public inbox for gcc-bugs@sourceware.org
help / color / mirror / Atom feed
* [Bug target/102552] New: i386 failing tests after r12-4038-g6de9f0c13b27c34336587da19d03200f8cc6bcd5
@ 2021-10-01 11:21 marxin at gcc dot gnu.org
2021-10-01 11:22 ` [Bug target/102552] " pinskia at gcc dot gnu.org
` (4 more replies)
0 siblings, 5 replies; 6+ messages in thread
From: marxin at gcc dot gnu.org @ 2021-10-01 11:21 UTC (permalink / raw)
To: gcc-bugs
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102552
Bug ID: 102552
Summary: i386 failing tests after
r12-4038-g6de9f0c13b27c34336587da19d03200f8cc6bcd5
Product: gcc
Version: 12.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: target
Assignee: unassigned at gcc dot gnu.org
Reporter: marxin at gcc dot gnu.org
Target Milestone: ---
The following 2 tests do fail after my revision:
FAIL: gcc.target/i386/avx512fp16-reduce-op-2.c scan-tree-dump-times optimized
".REDUC_MAX" 3
FAIL: gcc.target/i386/avx512fp16-reduce-op-2.c scan-tree-dump-times optimized
".REDUC_MIN" 3
FAIL: gcc.target/i386/avx512fp16-reduce-op-2.c scan-tree-dump-times optimized
".REDUC_PLUS" 3
FAIL: gcc.target/i386/intrinsics_4.c (test for excess errors)
Apparently, it does not happen on all x86_64 systems, strange:
$ cat reduc.c
_Float16
__attribute__((noipa, target("avx512fp16,avx512vl"), optimize("Ofast")))
reduc_add_128 (_Float16* p)
{
_Float16 sum = 0;
for (int i = 0; i != 8; i++)
sum += p[i];
return sum;
}
$ ./xgcc -B. reduc.c -Ofast -fdump-tree-optimized=/dev/stdout
;; Function reduc_add_128 (reduc_add_128, funcdef_no=0, decl_uid=1978,
cgraph_uid=1, symbol_order=0)
__attribute__((optimize ("Ofast"), target ("avx512fp16,avx512vl"), noipa,
noinline, noclone, no_icf))
_Float16 reduc_add_128 (_Float16 * p)
{
vector(8) _Float16 vect__4.6;
_Float16 _24;
<bb 2> [local count: 119292720]:
vect__4.6_5 = MEM <vector(8) _Float16> [(_Float16 *)p_9(D)];
_24 = .REDUC_PLUS (vect__4.6_5); [tail call]
return _24;
}
while on a different box it does:
$ ./xgcc -B. reduc.c -Ofast -fdump-tree-optimized=/dev/stdout
;; Function reduc_add_128 (reduc_add_128, funcdef_no=0, decl_uid=1978,
cgraph_uid=1, symbol_order=0)
__attribute__((optimize ("Ofast"), target ("avx512fp16,avx512vl"), noipa,
noinline, noclone, no_icf))
_Float16 reduc_add_128 (_Float16 * p)
{
_Float16 sum;
_Float16 _4;
_Float16 _5;
_Float16 _6;
_Float16 _7;
_Float16 _12;
_Float16 _13;
_Float16 _22;
_Float16 _23;
_Float16 _30;
_Float16 _40;
_Float16 _50;
_Float16 _60;
_Float16 _70;
_Float16 _80;
<bb 2> [local count: 119292720]:
_6 = *p_9(D);
_30 = MEM[(_Float16 *)p_9(D) + 2B];
_40 = MEM[(_Float16 *)p_9(D) + 4B];
_13 = _30 + _40;
_50 = MEM[(_Float16 *)p_9(D) + 6B];
_60 = MEM[(_Float16 *)p_9(D) + 8B];
_12 = _50 + _60;
_22 = _12 + _13;
_70 = MEM[(_Float16 *)p_9(D) + 10B];
_80 = MEM[(_Float16 *)p_9(D) + 12B];
_7 = _70 + _80;
_4 = MEM[(_Float16 *)p_9(D) + 14B];
_5 = _4 + _6;
_23 = _5 + _7;
sum_10 = _22 + _23;
return sum_10;
}
^ permalink raw reply [flat|nested] 6+ messages in thread
* [Bug target/102552] i386 failing tests after r12-4038-g6de9f0c13b27c34336587da19d03200f8cc6bcd5
2021-10-01 11:21 [Bug target/102552] New: i386 failing tests after r12-4038-g6de9f0c13b27c34336587da19d03200f8cc6bcd5 marxin at gcc dot gnu.org
@ 2021-10-01 11:22 ` pinskia at gcc dot gnu.org
2021-10-01 13:23 ` marxin at gcc dot gnu.org
` (3 subsequent siblings)
4 siblings, 0 replies; 6+ messages in thread
From: pinskia at gcc dot gnu.org @ 2021-10-01 11:22 UTC (permalink / raw)
To: gcc-bugs
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102552
--- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Maybe binutils version difference?
^ permalink raw reply [flat|nested] 6+ messages in thread
* [Bug target/102552] i386 failing tests after r12-4038-g6de9f0c13b27c34336587da19d03200f8cc6bcd5
2021-10-01 11:21 [Bug target/102552] New: i386 failing tests after r12-4038-g6de9f0c13b27c34336587da19d03200f8cc6bcd5 marxin at gcc dot gnu.org
2021-10-01 11:22 ` [Bug target/102552] " pinskia at gcc dot gnu.org
@ 2021-10-01 13:23 ` marxin at gcc dot gnu.org
2021-10-01 13:27 ` marxin at gcc dot gnu.org
` (2 subsequent siblings)
4 siblings, 0 replies; 6+ messages in thread
From: marxin at gcc dot gnu.org @ 2021-10-01 13:23 UTC (permalink / raw)
To: gcc-bugs
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102552
--- Comment #2 from Martin Liška <marxin at gcc dot gnu.org> ---
No, it's related to fact that:
bool
parse_optimize_options (tree args, bool attr_p)
{
...
/* Build up argv vector. Just in case the string is stored away, use garbage
collected strings. */
vec_safe_truncate (optimize_args, 0);
vec_safe_push (optimize_args, (const char *) NULL);
As seen we push NULL pointer as the first vector element. Why do we do that?
^ permalink raw reply [flat|nested] 6+ messages in thread
* [Bug target/102552] i386 failing tests after r12-4038-g6de9f0c13b27c34336587da19d03200f8cc6bcd5
2021-10-01 11:21 [Bug target/102552] New: i386 failing tests after r12-4038-g6de9f0c13b27c34336587da19d03200f8cc6bcd5 marxin at gcc dot gnu.org
2021-10-01 11:22 ` [Bug target/102552] " pinskia at gcc dot gnu.org
2021-10-01 13:23 ` marxin at gcc dot gnu.org
@ 2021-10-01 13:27 ` marxin at gcc dot gnu.org
2021-10-01 13:57 ` cvs-commit at gcc dot gnu.org
2021-10-01 13:57 ` marxin at gcc dot gnu.org
4 siblings, 0 replies; 6+ messages in thread
From: marxin at gcc dot gnu.org @ 2021-10-01 13:27 UTC (permalink / raw)
To: gcc-bugs
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102552
--- Comment #3 from Martin Liška <marxin at gcc dot gnu.org> ---
Oh, it's because it calls eventually decode_cmdline_options_to_array that skips
the first argument (program name).
^ permalink raw reply [flat|nested] 6+ messages in thread
* [Bug target/102552] i386 failing tests after r12-4038-g6de9f0c13b27c34336587da19d03200f8cc6bcd5
2021-10-01 11:21 [Bug target/102552] New: i386 failing tests after r12-4038-g6de9f0c13b27c34336587da19d03200f8cc6bcd5 marxin at gcc dot gnu.org
` (2 preceding siblings ...)
2021-10-01 13:27 ` marxin at gcc dot gnu.org
@ 2021-10-01 13:57 ` cvs-commit at gcc dot gnu.org
2021-10-01 13:57 ` marxin at gcc dot gnu.org
4 siblings, 0 replies; 6+ messages in thread
From: cvs-commit at gcc dot gnu.org @ 2021-10-01 13:57 UTC (permalink / raw)
To: gcc-bugs
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102552
--- Comment #4 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Martin Liska <marxin@gcc.gnu.org>:
https://gcc.gnu.org/g:5051fad8582fcbdd0844232b5a8c4e856be5e5a4
commit r12-4045-g5051fad8582fcbdd0844232b5a8c4e856be5e5a4
Author: Martin Liska <mliska@suse.cz>
Date: Fri Oct 1 15:37:59 2021 +0200
options: fix concat of options.
PR target/102552
gcc/c-family/ChangeLog:
* c-common.c (parse_optimize_options): decoded_options[0] is
used for program name, so merged_decoded_options should also
respect that.
^ permalink raw reply [flat|nested] 6+ messages in thread
* [Bug target/102552] i386 failing tests after r12-4038-g6de9f0c13b27c34336587da19d03200f8cc6bcd5
2021-10-01 11:21 [Bug target/102552] New: i386 failing tests after r12-4038-g6de9f0c13b27c34336587da19d03200f8cc6bcd5 marxin at gcc dot gnu.org
` (3 preceding siblings ...)
2021-10-01 13:57 ` cvs-commit at gcc dot gnu.org
@ 2021-10-01 13:57 ` marxin at gcc dot gnu.org
4 siblings, 0 replies; 6+ messages in thread
From: marxin at gcc dot gnu.org @ 2021-10-01 13:57 UTC (permalink / raw)
To: gcc-bugs
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102552
Martin Liška <marxin at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|UNCONFIRMED |RESOLVED
Resolution|--- |FIXED
--- Comment #5 from Martin Liška <marxin at gcc dot gnu.org> ---
Fixed now.
^ permalink raw reply [flat|nested] 6+ messages in thread
end of thread, other threads:[~2021-10-01 13:57 UTC | newest]
Thread overview: 6+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2021-10-01 11:21 [Bug target/102552] New: i386 failing tests after r12-4038-g6de9f0c13b27c34336587da19d03200f8cc6bcd5 marxin at gcc dot gnu.org
2021-10-01 11:22 ` [Bug target/102552] " pinskia at gcc dot gnu.org
2021-10-01 13:23 ` marxin at gcc dot gnu.org
2021-10-01 13:27 ` marxin at gcc dot gnu.org
2021-10-01 13:57 ` cvs-commit at gcc dot gnu.org
2021-10-01 13:57 ` marxin at gcc dot gnu.org
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).