1 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=x86-64 -x86-experimental-vector-shuffle-lowering | FileCheck %s --check-prefix=CHECK-SSE2 2 3 target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128" 4 target triple = "x86_64-unknown-unknown" 5 6 define <8 x i16> @shuffle_v8i16_01012323(<8 x i16> %a, <8 x i16> %b) { 7 ; CHECK-SSE2-LABEL: @shuffle_v8i16_01012323 8 ; CHECK-SSE2: # BB#0: 9 ; CHECK-SSE2-NEXT: pshufd {{.*}} # xmm0 = xmm0[0,0,1,1] 10 ; CHECK-SSE2-NEXT: retq 11 %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 0, i32 1, i32 0, i32 1, i32 2, i32 3, i32 2, i32 3> 12 ret <8 x i16> %shuffle 13 } 14 define <8 x i16> @shuffle_v8i16_67452301(<8 x i16> %a, <8 x i16> %b) { 15 ; CHECK-SSE2-LABEL: @shuffle_v8i16_67452301 16 ; CHECK-SSE2: # BB#0: 17 ; CHECK-SSE2-NEXT: pshufd {{.*}} # xmm0 = xmm0[3,2,1,0] 18 ; CHECK-SSE2-NEXT: retq 19 %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 6, i32 7, i32 4, i32 5, i32 2, i32 3, i32 0, i32 1> 20 ret <8 x i16> %shuffle 21 } 22 define <8 x i16> @shuffle_v8i16_456789AB(<8 x i16> %a, <8 x i16> %b) { 23 ; CHECK-SSE2-LABEL: @shuffle_v8i16_456789AB 24 ; CHECK-SSE2: # BB#0: 25 ; CHECK-SSE2: shufpd {{.*}} # xmm0 = xmm0[1],xmm1[0] 26 ; CHECK-SSE2-NEXT: retq 27 %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11> 28 ret <8 x i16> %shuffle 29 } 30 31 define <8 x i16> @shuffle_v8i16_00000000(<8 x i16> %a, <8 x i16> %b) { 32 ; CHECK-SSE2-LABEL: @shuffle_v8i16_00000000 33 ; CHECK-SSE2: # BB#0: 34 ; CHECK-SSE2-NEXT: pshufd {{.*}} # xmm0 = xmm0[0,1,0,3] 35 ; CHECK-SSE2-NEXT: pshuflw {{.*}} # xmm0 = xmm0[0,0,0,0,4,5,6,7] 36 ; CHECK-SSE2-NEXT: pshufhw {{.*}} # xmm0 = xmm0[0,1,2,3,4,4,4,4] 37 ; CHECK-SSE2-NEXT: retq 38 %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0> 39 ret <8 x i16> %shuffle 40 } 41 define <8 x i16> @shuffle_v8i16_00004444(<8 x i16> %a, <8 x i16> %b) { 42 ; CHECK-SSE2-LABEL: @shuffle_v8i16_00004444 43 ; CHECK-SSE2: # BB#0: 44 ; CHECK-SSE2-NEXT: pshuflw {{.*}} # xmm0 = xmm0[0,0,0,0,4,5,6,7] 45 ; CHECK-SSE2-NEXT: pshufhw {{.*}} # xmm0 = xmm0[0,1,2,3,4,4,4,4] 46 ; CHECK-SSE2-NEXT: retq 47 %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 0, i32 0, i32 0, i32 0, i32 4, i32 4, i32 4, i32 4> 48 ret <8 x i16> %shuffle 49 } 50 define <8 x i16> @shuffle_v8i16_31206745(<8 x i16> %a, <8 x i16> %b) { 51 ; CHECK-SSE2-LABEL: @shuffle_v8i16_31206745 52 ; CHECK-SSE2: # BB#0: 53 ; CHECK-SSE2-NEXT: pshuflw {{.*}} # xmm0 = xmm0[3,1,2,0,4,5,6,7] 54 ; CHECK-SSE2-NEXT: pshufd {{.*}} # xmm0 = xmm0[0,1,3,2] 55 ; CHECK-SSE2-NEXT: retq 56 %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 3, i32 1, i32 2, i32 0, i32 6, i32 7, i32 4, i32 5> 57 ret <8 x i16> %shuffle 58 } 59 define <8 x i16> @shuffle_v8i16_44440000(<8 x i16> %a, <8 x i16> %b) { 60 ; CHECK-SSE2-LABEL: @shuffle_v8i16_44440000 61 ; CHECK-SSE2: # BB#0: 62 ; CHECK-SSE2-NEXT: pshufd {{.*}} # xmm0 = xmm0[2,1,0,3] 63 ; CHECK-SSE2-NEXT: pshuflw {{.*}} # xmm0 = xmm0[0,0,0,0,4,5,6,7] 64 ; CHECK-SSE2-NEXT: pshufhw {{.*}} # xmm0 = xmm0[0,1,2,3,4,4,4,4] 65 ; CHECK-SSE2-NEXT: retq 66 %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 4, i32 4, i32 4, i32 4, i32 0, i32 0, i32 0, i32 0> 67 ret <8 x i16> %shuffle 68 } 69 define <8 x i16> @shuffle_v8i16_75643120(<8 x i16> %a, <8 x i16> %b) { 70 ; CHECK-SSE2-LABEL: @shuffle_v8i16_75643120 71 ; CHECK-SSE2: # BB#0: 72 ; CHECK-SSE2-NEXT: pshufd {{.*}} # xmm0 = xmm0[2,3,0,1] 73 ; CHECK-SSE2-NEXT: pshuflw {{.*}} # xmm0 = xmm0[3,1,2,0,4,5,6,7] 74 ; CHECK-SSE2-NEXT: pshufhw {{.*}} # xmm0 = xmm0[0,1,2,3,7,5,6,4] 75 ; CHECK-SSE2-NEXT: retq 76 %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 7, i32 5, i32 6, i32 4, i32 3, i32 1, i32 2, i32 0> 77 ret <8 x i16> %shuffle 78 } 79 80 define <8 x i16> @shuffle_v8i16_10545410(<8 x i16> %a, <8 x i16> %b) { 81 ; CHECK-SSE2-LABEL: @shuffle_v8i16_10545410 82 ; CHECK-SSE2: # BB#0: 83 ; CHECK-SSE2-NEXT: pshufd {{.*}} # xmm0 = xmm0[0,2,2,0] 84 ; CHECK-SSE2-NEXT: pshuflw {{.*}} # xmm0 = xmm0[1,0,3,2,4,5,6,7] 85 ; CHECK-SSE2-NEXT: pshufhw {{.*}} # xmm0 = xmm0[0,1,2,3,5,4,7,6] 86 ; CHECK-SSE2-NEXT: retq 87 %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 1, i32 0, i32 5, i32 4, i32 5, i32 4, i32 1, i32 0> 88 ret <8 x i16> %shuffle 89 } 90 define <8 x i16> @shuffle_v8i16_54105410(<8 x i16> %a, <8 x i16> %b) { 91 ; CHECK-SSE2-LABEL: @shuffle_v8i16_54105410 92 ; CHECK-SSE2: # BB#0: 93 ; CHECK-SSE2-NEXT: pshufd {{.*}} # xmm0 = xmm0[0,2,2,0] 94 ; CHECK-SSE2-NEXT: pshuflw {{.*}} # xmm0 = xmm0[3,2,1,0,4,5,6,7] 95 ; CHECK-SSE2-NEXT: pshufhw {{.*}} # xmm0 = xmm0[0,1,2,3,5,4,7,6] 96 ; CHECK-SSE2-NEXT: retq 97 %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 5, i32 4, i32 1, i32 0, i32 5, i32 4, i32 1, i32 0> 98 ret <8 x i16> %shuffle 99 } 100 define <8 x i16> @shuffle_v8i16_54101054(<8 x i16> %a, <8 x i16> %b) { 101 ; CHECK-SSE2-LABEL: @shuffle_v8i16_54101054 102 ; CHECK-SSE2: # BB#0: 103 ; CHECK-SSE2-NEXT: pshufd {{.*}} # xmm0 = xmm0[0,2,2,0] 104 ; CHECK-SSE2-NEXT: pshuflw {{.*}} # xmm0 = xmm0[3,2,1,0,4,5,6,7] 105 ; CHECK-SSE2-NEXT: pshufhw {{.*}} # xmm0 = xmm0[0,1,2,3,7,6,5,4] 106 ; CHECK-SSE2-NEXT: retq 107 %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 5, i32 4, i32 1, i32 0, i32 1, i32 0, i32 5, i32 4> 108 ret <8 x i16> %shuffle 109 } 110 define <8 x i16> @shuffle_v8i16_04400440(<8 x i16> %a, <8 x i16> %b) { 111 ; CHECK-SSE2-LABEL: @shuffle_v8i16_04400440 112 ; CHECK-SSE2: # BB#0: 113 ; CHECK-SSE2-NEXT: pshufd {{.*}} # xmm0 = xmm0[0,2,2,0] 114 ; CHECK-SSE2-NEXT: pshuflw {{.*}} # xmm0 = xmm0[0,2,2,0,4,5,6,7] 115 ; CHECK-SSE2-NEXT: pshufhw {{.*}} # xmm0 = xmm0[0,1,2,3,6,4,4,6] 116 ; CHECK-SSE2-NEXT: retq 117 %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 0, i32 4, i32 4, i32 0, i32 0, i32 4, i32 4, i32 0> 118 ret <8 x i16> %shuffle 119 } 120 define <8 x i16> @shuffle_v8i16_40044004(<8 x i16> %a, <8 x i16> %b) { 121 ; CHECK-SSE2-LABEL: @shuffle_v8i16_40044004 122 ; CHECK-SSE2: # BB#0: 123 ; CHECK-SSE2-NEXT: pshufd {{.*}} # xmm0 = xmm0[0,2,2,0] 124 ; CHECK-SSE2-NEXT: pshuflw {{.*}} # xmm0 = xmm0[2,0,0,2,4,5,6,7] 125 ; CHECK-SSE2-NEXT: pshufhw {{.*}} # xmm0 = xmm0[0,1,2,3,4,6,6,4] 126 ; CHECK-SSE2-NEXT: retq 127 %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 4, i32 0, i32 0, i32 4, i32 4, i32 0, i32 0, i32 4> 128 ret <8 x i16> %shuffle 129 } 130 131 define <8 x i16> @shuffle_v8i16_26405173(<8 x i16> %a, <8 x i16> %b) { 132 ; CHECK-SSE2-LABEL: @shuffle_v8i16_26405173 133 ; CHECK-SSE2: # BB#0: 134 ; CHECK-SSE2-NEXT: pshuflw {{.*}} # xmm0 = xmm0[0,2,1,3,4,5,6,7] 135 ; CHECK-SSE2-NEXT: pshufhw {{.*}} # xmm0 = xmm0[0,1,2,3,7,5,4,6] 136 ; CHECK-SSE2-NEXT: pshufd {{.*}} # xmm0 = xmm0[0,3,2,1] 137 ; CHECK-SSE2-NEXT: pshuflw {{.*}} # xmm0 = xmm0[1,3,2,0,4,5,6,7] 138 ; CHECK-SSE2-NEXT: pshufhw {{.*}} # xmm0 = xmm0[0,1,2,3,5,6,4,7] 139 ; CHECK-SSE2-NEXT: retq 140 %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 2, i32 6, i32 4, i32 0, i32 5, i32 1, i32 7, i32 3> 141 ret <8 x i16> %shuffle 142 } 143 define <8 x i16> @shuffle_v8i16_20645173(<8 x i16> %a, <8 x i16> %b) { 144 ; CHECK-SSE2-LABEL: @shuffle_v8i16_20645173 145 ; CHECK-SSE2: # BB#0: 146 ; CHECK-SSE2-NEXT: pshuflw {{.*}} # xmm0 = xmm0[0,2,1,3,4,5,6,7] 147 ; CHECK-SSE2-NEXT: pshufhw {{.*}} # xmm0 = xmm0[0,1,2,3,7,5,4,6] 148 ; CHECK-SSE2-NEXT: pshufd {{.*}} # xmm0 = xmm0[0,3,2,1] 149 ; CHECK-SSE2-NEXT: pshuflw {{.*}} # xmm0 = xmm0[1,0,3,2,4,5,6,7] 150 ; CHECK-SSE2-NEXT: pshufhw {{.*}} # xmm0 = xmm0[0,1,2,3,5,6,4,7] 151 ; CHECK-SSE2-NEXT: retq 152 %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 2, i32 0, i32 6, i32 4, i32 5, i32 1, i32 7, i32 3> 153 ret <8 x i16> %shuffle 154 } 155 define <8 x i16> @shuffle_v8i16_26401375(<8 x i16> %a, <8 x i16> %b) { 156 ; CHECK-SSE2-LABEL: @shuffle_v8i16_26401375 157 ; CHECK-SSE2: # BB#0: 158 ; CHECK-SSE2-NEXT: pshuflw {{.*}} # xmm0 = xmm0[0,2,1,3,4,5,6,7] 159 ; CHECK-SSE2-NEXT: pshufhw {{.*}} # xmm0 = xmm0[0,1,2,3,7,5,4,6] 160 ; CHECK-SSE2-NEXT: pshufd {{.*}} # xmm0 = xmm0[0,3,1,2] 161 ; CHECK-SSE2-NEXT: pshuflw {{.*}} # xmm0 = xmm0[1,3,2,0,4,5,6,7] 162 ; CHECK-SSE2-NEXT: retq 163 %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 2, i32 6, i32 4, i32 0, i32 1, i32 3, i32 7, i32 5> 164 ret <8 x i16> %shuffle 165 } 166 167 define <8 x i16> @shuffle_v8i16_00444444(<8 x i16> %a, <8 x i16> %b) { 168 ; CHECK-SSE2-LABEL: @shuffle_v8i16_00444444 169 ; CHECK-SSE2: # BB#0: 170 ; CHECK-SSE2-NEXT: pshufd {{.*}} # xmm0 = xmm0[0,2,2,3] 171 ; CHECK-SSE2-NEXT: pshuflw {{.*}} # xmm0 = xmm0[0,0,2,2,4,5,6,7] 172 ; CHECK-SSE2-NEXT: pshufhw {{.*}} # xmm0 = xmm0[0,1,2,3,4,4,4,4] 173 ; CHECK-SSE2-NEXT: retq 174 %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 0, i32 0, i32 4, i32 4, i32 4, i32 4, i32 4, i32 4> 175 ret <8 x i16> %shuffle 176 } 177 define <8 x i16> @shuffle_v8i16_44004444(<8 x i16> %a, <8 x i16> %b) { 178 ; CHECK-SSE2-LABEL: @shuffle_v8i16_44004444 179 ; CHECK-SSE2: # BB#0: 180 ; CHECK-SSE2-NEXT: pshufd {{.*}} # xmm0 = xmm0[0,2,2,3] 181 ; CHECK-SSE2-NEXT: pshuflw {{.*}} # xmm0 = xmm0[2,2,0,0,4,5,6,7] 182 ; CHECK-SSE2-NEXT: pshufhw {{.*}} # xmm0 = xmm0[0,1,2,3,4,4,4,4] 183 ; CHECK-SSE2-NEXT: retq 184 %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 4, i32 4, i32 0, i32 0, i32 4, i32 4, i32 4, i32 4> 185 ret <8 x i16> %shuffle 186 } 187 define <8 x i16> @shuffle_v8i16_04404444(<8 x i16> %a, <8 x i16> %b) { 188 ; CHECK-SSE2-LABEL: @shuffle_v8i16_04404444 189 ; CHECK-SSE2: # BB#0: 190 ; CHECK-SSE2-NEXT: pshufd {{.*}} # xmm0 = xmm0[0,2,2,3] 191 ; CHECK-SSE2-NEXT: pshuflw {{.*}} # xmm0 = xmm0[0,2,2,0,4,5,6,7] 192 ; CHECK-SSE2-NEXT: pshufhw {{.*}} # xmm0 = xmm0[0,1,2,3,4,4,4,4] 193 ; CHECK-SSE2-NEXT: retq 194 %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 0, i32 4, i32 4, i32 0, i32 4, i32 4, i32 4, i32 4> 195 ret <8 x i16> %shuffle 196 } 197 define <8 x i16> @shuffle_v8i16_04400000(<8 x i16> %a, <8 x i16> %b) { 198 ; CHECK-SSE2-LABEL: @shuffle_v8i16_04400000 199 ; CHECK-SSE2: # BB#0: 200 ; CHECK-SSE2-NEXT: pshufd {{.*}} # xmm0 = xmm0[0,2,0,3] 201 ; CHECK-SSE2-NEXT: pshuflw {{.*}} # xmm0 = xmm0[0,2,2,0,4,5,6,7] 202 ; CHECK-SSE2-NEXT: pshufhw {{.*}} # xmm0 = xmm0[0,1,2,3,4,4,4,4] 203 ; CHECK-SSE2-NEXT: retq 204 %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 0, i32 4, i32 4, i32 0, i32 0, i32 0, i32 0, i32 0> 205 ret <8 x i16> %shuffle 206 } 207 define <8 x i16> @shuffle_v8i16_04404567(<8 x i16> %a, <8 x i16> %b) { 208 ; CHECK-SSE2-LABEL: @shuffle_v8i16_04404567 209 ; CHECK-SSE2: # BB#0: 210 ; CHECK-SSE2-NEXT: pshufd {{.*}} # xmm0 = xmm0[0,2,2,3] 211 ; CHECK-SSE2-NEXT: pshuflw {{.*}} # xmm0 = xmm0[0,2,2,0,4,5,6,7] 212 ; CHECK-SSE2-NEXT: retq 213 %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 0, i32 4, i32 4, i32 0, i32 4, i32 5, i32 6, i32 7> 214 ret <8 x i16> %shuffle 215 } 216 217 define <8 x i16> @shuffle_v8i16_0X444444(<8 x i16> %a, <8 x i16> %b) { 218 ; CHECK-SSE2-LABEL: @shuffle_v8i16_0X444444 219 ; CHECK-SSE2: # BB#0: 220 ; CHECK-SSE2-NEXT: pshufd {{.*}} # xmm0 = xmm0[0,2,2,3] 221 ; CHECK-SSE2-NEXT: pshuflw {{.*}} # xmm0 = xmm0[0,1,2,2,4,5,6,7] 222 ; CHECK-SSE2-NEXT: pshufhw {{.*}} # xmm0 = xmm0[0,1,2,3,4,4,4,4] 223 ; CHECK-SSE2-NEXT: retq 224 %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 0, i32 undef, i32 4, i32 4, i32 4, i32 4, i32 4, i32 4> 225 ret <8 x i16> %shuffle 226 } 227 define <8 x i16> @shuffle_v8i16_44X04444(<8 x i16> %a, <8 x i16> %b) { 228 ; CHECK-SSE2-LABEL: @shuffle_v8i16_44X04444 229 ; CHECK-SSE2: # BB#0: 230 ; CHECK-SSE2-NEXT: pshufd {{.*}} # xmm0 = xmm0[0,2,2,3] 231 ; CHECK-SSE2-NEXT: pshuflw {{.*}} # xmm0 = xmm0[2,2,2,0,4,5,6,7] 232 ; CHECK-SSE2-NEXT: pshufhw {{.*}} # xmm0 = xmm0[0,1,2,3,4,4,4,4] 233 ; CHECK-SSE2-NEXT: retq 234 %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 4, i32 4, i32 undef, i32 0, i32 4, i32 4, i32 4, i32 4> 235 ret <8 x i16> %shuffle 236 } 237 define <8 x i16> @shuffle_v8i16_X4404444(<8 x i16> %a, <8 x i16> %b) { 238 ; CHECK-SSE2-LABEL: @shuffle_v8i16_X4404444 239 ; CHECK-SSE2: # BB#0: 240 ; CHECK-SSE2-NEXT: pshufd {{.*}} # xmm0 = xmm0[0,2,2,3] 241 ; CHECK-SSE2-NEXT: pshuflw {{.*}} # xmm0 = xmm0[0,2,2,0,4,5,6,7] 242 ; CHECK-SSE2-NEXT: pshufhw {{.*}} # xmm0 = xmm0[0,1,2,3,4,4,4,4] 243 ; CHECK-SSE2-NEXT: retq 244 %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 undef, i32 4, i32 4, i32 0, i32 4, i32 4, i32 4, i32 4> 245 ret <8 x i16> %shuffle 246 } 247 248 define <8 x i16> @shuffle_v8i16_0127XXXX(<8 x i16> %a, <8 x i16> %b) { 249 ; CHECK-SSE2-LABEL: @shuffle_v8i16_0127XXXX 250 ; CHECK-SSE2: # BB#0: 251 ; CHECK-SSE2-NEXT: pshufd {{.*}} # xmm0 = xmm0[0,2,1,3] 252 ; CHECK-SSE2-NEXT: pshufhw {{.*}} # xmm0 = xmm0[0,1,2,3,4,7,6,7] 253 ; CHECK-SSE2-NEXT: pshufd {{.*}} # xmm0 = xmm0[0,2,2,3] 254 ; CHECK-SSE2-NEXT: retq 255 %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 0, i32 1, i32 2, i32 7, i32 undef, i32 undef, i32 undef, i32 undef> 256 ret <8 x i16> %shuffle 257 } 258 259 define <8 x i16> @shuffle_v8i16_XXXX4563(<8 x i16> %a, <8 x i16> %b) { 260 ; CHECK-SSE2-LABEL: @shuffle_v8i16_XXXX4563 261 ; CHECK-SSE2: # BB#0: 262 ; CHECK-SSE2-NEXT: pshufd {{.*}} # xmm0 = xmm0[3,1,2,0] 263 ; CHECK-SSE2-NEXT: pshuflw {{.*}} # xmm0 = xmm0[0,3,2,3,4,5,6,7] 264 ; CHECK-SSE2-NEXT: pshufd {{.*}} # xmm0 = xmm0[0,1,2,0] 265 ; CHECK-SSE2-NEXT: retq 266 %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 undef, i32 undef, i32 undef, i32 undef, i32 4, i32 5, i32 6, i32 3> 267 ret <8 x i16> %shuffle 268 } 269 270 define <8 x i16> @shuffle_v8i16_4563XXXX(<8 x i16> %a, <8 x i16> %b) { 271 ; CHECK-SSE2-LABEL: @shuffle_v8i16_4563XXXX 272 ; CHECK-SSE2: # BB#0: 273 ; CHECK-SSE2-NEXT: pshufd {{.*}} # xmm0 = xmm0[3,1,2,0] 274 ; CHECK-SSE2-NEXT: pshuflw {{.*}} # xmm0 = xmm0[0,3,2,3,4,5,6,7] 275 ; CHECK-SSE2-NEXT: pshufd {{.*}} # xmm0 = xmm0[2,0,2,3] 276 ; CHECK-SSE2-NEXT: retq 277 %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 4, i32 5, i32 6, i32 3, i32 undef, i32 undef, i32 undef, i32 undef> 278 ret <8 x i16> %shuffle 279 } 280 281 define <8 x i16> @shuffle_v8i16_01274563(<8 x i16> %a, <8 x i16> %b) { 282 ; CHECK-SSE2-LABEL: @shuffle_v8i16_01274563 283 ; CHECK-SSE2: # BB#0: 284 ; CHECK-SSE2-NEXT: pshufd {{.*}} # xmm0 = xmm0[0,2,1,3] 285 ; CHECK-SSE2-NEXT: pshufhw {{.*}} # xmm0 = xmm0[0,1,2,3,6,5,4,7] 286 ; CHECK-SSE2-NEXT: pshufd {{.*}} # xmm0 = xmm0[0,3,1,2] 287 ; CHECK-SSE2-NEXT: retq 288 %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 0, i32 1, i32 2, i32 7, i32 4, i32 5, i32 6, i32 3> 289 ret <8 x i16> %shuffle 290 } 291 292 define <8 x i16> @shuffle_v8i16_45630127(<8 x i16> %a, <8 x i16> %b) { 293 ; CHECK-SSE2-LABEL: @shuffle_v8i16_45630127 294 ; CHECK-SSE2: # BB#0: 295 ; CHECK-SSE2-NEXT: pshufd {{.*}} # xmm0 = xmm0[3,1,2,0] 296 ; CHECK-SSE2-NEXT: pshuflw {{.*}} # xmm0 = xmm0[0,3,1,2,4,5,6,7] 297 ; CHECK-SSE2-NEXT: pshufd {{.*}} # xmm0 = xmm0[2,0,1,3] 298 ; CHECK-SSE2-NEXT: pshufhw {{.*}} # xmm0 = xmm0[0,1,2,3,6,7,5,4] 299 ; CHECK-SSE2-NEXT: retq 300 %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 4, i32 5, i32 6, i32 3, i32 0, i32 1, i32 2, i32 7> 301 ret <8 x i16> %shuffle 302 } 303 304 define <8 x i16> @shuffle_v8i16_08192a3b(<8 x i16> %a, <8 x i16> %b) { 305 ; CHECK-SSE2-LABEL: @shuffle_v8i16_08192a3b 306 ; CHECK-SSE2: # BB#0: 307 ; CHECK-SSE2-NEXT: punpcklwd %xmm1, %xmm0 308 ; CHECK-SSE2-NEXT: retq 309 %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11> 310 ret <8 x i16> %shuffle 311 } 312 313 define <8 x i16> @shuffle_v8i16_0c1d2e3f(<8 x i16> %a, <8 x i16> %b) { 314 ; CHECK-SSE2-LABEL: @shuffle_v8i16_0c1d2e3f 315 ; CHECK-SSE2: # BB#0: 316 ; CHECK-SSE2-NEXT: pshufd {{.*}} # xmm1 = xmm1[2,3,2,3] 317 ; CHECK-SSE2-NEXT: punpcklwd %xmm1, %xmm0 318 ; CHECK-SSE2-NEXT: retq 319 %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 0, i32 12, i32 1, i32 13, i32 2, i32 14, i32 3, i32 15> 320 ret <8 x i16> %shuffle 321 } 322 323 define <8 x i16> @shuffle_v8i16_4c5d6e7f(<8 x i16> %a, <8 x i16> %b) { 324 ; CHECK-SSE2-LABEL: @shuffle_v8i16_4c5d6e7f 325 ; CHECK-SSE2: # BB#0: 326 ; CHECK-SSE2-NEXT: pshufd {{.*}} # xmm1 = xmm1[2,3,2,3] 327 ; CHECK-SSE2-NEXT: pshufd {{.*}} # xmm0 = xmm0[2,3,2,3] 328 ; CHECK-SSE2-NEXT: punpcklwd %xmm1, %xmm0 329 ; CHECK-SSE2-NEXT: retq 330 %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15> 331 ret <8 x i16> %shuffle 332 } 333 334 define <8 x i16> @shuffle_v8i16_48596a7b(<8 x i16> %a, <8 x i16> %b) { 335 ; CHECK-SSE2-LABEL: @shuffle_v8i16_48596a7b 336 ; CHECK-SSE2: # BB#0: 337 ; CHECK-SSE2-NEXT: pshufd {{.*}} # xmm0 = xmm0[2,3,2,3] 338 ; CHECK-SSE2-NEXT: punpcklwd %xmm1, %xmm0 339 ; CHECK-SSE2-NEXT: retq 340 %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 4, i32 8, i32 5, i32 9, i32 6, i32 10, i32 7, i32 11> 341 ret <8 x i16> %shuffle 342 } 343 344 define <8 x i16> @shuffle_v8i16_08196e7f(<8 x i16> %a, <8 x i16> %b) { 345 ; CHECK-SSE2-LABEL: @shuffle_v8i16_08196e7f 346 ; CHECK-SSE2: # BB#0: 347 ; CHECK-SSE2-NEXT: pshufd {{.*}} # xmm1 = xmm1[0,3,2,3] 348 ; CHECK-SSE2-NEXT: pshufd {{.*}} # xmm0 = xmm0[0,3,2,3] 349 ; CHECK-SSE2-NEXT: punpcklwd %xmm1, %xmm0 350 ; CHECK-SSE2-NEXT: retq 351 %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 0, i32 8, i32 1, i32 9, i32 6, i32 14, i32 7, i32 15> 352 ret <8 x i16> %shuffle 353 } 354 355 define <8 x i16> @shuffle_v8i16_0c1d6879(<8 x i16> %a, <8 x i16> %b) { 356 ; CHECK-SSE2-LABEL: @shuffle_v8i16_0c1d6879 357 ; CHECK-SSE2: # BB#0: 358 ; CHECK-SSE2-NEXT: pshufd {{.*}} # xmm1 = xmm1[2,0,2,3] 359 ; CHECK-SSE2-NEXT: pshufd {{.*}} # xmm0 = xmm0[0,3,2,3] 360 ; CHECK-SSE2-NEXT: punpcklwd %xmm1, %xmm0 361 ; CHECK-SSE2-NEXT: retq 362 %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 0, i32 12, i32 1, i32 13, i32 6, i32 8, i32 7, i32 9> 363 ret <8 x i16> %shuffle 364 } 365 366 define <8 x i16> @shuffle_v8i16_109832ba(<8 x i16> %a, <8 x i16> %b) { 367 ; CHECK-SSE2-LABEL: @shuffle_v8i16_109832ba 368 ; CHECK-SSE2: # BB#0: 369 ; CHECK-SSE2-NEXT: punpcklwd %xmm1, %xmm0 370 ; CHECK-SSE2-NEXT: pshuflw {{.*}} # xmm1 = xmm0[2,0,3,1,4,5,6,7] 371 ; CHECK-SSE2-NEXT: pshufd {{.*}} # xmm0 = xmm0[2,3,2,3] 372 ; CHECK-SSE2-NEXT: pshuflw {{.*}} # xmm0 = xmm0[2,0,3,1,4,5,6,7] 373 ; CHECK-SSE2-NEXT: punpcklqdq %xmm0, %xmm1 374 ; CHECK-SSE2-NEXT: movdqa %xmm1, %xmm0 375 ; CHECK-SSE2-NEXT: retq 376 %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 1, i32 0, i32 9, i32 8, i32 3, i32 2, i32 11, i32 10> 377 ret <8 x i16> %shuffle 378 } 379 380 define <8 x i16> @shuffle_v8i16_8091a2b3(<8 x i16> %a, <8 x i16> %b) { 381 ; CHECK-SSE2-LABEL: @shuffle_v8i16_8091a2b3 382 ; CHECK-SSE2: # BB#0: 383 ; CHECK-SSE2-NEXT: punpcklwd %xmm0, %xmm1 384 ; CHECK-SSE2-NEXT: movdqa %xmm1, %xmm0 385 ; CHECK-SSE2-NEXT: retq 386 %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 8, i32 0, i32 9, i32 1, i32 10, i32 2, i32 11, i32 3> 387 ret <8 x i16> %shuffle 388 } 389 define <8 x i16> @shuffle_v8i16_c4d5e6f7(<8 x i16> %a, <8 x i16> %b) { 390 ; CHECK-SSE2-LABEL: @shuffle_v8i16_c4d5e6f7 391 ; CHECK-SSE2: # BB#0: 392 ; CHECK-SSE2-NEXT: pshufd {{.*}} # xmm2 = xmm0[2,3,2,3] 393 ; CHECK-SSE2-NEXT: pshufd {{.*}} # xmm0 = xmm1[2,3,2,3] 394 ; CHECK-SSE2-NEXT: punpcklwd %xmm2, %xmm0 395 ; CHECK-SSE2-NEXT: retq 396 %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 12, i32 4, i32 13, i32 5, i32 14, i32 6, i32 15, i32 7> 397 ret <8 x i16> %shuffle 398 } 399 400 define <8 x i16> @shuffle_v8i16_0213cedf(<8 x i16> %a, <8 x i16> %b) { 401 ; CHECK-SSE2-LABEL: @shuffle_v8i16_0213cedf 402 ; CHECK-SSE2: # BB#0: 403 ; CHECK-SSE2-NEXT: pshuflw {{.*}} # xmm0 = xmm0[0,2,1,3,4,5,6,7] 404 ; CHECK-SSE2-NEXT: pshufd {{.*}} # xmm1 = xmm1[2,3,2,3] 405 ; CHECK-SSE2-NEXT: pshuflw {{.*}} # xmm1 = xmm1[0,2,1,3,4,5,6,7] 406 ; CHECK-SSE2-NEXT: punpcklqdq %xmm1, %xmm0 407 ; CHECK-SSE2-NEXT: retq 408 %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 0, i32 2, i32 1, i32 3, i32 12, i32 14, i32 13, i32 15> 409 ret <8 x i16> %shuffle 410 } 411 412 define <8 x i16> @shuffle_v8i16_032dXXXX(<8 x i16> %a, <8 x i16> %b) { 413 ; CHECK-SSE2-LABEL: @shuffle_v8i16_032dXXXX 414 ; CHECK-SSE2: # BB#0: 415 ; CHECK-SSE2-NEXT: pshufd {{.*}} # xmm1 = xmm1[2,1,2,3] 416 ; CHECK-SSE2-NEXT: punpcklwd %xmm1, %xmm0 417 ; CHECK-SSE2-NEXT: pshuflw {{.*}} # xmm0 = xmm0[0,3,2,3,4,5,6,7] 418 ; CHECK-SSE2-NEXT: pshufhw {{.*}} # xmm0 = xmm0[0,1,2,3,4,6,6,7] 419 ; CHECK-SSE2-NEXT: pshufd {{.*}} # xmm0 = xmm0[0,2,2,3] 420 ; CHECK-SSE2-NEXT: pshuflw {{.*}} # xmm0 = xmm0[0,3,2,1,4,5,6,7] 421 ; CHECK-SSE2-NEXT: retq 422 %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 0, i32 3, i32 2, i32 13, i32 undef, i32 undef, i32 undef, i32 undef> 423 ret <8 x i16> %shuffle 424 } 425 define <8 x i16> @shuffle_v8i16_XXXcXXXX(<8 x i16> %a, <8 x i16> %b) { 426 ; CHECK-SSE2-LABEL: @shuffle_v8i16_XXXcXXXX 427 ; CHECK-SSE2: # BB#0: 428 ; CHECK-SSE2-NEXT: pshufd {{.*}} # xmm0 = xmm1[2,1,2,3] 429 ; CHECK-SSE2-NEXT: pshuflw {{.*}} # xmm0 = xmm0[0,1,2,1,4,5,6,7] 430 ; CHECK-SSE2-NEXT: retq 431 %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 undef, i32 undef, i32 undef, i32 13, i32 undef, i32 undef, i32 undef, i32 undef> 432 ret <8 x i16> %shuffle 433 } 434 435 define <8 x i16> @shuffle_v8i16_012dXXXX(<8 x i16> %a, <8 x i16> %b) { 436 ; CHECK-SSE2-LABEL: @shuffle_v8i16_012dXXXX 437 ; CHECK-SSE2: # BB#0: 438 ; CHECK-SSE2-NEXT: pshufd {{.*}} # xmm1 = xmm1[2,1,2,3] 439 ; CHECK-SSE2-NEXT: punpcklwd %xmm1, %xmm0 440 ; CHECK-SSE2-NEXT: pshufd {{.*}} # xmm0 = xmm0[3,1,2,0] 441 ; CHECK-SSE2-NEXT: pshufhw {{.*}} # xmm0 = xmm0[0,1,2,3,4,6,6,7] 442 ; CHECK-SSE2-NEXT: pshufd {{.*}} # xmm0 = xmm0[2,1,2,3] 443 ; CHECK-SSE2-NEXT: pshuflw {{.*}} # xmm0 = xmm0[1,2,0,3,4,5,6,7] 444 ; CHECK-SSE2-NEXT: retq 445 %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 0, i32 1, i32 2, i32 13, i32 undef, i32 undef, i32 undef, i32 undef> 446 ret <8 x i16> %shuffle 447 } 448 449 define <8 x i16> @shuffle_v8i16_XXXXcde3(<8 x i16> %a, <8 x i16> %b) { 450 ; CHECK-SSE2-LABEL: @shuffle_v8i16_XXXXcde3 451 ; CHECK-SSE2: # BB#0: 452 ; CHECK-SSE2-NEXT: pshufd {{.*}} # xmm0 = xmm0[0,1,2,1] 453 ; CHECK-SSE2-NEXT: punpckhwd %xmm0, %xmm1 454 ; CHECK-SSE2-NEXT: pshuflw {{.*}} # xmm0 = xmm1[0,2,2,3,4,5,6,7] 455 ; CHECK-SSE2-NEXT: pshufhw {{.*}} # xmm0 = xmm0[0,1,2,3,4,7,6,7] 456 ; CHECK-SSE2-NEXT: pshufd {{.*}} # xmm0 = xmm0[0,1,0,2] 457 ; CHECK-SSE2-NEXT: retq 458 %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 undef, i32 undef, i32 undef, i32 undef, i32 12, i32 13, i32 14, i32 3> 459 ret <8 x i16> %shuffle 460 } 461 462 define <8 x i16> @shuffle_v8i16_cde3XXXX(<8 x i16> %a, <8 x i16> %b) { 463 ; CHECK-SSE2-LABEL: @shuffle_v8i16_cde3XXXX 464 ; CHECK-SSE2: # BB#0: 465 ; CHECK-SSE2-NEXT: pshufd {{.*}} # xmm0 = xmm0[0,1,2,1] 466 ; CHECK-SSE2-NEXT: punpckhwd %xmm0, %xmm1 467 ; CHECK-SSE2-NEXT: pshuflw {{.*}} # xmm0 = xmm1[0,2,2,3,4,5,6,7] 468 ; CHECK-SSE2-NEXT: pshufhw {{.*}} # xmm0 = xmm0[0,1,2,3,4,7,6,7] 469 ; CHECK-SSE2-NEXT: pshufd {{.*}} # xmm0 = xmm0[0,2,2,3] 470 ; CHECK-SSE2-NEXT: retq 471 %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 12, i32 13, i32 14, i32 3, i32 undef, i32 undef, i32 undef, i32 undef> 472 ret <8 x i16> %shuffle 473 } 474 475 define <8 x i16> @shuffle_v8i16_012dcde3(<8 x i16> %a, <8 x i16> %b) { 476 ; CHECK-SSE2-LABEL: @shuffle_v8i16_012dcde3 477 ; CHECK-SSE2: # BB#0: 478 ; CHECK-SSE2-NEXT: pshufd {{.*}} # xmm2 = xmm0[0,1,2,1] 479 ; CHECK-SSE2-NEXT: pshufd {{.*}} # xmm3 = xmm1[2,1,2,3] 480 ; CHECK-SSE2-NEXT: punpckhwd %xmm2, %xmm1 481 ; CHECK-SSE2-NEXT: pshuflw {{.*}} # xmm1 = xmm1[0,2,2,3,4,5,6,7] 482 ; CHECK-SSE2-NEXT: pshufhw {{.*}} # xmm1 = xmm1[0,1,2,3,4,7,6,7] 483 ; CHECK-SSE2-NEXT: pshufd {{.*}} # xmm1 = xmm1[0,2,2,3] 484 ; CHECK-SSE2-NEXT: punpcklwd %xmm3, %xmm0 485 ; CHECK-SSE2-NEXT: pshufd {{.*}} # xmm0 = xmm0[3,1,2,0] 486 ; CHECK-SSE2-NEXT: pshufhw {{.*}} # xmm0 = xmm0[0,1,2,3,4,6,6,7] 487 ; CHECK-SSE2-NEXT: pshufd {{.*}} # xmm0 = xmm0[2,1,2,3] 488 ; CHECK-SSE2-NEXT: pshuflw {{.*}} # xmm0 = xmm0[1,2,0,3,4,5,6,7] 489 ; CHECK-SSE2-NEXT: punpcklqdq %xmm1, %xmm0 490 ; CHECK-SSE2-NEXT: retq 491 %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 0, i32 1, i32 2, i32 13, i32 12, i32 13, i32 14, i32 3> 492 ret <8 x i16> %shuffle 493 } 494
