diff options
Diffstat (limited to 'llvm/test/CodeGen/X86/trunc-subvector.ll')
-rw-r--r-- | llvm/test/CodeGen/X86/trunc-subvector.ll | 134 |
1 files changed, 48 insertions, 86 deletions
diff --git a/llvm/test/CodeGen/X86/trunc-subvector.ll b/llvm/test/CodeGen/X86/trunc-subvector.ll index d52fe0f8ff8..05b2b6608ad 100644 --- a/llvm/test/CodeGen/X86/trunc-subvector.ll +++ b/llvm/test/CodeGen/X86/trunc-subvector.ll @@ -40,25 +40,14 @@ define <4 x i32> @test2(<8 x i32> %v) { define <2 x i32> @test3(<8 x i32> %v) { ; SSE2-LABEL: test3: ; SSE2: # %bb.0: -; SSE2-NEXT: movdqa %xmm1, %xmm0 -; SSE2-NEXT: pxor %xmm1, %xmm1 -; SSE2-NEXT: pcmpgtd %xmm0, %xmm1 -; SSE2-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1] +; SSE2-NEXT: movaps %xmm1, %xmm0 ; SSE2-NEXT: retq ; -; AVX2-LABEL: test3: -; AVX2: # %bb.0: -; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm0 -; AVX2-NEXT: vpmovsxdq %xmm0, %xmm0 -; AVX2-NEXT: vzeroupper -; AVX2-NEXT: retq -; -; AVX512-LABEL: test3: -; AVX512: # %bb.0: -; AVX512-NEXT: vpmovsxdq %ymm0, %zmm0 -; AVX512-NEXT: vextracti32x4 $2, %zmm0, %xmm0 -; AVX512-NEXT: vzeroupper -; AVX512-NEXT: retq +; AVX-LABEL: test3: +; AVX: # %bb.0: +; AVX-NEXT: vextractf128 $1, %ymm0, %xmm0 +; AVX-NEXT: vzeroupper +; AVX-NEXT: retq %x = sext <8 x i32> %v to <8 x i64> %s = shufflevector <8 x i64> %x, <8 x i64> undef, <2 x i32> <i32 4, i32 5> %t = trunc <2 x i64> %s to <2 x i32> @@ -68,23 +57,13 @@ define <2 x i32> @test3(<8 x i32> %v) { define <2 x i32> @test4(<8 x i32> %v) { ; SSE2-LABEL: test4: ; SSE2: # %bb.0: -; SSE2-NEXT: pxor %xmm1, %xmm1 -; SSE2-NEXT: pcmpgtd %xmm0, %xmm1 -; SSE2-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1] ; SSE2-NEXT: retq ; -; AVX2-LABEL: test4: -; AVX2: # %bb.0: -; AVX2-NEXT: vpmovsxdq %xmm0, %xmm0 -; AVX2-NEXT: vzeroupper -; AVX2-NEXT: retq -; -; AVX512-LABEL: test4: -; AVX512: # %bb.0: -; AVX512-NEXT: vpmovsxdq %ymm0, %zmm0 -; AVX512-NEXT: # kill: def $xmm0 killed $xmm0 killed $zmm0 -; AVX512-NEXT: vzeroupper -; AVX512-NEXT: retq +; AVX-LABEL: test4: +; AVX: # %bb.0: +; AVX-NEXT: # kill: def $xmm0 killed $xmm0 killed $ymm0 +; AVX-NEXT: vzeroupper +; AVX-NEXT: retq %x = sext <8 x i32> %v to <8 x i64> %s = shufflevector <8 x i64> %x, <8 x i64> undef, <2 x i32> <i32 0, i32 1> %t = trunc <2 x i64> %s to <2 x i32> @@ -94,32 +73,31 @@ define <2 x i32> @test4(<8 x i32> %v) { define <2 x i32> @test5(<8 x i32> %v) { ; SSE2-LABEL: test5: ; SSE2: # %bb.0: -; SSE2-NEXT: pxor %xmm2, %xmm2 -; SSE2-NEXT: pxor %xmm3, %xmm3 -; SSE2-NEXT: pcmpgtd %xmm1, %xmm3 -; SSE2-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1] -; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,3,0,1] -; SSE2-NEXT: pcmpgtd %xmm0, %xmm2 -; SSE2-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1] -; SSE2-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,3],xmm1[0,1] +; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[3,3,2,2] +; SSE2-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1] ; SSE2-NEXT: retq ; ; AVX2-LABEL: test5: ; AVX2: # %bb.0: -; AVX2-NEXT: vpmovsxdq %xmm0, %ymm1 +; AVX2-NEXT: vpmovzxdq {{.*#+}} ymm1 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero ; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm0 -; AVX2-NEXT: vpmovsxdq %xmm0, %xmm0 -; AVX2-NEXT: vextracti128 $1, %ymm1, %xmm1 -; AVX2-NEXT: vpalignr {{.*#+}} xmm0 = xmm1[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7] +; AVX2-NEXT: vpmovzxdq {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero +; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7] +; AVX2-NEXT: vpbroadcastq {{.*#+}} ymm1 = [6,6,6,6] +; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0 +; AVX2-NEXT: # kill: def $xmm0 killed $xmm0 killed $ymm0 ; AVX2-NEXT: vzeroupper ; AVX2-NEXT: retq ; ; AVX512-LABEL: test5: ; AVX512: # %bb.0: ; AVX512-NEXT: vpmovsxdq %ymm0, %zmm0 -; AVX512-NEXT: vextracti32x4 $2, %zmm0, %xmm1 -; AVX512-NEXT: vextracti128 $1, %ymm0, %xmm0 -; AVX512-NEXT: vpalignr {{.*#+}} xmm0 = xmm0[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7] +; AVX512-NEXT: vextracti128 $1, %ymm0, %xmm1 +; AVX512-NEXT: vpextrq $1, %xmm1, %rax +; AVX512-NEXT: vextracti32x4 $2, %zmm0, %xmm0 +; AVX512-NEXT: vmovq %xmm0, %rcx +; AVX512-NEXT: vmovd %eax, %xmm0 +; AVX512-NEXT: vpinsrd $1, %ecx, %xmm0, %xmm0 ; AVX512-NEXT: vzeroupper ; AVX512-NEXT: retq %x = sext <8 x i32> %v to <8 x i64> @@ -165,23 +143,13 @@ define <2 x i32> @test8(<8 x i32> %v) { ; SSE2-LABEL: test8: ; SSE2: # %bb.0: ; SSE2-NEXT: movaps %xmm1, %xmm0 -; SSE2-NEXT: xorps %xmm1, %xmm1 -; SSE2-NEXT: unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1] ; SSE2-NEXT: retq ; -; AVX2-LABEL: test8: -; AVX2: # %bb.0: -; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm0 -; AVX2-NEXT: vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero -; AVX2-NEXT: vzeroupper -; AVX2-NEXT: retq -; -; AVX512-LABEL: test8: -; AVX512: # %bb.0: -; AVX512-NEXT: vpmovzxdq {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero -; AVX512-NEXT: vextracti32x4 $2, %zmm0, %xmm0 -; AVX512-NEXT: vzeroupper -; AVX512-NEXT: retq +; AVX-LABEL: test8: +; AVX: # %bb.0: +; AVX-NEXT: vextractf128 $1, %ymm0, %xmm0 +; AVX-NEXT: vzeroupper +; AVX-NEXT: retq %x = zext <8 x i32> %v to <8 x i64> %s = shufflevector <8 x i64> %x, <8 x i64> undef, <2 x i32> <i32 4, i32 5> %t = trunc <2 x i64> %s to <2 x i32> @@ -191,22 +159,13 @@ define <2 x i32> @test8(<8 x i32> %v) { define <2 x i32> @test9(<8 x i32> %v) { ; SSE2-LABEL: test9: ; SSE2: # %bb.0: -; SSE2-NEXT: xorps %xmm1, %xmm1 -; SSE2-NEXT: unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1] ; SSE2-NEXT: retq ; -; AVX2-LABEL: test9: -; AVX2: # %bb.0: -; AVX2-NEXT: vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero -; AVX2-NEXT: vzeroupper -; AVX2-NEXT: retq -; -; AVX512-LABEL: test9: -; AVX512: # %bb.0: -; AVX512-NEXT: vpmovzxdq {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero -; AVX512-NEXT: # kill: def $xmm0 killed $xmm0 killed $zmm0 -; AVX512-NEXT: vzeroupper -; AVX512-NEXT: retq +; AVX-LABEL: test9: +; AVX: # %bb.0: +; AVX-NEXT: # kill: def $xmm0 killed $xmm0 killed $ymm0 +; AVX-NEXT: vzeroupper +; AVX-NEXT: retq %x = zext <8 x i32> %v to <8 x i64> %s = shufflevector <8 x i64> %x, <8 x i64> undef, <2 x i32> <i32 0, i32 1> %t = trunc <2 x i64> %s to <2 x i32> @@ -216,28 +175,31 @@ define <2 x i32> @test9(<8 x i32> %v) { define <2 x i32> @test10(<8 x i32> %v) { ; SSE2-LABEL: test10: ; SSE2: # %bb.0: -; SSE2-NEXT: xorps %xmm2, %xmm2 -; SSE2-NEXT: unpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1] -; SSE2-NEXT: unpckhps {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3] -; SSE2-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,3],xmm1[0,1] +; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[3,3,2,2] +; SSE2-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1] ; SSE2-NEXT: retq ; ; AVX2-LABEL: test10: ; AVX2: # %bb.0: ; AVX2-NEXT: vpmovzxdq {{.*#+}} ymm1 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero ; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm0 -; AVX2-NEXT: vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero -; AVX2-NEXT: vextracti128 $1, %ymm1, %xmm1 -; AVX2-NEXT: vpalignr {{.*#+}} xmm0 = xmm1[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7] +; AVX2-NEXT: vpmovzxdq {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero +; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7] +; AVX2-NEXT: vpbroadcastq {{.*#+}} ymm1 = [6,6,6,6] +; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0 +; AVX2-NEXT: # kill: def $xmm0 killed $xmm0 killed $ymm0 ; AVX2-NEXT: vzeroupper ; AVX2-NEXT: retq ; ; AVX512-LABEL: test10: ; AVX512: # %bb.0: ; AVX512-NEXT: vpmovzxdq {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero -; AVX512-NEXT: vextracti32x4 $2, %zmm0, %xmm1 -; AVX512-NEXT: vextracti128 $1, %ymm0, %xmm0 -; AVX512-NEXT: vpalignr {{.*#+}} xmm0 = xmm0[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7] +; AVX512-NEXT: vextracti128 $1, %ymm0, %xmm1 +; AVX512-NEXT: vpextrq $1, %xmm1, %rax +; AVX512-NEXT: vextracti32x4 $2, %zmm0, %xmm0 +; AVX512-NEXT: vmovq %xmm0, %rcx +; AVX512-NEXT: vmovd %eax, %xmm0 +; AVX512-NEXT: vpinsrd $1, %ecx, %xmm0, %xmm0 ; AVX512-NEXT: vzeroupper ; AVX512-NEXT: retq %x = zext <8 x i32> %v to <8 x i64> |