| ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py |
| ; RUN: llc < %s -mtriple=x86_64-unknown-unknown | FileCheck %s --check-prefix=SSE --check-prefix=SSE2 |
| ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse4.2 | FileCheck %s --check-prefix=SSE --check-prefix=SSE42 |
| ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx | FileCheck %s --check-prefix=AVX --check-prefix=AVX1 |
| ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefix=AVX --check-prefix=AVX2 |
| |
| ; |
| ; PR6455 'Clear Upper Bits' Patterns |
| ; |
| |
| define <2 x i64> @_clearupper2xi64a(<2 x i64>) nounwind { |
| ; SSE2-LABEL: _clearupper2xi64a: |
| ; SSE2: # %bb.0: |
| ; SSE2-NEXT: andps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0 |
| ; SSE2-NEXT: retq |
| ; |
| ; SSE42-LABEL: _clearupper2xi64a: |
| ; SSE42: # %bb.0: |
| ; SSE42-NEXT: xorps %xmm1, %xmm1 |
| ; SSE42-NEXT: blendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2],xmm1[3] |
| ; SSE42-NEXT: retq |
| ; |
| ; AVX-LABEL: _clearupper2xi64a: |
| ; AVX: # %bb.0: |
| ; AVX-NEXT: vxorps %xmm1, %xmm1, %xmm1 |
| ; AVX-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2],xmm1[3] |
| ; AVX-NEXT: retq |
| %x0 = extractelement <2 x i64> %0, i32 0 |
| %x1 = extractelement <2 x i64> %0, i32 1 |
| %trunc0 = trunc i64 %x0 to i32 |
| %trunc1 = trunc i64 %x1 to i32 |
| %ext0 = zext i32 %trunc0 to i64 |
| %ext1 = zext i32 %trunc1 to i64 |
| %v0 = insertelement <2 x i64> undef, i64 %ext0, i32 0 |
| %v1 = insertelement <2 x i64> %v0, i64 %ext1, i32 1 |
| ret <2 x i64> %v1 |
| } |
| |
| define <4 x i64> @_clearupper4xi64a(<4 x i64>) nounwind { |
| ; SSE2-LABEL: _clearupper4xi64a: |
| ; SSE2: # %bb.0: |
| ; SSE2-NEXT: movaps {{.*#+}} xmm2 = [4294967295,4294967295] |
| ; SSE2-NEXT: andps %xmm2, %xmm0 |
| ; SSE2-NEXT: andps %xmm2, %xmm1 |
| ; SSE2-NEXT: retq |
| ; |
| ; SSE42-LABEL: _clearupper4xi64a: |
| ; SSE42: # %bb.0: |
| ; SSE42-NEXT: xorps %xmm2, %xmm2 |
| ; SSE42-NEXT: blendps {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2],xmm2[3] |
| ; SSE42-NEXT: blendps {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2],xmm2[3] |
| ; SSE42-NEXT: retq |
| ; |
| ; AVX-LABEL: _clearupper4xi64a: |
| ; AVX: # %bb.0: |
| ; AVX-NEXT: vxorps %xmm1, %xmm1, %xmm1 |
| ; AVX-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7] |
| ; AVX-NEXT: retq |
| %x0 = extractelement <4 x i64> %0, i32 0 |
| %x1 = extractelement <4 x i64> %0, i32 1 |
| %x2 = extractelement <4 x i64> %0, i32 2 |
| %x3 = extractelement <4 x i64> %0, i32 3 |
| %trunc0 = trunc i64 %x0 to i32 |
| %trunc1 = trunc i64 %x1 to i32 |
| %trunc2 = trunc i64 %x2 to i32 |
| %trunc3 = trunc i64 %x3 to i32 |
| %ext0 = zext i32 %trunc0 to i64 |
| %ext1 = zext i32 %trunc1 to i64 |
| %ext2 = zext i32 %trunc2 to i64 |
| %ext3 = zext i32 %trunc3 to i64 |
| %v0 = insertelement <4 x i64> undef, i64 %ext0, i32 0 |
| %v1 = insertelement <4 x i64> %v0, i64 %ext1, i32 1 |
| %v2 = insertelement <4 x i64> %v1, i64 %ext2, i32 2 |
| %v3 = insertelement <4 x i64> %v2, i64 %ext3, i32 3 |
| ret <4 x i64> %v3 |
| } |
| |
| define <4 x i32> @_clearupper4xi32a(<4 x i32>) nounwind { |
| ; SSE2-LABEL: _clearupper4xi32a: |
| ; SSE2: # %bb.0: |
| ; SSE2-NEXT: andps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0 |
| ; SSE2-NEXT: retq |
| ; |
| ; SSE42-LABEL: _clearupper4xi32a: |
| ; SSE42: # %bb.0: |
| ; SSE42-NEXT: pxor %xmm1, %xmm1 |
| ; SSE42-NEXT: pblendw {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2],xmm1[3],xmm0[4],xmm1[5],xmm0[6],xmm1[7] |
| ; SSE42-NEXT: retq |
| ; |
| ; AVX-LABEL: _clearupper4xi32a: |
| ; AVX: # %bb.0: |
| ; AVX-NEXT: vpxor %xmm1, %xmm1, %xmm1 |
| ; AVX-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2],xmm1[3],xmm0[4],xmm1[5],xmm0[6],xmm1[7] |
| ; AVX-NEXT: retq |
| %x0 = extractelement <4 x i32> %0, i32 0 |
| %x1 = extractelement <4 x i32> %0, i32 1 |
| %x2 = extractelement <4 x i32> %0, i32 2 |
| %x3 = extractelement <4 x i32> %0, i32 3 |
| %trunc0 = trunc i32 %x0 to i16 |
| %trunc1 = trunc i32 %x1 to i16 |
| %trunc2 = trunc i32 %x2 to i16 |
| %trunc3 = trunc i32 %x3 to i16 |
| %ext0 = zext i16 %trunc0 to i32 |
| %ext1 = zext i16 %trunc1 to i32 |
| %ext2 = zext i16 %trunc2 to i32 |
| %ext3 = zext i16 %trunc3 to i32 |
| %v0 = insertelement <4 x i32> undef, i32 %ext0, i32 0 |
| %v1 = insertelement <4 x i32> %v0, i32 %ext1, i32 1 |
| %v2 = insertelement <4 x i32> %v1, i32 %ext2, i32 2 |
| %v3 = insertelement <4 x i32> %v2, i32 %ext3, i32 3 |
| ret <4 x i32> %v3 |
| } |
| |
| define <8 x i32> @_clearupper8xi32a(<8 x i32>) nounwind { |
| ; SSE2-LABEL: _clearupper8xi32a: |
| ; SSE2: # %bb.0: |
| ; SSE2-NEXT: movaps {{.*#+}} xmm2 = [65535,65535,65535,65535] |
| ; SSE2-NEXT: andps %xmm2, %xmm0 |
| ; SSE2-NEXT: andps %xmm2, %xmm1 |
| ; SSE2-NEXT: retq |
| ; |
| ; SSE42-LABEL: _clearupper8xi32a: |
| ; SSE42: # %bb.0: |
| ; SSE42-NEXT: pxor %xmm2, %xmm2 |
| ; SSE42-NEXT: pblendw {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2],xmm2[3],xmm0[4],xmm2[5],xmm0[6],xmm2[7] |
| ; SSE42-NEXT: pblendw {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2],xmm2[3],xmm1[4],xmm2[5],xmm1[6],xmm2[7] |
| ; SSE42-NEXT: retq |
| ; |
| ; AVX1-LABEL: _clearupper8xi32a: |
| ; AVX1: # %bb.0: |
| ; AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm0 |
| ; AVX1-NEXT: retq |
| ; |
| ; AVX2-LABEL: _clearupper8xi32a: |
| ; AVX2: # %bb.0: |
| ; AVX2-NEXT: vpxor %xmm1, %xmm1, %xmm1 |
| ; AVX2-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7],ymm0[8],ymm1[9],ymm0[10],ymm1[11],ymm0[12],ymm1[13],ymm0[14],ymm1[15] |
| ; AVX2-NEXT: retq |
| %x0 = extractelement <8 x i32> %0, i32 0 |
| %x1 = extractelement <8 x i32> %0, i32 1 |
| %x2 = extractelement <8 x i32> %0, i32 2 |
| %x3 = extractelement <8 x i32> %0, i32 3 |
| %x4 = extractelement <8 x i32> %0, i32 4 |
| %x5 = extractelement <8 x i32> %0, i32 5 |
| %x6 = extractelement <8 x i32> %0, i32 6 |
| %x7 = extractelement <8 x i32> %0, i32 7 |
| %trunc0 = trunc i32 %x0 to i16 |
| %trunc1 = trunc i32 %x1 to i16 |
| %trunc2 = trunc i32 %x2 to i16 |
| %trunc3 = trunc i32 %x3 to i16 |
| %trunc4 = trunc i32 %x4 to i16 |
| %trunc5 = trunc i32 %x5 to i16 |
| %trunc6 = trunc i32 %x6 to i16 |
| %trunc7 = trunc i32 %x7 to i16 |
| %ext0 = zext i16 %trunc0 to i32 |
| %ext1 = zext i16 %trunc1 to i32 |
| %ext2 = zext i16 %trunc2 to i32 |
| %ext3 = zext i16 %trunc3 to i32 |
| %ext4 = zext i16 %trunc4 to i32 |
| %ext5 = zext i16 %trunc5 to i32 |
| %ext6 = zext i16 %trunc6 to i32 |
| %ext7 = zext i16 %trunc7 to i32 |
| %v0 = insertelement <8 x i32> undef, i32 %ext0, i32 0 |
| %v1 = insertelement <8 x i32> %v0, i32 %ext1, i32 1 |
| %v2 = insertelement <8 x i32> %v1, i32 %ext2, i32 2 |
| %v3 = insertelement <8 x i32> %v2, i32 %ext3, i32 3 |
| %v4 = insertelement <8 x i32> %v3, i32 %ext4, i32 4 |
| %v5 = insertelement <8 x i32> %v4, i32 %ext5, i32 5 |
| %v6 = insertelement <8 x i32> %v5, i32 %ext6, i32 6 |
| %v7 = insertelement <8 x i32> %v6, i32 %ext7, i32 7 |
| ret <8 x i32> %v7 |
| } |
| |
| define <8 x i16> @_clearupper8xi16a(<8 x i16>) nounwind { |
| ; SSE-LABEL: _clearupper8xi16a: |
| ; SSE: # %bb.0: |
| ; SSE-NEXT: andps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0 |
| ; SSE-NEXT: retq |
| ; |
| ; AVX-LABEL: _clearupper8xi16a: |
| ; AVX: # %bb.0: |
| ; AVX-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0 |
| ; AVX-NEXT: retq |
| %x0 = extractelement <8 x i16> %0, i32 0 |
| %x1 = extractelement <8 x i16> %0, i32 1 |
| %x2 = extractelement <8 x i16> %0, i32 2 |
| %x3 = extractelement <8 x i16> %0, i32 3 |
| %x4 = extractelement <8 x i16> %0, i32 4 |
| %x5 = extractelement <8 x i16> %0, i32 5 |
| %x6 = extractelement <8 x i16> %0, i32 6 |
| %x7 = extractelement <8 x i16> %0, i32 7 |
| %trunc0 = trunc i16 %x0 to i8 |
| %trunc1 = trunc i16 %x1 to i8 |
| %trunc2 = trunc i16 %x2 to i8 |
| %trunc3 = trunc i16 %x3 to i8 |
| %trunc4 = trunc i16 %x4 to i8 |
| %trunc5 = trunc i16 %x5 to i8 |
| %trunc6 = trunc i16 %x6 to i8 |
| %trunc7 = trunc i16 %x7 to i8 |
| %ext0 = zext i8 %trunc0 to i16 |
| %ext1 = zext i8 %trunc1 to i16 |
| %ext2 = zext i8 %trunc2 to i16 |
| %ext3 = zext i8 %trunc3 to i16 |
| %ext4 = zext i8 %trunc4 to i16 |
| %ext5 = zext i8 %trunc5 to i16 |
| %ext6 = zext i8 %trunc6 to i16 |
| %ext7 = zext i8 %trunc7 to i16 |
| %v0 = insertelement <8 x i16> undef, i16 %ext0, i32 0 |
| %v1 = insertelement <8 x i16> %v0, i16 %ext1, i32 1 |
| %v2 = insertelement <8 x i16> %v1, i16 %ext2, i32 2 |
| %v3 = insertelement <8 x i16> %v2, i16 %ext3, i32 3 |
| %v4 = insertelement <8 x i16> %v3, i16 %ext4, i32 4 |
| %v5 = insertelement <8 x i16> %v4, i16 %ext5, i32 5 |
| %v6 = insertelement <8 x i16> %v5, i16 %ext6, i32 6 |
| %v7 = insertelement <8 x i16> %v6, i16 %ext7, i32 7 |
| ret <8 x i16> %v7 |
| } |
| |
| define <16 x i16> @_clearupper16xi16a(<16 x i16>) nounwind { |
| ; SSE-LABEL: _clearupper16xi16a: |
| ; SSE: # %bb.0: |
| ; SSE-NEXT: movaps {{.*#+}} xmm2 = [255,255,255,255,255,255,255,255] |
| ; SSE-NEXT: andps %xmm2, %xmm0 |
| ; SSE-NEXT: andps %xmm2, %xmm1 |
| ; SSE-NEXT: retq |
| ; |
| ; AVX-LABEL: _clearupper16xi16a: |
| ; AVX: # %bb.0: |
| ; AVX-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm0 |
| ; AVX-NEXT: retq |
| %x0 = extractelement <16 x i16> %0, i32 0 |
| %x1 = extractelement <16 x i16> %0, i32 1 |
| %x2 = extractelement <16 x i16> %0, i32 2 |
| %x3 = extractelement <16 x i16> %0, i32 3 |
| %x4 = extractelement <16 x i16> %0, i32 4 |
| %x5 = extractelement <16 x i16> %0, i32 5 |
| %x6 = extractelement <16 x i16> %0, i32 6 |
| %x7 = extractelement <16 x i16> %0, i32 7 |
| %x8 = extractelement <16 x i16> %0, i32 8 |
| %x9 = extractelement <16 x i16> %0, i32 9 |
| %x10 = extractelement <16 x i16> %0, i32 10 |
| %x11 = extractelement <16 x i16> %0, i32 11 |
| %x12 = extractelement <16 x i16> %0, i32 12 |
| %x13 = extractelement <16 x i16> %0, i32 13 |
| %x14 = extractelement <16 x i16> %0, i32 14 |
| %x15 = extractelement <16 x i16> %0, i32 15 |
| %trunc0 = trunc i16 %x0 to i8 |
| %trunc1 = trunc i16 %x1 to i8 |
| %trunc2 = trunc i16 %x2 to i8 |
| %trunc3 = trunc i16 %x3 to i8 |
| %trunc4 = trunc i16 %x4 to i8 |
| %trunc5 = trunc i16 %x5 to i8 |
| %trunc6 = trunc i16 %x6 to i8 |
| %trunc7 = trunc i16 %x7 to i8 |
| %trunc8 = trunc i16 %x8 to i8 |
| %trunc9 = trunc i16 %x9 to i8 |
| %trunc10 = trunc i16 %x10 to i8 |
| %trunc11 = trunc i16 %x11 to i8 |
| %trunc12 = trunc i16 %x12 to i8 |
| %trunc13 = trunc i16 %x13 to i8 |
| %trunc14 = trunc i16 %x14 to i8 |
| %trunc15 = trunc i16 %x15 to i8 |
| %ext0 = zext i8 %trunc0 to i16 |
| %ext1 = zext i8 %trunc1 to i16 |
| %ext2 = zext i8 %trunc2 to i16 |
| %ext3 = zext i8 %trunc3 to i16 |
| %ext4 = zext i8 %trunc4 to i16 |
| %ext5 = zext i8 %trunc5 to i16 |
| %ext6 = zext i8 %trunc6 to i16 |
| %ext7 = zext i8 %trunc7 to i16 |
| %ext8 = zext i8 %trunc8 to i16 |
| %ext9 = zext i8 %trunc9 to i16 |
| %ext10 = zext i8 %trunc10 to i16 |
| %ext11 = zext i8 %trunc11 to i16 |
| %ext12 = zext i8 %trunc12 to i16 |
| %ext13 = zext i8 %trunc13 to i16 |
| %ext14 = zext i8 %trunc14 to i16 |
| %ext15 = zext i8 %trunc15 to i16 |
| %v0 = insertelement <16 x i16> undef, i16 %ext0, i32 0 |
| %v1 = insertelement <16 x i16> %v0, i16 %ext1, i32 1 |
| %v2 = insertelement <16 x i16> %v1, i16 %ext2, i32 2 |
| %v3 = insertelement <16 x i16> %v2, i16 %ext3, i32 3 |
| %v4 = insertelement <16 x i16> %v3, i16 %ext4, i32 4 |
| %v5 = insertelement <16 x i16> %v4, i16 %ext5, i32 5 |
| %v6 = insertelement <16 x i16> %v5, i16 %ext6, i32 6 |
| %v7 = insertelement <16 x i16> %v6, i16 %ext7, i32 7 |
| %v8 = insertelement <16 x i16> %v7, i16 %ext8, i32 8 |
| %v9 = insertelement <16 x i16> %v8, i16 %ext9, i32 9 |
| %v10 = insertelement <16 x i16> %v9, i16 %ext10, i32 10 |
| %v11 = insertelement <16 x i16> %v10, i16 %ext11, i32 11 |
| %v12 = insertelement <16 x i16> %v11, i16 %ext12, i32 12 |
| %v13 = insertelement <16 x i16> %v12, i16 %ext13, i32 13 |
| %v14 = insertelement <16 x i16> %v13, i16 %ext14, i32 14 |
| %v15 = insertelement <16 x i16> %v14, i16 %ext15, i32 15 |
| ret <16 x i16> %v15 |
| } |
| |
| define <16 x i8> @_clearupper16xi8a(<16 x i8>) nounwind { |
| ; SSE-LABEL: _clearupper16xi8a: |
| ; SSE: # %bb.0: |
| ; SSE-NEXT: andps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0 |
| ; SSE-NEXT: retq |
| ; |
| ; AVX-LABEL: _clearupper16xi8a: |
| ; AVX: # %bb.0: |
| ; AVX-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0 |
| ; AVX-NEXT: retq |
| %x0 = extractelement <16 x i8> %0, i32 0 |
| %x1 = extractelement <16 x i8> %0, i32 1 |
| %x2 = extractelement <16 x i8> %0, i32 2 |
| %x3 = extractelement <16 x i8> %0, i32 3 |
| %x4 = extractelement <16 x i8> %0, i32 4 |
| %x5 = extractelement <16 x i8> %0, i32 5 |
| %x6 = extractelement <16 x i8> %0, i32 6 |
| %x7 = extractelement <16 x i8> %0, i32 7 |
| %x8 = extractelement <16 x i8> %0, i32 8 |
| %x9 = extractelement <16 x i8> %0, i32 9 |
| %x10 = extractelement <16 x i8> %0, i32 10 |
| %x11 = extractelement <16 x i8> %0, i32 11 |
| %x12 = extractelement <16 x i8> %0, i32 12 |
| %x13 = extractelement <16 x i8> %0, i32 13 |
| %x14 = extractelement <16 x i8> %0, i32 14 |
| %x15 = extractelement <16 x i8> %0, i32 15 |
| %trunc0 = trunc i8 %x0 to i4 |
| %trunc1 = trunc i8 %x1 to i4 |
| %trunc2 = trunc i8 %x2 to i4 |
| %trunc3 = trunc i8 %x3 to i4 |
| %trunc4 = trunc i8 %x4 to i4 |
| %trunc5 = trunc i8 %x5 to i4 |
| %trunc6 = trunc i8 %x6 to i4 |
| %trunc7 = trunc i8 %x7 to i4 |
| %trunc8 = trunc i8 %x8 to i4 |
| %trunc9 = trunc i8 %x9 to i4 |
| %trunc10 = trunc i8 %x10 to i4 |
| %trunc11 = trunc i8 %x11 to i4 |
| %trunc12 = trunc i8 %x12 to i4 |
| %trunc13 = trunc i8 %x13 to i4 |
| %trunc14 = trunc i8 %x14 to i4 |
| %trunc15 = trunc i8 %x15 to i4 |
| %ext0 = zext i4 %trunc0 to i8 |
| %ext1 = zext i4 %trunc1 to i8 |
| %ext2 = zext i4 %trunc2 to i8 |
| %ext3 = zext i4 %trunc3 to i8 |
| %ext4 = zext i4 %trunc4 to i8 |
| %ext5 = zext i4 %trunc5 to i8 |
| %ext6 = zext i4 %trunc6 to i8 |
| %ext7 = zext i4 %trunc7 to i8 |
| %ext8 = zext i4 %trunc8 to i8 |
| %ext9 = zext i4 %trunc9 to i8 |
| %ext10 = zext i4 %trunc10 to i8 |
| %ext11 = zext i4 %trunc11 to i8 |
| %ext12 = zext i4 %trunc12 to i8 |
| %ext13 = zext i4 %trunc13 to i8 |
| %ext14 = zext i4 %trunc14 to i8 |
| %ext15 = zext i4 %trunc15 to i8 |
| %v0 = insertelement <16 x i8> undef, i8 %ext0, i32 0 |
| %v1 = insertelement <16 x i8> %v0, i8 %ext1, i32 1 |
| %v2 = insertelement <16 x i8> %v1, i8 %ext2, i32 2 |
| %v3 = insertelement <16 x i8> %v2, i8 %ext3, i32 3 |
| %v4 = insertelement <16 x i8> %v3, i8 %ext4, i32 4 |
| %v5 = insertelement <16 x i8> %v4, i8 %ext5, i32 5 |
| %v6 = insertelement <16 x i8> %v5, i8 %ext6, i32 6 |
| %v7 = insertelement <16 x i8> %v6, i8 %ext7, i32 7 |
| %v8 = insertelement <16 x i8> %v7, i8 %ext8, i32 8 |
| %v9 = insertelement <16 x i8> %v8, i8 %ext9, i32 9 |
| %v10 = insertelement <16 x i8> %v9, i8 %ext10, i32 10 |
| %v11 = insertelement <16 x i8> %v10, i8 %ext11, i32 11 |
| %v12 = insertelement <16 x i8> %v11, i8 %ext12, i32 12 |
| %v13 = insertelement <16 x i8> %v12, i8 %ext13, i32 13 |
| %v14 = insertelement <16 x i8> %v13, i8 %ext14, i32 14 |
| %v15 = insertelement <16 x i8> %v14, i8 %ext15, i32 15 |
| ret <16 x i8> %v15 |
| } |
| |
| define <32 x i8> @_clearupper32xi8a(<32 x i8>) nounwind { |
| ; SSE-LABEL: _clearupper32xi8a: |
| ; SSE: # %bb.0: |
| ; SSE-NEXT: movaps {{.*#+}} xmm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15] |
| ; SSE-NEXT: andps %xmm2, %xmm0 |
| ; SSE-NEXT: andps %xmm2, %xmm1 |
| ; SSE-NEXT: retq |
| ; |
| ; AVX-LABEL: _clearupper32xi8a: |
| ; AVX: # %bb.0: |
| ; AVX-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm0 |
| ; AVX-NEXT: retq |
| %x0 = extractelement <32 x i8> %0, i32 0 |
| %x1 = extractelement <32 x i8> %0, i32 1 |
| %x2 = extractelement <32 x i8> %0, i32 2 |
| %x3 = extractelement <32 x i8> %0, i32 3 |
| %x4 = extractelement <32 x i8> %0, i32 4 |
| %x5 = extractelement <32 x i8> %0, i32 5 |
| %x6 = extractelement <32 x i8> %0, i32 6 |
| %x7 = extractelement <32 x i8> %0, i32 7 |
| %x8 = extractelement <32 x i8> %0, i32 8 |
| %x9 = extractelement <32 x i8> %0, i32 9 |
| %x10 = extractelement <32 x i8> %0, i32 10 |
| %x11 = extractelement <32 x i8> %0, i32 11 |
| %x12 = extractelement <32 x i8> %0, i32 12 |
| %x13 = extractelement <32 x i8> %0, i32 13 |
| %x14 = extractelement <32 x i8> %0, i32 14 |
| %x15 = extractelement <32 x i8> %0, i32 15 |
| %x16 = extractelement <32 x i8> %0, i32 16 |
| %x17 = extractelement <32 x i8> %0, i32 17 |
| %x18 = extractelement <32 x i8> %0, i32 18 |
| %x19 = extractelement <32 x i8> %0, i32 19 |
| %x20 = extractelement <32 x i8> %0, i32 20 |
| %x21 = extractelement <32 x i8> %0, i32 21 |
| %x22 = extractelement <32 x i8> %0, i32 22 |
| %x23 = extractelement <32 x i8> %0, i32 23 |
| %x24 = extractelement <32 x i8> %0, i32 24 |
| %x25 = extractelement <32 x i8> %0, i32 25 |
| %x26 = extractelement <32 x i8> %0, i32 26 |
| %x27 = extractelement <32 x i8> %0, i32 27 |
| %x28 = extractelement <32 x i8> %0, i32 28 |
| %x29 = extractelement <32 x i8> %0, i32 29 |
| %x30 = extractelement <32 x i8> %0, i32 30 |
| %x31 = extractelement <32 x i8> %0, i32 31 |
| %trunc0 = trunc i8 %x0 to i4 |
| %trunc1 = trunc i8 %x1 to i4 |
| %trunc2 = trunc i8 %x2 to i4 |
| %trunc3 = trunc i8 %x3 to i4 |
| %trunc4 = trunc i8 %x4 to i4 |
| %trunc5 = trunc i8 %x5 to i4 |
| %trunc6 = trunc i8 %x6 to i4 |
| %trunc7 = trunc i8 %x7 to i4 |
| %trunc8 = trunc i8 %x8 to i4 |
| %trunc9 = trunc i8 %x9 to i4 |
| %trunc10 = trunc i8 %x10 to i4 |
| %trunc11 = trunc i8 %x11 to i4 |
| %trunc12 = trunc i8 %x12 to i4 |
| %trunc13 = trunc i8 %x13 to i4 |
| %trunc14 = trunc i8 %x14 to i4 |
| %trunc15 = trunc i8 %x15 to i4 |
| %trunc16 = trunc i8 %x16 to i4 |
| %trunc17 = trunc i8 %x17 to i4 |
| %trunc18 = trunc i8 %x18 to i4 |
| %trunc19 = trunc i8 %x19 to i4 |
| %trunc20 = trunc i8 %x20 to i4 |
| %trunc21 = trunc i8 %x21 to i4 |
| %trunc22 = trunc i8 %x22 to i4 |
| %trunc23 = trunc i8 %x23 to i4 |
| %trunc24 = trunc i8 %x24 to i4 |
| %trunc25 = trunc i8 %x25 to i4 |
| %trunc26 = trunc i8 %x26 to i4 |
| %trunc27 = trunc i8 %x27 to i4 |
| %trunc28 = trunc i8 %x28 to i4 |
| %trunc29 = trunc i8 %x29 to i4 |
| %trunc30 = trunc i8 %x30 to i4 |
| %trunc31 = trunc i8 %x31 to i4 |
| %ext0 = zext i4 %trunc0 to i8 |
| %ext1 = zext i4 %trunc1 to i8 |
| %ext2 = zext i4 %trunc2 to i8 |
| %ext3 = zext i4 %trunc3 to i8 |
| %ext4 = zext i4 %trunc4 to i8 |
| %ext5 = zext i4 %trunc5 to i8 |
| %ext6 = zext i4 %trunc6 to i8 |
| %ext7 = zext i4 %trunc7 to i8 |
| %ext8 = zext i4 %trunc8 to i8 |
| %ext9 = zext i4 %trunc9 to i8 |
| %ext10 = zext i4 %trunc10 to i8 |
| %ext11 = zext i4 %trunc11 to i8 |
| %ext12 = zext i4 %trunc12 to i8 |
| %ext13 = zext i4 %trunc13 to i8 |
| %ext14 = zext i4 %trunc14 to i8 |
| %ext15 = zext i4 %trunc15 to i8 |
| %ext16 = zext i4 %trunc16 to i8 |
| %ext17 = zext i4 %trunc17 to i8 |
| %ext18 = zext i4 %trunc18 to i8 |
| %ext19 = zext i4 %trunc19 to i8 |
| %ext20 = zext i4 %trunc20 to i8 |
| %ext21 = zext i4 %trunc21 to i8 |
| %ext22 = zext i4 %trunc22 to i8 |
| %ext23 = zext i4 %trunc23 to i8 |
| %ext24 = zext i4 %trunc24 to i8 |
| %ext25 = zext i4 %trunc25 to i8 |
| %ext26 = zext i4 %trunc26 to i8 |
| %ext27 = zext i4 %trunc27 to i8 |
| %ext28 = zext i4 %trunc28 to i8 |
| %ext29 = zext i4 %trunc29 to i8 |
| %ext30 = zext i4 %trunc30 to i8 |
| %ext31 = zext i4 %trunc31 to i8 |
| %v0 = insertelement <32 x i8> undef, i8 %ext0, i32 0 |
| %v1 = insertelement <32 x i8> %v0, i8 %ext1, i32 1 |
| %v2 = insertelement <32 x i8> %v1, i8 %ext2, i32 2 |
| %v3 = insertelement <32 x i8> %v2, i8 %ext3, i32 3 |
| %v4 = insertelement <32 x i8> %v3, i8 %ext4, i32 4 |
| %v5 = insertelement <32 x i8> %v4, i8 %ext5, i32 5 |
| %v6 = insertelement <32 x i8> %v5, i8 %ext6, i32 6 |
| %v7 = insertelement <32 x i8> %v6, i8 %ext7, i32 7 |
| %v8 = insertelement <32 x i8> %v7, i8 %ext8, i32 8 |
| %v9 = insertelement <32 x i8> %v8, i8 %ext9, i32 9 |
| %v10 = insertelement <32 x i8> %v9, i8 %ext10, i32 10 |
| %v11 = insertelement <32 x i8> %v10, i8 %ext11, i32 11 |
| %v12 = insertelement <32 x i8> %v11, i8 %ext12, i32 12 |
| %v13 = insertelement <32 x i8> %v12, i8 %ext13, i32 13 |
| %v14 = insertelement <32 x i8> %v13, i8 %ext14, i32 14 |
| %v15 = insertelement <32 x i8> %v14, i8 %ext15, i32 15 |
| %v16 = insertelement <32 x i8> %v15, i8 %ext16, i32 16 |
| %v17 = insertelement <32 x i8> %v16, i8 %ext17, i32 17 |
| %v18 = insertelement <32 x i8> %v17, i8 %ext18, i32 18 |
| %v19 = insertelement <32 x i8> %v18, i8 %ext19, i32 19 |
| %v20 = insertelement <32 x i8> %v19, i8 %ext20, i32 20 |
| %v21 = insertelement <32 x i8> %v20, i8 %ext21, i32 21 |
| %v22 = insertelement <32 x i8> %v21, i8 %ext22, i32 22 |
| %v23 = insertelement <32 x i8> %v22, i8 %ext23, i32 23 |
| %v24 = insertelement <32 x i8> %v23, i8 %ext24, i32 24 |
| %v25 = insertelement <32 x i8> %v24, i8 %ext25, i32 25 |
| %v26 = insertelement <32 x i8> %v25, i8 %ext26, i32 26 |
| %v27 = insertelement <32 x i8> %v26, i8 %ext27, i32 27 |
| %v28 = insertelement <32 x i8> %v27, i8 %ext28, i32 28 |
| %v29 = insertelement <32 x i8> %v28, i8 %ext29, i32 29 |
| %v30 = insertelement <32 x i8> %v29, i8 %ext30, i32 30 |
| %v31 = insertelement <32 x i8> %v30, i8 %ext31, i32 31 |
| ret <32 x i8> %v31 |
| } |
| |
| define <2 x i64> @_clearupper2xi64b(<2 x i64>) nounwind { |
| ; SSE2-LABEL: _clearupper2xi64b: |
| ; SSE2: # %bb.0: |
| ; SSE2-NEXT: andps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0 |
| ; SSE2-NEXT: retq |
| ; |
| ; SSE42-LABEL: _clearupper2xi64b: |
| ; SSE42: # %bb.0: |
| ; SSE42-NEXT: xorps %xmm1, %xmm1 |
| ; SSE42-NEXT: blendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2],xmm1[3] |
| ; SSE42-NEXT: retq |
| ; |
| ; AVX-LABEL: _clearupper2xi64b: |
| ; AVX: # %bb.0: |
| ; AVX-NEXT: vxorps %xmm1, %xmm1, %xmm1 |
| ; AVX-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2],xmm1[3] |
| ; AVX-NEXT: retq |
| %x32 = bitcast <2 x i64> %0 to <4 x i32> |
| %r0 = insertelement <4 x i32> %x32, i32 zeroinitializer, i32 1 |
| %r1 = insertelement <4 x i32> %r0, i32 zeroinitializer, i32 3 |
| %r = bitcast <4 x i32> %r1 to <2 x i64> |
| ret <2 x i64> %r |
| } |
| |
| define <4 x i64> @_clearupper4xi64b(<4 x i64>) nounwind { |
| ; SSE2-LABEL: _clearupper4xi64b: |
| ; SSE2: # %bb.0: |
| ; SSE2-NEXT: movaps {{.*#+}} xmm2 = [NaN,0.0E+0,NaN,0.0E+0] |
| ; SSE2-NEXT: andps %xmm2, %xmm0 |
| ; SSE2-NEXT: andps %xmm2, %xmm1 |
| ; SSE2-NEXT: retq |
| ; |
| ; SSE42-LABEL: _clearupper4xi64b: |
| ; SSE42: # %bb.0: |
| ; SSE42-NEXT: xorps %xmm2, %xmm2 |
| ; SSE42-NEXT: blendps {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2],xmm2[3] |
| ; SSE42-NEXT: blendps {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2],xmm2[3] |
| ; SSE42-NEXT: retq |
| ; |
| ; AVX-LABEL: _clearupper4xi64b: |
| ; AVX: # %bb.0: |
| ; AVX-NEXT: vxorps %xmm1, %xmm1, %xmm1 |
| ; AVX-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7] |
| ; AVX-NEXT: retq |
| %x32 = bitcast <4 x i64> %0 to <8 x i32> |
| %r0 = insertelement <8 x i32> %x32, i32 zeroinitializer, i32 1 |
| %r1 = insertelement <8 x i32> %r0, i32 zeroinitializer, i32 3 |
| %r2 = insertelement <8 x i32> %r1, i32 zeroinitializer, i32 5 |
| %r3 = insertelement <8 x i32> %r2, i32 zeroinitializer, i32 7 |
| %r = bitcast <8 x i32> %r3 to <4 x i64> |
| ret <4 x i64> %r |
| } |
| |
| define <4 x i32> @_clearupper4xi32b(<4 x i32>) nounwind { |
| ; SSE2-LABEL: _clearupper4xi32b: |
| ; SSE2: # %bb.0: |
| ; SSE2-NEXT: andps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0 |
| ; SSE2-NEXT: retq |
| ; |
| ; SSE42-LABEL: _clearupper4xi32b: |
| ; SSE42: # %bb.0: |
| ; SSE42-NEXT: pxor %xmm1, %xmm1 |
| ; SSE42-NEXT: pblendw {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2],xmm1[3],xmm0[4],xmm1[5],xmm0[6],xmm1[7] |
| ; SSE42-NEXT: retq |
| ; |
| ; AVX-LABEL: _clearupper4xi32b: |
| ; AVX: # %bb.0: |
| ; AVX-NEXT: vpxor %xmm1, %xmm1, %xmm1 |
| ; AVX-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2],xmm1[3],xmm0[4],xmm1[5],xmm0[6],xmm1[7] |
| ; AVX-NEXT: retq |
| %x16 = bitcast <4 x i32> %0 to <8 x i16> |
| %r0 = insertelement <8 x i16> %x16, i16 zeroinitializer, i32 1 |
| %r1 = insertelement <8 x i16> %r0, i16 zeroinitializer, i32 3 |
| %r2 = insertelement <8 x i16> %r1, i16 zeroinitializer, i32 5 |
| %r3 = insertelement <8 x i16> %r2, i16 zeroinitializer, i32 7 |
| %r = bitcast <8 x i16> %r3 to <4 x i32> |
| ret <4 x i32> %r |
| } |
| |
| define <8 x i32> @_clearupper8xi32b(<8 x i32>) nounwind { |
| ; SSE2-LABEL: _clearupper8xi32b: |
| ; SSE2: # %bb.0: |
| ; SSE2-NEXT: movaps {{.*#+}} xmm2 = [65535,0,65535,0,65535,0,65535,0] |
| ; SSE2-NEXT: andps %xmm2, %xmm0 |
| ; SSE2-NEXT: andps %xmm2, %xmm1 |
| ; SSE2-NEXT: retq |
| ; |
| ; SSE42-LABEL: _clearupper8xi32b: |
| ; SSE42: # %bb.0: |
| ; SSE42-NEXT: pxor %xmm2, %xmm2 |
| ; SSE42-NEXT: pblendw {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2],xmm2[3],xmm0[4],xmm2[5],xmm0[6],xmm2[7] |
| ; SSE42-NEXT: pblendw {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2],xmm2[3],xmm1[4],xmm2[5],xmm1[6],xmm2[7] |
| ; SSE42-NEXT: retq |
| ; |
| ; AVX1-LABEL: _clearupper8xi32b: |
| ; AVX1: # %bb.0: |
| ; AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm0 |
| ; AVX1-NEXT: retq |
| ; |
| ; AVX2-LABEL: _clearupper8xi32b: |
| ; AVX2: # %bb.0: |
| ; AVX2-NEXT: vpxor %xmm1, %xmm1, %xmm1 |
| ; AVX2-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7],ymm0[8],ymm1[9],ymm0[10],ymm1[11],ymm0[12],ymm1[13],ymm0[14],ymm1[15] |
| ; AVX2-NEXT: retq |
| %x16 = bitcast <8 x i32> %0 to <16 x i16> |
| %r0 = insertelement <16 x i16> %x16, i16 zeroinitializer, i32 1 |
| %r1 = insertelement <16 x i16> %r0, i16 zeroinitializer, i32 3 |
| %r2 = insertelement <16 x i16> %r1, i16 zeroinitializer, i32 5 |
| %r3 = insertelement <16 x i16> %r2, i16 zeroinitializer, i32 7 |
| %r4 = insertelement <16 x i16> %r3, i16 zeroinitializer, i32 9 |
| %r5 = insertelement <16 x i16> %r4, i16 zeroinitializer, i32 11 |
| %r6 = insertelement <16 x i16> %r5, i16 zeroinitializer, i32 13 |
| %r7 = insertelement <16 x i16> %r6, i16 zeroinitializer, i32 15 |
| %r = bitcast <16 x i16> %r7 to <8 x i32> |
| ret <8 x i32> %r |
| } |
| |
| define <8 x i16> @_clearupper8xi16b(<8 x i16>) nounwind { |
| ; SSE-LABEL: _clearupper8xi16b: |
| ; SSE: # %bb.0: |
| ; SSE-NEXT: andps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0 |
| ; SSE-NEXT: retq |
| ; |
| ; AVX-LABEL: _clearupper8xi16b: |
| ; AVX: # %bb.0: |
| ; AVX-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0 |
| ; AVX-NEXT: retq |
| %x8 = bitcast <8 x i16> %0 to <16 x i8> |
| %r0 = insertelement <16 x i8> %x8, i8 zeroinitializer, i32 1 |
| %r1 = insertelement <16 x i8> %r0, i8 zeroinitializer, i32 3 |
| %r2 = insertelement <16 x i8> %r1, i8 zeroinitializer, i32 5 |
| %r3 = insertelement <16 x i8> %r2, i8 zeroinitializer, i32 7 |
| %r4 = insertelement <16 x i8> %r3, i8 zeroinitializer, i32 9 |
| %r5 = insertelement <16 x i8> %r4, i8 zeroinitializer, i32 11 |
| %r6 = insertelement <16 x i8> %r5, i8 zeroinitializer, i32 13 |
| %r7 = insertelement <16 x i8> %r6, i8 zeroinitializer, i32 15 |
| %r = bitcast <16 x i8> %r7 to <8 x i16> |
| ret <8 x i16> %r |
| } |
| |
| define <16 x i16> @_clearupper16xi16b(<16 x i16>) nounwind { |
| ; SSE-LABEL: _clearupper16xi16b: |
| ; SSE: # %bb.0: |
| ; SSE-NEXT: movaps {{.*#+}} xmm2 = [255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0] |
| ; SSE-NEXT: andps %xmm2, %xmm0 |
| ; SSE-NEXT: andps %xmm2, %xmm1 |
| ; SSE-NEXT: retq |
| ; |
| ; AVX-LABEL: _clearupper16xi16b: |
| ; AVX: # %bb.0: |
| ; AVX-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm0 |
| ; AVX-NEXT: retq |
| %x8 = bitcast <16 x i16> %0 to <32 x i8> |
| %r0 = insertelement <32 x i8> %x8, i8 zeroinitializer, i32 1 |
| %r1 = insertelement <32 x i8> %r0, i8 zeroinitializer, i32 3 |
| %r2 = insertelement <32 x i8> %r1, i8 zeroinitializer, i32 5 |
| %r3 = insertelement <32 x i8> %r2, i8 zeroinitializer, i32 7 |
| %r4 = insertelement <32 x i8> %r3, i8 zeroinitializer, i32 9 |
| %r5 = insertelement <32 x i8> %r4, i8 zeroinitializer, i32 11 |
| %r6 = insertelement <32 x i8> %r5, i8 zeroinitializer, i32 13 |
| %r7 = insertelement <32 x i8> %r6, i8 zeroinitializer, i32 15 |
| %r8 = insertelement <32 x i8> %r7, i8 zeroinitializer, i32 17 |
| %r9 = insertelement <32 x i8> %r8, i8 zeroinitializer, i32 19 |
| %r10 = insertelement <32 x i8> %r9, i8 zeroinitializer, i32 21 |
| %r11 = insertelement <32 x i8> %r10, i8 zeroinitializer, i32 23 |
| %r12 = insertelement <32 x i8> %r11, i8 zeroinitializer, i32 25 |
| %r13 = insertelement <32 x i8> %r12, i8 zeroinitializer, i32 27 |
| %r14 = insertelement <32 x i8> %r13, i8 zeroinitializer, i32 29 |
| %r15 = insertelement <32 x i8> %r14, i8 zeroinitializer, i32 31 |
| %r = bitcast <32 x i8> %r15 to <16 x i16> |
| ret <16 x i16> %r |
| } |
| |
| define <16 x i8> @_clearupper16xi8b(<16 x i8>) nounwind { |
| ; SSE2-LABEL: _clearupper16xi8b: |
| ; SSE2: # %bb.0: |
| ; SSE2-NEXT: pushq %rbx |
| ; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm0[2,3,2,3] |
| ; SSE2-NEXT: movq %xmm1, %r10 |
| ; SSE2-NEXT: movq %r10, %r8 |
| ; SSE2-NEXT: shrq $56, %r8 |
| ; SSE2-NEXT: andl $15, %r8d |
| ; SSE2-NEXT: movq %r10, %r9 |
| ; SSE2-NEXT: shrq $48, %r9 |
| ; SSE2-NEXT: andl $15, %r9d |
| ; SSE2-NEXT: movq %r10, %rsi |
| ; SSE2-NEXT: shrq $40, %rsi |
| ; SSE2-NEXT: andl $15, %esi |
| ; SSE2-NEXT: movq %r10, %r11 |
| ; SSE2-NEXT: shrq $32, %r11 |
| ; SSE2-NEXT: andl $15, %r11d |
| ; SSE2-NEXT: movq %xmm0, %rax |
| ; SSE2-NEXT: movq %rax, %rdx |
| ; SSE2-NEXT: shrq $56, %rdx |
| ; SSE2-NEXT: andl $15, %edx |
| ; SSE2-NEXT: movq %rax, %rcx |
| ; SSE2-NEXT: shrq $48, %rcx |
| ; SSE2-NEXT: andl $15, %ecx |
| ; SSE2-NEXT: movq %rax, %rdi |
| ; SSE2-NEXT: shrq $40, %rdi |
| ; SSE2-NEXT: andl $15, %edi |
| ; SSE2-NEXT: movq %rax, %rbx |
| ; SSE2-NEXT: shrq $32, %rbx |
| ; SSE2-NEXT: andl $15, %ebx |
| ; SSE2-NEXT: shlq $32, %rbx |
| ; SSE2-NEXT: andl $252645135, %eax # imm = 0xF0F0F0F |
| ; SSE2-NEXT: orq %rbx, %rax |
| ; SSE2-NEXT: shlq $40, %rdi |
| ; SSE2-NEXT: orq %rax, %rdi |
| ; SSE2-NEXT: shlq $48, %rcx |
| ; SSE2-NEXT: orq %rdi, %rcx |
| ; SSE2-NEXT: shlq $56, %rdx |
| ; SSE2-NEXT: orq %rcx, %rdx |
| ; SSE2-NEXT: shlq $32, %r11 |
| ; SSE2-NEXT: andl $252645135, %r10d # imm = 0xF0F0F0F |
| ; SSE2-NEXT: orq %r11, %r10 |
| ; SSE2-NEXT: shlq $40, %rsi |
| ; SSE2-NEXT: orq %r10, %rsi |
| ; SSE2-NEXT: shlq $48, %r9 |
| ; SSE2-NEXT: orq %rsi, %r9 |
| ; SSE2-NEXT: shlq $56, %r8 |
| ; SSE2-NEXT: orq %r9, %r8 |
| ; SSE2-NEXT: movq %rdx, %xmm0 |
| ; SSE2-NEXT: movq %r8, %xmm1 |
| ; SSE2-NEXT: punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0] |
| ; SSE2-NEXT: popq %rbx |
| ; SSE2-NEXT: retq |
| ; |
| ; SSE42-LABEL: _clearupper16xi8b: |
| ; SSE42: # %bb.0: |
| ; SSE42-NEXT: pushq %rbx |
| ; SSE42-NEXT: pextrq $1, %xmm0, %r10 |
| ; SSE42-NEXT: movq %r10, %r8 |
| ; SSE42-NEXT: shrq $56, %r8 |
| ; SSE42-NEXT: andl $15, %r8d |
| ; SSE42-NEXT: movq %r10, %r9 |
| ; SSE42-NEXT: shrq $48, %r9 |
| ; SSE42-NEXT: andl $15, %r9d |
| ; SSE42-NEXT: movq %r10, %rsi |
| ; SSE42-NEXT: shrq $40, %rsi |
| ; SSE42-NEXT: andl $15, %esi |
| ; SSE42-NEXT: movq %r10, %r11 |
| ; SSE42-NEXT: shrq $32, %r11 |
| ; SSE42-NEXT: andl $15, %r11d |
| ; SSE42-NEXT: movq %xmm0, %rax |
| ; SSE42-NEXT: movq %rax, %rdx |
| ; SSE42-NEXT: shrq $56, %rdx |
| ; SSE42-NEXT: andl $15, %edx |
| ; SSE42-NEXT: movq %rax, %rcx |
| ; SSE42-NEXT: shrq $48, %rcx |
| ; SSE42-NEXT: andl $15, %ecx |
| ; SSE42-NEXT: movq %rax, %rdi |
| ; SSE42-NEXT: shrq $40, %rdi |
| ; SSE42-NEXT: andl $15, %edi |
| ; SSE42-NEXT: movq %rax, %rbx |
| ; SSE42-NEXT: shrq $32, %rbx |
| ; SSE42-NEXT: andl $15, %ebx |
| ; SSE42-NEXT: shlq $32, %rbx |
| ; SSE42-NEXT: andl $252645135, %eax # imm = 0xF0F0F0F |
| ; SSE42-NEXT: orq %rbx, %rax |
| ; SSE42-NEXT: shlq $40, %rdi |
| ; SSE42-NEXT: orq %rax, %rdi |
| ; SSE42-NEXT: shlq $48, %rcx |
| ; SSE42-NEXT: orq %rdi, %rcx |
| ; SSE42-NEXT: shlq $56, %rdx |
| ; SSE42-NEXT: orq %rcx, %rdx |
| ; SSE42-NEXT: shlq $32, %r11 |
| ; SSE42-NEXT: andl $252645135, %r10d # imm = 0xF0F0F0F |
| ; SSE42-NEXT: orq %r11, %r10 |
| ; SSE42-NEXT: shlq $40, %rsi |
| ; SSE42-NEXT: orq %r10, %rsi |
| ; SSE42-NEXT: shlq $48, %r9 |
| ; SSE42-NEXT: orq %rsi, %r9 |
| ; SSE42-NEXT: shlq $56, %r8 |
| ; SSE42-NEXT: orq %r9, %r8 |
| ; SSE42-NEXT: movq %r8, %xmm1 |
| ; SSE42-NEXT: movq %rdx, %xmm0 |
| ; SSE42-NEXT: punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0] |
| ; SSE42-NEXT: popq %rbx |
| ; SSE42-NEXT: retq |
| ; |
| ; AVX-LABEL: _clearupper16xi8b: |
| ; AVX: # %bb.0: |
| ; AVX-NEXT: pushq %rbx |
| ; AVX-NEXT: vmovaps %xmm0, -{{[0-9]+}}(%rsp) |
| ; AVX-NEXT: movq -{{[0-9]+}}(%rsp), %r9 |
| ; AVX-NEXT: movq -{{[0-9]+}}(%rsp), %rcx |
| ; AVX-NEXT: movq %r9, %r8 |
| ; AVX-NEXT: shrq $56, %r8 |
| ; AVX-NEXT: andl $15, %r8d |
| ; AVX-NEXT: movq %r9, %r10 |
| ; AVX-NEXT: shrq $48, %r10 |
| ; AVX-NEXT: andl $15, %r10d |
| ; AVX-NEXT: movq %rcx, %rdx |
| ; AVX-NEXT: shldq $24, %r9, %rdx |
| ; AVX-NEXT: andl $15, %edx |
| ; AVX-NEXT: movq %r9, %r11 |
| ; AVX-NEXT: shrq $32, %r11 |
| ; AVX-NEXT: andl $15, %r11d |
| ; AVX-NEXT: movq %rcx, %rdi |
| ; AVX-NEXT: shrq $56, %rdi |
| ; AVX-NEXT: andl $15, %edi |
| ; AVX-NEXT: movq %rcx, %rsi |
| ; AVX-NEXT: shrq $48, %rsi |
| ; AVX-NEXT: andl $15, %esi |
| ; AVX-NEXT: movq %rcx, %rax |
| ; AVX-NEXT: shrq $40, %rax |
| ; AVX-NEXT: andl $15, %eax |
| ; AVX-NEXT: movq %rcx, %rbx |
| ; AVX-NEXT: shrq $32, %rbx |
| ; AVX-NEXT: andl $15, %ebx |
| ; AVX-NEXT: shlq $32, %rbx |
| ; AVX-NEXT: andl $252645135, %ecx # imm = 0xF0F0F0F |
| ; AVX-NEXT: orq %rbx, %rcx |
| ; AVX-NEXT: shlq $40, %rax |
| ; AVX-NEXT: orq %rcx, %rax |
| ; AVX-NEXT: shlq $48, %rsi |
| ; AVX-NEXT: orq %rax, %rsi |
| ; AVX-NEXT: shlq $56, %rdi |
| ; AVX-NEXT: orq %rsi, %rdi |
| ; AVX-NEXT: movq %rdi, -{{[0-9]+}}(%rsp) |
| ; AVX-NEXT: shlq $32, %r11 |
| ; AVX-NEXT: andl $252645135, %r9d # imm = 0xF0F0F0F |
| ; AVX-NEXT: orq %r11, %r9 |
| ; AVX-NEXT: shlq $40, %rdx |
| ; AVX-NEXT: orq %r9, %rdx |
| ; AVX-NEXT: shlq $48, %r10 |
| ; AVX-NEXT: orq %rdx, %r10 |
| ; AVX-NEXT: shlq $56, %r8 |
| ; AVX-NEXT: orq %r10, %r8 |
| ; AVX-NEXT: movq %r8, -{{[0-9]+}}(%rsp) |
| ; AVX-NEXT: vmovaps -{{[0-9]+}}(%rsp), %xmm0 |
| ; AVX-NEXT: popq %rbx |
| ; AVX-NEXT: retq |
| %x4 = bitcast <16 x i8> %0 to <32 x i4> |
| %r0 = insertelement <32 x i4> %x4, i4 zeroinitializer, i32 1 |
| %r1 = insertelement <32 x i4> %r0, i4 zeroinitializer, i32 3 |
| %r2 = insertelement <32 x i4> %r1, i4 zeroinitializer, i32 5 |
| %r3 = insertelement <32 x i4> %r2, i4 zeroinitializer, i32 7 |
| %r4 = insertelement <32 x i4> %r3, i4 zeroinitializer, i32 9 |
| %r5 = insertelement <32 x i4> %r4, i4 zeroinitializer, i32 11 |
| %r6 = insertelement <32 x i4> %r5, i4 zeroinitializer, i32 13 |
| %r7 = insertelement <32 x i4> %r6, i4 zeroinitializer, i32 15 |
| %r8 = insertelement <32 x i4> %r7, i4 zeroinitializer, i32 17 |
| %r9 = insertelement <32 x i4> %r8, i4 zeroinitializer, i32 19 |
| %r10 = insertelement <32 x i4> %r9, i4 zeroinitializer, i32 21 |
| %r11 = insertelement <32 x i4> %r10, i4 zeroinitializer, i32 23 |
| %r12 = insertelement <32 x i4> %r11, i4 zeroinitializer, i32 25 |
| %r13 = insertelement <32 x i4> %r12, i4 zeroinitializer, i32 27 |
| %r14 = insertelement <32 x i4> %r13, i4 zeroinitializer, i32 29 |
| %r15 = insertelement <32 x i4> %r14, i4 zeroinitializer, i32 31 |
| %r = bitcast <32 x i4> %r15 to <16 x i8> |
| ret <16 x i8> %r |
| } |
| |
| define <32 x i8> @_clearupper32xi8b(<32 x i8>) nounwind { |
| ; SSE2-LABEL: _clearupper32xi8b: |
| ; SSE2: # %bb.0: |
| ; SSE2-NEXT: pushq %rbx |
| ; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm0[2,3,2,3] |
| ; SSE2-NEXT: movq %xmm2, %r10 |
| ; SSE2-NEXT: movq %r10, %r8 |
| ; SSE2-NEXT: shrq $56, %r8 |
| ; SSE2-NEXT: andl $15, %r8d |
| ; SSE2-NEXT: movq %r10, %r9 |
| ; SSE2-NEXT: shrq $48, %r9 |
| ; SSE2-NEXT: andl $15, %r9d |
| ; SSE2-NEXT: movq %r10, %rsi |
| ; SSE2-NEXT: shrq $40, %rsi |
| ; SSE2-NEXT: andl $15, %esi |
| ; SSE2-NEXT: movq %r10, %r11 |
| ; SSE2-NEXT: shrq $32, %r11 |
| ; SSE2-NEXT: andl $15, %r11d |
| ; SSE2-NEXT: movq %xmm0, %rax |
| ; SSE2-NEXT: movq %rax, %rdx |
| ; SSE2-NEXT: shrq $56, %rdx |
| ; SSE2-NEXT: andl $15, %edx |
| ; SSE2-NEXT: movq %rax, %rcx |
| ; SSE2-NEXT: shrq $48, %rcx |
| ; SSE2-NEXT: andl $15, %ecx |
| ; SSE2-NEXT: movq %rax, %rdi |
| ; SSE2-NEXT: shrq $40, %rdi |
| ; SSE2-NEXT: andl $15, %edi |
| ; SSE2-NEXT: movq %rax, %rbx |
| ; SSE2-NEXT: shrq $32, %rbx |
| ; SSE2-NEXT: andl $15, %ebx |
| ; SSE2-NEXT: shlq $32, %rbx |
| ; SSE2-NEXT: andl $252645135, %eax # imm = 0xF0F0F0F |
| ; SSE2-NEXT: orq %rbx, %rax |
| ; SSE2-NEXT: shlq $40, %rdi |
| ; SSE2-NEXT: orq %rax, %rdi |
| ; SSE2-NEXT: shlq $48, %rcx |
| ; SSE2-NEXT: orq %rdi, %rcx |
| ; SSE2-NEXT: shlq $56, %rdx |
| ; SSE2-NEXT: orq %rcx, %rdx |
| ; SSE2-NEXT: shlq $32, %r11 |
| ; SSE2-NEXT: andl $252645135, %r10d # imm = 0xF0F0F0F |
| ; SSE2-NEXT: orq %r11, %r10 |
| ; SSE2-NEXT: shlq $40, %rsi |
| ; SSE2-NEXT: orq %r10, %rsi |
| ; SSE2-NEXT: shlq $48, %r9 |
| ; SSE2-NEXT: orq %rsi, %r9 |
| ; SSE2-NEXT: shlq $56, %r8 |
| ; SSE2-NEXT: orq %r9, %r8 |
| ; SSE2-NEXT: movq %rdx, %xmm0 |
| ; SSE2-NEXT: movq %r8, %xmm2 |
| ; SSE2-NEXT: punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm2[0] |
| ; SSE2-NEXT: popq %rbx |
| ; SSE2-NEXT: retq |
| ; |
| ; SSE42-LABEL: _clearupper32xi8b: |
| ; SSE42: # %bb.0: |
| ; SSE42-NEXT: pushq %rbx |
| ; SSE42-NEXT: pextrq $1, %xmm0, %r10 |
| ; SSE42-NEXT: movq %r10, %r8 |
| ; SSE42-NEXT: shrq $56, %r8 |
| ; SSE42-NEXT: andl $15, %r8d |
| ; SSE42-NEXT: movq %r10, %r9 |
| ; SSE42-NEXT: shrq $48, %r9 |
| ; SSE42-NEXT: andl $15, %r9d |
| ; SSE42-NEXT: movq %r10, %rsi |
| ; SSE42-NEXT: shrq $40, %rsi |
| ; SSE42-NEXT: andl $15, %esi |
| ; SSE42-NEXT: movq %r10, %r11 |
| ; SSE42-NEXT: shrq $32, %r11 |
| ; SSE42-NEXT: andl $15, %r11d |
| ; SSE42-NEXT: movq %xmm0, %rax |
| ; SSE42-NEXT: movq %rax, %rdx |
| ; SSE42-NEXT: shrq $56, %rdx |
| ; SSE42-NEXT: andl $15, %edx |
| ; SSE42-NEXT: movq %rax, %rcx |
| ; SSE42-NEXT: shrq $48, %rcx |
| ; SSE42-NEXT: andl $15, %ecx |
| ; SSE42-NEXT: movq %rax, %rdi |
| ; SSE42-NEXT: shrq $40, %rdi |
| ; SSE42-NEXT: andl $15, %edi |
| ; SSE42-NEXT: movq %rax, %rbx |
| ; SSE42-NEXT: shrq $32, %rbx |
| ; SSE42-NEXT: andl $15, %ebx |
| ; SSE42-NEXT: shlq $32, %rbx |
| ; SSE42-NEXT: andl $252645135, %eax # imm = 0xF0F0F0F |
| ; SSE42-NEXT: orq %rbx, %rax |
| ; SSE42-NEXT: shlq $40, %rdi |
| ; SSE42-NEXT: orq %rax, %rdi |
| ; SSE42-NEXT: shlq $48, %rcx |
| ; SSE42-NEXT: orq %rdi, %rcx |
| ; SSE42-NEXT: shlq $56, %rdx |
| ; SSE42-NEXT: orq %rcx, %rdx |
| ; SSE42-NEXT: shlq $32, %r11 |
| ; SSE42-NEXT: andl $252645135, %r10d # imm = 0xF0F0F0F |
| ; SSE42-NEXT: orq %r11, %r10 |
| ; SSE42-NEXT: shlq $40, %rsi |
| ; SSE42-NEXT: orq %r10, %rsi |
| ; SSE42-NEXT: shlq $48, %r9 |
| ; SSE42-NEXT: orq %rsi, %r9 |
| ; SSE42-NEXT: shlq $56, %r8 |
| ; SSE42-NEXT: orq %r9, %r8 |
| ; SSE42-NEXT: movq %r8, %xmm2 |
| ; SSE42-NEXT: movq %rdx, %xmm0 |
| ; SSE42-NEXT: punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm2[0] |
| ; SSE42-NEXT: popq %rbx |
| ; SSE42-NEXT: retq |
| ; |
| ; AVX1-LABEL: _clearupper32xi8b: |
| ; AVX1: # %bb.0: |
| ; AVX1-NEXT: vmovaps %xmm0, -{{[0-9]+}}(%rsp) |
| ; AVX1-NEXT: movq -{{[0-9]+}}(%rsp), %rax |
| ; AVX1-NEXT: movq %rax, %r8 |
| ; AVX1-NEXT: movq %rax, %rdx |
| ; AVX1-NEXT: movq %rax, %rsi |
| ; AVX1-NEXT: movq %rax, %rdi |
| ; AVX1-NEXT: movq %rax, %rcx |
| ; AVX1-NEXT: shrq $32, %rcx |
| ; AVX1-NEXT: andl $15, %ecx |
| ; AVX1-NEXT: shlq $32, %rcx |
| ; AVX1-NEXT: andl $252645135, %eax # imm = 0xF0F0F0F |
| ; AVX1-NEXT: orq %rcx, %rax |
| ; AVX1-NEXT: shrq $40, %rdi |
| ; AVX1-NEXT: andl $15, %edi |
| ; AVX1-NEXT: shlq $40, %rdi |
| ; AVX1-NEXT: orq %rax, %rdi |
| ; AVX1-NEXT: movq -{{[0-9]+}}(%rsp), %rax |
| ; AVX1-NEXT: shrq $48, %rsi |
| ; AVX1-NEXT: andl $15, %esi |
| ; AVX1-NEXT: shlq $48, %rsi |
| ; AVX1-NEXT: orq %rdi, %rsi |
| ; AVX1-NEXT: movq %rax, %rcx |
| ; AVX1-NEXT: shrq $56, %rdx |
| ; AVX1-NEXT: andl $15, %edx |
| ; AVX1-NEXT: shlq $56, %rdx |
| ; AVX1-NEXT: orq %rsi, %rdx |
| ; AVX1-NEXT: movq %rax, %rsi |
| ; AVX1-NEXT: shldq $24, %rax, %r8 |
| ; AVX1-NEXT: movq %rdx, -{{[0-9]+}}(%rsp) |
| ; AVX1-NEXT: movq %rax, %rdx |
| ; AVX1-NEXT: shrq $32, %rdx |
| ; AVX1-NEXT: andl $15, %edx |
| ; AVX1-NEXT: shlq $32, %rdx |
| ; AVX1-NEXT: andl $252645135, %eax # imm = 0xF0F0F0F |
| ; AVX1-NEXT: orq %rdx, %rax |
| ; AVX1-NEXT: andl $15, %r8d |
| ; AVX1-NEXT: shlq $40, %r8 |
| ; AVX1-NEXT: orq %rax, %r8 |
| ; AVX1-NEXT: shrq $48, %rsi |
| ; AVX1-NEXT: andl $15, %esi |
| ; AVX1-NEXT: shlq $48, %rsi |
| ; AVX1-NEXT: orq %r8, %rsi |
| ; AVX1-NEXT: shrq $56, %rcx |
| ; AVX1-NEXT: andl $15, %ecx |
| ; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm0 |
| ; AVX1-NEXT: shlq $56, %rcx |
| ; AVX1-NEXT: orq %rsi, %rcx |
| ; AVX1-NEXT: vmovq %xmm0, %rax |
| ; AVX1-NEXT: movq %rcx, -{{[0-9]+}}(%rsp) |
| ; AVX1-NEXT: movl %eax, %ecx |
| ; AVX1-NEXT: shrl $8, %ecx |
| ; AVX1-NEXT: vmovd %eax, %xmm1 |
| ; AVX1-NEXT: vpinsrb $1, %ecx, %xmm1, %xmm1 |
| ; AVX1-NEXT: movl %eax, %ecx |
| ; AVX1-NEXT: shrl $16, %ecx |
| ; AVX1-NEXT: vpinsrb $2, %ecx, %xmm1, %xmm1 |
| ; AVX1-NEXT: movl %eax, %ecx |
| ; AVX1-NEXT: shrl $24, %ecx |
| ; AVX1-NEXT: vpinsrb $3, %ecx, %xmm1, %xmm1 |
| ; AVX1-NEXT: movq %rax, %rcx |
| ; AVX1-NEXT: shrq $32, %rcx |
| ; AVX1-NEXT: vpinsrb $4, %ecx, %xmm1, %xmm1 |
| ; AVX1-NEXT: movq %rax, %rcx |
| ; AVX1-NEXT: shrq $40, %rcx |
| ; AVX1-NEXT: vpinsrb $5, %ecx, %xmm1, %xmm1 |
| ; AVX1-NEXT: movq %rax, %rcx |
| ; AVX1-NEXT: shrq $48, %rcx |
| ; AVX1-NEXT: vpinsrb $6, %ecx, %xmm1, %xmm1 |
| ; AVX1-NEXT: vpextrq $1, %xmm0, %rcx |
| ; AVX1-NEXT: shrq $56, %rax |
| ; AVX1-NEXT: vpinsrb $7, %eax, %xmm1, %xmm0 |
| ; AVX1-NEXT: movl %ecx, %eax |
| ; AVX1-NEXT: shrl $8, %eax |
| ; AVX1-NEXT: vpinsrb $8, %ecx, %xmm0, %xmm0 |
| ; AVX1-NEXT: vpinsrb $9, %eax, %xmm0, %xmm0 |
| ; AVX1-NEXT: movl %ecx, %eax |
| ; AVX1-NEXT: shrl $16, %eax |
| ; AVX1-NEXT: vpinsrb $10, %eax, %xmm0, %xmm0 |
| ; AVX1-NEXT: movl %ecx, %eax |
| ; AVX1-NEXT: shrl $24, %eax |
| ; AVX1-NEXT: vpinsrb $11, %eax, %xmm0, %xmm0 |
| ; AVX1-NEXT: movq %rcx, %rax |
| ; AVX1-NEXT: shrq $32, %rax |
| ; AVX1-NEXT: vpinsrb $12, %eax, %xmm0, %xmm0 |
| ; AVX1-NEXT: movq %rcx, %rax |
| ; AVX1-NEXT: shrq $40, %rax |
| ; AVX1-NEXT: vpinsrb $13, %eax, %xmm0, %xmm0 |
| ; AVX1-NEXT: movq %rcx, %rax |
| ; AVX1-NEXT: shrq $48, %rax |
| ; AVX1-NEXT: vpinsrb $14, %eax, %xmm0, %xmm0 |
| ; AVX1-NEXT: shrq $56, %rcx |
| ; AVX1-NEXT: vpinsrb $15, %ecx, %xmm0, %xmm0 |
| ; AVX1-NEXT: vmovaps -{{[0-9]+}}(%rsp), %xmm1 |
| ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0 |
| ; AVX1-NEXT: retq |
| ; |
| ; AVX2-LABEL: _clearupper32xi8b: |
| ; AVX2: # %bb.0: |
| ; AVX2-NEXT: vmovdqa %xmm0, -{{[0-9]+}}(%rsp) |
| ; AVX2-NEXT: movq -{{[0-9]+}}(%rsp), %rax |
| ; AVX2-NEXT: movq %rax, %r8 |
| ; AVX2-NEXT: movq %rax, %rdx |
| ; AVX2-NEXT: movq %rax, %rsi |
| ; AVX2-NEXT: movq %rax, %rdi |
| ; AVX2-NEXT: movq %rax, %rcx |
| ; AVX2-NEXT: shrq $32, %rcx |
| ; AVX2-NEXT: andl $15, %ecx |
| ; AVX2-NEXT: shlq $32, %rcx |
| ; AVX2-NEXT: andl $252645135, %eax # imm = 0xF0F0F0F |
| ; AVX2-NEXT: orq %rcx, %rax |
| ; AVX2-NEXT: shrq $40, %rdi |
| ; AVX2-NEXT: andl $15, %edi |
| ; AVX2-NEXT: shlq $40, %rdi |
| ; AVX2-NEXT: orq %rax, %rdi |
| ; AVX2-NEXT: movq -{{[0-9]+}}(%rsp), %rax |
| ; AVX2-NEXT: shrq $48, %rsi |
| ; AVX2-NEXT: andl $15, %esi |
| ; AVX2-NEXT: shlq $48, %rsi |
| ; AVX2-NEXT: orq %rdi, %rsi |
| ; AVX2-NEXT: movq %rax, %rcx |
| ; AVX2-NEXT: shrq $56, %rdx |
| ; AVX2-NEXT: andl $15, %edx |
| ; AVX2-NEXT: shlq $56, %rdx |
| ; AVX2-NEXT: orq %rsi, %rdx |
| ; AVX2-NEXT: movq %rax, %rsi |
| ; AVX2-NEXT: shldq $24, %rax, %r8 |
| ; AVX2-NEXT: movq %rdx, -{{[0-9]+}}(%rsp) |
| ; AVX2-NEXT: movq %rax, %rdx |
| ; AVX2-NEXT: shrq $32, %rdx |
| ; AVX2-NEXT: andl $15, %edx |
| ; AVX2-NEXT: shlq $32, %rdx |
| ; AVX2-NEXT: andl $252645135, %eax # imm = 0xF0F0F0F |
| ; AVX2-NEXT: orq %rdx, %rax |
| ; AVX2-NEXT: andl $15, %r8d |
| ; AVX2-NEXT: shlq $40, %r8 |
| ; AVX2-NEXT: orq %rax, %r8 |
| ; AVX2-NEXT: shrq $48, %rsi |
| ; AVX2-NEXT: andl $15, %esi |
| ; AVX2-NEXT: shlq $48, %rsi |
| ; AVX2-NEXT: orq %r8, %rsi |
| ; AVX2-NEXT: shrq $56, %rcx |
| ; AVX2-NEXT: andl $15, %ecx |
| ; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm0 |
| ; AVX2-NEXT: shlq $56, %rcx |
| ; AVX2-NEXT: orq %rsi, %rcx |
| ; AVX2-NEXT: vmovq %xmm0, %rax |
| ; AVX2-NEXT: movq %rcx, -{{[0-9]+}}(%rsp) |
| ; AVX2-NEXT: movl %eax, %ecx |
| ; AVX2-NEXT: shrl $8, %ecx |
| ; AVX2-NEXT: vmovd %eax, %xmm1 |
| ; AVX2-NEXT: vpinsrb $1, %ecx, %xmm1, %xmm1 |
| ; AVX2-NEXT: movl %eax, %ecx |
| ; AVX2-NEXT: shrl $16, %ecx |
| ; AVX2-NEXT: vpinsrb $2, %ecx, %xmm1, %xmm1 |
| ; AVX2-NEXT: movl %eax, %ecx |
| ; AVX2-NEXT: shrl $24, %ecx |
| ; AVX2-NEXT: vpinsrb $3, %ecx, %xmm1, %xmm1 |
| ; AVX2-NEXT: movq %rax, %rcx |
| ; AVX2-NEXT: shrq $32, %rcx |
| ; AVX2-NEXT: vpinsrb $4, %ecx, %xmm1, %xmm1 |
| ; AVX2-NEXT: movq %rax, %rcx |
| ; AVX2-NEXT: shrq $40, %rcx |
| ; AVX2-NEXT: vpinsrb $5, %ecx, %xmm1, %xmm1 |
| ; AVX2-NEXT: movq %rax, %rcx |
| ; AVX2-NEXT: shrq $48, %rcx |
| ; AVX2-NEXT: vpinsrb $6, %ecx, %xmm1, %xmm1 |
| ; AVX2-NEXT: vpextrq $1, %xmm0, %rcx |
| ; AVX2-NEXT: shrq $56, %rax |
| ; AVX2-NEXT: vpinsrb $7, %eax, %xmm1, %xmm0 |
| ; AVX2-NEXT: movl %ecx, %eax |
| ; AVX2-NEXT: shrl $8, %eax |
| ; AVX2-NEXT: vpinsrb $8, %ecx, %xmm0, %xmm0 |
| ; AVX2-NEXT: vpinsrb $9, %eax, %xmm0, %xmm0 |
| ; AVX2-NEXT: movl %ecx, %eax |
| ; AVX2-NEXT: shrl $16, %eax |
| ; AVX2-NEXT: vpinsrb $10, %eax, %xmm0, %xmm0 |
| ; AVX2-NEXT: movl %ecx, %eax |
| ; AVX2-NEXT: shrl $24, %eax |
| ; AVX2-NEXT: vpinsrb $11, %eax, %xmm0, %xmm0 |
| ; AVX2-NEXT: movq %rcx, %rax |
| ; AVX2-NEXT: shrq $32, %rax |
| ; AVX2-NEXT: vpinsrb $12, %eax, %xmm0, %xmm0 |
| ; AVX2-NEXT: movq %rcx, %rax |
| ; AVX2-NEXT: shrq $40, %rax |
| ; AVX2-NEXT: vpinsrb $13, %eax, %xmm0, %xmm0 |
| ; AVX2-NEXT: movq %rcx, %rax |
| ; AVX2-NEXT: shrq $48, %rax |
| ; AVX2-NEXT: vpinsrb $14, %eax, %xmm0, %xmm0 |
| ; AVX2-NEXT: shrq $56, %rcx |
| ; AVX2-NEXT: vpinsrb $15, %ecx, %xmm0, %xmm0 |
| ; AVX2-NEXT: vmovdqa -{{[0-9]+}}(%rsp), %xmm1 |
| ; AVX2-NEXT: vinserti128 $1, %xmm0, %ymm1, %ymm0 |
| ; AVX2-NEXT: retq |
| %x4 = bitcast <32 x i8> %0 to <64 x i4> |
| %r0 = insertelement <64 x i4> %x4, i4 zeroinitializer, i32 1 |
| %r1 = insertelement <64 x i4> %r0, i4 zeroinitializer, i32 3 |
| %r2 = insertelement <64 x i4> %r1, i4 zeroinitializer, i32 5 |
| %r3 = insertelement <64 x i4> %r2, i4 zeroinitializer, i32 7 |
| %r4 = insertelement <64 x i4> %r3, i4 zeroinitializer, i32 9 |
| %r5 = insertelement <64 x i4> %r4, i4 zeroinitializer, i32 11 |
| %r6 = insertelement <64 x i4> %r5, i4 zeroinitializer, i32 13 |
| %r7 = insertelement <64 x i4> %r6, i4 zeroinitializer, i32 15 |
| %r8 = insertelement <64 x i4> %r7, i4 zeroinitializer, i32 17 |
| %r9 = insertelement <64 x i4> %r8, i4 zeroinitializer, i32 19 |
| %r10 = insertelement <64 x i4> %r9, i4 zeroinitializer, i32 21 |
| %r11 = insertelement <64 x i4> %r10, i4 zeroinitializer, i32 23 |
| %r12 = insertelement <64 x i4> %r11, i4 zeroinitializer, i32 25 |
| %r13 = insertelement <64 x i4> %r12, i4 zeroinitializer, i32 27 |
| %r14 = insertelement <64 x i4> %r13, i4 zeroinitializer, i32 29 |
| %r15 = insertelement <64 x i4> %r14, i4 zeroinitializer, i32 31 |
| %r16 = insertelement <64 x i4> %r15, i4 zeroinitializer, i32 33 |
| %r17 = insertelement <64 x i4> %r16, i4 zeroinitializer, i32 35 |
| %r18 = insertelement <64 x i4> %r17, i4 zeroinitializer, i32 37 |
| %r19 = insertelement <64 x i4> %r18, i4 zeroinitializer, i32 39 |
| %r20 = insertelement <64 x i4> %r19, i4 zeroinitializer, i32 41 |
| %r21 = insertelement <64 x i4> %r20, i4 zeroinitializer, i32 43 |
| %r22 = insertelement <64 x i4> %r21, i4 zeroinitializer, i32 45 |
| %r23 = insertelement <64 x i4> %r22, i4 zeroinitializer, i32 47 |
| %r24 = insertelement <64 x i4> %r23, i4 zeroinitializer, i32 49 |
| %r25 = insertelement <64 x i4> %r24, i4 zeroinitializer, i32 51 |
| %r26 = insertelement <64 x i4> %r25, i4 zeroinitializer, i32 53 |
| %r27 = insertelement <64 x i4> %r26, i4 zeroinitializer, i32 55 |
| %r28 = insertelement <64 x i4> %r27, i4 zeroinitializer, i32 57 |
| %r29 = insertelement <64 x i4> %r28, i4 zeroinitializer, i32 59 |
| %r30 = insertelement <64 x i4> %r29, i4 zeroinitializer, i32 61 |
| %r31 = insertelement <64 x i4> %r30, i4 zeroinitializer, i32 63 |
| %r = bitcast <64 x i4> %r15 to <32 x i8> |
| ret <32 x i8> %r |
| } |
| |
| define <2 x i64> @_clearupper2xi64c(<2 x i64>) nounwind { |
| ; SSE2-LABEL: _clearupper2xi64c: |
| ; SSE2: # %bb.0: |
| ; SSE2-NEXT: andps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0 |
| ; SSE2-NEXT: retq |
| ; |
| ; SSE42-LABEL: _clearupper2xi64c: |
| ; SSE42: # %bb.0: |
| ; SSE42-NEXT: xorps %xmm1, %xmm1 |
| ; SSE42-NEXT: blendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2],xmm1[3] |
| ; SSE42-NEXT: retq |
| ; |
| ; AVX-LABEL: _clearupper2xi64c: |
| ; AVX: # %bb.0: |
| ; AVX-NEXT: vxorps %xmm1, %xmm1, %xmm1 |
| ; AVX-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2],xmm1[3] |
| ; AVX-NEXT: retq |
| %r = and <2 x i64> <i64 4294967295, i64 4294967295>, %0 |
| ret <2 x i64> %r |
| } |
| |
| define <4 x i64> @_clearupper4xi64c(<4 x i64>) nounwind { |
| ; SSE2-LABEL: _clearupper4xi64c: |
| ; SSE2: # %bb.0: |
| ; SSE2-NEXT: movaps {{.*#+}} xmm2 = [4294967295,0,4294967295,0] |
| ; SSE2-NEXT: andps %xmm2, %xmm0 |
| ; SSE2-NEXT: andps %xmm2, %xmm1 |
| ; SSE2-NEXT: retq |
| ; |
| ; SSE42-LABEL: _clearupper4xi64c: |
| ; SSE42: # %bb.0: |
| ; SSE42-NEXT: xorps %xmm2, %xmm2 |
| ; SSE42-NEXT: blendps {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2],xmm2[3] |
| ; SSE42-NEXT: blendps {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2],xmm2[3] |
| ; SSE42-NEXT: retq |
| ; |
| ; AVX-LABEL: _clearupper4xi64c: |
| ; AVX: # %bb.0: |
| ; AVX-NEXT: vxorps %xmm1, %xmm1, %xmm1 |
| ; AVX-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7] |
| ; AVX-NEXT: retq |
| %r = and <4 x i64> <i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295>, %0 |
| ret <4 x i64> %r |
| } |
| |
| define <4 x i32> @_clearupper4xi32c(<4 x i32>) nounwind { |
| ; SSE2-LABEL: _clearupper4xi32c: |
| ; SSE2: # %bb.0: |
| ; SSE2-NEXT: andps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0 |
| ; SSE2-NEXT: retq |
| ; |
| ; SSE42-LABEL: _clearupper4xi32c: |
| ; SSE42: # %bb.0: |
| ; SSE42-NEXT: pxor %xmm1, %xmm1 |
| ; SSE42-NEXT: pblendw {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2],xmm1[3],xmm0[4],xmm1[5],xmm0[6],xmm1[7] |
| ; SSE42-NEXT: retq |
| ; |
| ; AVX-LABEL: _clearupper4xi32c: |
| ; AVX: # %bb.0: |
| ; AVX-NEXT: vpxor %xmm1, %xmm1, %xmm1 |
| ; AVX-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2],xmm1[3],xmm0[4],xmm1[5],xmm0[6],xmm1[7] |
| ; AVX-NEXT: retq |
| %r = and <4 x i32> <i32 65535, i32 65535, i32 65535, i32 65535>, %0 |
| ret <4 x i32> %r |
| } |
| |
| define <8 x i32> @_clearupper8xi32c(<8 x i32>) nounwind { |
| ; SSE2-LABEL: _clearupper8xi32c: |
| ; SSE2: # %bb.0: |
| ; SSE2-NEXT: movaps {{.*#+}} xmm2 = [65535,0,65535,0,65535,0,65535,0] |
| ; SSE2-NEXT: andps %xmm2, %xmm0 |
| ; SSE2-NEXT: andps %xmm2, %xmm1 |
| ; SSE2-NEXT: retq |
| ; |
| ; SSE42-LABEL: _clearupper8xi32c: |
| ; SSE42: # %bb.0: |
| ; SSE42-NEXT: pxor %xmm2, %xmm2 |
| ; SSE42-NEXT: pblendw {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2],xmm2[3],xmm0[4],xmm2[5],xmm0[6],xmm2[7] |
| ; SSE42-NEXT: pblendw {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2],xmm2[3],xmm1[4],xmm2[5],xmm1[6],xmm2[7] |
| ; SSE42-NEXT: retq |
| ; |
| ; AVX1-LABEL: _clearupper8xi32c: |
| ; AVX1: # %bb.0: |
| ; AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm0 |
| ; AVX1-NEXT: retq |
| ; |
| ; AVX2-LABEL: _clearupper8xi32c: |
| ; AVX2: # %bb.0: |
| ; AVX2-NEXT: vpxor %xmm1, %xmm1, %xmm1 |
| ; AVX2-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7],ymm0[8],ymm1[9],ymm0[10],ymm1[11],ymm0[12],ymm1[13],ymm0[14],ymm1[15] |
| ; AVX2-NEXT: retq |
| %r = and <8 x i32> <i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535>, %0 |
| ret <8 x i32> %r |
| } |
| |
| define <8 x i16> @_clearupper8xi16c(<8 x i16>) nounwind { |
| ; SSE-LABEL: _clearupper8xi16c: |
| ; SSE: # %bb.0: |
| ; SSE-NEXT: andps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0 |
| ; SSE-NEXT: retq |
| ; |
| ; AVX-LABEL: _clearupper8xi16c: |
| ; AVX: # %bb.0: |
| ; AVX-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0 |
| ; AVX-NEXT: retq |
| %r = and <8 x i16> <i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255>, %0 |
| ret <8 x i16> %r |
| } |
| |
| define <16 x i16> @_clearupper16xi16c(<16 x i16>) nounwind { |
| ; SSE-LABEL: _clearupper16xi16c: |
| ; SSE: # %bb.0: |
| ; SSE-NEXT: movaps {{.*#+}} xmm2 = [255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0] |
| ; SSE-NEXT: andps %xmm2, %xmm0 |
| ; SSE-NEXT: andps %xmm2, %xmm1 |
| ; SSE-NEXT: retq |
| ; |
| ; AVX-LABEL: _clearupper16xi16c: |
| ; AVX: # %bb.0: |
| ; AVX-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm0 |
| ; AVX-NEXT: retq |
| %r = and <16 x i16> <i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255>, %0 |
| ret <16 x i16> %r |
| } |
| |
| define <16 x i8> @_clearupper16xi8c(<16 x i8>) nounwind { |
| ; SSE-LABEL: _clearupper16xi8c: |
| ; SSE: # %bb.0: |
| ; SSE-NEXT: andps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0 |
| ; SSE-NEXT: retq |
| ; |
| ; AVX-LABEL: _clearupper16xi8c: |
| ; AVX: # %bb.0: |
| ; AVX-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0 |
| ; AVX-NEXT: retq |
| %r = and <16 x i8> <i8 15, i8 15, i8 15, i8 15, i8 15, i8 15, i8 15, i8 15, i8 15, i8 15, i8 15, i8 15, i8 15, i8 15, i8 15, i8 15>, %0 |
| ret <16 x i8> %r |
| } |
| |
| define <32 x i8> @_clearupper32xi8c(<32 x i8>) nounwind { |
| ; SSE-LABEL: _clearupper32xi8c: |
| ; SSE: # %bb.0: |
| ; SSE-NEXT: movaps {{.*#+}} xmm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15] |
| ; SSE-NEXT: andps %xmm2, %xmm0 |
| ; SSE-NEXT: andps %xmm2, %xmm1 |
| ; SSE-NEXT: retq |
| ; |
| ; AVX-LABEL: _clearupper32xi8c: |
| ; AVX: # %bb.0: |
| ; AVX-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm0 |
| ; AVX-NEXT: retq |
| %r = and <32 x i8> <i8 15, i8 15, i8 15, i8 15, i8 15, i8 15, i8 15, i8 15, i8 15, i8 15, i8 15, i8 15, i8 15, i8 15, i8 15, i8 15, i8 15, i8 15, i8 15, i8 15, i8 15, i8 15, i8 15, i8 15, i8 15, i8 15, i8 15, i8 15, i8 15, i8 15, i8 15, i8 15>, %0 |
| ret <32 x i8> %r |
| } |