1 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py 2 ; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+sse2 | FileCheck %s --check-prefixes=SSE,SSE2 3 ; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+ssse3 | FileCheck %s --check-prefixes=SSE,SSSE3 4 ; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+avx | FileCheck %s --check-prefixes=AVX 5 6 define float @pr26491(<4 x float> %a0) { 7 ; SSE2-LABEL: pr26491: 8 ; SSE2: # %bb.0: 9 ; SSE2-NEXT: movaps %xmm0, %xmm1 10 ; SSE2-NEXT: shufps {{.*#+}} xmm1 = xmm1[1,1],xmm0[3,3] 11 ; SSE2-NEXT: addps %xmm0, %xmm1 12 ; SSE2-NEXT: movaps %xmm1, %xmm0 13 ; SSE2-NEXT: movhlps {{.*#+}} xmm0 = xmm1[1],xmm0[1] 14 ; SSE2-NEXT: addss %xmm1, %xmm0 15 ; SSE2-NEXT: retq 16 ; 17 ; SSSE3-LABEL: pr26491: 18 ; SSSE3: # %bb.0: 19 ; SSSE3-NEXT: movshdup {{.*#+}} xmm1 = xmm0[1,1,3,3] 20 ; SSSE3-NEXT: addps %xmm0, %xmm1 21 ; SSSE3-NEXT: movaps %xmm1, %xmm0 22 ; SSSE3-NEXT: movhlps {{.*#+}} xmm0 = xmm1[1],xmm0[1] 23 ; SSSE3-NEXT: addss %xmm1, %xmm0 24 ; SSSE3-NEXT: retq 25 ; 26 ; AVX-LABEL: pr26491: 27 ; AVX: # %bb.0: 28 ; AVX-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3] 29 ; AVX-NEXT: vaddps %xmm0, %xmm1, %xmm0 30 ; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0] 31 ; AVX-NEXT: vaddss %xmm0, %xmm1, %xmm0 32 ; AVX-NEXT: retq 33 %1 = shufflevector <4 x float> %a0, <4 x float> undef, <4 x i32> <i32 1, i32 1, i32 3, i32 3> 34 %2 = fadd <4 x float> %1, %a0 35 %3 = extractelement <4 x float> %2, i32 2 36 %4 = extractelement <4 x float> %2, i32 0 37 %5 = fadd float %3, %4 38 ret float %5 39 } 40
