52 lines · plain
1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 42; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx | FileCheck %s --check-prefixes=AVX13; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx2 | FileCheck %s --check-prefixes=AVX24 5; PR90847 - failure to peek through FREEZE(SETCC()) results in VPMOVSMSKB(TRUNC()) instead of VMOVMSKPS6 7define i32 @PR90847(<8 x float> %x) nounwind {8; AVX1-LABEL: PR90847:9; AVX1: # %bb.0: # %entry10; AVX1-NEXT: vshufps {{.*#+}} ymm1 = ymm0[1,0,3,2,5,4,7,6]11; AVX1-NEXT: vminps %ymm1, %ymm0, %ymm112; AVX1-NEXT: vshufpd {{.*#+}} ymm2 = ymm1[1,0,3,2]13; AVX1-NEXT: vminps %ymm2, %ymm1, %ymm114; AVX1-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm1[2,3,0,1]15; AVX1-NEXT: vminps %ymm2, %ymm1, %ymm116; AVX1-NEXT: vcmpeqps %ymm0, %ymm1, %ymm017; AVX1-NEXT: vmovmskps %ymm0, %ecx18; AVX1-NEXT: movl $32, %eax19; AVX1-NEXT: rep bsfl %ecx, %eax20; AVX1-NEXT: vzeroupper21; AVX1-NEXT: retq22;23; AVX2-LABEL: PR90847:24; AVX2: # %bb.0: # %entry25; AVX2-NEXT: vshufps {{.*#+}} ymm1 = ymm0[1,0,3,2,5,4,7,6]26; AVX2-NEXT: vminps %ymm1, %ymm0, %ymm127; AVX2-NEXT: vshufpd {{.*#+}} ymm2 = ymm1[1,0,3,2]28; AVX2-NEXT: vminps %ymm2, %ymm1, %ymm129; AVX2-NEXT: vpermpd {{.*#+}} ymm2 = ymm1[2,3,0,1]30; AVX2-NEXT: vminps %ymm2, %ymm1, %ymm131; AVX2-NEXT: vcmpeqps %ymm0, %ymm1, %ymm032; AVX2-NEXT: vmovmskps %ymm0, %ecx33; AVX2-NEXT: movl $32, %eax34; AVX2-NEXT: rep bsfl %ecx, %eax35; AVX2-NEXT: vzeroupper36; AVX2-NEXT: retq37entry:38 %shuf1 = shufflevector <8 x float> %x, <8 x float> poison, <8 x i32> <i32 1, i32 0, i32 3, i32 2, i32 5, i32 4, i32 7, i32 6>39 %min1 = tail call noundef <8 x float> @llvm.x86.avx.min.ps.256(<8 x float> %x, <8 x float> %shuf1)40 %shuf2 = shufflevector <8 x float> %min1, <8 x float> poison, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 6, i32 7, i32 4, i32 5>41 %min2 = tail call noundef <8 x float> @llvm.x86.avx.min.ps.256(<8 x float> %min1, <8 x float> %shuf2)42 %shuf3 = shufflevector <8 x float> %min2, <8 x float> poison, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>43 %min3 = tail call noundef <8 x float> @llvm.x86.avx.min.ps.256(<8 x float> %min2, <8 x float> %shuf3)44 %fcmp = fcmp oeq <8 x float> %min3, %x45 %mask = bitcast <8 x i1> %fcmp to i846 %zext = zext i8 %mask to i3247 %cmp = icmp eq i8 %mask, 048 %tz = tail call range(i32 0, 33) i32 @llvm.cttz.i32(i32 %zext, i1 false)49 %conv = select i1 %cmp, i32 undef, i32 %tz50 ret i32 %conv51}52