https://llvm.org/bugs/show_bug.cgi?id=31351
Bug ID: 31351
Summary: Lower shuffles using AVX-512 EXPAND* instructions
Product: libraries
Version: trunk
Hardware: PC
OS: Windows NT
Status: NEW
Severity: normal
Priority: P
Component: Backend: X86
Assignee: unassignedb...@nondot.org
Reporter: zvi.racko...@intel.com
CC: llvm-bugs@lists.llvm.org
Classification: Unclassified
define <8 x float> @expand(<4 x float> %a) {
%res = shufflevector <4 x float> %a, <4 x float> zeroinitializer, <8 x i32>
<i32 0, i32 5, i32 1, i32 5, i32 2, i32 5, i32 3, i32 5>
ret <8 x float> %res
}
We currently generate:
vmovaps .LCPI0_0(%rip), %ymm1 # ymm1 = <0,u,1,u,2,u,3,u>
vpermps %ymm0, %ymm1, %ymm0
vxorps %ymm1, %ymm1, %ymm1
vblendps $170, %ymm1, %ymm0, %ymm0 # ymm0 =
ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
This would be better:
mov $mask, %eax
mov $eax, %k1
expandps %ymm0, %ymm0 {%k1}{z}:
--
You are receiving this mail because:
You are on the CC list for the bug.
_______________________________________________
llvm-bugs mailing list
llvm-bugs@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-bugs
