https://gcc.gnu.org/bugzilla/show_bug.cgi?id=117714
Richard Biener <rguenth at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Keywords| |missed-optimization,
| |testsuite-fail
Status|UNCONFIRMED |NEW
Blocks| |116578
Last reconfirmed| |2024-11-20
CC| |rsandifo at gcc dot gnu.org
Ever confirmed|0 |1
--- Comment #2 from Richard Biener <rguenth at gcc dot gnu.org> ---
So on x86 we elide the permute on the load for the SLP reduction but on sparc
this somehow doesn't work according to your dump, SLP optimize partitions
rejects both layouts while with SSE it works out.
The dump gives no hint as to why, so I'm lost here.
On sparc we then fail because we cannot perform the permutation for the
reversal.
sparc seems to miss any kind of useful permute (vec_perm_const), it cannot
even do interleave lo/hi of its V2SImode vectors.
Referenced Bugs:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=116578
[Bug 116578] vectorizer SLP transition issues / dependencies
