0, b1, d2, 0, b2);
+}
+
static void
s390_format_SIL_RDI(const HChar *(*irgen)(UShort i2, IRTemp op1addr),
UChar b1, UShort d1, UShort i2)
@@ -13667,6 +13686,31 @@ s390_irgen_MVCIN(UChar length, IRTemp start1, IRTemp start2)
return "mvcin";
}
+static const HChar *
+s390_irgen_MVCRL(IRTemp op1addr, IRTemp op2addr)
+{
+ IRTemp counter = newTemp(Ity_I64);
+ IRTemp offset = newTemp(Ity_I64);
+
+ assign(counter, get_counter_dw0());
+ /* offset = length - 1 - counter, where length-1 is specified in r0 */
+ assign(offset,
+ binop(Iop_Sub64,
+ unop(Iop_16Uto64,
+ binop(Iop_And16, get_gpr_hw3(0), mkU16(0xfff))),
+ mkexpr(counter)));
+
+ store(binop(Iop_Add64, mkexpr(op1addr), mkexpr(offset)),
+ load(Ity_I8, binop(Iop_Add64, mkexpr(op2addr), mkexpr(offset))));
+
+ /* Check for end of field */
+ put_counter_dw0(binop(Iop_Add64, mkexpr(counter), mkU64(1)));
+ iterate_if(binop(Iop_CmpNE64, mkexpr(offset), mkU64(0)));
+ put_counter_dw0(mkU64(0));
+
+ return "mvcrl";
+}
+
static const HChar *
s390_irgen_MVCL(UChar r1, UChar r2)
{
@@ -22217,6 +22261,9 @@ s390_decode_6byte_and_irgen(const UChar *bytes)
case 0xe500ULL: /* LASP */ goto unimplemented;
case 0xe501ULL: /* TPROT */ goto unimplemented;
case 0xe502ULL: /* STRAG */ goto unimplemented;
+ case 0xe50aULL: s390_format_SSE_RDRD(s390_irgen_MVCRL,
+ SS_b1(ovl), SS_d1(ovl),
+ SS_b2(ovl), SS_d2(ovl)); goto ok;
case 0xe50eULL: /* MVCSK */ goto unimplemented;
case 0xe50fULL: /* MVCDK */ goto unimplemented;
case 0xe544ULL: s390_format_SIL_RDI(s390_irgen_MVHHI, SIL_b1(ovl),
--
2.23.0
From 6cc4d66cc3a999253d9a57e2b5c75aeb67f77918 Mon Sep 17 00:00:00 2001
From: Andreas Arnez <arnez@linux.ibm.com>
Date: Tue, 2 Feb 2021 20:15:02 +0100
Subject: [PATCH 05/13] s390x: Misc-insn-3, test case
Add a test case for the new instructions in the miscellaneous instruction
extensions facitility 3.
---
.gitignore | 1 +
none/tests/s390x/Makefile.am | 3 +-
none/tests/s390x/misc3.c | 182 ++++++++++++++++++++++++++++++
none/tests/s390x/misc3.stderr.exp | 2 +
none/tests/s390x/misc3.stdout.exp | 103 +++++++++++++++++
none/tests/s390x/misc3.vgtest | 1 +
6 files changed, 291 insertions(+), 1 deletion(-)
create mode 100644 none/tests/s390x/misc3.c
create mode 100644 none/tests/s390x/misc3.stderr.exp
create mode 100644 none/tests/s390x/misc3.stdout.exp
create mode 100644 none/tests/s390x/misc3.vgtest
diff --git a/none/tests/s390x/Makefile.am b/none/tests/s390x/Makefile.am
index a0fb92ef5..2fd45ec1e 100644
--- a/none/tests/s390x/Makefile.am
+++ b/none/tests/s390x/Makefile.am
@@ -19,7 +19,8 @@ INSN_TESTS = clc clcle cvb cvd icm lpr tcxb lam_stam xc mvst add sub mul \
spechelper-ltr spechelper-or \
spechelper-icm-1 spechelper-icm-2 spechelper-tmll \
spechelper-tm laa vector lsc2 ppno vector_string vector_integer \
- vector_float add-z14 sub-z14 mul-z14 bic
+ vector_float add-z14 sub-z14 mul-z14 bic \
+ misc3
if BUILD_DFP_TESTS
INSN_TESTS += dfp-1 dfp-2 dfp-3 dfp-4 dfptest dfpext dfpconv srnmt pfpo
diff --git a/none/tests/s390x/misc3.c b/none/tests/s390x/misc3.c
new file mode 100644
index 000000000..ae6e8d4c2
--- /dev/null
+++ b/none/tests/s390x/misc3.c
@@ -0,0 +1,182 @@
+#include <stdio.h>
+
+/* -- Logical instructions -- */
+
+#define TEST_GENERATE(opcode,insn) \
+ static void test_##insn(unsigned long a, unsigned long b) \
+ { \
+ unsigned long out = 0xdecaffee42424242; \
+ int cc; \
+ \
+ __asm__( \
+ "cr 0,0\n\t" /* Clear CC */ \
+ ".insn rrf,0x" #opcode "0000,%[out],%[a],%[b],0\n\t" \
+ "ipm %[cc]\n\t" \
+ "srl %[cc],28\n" \
+ : [out] "+d" (out), \
+ [cc] "=d" (cc) \
+ : [a] "d" (a), \
+ [b] "d" (b) \
+ : "cc"); \
+ \
+ printf("\t%016lx %016lx -> %016lx cc=%d\n", \
+ a, b, out, cc); \
+ }
+
+#define TEST_EXEC(opcode,insn) \
+ do { \
+ puts(#insn); \
+ test_##insn(0, 0); \
+ test_##insn(0, -1); \
+ test_##insn(-1, 0); \
+ test_##insn(-1, -1); \
+ test_##insn(0x012345678abcdef, 0); \
+ test_##insn(0x012345678abcdef, -1); \
+ test_##insn(0x55555555aaaaaaaa, 0xaaaaaaaa55555555); \
+ } while (0)
+
+#define INSNS \
+ XTEST(b9f5,ncrk); \
+ XTEST(b9e5,ncgrk); \
+ XTEST(b974,nnrk); \
+ XTEST(b964,nngrk); \
+ XTEST(b976,nork); \
+ XTEST(b966,nogrk); \
+ XTEST(b977,nxrk); \
+ XTEST(b967,nxgrk); \
+ XTEST(b975,ocrk); \
+ XTEST(b965,ocgrk);
+
+#define XTEST TEST_GENERATE
+INSNS
+#undef XTEST
+
+static void test_all_logical_insns()
+{
+#define XTEST TEST_EXEC
+ INSNS
+#undef XTEST
+}
+#undef INSNS
+#undef TEST_GENERATE
+#undef TEST_EXEC
+
+
+/* -- Full population count -- */
+
+static void test_popcnt(unsigned long op2)
+{
+ unsigned long result;
+ int cc;
+
+ __asm__(".insn rrf,0xb9e10000,%[result],%[op2],8,0\n\t"
+ "ipm %[cc]\n\t"
+ "srl %[cc],28\n"
+ : [result]"=d" (result),
+ [cc]"=d" (cc)
+ : [op2]"d" (op2)
+ : "cc");
+ printf("\t%016lx -> %2lu cc=%d\n", op2, result, cc);
+}
+
+static int test_all_popcnt()
+{
+ puts("popcnt");
+ test_popcnt(0);
+ test_popcnt(1);
+ test_popcnt(0x8000000000000000);
+ test_popcnt(-1UL);
+ test_popcnt(0xff427e3800556bcd);
+ return 0;
+}
+
+/* -- Select -- */
+
+#define TEST_GENERATE(opcode,insn) \
+ static void test_##insn(unsigned long a, unsigned long b) \
+ { \
+ unsigned long out0 = 0x0cafebad0badcafe; \
+ unsigned long out1 = 0x0badcafe0cafebad; \
+ \
+ __asm__( \
+ "cr 0,0\n\t" /* Clear CC */ \
+ ".insn rrf,0x" #opcode "0000,%[out0],%[a],%[b],8\n\t" \
+ ".insn rrf,0x" #opcode "0000,%[out1],%[a],%[b],7\n\t" \
+ : [out0] "+d" (out0), \
+ [out1] "+d" (out1) \
+ : [a] "d" (a), \
+ [b] "d" (b) \
+ : ); \
+ \
+ printf("\t%016lx %016lx -> %016lx %016lx\n", \
+ a, b, out0, out1); \
+ }
+
+#define TEST_EXEC(opcode,insn) \
+ do { \
+ puts(#insn); \
+ test_##insn(-1, 0); \
+ test_##insn(0, -1); \
+ test_##insn(0x1234567890abcdef, 0xfedcba9876543210); \
+ } while (0)
+
+#define INSNS \
+ XTEST(b9f0,selr); \
+ XTEST(b9e3,selgr); \
+ XTEST(b9c0,selfhr);
+
+#define XTEST TEST_GENERATE
+INSNS
+#undef XTEST
+
+static void test_all_select()
+{
+#define XTEST TEST_EXEC
+ INSNS
+#undef XTEST
+}
+#undef INSNS
+#undef TEST_GENERATE
+#undef TEST_EXEC
+
+
+/* -- Move right to left -- */
+
+static void test_mvcrl(void *to, void *from, size_t len)
+{
+ len -= 1;
+ __asm__("lgr 0,%[len]\n\t"
+ ".insn sse,0xe50a00000000,%[to],%[from]\n\t"
+ : [to] "+Q" (*(struct { char c[len]; } *) to)
+ : [from] "Q" (*(struct { char c[len]; } *) from),
+ [len] "d" (len)
+ : );
+}
+
+static void test_all_mvcrl()
+{
+ static const char pattern[] =
+ "abcdefghijklmnopqrstuvwxyz-0123456789.ABCDEFGHIJKLMNOPQRSTUVWXYZ";
+ char buf[4 * sizeof(pattern) - 2];
+
+ test_mvcrl(buf, (char *) pattern, sizeof(pattern));
+ test_mvcrl(buf + sizeof(pattern) - 1, buf, sizeof(pattern));
+ test_mvcrl(buf + 2 * sizeof(pattern) - 2, buf, 2 * sizeof(pattern) - 1);
+ test_mvcrl(buf + 32, buf + 10, 63);
+ test_mvcrl(buf + 2, buf + 1, 256);
+ test_mvcrl(buf + 254, buf + 256, 2);
+ puts("mvcrl");
+ for (int i = 0; i < 256; i += 64) {
+ printf("\t%.64s\n", buf + i);
+ }
+}
+
+
+int main()
+{
+ test_all_logical_insns();
+ test_all_popcnt();
+ test_all_select();
+ test_all_mvcrl();
+ return 0;
+}
diff --git a/none/tests/s390x/misc3.stderr.exp b/none/tests/s390x/misc3.stderr.exp
new file mode 100644
index 000000000..139597f9c
--- /dev/null
+++ b/none/tests/s390x/misc3.stderr.exp
@@ -0,0 +1,2 @@
+
+
diff --git a/none/tests/s390x/misc3.stdout.exp b/none/tests/s390x/misc3.stdout.exp
new file mode 100644
index 000000000..caaba4960
--- /dev/null
+++ b/none/tests/s390x/misc3.stdout.exp
@@ -0,0 +1,103 @@
+ncrk
+ 0000000000000000 0000000000000000 -> decaffee00000000 cc=0
+ 0000000000000000 ffffffffffffffff -> decaffee00000000 cc=0
+ ffffffffffffffff 0000000000000000 -> decaffeeffffffff cc=1
+ ffffffffffffffff ffffffffffffffff -> decaffee00000000 cc=0
+ 0012345678abcdef 0000000000000000 -> decaffee78abcdef cc=1
+ 0012345678abcdef ffffffffffffffff -> decaffee00000000 cc=0
+ 55555555aaaaaaaa aaaaaaaa55555555 -> decaffeeaaaaaaaa cc=1
+ncgrk
+ 0000000000000000 0000000000000000 -> 0000000000000000 cc=0
+ 0000000000000000 ffffffffffffffff -> 0000000000000000 cc=0
+ ffffffffffffffff 0000000000000000 -> ffffffffffffffff cc=1
+ ffffffffffffffff ffffffffffffffff -> 0000000000000000 cc=0
+ 0012345678abcdef 0000000000000000 -> 0012345678abcdef cc=1
+ 0012345678abcdef ffffffffffffffff -> 0000000000000000 cc=0
+ 55555555aaaaaaaa aaaaaaaa55555555 -> 55555555aaaaaaaa cc=1
+nnrk
+ 0000000000000000 0000000000000000 -> decaffeeffffffff cc=1
+ 0000000000000000 ffffffffffffffff -> decaffeeffffffff cc=1
+ ffffffffffffffff 0000000000000000 -> decaffeeffffffff cc=1
+ ffffffffffffffff ffffffffffffffff -> decaffee00000000 cc=0
+ 0012345678abcdef 0000000000000000 -> decaffeeffffffff cc=1
+ 0012345678abcdef ffffffffffffffff -> decaffee87543210 cc=1
+ 55555555aaaaaaaa aaaaaaaa55555555 -> decaffeeffffffff cc=1
+nngrk
+ 0000000000000000 0000000000000000 -> ffffffffffffffff cc=1
+ 0000000000000000 ffffffffffffffff -> ffffffffffffffff cc=1
+ ffffffffffffffff 0000000000000000 -> ffffffffffffffff cc=1
+ ffffffffffffffff ffffffffffffffff -> 0000000000000000 cc=0
+ 0012345678abcdef 0000000000000000 -> ffffffffffffffff cc=1
+ 0012345678abcdef ffffffffffffffff -> ffedcba987543210 cc=1
+ 55555555aaaaaaaa aaaaaaaa55555555 -> ffffffffffffffff cc=1
+nork
+ 0000000000000000 0000000000000000 -> decaffeeffffffff cc=1
+ 0000000000000000 ffffffffffffffff -> decaffee00000000 cc=0
+ ffffffffffffffff 0000000000000000 -> decaffee00000000 cc=0
+ ffffffffffffffff ffffffffffffffff -> decaffee00000000 cc=0
+ 0012345678abcdef 0000000000000000 -> decaffee87543210 cc=1
+ 0012345678abcdef ffffffffffffffff -> decaffee00000000 cc=0
+ 55555555aaaaaaaa aaaaaaaa55555555 -> decaffee00000000 cc=0
+nogrk
+ 0000000000000000 0000000000000000 -> ffffffffffffffff cc=1
+ 0000000000000000 ffffffffffffffff -> 0000000000000000 cc=0
+ ffffffffffffffff 0000000000000000 -> 0000000000000000 cc=0
+ ffffffffffffffff ffffffffffffffff -> 0000000000000000 cc=0
+ 0012345678abcdef 0000000000000000 -> ffedcba987543210 cc=1
+ 0012345678abcdef ffffffffffffffff -> 0000000000000000 cc=0
+ 55555555aaaaaaaa aaaaaaaa55555555 -> 0000000000000000 cc=0
+nxrk
+ 0000000000000000 0000000000000000 -> decaffeeffffffff cc=1
+ 0000000000000000 ffffffffffffffff -> decaffee00000000 cc=0
+ ffffffffffffffff 0000000000000000 -> decaffee00000000 cc=0
+ ffffffffffffffff ffffffffffffffff -> decaffeeffffffff cc=1
+ 0012345678abcdef 0000000000000000 -> decaffee87543210 cc=1
+ 0012345678abcdef ffffffffffffffff -> decaffee78abcdef cc=1
+ 55555555aaaaaaaa aaaaaaaa55555555 -> decaffee00000000 cc=0
+nxgrk
+ 0000000000000000 0000000000000000 -> ffffffffffffffff cc=1
+ 0000000000000000 ffffffffffffffff -> 0000000000000000 cc=0
+ ffffffffffffffff 0000000000000000 -> 0000000000000000 cc=0
+ ffffffffffffffff ffffffffffffffff -> ffffffffffffffff cc=1
+ 0012345678abcdef 0000000000000000 -> ffedcba987543210 cc=1
+ 0012345678abcdef ffffffffffffffff -> 0012345678abcdef cc=1
+ 55555555aaaaaaaa aaaaaaaa55555555 -> 0000000000000000 cc=0
+ocrk
+ 0000000000000000 0000000000000000 -> decaffeeffffffff cc=1
+ 0000000000000000 ffffffffffffffff -> decaffee00000000 cc=0
+ ffffffffffffffff 0000000000000000 -> decaffeeffffffff cc=1
+ ffffffffffffffff ffffffffffffffff -> decaffeeffffffff cc=1
+ 0012345678abcdef 0000000000000000 -> decaffeeffffffff cc=1
+ 0012345678abcdef ffffffffffffffff -> decaffee78abcdef cc=1
+ 55555555aaaaaaaa aaaaaaaa55555555 -> decaffeeaaaaaaaa cc=1
+ocgrk
+ 0000000000000000 0000000000000000 -> ffffffffffffffff cc=1
+ 0000000000000000 ffffffffffffffff -> 0000000000000000 cc=0
+ ffffffffffffffff 0000000000000000 -> ffffffffffffffff cc=1
+ ffffffffffffffff ffffffffffffffff -> ffffffffffffffff cc=1
+ 0012345678abcdef 0000000000000000 -> ffffffffffffffff cc=1
+ 0012345678abcdef ffffffffffffffff -> 0012345678abcdef cc=1
+ 55555555aaaaaaaa aaaaaaaa55555555 -> 55555555aaaaaaaa cc=1
+popcnt
+ 0000000000000000 -> 0 cc=0
+ 0000000000000001 -> 1 cc=1
+ 8000000000000000 -> 1 cc=1
+ ffffffffffffffff -> 64 cc=1
+ ff427e3800556bcd -> 33 cc=1
+selr
+ ffffffffffffffff 0000000000000000 -> 0cafebadffffffff 0badcafe00000000
+ 0000000000000000 ffffffffffffffff -> 0cafebad00000000 0badcafeffffffff
+ 1234567890abcdef fedcba9876543210 -> 0cafebad90abcdef 0badcafe76543210
+selgr
+ ffffffffffffffff 0000000000000000 -> ffffffffffffffff 0000000000000000
+ 0000000000000000 ffffffffffffffff -> 0000000000000000 ffffffffffffffff
+ 1234567890abcdef fedcba9876543210 -> 1234567890abcdef fedcba9876543210
+selfhr
+ ffffffffffffffff 0000000000000000 -> ffffffff0badcafe 000000000cafebad
+ 0000000000000000 ffffffffffffffff -> 000000000badcafe ffffffff0cafebad
+ 1234567890abcdef fedcba9876543210 -> 123456780badcafe fedcba980cafebad
+mvcrl
+ abbcdefghijklmnopqrstuvwxyz-01234klmnopqrstuvwxyz-0123456789.ABC
+ DEFGHIJKLMNOPQRSTUVWXYZabcdefghi456789.ABCDEFGHIJKLMNOPQRSTUVWXY
+ Zabcdefghijklmnopqrstuvwxyz-0123456789.ABCDEFGHIJKLMNOPQRSTUVWXY
+ Zabcdefghijklmnopqrstuvwxyz-0123456789.ABCDEFGHIJKLMNOPQRSTUVWZ
diff --git a/none/tests/s390x/misc3.vgtest b/none/tests/s390x/misc3.vgtest
new file mode 100644
index 000000000..d051a06bd
--- /dev/null
+++ b/none/tests/s390x/misc3.vgtest
@@ -0,0 +1 @@
+prog: misc3
--
2.23.0
From 401b51d79886362d1962dc487db45ac91462eaa0 Mon Sep 17 00:00:00 2001
From: Andreas Arnez <arnez@linux.ibm.com>
Date: Wed, 7 Apr 2021 12:29:32 +0200
Subject: [PATCH 06/13] s390x: Vec-enh-2, extend VSL, VSRA, and VSRL
The vector-enhancements facility 2 extends the existing bitwise vector
shift instructions VSL, VSRA, and VSRL. Now they allow the shift
vector (the third operand) to contain different shift amounts for each
byte. Add support for these new forms.
---
VEX/priv/guest_s390_toIR.c | 58 ++++++++++++++++++++++++++++++--------
1 file changed, 47 insertions(+), 11 deletions(-)
diff --git a/VEX/priv/guest_s390_toIR.c b/VEX/priv/guest_s390_toIR.c
index 9f7d98f8c..622d5a02e 100644
--- a/VEX/priv/guest_s390_toIR.c
+++ b/VEX/priv/guest_s390_toIR.c
@@ -17983,30 +17983,66 @@ s390_irgen_VERLL(UChar v1, IRTemp op2addr, UChar v3, UChar m4)
static const HChar *
s390_irgen_VSL(UChar v1, UChar v2, UChar v3)
{
- IRTemp shift_amount = newTemp(Ity_I8);
- assign(shift_amount, binop(Iop_And8, get_vr_b7(v3), mkU8(0b00000111)));
-
- put_vr_qw(v1, binop(Iop_ShlV128, get_vr_qw(v2), mkexpr(shift_amount)));
+ IRTemp a = newTemp(Ity_V128);
+ IRTemp b = newTemp(Ity_V128);
+
+ assign(a, get_vr_qw(v2));
+ assign(b, get_vr_qw(v3));
+
+ put_vr_qw(v1,
+ binop(Iop_OrV128,
+ binop(Iop_Shl8x16, mkexpr(a), mkexpr(b)),
+ binop(Iop_Shr8x16,
+ binop(Iop_Shr8x16,
+ binop(Iop_ShlV128, mkexpr(a), mkU8(8)),
+ unop(Iop_NotV128, mkexpr(b))),
+ unop(Iop_Dup8x16, mkU8(1)))));
return "vsl";
}
static const HChar *
s390_irgen_VSRL(UChar v1, UChar v2, UChar v3)
{
- IRTemp shift_amount = newTemp(Ity_I8);
- assign(shift_amount, binop(Iop_And8, get_vr_b7(v3), mkU8(0b00000111)));
+ IRTemp a = newTemp(Ity_V128);
+ IRTemp b = newTemp(Ity_V128);
- put_vr_qw(v1, binop(Iop_ShrV128, get_vr_qw(v2), mkexpr(shift_amount)));
+ assign(a, get_vr_qw(v2));
+ assign(b, get_vr_qw(v3));
+
+ put_vr_qw(v1,
+ binop(Iop_OrV128,
+ binop(Iop_Shr8x16, mkexpr(a), mkexpr(b)),
+ binop(Iop_Shl8x16,
+ binop(Iop_Shl8x16,
+ binop(Iop_ShrV128, mkexpr(a), mkU8(8)),
+ unop(Iop_NotV128, mkexpr(b))),
+ unop(Iop_Dup8x16, mkU8(1)))));
return "vsrl";
}
static const HChar *
s390_irgen_VSRA(UChar v1, UChar v2, UChar v3)
{
- IRTemp shift_amount = newTemp(Ity_I8);
- assign(shift_amount, binop(Iop_And8, get_vr_b7(v3), mkU8(0b00000111)));
-
- put_vr_qw(v1, binop(Iop_SarV128, get_vr_qw(v2), mkexpr(shift_amount)));
+ IRTemp a = newTemp(Ity_V128);
+ IRTemp b = newTemp(Ity_V128);
+
+ assign(a, get_vr_qw(v2));
+ assign(b, get_vr_qw(v3));
+
+ /* Shift-right: first byte arithmetically, all others logically */
+ IRExpr* elems_shifted =
+ binop(Iop_Sar8x16,
+ binop(Iop_Shr8x16, mkexpr(a),
+ binop(Iop_AndV128, mkexpr(b), mkV128(0x7fff))),
+ binop(Iop_AndV128, mkexpr(b), mkV128(0x8000)));
+ /* Then OR the appropriate bits from the byte to the left */
+ put_vr_qw(v1,
+ binop(Iop_OrV128, elems_shifted,
+ binop(Iop_Shl8x16,
+ binop(Iop_Shl8x16,
+ binop(Iop_ShrV128, mkexpr(a), mkU8(8)),
+ unop(Iop_NotV128, mkexpr(b))),
+ unop(Iop_Dup8x16, mkU8(1)))));
return "vsra";
}
--
2.23.0
From 3fdf065d0bf26a02d6d93a812a6571a287379c36 Mon Sep 17 00:00:00 2001
From: Andreas Arnez <arnez@linux.ibm.com>
Date: Thu, 11 Feb 2021 20:02:03 +0100
Subject: [PATCH 07/13] s390x: Vec-enh-2, extend VCDG, VCDLG, VCGD, and VCLGD
The vector-enhancements facility 2 extends the vector floating-point
conversion instructions VCDG, VCDLG, VCGD, and VCLGD. In addition to
64-bit elements, they now also handle 32-bit elements. Add support for
these new forms.
---
VEX/priv/guest_s390_toIR.c | 36 ++++++++++++++++++++----------------
1 file changed, 20 insertions(+), 16 deletions(-)
diff --git a/VEX/priv/guest_s390_toIR.c b/VEX/priv/guest_s390_toIR.c
index 622d5a02e..11271a1c9 100644
--- a/VEX/priv/guest_s390_toIR.c
+++ b/VEX/priv/guest_s390_toIR.c
@@ -18794,44 +18794,48 @@ s390_vector_fp_convert(IROp op, IRType fromType, IRType toType, Bool rounding,
static const HChar *
s390_irgen_VCDG(UChar v1, UChar v2, UChar m3, UChar m4, UChar m5)
{
- s390_insn_assert("vcdg", m3 == 3);
-
- s390_vector_fp_convert(Iop_I64StoF64, Ity_I64, Ity_F64, True,
- v1, v2, m3, m4, m5);
+ s390_insn_assert("vcdg", m3 == 2 || m3 == 3);
+ s390_vector_fp_convert(m3 == 2 ? Iop_I32StoF32 : Iop_I64StoF64,
+ m3 == 2 ? Ity_I32 : Ity_I64,
+ m3 == 2 ? Ity_F32 : Ity_F64,
+ True, v1, v2, m3, m4, m5);
return "vcdg";
}
static const HChar *
s390_irgen_VCDLG(UChar v1, UChar v2, UChar m3, UChar m4, UChar m5)
{
- s390_insn_assert("vcdlg", m3 == 3);
-
- s390_vector_fp_convert(Iop_I64UtoF64, Ity_I64, Ity_F64, True,
- v1, v2, m3, m4, m5);
+ s390_insn_assert("vcdlg", m3 == 2 || m3 == 3);
+ s390_vector_fp_convert(m3 == 2 ? Iop_I32UtoF32 : Iop_I64UtoF64,
+ m3 == 2 ? Ity_I32 : Ity_I64,
+ m3 == 2 ? Ity_F32 : Ity_F64,
+ True, v1, v2, m3, m4, m5);
return "vcdlg";
}
static const HChar *
s390_irgen_VCGD(UChar v1, UChar v2, UChar m3, UChar m4, UChar m5)
{
- s390_insn_assert("vcgd", m3 == 3);
-
- s390_vector_fp_convert(Iop_F64toI64S, Ity_F64, Ity_I64, True,
- v1, v2, m3, m4, m5);
+ s390_insn_assert("vcgd", m3 == 2 || m3 == 3);
+ s390_vector_fp_convert(m3 == 2 ? Iop_F32toI32S : Iop_F64toI64S,
+ m3 == 2 ? Ity_F32 : Ity_F64,
+ m3 == 2 ? Ity_I32 : Ity_I64,
+ True, v1, v2, m3, m4, m5);
return "vcgd";
}
static const HChar *
s390_irgen_VCLGD(UChar v1, UChar v2, UChar m3, UChar m4, UChar m5)
{
- s390_insn_assert("vclgd", m3 == 3);
-
- s390_vector_fp_convert(Iop_F64toI64U, Ity_F64, Ity_I64, True,
- v1, v2, m3, m4, m5);
+ s390_insn_assert("vclgd", m3 == 2 || m3 == 3);
+ s390_vector_fp_convert(m3 == 2 ? Iop_F32toI32U : Iop_F64toI64U,
+ m3 == 2 ? Ity_F32 : Ity_F64,
+ m3 == 2 ? Ity_I32 : Ity_I64,
+ True, v1, v2, m3, m4, m5);
return "vclgd";
}
--
2.23.0
From d195bf17388572e85474c7ded4b5bd0e4774637d Mon Sep 17 00:00:00 2001
From: Andreas Arnez <arnez@linux.ibm.com>
Date: Tue, 16 Feb 2021 16:19:31 +0100
Subject: [PATCH 08/13] s390x: Vec-enh-2, VLBR and friends
Add support for the new byte- and element-swapping vector load/store
instructions VLEBRH, VLEBRG, VLEBRF, VLLEBRZ, VLBRREP, VLBR, VLER,
VSTEBRH, VSTEBRG, VSTEBRF, VSTBR, and VSTER.
---
VEX/priv/guest_s390_toIR.c | 256 +++++++++++++++++++++++++++++++++++++
VEX/priv/host_s390_isel.c | 9 ++
2 files changed, 265 insertions(+)
diff --git a/VEX/priv/guest_s390_toIR.c b/VEX/priv/guest_s390_toIR.c
index 11271a1c9..f65b42705 100644
--- a/VEX/priv/guest_s390_toIR.c
+++ b/VEX/priv/guest_s390_toIR.c
@@ -19388,6 +19388,209 @@ s390_irgen_SELFHR(UChar r3, UChar m4, UChar r1, UChar r2)
return "selfhr";
}
+/* Helper function that byte-swaps each element of its V128 input operand */
+static IRExpr *
+s390_byteswap_elements(IRExpr* v, UChar m)
+{
+ static const ULong perm[4][2] = {
+ { 0x0100030205040706, 0x09080b0a0d0c0f0e }, /* 2-byte elements */
+ { 0x0302010007060504, 0x0b0a09080f0e0d0c }, /* 4-byte elements */
+ { 0x0706050403020100, 0x0f0e0d0c0b0a0908 }, /* 8-byte elements */
+ { 0x0f0e0d0c0b0a0908, 0x0706050403020100 }, /* whole vector */
+ };
+ return binop(Iop_Perm8x16, v, binop(Iop_64HLtoV128,
+ mkU64(perm[m - 1][0]),
+ mkU64(perm[m - 1][1])));
+}
+
+/* Helper function that reverses the elements of its V128 input operand */
+static IRExpr *
+s390_reverse_elements(IRExpr* v, UChar m)
+{
+ static const ULong perm[3][2] = {
+ { 0x0e0f0c0d0a0b0809, 0x0607040502030001 }, /* 2-byte elements */
+ { 0x0c0d0e0f08090a0b, 0x0405060700010203 }, /* 4-byte elements */
+ { 0x08090a0b0c0d0e0f, 0x0001020304050607 }, /* 8-byte elements */
+ };
+ return binop(Iop_Perm8x16, v, binop(Iop_64HLtoV128,
+ mkU64(perm[m - 1][0]),
+ mkU64(perm[m - 1][1])));
+}
+
+static const HChar *
+s390_irgen_VLBR(UChar v1, IRTemp op2addr, UChar m3)
+{
+ s390_insn_assert("vlbr", m3 >= 1 && m3 <= 4);
+ put_vr_qw(v1, s390_byteswap_elements(load(Ity_V128, mkexpr(op2addr)), m3));
+ return "vlbr";
+}
+
+static const HChar *
+s390_irgen_VSTBR(UChar v1, IRTemp op2addr, UChar m3)
+{
+ s390_insn_assert("vstbr", m3 >= 1 && m3 <= 4);
+ store(mkexpr(op2addr), s390_byteswap_elements(get_vr_qw(v1), m3));
+ return "vstbr";
+}
+
+static const HChar *
+s390_irgen_VLER(UChar v1, IRTemp op2addr, UChar m3)
+{
+ s390_insn_assert("vler", m3 >= 1 && m3 <= 3);
+ put_vr_qw(v1, s390_reverse_elements(load(Ity_V128, mkexpr(op2addr)), m3));
+ return "vler";
+}
+
+static const HChar *
+s390_irgen_VSTER(UChar v1, IRTemp op2addr, UChar m3)
+{
+ s390_insn_assert("vstbr", m3 >= 1 && m3 <= 4);
+ store(mkexpr(op2addr), s390_reverse_elements(get_vr_qw(v1), m3));
+ return "vstbr";
+}
+
+/* Helper function that combines its two V128 operands by replacing element 'to'
+ in 'a' by byte-swapped element 'from' in 'b' */
+static IRExpr *
+s390_insert_byteswapped(IRExpr* a, IRExpr* b, UChar m, UChar to, UChar from)
+{
+ UInt elem_size = 1U << m;
+ UInt start = elem_size * to;
+ UInt end = start + elem_size - 1;
+ UInt offs = end + elem_size * from + 16;
+ UInt i;
+
+ ULong permH = 0;
+ for (i = 0; i < 8; i++) {
+ permH = (permH << 8) | (i >= start && i <= end ? offs - i : i);
+ }
+ ULong permL = 0;
+ for (i = 8; i < 16; i++) {
+ permL = (permL << 8) | (i >= start && i <= end ? offs - i : i);
+ }
+ return triop(Iop_Perm8x16x2, a, b, binop(Iop_64HLtoV128,
+ mkU64(permH), mkU64(permL)));
+}
+
+static const HChar *
+s390_irgen_VLEBRH(UChar v1, IRTemp op2addr, UChar m3)
+{
+ s390_insn_assert("vlebrh", m3 <= 7);
+ IRTemp op2 = newTemp(Ity_I16);
+ assign(op2, load(Ity_I16, mkexpr(op2addr)));
+ put_vr(v1, Ity_I16, m3, binop(Iop_Or16,
+ binop(Iop_Shl16, mkexpr(op2), mkU8(8)),
+ binop(Iop_Shr16, mkexpr(op2), mkU8(8))));
+ return "vlebrh";
+}
+
+static const HChar *
+s390_irgen_VLEBRF(UChar v1, IRTemp op2addr, UChar m3)
+{
+ s390_insn_assert("vlebrf", m3 <= 3);
+ IRTemp op1 = newTemp(Ity_V128);
+ assign(op1, get_vr_qw(v1));
+ IRTemp op2 = newTemp(Ity_I64);
+ assign(op2, unop(Iop_32Uto64, load(Ity_I32, mkexpr(op2addr))));
+ IRExpr* b = binop(Iop_64HLtoV128, mkexpr(op2), mkexpr(op2));
+ put_vr_qw(v1, s390_insert_byteswapped(mkexpr(op1), b, 2, m3, 3));
+ return "vlebrf";
+}
+
+static const HChar *
+s390_irgen_VLEBRG(UChar v1, IRTemp op2addr, UChar m3)
+{
+ s390_insn_assert("vlebrg", m3 <= 1);
+ IRTemp op1 = newTemp(Ity_V128);
+ assign(op1, get_vr_qw(v1));
+ IRTemp op2 = newTemp(Ity_I64);
+ assign(op2, load(Ity_I64, mkexpr(op2addr)));
+ IRExpr* b = binop(Iop_64HLtoV128, mkexpr(op2), mkexpr(op2));
+ put_vr_qw(v1, s390_insert_byteswapped(mkexpr(op1), b, 3, m3, 1));
+ return "vlebrg";
+}
+
+static const HChar *
+s390_irgen_VLBRREP(UChar v1, IRTemp op2addr, UChar m3)
+{
+ s390_insn_assert("vlbrrep", m3 >= 1 && m3 <= 3);
+ static const ULong perm[3] = {
+ 0x0f0e0f0e0f0e0f0e, /* 2-byte element */
+ 0x0f0e0d0c0f0e0d0c, /* 4-byte element */
+ 0x0f0e0d0c0b0a0908 /* 8-byte element */
+ };
+ IRExpr* permHL = mkU64(perm[m3 - 1]);
+ IRTemp op2 = newTemp(Ity_I64);
+ if (m3 == 3)
+ assign(op2, load(Ity_I64, mkexpr(op2addr)));
+ else
+ assign(op2, unop(m3 == 2 ? Iop_32Uto64 : Iop_16Uto64,
+ load(s390_vr_get_type(m3), mkexpr(op2addr))));
+ put_vr_qw(v1, binop(Iop_Perm8x16,
+ binop(Iop_64HLtoV128, mkexpr(op2), mkexpr(op2)),
+ binop(Iop_64HLtoV128, permHL, permHL)));
+ return "vlbrrep";
+}
+
+static const HChar *
+s390_irgen_VLLEBRZ(UChar v1, IRTemp op2addr, UChar m3)
+{
+ s390_insn_assert("vllebrz", (m3 >= 1 && m3 <= 3) || m3 == 6);
+ static const ULong perm[6] = {
+ 0x0000000000000f0e, /* 2-byte element */
+ 0x000000000f0e0d0c, /* 4-byte element */
+ 0x0f0e0d0c0b0a0908, /* 8-byte element */
+ 0, /* invalid (4) */
+ 0, /* invalid (5) */
+ 0x0f0e0d0c00000000, /* 4-byte element, left-aligned */
+ };
+ IRExpr* permH = mkU64(perm[m3 - 1]);
+ IRTemp op2 = newTemp(Ity_I64);
+ if (m3 == 3)
+ assign(op2, load(Ity_I64, mkexpr(op2addr)));
+ else
+ assign(op2, unop((m3 & 3) == 2 ? Iop_32Uto64 : Iop_16Uto64,
+ load(s390_vr_get_type(m3 & 3), mkexpr(op2addr))));
+ put_vr_qw(v1, binop(Iop_Perm8x16,
+ binop(Iop_64HLtoV128, mkU64(0), mkexpr(op2)),
+ binop(Iop_64HLtoV128, permH, mkU64(0))));
+ return "vllebrz";
+}
+
+static const HChar *
+s390_irgen_VSTEBRH(UChar v1, IRTemp op2addr, UChar m3)
+{
+ s390_insn_assert("vstebrh", m3 <= 7);
+ IRTemp op1 = newTemp(Ity_I16);
+ assign(op1, get_vr(v1, Ity_I16, m3));
+ store(mkexpr(op2addr), binop(Iop_Or16,
+ binop(Iop_Shl16, mkexpr(op1), mkU8(8)),
+ binop(Iop_Shr16, mkexpr(op1), mkU8(8))));
+ return "vstebrh";
+}
+
+static const HChar *
+s390_irgen_VSTEBRF(UChar v1, IRTemp op2addr, UChar m3)
+{
+ s390_insn_assert("vstebrf", m3 <= 3);
+ IRTemp op1 = newTemp(Ity_V128);
+ assign(op1, get_vr_qw(v1));
+ IRExpr* b = s390_insert_byteswapped(mkexpr(op1), mkexpr(op1), 2, 3, m3);
+ store(mkexpr(op2addr), unop(Iop_V128to32, b));
+ return "vstebrf";
+}
+
+static const HChar *
+s390_irgen_VSTEBRG(UChar v1, IRTemp op2addr, UChar m3)
+{
+ s390_insn_assert("vstebrg", m3 <= 1);
+ IRTemp op1 = newTemp(Ity_V128);
+ assign(op1, get_vr_qw(v1));
+ IRExpr* b = s390_insert_byteswapped(mkexpr(op1), mkexpr(op1), 3, 1, m3);
+ store(mkexpr(op2addr), unop(Iop_V128to64, b));
+ return "vstebrg";
+}
+
/* New insns are added here.
If an insn is contingent on a facility being installed also
check whether the list of supported facilities in function
@@ -21003,6 +21206,59 @@ s390_decode_6byte_and_irgen(const UChar *bytes)
RXY_x2(ovl), RXY_b2(ovl),
RXY_dl2(ovl),
RXY_dh2(ovl)); goto ok;
+ case 0xe60000000001ULL: s390_format_VRX_VRRDM(s390_irgen_VLEBRH, VRX_v1(ovl),
+ VRX_x2(ovl), VRX_b2(ovl),
+ VRX_d2(ovl), VRX_m3(ovl),
+ VRX_rxb(ovl)); goto ok;
+ case 0xe60000000002ULL: s390_format_VRX_VRRDM(s390_irgen_VLEBRG, VRX_v1(ovl),
+ VRX_x2(ovl), VRX_b2(ovl),
+ VRX_d2(ovl), VRX_m3(ovl),
+ VRX_rxb(ovl)); goto ok;
+ case 0xe60000000003ULL: s390_format_VRX_VRRDM(s390_irgen_VLEBRF, VRX_v1(ovl),
+ VRX_x2(ovl), VRX_b2(ovl),
+ VRX_d2(ovl), VRX_m3(ovl),
+ VRX_rxb(ovl)); goto ok;
+ case 0xe60000000004ULL: s390_format_VRX_VRRDM(s390_irgen_VLLEBRZ,
+ VRX_v1(ovl),
+ VRX_x2(ovl), VRX_b2(ovl),
+ VRX_d2(ovl), VRX_m3(ovl),
+ VRX_rxb(ovl)); goto ok;
+ case 0xe60000000005ULL: s390_format_VRX_VRRDM(s390_irgen_VLBRREP,
+ VRX_v1(ovl),
+ VRX_x2(ovl), VRX_b2(ovl),
+ VRX_d2(ovl), VRX_m3(ovl),
+ VRX_rxb(ovl)); goto ok;
+ case 0xe60000000006ULL: s390_format_VRX_VRRDM(s390_irgen_VLBR, VRX_v1(ovl),
+ VRX_x2(ovl), VRX_b2(ovl),
+ VRX_d2(ovl), VRX_m3(ovl),
+ VRX_rxb(ovl)); goto ok;
+ case 0xe60000000007ULL: s390_format_VRX_VRRDM(s390_irgen_VLER, VRX_v1(ovl),
+ VRX_x2(ovl), VRX_b2(ovl),
+ VRX_d2(ovl), VRX_m3(ovl),
+ VRX_rxb(ovl)); goto ok;
+ case 0xe60000000009ULL: s390_format_VRX_VRRDM(s390_irgen_VSTEBRH,
+ VRX_v1(ovl),
+ VRX_x2(ovl), VRX_b2(ovl),
+ VRX_d2(ovl), VRX_m3(ovl),
+ VRX_rxb(ovl)); goto ok;
+ case 0xe6000000000aULL: s390_format_VRX_VRRDM(s390_irgen_VSTEBRG,
+ VRX_v1(ovl),
+ VRX_x2(ovl), VRX_b2(ovl),
+ VRX_d2(ovl), VRX_m3(ovl),
+ VRX_rxb(ovl)); goto ok;
+ case 0xe6000000000bULL: s390_format_VRX_VRRDM(s390_irgen_VSTEBRF,
+ VRX_v1(ovl),
+ VRX_x2(ovl), VRX_b2(ovl),
+ VRX_d2(ovl), VRX_m3(ovl),
+ VRX_rxb(ovl)); goto ok;
+ case 0xe6000000000eULL: s390_format_VRX_VRRDM(s390_irgen_VSTBR, VRX_v1(ovl),
+ VRX_x2(ovl), VRX_b2(ovl),
+ VRX_d2(ovl), VRX_m3(ovl),
+ VRX_rxb(ovl)); goto ok;
+ case 0xe6000000000fULL: s390_format_VRX_VRRDM(s390_irgen_VSTER, VRX_v1(ovl),
+ VRX_x2(ovl), VRX_b2(ovl),
+ VRX_d2(ovl), VRX_m3(ovl),
+ VRX_rxb(ovl)); goto ok;
case 0xe60000000034ULL: /* VPKZ */ goto unimplemented;
case 0xe60000000035ULL: s390_format_VSI_URDV(s390_irgen_VLRL, VSI_v1(ovl),
VSI_b2(ovl), VSI_d2(ovl),
diff --git a/VEX/priv/host_s390_isel.c b/VEX/priv/host_s390_isel.c
index ee20c6711..06e195957 100644
--- a/VEX/priv/host_s390_isel.c
+++ b/VEX/priv/host_s390_isel.c
@@ -4189,6 +4189,15 @@ s390_isel_vec_expr_wrk(ISelEnv *env, IRExpr *expr)
return dst;
}
+ case Iop_Perm8x16:
+ size = 16;
+ reg1 = s390_isel_vec_expr(env, arg1);
+ reg2 = s390_isel_vec_expr(env, arg2);
+
+ addInstr(env, s390_insn_vec_triop(size, S390_VEC_PERM,
+ dst, reg1, reg1, reg2));
+ return dst;
+
case Iop_CmpEQ8x16:
size = 1;
vec_binop = S390_VEC_COMPARE_EQUAL;
--
2.23.0
From f7447f4c73b2d0fb4eb3827c3709f378f6c9c656 Mon Sep 17 00:00:00 2001
From: Andreas Arnez <arnez@linux.ibm.com>
Date: Tue, 23 Feb 2021 19:10:37 +0100
Subject: [PATCH 09/13] s390x: Vec-enh-2, VSLD and VSRD
Support the new "vector shift left/right double by bit" instructions VSLD
and VSRD.
---
VEX/priv/guest_s390_toIR.c | 50 ++++++++++++++++++++++++++++++++++++++
1 file changed, 50 insertions(+)
diff --git a/VEX/priv/guest_s390_toIR.c b/VEX/priv/guest_s390_toIR.c
index f65b42705..aa429d085 100644
--- a/VEX/priv/guest_s390_toIR.c
+++ b/VEX/priv/guest_s390_toIR.c
@@ -18228,6 +18228,48 @@ s390_irgen_VSLDB(UChar v1, UChar v2, UChar v3, UChar i4)
return "vsldb";
}
+static const HChar *
+s390_irgen_VSLD(UChar v1, UChar v2, UChar v3, UChar i4)
+{
+ s390_insn_assert("vsld", i4 <= 7);
+
+ if (i4 == 0) {
+ /* Just copy v2. */
+ put_vr_qw(v1, get_vr_qw(v2));
+ } else {
+ /* Concatenate v2's tail with v3's head. */
+ put_vr_qw(v1,
+ binop(Iop_OrV128,
+ binop(Iop_ShlV128, get_vr_qw(v2), mkU8(i4)),
+ binop(Iop_ShrV128, get_vr_qw(v3), mkU8(128 - i4))
+ )
+ );
+ }
+
+ return "vsld";
+}
+
+static const HChar *
+s390_irgen_VSRD(UChar v1, UChar v2, UChar v3, UChar i4)
+{
+ s390_insn_assert("vsrd", i4 <= 7);
+
+ if (i4 == 0) {
+ /* Just copy v3. */
+ put_vr_qw(v1, get_vr_qw(v3));
+ } else {
+ /* Concatenate v2's tail with v3's head. */
+ put_vr_qw(v1,
+ binop(Iop_OrV128,
+ binop(Iop_ShlV128, get_vr_qw(v2), mkU8(128 - i4)),
+ binop(Iop_ShrV128, get_vr_qw(v3), mkU8(i4))
+ )
+ );
+ }
+
+ return "vsrd";
+}
+
static const HChar *
s390_irgen_VMO(UChar v1, UChar v2, UChar v3, UChar m4)
{
@@ -21541,6 +21583,14 @@ s390_decode_6byte_and_irgen(const UChar *bytes)
case 0xe70000000085ULL: s390_format_VRR_VVV(s390_irgen_VBPERM, VRR_v1(ovl),
VRR_v2(ovl), VRR_r3(ovl),
VRR_rxb(ovl)); goto ok;
+ case 0xe70000000086ULL: s390_format_VRId_VVVI(s390_irgen_VSLD, VRId_v1(ovl),
+ VRId_v2(ovl), VRId_v3(ovl),
+ VRId_i4(ovl),
+ VRId_rxb(ovl)); goto ok;
+ case 0xe70000000087ULL: s390_format_VRId_VVVI(s390_irgen_VSRD, VRId_v1(ovl),
+ VRId_v2(ovl), VRId_v3(ovl),
+ VRId_i4(ovl),
+ VRId_rxb(ovl)); goto ok;
case 0xe7000000008aULL: s390_format_VRR_VVVVMM(s390_irgen_VSTRC, VRRd_v1(ovl),
VRRd_v2(ovl), VRRd_v3(ovl),
VRRd_v4(ovl), VRRd_m5(ovl),
--
2.23.0
From 388082bca7146f8a15814798dbfe570af2aab2a9 Mon Sep 17 00:00:00 2001
From: Andreas Arnez <arnez@linux.ibm.com>
Date: Wed, 10 Mar 2021 19:22:51 +0100
Subject: [PATCH 10/13] s390x: Vec-enh-2, VSTRS
Support the new "vector string search" instruction VSTRS. The
implementation is a full emulation and follows a similar approach as for
the other vector string instructions.
---
VEX/priv/guest_s390_toIR.c | 104 +++++++++++++++++++++++++++++++++++++
1 file changed, 104 insertions(+)
diff --git a/VEX/priv/guest_s390_toIR.c b/VEX/priv/guest_s390_toIR.c
index aa429d085..46a867475 100644
--- a/VEX/priv/guest_s390_toIR.c
+++ b/VEX/priv/guest_s390_toIR.c
@@ -17601,6 +17601,105 @@ s390_irgen_VSTRC(UChar v1, UChar v2, UChar v3, UChar v4, UChar m5, UChar m6)
return "vstrc";
}
+static const HChar *
+s390_irgen_VSTRS(UChar v1, UChar v2, UChar v3, UChar v4, UChar m5, UChar m6)
+{
+ s390_insn_assert("vstrs", m5 <= 2 && m6 == (m6 & 2));
+
+ IRTemp op2 = newTemp(Ity_V128);
+ IRTemp op3 = newTemp(Ity_V128);
+ IRTemp op4 = newTemp(Ity_I8);
+ IRTemp op2clean = newTemp(Ity_V128);
+ IRTemp op3mask = newTemp(Ity_V128);
+ IRTemp result = newTemp(Ity_V128);
+ IRTemp ccnomatch = newTemp(Ity_I64);
+ IRExpr* tmp;
+ IRExpr* match = NULL;
+ UChar elem_bits = 8 << m5;
+ IROp cmpeq_op = S390_VEC_OP3(m5, Iop_CmpEQ8x16,
+ Iop_CmpEQ16x8, Iop_CmpEQ32x4);
+
+ assign(op2, get_vr_qw(v2));
+ assign(op3, get_vr_qw(v3));
+ assign(op4, get_vr_b7(v4));
+
+ tmp = unop(Iop_Dup32x4,
+ unop(Iop_1Sto32, binop(Iop_CmpNE8, mkexpr(op4), mkU8(16))));
+ tmp = binop(Iop_ShrV128, tmp, binop(Iop_Shl8, mkexpr(op4), mkU8(3)));
+
+ if (s390_vr_is_zs_set(m6)) {
+ IRTemp op2eos = newTemp(Ity_V128);
+ IRExpr* t;
+ t = binop(cmpeq_op, mkexpr(op2), mkV128(0));
+ for (UChar i = m5; i < 4; i++) {
+ IRTemp s = newTemp(Ity_V128);
+ assign(s, t);
+ t = binop(Iop_OrV128, mkexpr(s), binop(Iop_ShrV128, mkexpr(s),
+ mkU8(8 << i)));
+ }
+ assign(op2eos, t);
+ assign(op2clean, binop(Iop_AndV128, mkexpr(op2),
+ unop(Iop_NotV128, mkexpr(op2eos))));
+ assign(ccnomatch, binop(Iop_And64, mkU64(1),
+ unop(Iop_V128to64, mkexpr(op2eos))));
+
+ t = binop(cmpeq_op, mkexpr(op3), mkV128(0));
+ for (UChar i = m5; i < 4; i++) {
+ IRTemp s = newTemp(Ity_V128);
+ assign(s, t);
+ t = binop(Iop_OrV128, mkexpr(s), binop(Iop_ShrV128, mkexpr(s),
+ mkU8(8 << i)));
+ }
+ tmp = binop(Iop_OrV128, tmp, t);
+ } else {
+ assign(op2clean, mkexpr(op2));
+ }
+ assign(op3mask, unop(Iop_NotV128, tmp));
+
+ for (UChar shift = 0; shift < 128; shift += elem_bits) {
+ IRTemp s = newTemp(Ity_V128);
+ tmp = unop(Iop_NotV128,
+ binop(cmpeq_op, mkexpr(op2clean),
+ binop(Iop_ShrV128, mkexpr(op3), mkU8(shift))));
+ assign(s, binop(Iop_CmpEQ64x2, mkV128(0),
+ binop(Iop_AndV128, mkexpr(op3mask),
+ binop(Iop_ShlV128, tmp, mkU8(shift)))));
+ tmp = mkexpr(s);
+ if (shift < 64) {
+ tmp = binop(Iop_AndV128, tmp,
+ unop(Iop_Dup16x8, binop(Iop_GetElem16x8, tmp, mkU8(4))));
+ }
+ tmp = binop(Iop_AndV128, tmp,
+ unop(Iop_Dup16x8, mkU16(1 << (15 - shift / 8))));
+ if (shift)
+ match = binop(Iop_OrV128, mkexpr(mktemp(Ity_V128, match)), tmp);
+ else
+ match = tmp;
+ }
+ assign(result, unop(Iop_ClzNat64,
+ binop(Iop_Or64,
+ unop(Iop_V128HIto64, match),
+ mkU64((1UL << 48) - 1))));
+ put_vr_qw(v1, binop(Iop_64HLtoV128, mkexpr(result), mkU64(0)));
+
+ /* Set condition code.
+ 0: no match, no string terminator in op2
+ 1: no match, string terminator found
+ 2: full match
+ 3: partial match */
+ IRTemp cc = newTemp(Ity_I64);
+ tmp = binop(Iop_CmpLE64U,
+ binop(Iop_Add64, mkexpr(result), unop(Iop_8Uto64, mkexpr(op4))),
+ mkU64(16));
+ assign(cc, mkite(binop(Iop_CmpEQ64, mkexpr(result), mkU64(16)),
+ s390_vr_is_zs_set(m6) ? mkexpr(ccnomatch) : mkU64(0),
+ mkite(tmp, mkU64(2), mkU64(3))));
+ s390_cc_set(cc);
+
+ dis_res->hint = Dis_HintVerbose;
+ return "vstrs";
+}
+
static const HChar *
s390_irgen_VNC(UChar v1, UChar v2, UChar v3)
{
@@ -21596,6 +21695,11 @@ s390_decode_6byte_and_irgen(const UChar *bytes)
VRRd_v4(ovl), VRRd_m5(ovl),
VRRd_m6(ovl),
VRRd_rxb(ovl)); goto ok;
+ case 0xe7000000008bULL: s390_format_VRR_VVVVMM(s390_irgen_VSTRS, VRRd_v1(ovl),
+ VRRd_v2(ovl), VRRd_v3(ovl),
+ VRRd_v4(ovl), VRRd_m5(ovl),
+ VRRd_m6(ovl),
+ VRRd_rxb(ovl)); goto ok;
case 0xe7000000008cULL: s390_format_VRR_VVVV(s390_irgen_VPERM, VRR_v1(ovl),
VRR_v2(ovl), VRR_r3(ovl),
VRR_m4(ovl), VRR_rxb(ovl)); goto ok;
--
2.23.0
From 8a079b405467fa127c6c311d7ae3c649e76106c6 Mon Sep 17 00:00:00 2001
From: Andreas Arnez <arnez@linux.ibm.com>
Date: Tue, 16 Feb 2021 17:52:09 +0100
Subject: [PATCH 11/13] s390x: Mark arch13 features as supported
Make the STFLE instruction report the miscellaneous-instruction-extensions
facility 3 and the vector-enhancements facility 2 as supported. Indicate
support for the latter in the HWCAP vector as well.
---
VEX/priv/guest_s390_helpers.c | 9 +++------
coregrind/m_initimg/initimg-linux.c | 3 ++-
include/vki/vki-s390x-linux.h | 1 +
3 files changed, 6 insertions(+), 7 deletions(-)
diff --git a/VEX/priv/guest_s390_helpers.c b/VEX/priv/guest_s390_helpers.c
index 1e04f601a..804b92a29 100644
--- a/VEX/priv/guest_s390_helpers.c
+++ b/VEX/priv/guest_s390_helpers.c
@@ -356,9 +356,7 @@ s390x_dirtyhelper_STFLE(VexGuestS390XState *guest_state, ULong *addr)
| s390_stfle_range(51, 55)
/* 56: unassigned */
/* 57: MSA5, not supported */
- | s390_stfle_range(58, 60)
- /* 61: miscellaneous-instruction 3, not supported */
- | s390_stfle_range(62, 63)),
+ | s390_stfle_range(58, 63)),
/* === 64 .. 127 === */
(s390_stfle_range(64, 72)
@@ -384,11 +382,10 @@ s390x_dirtyhelper_STFLE(VexGuestS390XState *guest_state, ULong *addr)
/* 143: unassigned */
| s390_stfle_range(144, 145)
/* 146: MSA8, not supported */
- | s390_stfle_range(147, 147)
- /* 148: vector-enhancements 2, not supported */
- | s390_stfle_range(149, 149)
+ | s390_stfle_range(147, 149)
/* 150: unassigned */
/* 151: DEFLATE-conversion, not supported */
+ /* 152: vector packed decimal enhancement, not supported */
/* 153: unassigned */
/* 154: unassigned */
/* 155: MSA9, not supported */
diff --git a/coregrind/m_initimg/initimg-linux.c b/coregrind/m_initimg/initimg-linux.c
index fc1a32ecf..37d005168 100644
--- a/coregrind/m_initimg/initimg-linux.c
+++ b/coregrind/m_initimg/initimg-linux.c
@@ -703,7 +703,8 @@ Addr setup_client_stack( void* init_sp,
itself, is not supported by Valgrind. */
auxv->u.a_val &= ((VKI_HWCAP_S390_TE - 1)
| VKI_HWCAP_S390_VXRS
- | VKI_HWCAP_S390_VXRS_EXT);
+ | VKI_HWCAP_S390_VXRS_EXT
+ | VKI_HWCAP_S390_VXRS_EXT2);
}
# elif defined(VGP_arm64_linux)
{
diff --git a/include/vki/vki-s390x-linux.h b/include/vki/vki-s390x-linux.h
index 4ab2d3334..71b363029 100644
--- a/include/vki/vki-s390x-linux.h
+++ b/include/vki/vki-s390x-linux.h
@@ -807,6 +807,7 @@ typedef vki_s390_regs vki_elf_gregset_t;
#define VKI_HWCAP_S390_TE 1024
#define VKI_HWCAP_S390_VXRS 2048
#define VKI_HWCAP_S390_VXRS_EXT 8192
+#define VKI_HWCAP_S390_VXRS_EXT2 32768
//----------------------------------------------------------------------
--
2.23.0
From 1461d9b8d0b12e55b648fbf50c5dcee30785afa2 Mon Sep 17 00:00:00 2001
From: Andreas Arnez <arnez@linux.ibm.com>
Date: Mon, 17 May 2021 15:34:15 +0200
Subject: [PATCH 12/13] s390x: Vec-enh-2, test cases
Add test cases for verifying the new/enhanced instructions in the
vector-enhancements facility 2. For "vector string search" VSTRS add a
memcheck test case.
---
.gitignore | 2 +
memcheck/tests/s390x/Makefile.am | 3 +-
memcheck/tests/s390x/vstrs.c | 68 ++++++
memcheck/tests/s390x/vstrs.stderr.exp | 16 ++
memcheck/tests/s390x/vstrs.stdout.exp | 0
memcheck/tests/s390x/vstrs.vgtest | 2 +
none/tests/s390x/Makefile.am | 3 +-
none/tests/s390x/vec2.c | 314 ++++++++++++++++++++++++++
none/tests/s390x/vec2.stderr.exp | 2 +
none/tests/s390x/vec2.stdout.exp | 168 ++++++++++++++
none/tests/s390x/vec2.vgtest | 2 +
tests/s390x_features.c | 4 +
12 files changed, 582 insertions(+), 2 deletions(-)
create mode 100644 memcheck/tests/s390x/vstrs.c
create mode 100644 memcheck/tests/s390x/vstrs.stderr.exp
create mode 100644 memcheck/tests/s390x/vstrs.stdout.exp
create mode 100644 memcheck/tests/s390x/vstrs.vgtest
create mode 100644 none/tests/s390x/vec2.c
create mode 100644 none/tests/s390x/vec2.stderr.exp
create mode 100644 none/tests/s390x/vec2.stdout.exp
create mode 100644 none/tests/s390x/vec2.vgtest
diff --git a/memcheck/tests/s390x/Makefile.am b/memcheck/tests/s390x/Makefile.am
index d183841ef..668fd9933 100644
--- a/memcheck/tests/s390x/Makefile.am
+++ b/memcheck/tests/s390x/Makefile.am
@@ -2,7 +2,7 @@ include $(top_srcdir)/Makefile.tool-tests.am
dist_noinst_SCRIPTS = filter_stderr
-INSN_TESTS = cdsg cu21 cu42 ltgjhe vstrc vfae vistr
+INSN_TESTS = cdsg cu21 cu42 ltgjhe vstrc vfae vistr vstrs
check_PROGRAMS = $(INSN_TESTS)
@@ -18,3 +18,4 @@ AM_CCASFLAGS += @FLAG_M64@
vstrc_CFLAGS = $(AM_CFLAGS) -march=z13
vfae_CFLAGS = $(AM_CFLAGS) -march=z13
vistr_CFLAGS = $(AM_CFLAGS) -march=z13
+vstrs_CFLAGS = $(AM_CFLAGS) -march=z13
diff --git a/memcheck/tests/s390x/vstrs.c b/memcheck/tests/s390x/vstrs.c
new file mode 100644
index 000000000..3354c2e53
--- /dev/null
+++ b/memcheck/tests/s390x/vstrs.c
@@ -0,0 +1,68 @@
+#include <stdio.h>
+#include <string.h>
+
+#define VECTOR __attribute__ ((vector_size (16)))
+
+typedef char VECTOR char_v;
+
+volatile char tmp;
+static const char *hex_digit = "0123456789abcdefGHIJKLMNOPQRSTUV";
+
+static char_v to_char_vec(const char *str)
+{
+ char buf[17];
+ char_v v;
+
+ for (int i = 0; i < sizeof(buf); i++) {
+ char ch = str[i];
+ if (ch == '\0')
+ break;
+ else if (ch == '$')
+ buf[i] = '\0';
+ else if (ch != '~')
+ buf[i] = ch;
+ }
+ v = *(char_v *) buf;
+ return v;
+}
+
+static void test_vstrs_char(const char *haystack, const char *needle,
+ int expect_res, int expect_cc)
+{
+ int cc;
+ char_v v2val = to_char_vec(haystack);
+ char_v v3val = to_char_vec(needle);
+
+ register unsigned long VECTOR v4 __asm__("v4") = { strlen(needle), 0 };
+ register char_v v1 __asm__("v1");
+ register char_v v2 __asm__("v2") = v2val;
+ register char_v v3 __asm__("v3") = v3val;
+
+ __asm__(
+ "cr 0,0\n\t" /* Clear CC */
+ ".short 0xe712,0x3020,0x408b\n\t" /* vstrs %v1,%v2,%v3,%v4,0,2 */
+ "ipm %[cc]\n\t"
+ "srl %[cc],28"
+ : "=v" (v1), [cc] "=d" (cc)
+ : "v" (v2), "v" (v3), "v" (v4)
+ : "cc");
+
+ tmp = hex_digit[v1[7] & 0x1f];
+ if (expect_res >= 0 && v1[7] != expect_res)
+ printf("result %u != %d\n", v1[7], expect_res);
+
+ tmp = hex_digit[cc & 0xf];
+ if (expect_cc >= 0 && cc != expect_cc)
+ printf("CC %d != %d\n", cc, expect_cc);
+}
+
+int main()
+{
+ test_vstrs_char("haystack$needle", "needle$haystack", 16, 1);
+ test_vstrs_char("haystack, needle", "needle, haystack", 10, 3);
+ test_vstrs_char("ABCDEFGH", "DEFGHI", -1, -1);
+ test_vstrs_char("match in UNDEF", "UN", 9, 2);
+ test_vstrs_char("after ~ UNDEF", "DEF", -1, -1);
+ test_vstrs_char("", "", 0, 2);
+ return 0;
+}
diff --git a/memcheck/tests/s390x/vstrs.stderr.exp b/memcheck/tests/s390x/vstrs.stderr.exp
new file mode 100644
index 000000000..c5c3ef705
--- /dev/null
+++ b/memcheck/tests/s390x/vstrs.stderr.exp
@@ -0,0 +1,16 @@
+Use of uninitialised value of size 8
+ at 0x........: test_vstrs_char (vstrs.c:50)
+ by 0x........: main (vstrs.c:63)
+
+Use of uninitialised value of size 8
+ at 0x........: test_vstrs_char (vstrs.c:54)
+ by 0x........: main (vstrs.c:63)
+
+Use of uninitialised value of size 8
+ at 0x........: test_vstrs_char (vstrs.c:50)
+ by 0x........: main (vstrs.c:65)
+
+Use of uninitialised value of size 8
+ at 0x........: test_vstrs_char (vstrs.c:54)
+ by 0x........: main (vstrs.c:65)
+
diff --git a/memcheck/tests/s390x/vstrs.stdout.exp b/memcheck/tests/s390x/vstrs.stdout.exp
new file mode 100644
index 000000000..e69de29bb
diff --git a/memcheck/tests/s390x/vstrs.vgtest b/memcheck/tests/s390x/vstrs.vgtest
new file mode 100644
index 000000000..fd2a29873
--- /dev/null
+++ b/memcheck/tests/s390x/vstrs.vgtest
@@ -0,0 +1,2 @@
+prog: vstrs
+vgopts: -q
diff --git a/none/tests/s390x/Makefile.am b/none/tests/s390x/Makefile.am
index 2fd45ec1e..ca38db935 100644
--- a/none/tests/s390x/Makefile.am
+++ b/none/tests/s390x/Makefile.am
@@ -20,7 +20,7 @@ INSN_TESTS = clc clcle cvb cvd icm lpr tcxb lam_stam xc mvst add sub mul \
spechelper-icm-1 spechelper-icm-2 spechelper-tmll \
spechelper-tm laa vector lsc2 ppno vector_string vector_integer \
vector_float add-z14 sub-z14 mul-z14 bic \
- misc3
+ misc3 vec2
if BUILD_DFP_TESTS
INSN_TESTS += dfp-1 dfp-2 dfp-3 dfp-4 dfptest dfpext dfpconv srnmt pfpo
@@ -74,3 +74,4 @@ lsc2_CFLAGS = -march=z13 -DS390_TESTS_NOCOLOR
vector_string_CFLAGS = $(AM_CFLAGS) -march=z13 -DS390_TEST_COUNT=5
vector_integer_CFLAGS = $(AM_CFLAGS) -march=z13 -DS390_TEST_COUNT=4
vector_float_CFLAGS = $(AM_CFLAGS) -march=z13 -DS390_TEST_COUNT=4
+vec2_CFLAGS = $(AM_CFLAGS) -march=z13
diff --git a/none/tests/s390x/vec2.c b/none/tests/s390x/vec2.c
new file mode 100644
index 000000000..73b04dee4
--- /dev/null
+++ b/none/tests/s390x/vec2.c
@@ -0,0 +1,314 @@
+#include <stdio.h>
+
+#define VECTOR __attribute__ ((vector_size (16)))
+
+typedef unsigned long VECTOR ulong_v;
+typedef float VECTOR float_v;
+
+static const ulong_v vec_a = { 0x0123456789abcdef, 0xfedcba9876543210 };
+static const ulong_v vec_b = { 0xfedcba9876543210, 0x0123456789abcdef };
+static const ulong_v vec_c = { 0x8040201008040201, 0x7fbfdfeff7fbfdfe };
+static const ulong_v vec_one = { -1, -1 };
+static const ulong_v vec_ini = { 0x0112233445566778, 0x899aabbccddeeff0 };
+
+static const float_v vec_fa = { 16777215., -16777215., 42.5, 10000. };
+static const float_v vec_fb = { 4., 3., 2., 1. };
+
+/* -- Vector shift -- */
+
+#define TEST_GENERATE(insn) \
+ static void test_##insn(ulong_v a, ulong_v b) \
+ { \
+ ulong_v out; \
+ __asm__( \
+ #insn " %[out],%[a],%[b]" \
+ : [out] "=v" (out) \
+ : [a] "v" (a), \
+ [b] "v" (b) \
+ : ); \
+ printf("\t%016lx %016lx\n", out[0], out[1]); \
+ }
+
+#define TEST_EXEC(insn) \
+ do { \
+ puts(#insn); \
+ test_##insn(vec_a, vec_b); \
+ test_##insn(vec_b, vec_a); \
+ test_##insn(vec_c, vec_a); \
+ test_##insn(vec_one, vec_b); \
+ } while (0)
+
+#define INSNS \
+ XTEST(vsl); \
+ XTEST(vsrl); \
+ XTEST(vsra);
+
+#define XTEST TEST_GENERATE
+INSNS
+#undef XTEST
+
+static void test_all_single_bitshifts()
+{
+#define XTEST TEST_EXEC
+ INSNS
+#undef XTEST
+}
+#undef INSNS
+#undef TEST_EXEC
+#undef TEST_GENERATE
+
+/* -- Vector load element-/byte-swapped -- */
+
+#define TEST_EXEC(opc1,opc2,insn,m3) \
+ do { \
+ puts(#insn " " #m3); \
+ test_##insn##_##m3(vec_a); \
+ test_##insn##_##m3(vec_b); \
+ } while (0)
+
+#define TEST_GENERATE(opc1,opc2,insn,m3) \
+ static void test_##insn##_##m3(ulong_v a) \
+ { \
+ ulong_v out = vec_ini; \
+ __asm__( \
+ ".insn vrx,0x" #opc1 "00000000" #opc2 ",%[out],%[a]," #m3 \
+ : [out] "+v" (out) \
+ : [a] "R" (a) \
+ : ); \
+ printf("\t%016lx %016lx\n", out[0], out[1]); \
+ }
+
+#define INSNS \
+ XTEST(e6,01, vlebrh, 0); \
+ XTEST(e6,01, vlebrh, 7); \
+ XTEST(e6,01, vlebrh, 2); \
+ XTEST(e6,03, vlebrf, 0); \
+ XTEST(e6,03, vlebrf, 3); \
+ XTEST(e6,03, vlebrf, 1); \
+ XTEST(e6,02, vlebrg, 0); \
+ XTEST(e6,02, vlebrg, 1); \
+ XTEST(e6,04, vllebrz, 1); \
+ XTEST(e6,04, vllebrz, 2); \
+ XTEST(e6,04, vllebrz, 3); \
+ XTEST(e6,04, vllebrz, 6); \
+ XTEST(e6,05, vlbrrep, 1); \
+ XTEST(e6,05, vlbrrep, 2); \
+ XTEST(e6,05, vlbrrep, 3); \
+ XTEST(e6,06, vlbr, 1); \
+ XTEST(e6,06, vlbr, 2); \
+ XTEST(e6,06, vlbr, 3); \
+ XTEST(e6,06, vlbr, 4); \
+ XTEST(e6,07, vler, 1); \
+ XTEST(e6,07, vler, 2); \
+ XTEST(e6,07, vler, 3);
+
+#define XTEST TEST_GENERATE
+INSNS
+#undef XTEST
+
+static void test_all_swapped_loads()
+{
+#define XTEST TEST_EXEC
+ INSNS
+#undef XTEST
+}
+
+#undef INSNS
+#undef TEST_GENERATE
+
+/* -- Vector store element-/byte-swapped -- */
+
+#define TEST_GENERATE(opc1,opc2,insn,m3) \
+ static void test_##insn##_##m3(ulong_v a) \
+ { \
+ ulong_v out = vec_ini; \
+ __asm__( \
+ ".insn vrx,0x" #opc1 "00000000" #opc2 ",%[a],%[out]," #m3 \
+ : [out] "+R" (out) \
+ : [a] "v" (a) \
+ : ); \
+ printf("\t%016lx %016lx\n", out[0], out[1]); \
+ }
+
+#define INSNS \
+ XTEST(e6,09, vstebrh, 0); \
+ XTEST(e6,09, vstebrh, 7); \
+ XTEST(e6,09, vstebrh, 2); \
+ XTEST(e6,0b, vstebrf, 0); \
+ XTEST(e6,0b, vstebrf, 3); \
+ XTEST(e6,0b, vstebrf, 1); \
+ XTEST(e6,0a, vstebrg, 0); \
+ XTEST(e6,0a, vstebrg, 1); \
+ XTEST(e6,0e, vstbr, 1); \
+ XTEST(e6,0e, vstbr, 2); \
+ XTEST(e6,0e, vstbr, 3); \
+ XTEST(e6,0e, vstbr, 4); \
+ XTEST(e6,0f, vster, 1); \
+ XTEST(e6,0f, vster, 2); \
+ XTEST(e6,0f, vster, 3);
+
+#define XTEST TEST_GENERATE
+INSNS
+#undef XTEST
+
+static void test_all_swapped_stores()
+{
+#define XTEST TEST_EXEC
+ INSNS
+#undef XTEST
+}
+
+#undef INSNS
+#undef TEST_EXEC
+#undef TEST_GENERATE
+
+/* -- Vector shift double by bit -- */
+
+#define TEST_GENERATE(opc1,opc2,insn,i4) \
+ static void test_##insn##_##i4(ulong_v a, ulong_v b) \
+ { \
+ ulong_v out = vec_ini; \
+ __asm__( \
+ ".insn vrr,0x" #opc1 "00000000" #opc2 \
+ ",%[out],%[a],%[b],0," #i4 ",0" \
+ : [out] "+v" (out) \
+ : [a] "v" (a), \
+ [b] "v" (b) \
+ : ); \
+ printf("\t%016lx %016lx\n", out[0], out[1]); \
+ }
+
+#define TEST_EXEC(opc1,opc2,insn,i4) \
+ do { \
+ puts(#insn " " #i4); \
+ test_##insn##_##i4(vec_a, vec_one); \
+ test_##insn##_##i4(vec_b, vec_a); \
+ } while (0)
+
+#define INSNS \
+ XTEST(e7,86,vsld,0); \
+ XTEST(e7,86,vsld,7); \
+ XTEST(e7,86,vsld,4); \
+ XTEST(e7,87,vsrd,0); \
+ XTEST(e7,87,vsrd,7); \
+ XTEST(e7,87,vsrd,4);
+
+#define XTEST TEST_GENERATE
+INSNS
+#undef XTEST
+
+static void test_all_double_bitshifts()
+{
+#define XTEST TEST_EXEC
+ INSNS
+#undef XTEST
+}
+
+#undef INSNS
+#undef TEST_EXEC
+#undef TEST_GENERATE
+
+/* -- Vector integer -> FP conversions -- */
+
+#define TEST_GENERATE(opc1,opc2,insn,m4) \
+ static void test_##insn##_##m4(ulong_v a) \
+ { \
+ float_v out; \
+ __asm__( \
+ ".insn vrr,0x" #opc1 "00000000" #opc2 \
+ ",%[out],%[a],0,2," #m4 ",0" \
+ : [out] "=v" (out) \
+ : [a] "v" (a) \
+ : ); \
+ if (m4 & 8) \
+ printf("\t%a - - -\n", out[0]); \
+ else \
+ printf("\t%a %a %a %a\n", out[0], out[1], out[2], out[3]); \
+ }
+
+#define TEST_EXEC(opc1,opc2,insn,m4) \
+ do { \
+ puts(#insn " " #m4); \
+ test_##insn##_##m4(vec_a); \
+ test_##insn##_##m4(vec_c); \
+ } while (0)
+
+#define INSNS \
+ XTEST(e7,c1,vcfpl,0); \
+ XTEST(e7,c1,vcfpl,8); \
+ XTEST(e7,c3,vcfps,0); \
+ XTEST(e7,c3,vcfps,8);
+
+#define XTEST TEST_GENERATE
+INSNS
+#undef XTEST
+
+static void test_all_int_fp_conversions()
+{
+#define XTEST TEST_EXEC
+ INSNS
+#undef XTEST
+}
+
+#undef INSNS
+#undef TEST_EXEC
+#undef TEST_GENERATE
+
+/* -- Vector FP -> integer conversions -- */
+
+#define TEST_GENERATE(opc1,opc2,insn,m4) \
+ static void test_##insn##_##m4(float_v a) \
+ { \
+ unsigned int VECTOR out; \
+ __asm__( \
+ ".insn vrr,0x" #opc1 "00000000" #opc2 \
+ ",%[out],%[a],0,2," #m4 ",0" \
+ : [out] "=v" (out) \
+ : [a] "v" (a) \
+ : ); \
+ if (m4 & 8) \
+ printf("\t%08x - - -\n", out[0]); \
+ else \
+ printf("\t%08x %08x %08x %08x\n", \
+ out[0], out[1], out[2], out[3]); \
+ }
+
+#define TEST_EXEC(opc1,opc2,insn,m4) \
+ do { \
+ puts(#insn " " #m4); \
+ test_##insn##_##m4(vec_fa); \
+ test_##insn##_##m4(vec_fb); \
+ } while (0)
+
+#define INSNS \
+ XTEST(e7,c0,vclfp,0); \
+ XTEST(e7,c0,vclfp,8); \
+ XTEST(e7,c2,vcsfp,0); \
+ XTEST(e7,c2,vcsfp,8);
+
+#define XTEST TEST_GENERATE
+INSNS
+#undef XTEST
+
+static void test_all_fp_int_conversions()
+{
+#define XTEST TEST_EXEC
+ INSNS
+#undef XTEST
+}
+
+#undef INSNS
+#undef TEST_EXEC
+#undef TEST_GENERATE
+
+
+int main()
+{
+ test_all_single_bitshifts();
+ test_all_swapped_loads();
+ test_all_swapped_stores();
+ test_all_double_bitshifts();
+ test_all_int_fp_conversions();
+ test_all_fp_int_conversions();
+ return 0;
+}
diff --git a/none/tests/s390x/vec2.stderr.exp b/none/tests/s390x/vec2.stderr.exp
new file mode 100644
index 000000000..139597f9c
--- /dev/null
+++ b/none/tests/s390x/vec2.stderr.exp
@@ -0,0 +1,2 @@
+
+
diff --git a/none/tests/s390x/vec2.stdout.exp b/none/tests/s390x/vec2.stdout.exp
new file mode 100644
index 000000000..b32cbe1bc
--- /dev/null
+++ b/none/tests/s390x/vec2.stdout.exp
@@ -0,0 +1,168 @@
+vsl
+ 483415676abc37ef fde5533beca14200
+ fde5533beca14200 483415676abc37ef
+ 00010204102040bf effd7feffebff7fe
+ ffffffffffffffff ffffffffffffff80
+vsrl
+ 0012d1679e9af3ef ffdbe5753bcaa164
+ 7fdbe5753bcaa164 4012d1679e9af3ef
+ 4008014004002004 05fbf7efbf7ffffe
+ 03ffffffffffffff ffffffffffffffff
+vsra
+ 0012d1679e9af3ef ffdbe5753bcaa164
+ ffdbe5753bcaa164 4012d1679e9af3ef
+ c008014004002004 05fbf7efbf7ffffe
+ ffffffffffffffff ffffffffffffffff
+vlebrh 0
+ 2301233445566778 899aabbccddeeff0
+ dcfe233445566778 899aabbccddeeff0
+vlebrh 7
+ 0112233445566778 899aabbccdde2301
+ 0112233445566778 899aabbccddedcfe
+vlebrh 2
+ 0112233423016778 899aabbccddeeff0
+ 01122334dcfe6778 899aabbccddeeff0
+vlebrf 0
+ 6745230145566778 899aabbccddeeff0
+ 98badcfe45566778 899aabbccddeeff0
+vlebrf 3
+ 0112233445566778 899aabbc67452301
+ 0112233445566778 899aabbc98badcfe
+vlebrf 1
+ 0112233467452301 899aabbccddeeff0
+ 0112233498badcfe 899aabbccddeeff0
+vlebrg 0
+ efcdab8967452301 899aabbccddeeff0
+ 1032547698badcfe 899aabbccddeeff0
+vlebrg 1
+ 0112233445566778 efcdab8967452301
+ 0112233445566778 1032547698badcfe
+vllebrz 1
+ 0000000000002301 0000000000000000
+ 000000000000dcfe 0000000000000000
+vllebrz 2
+ 0000000067452301 0000000000000000
+ 0000000098badcfe 0000000000000000
+vllebrz 3
+ efcdab8967452301 0000000000000000
+ 1032547698badcfe 0000000000000000
+vllebrz 6
+ 6745230100000000 0000000000000000
+ 98badcfe00000000 0000000000000000
+vlbrrep 1
+ 2301230123012301 2301230123012301
+ dcfedcfedcfedcfe dcfedcfedcfedcfe
+vlbrrep 2
+ 6745230167452301 6745230167452301
+ 98badcfe98badcfe 98badcfe98badcfe
+vlbrrep 3
+ efcdab8967452301 efcdab8967452301
+ 1032547698badcfe 1032547698badcfe
+vlbr 1
+ 23016745ab89efcd dcfe98ba54761032
+ dcfe98ba54761032 23016745ab89efcd
+vlbr 2
+ 67452301efcdab89 98badcfe10325476
+ 98badcfe10325476 67452301efcdab89
+vlbr 3
+ efcdab8967452301 1032547698badcfe
+ 1032547698badcfe efcdab8967452301
+vlbr 4
+ 1032547698badcfe efcdab8967452301
+ efcdab8967452301 1032547698badcfe
+vler 1
+ 32107654ba98fedc cdef89ab45670123
+ cdef89ab45670123 32107654ba98fedc
+vler 2
+ 76543210fedcba98 89abcdef01234567
+ 89abcdef01234567 76543210fedcba98
+vler 3
+ fedcba9876543210 0123456789abcdef
+ 0123456789abcdef fedcba9876543210
+vstebrh 0
+ 2301233445566778 899aabbccddeeff0
+ dcfe233445566778 899aabbccddeeff0
+vstebrh 7
+ 1032233445566778 899aabbccddeeff0
+ efcd233445566778 899aabbccddeeff0
+vstebrh 2
+ ab89233445566778 899aabbccddeeff0
+ 5476233445566778 899aabbccddeeff0
+vstebrf 0
+ 6745230145566778 899aabbccddeeff0
+ 98badcfe45566778 899aabbccddeeff0
+vstebrf 3
+ 1032547645566778 899aabbccddeeff0
+ efcdab8945566778 899aabbccddeeff0
+vstebrf 1
+ efcdab8945566778 899aabbccddeeff0
+ 1032547645566778 899aabbccddeeff0
+vstebrg 0
+ efcdab8967452301 899aabbccddeeff0
+ 1032547698badcfe 899aabbccddeeff0
+vstebrg 1
+ 1032547698badcfe 899aabbccddeeff0
+ efcdab8967452301 899aabbccddeeff0
+vstbr 1
+ 23016745ab89efcd dcfe98ba54761032
+ dcfe98ba54761032 23016745ab89efcd
+vstbr 2
+ 67452301efcdab89 98badcfe10325476
+ 98badcfe10325476 67452301efcdab89
+vstbr 3
+ efcdab8967452301 1032547698badcfe
+ 1032547698badcfe efcdab8967452301
+vstbr 4
+ 1032547698badcfe efcdab8967452301
+ efcdab8967452301 1032547698badcfe
+vster 1
+ 32107654ba98fedc cdef89ab45670123
+ cdef89ab45670123 32107654ba98fedc
+vster 2
+ 76543210fedcba98 89abcdef01234567
+ 89abcdef01234567 76543210fedcba98
+vster 3
+ fedcba9876543210 0123456789abcdef
+ 0123456789abcdef fedcba9876543210
+vsld 0
+ 0123456789abcdef fedcba9876543210
+ fedcba9876543210 0123456789abcdef
+vsld 7
+ 91a2b3c4d5e6f7ff 6e5d4c3b2a19087f
+ 6e5d4c3b2a190800 91a2b3c4d5e6f780
+vsld 4
+ 123456789abcdeff edcba9876543210f
+ edcba98765432100 123456789abcdef0
+vsrd 0
+ ffffffffffffffff ffffffffffffffff
+ 0123456789abcdef fedcba9876543210
+vsrd 7
+ 21ffffffffffffff ffffffffffffffff
+ de02468acf13579b dffdb97530eca864
+vsrd 4
+ 0fffffffffffffff ffffffffffffffff
+ f0123456789abcde ffedcba987654321
+vcfpl 0
+ 0x1.234568p+24 0x1.13579cp+31 0x1.fdb976p+31 0x1.d950c8p+30
+ 0x1.00804p+31 0x1.00804p+27 0x1.feff8p+30 0x1.eff7fcp+31
+vcfpl 8
+ 0x1.234568p+24 - - -
+ 0x1.00804p+31 - - -
+vcfps 0
+ 0x1.234568p+24 -0x1.d950c8p+30 -0x1.234568p+24 0x1.d950c8p+30
+ -0x1.feff8p+30 0x1.00804p+27 0x1.feff8p+30 -0x1.00804p+27
+vcfps 8
+ 0x1.234568p+24 - - -
+ -0x1.feff8p+30 - - -
+vclfp 0
+ 00ffffff 00000000 0000002a 00002710
+ 00000004 00000003 00000002 00000001
+vclfp 8
+ 00ffffff - - -
+ 00000004 - - -
+vcsfp 0
+ 00ffffff ff000001 0000002a 00002710
+ 00000004 00000003 00000002 00000001
+vcsfp 8
+ 00ffffff - - -
+ 00000004 - - -
diff --git a/none/tests/s390x/vec2.vgtest b/none/tests/s390x/vec2.vgtest
new file mode 100644
index 000000000..45e942e64
--- /dev/null
+++ b/none/tests/s390x/vec2.vgtest
@@ -0,0 +1,2 @@
+prog: vec2
+prereq: test -e vec2 && ../../../tests/s390x_features s390x-vx
diff --git a/tests/s390x_features.c b/tests/s390x_features.c
index 25b98f3a3..e7939c463 100644
--- a/tests/s390x_features.c
+++ b/tests/s390x_features.c
@@ -270,6 +270,10 @@ static int go(char *feature, char *cpu)
match = facilities[0] & FAC_BIT(57); /* message security assist 5 facility */
} else if (strcmp(feature, "s390x-mi2") == 0 ) {
match = facilities[0] & FAC_BIT(58);
+ } else if (strcmp(feature, "s390x-mi3") == 0 ) {
+ match = facilities[0] & FAC_BIT(61);
+ } else if (strcmp(feature, "s390x-vx2") == 0 ) {
+ match = facilities[2] & FAC_BIT(20);
} else {
return 2; // Unrecognised feature.
}
--
2.23.0
From d9364bc90ee894c43ee742840f806571edc08ab3 Mon Sep 17 00:00:00 2001
From: Andreas Arnez <arnez@linux.ibm.com>
Date: Tue, 18 May 2021 19:59:32 +0200
Subject: [PATCH 13/13] s390x: Wrap up misc-insn-3 and vec-enh-2 support
Wrap up support for the miscellaneous-instruction-extensions facility 3
and the vector-enhancements facility 2: Add 'case' statements for the
remaining unhandled arch13 instructions to 'guest_s390_toIR.c', document
the new support in 's390-opcodes.csv', adjust 's390-check-opcodes.pl', and
announce the new feature in 'NEWS'.
---
NEWS | 5 ++
VEX/priv/guest_s390_toIR.c | 5 +-
auxprogs/s390-check-opcodes.pl | 22 ++++++++-
docs/internals/s390-opcodes.csv | 81 +++++++++++++++++++++++++++++++--
4 files changed, 108 insertions(+), 5 deletions(-)
diff --git a/VEX/priv/guest_s390_toIR.c b/VEX/priv/guest_s390_toIR.c
index 46a867475..1bd18f760 100644
--- a/VEX/priv/guest_s390_toIR.c
+++ b/VEX/priv/guest_s390_toIR.c
@@ -8,7 +8,7 @@
This file is part of Valgrind, a dynamic binary instrumentation
framework.
- Copyright IBM Corp. 2010-2020
+ Copyright IBM Corp. 2010-2021
This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License as
@@ -20503,6 +20503,9 @@ s390_decode_4byte_and_irgen(const UChar *bytes)
RRE_r2(ovl)); goto ok;
case 0xb931: s390_format_RRE_RR(s390_irgen_CLGFR, RRE_r1(ovl),
RRE_r2(ovl)); goto ok;
+ case 0xb938: /* SORTL */ goto unimplemented;
+ case 0xb939: /* DFLTCC */ goto unimplemented;
+ case 0xb93a: /* KDSA */ goto unimplemented;
case 0xb93c: s390_format_RRE_RR(s390_irgen_PPNO, RRE_r1(ovl),
RRE_r2(ovl)); goto ok;
case 0xb93e: /* KIMD */ goto unimplemented;
--
2.23.0