This is the mail archive of the
gcc-bugs@gcc.gnu.org
mailing list for the GCC project.
[Bug target/52607] v4df __builtin_shuffle with {0,2,1,3} or {1,3,0,2}
- From: "jakub at gcc dot gnu.org" <gcc-bugzilla at gcc dot gnu dot org>
- To: gcc-bugs at gcc dot gnu dot org
- Date: Mon, 19 Mar 2012 11:13:52 +0000
- Subject: [Bug target/52607] v4df __builtin_shuffle with {0,2,1,3} or {1,3,0,2}
- Auto-submitted: auto-generated
- References: <bug-52607-4@http.gcc.gnu.org/bugzilla/>
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52607
--- Comment #7 from Jakub Jelinek <jakub at gcc dot gnu.org> 2012-03-19 11:13:52 UTC ---
Created attachment 26915
--> http://gcc.gnu.org/bugzilla/attachment.cgi?id=26915
gcc48-pr52607.patch
AVX2 changes.
This improves:
1) for {x, x, x, x} V4DFmode permutations we now don't generate
vpermpd+vperm2f128, and just emit vpermpd (for all other V4DFmode permutations
we were already emitting just vpermpd)
2) for most of V8SFmode permutations we now emit code like
vmovdqa .LC0(%rip), %ymm1; vpermps %ymm0, %ymm1, %ymm0 instead of
vperm2f128 $0, %ymm0, %ymm0, %ymm0; vpermilps .LC0(%rip), %ymm0, %ymm0
3) broadcast permutations improvements for V8SFmode (using vbroadcastss) as
well
as some integer modes