OSchip
/
llvm-project
mirror of https://github.com/THU-DSP-LAB/llvm-project.git
13
0
Fork 6
Code Issues Packages Projects Releases Wiki Activity

[AVX512] Use alignedstore256 in a pattern that's emitting a 256-bit movaps from an extract subvector operation. 

llvm-svn: 311263
This commit is contained in:
Craig Topper 2017-08-19 22:02:02 +00:00
parent ee3292c5e4
commit a0319bb434
2 changed files with 3 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
llvm/lib/Target/X86/X86InstrAVX512.td
View File

@ -3721,8 +3721,8 @@ let Predicates = [HasVLX] in {
def : Pat<(alignedstore256 (v4f64 (extract_subvector def : Pat<(alignedstore256 (v4f64 (extract_subvector
(v8f64 VR512:$src), (iPTR 0))), addr:$dst), (v8f64 VR512:$src), (iPTR 0))), addr:$dst),
(VMOVAPDZ256mr addr:$dst, (v4f64 (EXTRACT_SUBREG VR512:$src,sub_ymm)))>; (VMOVAPDZ256mr addr:$dst, (v4f64 (EXTRACT_SUBREG VR512:$src,sub_ymm)))>;
def : Pat<(alignedstore (v8f32 (extract_subvector def : Pat<(alignedstore256 (v8f32 (extract_subvector
(v16f32 VR512:$src), (iPTR 0))), addr:$dst), (v16f32 VR512:$src), (iPTR 0))), addr:$dst),
(VMOVAPSZ256mr addr:$dst, (v8f32 (EXTRACT_SUBREG VR512:$src,sub_ymm)))>; (VMOVAPSZ256mr addr:$dst, (v8f32 (EXTRACT_SUBREG VR512:$src,sub_ymm)))>;
def : Pat<(alignedstore256 (v4i64 (extract_subvector def : Pat<(alignedstore256 (v4i64 (extract_subvector
(v8i64 VR512:$src), (iPTR 0))), addr:$dst), (v8i64 VR512:$src), (iPTR 0))), addr:$dst),

2
llvm/test/CodeGen/X86/avx512-extract-subvector.ll
View File

@ -493,7 +493,7 @@ entry:
define void @extract_subvector512_v8f32_store_lo_align_16(float* nocapture %addr, <16 x float> %a) nounwind uwtable ssp { define void @extract_subvector512_v8f32_store_lo_align_16(float* nocapture %addr, <16 x float> %a) nounwind uwtable ssp {
; SKX-LABEL: extract_subvector512_v8f32_store_lo_align_16: ; SKX-LABEL: extract_subvector512_v8f32_store_lo_align_16:
; SKX: ## BB#0: ## %entry ; SKX: ## BB#0: ## %entry
; SKX-NEXT: vmovaps %ymm0, (%rdi) ; SKX-NEXT: vmovups %ymm0, (%rdi)
; SKX-NEXT: vzeroupper ; SKX-NEXT: vzeroupper
; SKX-NEXT: retq ; SKX-NEXT: retq
entry: entry: