horizontal-reduce-add.ll 16.4 KB
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+sse2            | FileCheck %s --check-prefixes=SSE,SSE2
; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+ssse3           | FileCheck %s --check-prefixes=SSE,SSSE3,SSSE3-SLOW
; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+ssse3,fast-hops | FileCheck %s --check-prefixes=SSE,SSSE3,SSSE3-FAST
; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+avx             | FileCheck %s --check-prefixes=AVX,AVX1,AVX1-SLOW
; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+avx,fast-hops   | FileCheck %s --check-prefixes=AVX,AVX1,AVX1-FAST
; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+avx2            | FileCheck %s --check-prefixes=AVX,AVX2

; PR37890 - subvector reduction followed by shuffle reduction

define i32 @PR37890_v4i32(<4 x i32> %a)  {
; SSE2-LABEL: PR37890_v4i32:
; SSE2:       # %bb.0:
; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
; SSE2-NEXT:    paddd %xmm0, %xmm1
; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[1,1,2,3]
; SSE2-NEXT:    paddd %xmm1, %xmm0
; SSE2-NEXT:    movd %xmm0, %eax
; SSE2-NEXT:    retq
;
; SSSE3-SLOW-LABEL: PR37890_v4i32:
; SSSE3-SLOW:       # %bb.0:
; SSSE3-SLOW-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
; SSSE3-SLOW-NEXT:    paddd %xmm0, %xmm1
; SSSE3-SLOW-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[1,1,2,3]
; SSSE3-SLOW-NEXT:    paddd %xmm1, %xmm0
; SSSE3-SLOW-NEXT:    movd %xmm0, %eax
; SSSE3-SLOW-NEXT:    retq
;
; SSSE3-FAST-LABEL: PR37890_v4i32:
; SSSE3-FAST:       # %bb.0:
; SSSE3-FAST-NEXT:    phaddd %xmm0, %xmm0
; SSSE3-FAST-NEXT:    phaddd %xmm0, %xmm0
; SSSE3-FAST-NEXT:    movd %xmm0, %eax
; SSSE3-FAST-NEXT:    retq
;
; AVX1-SLOW-LABEL: PR37890_v4i32:
; AVX1-SLOW:       # %bb.0:
; AVX1-SLOW-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
; AVX1-SLOW-NEXT:    vpaddd %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
; AVX1-SLOW-NEXT:    vpaddd %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT:    vmovd %xmm0, %eax
; AVX1-SLOW-NEXT:    retq
;
; AVX1-FAST-LABEL: PR37890_v4i32:
; AVX1-FAST:       # %bb.0:
; AVX1-FAST-NEXT:    vphaddd %xmm0, %xmm0, %xmm0
; AVX1-FAST-NEXT:    vphaddd %xmm0, %xmm0, %xmm0
; AVX1-FAST-NEXT:    vmovd %xmm0, %eax
; AVX1-FAST-NEXT:    retq
;
; AVX2-LABEL: PR37890_v4i32:
; AVX2:       # %bb.0:
; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
; AVX2-NEXT:    vpaddd %xmm1, %xmm0, %xmm0
; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
; AVX2-NEXT:    vpaddd %xmm1, %xmm0, %xmm0
; AVX2-NEXT:    vmovd %xmm0, %eax
; AVX2-NEXT:    retq
  %hi0 = shufflevector <4 x i32> %a, <4 x i32> undef, <2 x i32> <i32 2, i32 3>
  %lo0 = shufflevector <4 x i32> %a, <4 x i32> undef, <2 x i32> <i32 0, i32 1>
  %sum0 = add <2 x i32> %lo0, %hi0
  %hi1 = shufflevector <2 x i32> %sum0, <2 x i32> undef, <2 x i32> <i32 1, i32 undef>
  %sum1 = add <2 x i32> %sum0, %hi1
  %e = extractelement <2 x i32> %sum1, i32 0
  ret i32 %e
}

define i16 @PR37890_v8i16(<8 x i16> %a)  {
; SSE2-LABEL: PR37890_v8i16:
; SSE2:       # %bb.0:
; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
; SSE2-NEXT:    paddw %xmm0, %xmm1
; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[1,1,2,3]
; SSE2-NEXT:    paddw %xmm1, %xmm0
; SSE2-NEXT:    movdqa %xmm0, %xmm1
; SSE2-NEXT:    psrld $16, %xmm1
; SSE2-NEXT:    paddw %xmm0, %xmm1
; SSE2-NEXT:    movd %xmm1, %eax
; SSE2-NEXT:    # kill: def $ax killed $ax killed $eax
; SSE2-NEXT:    retq
;
; SSSE3-SLOW-LABEL: PR37890_v8i16:
; SSSE3-SLOW:       # %bb.0:
; SSSE3-SLOW-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
; SSSE3-SLOW-NEXT:    paddw %xmm0, %xmm1
; SSSE3-SLOW-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[1,1,2,3]
; SSSE3-SLOW-NEXT:    paddw %xmm1, %xmm0
; SSSE3-SLOW-NEXT:    movdqa %xmm0, %xmm1
; SSSE3-SLOW-NEXT:    psrld $16, %xmm1
; SSSE3-SLOW-NEXT:    paddw %xmm0, %xmm1
; SSSE3-SLOW-NEXT:    movd %xmm1, %eax
; SSSE3-SLOW-NEXT:    # kill: def $ax killed $ax killed $eax
; SSSE3-SLOW-NEXT:    retq
;
; SSSE3-FAST-LABEL: PR37890_v8i16:
; SSSE3-FAST:       # %bb.0:
; SSSE3-FAST-NEXT:    phaddw %xmm0, %xmm0
; SSSE3-FAST-NEXT:    phaddw %xmm0, %xmm0
; SSSE3-FAST-NEXT:    phaddw %xmm0, %xmm0
; SSSE3-FAST-NEXT:    movd %xmm0, %eax
; SSSE3-FAST-NEXT:    # kill: def $ax killed $ax killed $eax
; SSSE3-FAST-NEXT:    retq
;
; AVX1-SLOW-LABEL: PR37890_v8i16:
; AVX1-SLOW:       # %bb.0:
; AVX1-SLOW-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
; AVX1-SLOW-NEXT:    vpaddw %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
; AVX1-SLOW-NEXT:    vpaddw %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT:    vpsrld $16, %xmm0, %xmm1
; AVX1-SLOW-NEXT:    vpaddw %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT:    vmovd %xmm0, %eax
; AVX1-SLOW-NEXT:    # kill: def $ax killed $ax killed $eax
; AVX1-SLOW-NEXT:    retq
;
; AVX1-FAST-LABEL: PR37890_v8i16:
; AVX1-FAST:       # %bb.0:
; AVX1-FAST-NEXT:    vphaddw %xmm0, %xmm0, %xmm0
; AVX1-FAST-NEXT:    vphaddw %xmm0, %xmm0, %xmm0
; AVX1-FAST-NEXT:    vphaddw %xmm0, %xmm0, %xmm0
; AVX1-FAST-NEXT:    vmovd %xmm0, %eax
; AVX1-FAST-NEXT:    # kill: def $ax killed $ax killed $eax
; AVX1-FAST-NEXT:    retq
;
; AVX2-LABEL: PR37890_v8i16:
; AVX2:       # %bb.0:
; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
; AVX2-NEXT:    vpaddw %xmm1, %xmm0, %xmm0
; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
; AVX2-NEXT:    vpaddw %xmm1, %xmm0, %xmm0
; AVX2-NEXT:    vpsrld $16, %xmm0, %xmm1
; AVX2-NEXT:    vpaddw %xmm1, %xmm0, %xmm0
; AVX2-NEXT:    vmovd %xmm0, %eax
; AVX2-NEXT:    # kill: def $ax killed $ax killed $eax
; AVX2-NEXT:    retq
  %hi0 = shufflevector <8 x i16> %a, <8 x i16> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
  %lo0 = shufflevector <8 x i16> %a, <8 x i16> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %sum0 = add <4 x i16> %lo0, %hi0
  %hi1 = shufflevector <4 x i16> %sum0, <4 x i16> undef, <2 x i32> <i32 2, i32 3>
  %lo1 = shufflevector <4 x i16> %sum0, <4 x i16> undef, <2 x i32> <i32 0, i32 1>
  %sum1 = add <2 x i16> %lo1, %hi1
  %hi2 = shufflevector <2 x i16> %sum1, <2 x i16> undef, <2 x i32> <i32 1, i32 undef>
  %sum2 = add <2 x i16> %sum1, %hi2
  %e = extractelement <2 x i16> %sum2, i32 0
  ret i16 %e
}

define i32 @PR37890_v8i32(<8 x i32> %a)  {
; SSE2-LABEL: PR37890_v8i32:
; SSE2:       # %bb.0:
; SSE2-NEXT:    paddd %xmm1, %xmm0
; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
; SSE2-NEXT:    paddd %xmm0, %xmm1
; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[1,1,2,3]
; SSE2-NEXT:    paddd %xmm1, %xmm0
; SSE2-NEXT:    movd %xmm0, %eax
; SSE2-NEXT:    retq
;
; SSSE3-SLOW-LABEL: PR37890_v8i32:
; SSSE3-SLOW:       # %bb.0:
; SSSE3-SLOW-NEXT:    paddd %xmm1, %xmm0
; SSSE3-SLOW-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
; SSSE3-SLOW-NEXT:    paddd %xmm0, %xmm1
; SSSE3-SLOW-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[1,1,2,3]
; SSSE3-SLOW-NEXT:    paddd %xmm1, %xmm0
; SSSE3-SLOW-NEXT:    movd %xmm0, %eax
; SSSE3-SLOW-NEXT:    retq
;
; SSSE3-FAST-LABEL: PR37890_v8i32:
; SSSE3-FAST:       # %bb.0:
; SSSE3-FAST-NEXT:    paddd %xmm1, %xmm0
; SSSE3-FAST-NEXT:    phaddd %xmm0, %xmm0
; SSSE3-FAST-NEXT:    phaddd %xmm0, %xmm0
; SSSE3-FAST-NEXT:    movd %xmm0, %eax
; SSSE3-FAST-NEXT:    retq
;
; AVX1-SLOW-LABEL: PR37890_v8i32:
; AVX1-SLOW:       # %bb.0:
; AVX1-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm1
; AVX1-SLOW-NEXT:    vpaddd %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
; AVX1-SLOW-NEXT:    vpaddd %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
; AVX1-SLOW-NEXT:    vpaddd %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT:    vmovd %xmm0, %eax
; AVX1-SLOW-NEXT:    vzeroupper
; AVX1-SLOW-NEXT:    retq
;
; AVX1-FAST-LABEL: PR37890_v8i32:
; AVX1-FAST:       # %bb.0:
; AVX1-FAST-NEXT:    vextractf128 $1, %ymm0, %xmm1
; AVX1-FAST-NEXT:    vphaddd %xmm0, %xmm1, %xmm0
; AVX1-FAST-NEXT:    vphaddd %xmm0, %xmm0, %xmm0
; AVX1-FAST-NEXT:    vphaddd %xmm0, %xmm0, %xmm0
; AVX1-FAST-NEXT:    vmovd %xmm0, %eax
; AVX1-FAST-NEXT:    vzeroupper
; AVX1-FAST-NEXT:    retq
;
; AVX2-LABEL: PR37890_v8i32:
; AVX2:       # %bb.0:
; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
; AVX2-NEXT:    vpaddd %xmm1, %xmm0, %xmm0
; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
; AVX2-NEXT:    vpaddd %xmm1, %xmm0, %xmm0
; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
; AVX2-NEXT:    vpaddd %xmm1, %xmm0, %xmm0
; AVX2-NEXT:    vmovd %xmm0, %eax
; AVX2-NEXT:    vzeroupper
; AVX2-NEXT:    retq
  %hi0 = shufflevector <8 x i32> %a, <8 x i32> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
  %lo0 = shufflevector <8 x i32> %a, <8 x i32> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %sum0 = add <4 x i32> %lo0, %hi0
  %hi1 = shufflevector <4 x i32> %sum0, <4 x i32> undef, <2 x i32> <i32 2, i32 3>
  %lo1 = shufflevector <4 x i32> %sum0, <4 x i32> undef, <2 x i32> <i32 0, i32 1>
  %sum1 = add <2 x i32> %lo1, %hi1
  %hi2 = shufflevector <2 x i32> %sum1, <2 x i32> undef, <2 x i32> <i32 1, i32 undef>
  %sum2 = add <2 x i32> %sum1, %hi2
  %e = extractelement <2 x i32> %sum2, i32 0
  ret i32 %e
}

define i16 @PR37890_v16i16(<16 x i16> %a)  {
; SSE2-LABEL: PR37890_v16i16:
; SSE2:       # %bb.0:
; SSE2-NEXT:    paddw %xmm1, %xmm0
; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
; SSE2-NEXT:    paddw %xmm0, %xmm1
; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[1,1,2,3]
; SSE2-NEXT:    paddw %xmm1, %xmm0
; SSE2-NEXT:    movdqa %xmm0, %xmm1
; SSE2-NEXT:    psrld $16, %xmm1
; SSE2-NEXT:    paddw %xmm0, %xmm1
; SSE2-NEXT:    movd %xmm1, %eax
; SSE2-NEXT:    # kill: def $ax killed $ax killed $eax
; SSE2-NEXT:    retq
;
; SSSE3-SLOW-LABEL: PR37890_v16i16:
; SSSE3-SLOW:       # %bb.0:
; SSSE3-SLOW-NEXT:    paddw %xmm1, %xmm0
; SSSE3-SLOW-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
; SSSE3-SLOW-NEXT:    paddw %xmm0, %xmm1
; SSSE3-SLOW-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[1,1,2,3]
; SSSE3-SLOW-NEXT:    paddw %xmm1, %xmm0
; SSSE3-SLOW-NEXT:    movdqa %xmm0, %xmm1
; SSSE3-SLOW-NEXT:    psrld $16, %xmm1
; SSSE3-SLOW-NEXT:    paddw %xmm0, %xmm1
; SSSE3-SLOW-NEXT:    movd %xmm1, %eax
; SSSE3-SLOW-NEXT:    # kill: def $ax killed $ax killed $eax
; SSSE3-SLOW-NEXT:    retq
;
; SSSE3-FAST-LABEL: PR37890_v16i16:
; SSSE3-FAST:       # %bb.0:
; SSSE3-FAST-NEXT:    paddw %xmm1, %xmm0
; SSSE3-FAST-NEXT:    phaddw %xmm0, %xmm0
; SSSE3-FAST-NEXT:    phaddw %xmm0, %xmm0
; SSSE3-FAST-NEXT:    phaddw %xmm0, %xmm0
; SSSE3-FAST-NEXT:    movd %xmm0, %eax
; SSSE3-FAST-NEXT:    # kill: def $ax killed $ax killed $eax
; SSSE3-FAST-NEXT:    retq
;
; AVX1-SLOW-LABEL: PR37890_v16i16:
; AVX1-SLOW:       # %bb.0:
; AVX1-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm1
; AVX1-SLOW-NEXT:    vpaddw %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
; AVX1-SLOW-NEXT:    vpaddw %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
; AVX1-SLOW-NEXT:    vpaddw %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT:    vpsrld $16, %xmm0, %xmm1
; AVX1-SLOW-NEXT:    vpaddw %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT:    vmovd %xmm0, %eax
; AVX1-SLOW-NEXT:    # kill: def $ax killed $ax killed $eax
; AVX1-SLOW-NEXT:    vzeroupper
; AVX1-SLOW-NEXT:    retq
;
; AVX1-FAST-LABEL: PR37890_v16i16:
; AVX1-FAST:       # %bb.0:
; AVX1-FAST-NEXT:    vextractf128 $1, %ymm0, %xmm1
; AVX1-FAST-NEXT:    vphaddw %xmm0, %xmm1, %xmm0
; AVX1-FAST-NEXT:    vphaddw %xmm0, %xmm0, %xmm0
; AVX1-FAST-NEXT:    vphaddw %xmm0, %xmm0, %xmm0
; AVX1-FAST-NEXT:    vphaddw %xmm0, %xmm0, %xmm0
; AVX1-FAST-NEXT:    vmovd %xmm0, %eax
; AVX1-FAST-NEXT:    # kill: def $ax killed $ax killed $eax
; AVX1-FAST-NEXT:    vzeroupper
; AVX1-FAST-NEXT:    retq
;
; AVX2-LABEL: PR37890_v16i16:
; AVX2:       # %bb.0:
; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
; AVX2-NEXT:    vpaddw %xmm1, %xmm0, %xmm0
; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
; AVX2-NEXT:    vpaddw %xmm1, %xmm0, %xmm0
; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
; AVX2-NEXT:    vpaddw %xmm1, %xmm0, %xmm0
; AVX2-NEXT:    vpsrld $16, %xmm0, %xmm1
; AVX2-NEXT:    vpaddw %xmm1, %xmm0, %xmm0
; AVX2-NEXT:    vmovd %xmm0, %eax
; AVX2-NEXT:    # kill: def $ax killed $ax killed $eax
; AVX2-NEXT:    vzeroupper
; AVX2-NEXT:    retq
  %hi0 = shufflevector <16 x i16> %a, <16 x i16> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
  %lo0 = shufflevector <16 x i16> %a, <16 x i16> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
  %sum0 = add <8 x i16> %lo0, %hi0
  %hi1 = shufflevector <8 x i16> %sum0, <8 x i16> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
  %lo1 = shufflevector <8 x i16> %sum0, <8 x i16> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %sum1 = add <4 x i16> %lo1, %hi1
  %hi2 = shufflevector <4 x i16> %sum1, <4 x i16> undef, <2 x i32> <i32 2, i32 3>
  %lo2 = shufflevector <4 x i16> %sum1, <4 x i16> undef, <2 x i32> <i32 0, i32 1>
  %sum2 = add <2 x i16> %lo2, %hi2
  %hi3 = shufflevector <2 x i16> %sum2, <2 x i16> undef, <2 x i32> <i32 1, i32 undef>
  %sum3 = add <2 x i16> %sum2, %hi3
  %e = extractelement <2 x i16> %sum3, i32 0
  ret i16 %e
}

define i32 @PR37890_v16i32(<16 x i32> %a)  {
; SSE2-LABEL: PR37890_v16i32:
; SSE2:       # %bb.0:
; SSE2-NEXT:    paddd %xmm3, %xmm1
; SSE2-NEXT:    paddd %xmm2, %xmm1
; SSE2-NEXT:    paddd %xmm0, %xmm1
; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[2,3,0,1]
; SSE2-NEXT:    paddd %xmm1, %xmm0
; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
; SSE2-NEXT:    paddd %xmm0, %xmm1
; SSE2-NEXT:    movd %xmm1, %eax
; SSE2-NEXT:    retq
;
; SSSE3-SLOW-LABEL: PR37890_v16i32:
; SSSE3-SLOW:       # %bb.0:
; SSSE3-SLOW-NEXT:    paddd %xmm3, %xmm1
; SSSE3-SLOW-NEXT:    paddd %xmm2, %xmm1
; SSSE3-SLOW-NEXT:    paddd %xmm0, %xmm1
; SSSE3-SLOW-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[2,3,0,1]
; SSSE3-SLOW-NEXT:    paddd %xmm1, %xmm0
; SSSE3-SLOW-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
; SSSE3-SLOW-NEXT:    paddd %xmm0, %xmm1
; SSSE3-SLOW-NEXT:    movd %xmm1, %eax
; SSSE3-SLOW-NEXT:    retq
;
; SSSE3-FAST-LABEL: PR37890_v16i32:
; SSSE3-FAST:       # %bb.0:
; SSSE3-FAST-NEXT:    paddd %xmm3, %xmm1
; SSSE3-FAST-NEXT:    paddd %xmm2, %xmm1
; SSSE3-FAST-NEXT:    paddd %xmm0, %xmm1
; SSSE3-FAST-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[2,3,0,1]
; SSSE3-FAST-NEXT:    paddd %xmm1, %xmm0
; SSSE3-FAST-NEXT:    phaddd %xmm0, %xmm0
; SSSE3-FAST-NEXT:    movd %xmm0, %eax
; SSSE3-FAST-NEXT:    retq
;
; AVX1-SLOW-LABEL: PR37890_v16i32:
; AVX1-SLOW:       # %bb.0:
; AVX1-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm2
; AVX1-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm3
; AVX1-SLOW-NEXT:    vpaddd %xmm2, %xmm3, %xmm2
; AVX1-SLOW-NEXT:    vpaddd %xmm2, %xmm1, %xmm1
; AVX1-SLOW-NEXT:    vpaddd %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
; AVX1-SLOW-NEXT:    vpaddd %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
; AVX1-SLOW-NEXT:    vpaddd %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT:    vmovd %xmm0, %eax
; AVX1-SLOW-NEXT:    vzeroupper
; AVX1-SLOW-NEXT:    retq
;
; AVX1-FAST-LABEL: PR37890_v16i32:
; AVX1-FAST:       # %bb.0:
; AVX1-FAST-NEXT:    vpaddd %xmm1, %xmm0, %xmm2
; AVX1-FAST-NEXT:    vextractf128 $1, %ymm1, %xmm1
; AVX1-FAST-NEXT:    vextractf128 $1, %ymm0, %xmm0
; AVX1-FAST-NEXT:    vpaddd %xmm1, %xmm0, %xmm0
; AVX1-FAST-NEXT:    vphaddd %xmm2, %xmm0, %xmm0
; AVX1-FAST-NEXT:    vphaddd %xmm0, %xmm0, %xmm0
; AVX1-FAST-NEXT:    vphaddd %xmm0, %xmm0, %xmm0
; AVX1-FAST-NEXT:    vmovd %xmm0, %eax
; AVX1-FAST-NEXT:    vzeroupper
; AVX1-FAST-NEXT:    retq
;
; AVX2-LABEL: PR37890_v16i32:
; AVX2:       # %bb.0:
; AVX2-NEXT:    vpaddd %ymm1, %ymm0, %ymm0
; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
; AVX2-NEXT:    vpaddd %xmm1, %xmm0, %xmm0
; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
; AVX2-NEXT:    vpaddd %xmm1, %xmm0, %xmm0
; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
; AVX2-NEXT:    vpaddd %xmm1, %xmm0, %xmm0
; AVX2-NEXT:    vmovd %xmm0, %eax
; AVX2-NEXT:    vzeroupper
; AVX2-NEXT:    retq
  %hi0 = shufflevector <16 x i32> %a, <16 x i32> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
  %lo0 = shufflevector <16 x i32> %a, <16 x i32> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
  %sum0 = add <8 x i32> %lo0, %hi0
  %hi1 = shufflevector <8 x i32> %sum0, <8 x i32> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
  %lo1 = shufflevector <8 x i32> %sum0, <8 x i32> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %sum1 = add <4 x i32> %lo1, %hi1
  %hi2 = shufflevector <4 x i32> %sum1, <4 x i32> undef, <2 x i32> <i32 2, i32 3>
  %lo2 = shufflevector <4 x i32> %sum1, <4 x i32> undef, <2 x i32> <i32 0, i32 1>
  %sum2 = add <2 x i32> %lo2, %hi2
  %hi3 = shufflevector <2 x i32> %sum2, <2 x i32> undef, <2 x i32> <i32 1, i32 undef>
  %sum3 = add <2 x i32> %sum2, %hi3
  %e = extractelement <2 x i32> %sum3, i32 0
  ret i32 %e
}