diff options
Diffstat (limited to 'llvm/test/CodeGen/X86/horizontal-reduce-smax.ll')
| -rw-r--r-- | llvm/test/CodeGen/X86/horizontal-reduce-smax.ll | 54 |
1 files changed, 22 insertions, 32 deletions
diff --git a/llvm/test/CodeGen/X86/horizontal-reduce-smax.ll b/llvm/test/CodeGen/X86/horizontal-reduce-smax.ll index 161095377cd..ce3210c477c 100644 --- a/llvm/test/CodeGen/X86/horizontal-reduce-smax.ll +++ b/llvm/test/CodeGen/X86/horizontal-reduce-smax.ll @@ -455,13 +455,11 @@ define i64 @test_reduce_v4i64(<4 x i64> %a0) { ; X86-AVX1-LABEL: test_reduce_v4i64: ; X86-AVX1: ## %bb.0: ; X86-AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1 -; X86-AVX1-NEXT: vpcmpgtq %xmm0, %xmm1, %xmm2 -; X86-AVX1-NEXT: vpcmpgtq %xmm1, %xmm0, %xmm3 -; X86-AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm3, %ymm2 -; X86-AVX1-NEXT: vblendvpd %ymm2, %ymm0, %ymm1, %ymm0 +; X86-AVX1-NEXT: vpcmpgtq %xmm1, %xmm0, %xmm2 +; X86-AVX1-NEXT: vblendvpd %xmm2, %xmm0, %xmm1, %xmm0 ; X86-AVX1-NEXT: vpermilps {{.*#+}} xmm1 = xmm0[2,3,0,1] ; X86-AVX1-NEXT: vpcmpgtq %xmm1, %xmm0, %xmm2 -; X86-AVX1-NEXT: vblendvpd %ymm2, %ymm0, %ymm1, %ymm0 +; X86-AVX1-NEXT: vblendvpd %xmm2, %xmm0, %xmm1, %xmm0 ; X86-AVX1-NEXT: vmovd %xmm0, %eax ; X86-AVX1-NEXT: vpextrd $1, %xmm0, %edx ; X86-AVX1-NEXT: vzeroupper @@ -531,13 +529,11 @@ define i64 @test_reduce_v4i64(<4 x i64> %a0) { ; X64-AVX1-LABEL: test_reduce_v4i64: ; X64-AVX1: ## %bb.0: ; X64-AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1 -; X64-AVX1-NEXT: vpcmpgtq %xmm0, %xmm1, %xmm2 -; X64-AVX1-NEXT: vpcmpgtq %xmm1, %xmm0, %xmm3 -; X64-AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm3, %ymm2 -; X64-AVX1-NEXT: vblendvpd %ymm2, %ymm0, %ymm1, %ymm0 +; X64-AVX1-NEXT: vpcmpgtq %xmm1, %xmm0, %xmm2 +; X64-AVX1-NEXT: vblendvpd %xmm2, %xmm0, %xmm1, %xmm0 ; X64-AVX1-NEXT: vpermilps {{.*#+}} xmm1 = xmm0[2,3,0,1] ; X64-AVX1-NEXT: vpcmpgtq %xmm1, %xmm0, %xmm2 -; X64-AVX1-NEXT: vblendvpd %ymm2, %ymm0, %ymm1, %ymm0 +; X64-AVX1-NEXT: vblendvpd %xmm2, %xmm0, %xmm1, %xmm0 ; X64-AVX1-NEXT: vmovq %xmm0, %rax ; X64-AVX1-NEXT: vzeroupper ; X64-AVX1-NEXT: retq @@ -1118,20 +1114,17 @@ define i64 @test_reduce_v8i64(<8 x i64> %a0) { ; ; X86-AVX1-LABEL: test_reduce_v8i64: ; X86-AVX1: ## %bb.0: -; X86-AVX1-NEXT: vextractf128 $1, %ymm1, %xmm2 -; X86-AVX1-NEXT: vextractf128 $1, %ymm0, %xmm3 -; X86-AVX1-NEXT: vpcmpgtq %xmm2, %xmm3, %xmm2 -; X86-AVX1-NEXT: vpcmpgtq %xmm1, %xmm0, %xmm3 -; X86-AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm3, %ymm2 -; X86-AVX1-NEXT: vblendvpd %ymm2, %ymm0, %ymm1, %ymm0 -; X86-AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1 ; X86-AVX1-NEXT: vpcmpgtq %xmm1, %xmm0, %xmm2 -; X86-AVX1-NEXT: vpcmpgtq %xmm0, %xmm1, %xmm3 -; X86-AVX1-NEXT: vinsertf128 $1, %xmm3, %ymm2, %ymm2 -; X86-AVX1-NEXT: vblendvpd %ymm2, %ymm0, %ymm1, %ymm0 +; X86-AVX1-NEXT: vextractf128 $1, %ymm1, %xmm3 +; X86-AVX1-NEXT: vextractf128 $1, %ymm0, %xmm4 +; X86-AVX1-NEXT: vpcmpgtq %xmm3, %xmm4, %xmm5 +; X86-AVX1-NEXT: vblendvpd %xmm5, %xmm4, %xmm3, %xmm3 +; X86-AVX1-NEXT: vblendvpd %xmm2, %xmm0, %xmm1, %xmm0 +; X86-AVX1-NEXT: vpcmpgtq %xmm3, %xmm0, %xmm1 +; X86-AVX1-NEXT: vblendvpd %xmm1, %xmm0, %xmm3, %xmm0 ; X86-AVX1-NEXT: vpermilps {{.*#+}} xmm1 = xmm0[2,3,0,1] ; X86-AVX1-NEXT: vpcmpgtq %xmm1, %xmm0, %xmm2 -; X86-AVX1-NEXT: vblendvpd %ymm2, %ymm0, %ymm1, %ymm0 +; X86-AVX1-NEXT: vblendvpd %xmm2, %xmm0, %xmm1, %xmm0 ; X86-AVX1-NEXT: vmovd %xmm0, %eax ; X86-AVX1-NEXT: vpextrd $1, %xmm0, %edx ; X86-AVX1-NEXT: vzeroupper @@ -1239,20 +1232,17 @@ define i64 @test_reduce_v8i64(<8 x i64> %a0) { ; ; X64-AVX1-LABEL: test_reduce_v8i64: ; X64-AVX1: ## %bb.0: -; X64-AVX1-NEXT: vextractf128 $1, %ymm1, %xmm2 -; X64-AVX1-NEXT: vextractf128 $1, %ymm0, %xmm3 -; X64-AVX1-NEXT: vpcmpgtq %xmm2, %xmm3, %xmm2 -; X64-AVX1-NEXT: vpcmpgtq %xmm1, %xmm0, %xmm3 -; X64-AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm3, %ymm2 -; X64-AVX1-NEXT: vblendvpd %ymm2, %ymm0, %ymm1, %ymm0 -; X64-AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1 ; X64-AVX1-NEXT: vpcmpgtq %xmm1, %xmm0, %xmm2 -; X64-AVX1-NEXT: vpcmpgtq %xmm0, %xmm1, %xmm3 -; X64-AVX1-NEXT: vinsertf128 $1, %xmm3, %ymm2, %ymm2 -; X64-AVX1-NEXT: vblendvpd %ymm2, %ymm0, %ymm1, %ymm0 +; X64-AVX1-NEXT: vextractf128 $1, %ymm1, %xmm3 +; X64-AVX1-NEXT: vextractf128 $1, %ymm0, %xmm4 +; X64-AVX1-NEXT: vpcmpgtq %xmm3, %xmm4, %xmm5 +; X64-AVX1-NEXT: vblendvpd %xmm5, %xmm4, %xmm3, %xmm3 +; X64-AVX1-NEXT: vblendvpd %xmm2, %xmm0, %xmm1, %xmm0 +; X64-AVX1-NEXT: vpcmpgtq %xmm3, %xmm0, %xmm1 +; X64-AVX1-NEXT: vblendvpd %xmm1, %xmm0, %xmm3, %xmm0 ; X64-AVX1-NEXT: vpermilps {{.*#+}} xmm1 = xmm0[2,3,0,1] ; X64-AVX1-NEXT: vpcmpgtq %xmm1, %xmm0, %xmm2 -; X64-AVX1-NEXT: vblendvpd %ymm2, %ymm0, %ymm1, %ymm0 +; X64-AVX1-NEXT: vblendvpd %xmm2, %xmm0, %xmm1, %xmm0 ; X64-AVX1-NEXT: vmovq %xmm0, %rax ; X64-AVX1-NEXT: vzeroupper ; X64-AVX1-NEXT: retq |
