pool/gcc13
SHA256
2
0
Fork 1
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

OBS-URL: https://build.opensuse.org/package/show/devel:gcc/gcc13?expand=0&rev=103

Browse Source
This commit is contained in:
Richard Biener 2024-04-04 07:46:22 +00:00 committed by Git OBS Bridge
parent 6df8046a37
commit d44b58ed57
2 changed files with 27 additions and 29 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

54
gcc13-pr101523.patch
View File

@ -1,48 +1,46 @@
From 839bc42772ba7af66af3bd16efed4a69511312ae Mon Sep 17 00:00:00 2001 From 5729ffad4db78b26f0d1f658a9fd695be5970550 Mon Sep 17 00:00:00 2001
From: Segher Boessenkool <segher@kernel.crashing.org> From: Richard Biener <rguenther@suse.de>
Date: Wed, 27 Mar 2024 14:09:52 +0000 Date: Wed, 3 Apr 2024 12:58:26 +0200
Subject: [PATCH] combine: Don't combine if I2 does not change Subject: [PATCH] rtl-optimization/101523 - avoid re-combine after noop 2->2
combination
To: gcc-patches@gcc.gnu.org To: gcc-patches@gcc.gnu.org
In some cases combine will "combine" an I2 and I3, but end up putting The following avoids re-walking and re-combining the instructions
exactly the same thing back as I2 as was there before. This is never between i2 and i3 when the pattern of i2 doesn't change.
progress, so we shouldn't do it, it will lead to oscillating behaviour
and the like.
If we want to canonicalise things, that's fine, but this is not the
way to do it.
2024-03-27 Segher Boessenkool <segher@kernel.crashing.org>
PR rtl-optimization/101523 PR rtl-optimization/101523
* combine.cc (try_combine): Don't do a 2-insn combination if * combine.cc (try_combine): When the pattern of i2 doesn't
it does not in fact change I2. change do not re-start combining at i2 or an earlier insn which
had links or notes added.
--- ---
gcc/combine.cc | 11 +++++++++++ gcc/combine.cc | 7 +++++++
1 file changed, 11 insertions(+) 1 file changed, 7 insertions(+)
diff --git a/gcc/combine.cc b/gcc/combine.cc diff --git a/gcc/combine.cc b/gcc/combine.cc
index a4479f8d836..745391016d0 100644 index a4479f8d836..ff25752cac4 100644
--- a/gcc/combine.cc --- a/gcc/combine.cc
+++ b/gcc/combine.cc +++ b/gcc/combine.cc
@@ -4186,6 +4186,17 @@ try_combine (rtx_insn *i3, rtx_insn *i2, rtx_insn *i1, rtx_insn *i0, @@ -4186,6 +4186,10 @@ try_combine (rtx_insn *i3, rtx_insn *i2, rtx_insn *i1, rtx_insn *i0,
adjust_for_new_dest (i3); adjust_for_new_dest (i3);
} }
+ /* If I2 didn't change, this is not a combination (but a simplification or + bool i2_unchanged = false;
+ canonicalisation with context), which should not be done here. Doing
+ it here explodes the algorithm. Don't. */
+ if (rtx_equal_p (newi2pat, PATTERN (i2))) + if (rtx_equal_p (newi2pat, PATTERN (i2)))
+ { + i2_unchanged = true;
+ if (dump_file)
+ fprintf (dump_file, "i2 didn't change, not doing this\n");
+ undo_all ();
+ return 0;
+ }
+ +
/* We now know that we can do this combination. Merge the insns and /* We now know that we can do this combination. Merge the insns and
update the status of registers and LOG_LINKS. */ update the status of registers and LOG_LINKS. */
@@ -4752,6 +4756,9 @@ try_combine (rtx_insn *i3, rtx_insn *i2, rtx_insn *i1, rtx_insn *i0,
combine_successes++;
undo_commit ();
+ if (i2_unchanged)
+ return i3;
+
rtx_insn *ret = newi2pat ? i2 : i3;
if (added_links_insn && DF_INSN_LUID (added_links_insn) < DF_INSN_LUID (ret))
ret = added_links_insn;
-- --
2.35.3 2.35.3

2
gcc13.changes
View File

@ -1,5 +1,5 @@
------------------------------------------------------------------- -------------------------------------------------------------------
Thu Mar 28 06:55:18 UTC 2024 - Richard Biener <rguenther@suse.com> Thu Apr 4 07:43:17 UTC 2024 - Richard Biener <rguenther@suse.com>
- Add gcc13-pr101523.patch to avoid combine spending too much - Add gcc13-pr101523.patch to avoid combine spending too much
compile-time and memory doing nothing on s390x. [boo#1188441] compile-time and memory doing nothing on s390x. [boo#1188441]