RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-10-23 13:02:52 +02:00
Code Issues Projects Releases Wiki Activity

[X86][SSE] Add test showing missed opportunity to combine to pshuflw

Browse Source 
We are combining shuffles to bit shifts before unary permutes, which means we can't fold loads plus the destination register is destructive

llvm-svn: 306976
This commit is contained in:
Simon Pilgrim 2017-07-02 12:56:10 +00:00
parent 1032df3c54
commit 66c4c00993
1 changed files with 18 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

18
test/CodeGen/X86/vector-shuffle-combining-ssse3.ll
View File

@ -445,6 +445,24 @@ define <16 x i8> @combine_pshufb_not_as_pshufw(<16 x i8> %a0) {
ret <16 x i8> %res1
}
; TODO - we could fold the load if we lowered to pshuflw instead.
define <16 x i8> @combine_vpshufb_as_pshuflw_not_pslld(<16 x i8> *%a0) {
; SSE-LABEL: combine_vpshufb_as_pshuflw_not_pslld:
; SSE: # BB#0:
; SSE-NEXT: movdqa (%rdi), %xmm0
; SSE-NEXT: pslld $16, %xmm0
; SSE-NEXT: retq
;
; AVX-LABEL: combine_vpshufb_as_pshuflw_not_pslld:
; AVX: # BB#0:
; AVX-NEXT: vmovdqa (%rdi), %xmm0
; AVX-NEXT: vpslld $16, %xmm0, %xmm0
; AVX-NEXT: retq
%res0 = load <16 x i8>, <16 x i8> *%a0, align 16
%res1 = call <16 x i8> @llvm.x86.ssse3.pshuf.b.128(<16 x i8> %res0, <16 x i8> <i8 undef, i8 undef, i8 0, i8 1, i8 undef, i8 undef, i8 4, i8 5, i8 undef, i8 undef, i8 undef, i8 undef, i8 undef, i8 undef, i8 undef, i8 undef>)
ret <16 x i8> %res1
}
define <16 x i8> @combine_pshufb_as_unary_unpcklbw(<16 x i8> %a0) {
; SSE-LABEL: combine_pshufb_as_unary_unpcklbw:
; SSE: # BB#0: