1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py 2; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx | FileCheck %s --check-prefix=ALL --check-prefix=AVX --check-prefix=AVX1 3; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefix=ALL --check-prefix=AVX --check-prefix=AVX2 4 5define <4 x i64> @testv4i64(<4 x i64> %in) nounwind { 6; AVX1-LABEL: testv4i64: 7; AVX1: # BB#0: 8; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1 9; AVX1-NEXT: vmovaps {{.*#+}} xmm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15] 10; AVX1-NEXT: vandps %xmm2, %xmm1, %xmm3 11; AVX1-NEXT: vmovdqa {{.*#+}} xmm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4] 12; AVX1-NEXT: vpshufb %xmm3, %xmm4, %xmm3 13; AVX1-NEXT: vpsrlw $4, %xmm1, %xmm1 14; AVX1-NEXT: vpand %xmm2, %xmm1, %xmm1 15; AVX1-NEXT: vpshufb %xmm1, %xmm4, %xmm1 16; AVX1-NEXT: vpaddb %xmm3, %xmm1, %xmm1 17; AVX1-NEXT: vpxor %xmm3, %xmm3, %xmm3 18; AVX1-NEXT: vpsadbw %xmm3, %xmm1, %xmm1 19; AVX1-NEXT: vandps %xmm2, %xmm0, %xmm5 20; AVX1-NEXT: vpshufb %xmm5, %xmm4, %xmm5 21; AVX1-NEXT: vpsrlw $4, %xmm0, %xmm0 22; AVX1-NEXT: vpand %xmm2, %xmm0, %xmm0 23; AVX1-NEXT: vpshufb %xmm0, %xmm4, %xmm0 24; AVX1-NEXT: vpaddb %xmm5, %xmm0, %xmm0 25; AVX1-NEXT: vpsadbw %xmm3, %xmm0, %xmm0 26; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 27; AVX1-NEXT: retq 28; 29; AVX2-LABEL: testv4i64: 30; AVX2: # BB#0: 31; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15] 32; AVX2-NEXT: vpand %ymm1, %ymm0, %ymm2 33; AVX2-NEXT: vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4] 34; AVX2-NEXT: vpshufb %ymm2, %ymm3, %ymm2 35; AVX2-NEXT: vpsrlw $4, %ymm0, %ymm0 36; AVX2-NEXT: vpand %ymm1, %ymm0, %ymm0 37; AVX2-NEXT: vpshufb %ymm0, %ymm3, %ymm0 38; AVX2-NEXT: vpaddb %ymm2, %ymm0, %ymm0 39; AVX2-NEXT: vpxor %ymm1, %ymm1, %ymm1 40; AVX2-NEXT: vpsadbw %ymm1, %ymm0, %ymm0 41; AVX2-NEXT: retq 42 %out = call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %in) 43 ret <4 x i64> %out 44} 45 46define <8 x i32> @testv8i32(<8 x i32> %in) nounwind { 47; AVX1-LABEL: testv8i32: 48; AVX1: # BB#0: 49; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1 50; AVX1-NEXT: vmovaps {{.*#+}} xmm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15] 51; AVX1-NEXT: vandps %xmm2, %xmm1, %xmm3 52; AVX1-NEXT: vmovdqa {{.*#+}} xmm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4] 53; AVX1-NEXT: vpshufb %xmm3, %xmm4, %xmm3 54; AVX1-NEXT: vpsrlw $4, %xmm1, %xmm1 55; AVX1-NEXT: vpand %xmm2, %xmm1, %xmm1 56; AVX1-NEXT: vpshufb %xmm1, %xmm4, %xmm1 57; AVX1-NEXT: vpaddb %xmm3, %xmm1, %xmm1 58; AVX1-NEXT: vpxor %xmm3, %xmm3, %xmm3 59; AVX1-NEXT: vpunpckhdq {{.*#+}} xmm5 = xmm1[2],xmm3[2],xmm1[3],xmm3[3] 60; AVX1-NEXT: vpsadbw %xmm3, %xmm5, %xmm5 61; AVX1-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1] 62; AVX1-NEXT: vpsadbw %xmm3, %xmm1, %xmm1 63; AVX1-NEXT: vpackuswb %xmm5, %xmm1, %xmm1 64; AVX1-NEXT: vandps %xmm2, %xmm0, %xmm5 65; AVX1-NEXT: vpshufb %xmm5, %xmm4, %xmm5 66; AVX1-NEXT: vpsrlw $4, %xmm0, %xmm0 67; AVX1-NEXT: vpand %xmm2, %xmm0, %xmm0 68; AVX1-NEXT: vpshufb %xmm0, %xmm4, %xmm0 69; AVX1-NEXT: vpaddb %xmm5, %xmm0, %xmm0 70; AVX1-NEXT: vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm3[2],xmm0[3],xmm3[3] 71; AVX1-NEXT: vpsadbw %xmm3, %xmm2, %xmm2 72; AVX1-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1] 73; AVX1-NEXT: vpsadbw %xmm3, %xmm0, %xmm0 74; AVX1-NEXT: vpackuswb %xmm2, %xmm0, %xmm0 75; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 76; AVX1-NEXT: retq 77; 78; AVX2-LABEL: testv8i32: 79; AVX2: # BB#0: 80; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15] 81; AVX2-NEXT: vpand %ymm1, %ymm0, %ymm2 82; AVX2-NEXT: vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4] 83; AVX2-NEXT: vpshufb %ymm2, %ymm3, %ymm2 84; AVX2-NEXT: vpsrlw $4, %ymm0, %ymm0 85; AVX2-NEXT: vpand %ymm1, %ymm0, %ymm0 86; AVX2-NEXT: vpshufb %ymm0, %ymm3, %ymm0 87; AVX2-NEXT: vpaddb %ymm2, %ymm0, %ymm0 88; AVX2-NEXT: vpxor %ymm1, %ymm1, %ymm1 89; AVX2-NEXT: vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7] 90; AVX2-NEXT: vpsadbw %ymm1, %ymm2, %ymm2 91; AVX2-NEXT: vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5] 92; AVX2-NEXT: vpsadbw %ymm1, %ymm0, %ymm0 93; AVX2-NEXT: vpackuswb %ymm2, %ymm0, %ymm0 94; AVX2-NEXT: retq 95 %out = call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %in) 96 ret <8 x i32> %out 97} 98 99define <16 x i16> @testv16i16(<16 x i16> %in) nounwind { 100; AVX1-LABEL: testv16i16: 101; AVX1: # BB#0: 102; AVX1-NEXT: vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15] 103; AVX1-NEXT: vpand %xmm1, %xmm0, %xmm2 104; AVX1-NEXT: vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4] 105; AVX1-NEXT: vpshufb %xmm2, %xmm3, %xmm2 106; AVX1-NEXT: vpsrlw $4, %xmm0, %xmm4 107; AVX1-NEXT: vpand %xmm1, %xmm4, %xmm4 108; AVX1-NEXT: vpshufb %xmm4, %xmm3, %xmm4 109; AVX1-NEXT: vpaddb %xmm2, %xmm4, %xmm2 110; AVX1-NEXT: vpsllw $8, %xmm2, %xmm4 111; AVX1-NEXT: vpaddb %xmm2, %xmm4, %xmm2 112; AVX1-NEXT: vpsrlw $8, %xmm2, %xmm2 113; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm0 114; AVX1-NEXT: vpand %xmm1, %xmm0, %xmm4 115; AVX1-NEXT: vpshufb %xmm4, %xmm3, %xmm4 116; AVX1-NEXT: vpsrlw $4, %xmm0, %xmm0 117; AVX1-NEXT: vpand %xmm1, %xmm0, %xmm0 118; AVX1-NEXT: vpshufb %xmm0, %xmm3, %xmm0 119; AVX1-NEXT: vpaddb %xmm4, %xmm0, %xmm0 120; AVX1-NEXT: vpsllw $8, %xmm0, %xmm1 121; AVX1-NEXT: vpaddb %xmm0, %xmm1, %xmm0 122; AVX1-NEXT: vpsrlw $8, %xmm0, %xmm0 123; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm2, %ymm0 124; AVX1-NEXT: retq 125; 126; AVX2-LABEL: testv16i16: 127; AVX2: # BB#0: 128; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15] 129; AVX2-NEXT: vpand %ymm1, %ymm0, %ymm2 130; AVX2-NEXT: vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4] 131; AVX2-NEXT: vpshufb %ymm2, %ymm3, %ymm2 132; AVX2-NEXT: vpsrlw $4, %ymm0, %ymm0 133; AVX2-NEXT: vpand %ymm1, %ymm0, %ymm0 134; AVX2-NEXT: vpshufb %ymm0, %ymm3, %ymm0 135; AVX2-NEXT: vpaddb %ymm2, %ymm0, %ymm0 136; AVX2-NEXT: vpsllw $8, %ymm0, %ymm1 137; AVX2-NEXT: vpaddb %ymm0, %ymm1, %ymm0 138; AVX2-NEXT: vpsrlw $8, %ymm0, %ymm0 139; AVX2-NEXT: retq 140 %out = call <16 x i16> @llvm.ctpop.v16i16(<16 x i16> %in) 141 ret <16 x i16> %out 142} 143 144define <32 x i8> @testv32i8(<32 x i8> %in) nounwind { 145; AVX1-LABEL: testv32i8: 146; AVX1: # BB#0: 147; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1 148; AVX1-NEXT: vmovaps {{.*#+}} xmm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15] 149; AVX1-NEXT: vandps %xmm2, %xmm1, %xmm3 150; AVX1-NEXT: vmovdqa {{.*#+}} xmm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4] 151; AVX1-NEXT: vpshufb %xmm3, %xmm4, %xmm3 152; AVX1-NEXT: vpsrlw $4, %xmm1, %xmm1 153; AVX1-NEXT: vpand %xmm2, %xmm1, %xmm1 154; AVX1-NEXT: vpshufb %xmm1, %xmm4, %xmm1 155; AVX1-NEXT: vpaddb %xmm3, %xmm1, %xmm1 156; AVX1-NEXT: vandps %xmm2, %xmm0, %xmm3 157; AVX1-NEXT: vpshufb %xmm3, %xmm4, %xmm3 158; AVX1-NEXT: vpsrlw $4, %xmm0, %xmm0 159; AVX1-NEXT: vpand %xmm2, %xmm0, %xmm0 160; AVX1-NEXT: vpshufb %xmm0, %xmm4, %xmm0 161; AVX1-NEXT: vpaddb %xmm3, %xmm0, %xmm0 162; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 163; AVX1-NEXT: retq 164; 165; AVX2-LABEL: testv32i8: 166; AVX2: # BB#0: 167; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15] 168; AVX2-NEXT: vpand %ymm1, %ymm0, %ymm2 169; AVX2-NEXT: vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4] 170; AVX2-NEXT: vpshufb %ymm2, %ymm3, %ymm2 171; AVX2-NEXT: vpsrlw $4, %ymm0, %ymm0 172; AVX2-NEXT: vpand %ymm1, %ymm0, %ymm0 173; AVX2-NEXT: vpshufb %ymm0, %ymm3, %ymm0 174; AVX2-NEXT: vpaddb %ymm2, %ymm0, %ymm0 175; AVX2-NEXT: retq 176 %out = call <32 x i8> @llvm.ctpop.v32i8(<32 x i8> %in) 177 ret <32 x i8> %out 178} 179 180define <4 x i64> @foldv4i64() nounwind { 181; ALL-LABEL: foldv4i64: 182; ALL: # BB#0: 183; ALL-NEXT: vmovaps {{.*#+}} ymm0 = [1,64,0,8] 184; ALL-NEXT: retq 185 %out = call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> <i64 256, i64 -1, i64 0, i64 255>) 186 ret <4 x i64> %out 187} 188 189define <8 x i32> @foldv8i32() nounwind { 190; ALL-LABEL: foldv8i32: 191; ALL: # BB#0: 192; ALL-NEXT: vmovaps {{.*#+}} ymm0 = [1,32,0,8,16,3,2,3] 193; ALL-NEXT: retq 194 %out = call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> <i32 256, i32 -1, i32 0, i32 255, i32 -65536, i32 7, i32 24, i32 88>) 195 ret <8 x i32> %out 196} 197 198define <16 x i16> @foldv16i16() nounwind { 199; ALL-LABEL: foldv16i16: 200; ALL: # BB#0: 201; ALL-NEXT: vmovaps {{.*#+}} ymm0 = [1,16,0,8,0,3,2,3,15,7,1,1,1,1,1,1] 202; ALL-NEXT: retq 203 %out = call <16 x i16> @llvm.ctpop.v16i16(<16 x i16> <i16 256, i16 -1, i16 0, i16 255, i16 -65536, i16 7, i16 24, i16 88, i16 -2, i16 254, i16 1, i16 2, i16 4, i16 8, i16 16, i16 32>) 204 ret <16 x i16> %out 205} 206 207define <32 x i8> @foldv32i8() nounwind { 208; ALL-LABEL: foldv32i8: 209; ALL: # BB#0: 210; ALL-NEXT: vmovaps {{.*#+}} ymm0 = [0,8,0,8,0,3,2,3,7,7,1,1,1,1,1,1,1,1,0,0,1,2,3,4,5,6,7,8,2,2,3,7] 211; ALL-NEXT: retq 212 %out = call <32 x i8> @llvm.ctpop.v32i8(<32 x i8> <i8 256, i8 -1, i8 0, i8 255, i8 -65536, i8 7, i8 24, i8 88, i8 -2, i8 254, i8 1, i8 2, i8 4, i8 8, i8 16, i8 32, i8 64, i8 128, i8 256, i8 -256, i8 -128, i8 -64, i8 -32, i8 -16, i8 -8, i8 -4, i8 -2, i8 -1, i8 3, i8 5, i8 7, i8 127>) 213 ret <32 x i8> %out 214} 215 216declare <4 x i64> @llvm.ctpop.v4i64(<4 x i64>) 217declare <8 x i32> @llvm.ctpop.v8i32(<8 x i32>) 218declare <16 x i16> @llvm.ctpop.v16i16(<16 x i16>) 219declare <32 x i8> @llvm.ctpop.v32i8(<32 x i8>) 220