https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91275
Bill Schmidt changed:
What|Removed |Added
Status|RESOLVED|CLOSED
--- Comment #20 from Bill Schmidt
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91275
Bill Schmidt changed:
What|Removed |Added
Status|NEW |RESOLVED
Known to work|
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91275
--- Comment #18 from Bill Schmidt ---
Author: wschmidt
Date: Mon Oct 7 20:50:05 2019
New Revision: 276678
URL: https://gcc.gnu.org/viewcvs?rev=276678=gcc=rev
Log:
[gcc]
2019-10-07 Bill Schmidt
Backport from mainline
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91275
--- Comment #17 from Bill Schmidt ---
Author: wschmidt
Date: Mon Oct 7 19:34:41 2019
New Revision: 276669
URL: https://gcc.gnu.org/viewcvs?rev=276669=gcc=rev
Log:
[gcc]
2019-10-07 Bill Schmidt
Backport from mainline
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91275
--- Comment #16 from Bill Schmidt ---
Author: wschmidt
Date: Mon Oct 7 18:23:20 2019
New Revision: 276667
URL: https://gcc.gnu.org/viewcvs?rev=276667=gcc=rev
Log:
[gcc]
2019-10-07 Bill Schmidt
Backport from mainline
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91275
--- Comment #15 from Bill Schmidt ---
Excellent! I'm glad you have a workaround for the time being.
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91275
--- Comment #14 from Lauri Kasanen ---
Inline asm works on the buggy versions. Thanks.
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91275
Bill Schmidt changed:
What|Removed |Added
Known to work||10.0
Known to fail|
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91275
--- Comment #12 from Bill Schmidt ---
Committed to trunk with 276410. Forgot to annotate the ChangeLog with the PR,
so here it is:
[gcc]
2019-10-01 Bill Schmidt
* config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Don't swap
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91275
--- Comment #11 from Bill Schmidt ---
I tested #pragma target -mno-optimize-swaps, and it doesn't help. The only
options that can be specified with #pragma target for Power are listed here:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91275
--- Comment #10 from Bill Schmidt ---
I don't *believe* that "#pragma target" works with -mno-optimize-swaps, but you
could try it. I think that mechanism only works for certain flags, but I
haven't tried that one.
I think inline asm should
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91275
--- Comment #9 from Lauri Kasanen ---
Can -mno-optimize-swaps be used per-function, in the code via some pragma?
Alternatively, does calling the instruction via inline asm prevent the
swapping?
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91275
--- Comment #8 from Segher Boessenkool ---
(In reply to Lauri Kasanen from comment #7)
> Are you sure about the smaller ones? To me they should not care about 64-bit
> swaps,
"swappable" here means you can swap the low and high half on all
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91275
--- Comment #7 from Lauri Kasanen ---
Thanks for working on this. Per my experience, this is fast for gcc ;)
Are you sure about the smaller ones? To me they should not care about 64-bit
swaps, but clang lists them all as lane-sensitive:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91275
--- Comment #6 from Bill Schmidt ---
Thanks for the report! Sorry for taking so long to look at this. Swap
optimization treats vpmsumd as swappable, but unlike vpmsumb, vpmsumh, and
vpmsumw, it is not. Testing a patch.
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91275
Bill Schmidt changed:
What|Removed |Added
Assignee|unassigned at gcc dot gnu.org |wschmidt at gcc dot
gnu.org
---
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91275
Segher Boessenkool changed:
What|Removed |Added
CC||wschmidt at gcc dot gnu.org
---
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91275
Segher Boessenkool changed:
What|Removed |Added
Status|UNCONFIRMED |NEW
Last reconfirmed|
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91275
Segher Boessenkool changed:
What|Removed |Added
Target|ppc64le |powerpc*-*-*
CC|
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91275
--- Comment #1 from Lauri Kasanen ---
clang 7.0.0 outputs the expected values, aka the gcc -O0 ones, at all
optimization levels. (it calls the builtin __builtin_altivec_crypto_vpmsumd,
but no other changes)
20 matches
Mail list logo
