https://gcc.gnu.org/bugzilla/show_bug.cgi?id=78762
John Keiser changed:
What|Removed |Added
CC||john at johnkeiser dot com
--- Comment
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=78762
Peter Cordes changed:
What|Removed |Added
CC||peter at cordes dot ca
--- Comment #16
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=78762
--- Comment #15 from Venkataramanan ---
Considering this PR, removing the tuning (splitting of unaligned avx256 loads)
for generic is suggested.
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=78762
--- Comment #14 from Venkataramanan ---
(In reply to Allan Jensen from comment #13)
> The question is if the unaligned store is still slow on Excavator and Ryzen
> which support AVX2. As far as I understand the bulldozer architectures just
>
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=78762
--- Comment #13 from Allan Jensen ---
The question is if the unaligned store is still slow on Excavator and Ryzen
which support AVX2. As far as I understand the bulldozer architectures just
prefer split AVX because it was basically emulating
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=78762
--- Comment #12 from Venkataramanan ---
(In reply to Allan Jensen from comment #11)
> Btw, did you benchmark store splitting on AMD? It is also enabled for BDVER
> and ZNVER1.
I have not done that.
As per SWOG for AMD15h (BDVER) it is
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=78762
--- Comment #11 from Allan Jensen ---
Btw, did you benchmark store splitting on AMD? It is also enabled for BDVER and
ZNVER1.
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=78762
--- Comment #10 from Allan Jensen ---
That would solve the problem, but also leave the behavior as Sandybridge only
(nehalem didn't have AVX).
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=78762
--- Comment #9 from Venkataramanan ---
(In reply to Jakub Jelinek from comment #6)
> Sure, the question is (raised several times over the last couple of years)
> is if the generic tuning should not adjust slightly based on the selected
> ISAs.
>
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=78762
Venkataramanan changed:
What|Removed |Added
CC||venkataramanan.kumar at amd
dot co
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=78762
Jakub Jelinek changed:
What|Removed |Added
CC||hjl.tools at gmail dot com
--- Comment
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=78762
Jakub Jelinek changed:
What|Removed |Added
CC||jakub at gcc dot gnu.org,
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=78762
--- Comment #5 from Marc Glisse ---
(In reply to Martin Liška from comment #4)
> I'm not much familiar with the vector instructions, however the change
> started with r239889.
Arguably r239889 fixed a bug where the tuning was ignored. Now this
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=78762
Martin Liška changed:
What|Removed |Added
Status|UNCONFIRMED |NEW
Last reconfirmed|
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=78762
--- Comment #3 from Allan Jensen ---
Created attachment 40298
--> https://gcc.gnu.org/bugzilla/attachment.cgi?id=40298=edit
Test compiled with gcc 6
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=78762
--- Comment #2 from Allan Jensen ---
Created attachment 40297
--> https://gcc.gnu.org/bugzilla/attachment.cgi?id=40297=edit
Test compiled with -march=haswell
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=78762
--- Comment #1 from Allan Jensen ---
Created attachment 40296
--> https://gcc.gnu.org/bugzilla/attachment.cgi?id=40296=edit
Test compiled with -mavx2
17 matches
Mail list logo
