From f15d40afe384becff8739472c13957abd24bba57 Mon Sep 17 00:00:00 2001 From: Simon Pilgrim Date: Tue, 21 Aug 2018 18:02:28 +0000 Subject: [PATCH] [X86][SSE] Add non-uniform udiv test that is mostly divide by 1. The test demonstrates over-complicated codegen for a udiv that only has one divisor that doesn't equal 1. This should have allowed the codegen to be a lot simpler (uniform shifts etc.) but only the SSE2 manages to make use of this...... llvm-svn: 340313 --- test/CodeGen/X86/combine-udiv.ll | 178 +++++++++++++++++++++++++++++++ 1 file changed, 178 insertions(+) diff --git a/test/CodeGen/X86/combine-udiv.ll b/test/CodeGen/X86/combine-udiv.ll index 5b8824e7595..ed05a7d981a 100644 --- a/test/CodeGen/X86/combine-udiv.ll +++ b/test/CodeGen/X86/combine-udiv.ll @@ -825,6 +825,184 @@ define <8 x i16> @combine_vec_udiv_nonuniform3(<8 x i16> %x) { ret <8 x i16> %1 } +define <16 x i8> @combine_vec_udiv_nonuniform4(<16 x i8> %x) { +; SSE2-LABEL: combine_vec_udiv_nonuniform4: +; SSE2: # %bb.0: +; SSE2-NEXT: pxor %xmm1, %xmm1 +; SSE2-NEXT: movdqa %xmm0, %xmm2 +; SSE2-NEXT: punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3],xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7] +; SSE2-NEXT: movl $255, %eax +; SSE2-NEXT: movd %eax, %xmm1 +; SSE2-NEXT: movl $171, %eax +; SSE2-NEXT: movd %eax, %xmm3 +; SSE2-NEXT: pand %xmm1, %xmm3 +; SSE2-NEXT: pmullw %xmm2, %xmm3 +; SSE2-NEXT: psrlw $8, %xmm3 +; SSE2-NEXT: pmullw {{.*}}(%rip), %xmm3 +; SSE2-NEXT: psrlw $8, %xmm3 +; SSE2-NEXT: pand %xmm1, %xmm3 +; SSE2-NEXT: pand {{.*}}(%rip), %xmm0 +; SSE2-NEXT: por %xmm3, %xmm0 +; SSE2-NEXT: retq +; +; SSE41-LABEL: combine_vec_udiv_nonuniform4: +; SSE41: # %bb.0: +; SSE41-NEXT: movdqa %xmm0, %xmm8 +; SSE41-NEXT: movdqa {{.*#+}} xmm2 = [8,8,8,8,8,8,8,8] +; SSE41-NEXT: movdqa %xmm2, %xmm0 +; SSE41-NEXT: psllw $12, %xmm0 +; SSE41-NEXT: psllw $4, %xmm2 +; SSE41-NEXT: por %xmm0, %xmm2 +; SSE41-NEXT: movl $171, %eax +; SSE41-NEXT: movd %eax, %xmm7 +; SSE41-NEXT: pmovzxbw {{.*#+}} xmm6 = xmm8[0],zero,xmm8[1],zero,xmm8[2],zero,xmm8[3],zero,xmm8[4],zero,xmm8[5],zero,xmm8[6],zero,xmm8[7],zero +; SSE41-NEXT: pmullw %xmm7, %xmm6 +; SSE41-NEXT: movdqa %xmm6, %xmm3 +; SSE41-NEXT: psrlw $8, %xmm3 +; SSE41-NEXT: movdqa %xmm2, %xmm0 +; SSE41-NEXT: pblendvb %xmm0, %xmm3, %xmm6 +; SSE41-NEXT: movdqa %xmm6, %xmm4 +; SSE41-NEXT: psrlw $4, %xmm4 +; SSE41-NEXT: movdqa %xmm2, %xmm3 +; SSE41-NEXT: paddw %xmm2, %xmm3 +; SSE41-NEXT: movdqa %xmm3, %xmm0 +; SSE41-NEXT: pblendvb %xmm0, %xmm4, %xmm6 +; SSE41-NEXT: movdqa %xmm6, %xmm5 +; SSE41-NEXT: psrlw $2, %xmm5 +; SSE41-NEXT: movdqa %xmm3, %xmm4 +; SSE41-NEXT: paddw %xmm3, %xmm4 +; SSE41-NEXT: movdqa %xmm4, %xmm0 +; SSE41-NEXT: pblendvb %xmm0, %xmm5, %xmm6 +; SSE41-NEXT: movdqa %xmm6, %xmm1 +; SSE41-NEXT: psrlw $1, %xmm1 +; SSE41-NEXT: movdqa %xmm4, %xmm5 +; SSE41-NEXT: paddw %xmm4, %xmm5 +; SSE41-NEXT: movdqa %xmm5, %xmm0 +; SSE41-NEXT: pblendvb %xmm0, %xmm1, %xmm6 +; SSE41-NEXT: pshufd {{.*#+}} xmm0 = xmm8[2,3,0,1] +; SSE41-NEXT: pmovzxbw {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero +; SSE41-NEXT: pshufd {{.*#+}} xmm1 = xmm7[2,3,0,1] +; SSE41-NEXT: pmovzxbw {{.*#+}} xmm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero +; SSE41-NEXT: pmullw %xmm0, %xmm1 +; SSE41-NEXT: movdqa %xmm1, %xmm7 +; SSE41-NEXT: psrlw $8, %xmm7 +; SSE41-NEXT: movdqa %xmm2, %xmm0 +; SSE41-NEXT: pblendvb %xmm0, %xmm7, %xmm1 +; SSE41-NEXT: movdqa %xmm1, %xmm2 +; SSE41-NEXT: psrlw $4, %xmm2 +; SSE41-NEXT: movdqa %xmm3, %xmm0 +; SSE41-NEXT: pblendvb %xmm0, %xmm2, %xmm1 +; SSE41-NEXT: movdqa %xmm1, %xmm2 +; SSE41-NEXT: psrlw $2, %xmm2 +; SSE41-NEXT: movdqa %xmm4, %xmm0 +; SSE41-NEXT: pblendvb %xmm0, %xmm2, %xmm1 +; SSE41-NEXT: movdqa %xmm1, %xmm2 +; SSE41-NEXT: psrlw $1, %xmm2 +; SSE41-NEXT: movdqa %xmm5, %xmm0 +; SSE41-NEXT: pblendvb %xmm0, %xmm2, %xmm1 +; SSE41-NEXT: packuswb %xmm1, %xmm6 +; SSE41-NEXT: pxor %xmm0, %xmm0 +; SSE41-NEXT: pxor %xmm1, %xmm1 +; SSE41-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm6[0],xmm1[1],xmm6[1],xmm1[2],xmm6[2],xmm1[3],xmm6[3],xmm1[4],xmm6[4],xmm1[5],xmm6[5],xmm1[6],xmm6[6],xmm1[7],xmm6[7] +; SSE41-NEXT: pmovzxbw {{.*#+}} xmm2 = xmm6[0],zero,xmm6[1],zero,xmm6[2],zero,xmm6[3],zero,xmm6[4],zero,xmm6[5],zero,xmm6[6],zero,xmm6[7],zero +; SSE41-NEXT: psllw $1, %xmm2 +; SSE41-NEXT: pblendw {{.*#+}} xmm2 = xmm2[0],xmm1[1,2,3,4,5,6,7] +; SSE41-NEXT: psrlw $8, %xmm2 +; SSE41-NEXT: punpckhbw {{.*#+}} xmm6 = xmm6[8],xmm0[8],xmm6[9],xmm0[9],xmm6[10],xmm0[10],xmm6[11],xmm0[11],xmm6[12],xmm0[12],xmm6[13],xmm0[13],xmm6[14],xmm0[14],xmm6[15],xmm0[15] +; SSE41-NEXT: packuswb %xmm6, %xmm2 +; SSE41-NEXT: movaps {{.*#+}} xmm0 = [0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; SSE41-NEXT: pblendvb %xmm0, %xmm8, %xmm2 +; SSE41-NEXT: movdqa %xmm2, %xmm0 +; SSE41-NEXT: retq +; +; AVX1-LABEL: combine_vec_udiv_nonuniform4: +; AVX1: # %bb.0: +; AVX1-NEXT: vmovdqa {{.*#+}} xmm1 = [8,8,8,8,8,8,8,8] +; AVX1-NEXT: vpsllw $12, %xmm1, %xmm2 +; AVX1-NEXT: vpsllw $4, %xmm1, %xmm1 +; AVX1-NEXT: vpor %xmm2, %xmm1, %xmm1 +; AVX1-NEXT: movl $171, %eax +; AVX1-NEXT: vmovd %eax, %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 %xmm2, %xmm3, %xmm3 +; AVX1-NEXT: vpsrlw $8, %xmm3, %xmm4 +; AVX1-NEXT: vpblendvb %xmm1, %xmm4, %xmm3, %xmm3 +; AVX1-NEXT: vpsrlw $4, %xmm3, %xmm4 +; AVX1-NEXT: vpaddw %xmm1, %xmm1, %xmm5 +; AVX1-NEXT: vpblendvb %xmm5, %xmm4, %xmm3, %xmm3 +; AVX1-NEXT: vpsrlw $2, %xmm3, %xmm4 +; AVX1-NEXT: vpaddw %xmm5, %xmm5, %xmm6 +; AVX1-NEXT: vpblendvb %xmm6, %xmm4, %xmm3, %xmm3 +; AVX1-NEXT: vpsrlw $1, %xmm3, %xmm4 +; AVX1-NEXT: vpaddw %xmm6, %xmm6, %xmm7 +; AVX1-NEXT: vpblendvb %xmm7, %xmm4, %xmm3, %xmm3 +; AVX1-NEXT: vpshufd {{.*#+}} xmm4 = xmm0[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: vpshufd {{.*#+}} xmm2 = xmm2[2,3,0,1] +; 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 %xmm2, %xmm4, %xmm2 +; AVX1-NEXT: vpsrlw $8, %xmm2, %xmm4 +; AVX1-NEXT: vpblendvb %xmm1, %xmm4, %xmm2, %xmm1 +; AVX1-NEXT: vpsrlw $4, %xmm1, %xmm2 +; AVX1-NEXT: vpblendvb %xmm5, %xmm2, %xmm1, %xmm1 +; AVX1-NEXT: vpsrlw $2, %xmm1, %xmm2 +; AVX1-NEXT: vpblendvb %xmm6, %xmm2, %xmm1, %xmm1 +; AVX1-NEXT: vpsrlw $1, %xmm1, %xmm2 +; AVX1-NEXT: vpblendvb %xmm7, %xmm2, %xmm1, %xmm1 +; AVX1-NEXT: vpackuswb %xmm1, %xmm3, %xmm1 +; AVX1-NEXT: vpxor %xmm2, %xmm2, %xmm2 +; AVX1-NEXT: vpunpcklbw {{.*#+}} xmm3 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3],xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7] +; AVX1-NEXT: vpmovzxbw {{.*#+}} xmm4 = 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: vpsllw $1, %xmm4, %xmm4 +; AVX1-NEXT: vpblendw {{.*#+}} xmm3 = xmm4[0],xmm3[1,2,3,4,5,6,7] +; AVX1-NEXT: vpsrlw $8, %xmm3, %xmm3 +; AVX1-NEXT: vpunpckhbw {{.*#+}} xmm1 = xmm1[8],xmm2[8],xmm1[9],xmm2[9],xmm1[10],xmm2[10],xmm1[11],xmm2[11],xmm1[12],xmm2[12],xmm1[13],xmm2[13],xmm1[14],xmm2[14],xmm1[15],xmm2[15] +; AVX1-NEXT: vpackuswb %xmm1, %xmm3, %xmm1 +; AVX1-NEXT: vmovdqa {{.*#+}} xmm2 = [0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX1-NEXT: vpblendvb %xmm2, %xmm0, %xmm1, %xmm0 +; AVX1-NEXT: retq +; +; AVX2-LABEL: combine_vec_udiv_nonuniform4: +; AVX2: # %bb.0: +; AVX2-NEXT: vpmovzxbw {{.*#+}} ymm1 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero +; AVX2-NEXT: movl $171, %eax +; AVX2-NEXT: vmovd %eax, %xmm2 +; AVX2-NEXT: vpmullw %ymm2, %ymm1, %ymm1 +; AVX2-NEXT: vpsrlw $8, %ymm1, %ymm1 +; AVX2-NEXT: vextracti128 $1, %ymm1, %xmm2 +; AVX2-NEXT: vpackuswb %xmm2, %xmm1, %xmm1 +; AVX2-NEXT: vpmovzxbw {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero,xmm1[8],zero,xmm1[9],zero,xmm1[10],zero,xmm1[11],zero,xmm1[12],zero,xmm1[13],zero,xmm1[14],zero,xmm1[15],zero +; AVX2-NEXT: vpmullw {{.*}}(%rip), %ymm1, %ymm1 +; AVX2-NEXT: vpsrlw $8, %ymm1, %ymm1 +; AVX2-NEXT: vextracti128 $1, %ymm1, %xmm2 +; AVX2-NEXT: vpackuswb %xmm2, %xmm1, %xmm1 +; AVX2-NEXT: vmovdqa {{.*#+}} xmm2 = [0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX2-NEXT: vpblendvb %xmm2, %xmm0, %xmm1, %xmm0 +; AVX2-NEXT: vzeroupper +; AVX2-NEXT: retq +; +; XOP-LABEL: combine_vec_udiv_nonuniform4: +; XOP: # %bb.0: +; XOP-NEXT: movl $171, %eax +; XOP-NEXT: vmovd %eax, %xmm1 +; XOP-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 +; XOP-NEXT: vpmullw %xmm1, %xmm2, %xmm2 +; XOP-NEXT: vpshufd {{.*#+}} xmm3 = xmm0[2,3,0,1] +; XOP-NEXT: vpmovzxbw {{.*#+}} xmm3 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero,xmm3[4],zero,xmm3[5],zero,xmm3[6],zero,xmm3[7],zero +; XOP-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[2,3,0,1] +; XOP-NEXT: vpmovzxbw {{.*#+}} xmm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero +; XOP-NEXT: vpmullw %xmm1, %xmm3, %xmm1 +; XOP-NEXT: vpperm {{.*#+}} xmm1 = xmm2[1,3,5,7,9,11,13,15],xmm1[1,3,5,7,9,11,13,15] +; XOP-NEXT: movl $249, %eax +; XOP-NEXT: vmovd %eax, %xmm2 +; XOP-NEXT: vpshlb %xmm2, %xmm1, %xmm1 +; XOP-NEXT: vmovdqa {{.*#+}} xmm2 = [0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; XOP-NEXT: vpblendvb %xmm2, %xmm0, %xmm1, %xmm0 +; XOP-NEXT: retq + %div = udiv <16 x i8> %x, + ret <16 x i8> %div +} + define <8 x i16> @pr38477(<8 x i16> %a0) { ; SSE2-LABEL: pr38477: ; SSE2: # %bb.0: