RPCS3/llvm
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm.git synced 2026-07-19 15:13:49 -04:00
Code Issues Packages Projects Releases 10 Wiki Activity

[X86][AVX] Add broadcast(v4f64 hadd) test

Browse Source 
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@363252 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
Simon Pilgrim
2019-06-13 11:42:32 +00:00
parent 0263b7d55f
commit 02dbbc259c
1 changed files with 54 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files
test/CodeGen/X86/haddsub-shuf.ll
+54
View File
@@ -263,6 +263,60 @@ define <4 x double> @hadd_v4f64_scalar_splat(<4 x double> %a) {
ret <4 x double> %shuf
}
define <4 x double> @hadd_v4f64_scalar_broadcast(<4 x double> %a) {
; SSSE3_SLOW-LABEL: hadd_v4f64_scalar_broadcast:
; SSSE3_SLOW: # %bb.0:
; SSSE3_SLOW-NEXT: movapd %xmm0, %xmm1
; SSSE3_SLOW-NEXT: unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSSE3_SLOW-NEXT: addsd %xmm0, %xmm1
; SSSE3_SLOW-NEXT: movddup {{.*#+}} xmm0 = xmm1[0,0]
; SSSE3_SLOW-NEXT: movapd %xmm0, %xmm1
; SSSE3_SLOW-NEXT: retq
;
; SSSE3_FAST-LABEL: hadd_v4f64_scalar_broadcast:
; SSSE3_FAST: # %bb.0:
; SSSE3_FAST-NEXT: haddpd %xmm0, %xmm0
; SSSE3_FAST-NEXT: movapd %xmm0, %xmm1
; SSSE3_FAST-NEXT: retq
;
; AVX1_SLOW-LABEL: hadd_v4f64_scalar_broadcast:
; AVX1_SLOW: # %bb.0:
; AVX1_SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX1_SLOW-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX1_SLOW-NEXT: vmovddup {{.*#+}} xmm0 = xmm0[0,0]
; AVX1_SLOW-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
; AVX1_SLOW-NEXT: retq
;
; AVX1_FAST-LABEL: hadd_v4f64_scalar_broadcast:
; AVX1_FAST: # %bb.0:
; AVX1_FAST-NEXT: vhaddpd %xmm0, %xmm0, %xmm0
; AVX1_FAST-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
; AVX1_FAST-NEXT: retq
;
; AVX2_SLOW-LABEL: hadd_v4f64_scalar_broadcast:
; AVX2_SLOW: # %bb.0:
; AVX2_SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX2_SLOW-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX2_SLOW-NEXT: vbroadcastsd %xmm0, %ymm0
; AVX2_SLOW-NEXT: retq
;
; AVX2_FAST-LABEL: hadd_v4f64_scalar_broadcast:
; AVX2_FAST: # %bb.0:
; AVX2_FAST-NEXT: vhaddpd %xmm0, %xmm0, %xmm0
; AVX2_FAST-NEXT: vbroadcastsd %xmm0, %ymm0
; AVX2_FAST-NEXT: retq
%a0 = extractelement <4 x double> %a, i32 0
%a1 = extractelement <4 x double> %a, i32 1
%hop0 = fadd double %a0, %a1
%a2 = extractelement <4 x double> %a, i32 2
%a3 = extractelement <4 x double> %a, i32 3
%hop1 = fadd double %a2, %a3
%ins = insertelement <4 x double> undef, double %hop0, i32 0
%ins2 = insertelement <4 x double> %ins, double %hop1, i32 2
%shuf = shufflevector <4 x double> %ins2, <4 x double> undef, <4 x i32> <i32 0, i32 0, i32 0, i32 0>
ret <4 x double> %shuf
}
define <4 x double> @hadd_v4f64(<4 x double> %a) {
; SSSE3_SLOW-LABEL: hadd_v4f64:
; SSSE3_SLOW: # %bb.0: