https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92140
Andrew Pinski changed:
What|Removed |Added
Target Milestone|--- |10.0
Status|NEW
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92140
--- Comment #32 from Paweł Bylica ---
For what it's worth, the original code is compiled the same as in Clang since
GCC 10. https://godbolt.org/z/vxorYW815
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92140
--- Comment #31 from Segher Boessenkool ---
(In reply to Segher Boessenkool from comment #29)
> Does it help the i386 port if we disallow a hard reg dest as well?
> RA should be able to handle that just fine as well.
I tried this out, and it cre
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92140
--- Comment #30 from Uroš Bizjak ---
(In reply to Segher Boessenkool from comment #29)
>
> Does it help the i386 port if we disallow a hard reg dest as well?
> RA should be able to handle that just fine as well.
I don't know from the top of my h
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92140
--- Comment #29 from Segher Boessenkool ---
(In reply to Uroš Bizjak from comment #27)
> FYI, these constraints were used in the past (when combine was allowed to
> propagate hard registers into combined insn) to prevent reload failures,
> where
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92140
--- Comment #28 from Jakub Jelinek ---
Author: jakub
Date: Sat Oct 19 12:46:57 2019
New Revision: 277203
URL: https://gcc.gnu.org/viewcvs?rev=277203&root=gcc&view=rev
Log:
PR target/92140
* config/i386/predicates.md (int_nonimmed
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92140
--- Comment #27 from Uroš Bizjak ---
(In reply to Jakub Jelinek from comment #26)
> Created attachment 47070 [details]
> gcc10-prereload-splitters.patch
>
> Updated patch for the pre-reload splitters, which uses a new predicate and
> additionall
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92140
Jakub Jelinek changed:
What|Removed |Added
Attachment #47069|0 |1
is obsolete|
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92140
--- Comment #25 from Jakub Jelinek ---
The define_insn part of define_insn_and_split needs constraints if it is meant
to match during or after reload, the patterns are just written with the
assumption that they are split before reload. At least
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92140
--- Comment #24 from Segher Boessenkool ---
A dumb question I'm sure, but I don't see it: if the rest of your
define_insn doesn't need constraints, why would the match_scratch
need some? (A define_split never uses constraints).
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92140
--- Comment #23 from Jakub Jelinek ---
(In reply to Segher Boessenkool from comment #22)
> Hrm, I don't see how this is nicer than just adding a scratch in the
> pattern? What makes that a worse option?
Most of the patterns don't have constrain
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92140
--- Comment #22 from Segher Boessenkool ---
Hrm, I don't see how this is nicer than just adding a scratch in the
pattern? What makes that a worse option?
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92140
--- Comment #21 from Jakub Jelinek ---
Created attachment 47069
--> https://gcc.gnu.org/bugzilla/attachment.cgi?id=47069&action=edit
gcc10-prereload-splitters.patch
Ah, apparently we already have for ~ 2 years a property to handle this safely.
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92140
--- Comment #20 from Segher Boessenkool ---
Ah, okay. So it is either one or two insns (zero can not be handled, but you
can do a noop, a move of a reg to itself, and that will be optimised away just
fine). Three insns is not something combine
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92140
--- Comment #19 from Uroš Bizjak ---
(In reply to Segher Boessenkool from comment #18)
> (In reply to Uroš Bizjak from comment #15)
> > Is it possible to lift the limitation from the combine pass, where the
> > combine tries to split the insn, bu
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92140
--- Comment #18 from Segher Boessenkool ---
(In reply to Uroš Bizjak from comment #15)
> Is it possible to lift the limitation from the combine pass, where the
> combine tries to split the insn, but expects exactly two new insn patterns
> to be g
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92140
--- Comment #17 from Jakub Jelinek ---
I've tried to change the patch to use define_split instead of
define_insn_and_split, with all of them changed, it creates worse code for
f8/f12/f15 (the last one is expected, because we split into 3 instruct
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92140
--- Comment #16 from Uroš Bizjak ---
(In reply to Jakub Jelinek from comment #13)
> Created attachment 47067 [details]
> gcc10-pr92140.patch
>
> So what about this version then? I've changed back a couple of
> to nonimmediate_operand and remov
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92140
--- Comment #15 from Uroš Bizjak ---
(In reply to Segher Boessenkool from comment #12)
> (In reply to Uroš Bizjak from comment #10)
> > Regarding reliability of pre-reload splitters, IIRC they should be safe, but
> > I'll leave the final verdict
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92140
--- Comment #14 from Jakub Jelinek ---
And as for the define_insn_and_split without constraints that don't expect to
be matched post split1, I think we can try to figure out something
incrementally and change all of them at once, e.g. a property
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92140
--- Comment #13 from Jakub Jelinek ---
Created attachment 47067
--> https://gcc.gnu.org/bugzilla/attachment.cgi?id=47067&action=edit
gcc10-pr92140.patch
So what about this version then? I've changed back a couple of
to nonimmediate_operand a
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92140
Segher Boessenkool changed:
What|Removed |Added
CC||segher at gcc dot gnu.org
--- Comme
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92140
--- Comment #11 from Segher Boessenkool ---
If an insn condition uses can_create_pseudo_p, the insn will suddenly stop
to match after reload --> kaboom.
If your insn always splits ("&& 1"), this means that if any of these:
NEXT_PASS (pass_
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92140
--- Comment #10 from Uroš Bizjak ---
(In reply to Jakub Jelinek from comment #9)
> Created attachment 47065 [details]
> gcc10-pr92140-wip.patch
>
> If pre-reload splitters are reliable, my patch can be greatly simplified and
> using the formatti
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92140
--- Comment #9 from Jakub Jelinek ---
Created attachment 47065
--> https://gcc.gnu.org/bugzilla/attachment.cgi?id=47065&action=edit
gcc10-pr92140-wip.patch
If pre-reload splitters are reliable, my patch can be greatly simplified and
using the
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92140
--- Comment #8 from Jakub Jelinek ---
Comparing the two patches, your patch handles f1-f4 in
/* PR target/92140 */
char c;
int v;
__attribute__((noipa)) void f1 (void) { v += c != 0; }
__attribute__((noipa)) void f2 (void) { v -= c != 0; }
__at
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92140
--- Comment #7 from Uroš Bizjak ---
Created attachment 47064
--> https://gcc.gnu.org/bugzilla/attachment.cgi?id=47064&action=edit
Proposed patch with pre-reload splitters
Maybe we should use pre-reload splitters as with the attached patch that
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92140
Jakub Jelinek changed:
What|Removed |Added
Attachment #47062|0 |1
is obsolete|
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92140
--- Comment #5 from Jakub Jelinek ---
The patch adds 144 define_insn and 144 define_split to tmp-mddump.md though, to
total 6217 define_insn and 733 define_split.
Maybe a better way to deal with it would be to have x86_ne_0_operator and
x86_eq_0_
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92140
--- Comment #4 from Jakub Jelinek ---
Created attachment 47062
--> https://gcc.gnu.org/bugzilla/attachment.cgi?id=47062&action=edit
gcc10-pr92140-wip.patch
Slightly extended untested patch, which handles all the cases in the testcase
at the st
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92140
--- Comment #3 from Jakub Jelinek ---
Untested patch that handles tst1 and tst2 and some more, but doesn't handle
tst3 yet and is still missing some patterns.
Unfortunately, it can result in quite a lot of define_insn_and_split patterns,
while fo
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92140
Jakub Jelinek changed:
What|Removed |Added
CC||jakub at gcc dot gnu.org
--- Comment #2
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92140
Uroš Bizjak changed:
What|Removed |Added
Status|UNCONFIRMED |NEW
Last reconfirmed|
33 matches
Mail list logo
