public inbox for gcc-bugs@sourceware.org help / color / mirror / Atom feed
* [Bug tree-optimization/98443] New: Failure to optimize out vector operations into a constant when possible
@ 2020-12-25 6:08 gabravier at gmail dot com
2021-01-02 8:55 ` [Bug tree-optimization/98443] " jakub at gcc dot gnu.org
` (2 more replies)
0 siblings, 3 replies; 4+ messages in thread
From: gabravier at gmail dot com @ 2020-12-25 6:08 UTC (permalink / raw)
To: gcc-bugs
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98443
Bug ID: 98443
Summary: Failure to optimize out vector operations into a
constant when possible
Product: gcc
Version: 11.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: tree-optimization
Assignee: unassigned at gcc dot gnu.org
Reporter: gabravier at gmail dot com
Target Milestone: ---
typedef char A __attribute__((vector_size(32)));
int foo(A w)
{
w *= (A){0, 1};
return w[0];
}
This can be optimized into `return 0;`. This transformation is done by LLVM,
but not by GCC.
^ permalink raw reply [flat|nested] 4+ messages in thread* [Bug tree-optimization/98443] Failure to optimize out vector operations into a constant when possible 2020-12-25 6:08 [Bug tree-optimization/98443] New: Failure to optimize out vector operations into a constant when possible gabravier at gmail dot com @ 2021-01-02 8:55 ` jakub at gcc dot gnu.org 2021-01-05 10:15 ` rguenth at gcc dot gnu.org 2021-08-29 5:17 ` pinskia at gcc dot gnu.org 2 siblings, 0 replies; 4+ messages in thread From: jakub at gcc dot gnu.org @ 2021-01-02 8:55 UTC (permalink / raw) To: gcc-bugs https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98443 Jakub Jelinek <jakub at gcc dot gnu.org> changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |jakub at gcc dot gnu.org --- Comment #1 from Jakub Jelinek <jakub at gcc dot gnu.org> --- We'd probably need VRP on the vector elements for this. ^ permalink raw reply [flat|nested] 4+ messages in thread
* [Bug tree-optimization/98443] Failure to optimize out vector operations into a constant when possible 2020-12-25 6:08 [Bug tree-optimization/98443] New: Failure to optimize out vector operations into a constant when possible gabravier at gmail dot com 2021-01-02 8:55 ` [Bug tree-optimization/98443] " jakub at gcc dot gnu.org @ 2021-01-05 10:15 ` rguenth at gcc dot gnu.org 2021-08-29 5:17 ` pinskia at gcc dot gnu.org 2 siblings, 0 replies; 4+ messages in thread From: rguenth at gcc dot gnu.org @ 2021-01-05 10:15 UTC (permalink / raw) To: gcc-bugs https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98443 Richard Biener <rguenth at gcc dot gnu.org> changed: What |Removed |Added ---------------------------------------------------------------------------- Ever confirmed|0 |1 Status|UNCONFIRMED |NEW Last reconfirmed| |2021-01-05 --- Comment #2 from Richard Biener <rguenth at gcc dot gnu.org> --- More like value numbering vector components and/or known-zero/one-bits. Possibly the CCP bit handling can be reasonably easy extended to handle this particular case (integer vectors). ^ permalink raw reply [flat|nested] 4+ messages in thread
* [Bug tree-optimization/98443] Failure to optimize out vector operations into a constant when possible 2020-12-25 6:08 [Bug tree-optimization/98443] New: Failure to optimize out vector operations into a constant when possible gabravier at gmail dot com 2021-01-02 8:55 ` [Bug tree-optimization/98443] " jakub at gcc dot gnu.org 2021-01-05 10:15 ` rguenth at gcc dot gnu.org @ 2021-08-29 5:17 ` pinskia at gcc dot gnu.org 2 siblings, 0 replies; 4+ messages in thread From: pinskia at gcc dot gnu.org @ 2021-08-29 5:17 UTC (permalink / raw) To: gcc-bugs https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98443 --- Comment #3 from Andrew Pinski <pinskia at gcc dot gnu.org> --- Or we could push the BIT_FIELD_REF down the chain: _1 = w_3(D) & { 0, -1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }; _2 = BIT_FIELD_REF <_1, 8, 0>; Which is exactly what LLVM does. Take: typedef char A __attribute__((vector_size(32))); int foo(A w) { w *= (A){2, 3,4,5}; return w[0]; } LLVM gets: vmovd %xmm0, %eax addb %al, %al movsbl %al, %eax vzeroupper retq ^ permalink raw reply [flat|nested] 4+ messages in thread
end of thread, other threads:[~2021-08-29 5:17 UTC | newest] Thread overview: 4+ messages (download: mbox.gz / follow: Atom feed) -- links below jump to the message on this page -- 2020-12-25 6:08 [Bug tree-optimization/98443] New: Failure to optimize out vector operations into a constant when possible gabravier at gmail dot com 2021-01-02 8:55 ` [Bug tree-optimization/98443] " jakub at gcc dot gnu.org 2021-01-05 10:15 ` rguenth at gcc dot gnu.org 2021-08-29 5:17 ` pinskia 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).