diff options
Diffstat (limited to 'llvm/test/CodeGen/X86/vector-reduce-umin.ll')
| -rw-r--r-- | llvm/test/CodeGen/X86/vector-reduce-umin.ll | 524 |
1 files changed, 142 insertions, 382 deletions
diff --git a/llvm/test/CodeGen/X86/vector-reduce-umin.ll b/llvm/test/CodeGen/X86/vector-reduce-umin.ll index 184d5452863..7225c270aef 100644 --- a/llvm/test/CodeGen/X86/vector-reduce-umin.ll +++ b/llvm/test/CodeGen/X86/vector-reduce-umin.ll @@ -725,87 +725,38 @@ define i64 @test_v16i64(<16 x i64> %a0) { define i32 @test_v2i32(<2 x i32> %a0) { ; SSE2-LABEL: test_v2i32: ; SSE2: # %bb.0: -; SSE2-NEXT: movdqa {{.*#+}} xmm1 = [4294967295,0,4294967295,0] -; SSE2-NEXT: pand %xmm0, %xmm1 -; SSE2-NEXT: pxor %xmm2, %xmm2 +; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm0[1,1,2,3] +; SSE2-NEXT: movdqa {{.*#+}} xmm2 = [2147483648,2147483648,2147483648,2147483648] ; SSE2-NEXT: movdqa %xmm0, %xmm3 -; SSE2-NEXT: punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm2[2],xmm3[3],xmm2[3] -; SSE2-NEXT: movdqa {{.*#+}} xmm2 = [2147483648,2147483648] -; SSE2-NEXT: pxor %xmm2, %xmm1 -; SSE2-NEXT: pxor %xmm3, %xmm2 -; SSE2-NEXT: movdqa %xmm2, %xmm4 -; SSE2-NEXT: pcmpgtd %xmm1, %xmm4 -; SSE2-NEXT: pshufd {{.*#+}} xmm5 = xmm4[0,0,2,2] -; SSE2-NEXT: pcmpeqd %xmm1, %xmm2 -; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm2[1,1,3,3] -; SSE2-NEXT: pand %xmm5, %xmm1 -; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm4[1,1,3,3] -; SSE2-NEXT: por %xmm1, %xmm2 +; SSE2-NEXT: pxor %xmm2, %xmm3 +; SSE2-NEXT: pxor %xmm1, %xmm2 +; SSE2-NEXT: pcmpgtd %xmm3, %xmm2 ; SSE2-NEXT: pand %xmm2, %xmm0 -; SSE2-NEXT: pandn %xmm3, %xmm2 +; SSE2-NEXT: pandn %xmm1, %xmm2 ; SSE2-NEXT: por %xmm0, %xmm2 ; SSE2-NEXT: movd %xmm2, %eax ; SSE2-NEXT: retq ; ; SSE41-LABEL: test_v2i32: ; SSE41: # %bb.0: -; SSE41-NEXT: pxor %xmm1, %xmm1 -; SSE41-NEXT: pblendw {{.*#+}} xmm1 = xmm0[0,1],xmm1[2,3],xmm0[4,5],xmm1[6,7] -; SSE41-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,3,0,1] -; SSE41-NEXT: pmovzxdq {{.*#+}} xmm2 = xmm0[0],zero,xmm0[1],zero -; SSE41-NEXT: movdqa {{.*#+}} xmm3 = [2147483648,2147483648] -; SSE41-NEXT: movdqa %xmm1, %xmm0 -; SSE41-NEXT: pxor %xmm3, %xmm0 -; SSE41-NEXT: pxor %xmm2, %xmm3 -; SSE41-NEXT: movdqa %xmm3, %xmm4 -; SSE41-NEXT: pcmpeqd %xmm0, %xmm4 -; SSE41-NEXT: pcmpgtd %xmm0, %xmm3 -; SSE41-NEXT: pshufd {{.*#+}} xmm0 = xmm3[0,0,2,2] -; SSE41-NEXT: pand %xmm4, %xmm0 -; SSE41-NEXT: por %xmm3, %xmm0 -; SSE41-NEXT: blendvpd %xmm0, %xmm1, %xmm2 -; SSE41-NEXT: movd %xmm2, %eax +; SSE41-NEXT: pshufd {{.*#+}} xmm1 = xmm0[1,1,2,3] +; SSE41-NEXT: pminud %xmm0, %xmm1 +; SSE41-NEXT: movd %xmm1, %eax ; SSE41-NEXT: retq ; -; AVX1-LABEL: test_v2i32: -; AVX1: # %bb.0: -; AVX1-NEXT: vpxor %xmm1, %xmm1, %xmm1 -; AVX1-NEXT: vpblendw {{.*#+}} xmm2 = xmm0[0,1],xmm1[2,3],xmm0[4,5],xmm1[6,7] -; AVX1-NEXT: vpunpckhdq {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3] -; AVX1-NEXT: vpcmpgtq %xmm2, %xmm0, %xmm1 -; AVX1-NEXT: vblendvpd %xmm1, %xmm2, %xmm0, %xmm0 -; AVX1-NEXT: vmovd %xmm0, %eax -; AVX1-NEXT: retq -; -; AVX2-LABEL: test_v2i32: -; AVX2: # %bb.0: -; AVX2-NEXT: vpxor %xmm1, %xmm1, %xmm1 -; AVX2-NEXT: vpblendd {{.*#+}} xmm2 = xmm0[0],xmm1[1],xmm0[2],xmm1[3] -; AVX2-NEXT: vpunpckhdq {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3] -; AVX2-NEXT: vpcmpgtq %xmm2, %xmm0, %xmm1 -; AVX2-NEXT: vblendvpd %xmm1, %xmm2, %xmm0, %xmm0 -; AVX2-NEXT: vmovd %xmm0, %eax -; AVX2-NEXT: retq -; -; AVX512BW-LABEL: test_v2i32: -; AVX512BW: # %bb.0: -; AVX512BW-NEXT: vpxor %xmm1, %xmm1, %xmm1 -; AVX512BW-NEXT: vpblendd {{.*#+}} xmm2 = xmm0[0],xmm1[1],xmm0[2],xmm1[3] -; AVX512BW-NEXT: vpunpckhdq {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3] -; AVX512BW-NEXT: vpminuq %zmm0, %zmm2, %zmm0 -; AVX512BW-NEXT: vmovd %xmm0, %eax -; AVX512BW-NEXT: vzeroupper -; AVX512BW-NEXT: retq +; AVX-LABEL: test_v2i32: +; AVX: # %bb.0: +; AVX-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3] +; AVX-NEXT: vpminud %xmm1, %xmm0, %xmm0 +; AVX-NEXT: vmovd %xmm0, %eax +; AVX-NEXT: retq ; -; AVX512VL-LABEL: test_v2i32: -; AVX512VL: # %bb.0: -; AVX512VL-NEXT: vpxor %xmm1, %xmm1, %xmm1 -; AVX512VL-NEXT: vpblendd {{.*#+}} xmm1 = xmm0[0],xmm1[1],xmm0[2],xmm1[3] -; AVX512VL-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[2,3,0,1] -; AVX512VL-NEXT: vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero -; AVX512VL-NEXT: vpminuq %xmm0, %xmm1, %xmm0 -; AVX512VL-NEXT: vmovd %xmm0, %eax -; AVX512VL-NEXT: retq +; AVX512-LABEL: test_v2i32: +; AVX512: # %bb.0: +; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3] +; AVX512-NEXT: vpminud %xmm1, %xmm0, %xmm0 +; AVX512-NEXT: vmovd %xmm0, %eax +; AVX512-NEXT: retq %1 = call i32 @llvm.experimental.vector.reduce.umin.v2i32(<2 x i32> %a0) ret i32 %1 } @@ -1207,85 +1158,41 @@ define i32 @test_v32i32(<32 x i32> %a0) { define i16 @test_v2i16(<2 x i16> %a0) { ; SSE2-LABEL: test_v2i16: ; SSE2: # %bb.0: -; SSE2-NEXT: movdqa {{.*#+}} xmm1 = [65535,0,0,0,65535,0,0,0] -; SSE2-NEXT: pand %xmm0, %xmm1 -; SSE2-NEXT: pxor %xmm2, %xmm2 -; SSE2-NEXT: movdqa %xmm0, %xmm3 -; SSE2-NEXT: punpckhwd {{.*#+}} xmm3 = xmm3[4],xmm2[4],xmm3[5],xmm2[5],xmm3[6],xmm2[6],xmm3[7],xmm2[7] -; SSE2-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1] -; SSE2-NEXT: movdqa {{.*#+}} xmm2 = [2147483648,2147483648] -; SSE2-NEXT: por %xmm2, %xmm1 -; SSE2-NEXT: por %xmm3, %xmm2 -; SSE2-NEXT: movdqa %xmm2, %xmm4 -; SSE2-NEXT: pcmpgtd %xmm1, %xmm4 -; SSE2-NEXT: pshufd {{.*#+}} xmm5 = xmm4[0,0,2,2] -; SSE2-NEXT: pcmpeqd %xmm1, %xmm2 -; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm2[1,1,3,3] -; SSE2-NEXT: pand %xmm5, %xmm1 -; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm4[1,1,3,3] -; SSE2-NEXT: por %xmm1, %xmm2 -; SSE2-NEXT: pand %xmm2, %xmm0 -; SSE2-NEXT: pandn %xmm3, %xmm2 -; SSE2-NEXT: por %xmm0, %xmm2 -; SSE2-NEXT: movd %xmm2, %eax +; SSE2-NEXT: movdqa %xmm0, %xmm1 +; SSE2-NEXT: psrld $16, %xmm1 +; SSE2-NEXT: movdqa {{.*#+}} xmm2 = [32768,32768,32768,32768,32768,32768,32768,32768] +; SSE2-NEXT: pxor %xmm2, %xmm0 +; SSE2-NEXT: pxor %xmm2, %xmm1 +; SSE2-NEXT: pminsw %xmm0, %xmm1 +; SSE2-NEXT: movd %xmm1, %eax +; SSE2-NEXT: xorl $32768, %eax # imm = 0x8000 ; SSE2-NEXT: # kill: def $ax killed $ax killed $eax ; SSE2-NEXT: retq ; ; SSE41-LABEL: test_v2i16: ; SSE41: # %bb.0: -; SSE41-NEXT: pxor %xmm1, %xmm1 -; SSE41-NEXT: pblendw {{.*#+}} xmm1 = xmm0[0],xmm1[1,2,3],xmm0[4],xmm1[5,6,7] -; SSE41-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,3,0,1] -; SSE41-NEXT: pmovzxwq {{.*#+}} xmm2 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero -; SSE41-NEXT: movdqa {{.*#+}} xmm3 = [2147483648,2147483648] -; SSE41-NEXT: movdqa %xmm1, %xmm0 -; SSE41-NEXT: por %xmm3, %xmm0 -; SSE41-NEXT: por %xmm2, %xmm3 -; SSE41-NEXT: movdqa %xmm3, %xmm4 -; SSE41-NEXT: pcmpeqd %xmm0, %xmm4 -; SSE41-NEXT: pcmpgtd %xmm0, %xmm3 -; SSE41-NEXT: pshufd {{.*#+}} xmm0 = xmm3[0,0,2,2] -; SSE41-NEXT: pand %xmm4, %xmm0 -; SSE41-NEXT: por %xmm3, %xmm0 -; SSE41-NEXT: blendvpd %xmm0, %xmm1, %xmm2 -; SSE41-NEXT: movd %xmm2, %eax +; SSE41-NEXT: movdqa %xmm0, %xmm1 +; SSE41-NEXT: psrld $16, %xmm1 +; SSE41-NEXT: pminuw %xmm0, %xmm1 +; SSE41-NEXT: movd %xmm1, %eax ; SSE41-NEXT: # kill: def $ax killed $ax killed $eax ; SSE41-NEXT: retq ; ; AVX-LABEL: test_v2i16: ; AVX: # %bb.0: -; AVX-NEXT: vpxor %xmm1, %xmm1, %xmm1 -; AVX-NEXT: vpblendw {{.*#+}} xmm1 = xmm0[0],xmm1[1,2,3],xmm0[4],xmm1[5,6,7] -; AVX-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[2,3,0,1] -; AVX-NEXT: vpmovzxwq {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero -; AVX-NEXT: vpcmpgtq %xmm1, %xmm0, %xmm2 -; AVX-NEXT: vblendvpd %xmm2, %xmm1, %xmm0, %xmm0 +; AVX-NEXT: vpsrld $16, %xmm0, %xmm1 +; AVX-NEXT: vpminuw %xmm1, %xmm0, %xmm0 ; AVX-NEXT: vmovd %xmm0, %eax ; AVX-NEXT: # kill: def $ax killed $ax killed $eax ; AVX-NEXT: retq ; -; AVX512BW-LABEL: test_v2i16: -; AVX512BW: # %bb.0: -; AVX512BW-NEXT: vpxor %xmm1, %xmm1, %xmm1 -; AVX512BW-NEXT: vpblendw {{.*#+}} xmm1 = xmm0[0],xmm1[1,2,3],xmm0[4],xmm1[5,6,7] -; AVX512BW-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[2,3,0,1] -; AVX512BW-NEXT: vpmovzxwq {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero -; AVX512BW-NEXT: vpminuq %zmm0, %zmm1, %zmm0 -; AVX512BW-NEXT: vmovd %xmm0, %eax -; AVX512BW-NEXT: # kill: def $ax killed $ax killed $eax -; AVX512BW-NEXT: vzeroupper -; AVX512BW-NEXT: retq -; -; AVX512VL-LABEL: test_v2i16: -; AVX512VL: # %bb.0: -; AVX512VL-NEXT: vpxor %xmm1, %xmm1, %xmm1 -; AVX512VL-NEXT: vpblendw {{.*#+}} xmm1 = xmm0[0],xmm1[1,2,3],xmm0[4],xmm1[5,6,7] -; AVX512VL-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[2,3,0,1] -; AVX512VL-NEXT: vpmovzxwq {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero -; AVX512VL-NEXT: vpminuq %xmm0, %xmm1, %xmm0 -; AVX512VL-NEXT: vmovd %xmm0, %eax -; AVX512VL-NEXT: # kill: def $ax killed $ax killed $eax -; AVX512VL-NEXT: retq +; AVX512-LABEL: test_v2i16: +; AVX512: # %bb.0: +; AVX512-NEXT: vpsrld $16, %xmm0, %xmm1 +; AVX512-NEXT: vpminuw %xmm1, %xmm0, %xmm0 +; AVX512-NEXT: vmovd %xmm0, %eax +; AVX512-NEXT: # kill: def $ax killed $ax killed $eax +; AVX512-NEXT: retq %1 = call i16 @llvm.experimental.vector.reduce.umin.v2i16(<2 x i16> %a0) ret i16 %1 } @@ -1293,70 +1200,48 @@ define i16 @test_v2i16(<2 x i16> %a0) { define i16 @test_v4i16(<4 x i16> %a0) { ; SSE2-LABEL: test_v4i16: ; SSE2: # %bb.0: -; SSE2-NEXT: movdqa {{.*#+}} xmm1 = [65535,0,65535,0,65535,0,65535,0] -; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm0[2,3,0,1] -; SSE2-NEXT: pand %xmm1, %xmm0 -; SSE2-NEXT: pand {{.*}}(%rip), %xmm2 -; SSE2-NEXT: movdqa %xmm2, %xmm3 -; SSE2-NEXT: pcmpgtd %xmm0, %xmm3 -; SSE2-NEXT: pand %xmm3, %xmm0 -; SSE2-NEXT: pandn %xmm2, %xmm3 -; SSE2-NEXT: por %xmm0, %xmm3 -; SSE2-NEXT: pand %xmm3, %xmm1 -; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm3[1,1,2,3] -; SSE2-NEXT: pxor %xmm2, %xmm2 -; SSE2-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3] -; SSE2-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1] -; SSE2-NEXT: movdqa %xmm0, %xmm2 -; SSE2-NEXT: pcmpgtd %xmm1, %xmm2 -; SSE2-NEXT: pand %xmm2, %xmm3 -; SSE2-NEXT: pandn %xmm0, %xmm2 -; SSE2-NEXT: por %xmm3, %xmm2 -; SSE2-NEXT: movd %xmm2, %eax +; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm0[1,1,2,3] +; SSE2-NEXT: movdqa {{.*#+}} xmm2 = [32768,32768,32768,32768,32768,32768,32768,32768] +; SSE2-NEXT: pxor %xmm2, %xmm0 +; SSE2-NEXT: pxor %xmm2, %xmm1 +; SSE2-NEXT: pminsw %xmm0, %xmm1 +; SSE2-NEXT: movdqa %xmm1, %xmm0 +; SSE2-NEXT: pxor %xmm2, %xmm0 +; SSE2-NEXT: psrld $16, %xmm0 +; SSE2-NEXT: pxor %xmm2, %xmm0 +; SSE2-NEXT: pminsw %xmm1, %xmm0 +; SSE2-NEXT: movd %xmm0, %eax +; SSE2-NEXT: xorl $32768, %eax # imm = 0x8000 ; SSE2-NEXT: # kill: def $ax killed $ax killed $eax ; SSE2-NEXT: retq ; ; SSE41-LABEL: test_v4i16: ; SSE41: # %bb.0: -; SSE41-NEXT: pxor %xmm1, %xmm1 -; SSE41-NEXT: pshufd {{.*#+}} xmm2 = xmm0[2,3,0,1] -; SSE41-NEXT: pblendw {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2],xmm1[3],xmm0[4],xmm1[5],xmm0[6],xmm1[7] -; SSE41-NEXT: pblendw {{.*#+}} xmm2 = xmm2[0],xmm1[1],xmm2[2],xmm1[3,4,5,6,7] -; SSE41-NEXT: pminud %xmm0, %xmm2 -; SSE41-NEXT: pblendw {{.*#+}} xmm1 = xmm2[0],xmm1[1],xmm2[2],xmm1[3],xmm2[4],xmm1[5],xmm2[6],xmm1[7] -; SSE41-NEXT: pshufd {{.*#+}} xmm0 = xmm2[1,1,2,3] -; SSE41-NEXT: pmovzxwq {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero -; SSE41-NEXT: pminud %xmm1, %xmm0 +; SSE41-NEXT: pshufd {{.*#+}} xmm1 = xmm0[1,1,2,3] +; SSE41-NEXT: pminuw %xmm0, %xmm1 +; SSE41-NEXT: movdqa %xmm1, %xmm0 +; SSE41-NEXT: psrld $16, %xmm0 +; SSE41-NEXT: pminuw %xmm1, %xmm0 ; SSE41-NEXT: movd %xmm0, %eax ; SSE41-NEXT: # kill: def $ax killed $ax killed $eax ; SSE41-NEXT: retq ; ; AVX-LABEL: test_v4i16: ; AVX: # %bb.0: -; AVX-NEXT: vpxor %xmm1, %xmm1, %xmm1 -; AVX-NEXT: vpblendw {{.*#+}} xmm2 = xmm0[0],xmm1[1],xmm0[2],xmm1[3],xmm0[4],xmm1[5],xmm0[6],xmm1[7] -; AVX-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[2,3,0,1] -; AVX-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2],xmm1[3,4,5,6,7] -; AVX-NEXT: vpminud %xmm0, %xmm2, %xmm0 -; AVX-NEXT: vpblendw {{.*#+}} xmm1 = xmm0[0],xmm1[1],xmm0[2],xmm1[3],xmm0[4],xmm1[5],xmm0[6],xmm1[7] -; AVX-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[1,1,2,3] -; AVX-NEXT: vpmovzxwq {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero -; AVX-NEXT: vpminud %xmm0, %xmm1, %xmm0 +; AVX-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3] +; AVX-NEXT: vpminuw %xmm1, %xmm0, %xmm0 +; AVX-NEXT: vpsrld $16, %xmm0, %xmm1 +; AVX-NEXT: vpminuw %xmm1, %xmm0, %xmm0 ; AVX-NEXT: vmovd %xmm0, %eax ; AVX-NEXT: # kill: def $ax killed $ax killed $eax ; AVX-NEXT: retq ; ; AVX512-LABEL: test_v4i16: ; AVX512: # %bb.0: -; AVX512-NEXT: vpxor %xmm1, %xmm1, %xmm1 -; AVX512-NEXT: vpblendw {{.*#+}} xmm2 = xmm0[0],xmm1[1],xmm0[2],xmm1[3],xmm0[4],xmm1[5],xmm0[6],xmm1[7] -; AVX512-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[2,3,0,1] -; AVX512-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2],xmm1[3,4,5,6,7] -; AVX512-NEXT: vpminud %xmm0, %xmm2, %xmm0 -; AVX512-NEXT: vpblendw {{.*#+}} xmm1 = xmm0[0],xmm1[1],xmm0[2],xmm1[3],xmm0[4],xmm1[5],xmm0[6],xmm1[7] -; AVX512-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[1,1,2,3] -; AVX512-NEXT: vpmovzxwq {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero -; AVX512-NEXT: vpminud %xmm0, %xmm1, %xmm0 +; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3] +; AVX512-NEXT: vpminuw %xmm1, %xmm0, %xmm0 +; AVX512-NEXT: vpsrld $16, %xmm0, %xmm1 +; AVX512-NEXT: vpminuw %xmm1, %xmm0, %xmm0 ; AVX512-NEXT: vmovd %xmm0, %eax ; AVX512-NEXT: # kill: def $ax killed $ax killed $eax ; AVX512-NEXT: retq @@ -1667,83 +1552,37 @@ define i16 @test_v64i16(<64 x i16> %a0) { define i8 @test_v2i8(<2 x i8> %a0) { ; SSE2-LABEL: test_v2i8: ; SSE2: # %bb.0: -; SSE2-NEXT: movdqa {{.*#+}} xmm1 = [255,0,0,0,0,0,0,0,255,0,0,0,0,0,0,0] -; SSE2-NEXT: pand %xmm0, %xmm1 -; SSE2-NEXT: pxor %xmm2, %xmm2 -; SSE2-NEXT: movdqa %xmm0, %xmm3 -; SSE2-NEXT: punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm2[8],xmm3[9],xmm2[9],xmm3[10],xmm2[10],xmm3[11],xmm2[11],xmm3[12],xmm2[12],xmm3[13],xmm2[13],xmm3[14],xmm2[14],xmm3[15],xmm2[15] -; SSE2-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3] -; SSE2-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1] -; SSE2-NEXT: movdqa {{.*#+}} xmm2 = [2147483648,2147483648] -; SSE2-NEXT: por %xmm2, %xmm1 -; SSE2-NEXT: por %xmm3, %xmm2 -; SSE2-NEXT: movdqa %xmm2, %xmm4 -; SSE2-NEXT: pcmpgtd %xmm1, %xmm4 -; SSE2-NEXT: pshufd {{.*#+}} xmm5 = xmm4[0,0,2,2] -; SSE2-NEXT: pcmpeqd %xmm1, %xmm2 -; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm2[1,1,3,3] -; SSE2-NEXT: pand %xmm5, %xmm1 -; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm4[1,1,3,3] -; SSE2-NEXT: por %xmm1, %xmm2 -; SSE2-NEXT: pand %xmm2, %xmm0 -; SSE2-NEXT: pandn %xmm3, %xmm2 -; SSE2-NEXT: por %xmm0, %xmm2 -; SSE2-NEXT: movd %xmm2, %eax +; SSE2-NEXT: movdqa %xmm0, %xmm1 +; SSE2-NEXT: psrlw $8, %xmm1 +; SSE2-NEXT: pminub %xmm0, %xmm1 +; SSE2-NEXT: movd %xmm1, %eax ; SSE2-NEXT: # kill: def $al killed $al killed $eax ; SSE2-NEXT: retq ; ; SSE41-LABEL: test_v2i8: ; SSE41: # %bb.0: ; SSE41-NEXT: movdqa %xmm0, %xmm1 -; SSE41-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,3,0,1] -; SSE41-NEXT: pand {{.*}}(%rip), %xmm1 -; SSE41-NEXT: pmovzxbq {{.*#+}} xmm2 = xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero -; SSE41-NEXT: movdqa {{.*#+}} xmm3 = [2147483648,2147483648] -; SSE41-NEXT: movdqa %xmm1, %xmm0 -; SSE41-NEXT: por %xmm3, %xmm0 -; SSE41-NEXT: por %xmm2, %xmm3 -; SSE41-NEXT: movdqa %xmm3, %xmm4 -; SSE41-NEXT: pcmpeqd %xmm0, %xmm4 -; SSE41-NEXT: pcmpgtd %xmm0, %xmm3 -; SSE41-NEXT: pshufd {{.*#+}} xmm0 = xmm3[0,0,2,2] -; SSE41-NEXT: pand %xmm4, %xmm0 -; SSE41-NEXT: por %xmm3, %xmm0 -; SSE41-NEXT: blendvpd %xmm0, %xmm1, %xmm2 -; SSE41-NEXT: pextrb $0, %xmm2, %eax +; SSE41-NEXT: psrlw $8, %xmm1 +; SSE41-NEXT: pminub %xmm0, %xmm1 +; SSE41-NEXT: pextrb $0, %xmm1, %eax ; SSE41-NEXT: # kill: def $al killed $al killed $eax ; SSE41-NEXT: retq ; ; AVX-LABEL: test_v2i8: ; AVX: # %bb.0: -; AVX-NEXT: vpand {{.*}}(%rip), %xmm0, %xmm1 -; AVX-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[2,3,0,1] -; AVX-NEXT: vpmovzxbq {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero -; AVX-NEXT: vpcmpgtq %xmm1, %xmm0, %xmm2 -; AVX-NEXT: vblendvpd %xmm2, %xmm1, %xmm0, %xmm0 +; AVX-NEXT: vpsrlw $8, %xmm0, %xmm1 +; AVX-NEXT: vpminub %xmm1, %xmm0, %xmm0 ; AVX-NEXT: vpextrb $0, %xmm0, %eax ; AVX-NEXT: # kill: def $al killed $al killed $eax ; AVX-NEXT: retq ; -; AVX512BW-LABEL: test_v2i8: -; AVX512BW: # %bb.0: -; AVX512BW-NEXT: vpand {{.*}}(%rip), %xmm0, %xmm1 -; AVX512BW-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[2,3,0,1] -; AVX512BW-NEXT: vpmovzxbq {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero -; AVX512BW-NEXT: vpminuq %zmm0, %zmm1, %zmm0 -; AVX512BW-NEXT: vpextrb $0, %xmm0, %eax -; AVX512BW-NEXT: # kill: def $al killed $al killed $eax -; AVX512BW-NEXT: vzeroupper -; AVX512BW-NEXT: retq -; -; AVX512VL-LABEL: test_v2i8: -; AVX512VL: # %bb.0: -; AVX512VL-NEXT: vpand {{.*}}(%rip), %xmm0, %xmm1 -; AVX512VL-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[2,3,0,1] -; AVX512VL-NEXT: vpmovzxbq {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero -; AVX512VL-NEXT: vpminuq %xmm0, %xmm1, %xmm0 -; AVX512VL-NEXT: vpextrb $0, %xmm0, %eax -; AVX512VL-NEXT: # kill: def $al killed $al killed $eax -; AVX512VL-NEXT: retq +; AVX512-LABEL: test_v2i8: +; AVX512: # %bb.0: +; AVX512-NEXT: vpsrlw $8, %xmm0, %xmm1 +; AVX512-NEXT: vpminub %xmm1, %xmm0, %xmm0 +; AVX512-NEXT: vpextrb $0, %xmm0, %eax +; AVX512-NEXT: # kill: def $al killed $al killed $eax +; AVX512-NEXT: retq %1 = call i8 @llvm.experimental.vector.reduce.umin.v2i8(<2 x i8> %a0) ret i8 %1 } @@ -1751,83 +1590,44 @@ define i8 @test_v2i8(<2 x i8> %a0) { define i8 @test_v4i8(<4 x i8> %a0) { ; SSE2-LABEL: test_v4i8: ; SSE2: # %bb.0: -; SSE2-NEXT: movdqa {{.*#+}} xmm1 = [255,0,0,0,255,0,0,0,255,0,0,0,255,0,0,0] -; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm0[2,3,0,1] -; SSE2-NEXT: pand %xmm1, %xmm0 -; SSE2-NEXT: pand {{.*}}(%rip), %xmm2 -; SSE2-NEXT: movdqa %xmm2, %xmm3 -; SSE2-NEXT: pcmpgtd %xmm0, %xmm3 -; SSE2-NEXT: pand %xmm3, %xmm0 -; SSE2-NEXT: pandn %xmm2, %xmm3 -; SSE2-NEXT: por %xmm0, %xmm3 -; SSE2-NEXT: pand %xmm3, %xmm1 -; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm3[1,1,2,3] -; SSE2-NEXT: pxor %xmm2, %xmm2 -; SSE2-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3],xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7] -; SSE2-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3] -; SSE2-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1] -; SSE2-NEXT: movdqa %xmm0, %xmm2 -; SSE2-NEXT: pcmpgtd %xmm1, %xmm2 -; SSE2-NEXT: pand %xmm2, %xmm3 -; SSE2-NEXT: pandn %xmm0, %xmm2 -; SSE2-NEXT: por %xmm3, %xmm2 -; SSE2-NEXT: movd %xmm2, %eax +; SSE2-NEXT: movdqa %xmm0, %xmm1 +; SSE2-NEXT: psrld $16, %xmm1 +; SSE2-NEXT: pminub %xmm0, %xmm1 +; SSE2-NEXT: movdqa %xmm1, %xmm0 +; SSE2-NEXT: psrlw $8, %xmm0 +; SSE2-NEXT: pminub %xmm1, %xmm0 +; SSE2-NEXT: movd %xmm0, %eax ; SSE2-NEXT: # kill: def $al killed $al killed $eax ; SSE2-NEXT: retq ; ; SSE41-LABEL: test_v4i8: ; SSE41: # %bb.0: -; SSE41-NEXT: movdqa {{.*#+}} xmm1 = [255,0,0,0,255,0,0,0,255,0,0,0,255,0,0,0] -; SSE41-NEXT: movdqa %xmm0, %xmm2 -; SSE41-NEXT: pand %xmm1, %xmm2 -; SSE41-NEXT: pshufb {{.*#+}} xmm0 = xmm0[8],zero,zero,zero,xmm0[12],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero -; SSE41-NEXT: pminud %xmm2, %xmm0 -; SSE41-NEXT: pand %xmm0, %xmm1 -; SSE41-NEXT: pshufd {{.*#+}} xmm0 = xmm0[1,1,2,3] -; SSE41-NEXT: pmovzxbq {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero -; SSE41-NEXT: pminud %xmm1, %xmm0 +; SSE41-NEXT: movdqa %xmm0, %xmm1 +; SSE41-NEXT: psrld $16, %xmm1 +; SSE41-NEXT: pminub %xmm0, %xmm1 +; SSE41-NEXT: movdqa %xmm1, %xmm0 +; SSE41-NEXT: psrlw $8, %xmm0 +; SSE41-NEXT: pminub %xmm1, %xmm0 ; SSE41-NEXT: pextrb $0, %xmm0, %eax ; SSE41-NEXT: # kill: def $al killed $al killed $eax ; SSE41-NEXT: retq ; -; AVX1-LABEL: test_v4i8: -; AVX1: # %bb.0: -; AVX1-NEXT: vbroadcastss {{.*#+}} xmm1 = [3.57331108E-43,3.57331108E-43,3.57331108E-43,3.57331108E-43] -; AVX1-NEXT: vandps %xmm1, %xmm0, %xmm2 -; AVX1-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[8],zero,zero,zero,xmm0[12],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero -; AVX1-NEXT: vpminud %xmm0, %xmm2, %xmm0 -; AVX1-NEXT: vpand %xmm1, %xmm0, %xmm1 -; AVX1-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[1,1,2,3] -; AVX1-NEXT: vpmovzxbq {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero -; AVX1-NEXT: vpminud %xmm0, %xmm1, %xmm0 -; AVX1-NEXT: vpextrb $0, %xmm0, %eax -; AVX1-NEXT: # kill: def $al killed $al killed $eax -; AVX1-NEXT: retq -; -; AVX2-LABEL: test_v4i8: -; AVX2: # %bb.0: -; AVX2-NEXT: vpbroadcastd {{.*#+}} xmm1 = [255,255,255,255] -; AVX2-NEXT: vpand %xmm1, %xmm0, %xmm2 -; AVX2-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[8],zero,zero,zero,xmm0[12],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero -; AVX2-NEXT: vpminud %xmm0, %xmm2, %xmm0 -; AVX2-NEXT: vpand %xmm1, %xmm0, %xmm1 -; AVX2-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[1,1,2,3] -; AVX2-NEXT: vpmovzxbq {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero -; AVX2-NEXT: vpminud %xmm0, %xmm1, %xmm0 -; AVX2-NEXT: vpextrb $0, %xmm0, %eax -; AVX2-NEXT: # kill: def $al killed $al killed $eax -; AVX2-NEXT: retq +; AVX-LABEL: test_v4i8: +; AVX: # %bb.0: +; AVX-NEXT: vpsrld $16, %xmm0, %xmm1 +; AVX-NEXT: vpminub %xmm1, %xmm0, %xmm0 +; AVX-NEXT: vpsrlw $8, %xmm0, %xmm1 +; AVX-NEXT: vpminub %xmm1, %xmm0, %xmm0 +; AVX-NEXT: vpextrb $0, %xmm0, %eax +; AVX-NEXT: # kill: def $al killed $al killed $eax +; AVX-NEXT: retq ; ; AVX512-LABEL: test_v4i8: ; AVX512: # %bb.0: -; AVX512-NEXT: vpbroadcastd {{.*#+}} xmm1 = [255,255,255,255] -; AVX512-NEXT: vpand %xmm1, %xmm0, %xmm2 -; AVX512-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[8],zero,zero,zero,xmm0[12],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero -; AVX512-NEXT: vpminud %xmm0, %xmm2, %xmm0 -; AVX512-NEXT: vpand %xmm1, %xmm0, %xmm1 -; AVX512-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[1,1,2,3] -; AVX512-NEXT: vpmovzxbq {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero -; AVX512-NEXT: vpminud %xmm0, %xmm1, %xmm0 +; AVX512-NEXT: vpsrld $16, %xmm0, %xmm1 +; AVX512-NEXT: vpminub %xmm1, %xmm0, %xmm0 +; AVX512-NEXT: vpsrlw $8, %xmm0, %xmm1 +; AVX512-NEXT: vpminub %xmm1, %xmm0, %xmm0 ; AVX512-NEXT: vpextrb $0, %xmm0, %eax ; AVX512-NEXT: # kill: def $al killed $al killed $eax ; AVX512-NEXT: retq @@ -1838,92 +1638,52 @@ define i8 @test_v4i8(<4 x i8> %a0) { define i8 @test_v8i8(<8 x i8> %a0) { ; SSE2-LABEL: test_v8i8: ; SSE2: # %bb.0: -; SSE2-NEXT: movdqa {{.*#+}} xmm1 = [255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0] -; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm0[2,3,0,1] -; SSE2-NEXT: pand %xmm1, %xmm0 -; SSE2-NEXT: pand {{.*}}(%rip), %xmm2 -; SSE2-NEXT: pminsw %xmm0, %xmm2 -; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm2[1,1,2,3] -; SSE2-NEXT: pand %xmm1, %xmm2 -; SSE2-NEXT: pand {{.*}}(%rip), %xmm0 -; SSE2-NEXT: pminsw %xmm2, %xmm0 -; SSE2-NEXT: pand %xmm0, %xmm1 +; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm0[1,1,2,3] +; SSE2-NEXT: pminub %xmm0, %xmm1 +; SSE2-NEXT: movdqa %xmm1, %xmm0 ; SSE2-NEXT: psrld $16, %xmm0 -; SSE2-NEXT: pxor %xmm2, %xmm2 -; SSE2-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3],xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7] -; SSE2-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3] -; SSE2-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1] -; SSE2-NEXT: pminsw %xmm1, %xmm0 -; SSE2-NEXT: movd %xmm0, %eax +; SSE2-NEXT: pminub %xmm1, %xmm0 +; SSE2-NEXT: movdqa %xmm0, %xmm1 +; SSE2-NEXT: psrlw $8, %xmm1 +; SSE2-NEXT: pminub %xmm0, %xmm1 +; SSE2-NEXT: movd %xmm1, %eax ; SSE2-NEXT: # kill: def $al killed $al killed $eax ; SSE2-NEXT: retq ; ; SSE41-LABEL: test_v8i8: ; SSE41: # %bb.0: -; SSE41-NEXT: movdqa {{.*#+}} xmm1 = [255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0] -; SSE41-NEXT: movdqa %xmm0, %xmm2 -; SSE41-NEXT: pand %xmm1, %xmm2 -; SSE41-NEXT: pshufb {{.*#+}} xmm0 = xmm0[8],zero,xmm0[10],zero,xmm0[12],zero,xmm0[14],zero,zero,zero,zero,zero,zero,zero,zero,zero -; SSE41-NEXT: pminuw %xmm2, %xmm0 -; SSE41-NEXT: movdqa %xmm0, %xmm2 -; SSE41-NEXT: pand %xmm1, %xmm2 -; SSE41-NEXT: pshufb {{.*#+}} xmm0 = xmm0[4],zero,xmm0[6],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero -; SSE41-NEXT: pminuw %xmm2, %xmm0 -; SSE41-NEXT: pand %xmm0, %xmm1 +; SSE41-NEXT: pshufd {{.*#+}} xmm1 = xmm0[1,1,2,3] +; SSE41-NEXT: pminub %xmm0, %xmm1 +; SSE41-NEXT: movdqa %xmm1, %xmm0 ; SSE41-NEXT: psrld $16, %xmm0 -; SSE41-NEXT: pmovzxbq {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero -; SSE41-NEXT: pminuw %xmm1, %xmm0 -; SSE41-NEXT: pextrb $0, %xmm0, %eax +; SSE41-NEXT: pminub %xmm1, %xmm0 +; SSE41-NEXT: movdqa %xmm0, %xmm1 +; SSE41-NEXT: psrlw $8, %xmm1 +; SSE41-NEXT: pminub %xmm0, %xmm1 +; SSE41-NEXT: pextrb $0, %xmm1, %eax ; SSE41-NEXT: # kill: def $al killed $al killed $eax ; SSE41-NEXT: retq ; -; AVX1-LABEL: test_v8i8: -; AVX1: # %bb.0: -; AVX1-NEXT: vmovdqa {{.*#+}} xmm1 = [255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0] -; AVX1-NEXT: vpand %xmm1, %xmm0, %xmm2 -; AVX1-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[8],zero,xmm0[10],zero,xmm0[12],zero,xmm0[14],zero,zero,zero,zero,zero,zero,zero,zero,zero -; AVX1-NEXT: vpminuw %xmm0, %xmm2, %xmm0 -; AVX1-NEXT: vpand %xmm1, %xmm0, %xmm2 -; AVX1-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[4],zero,xmm0[6],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero -; AVX1-NEXT: vpminuw %xmm0, %xmm2, %xmm0 -; AVX1-NEXT: vpand %xmm1, %xmm0, %xmm1 -; AVX1-NEXT: vpsrld $16, %xmm0, %xmm0 -; AVX1-NEXT: vpmovzxbq {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero -; AVX1-NEXT: vpminuw %xmm0, %xmm1, %xmm0 -; AVX1-NEXT: vpextrb $0, %xmm0, %eax -; AVX1-NEXT: # kill: def $al killed $al killed $eax -; AVX1-NEXT: retq -; -; AVX2-LABEL: test_v8i8: -; AVX2: # %bb.0: -; AVX2-NEXT: vpbroadcastw {{.*#+}} xmm1 = [255,255,255,255,255,255,255,255] -; AVX2-NEXT: vpand %xmm1, %xmm0, %xmm2 -; AVX2-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[8],zero,xmm0[10],zero,xmm0[12],zero,xmm0[14],zero,zero,zero,zero,zero,zero,zero,zero,zero -; AVX2-NEXT: vpminuw %xmm0, %xmm2, %xmm0 -; AVX2-NEXT: vpand %xmm1, %xmm0, %xmm2 -; AVX2-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[4],zero,xmm0[6],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero -; AVX2-NEXT: vpminuw %xmm0, %xmm2, %xmm0 -; AVX2-NEXT: vpand %xmm1, %xmm0, %xmm1 -; AVX2-NEXT: vpsrld $16, %xmm0, %xmm0 -; AVX2-NEXT: vpmovzxbq {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero -; AVX2-NEXT: vpminuw %xmm0, %xmm1, %xmm0 -; AVX2-NEXT: vpextrb $0, %xmm0, %eax -; AVX2-NEXT: # kill: def $al killed $al killed $eax -; AVX2-NEXT: retq +; AVX-LABEL: test_v8i8: +; AVX: # %bb.0: +; AVX-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3] +; AVX-NEXT: vpminub %xmm1, %xmm0, %xmm0 +; AVX-NEXT: vpsrld $16, %xmm0, %xmm1 +; AVX-NEXT: vpminub %xmm1, %xmm0, %xmm0 +; AVX-NEXT: vpsrlw $8, %xmm0, %xmm1 +; AVX-NEXT: vpminub %xmm1, %xmm0, %xmm0 +; AVX-NEXT: vpextrb $0, %xmm0, %eax +; AVX-NEXT: # kill: def $al killed $al killed $eax +; AVX-NEXT: retq ; ; AVX512-LABEL: test_v8i8: ; AVX512: # %bb.0: -; AVX512-NEXT: vpbroadcastw {{.*#+}} xmm1 = [255,255,255,255,255,255,255,255] -; AVX512-NEXT: vpand %xmm1, %xmm0, %xmm2 -; AVX512-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[8],zero,xmm0[10],zero,xmm0[12],zero,xmm0[14],zero,zero,zero,zero,zero,zero,zero,zero,zero -; AVX512-NEXT: vpminuw %xmm0, %xmm2, %xmm0 -; AVX512-NEXT: vpand %xmm1, %xmm0, %xmm2 -; AVX512-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[4],zero,xmm0[6],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero -; AVX512-NEXT: vpminuw %xmm0, %xmm2, %xmm0 -; AVX512-NEXT: vpand %xmm1, %xmm0, %xmm1 -; AVX512-NEXT: vpsrld $16, %xmm0, %xmm0 -; AVX512-NEXT: vpmovzxbq {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero -; AVX512-NEXT: vpminuw %xmm0, %xmm1, %xmm0 +; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3] +; AVX512-NEXT: vpminub %xmm1, %xmm0, %xmm0 +; AVX512-NEXT: vpsrld $16, %xmm0, %xmm1 +; AVX512-NEXT: vpminub %xmm1, %xmm0, %xmm0 +; AVX512-NEXT: vpsrlw $8, %xmm0, %xmm1 +; AVX512-NEXT: vpminub %xmm1, %xmm0, %xmm0 ; AVX512-NEXT: vpextrb $0, %xmm0, %eax ; AVX512-NEXT: # kill: def $al killed $al killed $eax ; AVX512-NEXT: retq |
