gcc/gcc14-pr101523.patch

77 lines
2.5 KiB
Diff
Raw Normal View History

2024-05-20 15:51:07 +00:00
Based on Richi's patch:
<https://gcc.gnu.org/pipermail/gcc-patches/2024-April/648725.html>
~~
The following avoids re-walking and re-combining the instructions
between i2 and i3 when the pattern of i2 doesn't change.
Bootstrap and regtest running ontop of a reversal of
r14-9692-g839bc42772ba7a.
2024-06-10 17:54:05 +00:00
It brings down memory use from 9GB to 400MB and compile-time from
2024-05-20 15:51:07 +00:00
80s to 3.5s. r14-9692-g839bc42772ba7a does better in both metrics
2024-06-10 17:54:05 +00:00
but has shown code generation regressions across architectures.
2024-05-20 15:51:07 +00:00
PR rtl-optimization/101523
* combine.cc (try_combine): When the pattern of i2 doesn't
change do not re-start combining at i2 or an earlier insn which
had links or notes added.
~~
But, since the patch affects code generation (for instance,
libstdc++-v3/src/c++17/floating_from_chars.o), we limit the bailing out
only when I2 hasn't been changed 1000x. I've measured how many times
at most is I2 unchanged during a bootstrap + regtest.
x86: 134
aarch64: 736 (gimple-match-1.cc)
s390x: 635 (gimple-match-*)
ppc64le: 620 (gimple-match-*)
while certain pathological testcases trigger it more than 10,000 times.
With the limit in place this patch doesn't affect common code.
--- a/gcc/combine.cc
+++ b/gcc/combine.cc
@@ -92,6 +92,11 @@ along with GCC; see the file COPYING3. If not see
#include "function-abi.h"
#include "rtlanal.h"
+/* Number of times I2 didn't change in try_combine. Used to prevent a
+ combinatorial explosion. */
+
2024-06-10 17:54:05 +00:00
+static int combine_i2_unchanged;
2024-05-20 15:51:07 +00:00
+
/* Number of attempts to combine instructions in this function. */
static int combine_attempts;
@@ -1127,6 +1132,7 @@ combine_instructions (rtx_insn *f, unsigned int nregs)
return false;
combine_attempts = 0;
2024-06-10 17:54:05 +00:00
+ combine_i2_unchanged = 0;
2024-05-20 15:51:07 +00:00
combine_merges = 0;
combine_extras = 0;
combine_successes = 0;
@@ -4196,6 +4201,10 @@ try_combine (rtx_insn *i3, rtx_insn *i2, rtx_insn *i1, rtx_insn *i0,
adjust_for_new_dest (i3);
}
2024-06-10 17:54:05 +00:00
+ bool i2_i2_unchanged = false;
2024-05-20 15:51:07 +00:00
+ if (rtx_equal_p (newi2pat, PATTERN (i2)))
2024-06-10 17:54:05 +00:00
+ i2_i2_unchanged = true;
2024-05-20 15:51:07 +00:00
+
/* We now know that we can do this combination. Merge the insns and
update the status of registers and LOG_LINKS. */
@@ -4762,6 +4771,13 @@ try_combine (rtx_insn *i3, rtx_insn *i2, rtx_insn *i1, rtx_insn *i0,
combine_successes++;
undo_commit ();
2024-06-10 17:54:05 +00:00
+ if (i2_i2_unchanged)
2024-05-20 15:51:07 +00:00
+ {
2024-06-10 17:54:05 +00:00
+ if (combine_i2_unchanged == 1000)
2024-05-20 15:51:07 +00:00
+ return i3;
2024-06-10 17:54:05 +00:00
+ ++combine_i2_unchanged;
2024-05-20 15:51:07 +00:00
+ }
+
rtx_insn *ret = newi2pat ? i2 : i3;
if (added_links_insn && DF_INSN_LUID (added_links_insn) < DF_INSN_LUID (ret))
ret = added_links_insn;