diff options
Diffstat (limited to 'llvm/test/CodeGen/X86/vector-idiv-udiv-256.ll')
| -rw-r--r-- | llvm/test/CodeGen/X86/vector-idiv-udiv-256.ll | 74 |
1 files changed, 36 insertions, 38 deletions
diff --git a/llvm/test/CodeGen/X86/vector-idiv-udiv-256.ll b/llvm/test/CodeGen/X86/vector-idiv-udiv-256.ll index 602f050935d..81d93984e26 100644 --- a/llvm/test/CodeGen/X86/vector-idiv-udiv-256.ll +++ b/llvm/test/CodeGen/X86/vector-idiv-udiv-256.ll @@ -463,59 +463,57 @@ define <16 x i16> @test_rem7_16i16(<16 x i16> %a) nounwind { define <32 x i8> @test_rem7_32i8(<32 x i8> %a) nounwind { ; AVX1-LABEL: test_rem7_32i8: ; AVX1: # %bb.0: -; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm2 -; AVX1-NEXT: vpmovzxbw {{.*#+}} xmm3 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero -; AVX1-NEXT: vmovdqa {{.*#+}} xmm1 = [37,37,37,37,37,37,37,37] -; AVX1-NEXT: vpmullw %xmm1, %xmm3, %xmm3 -; AVX1-NEXT: vpsrlw $8, %xmm3, %xmm3 -; AVX1-NEXT: vpshufd {{.*#+}} xmm4 = xmm2[2,3,0,1] +; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1 +; AVX1-NEXT: vpmovzxbw {{.*#+}} xmm2 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero +; AVX1-NEXT: vmovdqa {{.*#+}} xmm3 = [37,37,37,37,37,37,37,37] +; AVX1-NEXT: vpmullw %xmm3, %xmm2, %xmm2 +; AVX1-NEXT: vpsrlw $8, %xmm2, %xmm2 +; AVX1-NEXT: vpshufd {{.*#+}} xmm4 = xmm1[2,3,0,1] ; AVX1-NEXT: vpmovzxbw {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero,xmm4[4],zero,xmm4[5],zero,xmm4[6],zero,xmm4[7],zero -; AVX1-NEXT: vpmullw %xmm1, %xmm4, %xmm4 +; AVX1-NEXT: vpmullw %xmm3, %xmm4, %xmm4 ; AVX1-NEXT: vpsrlw $8, %xmm4, %xmm4 -; AVX1-NEXT: vpackuswb %xmm4, %xmm3, %xmm3 -; AVX1-NEXT: vpsubb %xmm3, %xmm2, %xmm4 +; AVX1-NEXT: vpackuswb %xmm4, %xmm2, %xmm2 +; AVX1-NEXT: vpsubb %xmm2, %xmm1, %xmm4 ; AVX1-NEXT: vpsrlw $1, %xmm4, %xmm4 ; AVX1-NEXT: vmovdqa {{.*#+}} xmm8 = [127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127] ; AVX1-NEXT: vpand %xmm8, %xmm4, %xmm4 -; AVX1-NEXT: vpaddb %xmm3, %xmm4, %xmm3 -; AVX1-NEXT: vpsrlw $2, %xmm3, %xmm3 +; AVX1-NEXT: vpaddb %xmm2, %xmm4, %xmm2 +; AVX1-NEXT: vpsrlw $2, %xmm2, %xmm2 ; AVX1-NEXT: vmovdqa {{.*#+}} xmm4 = [63,63,63,63,63,63,63,63,63,63,63,63,63,63,63,63] -; AVX1-NEXT: vpand %xmm4, %xmm3, %xmm3 -; AVX1-NEXT: vpmovsxbw %xmm3, %xmm6 -; AVX1-NEXT: vmovdqa {{.*#+}} xmm7 = [7,7,7,7,7,7,7,7] +; AVX1-NEXT: vpand %xmm4, %xmm2, %xmm2 +; AVX1-NEXT: vpunpckhbw {{.*#+}} xmm6 = xmm2[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] +; AVX1-NEXT: vmovdqa {{.*#+}} xmm7 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7] ; AVX1-NEXT: vpmullw %xmm7, %xmm6, %xmm6 ; AVX1-NEXT: vmovdqa {{.*#+}} xmm5 = [255,255,255,255,255,255,255,255] ; AVX1-NEXT: vpand %xmm5, %xmm6, %xmm6 -; AVX1-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[2,3,0,1] -; AVX1-NEXT: vpmovsxbw %xmm3, %xmm3 -; AVX1-NEXT: vpmullw %xmm7, %xmm3, %xmm3 -; AVX1-NEXT: vpand %xmm5, %xmm3, %xmm3 -; AVX1-NEXT: vpackuswb %xmm3, %xmm6, %xmm3 -; AVX1-NEXT: vpsubb %xmm3, %xmm2, %xmm2 -; AVX1-NEXT: vpmovzxbw {{.*#+}} xmm3 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero -; AVX1-NEXT: vpmullw %xmm1, %xmm3, %xmm3 -; AVX1-NEXT: vpsrlw $8, %xmm3, %xmm3 +; AVX1-NEXT: vpmovzxbw {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero +; AVX1-NEXT: vpmullw %xmm7, %xmm2, %xmm2 +; AVX1-NEXT: vpand %xmm5, %xmm2, %xmm2 +; AVX1-NEXT: vpackuswb %xmm6, %xmm2, %xmm2 +; AVX1-NEXT: vpsubb %xmm2, %xmm1, %xmm1 +; AVX1-NEXT: vpmovzxbw {{.*#+}} xmm2 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero +; AVX1-NEXT: vpmullw %xmm3, %xmm2, %xmm2 +; AVX1-NEXT: vpsrlw $8, %xmm2, %xmm2 ; AVX1-NEXT: vpshufd {{.*#+}} xmm6 = xmm0[2,3,0,1] ; AVX1-NEXT: vpmovzxbw {{.*#+}} xmm6 = xmm6[0],zero,xmm6[1],zero,xmm6[2],zero,xmm6[3],zero,xmm6[4],zero,xmm6[5],zero,xmm6[6],zero,xmm6[7],zero -; AVX1-NEXT: vpmullw %xmm1, %xmm6, %xmm1 -; AVX1-NEXT: vpsrlw $8, %xmm1, %xmm1 -; AVX1-NEXT: vpackuswb %xmm1, %xmm3, %xmm1 -; AVX1-NEXT: vpsubb %xmm1, %xmm0, %xmm3 +; AVX1-NEXT: vpmullw %xmm3, %xmm6, %xmm3 +; AVX1-NEXT: vpsrlw $8, %xmm3, %xmm3 +; AVX1-NEXT: vpackuswb %xmm3, %xmm2, %xmm2 +; AVX1-NEXT: vpsubb %xmm2, %xmm0, %xmm3 ; AVX1-NEXT: vpsrlw $1, %xmm3, %xmm3 ; AVX1-NEXT: vpand %xmm8, %xmm3, %xmm3 -; AVX1-NEXT: vpaddb %xmm1, %xmm3, %xmm1 -; AVX1-NEXT: vpsrlw $2, %xmm1, %xmm1 -; AVX1-NEXT: vpand %xmm4, %xmm1, %xmm1 -; AVX1-NEXT: vpmovsxbw %xmm1, %xmm3 +; AVX1-NEXT: vpaddb %xmm2, %xmm3, %xmm2 +; AVX1-NEXT: vpsrlw $2, %xmm2, %xmm2 +; AVX1-NEXT: vpand %xmm4, %xmm2, %xmm2 +; AVX1-NEXT: vpunpckhbw {{.*#+}} xmm3 = xmm2[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] ; AVX1-NEXT: vpmullw %xmm7, %xmm3, %xmm3 ; AVX1-NEXT: vpand %xmm5, %xmm3, %xmm3 -; AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[2,3,0,1] -; AVX1-NEXT: vpmovsxbw %xmm1, %xmm1 -; AVX1-NEXT: vpmullw %xmm7, %xmm1, %xmm1 -; AVX1-NEXT: vpand %xmm5, %xmm1, %xmm1 -; AVX1-NEXT: vpackuswb %xmm1, %xmm3, %xmm1 -; AVX1-NEXT: vpsubb %xmm1, %xmm0, %xmm0 -; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 +; AVX1-NEXT: vpmovzxbw {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero +; AVX1-NEXT: vpmullw %xmm7, %xmm2, %xmm2 +; AVX1-NEXT: vpand %xmm5, %xmm2, %xmm2 +; AVX1-NEXT: vpackuswb %xmm3, %xmm2, %xmm2 +; AVX1-NEXT: vpsubb %xmm2, %xmm0, %xmm0 +; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 ; AVX1-NEXT: retq ; ; AVX2NOBW-LABEL: test_rem7_32i8: |
