; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=x86_64-- -mattr=+sse2 | FileCheck %s --check-prefixes=SSE
; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx | FileCheck %s --check-prefixes=AVX
; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx2 | FileCheck %s --check-prefixes=AVX2
; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx2,+fast-variable-perlane-shuffle | FileCheck %s --check-prefixes=AVX2-FP
; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx2,+fast-variable-crosslane-shuffle,+fast-variable-perlane-shuffle | FileCheck %s --check-prefixes=AVX2-FCP
; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx512vl | FileCheck %s --check-prefixes=AVX512
; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx512vl,+fast-variable-crosslane-shuffle,+fast-variable-perlane-shuffle | FileCheck %s --check-prefixes=AVX512-FCP
; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx512vl,+avx512dq | FileCheck %s --check-prefixes=AVX512DQ
; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx512vl,+avx512dq,+fast-variable-crosslane-shuffle,+fast-variable-perlane-shuffle | FileCheck %s --check-prefixes=AVX512DQ-FCP
; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx512vl,+avx512bw | FileCheck %s --check-prefixes=AVX512BW
; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx512vl,+avx512bw,+fast-variable-crosslane-shuffle,+fast-variable-perlane-shuffle | FileCheck %s --check-prefixes=AVX512BW-FCP
; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx512vl,+avx512dq,+avx512bw | FileCheck %s --check-prefixes=AVX512DQ-BW
; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx512vl,+avx512dq,+avx512bw,+fast-variable-crosslane-shuffle,+fast-variable-perlane-shuffle | FileCheck %s --check-prefixes=AVX512DQ-BW-FCP
; These patterns are produced by LoopVectorizer for interleaved loads.
define void @load_i16_stride4_vf2(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3) nounwind {
; SSE-LABEL: load_i16_stride4_vf2:
; SSE: # %bb.0:
; SSE-NEXT: movdqa (%rdi), %xmm0
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm1[0,2,2,3,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[1,3,2,3,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[3,1,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm0[2,0,2,3,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[3,1,2,3,4,5,6,7]
; SSE-NEXT: movd %xmm2, (%rsi)
; SSE-NEXT: movd %xmm1, (%rdx)
; SSE-NEXT: movd %xmm3, (%rcx)
; SSE-NEXT: movd %xmm0, (%r8)
; SSE-NEXT: retq
;
; AVX-LABEL: load_i16_stride4_vf2:
; AVX: # %bb.0:
; AVX-NEXT: vmovdqa (%rdi), %xmm0
; AVX-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[0,2,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm2 = xmm1[0,2,2,3,4,5,6,7]
; AVX-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[1,3,2,3,4,5,6,7]
; AVX-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[3,1,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm3 = xmm0[2,0,2,3,4,5,6,7]
; AVX-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[3,1,2,3,4,5,6,7]
; AVX-NEXT: vmovd %xmm2, (%rsi)
; AVX-NEXT: vmovd %xmm1, (%rdx)
; AVX-NEXT: vmovd %xmm3, (%rcx)
; AVX-NEXT: vmovd %xmm0, (%r8)
; AVX-NEXT: retq
;
; AVX2-LABEL: load_i16_stride4_vf2:
; AVX2: # %bb.0:
; AVX2-NEXT: vmovdqa (%rdi), %xmm0
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[0,2,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm2 = xmm1[0,2,2,3,4,5,6,7]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[1,3,2,3,4,5,6,7]
; AVX2-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[3,1,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm3 = xmm0[2,0,2,3,4,5,6,7]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[3,1,2,3,4,5,6,7]
; AVX2-NEXT: vmovd %xmm2, (%rsi)
; AVX2-NEXT: vmovd %xmm1, (%rdx)
; AVX2-NEXT: vmovd %xmm3, (%rcx)
; AVX2-NEXT: vmovd %xmm0, (%r8)
; AVX2-NEXT: retq
;
; AVX2-FP-LABEL: load_i16_stride4_vf2:
; AVX2-FP: # %bb.0:
; AVX2-FP-NEXT: vmovdqa (%rdi), %xmm0
; AVX2-FP-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[0,2,2,3]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm2 = xmm1[0,2,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[1,3,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[3,1,2,3]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm3 = xmm0[2,0,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[3,1,2,3,4,5,6,7]
; AVX2-FP-NEXT: vmovd %xmm2, (%rsi)
; AVX2-FP-NEXT: vmovd %xmm1, (%rdx)
; AVX2-FP-NEXT: vmovd %xmm3, (%rcx)
; AVX2-FP-NEXT: vmovd %xmm0, (%r8)
; AVX2-FP-NEXT: retq
;
; AVX2-FCP-LABEL: load_i16_stride4_vf2:
; AVX2-FCP: # %bb.0:
; AVX2-FCP-NEXT: vmovdqa (%rdi), %xmm0
; AVX2-FCP-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[0,2,2,3]
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm2 = xmm1[0,2,2,3,4,5,6,7]
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[1,3,2,3,4,5,6,7]
; AVX2-FCP-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[3,1,2,3]
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm3 = xmm0[2,0,2,3,4,5,6,7]
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[3,1,2,3,4,5,6,7]
; AVX2-FCP-NEXT: vmovd %xmm2, (%rsi)
; AVX2-FCP-NEXT: vmovd %xmm1, (%rdx)
; AVX2-FCP-NEXT: vmovd %xmm3, (%rcx)
; AVX2-FCP-NEXT: vmovd %xmm0, (%r8)
; AVX2-FCP-NEXT: retq
;
; AVX512-LABEL: load_i16_stride4_vf2:
; AVX512: # %bb.0:
; AVX512-NEXT: vmovdqa (%rdi), %xmm0
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[0,2,2,3]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[1,3,2,3,4,5,6,7]
; AVX512-NEXT: vpshufd {{.*#+}} xmm2 = xmm0[3,1,2,3]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm3 = xmm2[2,0,2,3,4,5,6,7]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[3,1,2,3,4,5,6,7]
; AVX512-NEXT: vpmovqw %xmm0, (%rsi)
; AVX512-NEXT: vmovd %xmm1, (%rdx)
; AVX512-NEXT: vmovd %xmm3, (%rcx)
; AVX512-NEXT: vmovd %xmm2, (%r8)
; AVX512-NEXT: retq
;
; AVX512-FCP-LABEL: load_i16_stride4_vf2:
; AVX512-FCP: # %bb.0:
; AVX512-FCP-NEXT: vmovdqa (%rdi), %xmm0
; AVX512-FCP-NEXT: vpshufb {{.*#+}} xmm1 = xmm0[2,3,10,11,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512-FCP-NEXT: vpshufd {{.*#+}} xmm2 = xmm0[3,1,2,3]
; AVX512-FCP-NEXT: vpshuflw {{.*#+}} xmm3 = xmm2[2,0,2,3,4,5,6,7]
; AVX512-FCP-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[3,1,2,3,4,5,6,7]
; AVX512-FCP-NEXT: vpmovqw %xmm0, (%rsi)
; AVX512-FCP-NEXT: vmovd %xmm1, (%rdx)
; AVX512-FCP-NEXT: vmovd %xmm3, (%rcx)
; AVX512-FCP-NEXT: vmovd %xmm2, (%r8)
; AVX512-FCP-NEXT: retq
;
; AVX512DQ-LABEL: load_i16_stride4_vf2:
; AVX512DQ: # %bb.0:
; AVX512DQ-NEXT: vmovdqa (%rdi), %xmm0
; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[0,2,2,3]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[1,3,2,3,4,5,6,7]
; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm2 = xmm0[3,1,2,3]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm3 = xmm2[2,0,2,3,4,5,6,7]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[3,1,2,3,4,5,6,7]
; AVX512DQ-NEXT: vpmovqw %xmm0, (%rsi)
; AVX512DQ-NEXT: vmovd %xmm1, (%rdx)
; AVX512DQ-NEXT: vmovd %xmm3, (%rcx)
; AVX512DQ-NEXT: vmovd %xmm2, (%r8)
; AVX512DQ-NEXT: retq
;
; AVX512DQ-FCP-LABEL: load_i16_stride4_vf2:
; AVX512DQ-FCP: # %bb.0:
; AVX512DQ-FCP-NEXT: vmovdqa (%rdi), %xmm0
; AVX512DQ-FCP-NEXT: vpshufb {{.*#+}} xmm1 = xmm0[2,3,10,11,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} xmm2 = xmm0[3,1,2,3]
; AVX512DQ-FCP-NEXT: vpshuflw {{.*#+}} xmm3 = xmm2[2,0,2,3,4,5,6,7]
; AVX512DQ-FCP-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[3,1,2,3,4,5,6,7]
; AVX512DQ-FCP-NEXT: vpmovqw %xmm0, (%rsi)
; AVX512DQ-FCP-NEXT: vmovd %xmm1, (%rdx)
; AVX512DQ-FCP-NEXT: vmovd %xmm3, (%rcx)
; AVX512DQ-FCP-NEXT: vmovd %xmm2, (%r8)
; AVX512DQ-FCP-NEXT: retq
;
; AVX512BW-LABEL: load_i16_stride4_vf2:
; AVX512BW: # %bb.0:
; AVX512BW-NEXT: vmovdqa (%rdi), %xmm0
; AVX512BW-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[0,2,2,3]
; AVX512BW-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[1,3,2,3,4,5,6,7]
; AVX512BW-NEXT: vpshufd {{.*#+}} xmm2 = xmm0[3,1,2,3]
; AVX512BW-NEXT: vpshuflw {{.*#+}} xmm3 = xmm2[2,0,2,3,4,5,6,7]
; AVX512BW-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[3,1,2,3,4,5,6,7]
; AVX512BW-NEXT: vpmovqw %xmm0, (%rsi)
; AVX512BW-NEXT: vmovd %xmm1, (%rdx)
; AVX512BW-NEXT: vmovd %xmm3, (%rcx)
; AVX512BW-NEXT: vmovd %xmm2, (%r8)
; AVX512BW-NEXT: retq
;
; AVX512BW-FCP-LABEL: load_i16_stride4_vf2:
; AVX512BW-FCP: # %bb.0:
; AVX512BW-FCP-NEXT: vmovdqa (%rdi), %xmm0
; AVX512BW-FCP-NEXT: vpshufb {{.*#+}} xmm1 = xmm0[2,3,10,11,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512BW-FCP-NEXT: vpshufd {{.*#+}} xmm2 = xmm0[3,1,2,3]
; AVX512BW-FCP-NEXT: vpshuflw {{.*#+}} xmm3 = xmm2[2,0,2,3,4,5,6,7]
; AVX512BW-FCP-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[3,1,2,3,4,5,6,7]
; AVX512BW-FCP-NEXT: vpmovqw %xmm0, (%rsi)
; AVX512BW-FCP-NEXT: vmovd %xmm1, (%rdx)
; AVX512BW-FCP-NEXT: vmovd %xmm3, (%rcx)
; AVX512BW-FCP-NEXT: vmovd %xmm2, (%r8)
; AVX512BW-FCP-NEXT: retq
;
; AVX512DQ-BW-LABEL: load_i16_stride4_vf2:
; AVX512DQ-BW: # %bb.0:
; AVX512DQ-BW-NEXT: vmovdqa (%rdi), %xmm0
; AVX512DQ-BW-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[0,2,2,3]
; AVX512DQ-BW-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[1,3,2,3,4,5,6,7]
; AVX512DQ-BW-NEXT: vpshufd {{.*#+}} xmm2 = xmm0[3,1,2,3]
; AVX512DQ-BW-NEXT: vpshuflw {{.*#+}} xmm3 = xmm2[2,0,2,3,4,5,6,7]
; AVX512DQ-BW-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[3,1,2,3,4,5,6,7]
; AVX512DQ-BW-NEXT: vpmovqw %xmm0, (%rsi)
; AVX512DQ-BW-NEXT: vmovd %xmm1, (%rdx)
; AVX512DQ-BW-NEXT: vmovd %xmm3, (%rcx)
; AVX512DQ-BW-NEXT: vmovd %xmm2, (%r8)
; AVX512DQ-BW-NEXT: retq
;
; AVX512DQ-BW-FCP-LABEL: load_i16_stride4_vf2:
; AVX512DQ-BW-FCP: # %bb.0:
; AVX512DQ-BW-FCP-NEXT: vmovdqa (%rdi), %xmm0
; AVX512DQ-BW-FCP-NEXT: vpshufb {{.*#+}} xmm1 = xmm0[2,3,10,11,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512DQ-BW-FCP-NEXT: vpshufd {{.*#+}} xmm2 = xmm0[3,1,2,3]
; AVX512DQ-BW-FCP-NEXT: vpshuflw {{.*#+}} xmm3 = xmm2[2,0,2,3,4,5,6,7]
; AVX512DQ-BW-FCP-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[3,1,2,3,4,5,6,7]
; AVX512DQ-BW-FCP-NEXT: vpmovqw %xmm0, (%rsi)
; AVX512DQ-BW-FCP-NEXT: vmovd %xmm1, (%rdx)
; AVX512DQ-BW-FCP-NEXT: vmovd %xmm3, (%rcx)
; AVX512DQ-BW-FCP-NEXT: vmovd %xmm2, (%r8)
; AVX512DQ-BW-FCP-NEXT: retq
%wide.vec = load <8 x i16>, ptr %in.vec, align 64
%strided.vec0 = shufflevector <8 x i16> %wide.vec, <8 x i16> poison, <2 x i32> <i32 0, i32 4>
%strided.vec1 = shufflevector <8 x i16> %wide.vec, <8 x i16> poison, <2 x i32> <i32 1, i32 5>
%strided.vec2 = shufflevector <8 x i16> %wide.vec, <8 x i16> poison, <2 x i32> <i32 2, i32 6>
%strided.vec3 = shufflevector <8 x i16> %wide.vec, <8 x i16> poison, <2 x i32> <i32 3, i32 7>
store <2 x i16> %strided.vec0, ptr %out.vec0, align 64
store <2 x i16> %strided.vec1, ptr %out.vec1, align 64
store <2 x i16> %strided.vec2, ptr %out.vec2, align 64
store <2 x i16> %strided.vec3, ptr %out.vec3, align 64
ret void
}
define void @load_i16_stride4_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3) nounwind {
; SSE-LABEL: load_i16_stride4_vf4:
; SSE: # %bb.0:
; SSE-NEXT: movdqa (%rdi), %xmm0
; SSE-NEXT: movdqa 16(%rdi), %xmm1
; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm1[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm2[0,2,2,3,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm0[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm5 = xmm4[0,2,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm3[0],xmm5[1],xmm3[1]
; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[1,3,2,3,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm4[1,3,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[3,1,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm1[2,0,2,3,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[3,1,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm4 = xmm0[2,0,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm2[0],xmm4[1],xmm2[1]
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[3,1,2,3,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[3,1,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT: movq %xmm5, (%rsi)
; SSE-NEXT: movq %xmm3, (%rdx)
; SSE-NEXT: movq %xmm4, (%rcx)
; SSE-NEXT: movq %xmm0, (%r8)
; SSE-NEXT: retq
;
; AVX-LABEL: load_i16_stride4_vf4:
; AVX: # %bb.0:
; AVX-NEXT: vpxor %xmm0, %xmm0, %xmm0
; AVX-NEXT: vmovdqa (%rdi), %xmm1
; AVX-NEXT: vmovdqa 16(%rdi), %xmm2
; AVX-NEXT: vpblendw {{.*#+}} xmm3 = xmm2[0],xmm0[1,2,3],xmm2[4],xmm0[5,6,7]
; AVX-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0],xmm0[1,2,3],xmm1[4],xmm0[5,6,7]
; AVX-NEXT: vpackusdw %xmm3, %xmm0, %xmm0
; AVX-NEXT: vpackusdw %xmm0, %xmm0, %xmm0
; AVX-NEXT: vpshufd {{.*#+}} xmm3 = xmm2[0,2,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[1,3,2,3,4,5,6,7]
; AVX-NEXT: vpshufd {{.*#+}} xmm4 = xmm1[0,2,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[1,3,2,3,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
; AVX-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[3,1,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm4 = xmm2[2,0,2,3,4,5,6,7]
; AVX-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[3,1,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm5 = xmm1[2,0,2,3,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm4 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
; AVX-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[3,1,2,3,4,5,6,7]
; AVX-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[3,1,2,3,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
; AVX-NEXT: vmovq %xmm0, (%rsi)
; AVX-NEXT: vmovq %xmm3, (%rdx)
; AVX-NEXT: vmovq %xmm4, (%rcx)
; AVX-NEXT: vmovq %xmm1, (%r8)
; AVX-NEXT: retq
;
; AVX2-LABEL: load_i16_stride4_vf4:
; AVX2: # %bb.0:
; AVX2-NEXT: vpxor %xmm0, %xmm0, %xmm0
; AVX2-NEXT: vmovdqa (%rdi), %xmm1
; AVX2-NEXT: vmovdqa 16(%rdi), %xmm2
; AVX2-NEXT: vpblendw {{.*#+}} xmm3 = xmm2[0],xmm0[1,2,3],xmm2[4],xmm0[5,6,7]
; AVX2-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0],xmm0[1,2,3],xmm1[4],xmm0[5,6,7]
; AVX2-NEXT: vpackusdw %xmm3, %xmm0, %xmm0
; AVX2-NEXT: vpackusdw %xmm0, %xmm0, %xmm0
; AVX2-NEXT: vpshufd {{.*#+}} xmm3 = xmm2[0,2,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[1,3,2,3,4,5,6,7]
; AVX2-NEXT: vpshufd {{.*#+}} xmm4 = xmm1[0,2,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[1,3,2,3,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
; AVX2-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[3,1,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm4 = xmm2[2,0,2,3,4,5,6,7]
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[3,1,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm5 = xmm1[2,0,2,3,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm4 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[3,1,2,3,4,5,6,7]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[3,1,2,3,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
; AVX2-NEXT: vmovq %xmm0, (%rsi)
; AVX2-NEXT: vmovq %xmm3, (%rdx)
; AVX2-NEXT: vmovq %xmm4, (%rcx)
; AVX2-NEXT: vmovq %xmm1, (%r8)
; AVX2-NEXT: retq
;
; AVX2-FP-LABEL: load_i16_stride4_vf4:
; AVX2-FP: # %bb.0:
; AVX2-FP-NEXT: vpxor %xmm0, %xmm0, %xmm0
; AVX2-FP-NEXT: vmovdqa (%rdi), %xmm1
; AVX2-FP-NEXT: vmovdqa 16(%rdi), %xmm2
; AVX2-FP-NEXT: vpblendw {{.*#+}} xmm3 = xmm2[0],xmm0[1,2,3],xmm2[4],xmm0[5,6,7]
; AVX2-FP-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0],xmm0[1,2,3],xmm1[4],xmm0[5,6,7]
; AVX2-FP-NEXT: vpackusdw %xmm3, %xmm0, %xmm0
; AVX2-FP-NEXT: vpackusdw %xmm0, %xmm0, %xmm0
; AVX2-FP-NEXT: vmovdqa {{.*#+}} xmm3 = [2,3,10,11,8,9,10,11,8,9,10,11,12,13,14,15]
; AVX2-FP-NEXT: vpshufb %xmm3, %xmm2, %xmm4
; AVX2-FP-NEXT: vpshufb %xmm3, %xmm1, %xmm3
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
; AVX2-FP-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[3,1,2,3]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm4 = xmm2[2,0,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[3,1,2,3]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm5 = xmm1[2,0,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm4 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[3,1,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[3,1,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
; AVX2-FP-NEXT: vmovq %xmm0, (%rsi)
; AVX2-FP-NEXT: vmovq %xmm3, (%rdx)
; AVX2-FP-NEXT: vmovq %xmm4, (%rcx)
; AVX2-FP-NEXT: vmovq %xmm1, (%r8)
; AVX2-FP-NEXT: retq
;
; AVX2-FCP-LABEL: load_i16_stride4_vf4:
; AVX2-FCP: # %bb.0:
; AVX2-FCP-NEXT: vpxor %xmm0, %xmm0, %xmm0
; AVX2-FCP-NEXT: vmovdqa (%rdi), %xmm1
; AVX2-FCP-NEXT: vmovdqa 16(%rdi), %xmm2
; AVX2-FCP-NEXT: vpblendw {{.*#+}} xmm3 = xmm2[0],xmm0[1,2,3],xmm2[4],xmm0[5,6,7]
; AVX2-FCP-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0],xmm0[1,2,3],xmm1[4],xmm0[5,6,7]
; AVX2-FCP-NEXT: vpackusdw %xmm3, %xmm0, %xmm0
; AVX2-FCP-NEXT: vpackusdw %xmm0, %xmm0, %xmm0
; AVX2-FCP-NEXT: vmovdqa {{.*#+}} xmm3 = [2,3,10,11,8,9,10,11,8,9,10,11,12,13,14,15]
; AVX2-FCP-NEXT: vpshufb %xmm3, %xmm2, %xmm4
; AVX2-FCP-NEXT: vpshufb %xmm3, %xmm1, %xmm3
; AVX2-FCP-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
; AVX2-FCP-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[3,1,2,3]
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm4 = xmm2[2,0,2,3,4,5,6,7]
; AVX2-FCP-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[3,1,2,3]
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm5 = xmm1[2,0,2,3,4,5,6,7]
; AVX2-FCP-NEXT: vpunpckldq {{.*#+}} xmm4 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[3,1,2,3,4,5,6,7]
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[3,1,2,3,4,5,6,7]
; AVX2-FCP-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
; AVX2-FCP-NEXT: vmovq %xmm0, (%rsi)
; AVX2-FCP-NEXT: vmovq %xmm3, (%rdx)
; AVX2-FCP-NEXT: vmovq %xmm4, (%rcx)
; AVX2-FCP-NEXT: vmovq %xmm1, (%r8)
; AVX2-FCP-NEXT: retq
;
; AVX512-LABEL: load_i16_stride4_vf4:
; AVX512: # %bb.0:
; AVX512-NEXT: vmovdqa (%rdi), %ymm0
; AVX512-NEXT: vpsrlq $16, %ymm0, %ymm1
; AVX512-NEXT: vpsrlq $32, %ymm0, %ymm2
; AVX512-NEXT: vpsrlq $48, %ymm0, %ymm3
; AVX512-NEXT: vpmovqw %ymm0, (%rsi)
; AVX512-NEXT: vpmovqw %ymm1, (%rdx)
; AVX512-NEXT: vpmovqw %ymm2, (%rcx)
; AVX512-NEXT: vpmovqw %ymm3, (%r8)
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
;
; AVX512-FCP-LABEL: load_i16_stride4_vf4:
; AVX512-FCP: # %bb.0:
; AVX512-FCP-NEXT: vmovdqa (%rdi), %ymm0
; AVX512-FCP-NEXT: vpsrlq $16, %ymm0, %ymm1
; AVX512-FCP-NEXT: vpsrlq $32, %ymm0, %ymm2
; AVX512-FCP-NEXT: vpsrlq $48, %ymm0, %ymm3
; AVX512-FCP-NEXT: vpmovqw %ymm0, (%rsi)
; AVX512-FCP-NEXT: vpmovqw %ymm1, (%rdx)
; AVX512-FCP-NEXT: vpmovqw %ymm2, (%rcx)
; AVX512-FCP-NEXT: vpmovqw %ymm3, (%r8)
; AVX512-FCP-NEXT: vzeroupper
; AVX512-FCP-NEXT: retq
;
; AVX512DQ-LABEL: load_i16_stride4_vf4:
; AVX512DQ: # %bb.0:
; AVX512DQ-NEXT: vmovdqa (%rdi), %ymm0
; AVX512DQ-NEXT: vpsrlq $16, %ymm0, %ymm1
; AVX512DQ-NEXT: vpsrlq $32, %ymm0, %ymm2
; AVX512DQ-NEXT: vpsrlq $48, %ymm0, %ymm3
; AVX512DQ-NEXT: vpmovqw %ymm0, (%rsi)
; AVX512DQ-NEXT: vpmovqw %ymm1, (%rdx)
; AVX512DQ-NEXT: vpmovqw %ymm2, (%rcx)
; AVX512DQ-NEXT: vpmovqw %ymm3, (%r8)
; AVX512DQ-NEXT: vzeroupper
; AVX512DQ-NEXT: retq
;
; AVX512DQ-FCP-LABEL: load_i16_stride4_vf4:
; AVX512DQ-FCP: # %bb.0:
; AVX512DQ-FCP-NEXT: vmovdqa (%rdi), %ymm0
; AVX512DQ-FCP-NEXT: vpsrlq $16, %ymm0, %ymm1
; AVX512DQ-FCP-NEXT: vpsrlq $32, %ymm0, %ymm2
; AVX512DQ-FCP-NEXT: vpsrlq $48, %ymm0, %ymm3
; AVX512DQ-FCP-NEXT: vpmovqw %ymm0, (%rsi)
; AVX512DQ-FCP-NEXT: vpmovqw %ymm1, (%rdx)
; AVX512DQ-FCP-NEXT: vpmovqw %ymm2, (%rcx)
; AVX512DQ-FCP-NEXT: vpmovqw %ymm3, (%r8)
; AVX512DQ-FCP-NEXT: vzeroupper
; AVX512DQ-FCP-NEXT: retq
;
; AVX512BW-LABEL: load_i16_stride4_vf4:
; AVX512BW: # %bb.0:
; AVX512BW-NEXT: vmovdqa (%rdi), %ymm0
; AVX512BW-NEXT: vpsrlq $16, %ymm0, %ymm1
; AVX512BW-NEXT: vpsrlq $32, %ymm0, %ymm2
; AVX512BW-NEXT: vpsrlq $48, %ymm0, %ymm3
; AVX512BW-NEXT: vpmovqw %ymm0, (%rsi)
; AVX512BW-NEXT: vpmovqw %ymm1, (%rdx)
; AVX512BW-NEXT: vpmovqw %ymm2, (%rcx)
; AVX512BW-NEXT: vpmovqw %ymm3, (%r8)
; AVX512BW-NEXT: vzeroupper
; AVX512BW-NEXT: retq
;
; AVX512BW-FCP-LABEL: load_i16_stride4_vf4:
; AVX512BW-FCP: # %bb.0:
; AVX512BW-FCP-NEXT: vmovdqa (%rdi), %ymm0
; AVX512BW-FCP-NEXT: vpsrlq $16, %ymm0, %ymm1
; AVX512BW-FCP-NEXT: vpsrlq $32, %ymm0, %ymm2
; AVX512BW-FCP-NEXT: vpsrlq $48, %ymm0, %ymm3
; AVX512BW-FCP-NEXT: vpmovqw %ymm0, (%rsi)
; AVX512BW-FCP-NEXT: vpmovqw %ymm1, (%rdx)
; AVX512BW-FCP-NEXT: vpmovqw %ymm2, (%rcx)
; AVX512BW-FCP-NEXT: vpmovqw %ymm3, (%r8)
; AVX512BW-FCP-NEXT: vzeroupper
; AVX512BW-FCP-NEXT: retq
;
; AVX512DQ-BW-LABEL: load_i16_stride4_vf4:
; AVX512DQ-BW: # %bb.0:
; AVX512DQ-BW-NEXT: vmovdqa (%rdi), %ymm0
; AVX512DQ-BW-NEXT: vpsrlq $16, %ymm0, %ymm1
; AVX512DQ-BW-NEXT: vpsrlq $32, %ymm0, %ymm2
; AVX512DQ-BW-NEXT: vpsrlq $48, %ymm0, %ymm3
; AVX512DQ-BW-NEXT: vpmovqw %ymm0, (%rsi)
; AVX512DQ-BW-NEXT: vpmovqw %ymm1, (%rdx)
; AVX512DQ-BW-NEXT: vpmovqw %ymm2, (%rcx)
; AVX512DQ-BW-NEXT: vpmovqw %ymm3, (%r8)
; AVX512DQ-BW-NEXT: vzeroupper
; AVX512DQ-BW-NEXT: retq
;
; AVX512DQ-BW-FCP-LABEL: load_i16_stride4_vf4:
; AVX512DQ-BW-FCP: # %bb.0:
; AVX512DQ-BW-FCP-NEXT: vmovdqa (%rdi), %ymm0
; AVX512DQ-BW-FCP-NEXT: vpsrlq $16, %ymm0, %ymm1
; AVX512DQ-BW-FCP-NEXT: vpsrlq $32, %ymm0, %ymm2
; AVX512DQ-BW-FCP-NEXT: vpsrlq $48, %ymm0, %ymm3
; AVX512DQ-BW-FCP-NEXT: vpmovqw %ymm0, (%rsi)
; AVX512DQ-BW-FCP-NEXT: vpmovqw %ymm1, (%rdx)
; AVX512DQ-BW-FCP-NEXT: vpmovqw %ymm2, (%rcx)
; AVX512DQ-BW-FCP-NEXT: vpmovqw %ymm3, (%r8)
; AVX512DQ-BW-FCP-NEXT: vzeroupper
; AVX512DQ-BW-FCP-NEXT: retq
%wide.vec = load <16 x i16>, ptr %in.vec, align 64
%strided.vec0 = shufflevector <16 x i16> %wide.vec, <16 x i16> poison, <4 x i32> <i32 0, i32 4, i32 8, i32 12>
%strided.vec1 = shufflevector <16 x i16> %wide.vec, <16 x i16> poison, <4 x i32> <i32 1, i32 5, i32 9, i32 13>
%strided.vec2 = shufflevector <16 x i16> %wide.vec, <16 x i16> poison, <4 x i32> <i32 2, i32 6, i32 10, i32 14>
%strided.vec3 = shufflevector <16 x i16> %wide.vec, <16 x i16> poison, <4 x i32> <i32 3, i32 7, i32 11, i32 15>
store <4 x i16> %strided.vec0, ptr %out.vec0, align 64
store <4 x i16> %strided.vec1, ptr %out.vec1, align 64
store <4 x i16> %strided.vec2, ptr %out.vec2, align 64
store <4 x i16> %strided.vec3, ptr %out.vec3, align 64
ret void
}
define void @load_i16_stride4_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3) nounwind {
; SSE-LABEL: load_i16_stride4_vf8:
; SSE: # %bb.0:
; SSE-NEXT: movdqa (%rdi), %xmm2
; SSE-NEXT: movdqa 16(%rdi), %xmm3
; SSE-NEXT: movdqa 32(%rdi), %xmm1
; SSE-NEXT: movdqa 48(%rdi), %xmm4
; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm3[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm5[0,2,2,3,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm2[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm7 = xmm6[0,2,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm0[0],xmm7[1],xmm0[1]
; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm4[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm9 = xmm8[0,1,0,2,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm10 = xmm1[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm10[0,1,0,2,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1]
; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm7[0],xmm0[1]
; SSE-NEXT: pshuflw {{.*#+}} xmm5 = xmm5[1,3,2,3,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm6 = xmm6[1,3,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm6 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
; SSE-NEXT: pshuflw {{.*#+}} xmm5 = xmm8[0,1,1,3,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm7 = xmm10[0,1,1,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm5[0],xmm7[1],xmm5[1]
; SSE-NEXT: movsd {{.*#+}} xmm7 = xmm6[0],xmm7[1]
; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm3[3,1,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm5 = xmm3[2,0,2,3,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[3,1,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm6 = xmm2[2,0,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm6 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm4[3,1,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm5 = xmm4[0,1,2,0,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[3,1,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm8 = xmm1[0,1,2,0,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm8 = xmm8[0],xmm5[0],xmm8[1],xmm5[1]
; SSE-NEXT: movsd {{.*#+}} xmm8 = xmm6[0],xmm8[1]
; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[3,1,2,3,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[3,1,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm4[0,1,3,1,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[0,1,3,1,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
; SSE-NEXT: movapd %xmm0, (%rsi)
; SSE-NEXT: movapd %xmm7, (%rdx)
; SSE-NEXT: movapd %xmm8, (%rcx)
; SSE-NEXT: movapd %xmm1, (%r8)
; SSE-NEXT: retq
;
; AVX-LABEL: load_i16_stride4_vf8:
; AVX: # %bb.0:
; AVX-NEXT: vpxor %xmm0, %xmm0, %xmm0
; AVX-NEXT: vmovdqa (%rdi), %xmm1
; AVX-NEXT: vmovdqa 16(%rdi), %xmm2
; AVX-NEXT: vmovdqa 32(%rdi), %xmm3
; AVX-NEXT: vmovdqa 48(%rdi), %xmm4
; AVX-NEXT: vpblendw {{.*#+}} xmm5 = xmm4[0],xmm0[1,2,3],xmm4[4],xmm0[5,6,7]
; AVX-NEXT: vpblendw {{.*#+}} xmm6 = xmm3[0],xmm0[1,2,3],xmm3[4],xmm0[5,6,7]
; AVX-NEXT: vpackusdw %xmm5, %xmm6, %xmm5
; AVX-NEXT: vpblendw {{.*#+}} xmm6 = xmm2[0],xmm0[1,2,3],xmm2[4],xmm0[5,6,7]
; AVX-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0],xmm0[1,2,3],xmm1[4],xmm0[5,6,7]
; AVX-NEXT: vpackusdw %xmm6, %xmm0, %xmm0
; AVX-NEXT: vpackusdw %xmm5, %xmm0, %xmm0
; AVX-NEXT: vpshufd {{.*#+}} xmm5 = xmm4[0,2,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm5 = xmm5[0,1,1,3,4,5,6,7]
; AVX-NEXT: vpshufd {{.*#+}} xmm6 = xmm3[0,2,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm6 = xmm6[0,1,1,3,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm5 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
; AVX-NEXT: vpshufd {{.*#+}} xmm6 = xmm2[0,2,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm6 = xmm6[1,3,2,3,4,5,6,7]
; AVX-NEXT: vpshufd {{.*#+}} xmm7 = xmm1[0,2,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm7 = xmm7[1,3,2,3,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm6 = xmm7[0],xmm6[0],xmm7[1],xmm6[1]
; AVX-NEXT: vpblendw {{.*#+}} xmm5 = xmm6[0,1,2,3],xmm5[4,5,6,7]
; AVX-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[3,1,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm6 = xmm4[0,1,2,0,4,5,6,7]
; AVX-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[3,1,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm7 = xmm3[0,1,2,0,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm6 = xmm7[0],xmm6[0],xmm7[1],xmm6[1]
; AVX-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[3,1,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm7 = xmm2[2,0,2,3,4,5,6,7]
; AVX-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[3,1,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm8 = xmm1[2,0,2,3,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm7 = xmm8[0],xmm7[0],xmm8[1],xmm7[1]
; AVX-NEXT: vpblendw {{.*#+}} xmm6 = xmm7[0,1,2,3],xmm6[4,5,6,7]
; AVX-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[0,1,3,1,4,5,6,7]
; AVX-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[0,1,3,1,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
; AVX-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[3,1,2,3,4,5,6,7]
; AVX-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[3,1,2,3,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
; AVX-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3],xmm3[4,5,6,7]
; AVX-NEXT: vmovdqa %xmm0, (%rsi)
; AVX-NEXT: vmovdqa %xmm5, (%rdx)
; AVX-NEXT: vmovdqa %xmm6, (%rcx)
; AVX-NEXT: vmovdqa %xmm1, (%r8)
; AVX-NEXT: retq
;
; AVX2-LABEL: load_i16_stride4_vf8:
; AVX2: # %bb.0:
; AVX2-NEXT: vpxor %xmm0, %xmm0, %xmm0
; AVX2-NEXT: vpblendw {{.*#+}} ymm1 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-NEXT: vextracti128 $1, %ymm1, %xmm2
; AVX2-NEXT: vpackusdw %xmm2, %xmm1, %xmm1
; AVX2-NEXT: vpblendw {{.*#+}} ymm0 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm2
; AVX2-NEXT: vpackusdw %xmm2, %xmm0, %xmm0
; AVX2-NEXT: vpackusdw %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vmovdqa (%rdi), %xmm1
; AVX2-NEXT: vmovdqa 16(%rdi), %xmm2
; AVX2-NEXT: vmovdqa 32(%rdi), %xmm3
; AVX2-NEXT: vmovdqa 48(%rdi), %xmm4
; AVX2-NEXT: vpshufd {{.*#+}} xmm5 = xmm4[0,2,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm5 = xmm5[0,1,1,3,4,5,6,7]
; AVX2-NEXT: vpshufd {{.*#+}} xmm6 = xmm3[0,2,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm6 = xmm6[0,1,1,3,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm5 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
; AVX2-NEXT: vpshufd {{.*#+}} xmm6 = xmm2[0,2,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm6 = xmm6[1,3,2,3,4,5,6,7]
; AVX2-NEXT: vpshufd {{.*#+}} xmm7 = xmm1[0,2,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm7 = xmm7[1,3,2,3,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm6 = xmm7[0],xmm6[0],xmm7[1],xmm6[1]
; AVX2-NEXT: vpblendd {{.*#+}} xmm5 = xmm6[0,1],xmm5[2,3]
; AVX2-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[3,1,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm6 = xmm4[0,1,2,0,4,5,6,7]
; AVX2-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[3,1,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm7 = xmm3[0,1,2,0,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm6 = xmm7[0],xmm6[0],xmm7[1],xmm6[1]
; AVX2-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[3,1,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm7 = xmm2[2,0,2,3,4,5,6,7]
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[3,1,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm8 = xmm1[2,0,2,3,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm7 = xmm8[0],xmm7[0],xmm8[1],xmm7[1]
; AVX2-NEXT: vpblendd {{.*#+}} xmm6 = xmm7[0,1],xmm6[2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[0,1,3,1,4,5,6,7]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[0,1,3,1,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[3,1,2,3,4,5,6,7]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[3,1,2,3,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
; AVX2-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0,1],xmm3[2,3]
; AVX2-NEXT: vmovdqa %xmm0, (%rsi)
; AVX2-NEXT: vmovdqa %xmm5, (%rdx)
; AVX2-NEXT: vmovdqa %xmm6, (%rcx)
; AVX2-NEXT: vmovdqa %xmm1, (%r8)
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
; AVX2-FP-LABEL: load_i16_stride4_vf8:
; AVX2-FP: # %bb.0:
; AVX2-FP-NEXT: vpxor %xmm0, %xmm0, %xmm0
; AVX2-FP-NEXT: vpblendw {{.*#+}} ymm1 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-FP-NEXT: vextracti128 $1, %ymm1, %xmm2
; AVX2-FP-NEXT: vpackusdw %xmm2, %xmm1, %xmm1
; AVX2-FP-NEXT: vpblendw {{.*#+}} ymm0 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-FP-NEXT: vextracti128 $1, %ymm0, %xmm2
; AVX2-FP-NEXT: vpackusdw %xmm2, %xmm0, %xmm0
; AVX2-FP-NEXT: vpackusdw %xmm1, %xmm0, %xmm0
; AVX2-FP-NEXT: vmovdqa (%rdi), %xmm1
; AVX2-FP-NEXT: vmovdqa 16(%rdi), %xmm2
; AVX2-FP-NEXT: vmovdqa 32(%rdi), %xmm3
; AVX2-FP-NEXT: vmovdqa 48(%rdi), %xmm4
; AVX2-FP-NEXT: vmovdqa {{.*#+}} xmm5 = [0,1,2,3,2,3,10,11,8,9,10,11,12,13,14,15]
; AVX2-FP-NEXT: vpshufb %xmm5, %xmm4, %xmm6
; AVX2-FP-NEXT: vpshufb %xmm5, %xmm3, %xmm5
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm5 = xmm5[0],xmm6[0],xmm5[1],xmm6[1]
; AVX2-FP-NEXT: vmovdqa {{.*#+}} xmm6 = [2,3,10,11,8,9,10,11,8,9,10,11,12,13,14,15]
; AVX2-FP-NEXT: vpshufb %xmm6, %xmm2, %xmm7
; AVX2-FP-NEXT: vpshufb %xmm6, %xmm1, %xmm6
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm6 = xmm6[0],xmm7[0],xmm6[1],xmm7[1]
; AVX2-FP-NEXT: vpblendd {{.*#+}} xmm5 = xmm6[0,1],xmm5[2,3]
; AVX2-FP-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[3,1,2,3]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm6 = xmm4[0,1,2,0,4,5,6,7]
; AVX2-FP-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[3,1,2,3]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm7 = xmm3[0,1,2,0,4,5,6,7]
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm6 = xmm7[0],xmm6[0],xmm7[1],xmm6[1]
; AVX2-FP-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[3,1,2,3]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm7 = xmm2[2,0,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[3,1,2,3]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm8 = xmm1[2,0,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm7 = xmm8[0],xmm7[0],xmm8[1],xmm7[1]
; AVX2-FP-NEXT: vpblendd {{.*#+}} xmm6 = xmm7[0,1],xmm6[2,3]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[0,1,3,1,4,5,6,7]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[0,1,3,1,4,5,6,7]
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[3,1,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[3,1,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
; AVX2-FP-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0,1],xmm3[2,3]
; AVX2-FP-NEXT: vmovdqa %xmm0, (%rsi)
; AVX2-FP-NEXT: vmovdqa %xmm5, (%rdx)
; AVX2-FP-NEXT: vmovdqa %xmm6, (%rcx)
; AVX2-FP-NEXT: vmovdqa %xmm1, (%r8)
; AVX2-FP-NEXT: vzeroupper
; AVX2-FP-NEXT: retq
;
; AVX2-FCP-LABEL: load_i16_stride4_vf8:
; AVX2-FCP: # %bb.0:
; AVX2-FCP-NEXT: vpxor %xmm0, %xmm0, %xmm0
; AVX2-FCP-NEXT: vpblendw {{.*#+}} ymm1 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-FCP-NEXT: vextracti128 $1, %ymm1, %xmm2
; AVX2-FCP-NEXT: vpackusdw %xmm2, %xmm1, %xmm1
; AVX2-FCP-NEXT: vpblendw {{.*#+}} ymm0 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-FCP-NEXT: vextracti128 $1, %ymm0, %xmm2
; AVX2-FCP-NEXT: vpackusdw %xmm2, %xmm0, %xmm0
; AVX2-FCP-NEXT: vpackusdw %xmm1, %xmm0, %xmm0
; AVX2-FCP-NEXT: vmovdqa (%rdi), %xmm1
; AVX2-FCP-NEXT: vmovdqa 16(%rdi), %xmm2
; AVX2-FCP-NEXT: vmovdqa 32(%rdi), %xmm3
; AVX2-FCP-NEXT: vmovdqa 48(%rdi), %xmm4
; AVX2-FCP-NEXT: vmovdqa {{.*#+}} xmm5 = [0,1,2,3,2,3,10,11,8,9,10,11,12,13,14,15]
; AVX2-FCP-NEXT: vpshufb %xmm5, %xmm4, %xmm6
; AVX2-FCP-NEXT: vpshufb %xmm5, %xmm3, %xmm5
; AVX2-FCP-NEXT: vpunpckldq {{.*#+}} xmm5 = xmm5[0],xmm6[0],xmm5[1],xmm6[1]
; AVX2-FCP-NEXT: vmovdqa {{.*#+}} xmm6 = [2,3,10,11,8,9,10,11,8,9,10,11,12,13,14,15]
; AVX2-FCP-NEXT: vpshufb %xmm6, %xmm2, %xmm7
; AVX2-FCP-NEXT: vpshufb %xmm6, %xmm1, %xmm6
; AVX2-FCP-NEXT: vpunpckldq {{.*#+}} xmm6 = xmm6[0],xmm7[0],xmm6[1],xmm7[1]
; AVX2-FCP-NEXT: vpblendd {{.*#+}} xmm5 = xmm6[0,1],xmm5[2,3]
; AVX2-FCP-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[3,1,2,3]
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm6 = xmm4[0,1,2,0,4,5,6,7]
; AVX2-FCP-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[3,1,2,3]
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm7 = xmm3[0,1,2,0,4,5,6,7]
; AVX2-FCP-NEXT: vpunpckldq {{.*#+}} xmm6 = xmm7[0],xmm6[0],xmm7[1],xmm6[1]
; AVX2-FCP-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[3,1,2,3]
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm7 = xmm2[2,0,2,3,4,5,6,7]
; AVX2-FCP-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[3,1,2,3]
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm8 = xmm1[2,0,2,3,4,5,6,7]
; AVX2-FCP-NEXT: vpunpckldq {{.*#+}} xmm7 = xmm8[0],xmm7[0],xmm8[1],xmm7[1]
; AVX2-FCP-NEXT: vpblendd {{.*#+}} xmm6 = xmm7[0,1],xmm6[2,3]
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[0,1,3,1,4,5,6,7]
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[0,1,3,1,4,5,6,7]
; AVX2-FCP-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[3,1,2,3,4,5,6,7]
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[3,1,2,3,4,5,6,7]
; AVX2-FCP-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
; AVX2-FCP-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0,1],xmm3[2,3]
; AVX2-FCP-NEXT: vmovdqa %xmm0, (%rsi)
; AVX2-FCP-NEXT: vmovdqa %xmm5, (%rdx)
; AVX2-FCP-NEXT: vmovdqa %xmm6, (%rcx)
; AVX2-FCP-NEXT: vmovdqa %xmm1, (%r8)
; AVX2-FCP-NEXT: vzeroupper
; AVX2-FCP-NEXT: retq
;
; AVX512-LABEL: load_i16_stride4_vf8:
; AVX512: # %bb.0:
; AVX512-NEXT: vmovdqa64 (%rdi), %zmm0
; AVX512-NEXT: vpsrlq $16, %zmm0, %zmm1
; AVX512-NEXT: vpsrlq $32, %zmm0, %zmm2
; AVX512-NEXT: vpsrlq $48, %zmm0, %zmm3
; AVX512-NEXT: vpmovqw %zmm0, (%rsi)
; AVX512-NEXT: vpmovqw %zmm1, (%rdx)
; AVX512-NEXT: vpmovqw %zmm2, (%rcx)
; AVX512-NEXT: vpmovqw %zmm3, (%r8)
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
;
; AVX512-FCP-LABEL: load_i16_stride4_vf8:
; AVX512-FCP: # %bb.0:
; AVX512-FCP-NEXT: vmovdqa64 (%rdi), %zmm0
; AVX512-FCP-NEXT: vpsrlq $16, %zmm0, %zmm1
; AVX512-FCP-NEXT: vpsrlq $32, %zmm0, %zmm2
; AVX512-FCP-NEXT: vpsrlq $48, %zmm0, %zmm3
; AVX512-FCP-NEXT: vpmovqw %zmm0, (%rsi)
; AVX512-FCP-NEXT: vpmovqw %zmm1, (%rdx)
; AVX512-FCP-NEXT: vpmovqw %zmm2, (%rcx)
; AVX512-FCP-NEXT: vpmovqw %zmm3, (%r8)
; AVX512-FCP-NEXT: vzeroupper
; AVX512-FCP-NEXT: retq
;
; AVX512DQ-LABEL: load_i16_stride4_vf8:
; AVX512DQ: # %bb.0:
; AVX512DQ-NEXT: vmovdqa64 (%rdi), %zmm0
; AVX512DQ-NEXT: vpsrlq $16, %zmm0, %zmm1
; AVX512DQ-NEXT: vpsrlq $32, %zmm0, %zmm2
; AVX512DQ-NEXT: vpsrlq $48, %zmm0, %zmm3
; AVX512DQ-NEXT: vpmovqw %zmm0, (%rsi)
; AVX512DQ-NEXT: vpmovqw %zmm1, (%rdx)
; AVX512DQ-NEXT: vpmovqw %zmm2, (%rcx)
; AVX512DQ-NEXT: vpmovqw %zmm3, (%r8)
; AVX512DQ-NEXT: vzeroupper
; AVX512DQ-NEXT: retq
;
; AVX512DQ-FCP-LABEL: load_i16_stride4_vf8:
; AVX512DQ-FCP: # %bb.0:
; AVX512DQ-FCP-NEXT: vmovdqa64 (%rdi), %zmm0
; AVX512DQ-FCP-NEXT: vpsrlq $16, %zmm0, %zmm1
; AVX512DQ-FCP-NEXT: vpsrlq $32, %zmm0, %zmm2
; AVX512DQ-FCP-NEXT: vpsrlq $48, %zmm0, %zmm3
; AVX512DQ-FCP-NEXT: vpmovqw %zmm0, (%rsi)
; AVX512DQ-FCP-NEXT: vpmovqw %zmm1, (%rdx)
; AVX512DQ-FCP-NEXT: vpmovqw %zmm2, (%rcx)
; AVX512DQ-FCP-NEXT: vpmovqw %zmm3, (%r8)
; AVX512DQ-FCP-NEXT: vzeroupper
; AVX512DQ-FCP-NEXT: retq
;
; AVX512BW-LABEL: load_i16_stride4_vf8:
; AVX512BW: # %bb.0:
; AVX512BW-NEXT: vmovdqa64 (%rdi), %zmm0
; AVX512BW-NEXT: vpsrlq $16, %zmm0, %zmm1
; AVX512BW-NEXT: vpsrlq $32, %zmm0, %zmm2
; AVX512BW-NEXT: vpsrlq $48, %zmm0, %zmm3
; AVX512BW-NEXT: vpmovqw %zmm0, (%rsi)
; AVX512BW-NEXT: vpmovqw %zmm1, (%rdx)
; AVX512BW-NEXT: vpmovqw %zmm2, (%rcx)
; AVX512BW-NEXT: vpmovqw %zmm3, (%r8)
; AVX512BW-NEXT: vzeroupper
; AVX512BW-NEXT: retq
;
; AVX512BW-FCP-LABEL: load_i16_stride4_vf8:
; AVX512BW-FCP: # %bb.0:
; AVX512BW-FCP-NEXT: vmovdqa64 (%rdi), %zmm0
; AVX512BW-FCP-NEXT: vpsrlq $16, %zmm0, %zmm1
; AVX512BW-FCP-NEXT: vpsrlq $32, %zmm0, %zmm2
; AVX512BW-FCP-NEXT: vpsrlq $48, %zmm0, %zmm3
; AVX512BW-FCP-NEXT: vpmovqw %zmm0, (%rsi)
; AVX512BW-FCP-NEXT: vpmovqw %zmm1, (%rdx)
; AVX512BW-FCP-NEXT: vpmovqw %zmm2, (%rcx)
; AVX512BW-FCP-NEXT: vpmovqw %zmm3, (%r8)
; AVX512BW-FCP-NEXT: vzeroupper
; AVX512BW-FCP-NEXT: retq
;
; AVX512DQ-BW-LABEL: load_i16_stride4_vf8:
; AVX512DQ-BW: # %bb.0:
; AVX512DQ-BW-NEXT: vmovdqa64 (%rdi), %zmm0
; AVX512DQ-BW-NEXT: vpsrlq $16, %zmm0, %zmm1
; AVX512DQ-BW-NEXT: vpsrlq $32, %zmm0, %zmm2
; AVX512DQ-BW-NEXT: vpsrlq $48, %zmm0, %zmm3
; AVX512DQ-BW-NEXT: vpmovqw %zmm0, (%rsi)
; AVX512DQ-BW-NEXT: vpmovqw %zmm1, (%rdx)
; AVX512DQ-BW-NEXT: vpmovqw %zmm2, (%rcx)
; AVX512DQ-BW-NEXT: vpmovqw %zmm3, (%r8)
; AVX512DQ-BW-NEXT: vzeroupper
; AVX512DQ-BW-NEXT: retq
;
; AVX512DQ-BW-FCP-LABEL: load_i16_stride4_vf8:
; AVX512DQ-BW-FCP: # %bb.0:
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 (%rdi), %zmm0
; AVX512DQ-BW-FCP-NEXT: vpsrlq $16, %zmm0, %zmm1
; AVX512DQ-BW-FCP-NEXT: vpsrlq $32, %zmm0, %zmm2
; AVX512DQ-BW-FCP-NEXT: vpsrlq $48, %zmm0, %zmm3
; AVX512DQ-BW-FCP-NEXT: vpmovqw %zmm0, (%rsi)
; AVX512DQ-BW-FCP-NEXT: vpmovqw %zmm1, (%rdx)
; AVX512DQ-BW-FCP-NEXT: vpmovqw %zmm2, (%rcx)
; AVX512DQ-BW-FCP-NEXT: vpmovqw %zmm3, (%r8)
; AVX512DQ-BW-FCP-NEXT: vzeroupper
; AVX512DQ-BW-FCP-NEXT: retq
%wide.vec = load <32 x i16>, ptr %in.vec, align 64
%strided.vec0 = shufflevector <32 x i16> %wide.vec, <32 x i16> poison, <8 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28>
%strided.vec1 = shufflevector <32 x i16> %wide.vec, <32 x i16> poison, <8 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29>
%strided.vec2 = shufflevector <32 x i16> %wide.vec, <32 x i16> poison, <8 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30>
%strided.vec3 = shufflevector <32 x i16> %wide.vec, <32 x i16> poison, <8 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31>
store <8 x i16> %strided.vec0, ptr %out.vec0, align 64
store <8 x i16> %strided.vec1, ptr %out.vec1, align 64
store <8 x i16> %strided.vec2, ptr %out.vec2, align 64
store <8 x i16> %strided.vec3, ptr %out.vec3, align 64
ret void
}
define void @load_i16_stride4_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3) nounwind {
; SSE-LABEL: load_i16_stride4_vf16:
; SSE: # %bb.0:
; SSE-NEXT: movdqa 96(%rdi), %xmm4
; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa 64(%rdi), %xmm1
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa 80(%rdi), %xmm3
; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa (%rdi), %xmm8
; SSE-NEXT: movdqa 16(%rdi), %xmm10
; SSE-NEXT: movdqa 32(%rdi), %xmm7
; SSE-NEXT: movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa 48(%rdi), %xmm6
; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm10[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm5[0,2,2,3,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm9 = xmm8[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm9[0,2,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT: pshufd {{.*#+}} xmm12 = xmm6[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm11 = xmm12[0,1,0,2,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm13 = xmm7[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm7 = xmm13[0,1,0,2,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm11[0],xmm7[1],xmm11[1]
; SSE-NEXT: movsd {{.*#+}} xmm7 = xmm2[0],xmm7[1]
; SSE-NEXT: pshufd {{.*#+}} xmm14 = xmm3[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm14[0,2,2,3,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm15 = xmm1[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm15[0,2,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
; SSE-NEXT: movdqa 112(%rdi), %xmm11
; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm11[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm3[0,1,0,2,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm4[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm4 = xmm2[0,1,0,2,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
; SSE-NEXT: movsd {{.*#+}} xmm4 = xmm0[0],xmm4[1]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm5[1,3,2,3,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm9[1,3,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm12[0,1,1,3,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm5 = xmm13[0,1,1,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1]
; SSE-NEXT: movsd {{.*#+}} xmm5 = xmm1[0],xmm5[1]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm14[1,3,2,3,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm15[1,3,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm3[0,1,1,3,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm9 = xmm2[0,1,1,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm9 = xmm9[0],xmm0[0],xmm9[1],xmm0[1]
; SSE-NEXT: movsd {{.*#+}} xmm9 = xmm1[0],xmm9[1]
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm10[3,1,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm0[2,0,2,3,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm8[3,1,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm2[2,0,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm6[3,1,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm8 = xmm1[0,1,2,0,4,5,6,7]
; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
; SSE-NEXT: # xmm10 = mem[3,1,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm6 = xmm10[0,1,2,0,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm6 = xmm6[0],xmm8[0],xmm6[1],xmm8[1]
; SSE-NEXT: movsd {{.*#+}} xmm6 = xmm3[0],xmm6[1]
; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; SSE-NEXT: # xmm3 = mem[3,1,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm8 = xmm3[2,0,2,3,4,5,6,7]
; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
; SSE-NEXT: # xmm12 = mem[3,1,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm13 = xmm12[2,0,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm13 = xmm13[0],xmm8[0],xmm13[1],xmm8[1]
; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm11[3,1,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm11 = xmm8[0,1,2,0,4,5,6,7]
; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
; SSE-NEXT: # xmm14 = mem[3,1,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm15 = xmm14[0,1,2,0,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm15 = xmm15[0],xmm11[0],xmm15[1],xmm11[1]
; SSE-NEXT: movsd {{.*#+}} xmm15 = xmm13[0],xmm15[1]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[3,1,2,3,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[3,1,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm1[0,1,3,1,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm10[0,1,3,1,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm3[3,1,2,3,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm12[3,1,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm8[0,1,3,1,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm14[0,1,3,1,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
; SSE-NEXT: movsd {{.*#+}} xmm3 = xmm2[0],xmm3[1]
; SSE-NEXT: movapd %xmm4, 16(%rsi)
; SSE-NEXT: movapd %xmm7, (%rsi)
; SSE-NEXT: movapd %xmm9, 16(%rdx)
; SSE-NEXT: movapd %xmm5, (%rdx)
; SSE-NEXT: movapd %xmm15, 16(%rcx)
; SSE-NEXT: movapd %xmm6, (%rcx)
; SSE-NEXT: movapd %xmm3, 16(%r8)
; SSE-NEXT: movapd %xmm1, (%r8)
; SSE-NEXT: retq
;
; AVX-LABEL: load_i16_stride4_vf16:
; AVX: # %bb.0:
; AVX-NEXT: vpxor %xmm1, %xmm1, %xmm1
; AVX-NEXT: vmovdqa 112(%rdi), %xmm5
; AVX-NEXT: vpblendw {{.*#+}} xmm0 = xmm5[0],xmm1[1,2,3],xmm5[4],xmm1[5,6,7]
; AVX-NEXT: vmovdqa 96(%rdi), %xmm6
; AVX-NEXT: vpblendw {{.*#+}} xmm2 = xmm6[0],xmm1[1,2,3],xmm6[4],xmm1[5,6,7]
; AVX-NEXT: vpackusdw %xmm0, %xmm2, %xmm0
; AVX-NEXT: vmovdqa 80(%rdi), %xmm7
; AVX-NEXT: vpblendw {{.*#+}} xmm2 = xmm7[0],xmm1[1,2,3],xmm7[4],xmm1[5,6,7]
; AVX-NEXT: vmovdqa 64(%rdi), %xmm8
; AVX-NEXT: vpblendw {{.*#+}} xmm3 = xmm8[0],xmm1[1,2,3],xmm8[4],xmm1[5,6,7]
; AVX-NEXT: vpackusdw %xmm2, %xmm3, %xmm2
; AVX-NEXT: vpackusdw %xmm0, %xmm2, %xmm0
; AVX-NEXT: vmovdqa (%rdi), %xmm2
; AVX-NEXT: vmovdqa 16(%rdi), %xmm3
; AVX-NEXT: vmovdqa 32(%rdi), %xmm4
; AVX-NEXT: vmovdqa 48(%rdi), %xmm9
; AVX-NEXT: vpblendw {{.*#+}} xmm10 = xmm9[0],xmm1[1,2,3],xmm9[4],xmm1[5,6,7]
; AVX-NEXT: vpblendw {{.*#+}} xmm11 = xmm4[0],xmm1[1,2,3],xmm4[4],xmm1[5,6,7]
; AVX-NEXT: vpackusdw %xmm10, %xmm11, %xmm10
; AVX-NEXT: vpblendw {{.*#+}} xmm11 = xmm3[0],xmm1[1,2,3],xmm3[4],xmm1[5,6,7]
; AVX-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0],xmm1[1,2,3],xmm2[4],xmm1[5,6,7]
; AVX-NEXT: vpackusdw %xmm11, %xmm1, %xmm1
; AVX-NEXT: vpackusdw %xmm10, %xmm1, %xmm1
; AVX-NEXT: vpshufd {{.*#+}} xmm10 = xmm5[0,2,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm10 = xmm10[0,1,1,3,4,5,6,7]
; AVX-NEXT: vpshufd {{.*#+}} xmm11 = xmm6[0,2,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm11 = xmm11[0,1,1,3,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm10 = xmm11[0],xmm10[0],xmm11[1],xmm10[1]
; AVX-NEXT: vpshufd {{.*#+}} xmm11 = xmm7[0,2,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm11 = xmm11[1,3,2,3,4,5,6,7]
; AVX-NEXT: vpshufd {{.*#+}} xmm12 = xmm8[0,2,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm12 = xmm12[1,3,2,3,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm11 = xmm12[0],xmm11[0],xmm12[1],xmm11[1]
; AVX-NEXT: vpblendw {{.*#+}} xmm10 = xmm11[0,1,2,3],xmm10[4,5,6,7]
; AVX-NEXT: vinsertf128 $1, %xmm10, %ymm0, %ymm10
; AVX-NEXT: vpshufd {{.*#+}} xmm11 = xmm9[0,2,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm11 = xmm11[0,1,1,3,4,5,6,7]
; AVX-NEXT: vpshufd {{.*#+}} xmm12 = xmm4[0,2,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm12 = xmm12[0,1,1,3,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm11 = xmm12[0],xmm11[0],xmm12[1],xmm11[1]
; AVX-NEXT: vpshufd {{.*#+}} xmm12 = xmm3[0,2,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm12 = xmm12[1,3,2,3,4,5,6,7]
; AVX-NEXT: vpshufd {{.*#+}} xmm13 = xmm2[0,2,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm13 = xmm13[1,3,2,3,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm12 = xmm13[0],xmm12[0],xmm13[1],xmm12[1]
; AVX-NEXT: vpblendw {{.*#+}} xmm11 = xmm12[0,1,2,3],xmm11[4,5,6,7]
; AVX-NEXT: vblendps {{.*#+}} ymm10 = ymm11[0,1,2,3],ymm10[4,5,6,7]
; AVX-NEXT: vpshufd {{.*#+}} xmm5 = xmm5[3,1,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm11 = xmm5[0,1,2,0,4,5,6,7]
; AVX-NEXT: vpshufd {{.*#+}} xmm6 = xmm6[3,1,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm12 = xmm6[0,1,2,0,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm11 = xmm12[0],xmm11[0],xmm12[1],xmm11[1]
; AVX-NEXT: vpshufd {{.*#+}} xmm7 = xmm7[3,1,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm12 = xmm7[2,0,2,3,4,5,6,7]
; AVX-NEXT: vpshufd {{.*#+}} xmm8 = xmm8[3,1,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm13 = xmm8[2,0,2,3,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm12 = xmm13[0],xmm12[0],xmm13[1],xmm12[1]
; AVX-NEXT: vpblendw {{.*#+}} xmm11 = xmm12[0,1,2,3],xmm11[4,5,6,7]
; AVX-NEXT: vinsertf128 $1, %xmm11, %ymm0, %ymm11
; AVX-NEXT: vpshufd {{.*#+}} xmm9 = xmm9[3,1,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm12 = xmm9[0,1,2,0,4,5,6,7]
; AVX-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[3,1,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm13 = xmm4[0,1,2,0,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm12 = xmm13[0],xmm12[0],xmm13[1],xmm12[1]
; AVX-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[3,1,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm13 = xmm3[2,0,2,3,4,5,6,7]
; AVX-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[3,1,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm14 = xmm2[2,0,2,3,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm13 = xmm14[0],xmm13[0],xmm14[1],xmm13[1]
; AVX-NEXT: vpblendw {{.*#+}} xmm12 = xmm13[0,1,2,3],xmm12[4,5,6,7]
; AVX-NEXT: vblendps {{.*#+}} ymm11 = ymm12[0,1,2,3],ymm11[4,5,6,7]
; AVX-NEXT: vpshuflw {{.*#+}} xmm5 = xmm5[0,1,3,1,4,5,6,7]
; AVX-NEXT: vpshuflw {{.*#+}} xmm6 = xmm6[0,1,3,1,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm5 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
; AVX-NEXT: vpshuflw {{.*#+}} xmm6 = xmm7[3,1,2,3,4,5,6,7]
; AVX-NEXT: vpshuflw {{.*#+}} xmm7 = xmm8[3,1,2,3,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm6 = xmm7[0],xmm6[0],xmm7[1],xmm6[1]
; AVX-NEXT: vpblendw {{.*#+}} xmm5 = xmm6[0,1,2,3],xmm5[4,5,6,7]
; AVX-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm5
; AVX-NEXT: vpshuflw {{.*#+}} xmm6 = xmm9[0,1,3,1,4,5,6,7]
; AVX-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[0,1,3,1,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm4 = xmm4[0],xmm6[0],xmm4[1],xmm6[1]
; AVX-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[3,1,2,3,4,5,6,7]
; AVX-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[3,1,2,3,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
; AVX-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3],xmm4[4,5,6,7]
; AVX-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm5[4,5,6,7]
; AVX-NEXT: vmovdqa %xmm1, (%rsi)
; AVX-NEXT: vmovdqa %xmm0, 16(%rsi)
; AVX-NEXT: vmovaps %ymm10, (%rdx)
; AVX-NEXT: vmovaps %ymm11, (%rcx)
; AVX-NEXT: vmovaps %ymm2, (%r8)
; AVX-NEXT: vzeroupper
; AVX-NEXT: retq
;
; AVX2-LABEL: load_i16_stride4_vf16:
; AVX2: # %bb.0:
; AVX2-NEXT: vpxor %xmm0, %xmm0, %xmm0
; AVX2-NEXT: vpblendw {{.*#+}} ymm1 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-NEXT: vextracti128 $1, %ymm1, %xmm2
; AVX2-NEXT: vpackusdw %xmm2, %xmm1, %xmm1
; AVX2-NEXT: vpblendw {{.*#+}} ymm2 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-NEXT: vextracti128 $1, %ymm2, %xmm3
; AVX2-NEXT: vpackusdw %xmm3, %xmm2, %xmm2
; AVX2-NEXT: vpackusdw %xmm1, %xmm2, %xmm1
; AVX2-NEXT: vpblendw {{.*#+}} ymm2 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-NEXT: vextracti128 $1, %ymm2, %xmm3
; AVX2-NEXT: vpackusdw %xmm3, %xmm2, %xmm2
; AVX2-NEXT: vpackusdw %xmm2, %xmm2, %xmm2
; AVX2-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX2-NEXT: vpblendw {{.*#+}} ymm0 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm3
; AVX2-NEXT: vpackusdw %xmm3, %xmm0, %xmm0
; AVX2-NEXT: vpackusdw %xmm0, %xmm0, %xmm0
; AVX2-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-NEXT: vmovdqa (%rdi), %xmm1
; AVX2-NEXT: vmovdqa 16(%rdi), %xmm2
; AVX2-NEXT: vmovdqa 32(%rdi), %xmm3
; AVX2-NEXT: vmovdqa 48(%rdi), %xmm5
; AVX2-NEXT: vmovdqa 112(%rdi), %xmm6
; AVX2-NEXT: vpshufd {{.*#+}} xmm4 = xmm6[0,2,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[0,1,1,3,4,5,6,7]
; AVX2-NEXT: vmovdqa 96(%rdi), %xmm7
; AVX2-NEXT: vpshufd {{.*#+}} xmm8 = xmm7[0,2,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm8 = xmm8[0,1,1,3,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm4 = xmm8[0],xmm4[0],xmm8[1],xmm4[1]
; AVX2-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4
; AVX2-NEXT: vmovdqa 80(%rdi), %xmm8
; AVX2-NEXT: vpshufd {{.*#+}} xmm9 = xmm8[0,2,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm9 = xmm9[1,3,2,3,4,5,6,7]
; AVX2-NEXT: vmovdqa 64(%rdi), %xmm10
; AVX2-NEXT: vpshufd {{.*#+}} xmm11 = xmm10[0,2,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm11 = xmm11[1,3,2,3,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm9 = xmm11[0],xmm9[0],xmm11[1],xmm9[1]
; AVX2-NEXT: vinserti128 $1, %xmm9, %ymm0, %ymm9
; AVX2-NEXT: vpblendd {{.*#+}} ymm4 = ymm9[0,1,2,3,4,5],ymm4[6,7]
; AVX2-NEXT: vpshufd {{.*#+}} xmm9 = xmm5[0,2,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm9 = xmm9[0,1,1,3,4,5,6,7]
; AVX2-NEXT: vpshufd {{.*#+}} xmm11 = xmm3[0,2,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm11 = xmm11[0,1,1,3,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm9 = xmm11[0],xmm9[0],xmm11[1],xmm9[1]
; AVX2-NEXT: vpshufd {{.*#+}} xmm11 = xmm2[0,2,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm11 = xmm11[1,3,2,3,4,5,6,7]
; AVX2-NEXT: vpshufd {{.*#+}} xmm12 = xmm1[0,2,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm12 = xmm12[1,3,2,3,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm11 = xmm12[0],xmm11[0],xmm12[1],xmm11[1]
; AVX2-NEXT: vpblendd {{.*#+}} xmm9 = xmm11[0,1],xmm9[2,3]
; AVX2-NEXT: vpblendd {{.*#+}} ymm4 = ymm9[0,1,2,3],ymm4[4,5,6,7]
; AVX2-NEXT: vpshufd {{.*#+}} xmm6 = xmm6[3,1,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm9 = xmm6[0,1,2,0,4,5,6,7]
; AVX2-NEXT: vpshufd {{.*#+}} xmm7 = xmm7[3,1,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm11 = xmm7[0,1,2,0,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm9 = xmm11[0],xmm9[0],xmm11[1],xmm9[1]
; AVX2-NEXT: vinserti128 $1, %xmm9, %ymm0, %ymm9
; AVX2-NEXT: vpshufd {{.*#+}} xmm8 = xmm8[3,1,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm11 = xmm8[2,0,2,3,4,5,6,7]
; AVX2-NEXT: vpshufd {{.*#+}} xmm10 = xmm10[3,1,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm12 = xmm10[2,0,2,3,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm11 = xmm12[0],xmm11[0],xmm12[1],xmm11[1]
; AVX2-NEXT: vinserti128 $1, %xmm11, %ymm0, %ymm11
; AVX2-NEXT: vpblendd {{.*#+}} ymm9 = ymm11[0,1,2,3,4,5],ymm9[6,7]
; AVX2-NEXT: vpshufd {{.*#+}} xmm5 = xmm5[3,1,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm11 = xmm5[0,1,2,0,4,5,6,7]
; AVX2-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[3,1,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm12 = xmm3[0,1,2,0,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm11 = xmm12[0],xmm11[0],xmm12[1],xmm11[1]
; AVX2-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[3,1,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm12 = xmm2[2,0,2,3,4,5,6,7]
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[3,1,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm13 = xmm1[2,0,2,3,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm12 = xmm13[0],xmm12[0],xmm13[1],xmm12[1]
; AVX2-NEXT: vpblendd {{.*#+}} xmm11 = xmm12[0,1],xmm11[2,3]
; AVX2-NEXT: vpblendd {{.*#+}} ymm9 = ymm11[0,1,2,3],ymm9[4,5,6,7]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm6 = xmm6[0,1,3,1,4,5,6,7]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm7 = xmm7[0,1,3,1,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm6 = xmm7[0],xmm6[0],xmm7[1],xmm6[1]
; AVX2-NEXT: vinserti128 $1, %xmm6, %ymm0, %ymm6
; AVX2-NEXT: vpshuflw {{.*#+}} xmm7 = xmm8[3,1,2,3,4,5,6,7]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm8 = xmm10[3,1,2,3,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm7 = xmm8[0],xmm7[0],xmm8[1],xmm7[1]
; AVX2-NEXT: vinserti128 $1, %xmm7, %ymm0, %ymm7
; AVX2-NEXT: vpblendd {{.*#+}} ymm6 = ymm7[0,1,2,3,4,5],ymm6[6,7]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm5 = xmm5[0,1,3,1,4,5,6,7]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[0,1,3,1,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm3[0],xmm5[0],xmm3[1],xmm5[1]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[3,1,2,3,4,5,6,7]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[3,1,2,3,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
; AVX2-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0,1],xmm3[2,3]
; AVX2-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm6[4,5,6,7]
; AVX2-NEXT: vmovdqa %ymm0, (%rsi)
; AVX2-NEXT: vmovdqa %ymm4, (%rdx)
; AVX2-NEXT: vmovdqa %ymm9, (%rcx)
; AVX2-NEXT: vmovdqa %ymm1, (%r8)
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
; AVX2-FP-LABEL: load_i16_stride4_vf16:
; AVX2-FP: # %bb.0:
; AVX2-FP-NEXT: vpxor %xmm0, %xmm0, %xmm0
; AVX2-FP-NEXT: vpblendw {{.*#+}} ymm1 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-FP-NEXT: vextracti128 $1, %ymm1, %xmm2
; AVX2-FP-NEXT: vpackusdw %xmm2, %xmm1, %xmm1
; AVX2-FP-NEXT: vpblendw {{.*#+}} ymm2 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-FP-NEXT: vextracti128 $1, %ymm2, %xmm3
; AVX2-FP-NEXT: vpackusdw %xmm3, %xmm2, %xmm2
; AVX2-FP-NEXT: vpackusdw %xmm1, %xmm2, %xmm1
; AVX2-FP-NEXT: vpblendw {{.*#+}} ymm2 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-FP-NEXT: vextracti128 $1, %ymm2, %xmm3
; AVX2-FP-NEXT: vpackusdw %xmm3, %xmm2, %xmm2
; AVX2-FP-NEXT: vpackusdw %xmm2, %xmm2, %xmm2
; AVX2-FP-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX2-FP-NEXT: vpblendw {{.*#+}} ymm0 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-FP-NEXT: vextracti128 $1, %ymm0, %xmm3
; AVX2-FP-NEXT: vpackusdw %xmm3, %xmm0, %xmm0
; AVX2-FP-NEXT: vpackusdw %xmm0, %xmm0, %xmm0
; AVX2-FP-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-FP-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
; AVX2-FP-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FP-NEXT: vmovdqa (%rdi), %xmm1
; AVX2-FP-NEXT: vmovdqa 16(%rdi), %xmm3
; AVX2-FP-NEXT: vmovdqa 32(%rdi), %xmm4
; AVX2-FP-NEXT: vmovdqa 48(%rdi), %xmm5
; AVX2-FP-NEXT: vmovdqa {{.*#+}} xmm2 = [0,1,2,3,2,3,10,11,8,9,10,11,12,13,14,15]
; AVX2-FP-NEXT: vmovdqa 112(%rdi), %xmm6
; AVX2-FP-NEXT: vpshufb %xmm2, %xmm6, %xmm7
; AVX2-FP-NEXT: vmovdqa 96(%rdi), %xmm8
; AVX2-FP-NEXT: vpshufb %xmm2, %xmm8, %xmm9
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm7 = xmm9[0],xmm7[0],xmm9[1],xmm7[1]
; AVX2-FP-NEXT: vinserti128 $1, %xmm7, %ymm0, %ymm7
; AVX2-FP-NEXT: vmovdqa {{.*#+}} xmm9 = [2,3,10,11,8,9,10,11,8,9,10,11,12,13,14,15]
; AVX2-FP-NEXT: vmovdqa 80(%rdi), %xmm10
; AVX2-FP-NEXT: vpshufb %xmm9, %xmm10, %xmm11
; AVX2-FP-NEXT: vmovdqa 64(%rdi), %xmm12
; AVX2-FP-NEXT: vpshufb %xmm9, %xmm12, %xmm13
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm11 = xmm13[0],xmm11[0],xmm13[1],xmm11[1]
; AVX2-FP-NEXT: vinserti128 $1, %xmm11, %ymm0, %ymm11
; AVX2-FP-NEXT: vpblendd {{.*#+}} ymm7 = ymm11[0,1,2,3,4,5],ymm7[6,7]
; AVX2-FP-NEXT: vpshufb %xmm2, %xmm5, %xmm11
; AVX2-FP-NEXT: vpshufb %xmm2, %xmm4, %xmm2
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm2[0],xmm11[0],xmm2[1],xmm11[1]
; AVX2-FP-NEXT: vpshufb %xmm9, %xmm3, %xmm11
; AVX2-FP-NEXT: vpshufb %xmm9, %xmm1, %xmm9
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm9 = xmm9[0],xmm11[0],xmm9[1],xmm11[1]
; AVX2-FP-NEXT: vpblendd {{.*#+}} xmm2 = xmm9[0,1],xmm2[2,3]
; AVX2-FP-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm7[4,5,6,7]
; AVX2-FP-NEXT: vpshufd {{.*#+}} xmm6 = xmm6[3,1,2,3]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm7 = xmm6[0,1,2,0,4,5,6,7]
; AVX2-FP-NEXT: vpshufd {{.*#+}} xmm8 = xmm8[3,1,2,3]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm9 = xmm8[0,1,2,0,4,5,6,7]
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm7 = xmm9[0],xmm7[0],xmm9[1],xmm7[1]
; AVX2-FP-NEXT: vinserti128 $1, %xmm7, %ymm0, %ymm7
; AVX2-FP-NEXT: vpshufd {{.*#+}} xmm9 = xmm10[3,1,2,3]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm10 = xmm9[2,0,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpshufd {{.*#+}} xmm11 = xmm12[3,1,2,3]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm12 = xmm11[2,0,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm10 = xmm12[0],xmm10[0],xmm12[1],xmm10[1]
; AVX2-FP-NEXT: vinserti128 $1, %xmm10, %ymm0, %ymm10
; AVX2-FP-NEXT: vpblendd {{.*#+}} ymm7 = ymm10[0,1,2,3,4,5],ymm7[6,7]
; AVX2-FP-NEXT: vpshufd {{.*#+}} xmm5 = xmm5[3,1,2,3]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm10 = xmm5[0,1,2,0,4,5,6,7]
; AVX2-FP-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[3,1,2,3]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm12 = xmm4[0,1,2,0,4,5,6,7]
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm10 = xmm12[0],xmm10[0],xmm12[1],xmm10[1]
; AVX2-FP-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[3,1,2,3]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm12 = xmm3[2,0,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[3,1,2,3]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm13 = xmm1[2,0,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm12 = xmm13[0],xmm12[0],xmm13[1],xmm12[1]
; AVX2-FP-NEXT: vpblendd {{.*#+}} xmm10 = xmm12[0,1],xmm10[2,3]
; AVX2-FP-NEXT: vpblendd {{.*#+}} ymm7 = ymm10[0,1,2,3],ymm7[4,5,6,7]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm6 = xmm6[0,1,3,1,4,5,6,7]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm8 = xmm8[0,1,3,1,4,5,6,7]
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm6 = xmm8[0],xmm6[0],xmm8[1],xmm6[1]
; AVX2-FP-NEXT: vinserti128 $1, %xmm6, %ymm0, %ymm6
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm8 = xmm9[3,1,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm9 = xmm11[3,1,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm8 = xmm9[0],xmm8[0],xmm9[1],xmm8[1]
; AVX2-FP-NEXT: vinserti128 $1, %xmm8, %ymm0, %ymm8
; AVX2-FP-NEXT: vpblendd {{.*#+}} ymm6 = ymm8[0,1,2,3,4,5],ymm6[6,7]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm5 = xmm5[0,1,3,1,4,5,6,7]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[0,1,3,1,4,5,6,7]
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm4 = xmm4[0],xmm5[0],xmm4[1],xmm5[1]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[3,1,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[3,1,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
; AVX2-FP-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0,1],xmm4[2,3]
; AVX2-FP-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm6[4,5,6,7]
; AVX2-FP-NEXT: vmovdqa %ymm0, (%rsi)
; AVX2-FP-NEXT: vmovdqa %ymm2, (%rdx)
; AVX2-FP-NEXT: vmovdqa %ymm7, (%rcx)
; AVX2-FP-NEXT: vmovdqa %ymm1, (%r8)
; AVX2-FP-NEXT: vzeroupper
; AVX2-FP-NEXT: retq
;
; AVX2-FCP-LABEL: load_i16_stride4_vf16:
; AVX2-FCP: # %bb.0:
; AVX2-FCP-NEXT: vmovdqa 64(%rdi), %ymm1
; AVX2-FCP-NEXT: vmovdqa 96(%rdi), %ymm2
; AVX2-FCP-NEXT: vpxor %xmm0, %xmm0, %xmm0
; AVX2-FCP-NEXT: vpblendw {{.*#+}} ymm3 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-FCP-NEXT: vextracti128 $1, %ymm3, %xmm4
; AVX2-FCP-NEXT: vpackusdw %xmm4, %xmm3, %xmm3
; AVX2-FCP-NEXT: vpblendw {{.*#+}} ymm0 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-FCP-NEXT: vextracti128 $1, %ymm0, %xmm4
; AVX2-FCP-NEXT: vpackusdw %xmm4, %xmm0, %xmm0
; AVX2-FCP-NEXT: vpackusdw %xmm3, %xmm0, %xmm0
; AVX2-FCP-NEXT: vpmovsxbd {{.*#+}} ymm3 = [0,2,2,3,0,2,4,6]
; AVX2-FCP-NEXT: vpermd %ymm2, %ymm3, %ymm4
; AVX2-FCP-NEXT: vmovdqa {{.*#+}} ymm5 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,16,17,20,21,24,25,28,29]
; AVX2-FCP-NEXT: vpshufb %ymm5, %ymm4, %ymm6
; AVX2-FCP-NEXT: vpermd %ymm1, %ymm3, %ymm3
; AVX2-FCP-NEXT: vpshufb %ymm5, %ymm3, %ymm7
; AVX2-FCP-NEXT: vpblendd {{.*#+}} ymm6 = ymm7[0,1,2,3,4,5],ymm6[6,7]
; AVX2-FCP-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm6[4,5,6,7]
; AVX2-FCP-NEXT: vmovdqa (%rdi), %xmm6
; AVX2-FCP-NEXT: vmovdqa 16(%rdi), %xmm7
; AVX2-FCP-NEXT: vmovdqa 32(%rdi), %xmm8
; AVX2-FCP-NEXT: vmovdqa 48(%rdi), %xmm9
; AVX2-FCP-NEXT: vmovdqa {{.*#+}} xmm10 = [0,1,2,3,2,3,10,11,8,9,10,11,12,13,14,15]
; AVX2-FCP-NEXT: vpshufb %xmm10, %xmm9, %xmm11
; AVX2-FCP-NEXT: vpshufb %xmm10, %xmm8, %xmm10
; AVX2-FCP-NEXT: vpunpckldq {{.*#+}} xmm10 = xmm10[0],xmm11[0],xmm10[1],xmm11[1]
; AVX2-FCP-NEXT: vmovdqa {{.*#+}} xmm11 = [2,3,10,11,8,9,10,11,8,9,10,11,12,13,14,15]
; AVX2-FCP-NEXT: vpshufb %xmm11, %xmm7, %xmm12
; AVX2-FCP-NEXT: vpshufb %xmm11, %xmm6, %xmm11
; AVX2-FCP-NEXT: vpunpckldq {{.*#+}} xmm11 = xmm11[0],xmm12[0],xmm11[1],xmm12[1]
; AVX2-FCP-NEXT: vpblendd {{.*#+}} xmm10 = xmm11[0,1],xmm10[2,3]
; AVX2-FCP-NEXT: vmovdqa {{.*#+}} ymm11 = [2,3,6,7,10,11,14,15,14,15,10,11,12,13,14,15,18,19,22,23,26,27,30,31,18,19,22,23,26,27,30,31]
; AVX2-FCP-NEXT: vpshufb %ymm11, %ymm4, %ymm4
; AVX2-FCP-NEXT: vpshufb %ymm11, %ymm3, %ymm3
; AVX2-FCP-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm4[6,7]
; AVX2-FCP-NEXT: vpblendd {{.*#+}} ymm3 = ymm10[0,1,2,3],ymm3[4,5,6,7]
; AVX2-FCP-NEXT: vpmovsxbd {{.*#+}} ymm4 = [1,3,2,3,1,3,5,7]
; AVX2-FCP-NEXT: vpermd %ymm2, %ymm4, %ymm2
; AVX2-FCP-NEXT: vpshufb %ymm5, %ymm2, %ymm10
; AVX2-FCP-NEXT: vpermd %ymm1, %ymm4, %ymm1
; AVX2-FCP-NEXT: vpshufb %ymm5, %ymm1, %ymm4
; AVX2-FCP-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm10[6,7]
; AVX2-FCP-NEXT: vpshufd {{.*#+}} xmm5 = xmm9[3,1,2,3]
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm9 = xmm5[0,1,2,0,4,5,6,7]
; AVX2-FCP-NEXT: vpshufd {{.*#+}} xmm8 = xmm8[3,1,2,3]
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm10 = xmm8[0,1,2,0,4,5,6,7]
; AVX2-FCP-NEXT: vpunpckldq {{.*#+}} xmm9 = xmm10[0],xmm9[0],xmm10[1],xmm9[1]
; AVX2-FCP-NEXT: vpshufd {{.*#+}} xmm7 = xmm7[3,1,2,3]
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm10 = xmm7[2,0,2,3,4,5,6,7]
; AVX2-FCP-NEXT: vpshufd {{.*#+}} xmm6 = xmm6[3,1,2,3]
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm12 = xmm6[2,0,2,3,4,5,6,7]
; AVX2-FCP-NEXT: vpunpckldq {{.*#+}} xmm10 = xmm12[0],xmm10[0],xmm12[1],xmm10[1]
; AVX2-FCP-NEXT: vpblendd {{.*#+}} xmm9 = xmm10[0,1],xmm9[2,3]
; AVX2-FCP-NEXT: vpblendd {{.*#+}} ymm4 = ymm9[0,1,2,3],ymm4[4,5,6,7]
; AVX2-FCP-NEXT: vpshufb %ymm11, %ymm2, %ymm2
; AVX2-FCP-NEXT: vpshufb %ymm11, %ymm1, %ymm1
; AVX2-FCP-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm2 = xmm5[0,1,3,1,4,5,6,7]
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm5 = xmm8[0,1,3,1,4,5,6,7]
; AVX2-FCP-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm5[0],xmm2[0],xmm5[1],xmm2[1]
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm5 = xmm7[3,1,2,3,4,5,6,7]
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm6 = xmm6[3,1,2,3,4,5,6,7]
; AVX2-FCP-NEXT: vpunpckldq {{.*#+}} xmm5 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
; AVX2-FCP-NEXT: vpblendd {{.*#+}} xmm2 = xmm5[0,1],xmm2[2,3]
; AVX2-FCP-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FCP-NEXT: vmovdqa %ymm0, (%rsi)
; AVX2-FCP-NEXT: vmovdqa %ymm3, (%rdx)
; AVX2-FCP-NEXT: vmovdqa %ymm4, (%rcx)
; AVX2-FCP-NEXT: vmovdqa %ymm1, (%r8)
; AVX2-FCP-NEXT: vzeroupper
; AVX2-FCP-NEXT: retq
;
; AVX512-LABEL: load_i16_stride4_vf16:
; AVX512: # %bb.0:
; AVX512-NEXT: vmovdqa 64(%rdi), %ymm0
; AVX512-NEXT: vpmovqw %ymm0, %xmm0
; AVX512-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX512-NEXT: vmovdqa 112(%rdi), %xmm1
; AVX512-NEXT: vpshufd {{.*#+}} xmm2 = xmm1[0,2,2,3]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm3 = xmm2[0,1,0,2,4,5,6,7]
; AVX512-NEXT: vmovdqa 96(%rdi), %xmm4
; AVX512-NEXT: vpshufd {{.*#+}} xmm5 = xmm4[0,2,2,3]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm6 = xmm5[0,1,0,2,4,5,6,7]
; AVX512-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm6[0],xmm3[0],xmm6[1],xmm3[1]
; AVX512-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
; AVX512-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm3[6,7]
; AVX512-NEXT: vmovdqa64 (%rdi), %zmm3
; AVX512-NEXT: vpmovqw %zmm3, %xmm6
; AVX512-NEXT: vpblendd {{.*#+}} ymm0 = ymm6[0,1,2,3],ymm0[4,5,6,7]
; AVX512-NEXT: vmovdqa 64(%rdi), %xmm6
; AVX512-NEXT: vmovdqa 80(%rdi), %xmm7
; AVX512-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[0,1,1,3,4,5,6,7]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm5 = xmm5[0,1,1,3,4,5,6,7]
; AVX512-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm5[0],xmm2[0],xmm5[1],xmm2[1]
; AVX512-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX512-NEXT: vpshufd {{.*#+}} xmm5 = xmm7[0,2,2,3]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm5 = xmm5[1,3,2,3,4,5,6,7]
; AVX512-NEXT: vpshufd {{.*#+}} xmm8 = xmm6[0,2,2,3]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm8 = xmm8[1,3,2,3,4,5,6,7]
; AVX512-NEXT: vpunpckldq {{.*#+}} xmm5 = xmm8[0],xmm5[0],xmm8[1],xmm5[1]
; AVX512-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
; AVX512-NEXT: vpblendd {{.*#+}} ymm2 = ymm5[0,1,2,3,4,5],ymm2[6,7]
; AVX512-NEXT: vpsrlq $16, %zmm3, %zmm5
; AVX512-NEXT: vpmovqw %zmm5, %xmm5
; AVX512-NEXT: vpblendd {{.*#+}} ymm2 = ymm5[0,1,2,3],ymm2[4,5,6,7]
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[3,1,2,3]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm5 = xmm1[0,1,2,0,4,5,6,7]
; AVX512-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[3,1,2,3]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm8 = xmm4[0,1,2,0,4,5,6,7]
; AVX512-NEXT: vpunpckldq {{.*#+}} xmm5 = xmm8[0],xmm5[0],xmm8[1],xmm5[1]
; AVX512-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
; AVX512-NEXT: vpshufd {{.*#+}} xmm7 = xmm7[3,1,2,3]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm8 = xmm7[2,0,2,3,4,5,6,7]
; AVX512-NEXT: vpshufd {{.*#+}} xmm6 = xmm6[3,1,2,3]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm9 = xmm6[2,0,2,3,4,5,6,7]
; AVX512-NEXT: vpunpckldq {{.*#+}} xmm8 = xmm9[0],xmm8[0],xmm9[1],xmm8[1]
; AVX512-NEXT: vinserti128 $1, %xmm8, %ymm0, %ymm8
; AVX512-NEXT: vpblendd {{.*#+}} ymm5 = ymm8[0,1,2,3,4,5],ymm5[6,7]
; AVX512-NEXT: vpsrlq $32, %zmm3, %zmm8
; AVX512-NEXT: vpmovqw %zmm8, %xmm8
; AVX512-NEXT: vpblendd {{.*#+}} ymm5 = ymm8[0,1,2,3],ymm5[4,5,6,7]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[0,1,3,1,4,5,6,7]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[0,1,3,1,4,5,6,7]
; AVX512-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
; AVX512-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX512-NEXT: vpshuflw {{.*#+}} xmm4 = xmm7[3,1,2,3,4,5,6,7]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm6 = xmm6[3,1,2,3,4,5,6,7]
; AVX512-NEXT: vpunpckldq {{.*#+}} xmm4 = xmm6[0],xmm4[0],xmm6[1],xmm4[1]
; AVX512-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4
; AVX512-NEXT: vpblendd {{.*#+}} ymm1 = ymm4[0,1,2,3,4,5],ymm1[6,7]
; AVX512-NEXT: vpsrlq $48, %zmm3, %zmm3
; AVX512-NEXT: vpmovqw %zmm3, %xmm3
; AVX512-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm1[4,5,6,7]
; AVX512-NEXT: vmovdqa %ymm0, (%rsi)
; AVX512-NEXT: vmovdqa %ymm2, (%rdx)
; AVX512-NEXT: vmovdqa %ymm5, (%rcx)
; AVX512-NEXT: vmovdqa %ymm1, (%r8)
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
;
; AVX512-FCP-LABEL: load_i16_stride4_vf16:
; AVX512-FCP: # %bb.0:
; AVX512-FCP-NEXT: vbroadcasti128 {{.*#+}} ymm0 = [16,17,20,21,0,0,0,0,24,25,28,29,0,0,0,0,16,17,20,21,0,0,0,0,24,25,28,29,0,0,0,0]
; AVX512-FCP-NEXT: # ymm0 = mem[0,1,0,1]
; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} ymm1 = [0,2,2,3,0,2,4,6]
; AVX512-FCP-NEXT: vmovdqa 96(%rdi), %ymm2
; AVX512-FCP-NEXT: vpermd %ymm2, %ymm1, %ymm3
; AVX512-FCP-NEXT: vpshufb %ymm0, %ymm3, %ymm4
; AVX512-FCP-NEXT: vmovdqa {{.*#+}} ymm5 = [0,1,4,5,u,u,u,u,8,9,12,13,12,13,14,15,16,17,20,21,u,u,u,u,24,25,28,29,u,u,u,u]
; AVX512-FCP-NEXT: vmovdqa 64(%rdi), %ymm6
; AVX512-FCP-NEXT: vpermd %ymm6, %ymm1, %ymm1
; AVX512-FCP-NEXT: vpshufb %ymm5, %ymm1, %ymm7
; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} ymm8 = [0,2,2,3,4,6,12,14]
; AVX512-FCP-NEXT: vpermt2d %ymm4, %ymm8, %ymm7
; AVX512-FCP-NEXT: vmovdqa64 (%rdi), %zmm4
; AVX512-FCP-NEXT: vpmovqw %zmm4, %xmm9
; AVX512-FCP-NEXT: vpblendd {{.*#+}} ymm7 = ymm9[0,1,2,3],ymm7[4,5,6,7]
; AVX512-FCP-NEXT: vmovdqa {{.*#+}} ymm9 = [2,3,6,7,10,11,14,15,14,15,10,11,12,13,14,15,18,19,22,23,26,27,30,31,18,19,22,23,26,27,30,31]
; AVX512-FCP-NEXT: vpshufb %ymm9, %ymm3, %ymm3
; AVX512-FCP-NEXT: vpshufb %ymm9, %ymm1, %ymm1
; AVX512-FCP-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm3[6,7]
; AVX512-FCP-NEXT: vpsrlq $16, %zmm4, %zmm3
; AVX512-FCP-NEXT: vpmovqw %zmm3, %xmm3
; AVX512-FCP-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm1[4,5,6,7]
; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} ymm3 = [1,3,2,3,1,3,5,7]
; AVX512-FCP-NEXT: vpermd %ymm2, %ymm3, %ymm2
; AVX512-FCP-NEXT: vpshufb %ymm0, %ymm2, %ymm0
; AVX512-FCP-NEXT: vpermd %ymm6, %ymm3, %ymm3
; AVX512-FCP-NEXT: vpshufb %ymm5, %ymm3, %ymm5
; AVX512-FCP-NEXT: vpermt2d %ymm0, %ymm8, %ymm5
; AVX512-FCP-NEXT: vpsrlq $32, %zmm4, %zmm0
; AVX512-FCP-NEXT: vpmovqw %zmm0, %xmm0
; AVX512-FCP-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm5[4,5,6,7]
; AVX512-FCP-NEXT: vpshufb %ymm9, %ymm2, %ymm2
; AVX512-FCP-NEXT: vpshufb %ymm9, %ymm3, %ymm3
; AVX512-FCP-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
; AVX512-FCP-NEXT: vpsrlq $48, %zmm4, %zmm3
; AVX512-FCP-NEXT: vpmovqw %zmm3, %xmm3
; AVX512-FCP-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
; AVX512-FCP-NEXT: vmovdqa %ymm7, (%rsi)
; AVX512-FCP-NEXT: vmovdqa %ymm1, (%rdx)
; AVX512-FCP-NEXT: vmovdqa %ymm0, (%rcx)
; AVX512-FCP-NEXT: vmovdqa %ymm2, (%r8)
; AVX512-FCP-NEXT: vzeroupper
; AVX512-FCP-NEXT: retq
;
; AVX512DQ-LABEL: load_i16_stride4_vf16:
; AVX512DQ: # %bb.0:
; AVX512DQ-NEXT: vmovdqa 64(%rdi), %ymm0
; AVX512DQ-NEXT: vpmovqw %ymm0, %xmm0
; AVX512DQ-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX512DQ-NEXT: vmovdqa 112(%rdi), %xmm1
; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm2 = xmm1[0,2,2,3]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm3 = xmm2[0,1,0,2,4,5,6,7]
; AVX512DQ-NEXT: vmovdqa 96(%rdi), %xmm4
; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm5 = xmm4[0,2,2,3]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm6 = xmm5[0,1,0,2,4,5,6,7]
; AVX512DQ-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm6[0],xmm3[0],xmm6[1],xmm3[1]
; AVX512DQ-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
; AVX512DQ-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm3[6,7]
; AVX512DQ-NEXT: vmovdqa64 (%rdi), %zmm3
; AVX512DQ-NEXT: vpmovqw %zmm3, %xmm6
; AVX512DQ-NEXT: vpblendd {{.*#+}} ymm0 = ymm6[0,1,2,3],ymm0[4,5,6,7]
; AVX512DQ-NEXT: vmovdqa 64(%rdi), %xmm6
; AVX512DQ-NEXT: vmovdqa 80(%rdi), %xmm7
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[0,1,1,3,4,5,6,7]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm5 = xmm5[0,1,1,3,4,5,6,7]
; AVX512DQ-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm5[0],xmm2[0],xmm5[1],xmm2[1]
; AVX512DQ-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm5 = xmm7[0,2,2,3]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm5 = xmm5[1,3,2,3,4,5,6,7]
; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm8 = xmm6[0,2,2,3]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm8 = xmm8[1,3,2,3,4,5,6,7]
; AVX512DQ-NEXT: vpunpckldq {{.*#+}} xmm5 = xmm8[0],xmm5[0],xmm8[1],xmm5[1]
; AVX512DQ-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
; AVX512DQ-NEXT: vpblendd {{.*#+}} ymm2 = ymm5[0,1,2,3,4,5],ymm2[6,7]
; AVX512DQ-NEXT: vpsrlq $16, %zmm3, %zmm5
; AVX512DQ-NEXT: vpmovqw %zmm5, %xmm5
; AVX512DQ-NEXT: vpblendd {{.*#+}} ymm2 = ymm5[0,1,2,3],ymm2[4,5,6,7]
; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[3,1,2,3]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm5 = xmm1[0,1,2,0,4,5,6,7]
; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[3,1,2,3]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm8 = xmm4[0,1,2,0,4,5,6,7]
; AVX512DQ-NEXT: vpunpckldq {{.*#+}} xmm5 = xmm8[0],xmm5[0],xmm8[1],xmm5[1]
; AVX512DQ-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm7 = xmm7[3,1,2,3]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm8 = xmm7[2,0,2,3,4,5,6,7]
; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm6 = xmm6[3,1,2,3]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm9 = xmm6[2,0,2,3,4,5,6,7]
; AVX512DQ-NEXT: vpunpckldq {{.*#+}} xmm8 = xmm9[0],xmm8[0],xmm9[1],xmm8[1]
; AVX512DQ-NEXT: vinserti128 $1, %xmm8, %ymm0, %ymm8
; AVX512DQ-NEXT: vpblendd {{.*#+}} ymm5 = ymm8[0,1,2,3,4,5],ymm5[6,7]
; AVX512DQ-NEXT: vpsrlq $32, %zmm3, %zmm8
; AVX512DQ-NEXT: vpmovqw %zmm8, %xmm8
; AVX512DQ-NEXT: vpblendd {{.*#+}} ymm5 = ymm8[0,1,2,3],ymm5[4,5,6,7]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[0,1,3,1,4,5,6,7]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[0,1,3,1,4,5,6,7]
; AVX512DQ-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
; AVX512DQ-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm4 = xmm7[3,1,2,3,4,5,6,7]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm6 = xmm6[3,1,2,3,4,5,6,7]
; AVX512DQ-NEXT: vpunpckldq {{.*#+}} xmm4 = xmm6[0],xmm4[0],xmm6[1],xmm4[1]
; AVX512DQ-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4
; AVX512DQ-NEXT: vpblendd {{.*#+}} ymm1 = ymm4[0,1,2,3,4,5],ymm1[6,7]
; AVX512DQ-NEXT: vpsrlq $48, %zmm3, %zmm3
; AVX512DQ-NEXT: vpmovqw %zmm3, %xmm3
; AVX512DQ-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm1[4,5,6,7]
; AVX512DQ-NEXT: vmovdqa %ymm0, (%rsi)
; AVX512DQ-NEXT: vmovdqa %ymm2, (%rdx)
; AVX512DQ-NEXT: vmovdqa %ymm5, (%rcx)
; AVX512DQ-NEXT: vmovdqa %ymm1, (%r8)
; AVX512DQ-NEXT: vzeroupper
; AVX512DQ-NEXT: retq
;
; AVX512DQ-FCP-LABEL: load_i16_stride4_vf16:
; AVX512DQ-FCP: # %bb.0:
; AVX512DQ-FCP-NEXT: vbroadcasti128 {{.*#+}} ymm0 = [16,17,20,21,0,0,0,0,24,25,28,29,0,0,0,0,16,17,20,21,0,0,0,0,24,25,28,29,0,0,0,0]
; AVX512DQ-FCP-NEXT: # ymm0 = mem[0,1,0,1]
; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} ymm1 = [0,2,2,3,0,2,4,6]
; AVX512DQ-FCP-NEXT: vmovdqa 96(%rdi), %ymm2
; AVX512DQ-FCP-NEXT: vpermd %ymm2, %ymm1, %ymm3
; AVX512DQ-FCP-NEXT: vpshufb %ymm0, %ymm3, %ymm4
; AVX512DQ-FCP-NEXT: vmovdqa {{.*#+}} ymm5 = [0,1,4,5,u,u,u,u,8,9,12,13,12,13,14,15,16,17,20,21,u,u,u,u,24,25,28,29,u,u,u,u]
; AVX512DQ-FCP-NEXT: vmovdqa 64(%rdi), %ymm6
; AVX512DQ-FCP-NEXT: vpermd %ymm6, %ymm1, %ymm1
; AVX512DQ-FCP-NEXT: vpshufb %ymm5, %ymm1, %ymm7
; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} ymm8 = [0,2,2,3,4,6,12,14]
; AVX512DQ-FCP-NEXT: vpermt2d %ymm4, %ymm8, %ymm7
; AVX512DQ-FCP-NEXT: vmovdqa64 (%rdi), %zmm4
; AVX512DQ-FCP-NEXT: vpmovqw %zmm4, %xmm9
; AVX512DQ-FCP-NEXT: vpblendd {{.*#+}} ymm7 = ymm9[0,1,2,3],ymm7[4,5,6,7]
; AVX512DQ-FCP-NEXT: vmovdqa {{.*#+}} ymm9 = [2,3,6,7,10,11,14,15,14,15,10,11,12,13,14,15,18,19,22,23,26,27,30,31,18,19,22,23,26,27,30,31]
; AVX512DQ-FCP-NEXT: vpshufb %ymm9, %ymm3, %ymm3
; AVX512DQ-FCP-NEXT: vpshufb %ymm9, %ymm1, %ymm1
; AVX512DQ-FCP-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm3[6,7]
; AVX512DQ-FCP-NEXT: vpsrlq $16, %zmm4, %zmm3
; AVX512DQ-FCP-NEXT: vpmovqw %zmm3, %xmm3
; AVX512DQ-FCP-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm1[4,5,6,7]
; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} ymm3 = [1,3,2,3,1,3,5,7]
; AVX512DQ-FCP-NEXT: vpermd %ymm2, %ymm3, %ymm2
; AVX512DQ-FCP-NEXT: vpshufb %ymm0, %ymm2, %ymm0
; AVX512DQ-FCP-NEXT: vpermd %ymm6, %ymm3, %ymm3
; AVX512DQ-FCP-NEXT: vpshufb %ymm5, %ymm3, %ymm5
; AVX512DQ-FCP-NEXT: vpermt2d %ymm0, %ymm8, %ymm5
; AVX512DQ-FCP-NEXT: vpsrlq $32, %zmm4, %zmm0
; AVX512DQ-FCP-NEXT: vpmovqw %zmm0, %xmm0
; AVX512DQ-FCP-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm5[4,5,6,7]
; AVX512DQ-FCP-NEXT: vpshufb %ymm9, %ymm2, %ymm2
; AVX512DQ-FCP-NEXT: vpshufb %ymm9, %ymm3, %ymm3
; AVX512DQ-FCP-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
; AVX512DQ-FCP-NEXT: vpsrlq $48, %zmm4, %zmm3
; AVX512DQ-FCP-NEXT: vpmovqw %zmm3, %xmm3
; AVX512DQ-FCP-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
; AVX512DQ-FCP-NEXT: vmovdqa %ymm7, (%rsi)
; AVX512DQ-FCP-NEXT: vmovdqa %ymm1, (%rdx)
; AVX512DQ-FCP-NEXT: vmovdqa %ymm0, (%rcx)
; AVX512DQ-FCP-NEXT: vmovdqa %ymm2, (%r8)
; AVX512DQ-FCP-NEXT: vzeroupper
; AVX512DQ-FCP-NEXT: retq
;
; AVX512BW-LABEL: load_i16_stride4_vf16:
; AVX512BW: # %bb.0:
; AVX512BW-NEXT: vpmovsxbw {{.*#+}} ymm0 = [0,4,8,12,16,20,24,28,32,36,40,44,48,52,56,60]
; AVX512BW-NEXT: vmovdqa64 (%rdi), %zmm1
; AVX512BW-NEXT: vmovdqa64 64(%rdi), %zmm2
; AVX512BW-NEXT: vpermi2w %zmm2, %zmm1, %zmm0
; AVX512BW-NEXT: vpmovsxbw {{.*#+}} ymm3 = [1,5,9,13,17,21,25,29,33,37,41,45,49,53,57,61]
; AVX512BW-NEXT: vpermi2w %zmm2, %zmm1, %zmm3
; AVX512BW-NEXT: vpmovsxbw {{.*#+}} ymm4 = [2,6,10,14,18,22,26,30,34,38,42,46,50,54,58,62]
; AVX512BW-NEXT: vpermi2w %zmm2, %zmm1, %zmm4
; AVX512BW-NEXT: vpmovsxbw {{.*#+}} ymm5 = [3,7,11,15,19,23,27,31,35,39,43,47,51,55,59,63]
; AVX512BW-NEXT: vpermi2w %zmm2, %zmm1, %zmm5
; AVX512BW-NEXT: vmovdqa %ymm0, (%rsi)
; AVX512BW-NEXT: vmovdqa %ymm3, (%rdx)
; AVX512BW-NEXT: vmovdqa %ymm4, (%rcx)
; AVX512BW-NEXT: vmovdqa %ymm5, (%r8)
; AVX512BW-NEXT: vzeroupper
; AVX512BW-NEXT: retq
;
; AVX512BW-FCP-LABEL: load_i16_stride4_vf16:
; AVX512BW-FCP: # %bb.0:
; AVX512BW-FCP-NEXT: vpmovsxbw {{.*#+}} ymm0 = [0,4,8,12,16,20,24,28,32,36,40,44,48,52,56,60]
; AVX512BW-FCP-NEXT: vmovdqa64 (%rdi), %zmm1
; AVX512BW-FCP-NEXT: vmovdqa64 64(%rdi), %zmm2
; AVX512BW-FCP-NEXT: vpermi2w %zmm2, %zmm1, %zmm0
; AVX512BW-FCP-NEXT: vpmovsxbw {{.*#+}} ymm3 = [1,5,9,13,17,21,25,29,33,37,41,45,49,53,57,61]
; AVX512BW-FCP-NEXT: vpermi2w %zmm2, %zmm1, %zmm3
; AVX512BW-FCP-NEXT: vpmovsxbw {{.*#+}} ymm4 = [2,6,10,14,18,22,26,30,34,38,42,46,50,54,58,62]
; AVX512BW-FCP-NEXT: vpermi2w %zmm2, %zmm1, %zmm4
; AVX512BW-FCP-NEXT: vpmovsxbw {{.*#+}} ymm5 = [3,7,11,15,19,23,27,31,35,39,43,47,51,55,59,63]
; AVX512BW-FCP-NEXT: vpermi2w %zmm2, %zmm1, %zmm5
; AVX512BW-FCP-NEXT: vmovdqa %ymm0, (%rsi)
; AVX512BW-FCP-NEXT: vmovdqa %ymm3, (%rdx)
; AVX512BW-FCP-NEXT: vmovdqa %ymm4, (%rcx)
; AVX512BW-FCP-NEXT: vmovdqa %ymm5, (%r8)
; AVX512BW-FCP-NEXT: vzeroupper
; AVX512BW-FCP-NEXT: retq
;
; AVX512DQ-BW-LABEL: load_i16_stride4_vf16:
; AVX512DQ-BW: # %bb.0:
; AVX512DQ-BW-NEXT: vpmovsxbw {{.*#+}} ymm0 = [0,4,8,12,16,20,24,28,32,36,40,44,48,52,56,60]
; AVX512DQ-BW-NEXT: vmovdqa64 (%rdi), %zmm1
; AVX512DQ-BW-NEXT: vmovdqa64 64(%rdi), %zmm2
; AVX512DQ-BW-NEXT: vpermi2w %zmm2, %zmm1, %zmm0
; AVX512DQ-BW-NEXT: vpmovsxbw {{.*#+}} ymm3 = [1,5,9,13,17,21,25,29,33,37,41,45,49,53,57,61]
; AVX512DQ-BW-NEXT: vpermi2w %zmm2, %zmm1, %zmm3
; AVX512DQ-BW-NEXT: vpmovsxbw {{.*#+}} ymm4 = [2,6,10,14,18,22,26,30,34,38,42,46,50,54,58,62]
; AVX512DQ-BW-NEXT: vpermi2w %zmm2, %zmm1, %zmm4
; AVX512DQ-BW-NEXT: vpmovsxbw {{.*#+}} ymm5 = [3,7,11,15,19,23,27,31,35,39,43,47,51,55,59,63]
; AVX512DQ-BW-NEXT: vpermi2w %zmm2, %zmm1, %zmm5
; AVX512DQ-BW-NEXT: vmovdqa %ymm0, (%rsi)
; AVX512DQ-BW-NEXT: vmovdqa %ymm3, (%rdx)
; AVX512DQ-BW-NEXT: vmovdqa %ymm4, (%rcx)
; AVX512DQ-BW-NEXT: vmovdqa %ymm5, (%r8)
; AVX512DQ-BW-NEXT: vzeroupper
; AVX512DQ-BW-NEXT: retq
;
; AVX512DQ-BW-FCP-LABEL: load_i16_stride4_vf16:
; AVX512DQ-BW-FCP: # %bb.0:
; AVX512DQ-BW-FCP-NEXT: vpmovsxbw {{.*#+}} ymm0 = [0,4,8,12,16,20,24,28,32,36,40,44,48,52,56,60]
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 (%rdi), %zmm1
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 64(%rdi), %zmm2
; AVX512DQ-BW-FCP-NEXT: vpermi2w %zmm2, %zmm1, %zmm0
; AVX512DQ-BW-FCP-NEXT: vpmovsxbw {{.*#+}} ymm3 = [1,5,9,13,17,21,25,29,33,37,41,45,49,53,57,61]
; AVX512DQ-BW-FCP-NEXT: vpermi2w %zmm2, %zmm1, %zmm3
; AVX512DQ-BW-FCP-NEXT: vpmovsxbw {{.*#+}} ymm4 = [2,6,10,14,18,22,26,30,34,38,42,46,50,54,58,62]
; AVX512DQ-BW-FCP-NEXT: vpermi2w %zmm2, %zmm1, %zmm4
; AVX512DQ-BW-FCP-NEXT: vpmovsxbw {{.*#+}} ymm5 = [3,7,11,15,19,23,27,31,35,39,43,47,51,55,59,63]
; AVX512DQ-BW-FCP-NEXT: vpermi2w %zmm2, %zmm1, %zmm5
; AVX512DQ-BW-FCP-NEXT: vmovdqa %ymm0, (%rsi)
; AVX512DQ-BW-FCP-NEXT: vmovdqa %ymm3, (%rdx)
; AVX512DQ-BW-FCP-NEXT: vmovdqa %ymm4, (%rcx)
; AVX512DQ-BW-FCP-NEXT: vmovdqa %ymm5, (%r8)
; AVX512DQ-BW-FCP-NEXT: vzeroupper
; AVX512DQ-BW-FCP-NEXT: retq
%wide.vec = load <64 x i16>, ptr %in.vec, align 64
%strided.vec0 = shufflevector <64 x i16> %wide.vec, <64 x i16> poison, <16 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28, i32 32, i32 36, i32 40, i32 44, i32 48, i32 52, i32 56, i32 60>
%strided.vec1 = shufflevector <64 x i16> %wide.vec, <64 x i16> poison, <16 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29, i32 33, i32 37, i32 41, i32 45, i32 49, i32 53, i32 57, i32 61>
%strided.vec2 = shufflevector <64 x i16> %wide.vec, <64 x i16> poison, <16 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30, i32 34, i32 38, i32 42, i32 46, i32 50, i32 54, i32 58, i32 62>
%strided.vec3 = shufflevector <64 x i16> %wide.vec, <64 x i16> poison, <16 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31, i32 35, i32 39, i32 43, i32 47, i32 51, i32 55, i32 59, i32 63>
store <16 x i16> %strided.vec0, ptr %out.vec0, align 64
store <16 x i16> %strided.vec1, ptr %out.vec1, align 64
store <16 x i16> %strided.vec2, ptr %out.vec2, align 64
store <16 x i16> %strided.vec3, ptr %out.vec3, align 64
ret void
}
define void @load_i16_stride4_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3) nounwind {
; SSE-LABEL: load_i16_stride4_vf32:
; SSE: # %bb.0:
; SSE-NEXT: subq $248, %rsp
; SSE-NEXT: movdqa 160(%rdi), %xmm3
; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa 128(%rdi), %xmm4
; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa 144(%rdi), %xmm5
; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa 96(%rdi), %xmm2
; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa 112(%rdi), %xmm6
; SSE-NEXT: movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa 64(%rdi), %xmm1
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa 80(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
; SSE-NEXT: movdqa %xmm1, (%rsp) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[0,2,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm6[0,2,2,3]
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,1,0,2,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm13 = xmm2[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm13[0,1,0,2,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm15 = xmm5[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm15[0,2,2,3,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm11 = xmm4[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm11[0,2,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT: movdqa 176(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm10 = xmm0[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm10[0,1,0,2,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm9 = xmm3[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm9[0,1,0,2,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa (%rdi), %xmm1
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa 16(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm0[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm8[0,2,2,3,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm12 = xmm1[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm12[0,2,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT: movdqa 32(%rdi), %xmm2
; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa 48(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm0[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm7[0,1,0,2,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm2[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm6[0,1,0,2,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa 192(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa 208(%rdi), %xmm1
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm1[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm5[0,2,2,3,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm0[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm4[0,2,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT: movdqa 224(%rdi), %xmm2
; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa 240(%rdi), %xmm1
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm1[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm3[0,1,0,2,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm14 = xmm2[0,1,0,2,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm14 = xmm14[0],xmm1[0],xmm14[1],xmm1[1]
; SSE-NEXT: movsd {{.*#+}} xmm14 = xmm0[0],xmm14[1]
; SSE-NEXT: movapd %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw $237, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[1,3,2,3,4,5,6,7]
; SSE-NEXT: pshuflw $237, (%rsp), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[1,3,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT: pshuflw $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[0,1,1,3,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm13 = xmm13[0,1,1,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm13 = xmm13[0],xmm0[0],xmm13[1],xmm0[1]
; SSE-NEXT: movsd {{.*#+}} xmm13 = xmm1[0],xmm13[1]
; SSE-NEXT: movapd %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm15[1,3,2,3,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm11[1,3,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm10[0,1,1,3,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm9 = xmm9[0,1,1,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm9 = xmm9[0],xmm0[0],xmm9[1],xmm0[1]
; SSE-NEXT: movsd {{.*#+}} xmm9 = xmm1[0],xmm9[1]
; SSE-NEXT: movapd %xmm9, (%rsp) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm8[1,3,2,3,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm12[1,3,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm7[0,1,1,3,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm13 = xmm6[0,1,1,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm13 = xmm13[0],xmm0[0],xmm13[1],xmm0[1]
; SSE-NEXT: movsd {{.*#+}} xmm13 = xmm1[0],xmm13[1]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm5[1,3,2,3,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm4[1,3,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm3[0,1,1,3,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm15 = xmm2[0,1,1,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm15 = xmm15[0],xmm0[0],xmm15[1],xmm0[1]
; SSE-NEXT: movsd {{.*#+}} xmm15 = xmm1[0],xmm15[1]
; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[3,1,2,3]
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[3,1,2,3]
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[2,0,2,3,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[2,0,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[3,1,2,3]
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
; SSE-NEXT: # xmm9 = mem[3,1,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,1,2,0,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm5 = xmm9[0,1,2,0,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1]
; SSE-NEXT: movsd {{.*#+}} xmm5 = xmm1[0],xmm5[1]
; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[3,1,2,3]
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[3,1,2,3]
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[2,0,2,3,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[2,0,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[3,1,2,3]
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
; SSE-NEXT: # xmm14 = mem[3,1,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,1,2,0,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm11 = xmm14[0,1,2,0,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm11 = xmm11[0],xmm0[0],xmm11[1],xmm0[1]
; SSE-NEXT: movsd {{.*#+}} xmm11 = xmm1[0],xmm11[1]
; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[3,1,2,3]
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[3,1,2,3]
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[2,0,2,3,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm1[2,0,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
; SSE-NEXT: # xmm12 = mem[3,1,2,3]
; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
; SSE-NEXT: # xmm10 = mem[3,1,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm12[0,1,2,0,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm4 = xmm10[0,1,2,0,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
; SSE-NEXT: movsd {{.*#+}} xmm4 = xmm3[0],xmm4[1]
; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
; SSE-NEXT: # xmm8 = mem[3,1,2,3]
; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
; SSE-NEXT: # xmm7 = mem[3,1,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm8[2,0,2,3,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm7[2,0,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
; SSE-NEXT: # xmm6 = mem[3,1,2,3]
; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; SSE-NEXT: # xmm3 = mem[3,1,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm6[0,1,2,0,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm3[0,1,2,0,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT: pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[3,1,2,3,4,5,6,7]
; SSE-NEXT: pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; SSE-NEXT: # xmm2 = mem[3,1,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT: pshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[0,1,3,1,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm9 = xmm9[0,1,3,1,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm9 = xmm9[0],xmm0[0],xmm9[1],xmm0[1]
; SSE-NEXT: movsd {{.*#+}} xmm9 = xmm2[0],xmm9[1]
; SSE-NEXT: pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[3,1,2,3,4,5,6,7]
; SSE-NEXT: pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; SSE-NEXT: # xmm2 = mem[3,1,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT: pshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[0,1,3,1,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm14 = xmm14[0,1,3,1,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm14 = xmm14[0],xmm0[0],xmm14[1],xmm0[1]
; SSE-NEXT: movsd {{.*#+}} xmm14 = xmm2[0],xmm14[1]
; SSE-NEXT: pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[3,1,2,3,4,5,6,7]
; SSE-NEXT: pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; SSE-NEXT: # xmm2 = mem[3,1,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm12[0,1,3,1,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm10 = xmm10[0,1,3,1,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm10 = xmm10[0],xmm0[0],xmm10[1],xmm0[1]
; SSE-NEXT: movsd {{.*#+}} xmm10 = xmm2[0],xmm10[1]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm8[3,1,2,3,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm7[3,1,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm6[0,1,3,1,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[0,1,3,1,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
; SSE-NEXT: movsd {{.*#+}} xmm3 = xmm2[0],xmm3[1]
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 48(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, (%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 32(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 16(%rsi)
; SSE-NEXT: movapd %xmm15, 48(%rdx)
; SSE-NEXT: movapd %xmm13, (%rdx)
; SSE-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 32(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 16(%rdx)
; SSE-NEXT: movapd %xmm1, 48(%rcx)
; SSE-NEXT: movapd %xmm4, 32(%rcx)
; SSE-NEXT: movapd %xmm11, 16(%rcx)
; SSE-NEXT: movapd %xmm5, (%rcx)
; SSE-NEXT: movapd %xmm3, 48(%r8)
; SSE-NEXT: movapd %xmm10, 32(%r8)
; SSE-NEXT: movapd %xmm14, 16(%r8)
; SSE-NEXT: movapd %xmm9, (%r8)
; SSE-NEXT: addq $248, %rsp
; SSE-NEXT: retq
;
; AVX-LABEL: load_i16_stride4_vf32:
; AVX: # %bb.0:
; AVX-NEXT: subq $280, %rsp # imm = 0x118
; AVX-NEXT: vpxor %xmm6, %xmm6, %xmm6
; AVX-NEXT: vmovdqa 112(%rdi), %xmm4
; AVX-NEXT: vpblendw {{.*#+}} xmm2 = xmm4[0],xmm6[1,2,3],xmm4[4],xmm6[5,6,7]
; AVX-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vmovdqa 96(%rdi), %xmm11
; AVX-NEXT: vpblendw {{.*#+}} xmm3 = xmm11[0],xmm6[1,2,3],xmm11[4],xmm6[5,6,7]
; AVX-NEXT: vpackusdw %xmm2, %xmm3, %xmm2
; AVX-NEXT: vmovdqa 80(%rdi), %xmm3
; AVX-NEXT: vpblendw {{.*#+}} xmm5 = xmm3[0],xmm6[1,2,3],xmm3[4],xmm6[5,6,7]
; AVX-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vmovdqa 64(%rdi), %xmm0
; AVX-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vpblendw {{.*#+}} xmm7 = xmm0[0],xmm6[1,2,3],xmm0[4],xmm6[5,6,7]
; AVX-NEXT: vpackusdw %xmm5, %xmm7, %xmm5
; AVX-NEXT: vpackusdw %xmm2, %xmm5, %xmm2
; AVX-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vmovdqa (%rdi), %xmm10
; AVX-NEXT: vmovdqa 16(%rdi), %xmm0
; AVX-NEXT: vmovdqa 32(%rdi), %xmm1
; AVX-NEXT: vmovdqa 48(%rdi), %xmm2
; AVX-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vpblendw {{.*#+}} xmm7 = xmm2[0],xmm6[1,2,3],xmm2[4],xmm6[5,6,7]
; AVX-NEXT: vpblendw {{.*#+}} xmm8 = xmm1[0],xmm6[1,2,3],xmm1[4],xmm6[5,6,7]
; AVX-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vpackusdw %xmm7, %xmm8, %xmm7
; AVX-NEXT: vpblendw {{.*#+}} xmm8 = xmm0[0],xmm6[1,2,3],xmm0[4],xmm6[5,6,7]
; AVX-NEXT: vmovdqa %xmm0, %xmm2
; AVX-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vpblendw {{.*#+}} xmm9 = xmm10[0],xmm6[1,2,3],xmm10[4],xmm6[5,6,7]
; AVX-NEXT: vmovdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vpackusdw %xmm8, %xmm9, %xmm8
; AVX-NEXT: vpackusdw %xmm7, %xmm8, %xmm7
; AVX-NEXT: vmovdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vmovdqa 240(%rdi), %xmm0
; AVX-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vpblendw {{.*#+}} xmm7 = xmm0[0],xmm6[1,2,3],xmm0[4],xmm6[5,6,7]
; AVX-NEXT: vmovdqa 224(%rdi), %xmm12
; AVX-NEXT: vpblendw {{.*#+}} xmm8 = xmm12[0],xmm6[1,2,3],xmm12[4],xmm6[5,6,7]
; AVX-NEXT: vmovdqa %xmm12, (%rsp) # 16-byte Spill
; AVX-NEXT: vpackusdw %xmm7, %xmm8, %xmm7
; AVX-NEXT: vmovdqa 208(%rdi), %xmm5
; AVX-NEXT: vpblendw {{.*#+}} xmm8 = xmm5[0],xmm6[1,2,3],xmm5[4],xmm6[5,6,7]
; AVX-NEXT: vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vmovdqa 192(%rdi), %xmm0
; AVX-NEXT: vpblendw {{.*#+}} xmm9 = xmm0[0],xmm6[1,2,3],xmm0[4],xmm6[5,6,7]
; AVX-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vpackusdw %xmm8, %xmm9, %xmm8
; AVX-NEXT: vpackusdw %xmm7, %xmm8, %xmm7
; AVX-NEXT: vmovdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vmovdqa 176(%rdi), %xmm14
; AVX-NEXT: vpblendw {{.*#+}} xmm7 = xmm14[0],xmm6[1,2,3],xmm14[4],xmm6[5,6,7]
; AVX-NEXT: vmovdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vmovdqa 160(%rdi), %xmm13
; AVX-NEXT: vpblendw {{.*#+}} xmm15 = xmm13[0],xmm6[1,2,3],xmm13[4],xmm6[5,6,7]
; AVX-NEXT: vmovdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vpackusdw %xmm7, %xmm15, %xmm7
; AVX-NEXT: vmovdqa 144(%rdi), %xmm8
; AVX-NEXT: vpblendw {{.*#+}} xmm15 = xmm8[0],xmm6[1,2,3],xmm8[4],xmm6[5,6,7]
; AVX-NEXT: vmovdqa %xmm8, %xmm9
; AVX-NEXT: vmovdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vmovdqa 128(%rdi), %xmm8
; AVX-NEXT: vmovdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vpblendw {{.*#+}} xmm6 = xmm8[0],xmm6[1,2,3],xmm8[4],xmm6[5,6,7]
; AVX-NEXT: vpackusdw %xmm15, %xmm6, %xmm6
; AVX-NEXT: vpackusdw %xmm7, %xmm6, %xmm6
; AVX-NEXT: vmovdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vpshufd {{.*#+}} xmm6 = xmm4[0,2,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm6 = xmm6[0,1,1,3,4,5,6,7]
; AVX-NEXT: vpshufd {{.*#+}} xmm7 = xmm11[0,2,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm7 = xmm7[0,1,1,3,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm6 = xmm7[0],xmm6[0],xmm7[1],xmm6[1]
; AVX-NEXT: vpshufd {{.*#+}} xmm7 = xmm3[0,2,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm7 = xmm7[1,3,2,3,4,5,6,7]
; AVX-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; AVX-NEXT: vpshufd {{.*#+}} xmm15 = xmm3[0,2,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm15 = xmm15[1,3,2,3,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm7 = xmm15[0],xmm7[0],xmm15[1],xmm7[1]
; AVX-NEXT: vpblendw {{.*#+}} xmm6 = xmm7[0,1,2,3],xmm6[4,5,6,7]
; AVX-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
; AVX-NEXT: vpshufd {{.*#+}} xmm7 = xmm4[0,2,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm7 = xmm7[0,1,1,3,4,5,6,7]
; AVX-NEXT: vpshufd {{.*#+}} xmm15 = xmm1[0,2,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm15 = xmm15[0,1,1,3,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm7 = xmm15[0],xmm7[0],xmm15[1],xmm7[1]
; AVX-NEXT: vpshufd {{.*#+}} xmm15 = xmm2[0,2,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm15 = xmm15[1,3,2,3,4,5,6,7]
; AVX-NEXT: vpshufd {{.*#+}} xmm8 = xmm10[0,2,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm8 = xmm8[1,3,2,3,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm8 = xmm8[0],xmm15[0],xmm8[1],xmm15[1]
; AVX-NEXT: vpblendw {{.*#+}} xmm7 = xmm8[0,1,2,3],xmm7[4,5,6,7]
; AVX-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6
; AVX-NEXT: vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3],ymm6[4,5,6,7]
; AVX-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
; AVX-NEXT: vpshufd {{.*#+}} xmm6 = xmm10[0,2,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm6 = xmm6[0,1,1,3,4,5,6,7]
; AVX-NEXT: vpshufd {{.*#+}} xmm7 = xmm12[0,2,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm7 = xmm7[0,1,1,3,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm6 = xmm7[0],xmm6[0],xmm7[1],xmm6[1]
; AVX-NEXT: vpshufd {{.*#+}} xmm7 = xmm5[0,2,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm7 = xmm7[1,3,2,3,4,5,6,7]
; AVX-NEXT: vpshufd {{.*#+}} xmm8 = xmm0[0,2,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm8 = xmm8[1,3,2,3,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm7 = xmm8[0],xmm7[0],xmm8[1],xmm7[1]
; AVX-NEXT: vpblendw {{.*#+}} xmm6 = xmm7[0,1,2,3],xmm6[4,5,6,7]
; AVX-NEXT: vpshufd {{.*#+}} xmm7 = xmm14[0,2,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm7 = xmm7[0,1,1,3,4,5,6,7]
; AVX-NEXT: vpshufd {{.*#+}} xmm8 = xmm13[0,2,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm8 = xmm8[0,1,1,3,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm7 = xmm8[0],xmm7[0],xmm8[1],xmm7[1]
; AVX-NEXT: vpshufd {{.*#+}} xmm8 = xmm9[0,2,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm8 = xmm8[1,3,2,3,4,5,6,7]
; AVX-NEXT: vpshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
; AVX-NEXT: # xmm15 = mem[0,2,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm15 = xmm15[1,3,2,3,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm8 = xmm15[0],xmm8[0],xmm15[1],xmm8[1]
; AVX-NEXT: vpblendw {{.*#+}} xmm7 = xmm8[0,1,2,3],xmm7[4,5,6,7]
; AVX-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6
; AVX-NEXT: vblendps {{.*#+}} ymm0 = ymm7[0,1,2,3],ymm6[4,5,6,7]
; AVX-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; AVX-NEXT: # xmm0 = mem[3,1,2,3]
; AVX-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vpshufd {{.*#+}} xmm1 = xmm11[3,1,2,3]
; AVX-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[0,1,2,0,4,5,6,7]
; AVX-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[0,1,2,0,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; AVX-NEXT: # xmm1 = mem[3,1,2,3]
; AVX-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vpshufd {{.*#+}} xmm2 = xmm3[3,1,2,3]
; AVX-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[2,0,2,3,4,5,6,7]
; AVX-NEXT: vpshuflw {{.*#+}} xmm3 = xmm2[2,0,2,3,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; AVX-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3],xmm0[4,5,6,7]
; AVX-NEXT: vpshufd {{.*#+}} xmm7 = xmm4[3,1,2,3]
; AVX-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
; AVX-NEXT: # xmm6 = mem[3,1,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm1 = xmm7[0,1,2,0,4,5,6,7]
; AVX-NEXT: vpshuflw {{.*#+}} xmm2 = xmm6[0,1,2,0,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; AVX-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
; AVX-NEXT: # xmm12 = mem[3,1,2,3]
; AVX-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
; AVX-NEXT: # xmm5 = mem[3,1,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm8 = xmm12[2,0,2,3,4,5,6,7]
; AVX-NEXT: vpshuflw {{.*#+}} xmm9 = xmm5[2,0,2,3,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm8 = xmm9[0],xmm8[0],xmm9[1],xmm8[1]
; AVX-NEXT: vpblendw {{.*#+}} xmm1 = xmm8[0,1,2,3],xmm1[4,5,6,7]
; AVX-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
; AVX-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT: vpshufd {{.*#+}} xmm13 = xmm10[3,1,2,3]
; AVX-NEXT: vpshufd $231, (%rsp), %xmm14 # 16-byte Folded Reload
; AVX-NEXT: # xmm14 = mem[3,1,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm0 = xmm13[0,1,2,0,4,5,6,7]
; AVX-NEXT: vpshuflw {{.*#+}} xmm1 = xmm14[0,1,2,0,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
; AVX-NEXT: # xmm10 = mem[3,1,2,3]
; AVX-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
; AVX-NEXT: # xmm11 = mem[3,1,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm1 = xmm10[2,0,2,3,4,5,6,7]
; AVX-NEXT: vpshuflw {{.*#+}} xmm8 = xmm11[2,0,2,3,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm8[0],xmm1[0],xmm8[1],xmm1[1]
; AVX-NEXT: vpblendw {{.*#+}} xmm3 = xmm1[0,1,2,3],xmm0[4,5,6,7]
; AVX-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
; AVX-NEXT: # xmm8 = mem[3,1,2,3]
; AVX-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
; AVX-NEXT: # xmm9 = mem[3,1,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm1 = xmm8[0,1,2,0,4,5,6,7]
; AVX-NEXT: vpshuflw {{.*#+}} xmm2 = xmm9[0,1,2,0,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; AVX-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; AVX-NEXT: # xmm4 = mem[3,1,2,3]
; AVX-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; AVX-NEXT: # xmm2 = mem[3,1,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm0 = xmm4[2,0,2,3,4,5,6,7]
; AVX-NEXT: vpshuflw {{.*#+}} xmm15 = xmm2[2,0,2,3,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm15[0],xmm0[0],xmm15[1],xmm0[1]
; AVX-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm1[4,5,6,7]
; AVX-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm1
; AVX-NEXT: vblendps {{.*#+}} ymm15 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX-NEXT: vpshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; AVX-NEXT: # xmm0 = mem[0,1,3,1,4,5,6,7]
; AVX-NEXT: vpshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; AVX-NEXT: # xmm1 = mem[0,1,3,1,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX-NEXT: vpshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; AVX-NEXT: # xmm1 = mem[3,1,2,3,4,5,6,7]
; AVX-NEXT: vpshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; AVX-NEXT: # xmm3 = mem[3,1,2,3,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; AVX-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3],xmm0[4,5,6,7]
; AVX-NEXT: vpshuflw {{.*#+}} xmm1 = xmm7[0,1,3,1,4,5,6,7]
; AVX-NEXT: vpshuflw {{.*#+}} xmm3 = xmm6[0,1,3,1,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; AVX-NEXT: vpshuflw {{.*#+}} xmm3 = xmm12[3,1,2,3,4,5,6,7]
; AVX-NEXT: vpshuflw {{.*#+}} xmm5 = xmm5[3,1,2,3,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm5[0],xmm3[0],xmm5[1],xmm3[1]
; AVX-NEXT: vpblendw {{.*#+}} xmm1 = xmm3[0,1,2,3],xmm1[4,5,6,7]
; AVX-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
; AVX-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX-NEXT: vpshuflw {{.*#+}} xmm1 = xmm13[0,1,3,1,4,5,6,7]
; AVX-NEXT: vpshuflw {{.*#+}} xmm3 = xmm14[0,1,3,1,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; AVX-NEXT: vpshuflw {{.*#+}} xmm3 = xmm10[3,1,2,3,4,5,6,7]
; AVX-NEXT: vpshuflw {{.*#+}} xmm5 = xmm11[3,1,2,3,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm5[0],xmm3[0],xmm5[1],xmm3[1]
; AVX-NEXT: vpblendw {{.*#+}} xmm1 = xmm3[0,1,2,3],xmm1[4,5,6,7]
; AVX-NEXT: vpshuflw {{.*#+}} xmm3 = xmm8[0,1,3,1,4,5,6,7]
; AVX-NEXT: vpshuflw {{.*#+}} xmm5 = xmm9[0,1,3,1,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm5[0],xmm3[0],xmm5[1],xmm3[1]
; AVX-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[3,1,2,3,4,5,6,7]
; AVX-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[3,1,2,3,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1]
; AVX-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3],xmm3[4,5,6,7]
; AVX-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; AVX-NEXT: vmovaps %xmm2, 32(%rsi)
; AVX-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; AVX-NEXT: vmovaps %xmm2, 48(%rsi)
; AVX-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; AVX-NEXT: vmovaps %xmm2, (%rsi)
; AVX-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; AVX-NEXT: vmovaps %xmm2, 16(%rsi)
; AVX-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX-NEXT: vmovaps %ymm2, 32(%rdx)
; AVX-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX-NEXT: vmovaps %ymm2, (%rdx)
; AVX-NEXT: vmovaps %ymm15, 32(%rcx)
; AVX-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX-NEXT: vmovaps %ymm2, (%rcx)
; AVX-NEXT: vmovaps %ymm1, 32(%r8)
; AVX-NEXT: vmovaps %ymm0, (%r8)
; AVX-NEXT: addq $280, %rsp # imm = 0x118
; AVX-NEXT: vzeroupper
; AVX-NEXT: retq
;
; AVX2-LABEL: load_i16_stride4_vf32:
; AVX2: # %bb.0:
; AVX2-NEXT: subq $168, %rsp
; AVX2-NEXT: vpxor %xmm0, %xmm0, %xmm0
; AVX2-NEXT: vpblendw {{.*#+}} ymm1 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-NEXT: vextracti128 $1, %ymm1, %xmm2
; AVX2-NEXT: vpackusdw %xmm2, %xmm1, %xmm1
; AVX2-NEXT: vpblendw {{.*#+}} ymm2 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-NEXT: vextracti128 $1, %ymm2, %xmm3
; AVX2-NEXT: vpackusdw %xmm3, %xmm2, %xmm2
; AVX2-NEXT: vpackusdw %xmm1, %xmm2, %xmm1
; AVX2-NEXT: vpblendw {{.*#+}} ymm2 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-NEXT: vextracti128 $1, %ymm2, %xmm3
; AVX2-NEXT: vpackusdw %xmm3, %xmm2, %xmm2
; AVX2-NEXT: vpackusdw %xmm2, %xmm2, %xmm2
; AVX2-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX2-NEXT: vpblendw {{.*#+}} ymm3 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-NEXT: vextracti128 $1, %ymm3, %xmm4
; AVX2-NEXT: vpackusdw %xmm4, %xmm3, %xmm3
; AVX2-NEXT: vpackusdw %xmm3, %xmm3, %xmm3
; AVX2-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
; AVX2-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
; AVX2-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX2-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT: vpblendw {{.*#+}} ymm1 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-NEXT: vextracti128 $1, %ymm1, %xmm2
; AVX2-NEXT: vpackusdw %xmm2, %xmm1, %xmm1
; AVX2-NEXT: vpblendw {{.*#+}} ymm2 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-NEXT: vextracti128 $1, %ymm2, %xmm3
; AVX2-NEXT: vpackusdw %xmm3, %xmm2, %xmm2
; AVX2-NEXT: vpackusdw %xmm1, %xmm2, %xmm1
; AVX2-NEXT: vpblendw {{.*#+}} ymm2 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-NEXT: vextracti128 $1, %ymm2, %xmm3
; AVX2-NEXT: vpackusdw %xmm3, %xmm2, %xmm2
; AVX2-NEXT: vpackusdw %xmm2, %xmm2, %xmm2
; AVX2-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX2-NEXT: vpblendw {{.*#+}} ymm0 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm3
; AVX2-NEXT: vpackusdw %xmm3, %xmm0, %xmm0
; AVX2-NEXT: vpackusdw %xmm0, %xmm0, %xmm0
; AVX2-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT: vmovdqa (%rdi), %xmm1
; AVX2-NEXT: vmovdqa %xmm1, (%rsp) # 16-byte Spill
; AVX2-NEXT: vmovdqa 16(%rdi), %xmm5
; AVX2-NEXT: vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-NEXT: vmovdqa 32(%rdi), %xmm6
; AVX2-NEXT: vmovdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-NEXT: vmovdqa 48(%rdi), %xmm7
; AVX2-NEXT: vmovdqa 112(%rdi), %xmm9
; AVX2-NEXT: vpshufd {{.*#+}} xmm0 = xmm9[0,2,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[0,1,1,3,4,5,6,7]
; AVX2-NEXT: vmovdqa 96(%rdi), %xmm13
; AVX2-NEXT: vpshufd {{.*#+}} xmm2 = xmm13[0,2,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[0,1,1,3,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; AVX2-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-NEXT: vmovdqa 80(%rdi), %xmm14
; AVX2-NEXT: vpshufd {{.*#+}} xmm2 = xmm14[0,2,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm3 = xmm2[1,3,2,3,4,5,6,7]
; AVX2-NEXT: vmovdqa 64(%rdi), %xmm2
; AVX2-NEXT: vpshufd {{.*#+}} xmm4 = xmm2[0,2,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[1,3,2,3,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
; AVX2-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
; AVX2-NEXT: vpshufd {{.*#+}} xmm3 = xmm7[0,2,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[0,1,1,3,4,5,6,7]
; AVX2-NEXT: vpshufd {{.*#+}} xmm4 = xmm6[0,2,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[0,1,1,3,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
; AVX2-NEXT: vpshufd {{.*#+}} xmm4 = xmm5[0,2,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[1,3,2,3,4,5,6,7]
; AVX2-NEXT: vpshufd {{.*#+}} xmm8 = xmm1[0,2,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm8 = xmm8[1,3,2,3,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm4 = xmm8[0],xmm4[0],xmm8[1],xmm4[1]
; AVX2-NEXT: vpblendd {{.*#+}} xmm3 = xmm4[0,1],xmm3[2,3]
; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7]
; AVX2-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT: vmovdqa 240(%rdi), %xmm0
; AVX2-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[0,1,1,3,4,5,6,7]
; AVX2-NEXT: vmovdqa 224(%rdi), %xmm10
; AVX2-NEXT: vpshufd {{.*#+}} xmm3 = xmm10[0,2,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[0,1,1,3,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
; AVX2-NEXT: vmovdqa 208(%rdi), %xmm11
; AVX2-NEXT: vpshufd {{.*#+}} xmm3 = xmm11[0,2,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[1,3,2,3,4,5,6,7]
; AVX2-NEXT: vmovdqa 192(%rdi), %xmm15
; AVX2-NEXT: vpshufd {{.*#+}} xmm4 = xmm15[0,2,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[1,3,2,3,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
; AVX2-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
; AVX2-NEXT: vmovdqa 176(%rdi), %xmm12
; AVX2-NEXT: vpshufd {{.*#+}} xmm3 = xmm12[0,2,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[0,1,1,3,4,5,6,7]
; AVX2-NEXT: vmovdqa 160(%rdi), %xmm4
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm4[0,2,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[0,1,1,3,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
; AVX2-NEXT: vmovdqa 144(%rdi), %xmm3
; AVX2-NEXT: vpshufd {{.*#+}} xmm5 = xmm3[0,2,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm5 = xmm5[1,3,2,3,4,5,6,7]
; AVX2-NEXT: vmovdqa 128(%rdi), %xmm6
; AVX2-NEXT: vmovdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-NEXT: vpshufd {{.*#+}} xmm6 = xmm6[0,2,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm6 = xmm6[1,3,2,3,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm5 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
; AVX2-NEXT: vpblendd {{.*#+}} xmm1 = xmm5[0,1],xmm1[2,3]
; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT: vpshufd {{.*#+}} xmm0 = xmm9[3,1,2,3]
; AVX2-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm13[3,1,2,3]
; AVX2-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[0,1,2,0,4,5,6,7]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[0,1,2,0,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm14[3,1,2,3]
; AVX2-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[3,1,2,3]
; AVX2-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[2,0,2,3,4,5,6,7]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm5 = xmm2[2,0,2,3,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm5[0],xmm1[0],xmm5[1],xmm1[1]
; AVX2-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX2-NEXT: vpblendd {{.*#+}} ymm2 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-NEXT: vpshufd {{.*#+}} xmm13 = xmm7[3,1,2,3]
; AVX2-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
; AVX2-NEXT: # xmm9 = mem[3,1,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm1 = xmm13[0,1,2,0,4,5,6,7]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm5 = xmm9[0,1,2,0,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm5[0],xmm1[0],xmm5[1],xmm1[1]
; AVX2-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
; AVX2-NEXT: # xmm7 = mem[3,1,2,3]
; AVX2-NEXT: vpshufd $231, (%rsp), %xmm6 # 16-byte Folded Reload
; AVX2-NEXT: # xmm6 = mem[3,1,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm0 = xmm7[2,0,2,3,4,5,6,7]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm8 = xmm6[2,0,2,3,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm8[0],xmm0[0],xmm8[1],xmm0[1]
; AVX2-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
; AVX2-NEXT: vmovdqu %ymm0, (%rsp) # 32-byte Spill
; AVX2-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
; AVX2-NEXT: # xmm8 = mem[3,1,2,3]
; AVX2-NEXT: vpshufd {{.*#+}} xmm10 = xmm10[3,1,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm0 = xmm8[0,1,2,0,4,5,6,7]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm1 = xmm10[0,1,2,0,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX2-NEXT: vpshufd {{.*#+}} xmm11 = xmm11[3,1,2,3]
; AVX2-NEXT: vpshufd {{.*#+}} xmm15 = xmm15[3,1,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm1 = xmm11[2,0,2,3,4,5,6,7]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm2 = xmm15[2,0,2,3,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; AVX2-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX2-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-NEXT: vpshufd {{.*#+}} xmm12 = xmm12[3,1,2,3]
; AVX2-NEXT: vpshufd {{.*#+}} xmm5 = xmm4[3,1,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm0 = xmm12[0,1,2,0,4,5,6,7]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm2 = xmm5[0,1,2,0,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; AVX2-NEXT: vpshufd {{.*#+}} xmm4 = xmm3[3,1,2,3]
; AVX2-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; AVX2-NEXT: # xmm3 = mem[3,1,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm0 = xmm4[2,0,2,3,4,5,6,7]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm14 = xmm3[2,0,2,3,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm14[0],xmm0[0],xmm14[1],xmm0[1]
; AVX2-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm2[2,3]
; AVX2-NEXT: vpblendd {{.*#+}} ymm14 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-NEXT: vpshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; AVX2-NEXT: # xmm0 = mem[0,1,3,1,4,5,6,7]
; AVX2-NEXT: vpshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; AVX2-NEXT: # xmm1 = mem[0,1,3,1,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX2-NEXT: vpshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; AVX2-NEXT: # xmm1 = mem[3,1,2,3,4,5,6,7]
; AVX2-NEXT: vpshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; AVX2-NEXT: # xmm2 = mem[3,1,2,3,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; AVX2-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm1 = xmm13[0,1,3,1,4,5,6,7]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm2 = xmm9[0,1,3,1,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm2 = xmm7[3,1,2,3,4,5,6,7]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm6 = xmm6[3,1,2,3,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm6[0],xmm2[0],xmm6[1],xmm2[1]
; AVX2-NEXT: vpblendd {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm1 = xmm8[0,1,3,1,4,5,6,7]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm2 = xmm10[0,1,3,1,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm2 = xmm11[3,1,2,3,4,5,6,7]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm6 = xmm15[3,1,2,3,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm6[0],xmm2[0],xmm6[1],xmm2[1]
; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX2-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX2-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm2 = xmm12[0,1,3,1,4,5,6,7]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm5 = xmm5[0,1,3,1,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm5[0],xmm2[0],xmm5[1],xmm2[1]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[3,1,2,3,4,5,6,7]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[3,1,2,3,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
; AVX2-NEXT: vpblendd {{.*#+}} xmm2 = xmm3[0,1],xmm2[2,3]
; AVX2-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX2-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-NEXT: vmovaps %ymm2, 32(%rsi)
; AVX2-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-NEXT: vmovaps %ymm2, (%rsi)
; AVX2-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-NEXT: vmovaps %ymm2, 32(%rdx)
; AVX2-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-NEXT: vmovaps %ymm2, (%rdx)
; AVX2-NEXT: vmovdqa %ymm14, 32(%rcx)
; AVX2-NEXT: vmovups (%rsp), %ymm2 # 32-byte Reload
; AVX2-NEXT: vmovaps %ymm2, (%rcx)
; AVX2-NEXT: vmovdqa %ymm1, 32(%r8)
; AVX2-NEXT: vmovdqa %ymm0, (%r8)
; AVX2-NEXT: addq $168, %rsp
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
; AVX2-FP-LABEL: load_i16_stride4_vf32:
; AVX2-FP: # %bb.0:
; AVX2-FP-NEXT: subq $184, %rsp
; AVX2-FP-NEXT: vpxor %xmm0, %xmm0, %xmm0
; AVX2-FP-NEXT: vpblendw {{.*#+}} ymm1 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-FP-NEXT: vextracti128 $1, %ymm1, %xmm2
; AVX2-FP-NEXT: vpackusdw %xmm2, %xmm1, %xmm1
; AVX2-FP-NEXT: vpblendw {{.*#+}} ymm2 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-FP-NEXT: vextracti128 $1, %ymm2, %xmm3
; AVX2-FP-NEXT: vpackusdw %xmm3, %xmm2, %xmm2
; AVX2-FP-NEXT: vpackusdw %xmm1, %xmm2, %xmm1
; AVX2-FP-NEXT: vpblendw {{.*#+}} ymm2 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-FP-NEXT: vextracti128 $1, %ymm2, %xmm3
; AVX2-FP-NEXT: vpackusdw %xmm3, %xmm2, %xmm2
; AVX2-FP-NEXT: vpackusdw %xmm2, %xmm2, %xmm2
; AVX2-FP-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX2-FP-NEXT: vpblendw {{.*#+}} ymm3 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-FP-NEXT: vextracti128 $1, %ymm3, %xmm4
; AVX2-FP-NEXT: vpackusdw %xmm4, %xmm3, %xmm3
; AVX2-FP-NEXT: vpackusdw %xmm3, %xmm3, %xmm3
; AVX2-FP-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
; AVX2-FP-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
; AVX2-FP-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX2-FP-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT: vpblendw {{.*#+}} ymm1 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-FP-NEXT: vextracti128 $1, %ymm1, %xmm2
; AVX2-FP-NEXT: vpackusdw %xmm2, %xmm1, %xmm1
; AVX2-FP-NEXT: vpblendw {{.*#+}} ymm2 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-FP-NEXT: vextracti128 $1, %ymm2, %xmm3
; AVX2-FP-NEXT: vpackusdw %xmm3, %xmm2, %xmm2
; AVX2-FP-NEXT: vpackusdw %xmm1, %xmm2, %xmm1
; AVX2-FP-NEXT: vpblendw {{.*#+}} ymm2 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-FP-NEXT: vextracti128 $1, %ymm2, %xmm3
; AVX2-FP-NEXT: vpackusdw %xmm3, %xmm2, %xmm2
; AVX2-FP-NEXT: vpackusdw %xmm2, %xmm2, %xmm2
; AVX2-FP-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX2-FP-NEXT: vpblendw {{.*#+}} ymm0 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-FP-NEXT: vextracti128 $1, %ymm0, %xmm3
; AVX2-FP-NEXT: vpackusdw %xmm3, %xmm0, %xmm0
; AVX2-FP-NEXT: vpackusdw %xmm0, %xmm0, %xmm0
; AVX2-FP-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-FP-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
; AVX2-FP-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FP-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT: vmovdqa (%rdi), %xmm13
; AVX2-FP-NEXT: vmovdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FP-NEXT: vmovdqa 16(%rdi), %xmm12
; AVX2-FP-NEXT: vmovdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FP-NEXT: vmovdqa 32(%rdi), %xmm14
; AVX2-FP-NEXT: vmovdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FP-NEXT: vmovdqa 48(%rdi), %xmm7
; AVX2-FP-NEXT: vmovdqa {{.*#+}} xmm10 = [0,1,2,3,2,3,10,11,8,9,10,11,12,13,14,15]
; AVX2-FP-NEXT: vmovdqa 112(%rdi), %xmm2
; AVX2-FP-NEXT: vpshufb %xmm10, %xmm2, %xmm0
; AVX2-FP-NEXT: vmovdqa 96(%rdi), %xmm4
; AVX2-FP-NEXT: vpshufb %xmm10, %xmm4, %xmm1
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX2-FP-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm3
; AVX2-FP-NEXT: vmovdqa {{.*#+}} xmm11 = [2,3,10,11,8,9,10,11,8,9,10,11,12,13,14,15]
; AVX2-FP-NEXT: vmovdqa 80(%rdi), %xmm5
; AVX2-FP-NEXT: vpshufb %xmm11, %xmm5, %xmm8
; AVX2-FP-NEXT: vmovdqa 64(%rdi), %xmm6
; AVX2-FP-NEXT: vpshufb %xmm11, %xmm6, %xmm9
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm8 = xmm9[0],xmm8[0],xmm9[1],xmm8[1]
; AVX2-FP-NEXT: vinserti128 $1, %xmm8, %ymm0, %ymm8
; AVX2-FP-NEXT: vpblendd {{.*#+}} ymm3 = ymm8[0,1,2,3,4,5],ymm3[6,7]
; AVX2-FP-NEXT: vpshufb %xmm10, %xmm7, %xmm8
; AVX2-FP-NEXT: vpshufb %xmm10, %xmm14, %xmm9
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm8 = xmm9[0],xmm8[0],xmm9[1],xmm8[1]
; AVX2-FP-NEXT: vpshufb %xmm11, %xmm12, %xmm9
; AVX2-FP-NEXT: vpshufb %xmm11, %xmm13, %xmm12
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm9 = xmm12[0],xmm9[0],xmm12[1],xmm9[1]
; AVX2-FP-NEXT: vpblendd {{.*#+}} xmm8 = xmm9[0,1],xmm8[2,3]
; AVX2-FP-NEXT: vpblendd {{.*#+}} ymm0 = ymm8[0,1,2,3],ymm3[4,5,6,7]
; AVX2-FP-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT: vmovdqa 240(%rdi), %xmm14
; AVX2-FP-NEXT: vpshufb %xmm10, %xmm14, %xmm9
; AVX2-FP-NEXT: vmovdqa 224(%rdi), %xmm8
; AVX2-FP-NEXT: vpshufb %xmm10, %xmm8, %xmm12
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm9 = xmm12[0],xmm9[0],xmm12[1],xmm9[1]
; AVX2-FP-NEXT: vmovdqa 208(%rdi), %xmm0
; AVX2-FP-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FP-NEXT: vpshufb %xmm11, %xmm0, %xmm12
; AVX2-FP-NEXT: vmovdqa 192(%rdi), %xmm15
; AVX2-FP-NEXT: vpshufb %xmm11, %xmm15, %xmm13
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm12 = xmm13[0],xmm12[0],xmm13[1],xmm12[1]
; AVX2-FP-NEXT: vinserti128 $1, %xmm9, %ymm0, %ymm9
; AVX2-FP-NEXT: vinserti128 $1, %xmm12, %ymm0, %ymm12
; AVX2-FP-NEXT: vpblendd {{.*#+}} ymm1 = ymm12[0,1,2,3,4,5],ymm9[6,7]
; AVX2-FP-NEXT: vmovdqa 160(%rdi), %xmm0
; AVX2-FP-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FP-NEXT: vmovdqa 176(%rdi), %xmm13
; AVX2-FP-NEXT: vpshufb %xmm10, %xmm13, %xmm12
; AVX2-FP-NEXT: vpshufb %xmm10, %xmm0, %xmm10
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm10 = xmm10[0],xmm12[0],xmm10[1],xmm12[1]
; AVX2-FP-NEXT: vmovdqa 128(%rdi), %xmm3
; AVX2-FP-NEXT: vmovdqa %xmm3, (%rsp) # 16-byte Spill
; AVX2-FP-NEXT: vmovdqa 144(%rdi), %xmm12
; AVX2-FP-NEXT: vpshufb %xmm11, %xmm12, %xmm0
; AVX2-FP-NEXT: vpshufb %xmm11, %xmm3, %xmm11
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm11[0],xmm0[0],xmm11[1],xmm0[1]
; AVX2-FP-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm10[2,3]
; AVX2-FP-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FP-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT: vpshufd {{.*#+}} xmm0 = xmm2[3,1,2,3]
; AVX2-FP-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FP-NEXT: vpshufd {{.*#+}} xmm1 = xmm4[3,1,2,3]
; AVX2-FP-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[0,1,2,0,4,5,6,7]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[0,1,2,0,4,5,6,7]
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX2-FP-NEXT: vpshufd {{.*#+}} xmm1 = xmm5[3,1,2,3]
; AVX2-FP-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FP-NEXT: vpshufd {{.*#+}} xmm2 = xmm6[3,1,2,3]
; AVX2-FP-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[2,0,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[2,0,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; AVX2-FP-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-FP-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX2-FP-NEXT: vpblendd {{.*#+}} ymm3 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-FP-NEXT: vpshufd {{.*#+}} xmm7 = xmm7[3,1,2,3]
; AVX2-FP-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
; AVX2-FP-NEXT: # xmm6 = mem[3,1,2,3]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm1 = xmm7[0,1,2,0,4,5,6,7]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm2 = xmm6[0,1,2,0,4,5,6,7]
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; AVX2-FP-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
; AVX2-FP-NEXT: # xmm5 = mem[3,1,2,3]
; AVX2-FP-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; AVX2-FP-NEXT: # xmm4 = mem[3,1,2,3]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm0 = xmm5[2,0,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm9 = xmm4[2,0,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm9[0],xmm0[0],xmm9[1],xmm0[1]
; AVX2-FP-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
; AVX2-FP-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
; AVX2-FP-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT: vpshufd {{.*#+}} xmm2 = xmm14[3,1,2,3]
; AVX2-FP-NEXT: vpshufd {{.*#+}} xmm3 = xmm8[3,1,2,3]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm0 = xmm2[0,1,2,0,4,5,6,7]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm8 = xmm3[0,1,2,0,4,5,6,7]
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm8[0],xmm0[0],xmm8[1],xmm0[1]
; AVX2-FP-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
; AVX2-FP-NEXT: # xmm8 = mem[3,1,2,3]
; AVX2-FP-NEXT: vpshufd {{.*#+}} xmm14 = xmm15[3,1,2,3]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm9 = xmm8[2,0,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm15 = xmm14[2,0,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm9 = xmm15[0],xmm9[0],xmm15[1],xmm9[1]
; AVX2-FP-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-FP-NEXT: vinserti128 $1, %xmm9, %ymm0, %ymm9
; AVX2-FP-NEXT: vpblendd {{.*#+}} ymm1 = ymm9[0,1,2,3,4,5],ymm0[6,7]
; AVX2-FP-NEXT: vpshufd {{.*#+}} xmm13 = xmm13[3,1,2,3]
; AVX2-FP-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
; AVX2-FP-NEXT: # xmm9 = mem[3,1,2,3]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm15 = xmm13[0,1,2,0,4,5,6,7]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm11 = xmm9[0,1,2,0,4,5,6,7]
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm11 = xmm11[0],xmm15[0],xmm11[1],xmm15[1]
; AVX2-FP-NEXT: vpshufd {{.*#+}} xmm15 = xmm12[3,1,2,3]
; AVX2-FP-NEXT: vpshufd $231, (%rsp), %xmm12 # 16-byte Folded Reload
; AVX2-FP-NEXT: # xmm12 = mem[3,1,2,3]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm0 = xmm15[2,0,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm10 = xmm12[2,0,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm10[0],xmm0[0],xmm10[1],xmm0[1]
; AVX2-FP-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm11[2,3]
; AVX2-FP-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FP-NEXT: vpshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; AVX2-FP-NEXT: # xmm1 = mem[0,1,3,1,4,5,6,7]
; AVX2-FP-NEXT: vpshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
; AVX2-FP-NEXT: # xmm10 = mem[0,1,3,1,4,5,6,7]
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm10[0],xmm1[0],xmm10[1],xmm1[1]
; AVX2-FP-NEXT: vpshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
; AVX2-FP-NEXT: # xmm10 = mem[3,1,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
; AVX2-FP-NEXT: # xmm11 = mem[3,1,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm10 = xmm11[0],xmm10[0],xmm11[1],xmm10[1]
; AVX2-FP-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX2-FP-NEXT: vinserti128 $1, %xmm10, %ymm0, %ymm10
; AVX2-FP-NEXT: vpblendd {{.*#+}} ymm1 = ymm10[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm7 = xmm7[0,1,3,1,4,5,6,7]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm6 = xmm6[0,1,3,1,4,5,6,7]
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm6 = xmm6[0],xmm7[0],xmm6[1],xmm7[1]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm5 = xmm5[3,1,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[3,1,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm4 = xmm4[0],xmm5[0],xmm4[1],xmm5[1]
; AVX2-FP-NEXT: vpblendd {{.*#+}} xmm4 = xmm4[0,1],xmm6[2,3]
; AVX2-FP-NEXT: vpblendd {{.*#+}} ymm1 = ymm4[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[0,1,3,1,4,5,6,7]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[0,1,3,1,4,5,6,7]
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm3 = xmm8[3,1,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm4 = xmm14[3,1,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
; AVX2-FP-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX2-FP-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
; AVX2-FP-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm3 = xmm13[0,1,3,1,4,5,6,7]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm4 = xmm9[0,1,3,1,4,5,6,7]
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm4 = xmm15[3,1,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm5 = xmm12[3,1,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm4 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
; AVX2-FP-NEXT: vpblendd {{.*#+}} xmm3 = xmm4[0,1],xmm3[2,3]
; AVX2-FP-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
; AVX2-FP-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-FP-NEXT: vmovaps %ymm3, 32(%rsi)
; AVX2-FP-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-FP-NEXT: vmovaps %ymm3, (%rsi)
; AVX2-FP-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-FP-NEXT: vmovaps %ymm3, 32(%rdx)
; AVX2-FP-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-FP-NEXT: vmovaps %ymm3, (%rdx)
; AVX2-FP-NEXT: vmovdqa %ymm0, 32(%rcx)
; AVX2-FP-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FP-NEXT: vmovaps %ymm0, (%rcx)
; AVX2-FP-NEXT: vmovdqa %ymm2, 32(%r8)
; AVX2-FP-NEXT: vmovdqa %ymm1, (%r8)
; AVX2-FP-NEXT: addq $184, %rsp
; AVX2-FP-NEXT: vzeroupper
; AVX2-FP-NEXT: retq
;
; AVX2-FCP-LABEL: load_i16_stride4_vf32:
; AVX2-FCP: # %bb.0:
; AVX2-FCP-NEXT: subq $104, %rsp
; AVX2-FCP-NEXT: vmovdqa 192(%rdi), %ymm3
; AVX2-FCP-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT: vmovdqa 224(%rdi), %ymm6
; AVX2-FCP-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT: vmovdqa 64(%rdi), %ymm9
; AVX2-FCP-NEXT: vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT: vmovdqa 96(%rdi), %ymm7
; AVX2-FCP-NEXT: vpxor %xmm0, %xmm0, %xmm0
; AVX2-FCP-NEXT: vpblendw {{.*#+}} ymm1 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-FCP-NEXT: vextracti128 $1, %ymm1, %xmm2
; AVX2-FCP-NEXT: vpackusdw %xmm2, %xmm1, %xmm1
; AVX2-FCP-NEXT: vpblendw {{.*#+}} ymm2 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-FCP-NEXT: vextracti128 $1, %ymm2, %xmm4
; AVX2-FCP-NEXT: vpackusdw %xmm4, %xmm2, %xmm2
; AVX2-FCP-NEXT: vpackusdw %xmm1, %xmm2, %xmm1
; AVX2-FCP-NEXT: vpmovsxbd {{.*#+}} ymm2 = [0,2,2,3,0,2,4,6]
; AVX2-FCP-NEXT: vpermd %ymm7, %ymm2, %ymm5
; AVX2-FCP-NEXT: vmovdqa {{.*#+}} ymm4 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,16,17,20,21,24,25,28,29]
; AVX2-FCP-NEXT: vpshufb %ymm4, %ymm5, %ymm8
; AVX2-FCP-NEXT: vpermd %ymm9, %ymm2, %ymm10
; AVX2-FCP-NEXT: vpshufb %ymm4, %ymm10, %ymm9
; AVX2-FCP-NEXT: vpblendd {{.*#+}} ymm8 = ymm9[0,1,2,3,4,5],ymm8[6,7]
; AVX2-FCP-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm8[4,5,6,7]
; AVX2-FCP-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT: vpblendw {{.*#+}} ymm1 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-FCP-NEXT: vextracti128 $1, %ymm1, %xmm8
; AVX2-FCP-NEXT: vpackusdw %xmm8, %xmm1, %xmm1
; AVX2-FCP-NEXT: vpblendw {{.*#+}} ymm0 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-FCP-NEXT: vextracti128 $1, %ymm0, %xmm8
; AVX2-FCP-NEXT: vpackusdw %xmm8, %xmm0, %xmm0
; AVX2-FCP-NEXT: vpackusdw %xmm1, %xmm0, %xmm0
; AVX2-FCP-NEXT: vpermd %ymm6, %ymm2, %ymm9
; AVX2-FCP-NEXT: vpshufb %ymm4, %ymm9, %ymm1
; AVX2-FCP-NEXT: vpermd %ymm3, %ymm2, %ymm6
; AVX2-FCP-NEXT: vpshufb %ymm4, %ymm6, %ymm8
; AVX2-FCP-NEXT: vpblendd {{.*#+}} ymm1 = ymm8[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FCP-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FCP-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT: vmovdqa 32(%rdi), %xmm14
; AVX2-FCP-NEXT: vmovdqa 48(%rdi), %xmm3
; AVX2-FCP-NEXT: vmovdqa {{.*#+}} xmm12 = [0,1,2,3,2,3,10,11,8,9,10,11,12,13,14,15]
; AVX2-FCP-NEXT: vpshufb %xmm12, %xmm3, %xmm0
; AVX2-FCP-NEXT: vpshufb %xmm12, %xmm14, %xmm8
; AVX2-FCP-NEXT: vpunpckldq {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1]
; AVX2-FCP-NEXT: vmovdqa (%rdi), %xmm13
; AVX2-FCP-NEXT: vmovdqa 16(%rdi), %xmm2
; AVX2-FCP-NEXT: vmovdqa {{.*#+}} xmm1 = [2,3,10,11,8,9,10,11,8,9,10,11,12,13,14,15]
; AVX2-FCP-NEXT: vpshufb %xmm1, %xmm2, %xmm4
; AVX2-FCP-NEXT: vpshufb %xmm1, %xmm13, %xmm11
; AVX2-FCP-NEXT: vpunpckldq {{.*#+}} xmm4 = xmm11[0],xmm4[0],xmm11[1],xmm4[1]
; AVX2-FCP-NEXT: vpblendd {{.*#+}} xmm4 = xmm4[0,1],xmm8[2,3]
; AVX2-FCP-NEXT: vmovdqa {{.*#+}} ymm0 = [2,3,6,7,10,11,14,15,14,15,10,11,12,13,14,15,18,19,22,23,26,27,30,31,18,19,22,23,26,27,30,31]
; AVX2-FCP-NEXT: vpshufb %ymm0, %ymm5, %ymm5
; AVX2-FCP-NEXT: vpshufb %ymm0, %ymm10, %ymm10
; AVX2-FCP-NEXT: vmovdqa %ymm0, %ymm8
; AVX2-FCP-NEXT: vpblendd {{.*#+}} ymm5 = ymm10[0,1,2,3,4,5],ymm5[6,7]
; AVX2-FCP-NEXT: vpblendd {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm5[4,5,6,7]
; AVX2-FCP-NEXT: vmovdqu %ymm0, (%rsp) # 32-byte Spill
; AVX2-FCP-NEXT: vmovdqa 160(%rdi), %xmm11
; AVX2-FCP-NEXT: vmovdqa 176(%rdi), %xmm15
; AVX2-FCP-NEXT: vpshufb %xmm12, %xmm15, %xmm4
; AVX2-FCP-NEXT: vpshufb %xmm12, %xmm11, %xmm10
; AVX2-FCP-NEXT: vpunpckldq {{.*#+}} xmm4 = xmm10[0],xmm4[0],xmm10[1],xmm4[1]
; AVX2-FCP-NEXT: vmovdqa 128(%rdi), %xmm10
; AVX2-FCP-NEXT: vmovdqa 144(%rdi), %xmm12
; AVX2-FCP-NEXT: vpshufb %xmm1, %xmm12, %xmm0
; AVX2-FCP-NEXT: vpshufb %xmm1, %xmm10, %xmm1
; AVX2-FCP-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX2-FCP-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm4[2,3]
; AVX2-FCP-NEXT: vpshufb %ymm8, %ymm9, %ymm1
; AVX2-FCP-NEXT: vpshufb %ymm8, %ymm6, %ymm4
; AVX2-FCP-NEXT: vpblendd {{.*#+}} ymm1 = ymm4[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FCP-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FCP-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT: vpmovsxbd {{.*#+}} ymm6 = [1,3,2,3,1,3,5,7]
; AVX2-FCP-NEXT: vpermd %ymm7, %ymm6, %ymm4
; AVX2-FCP-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm7 # 32-byte Folded Reload
; AVX2-FCP-NEXT: vmovdqa {{.*#+}} ymm5 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,16,17,20,21,24,25,28,29]
; AVX2-FCP-NEXT: vpshufb %ymm5, %ymm4, %ymm0
; AVX2-FCP-NEXT: vpshufb %ymm5, %ymm7, %ymm1
; AVX2-FCP-NEXT: vmovdqa %ymm5, %ymm8
; AVX2-FCP-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-FCP-NEXT: vpshufd {{.*#+}} xmm1 = xmm3[3,1,2,3]
; AVX2-FCP-NEXT: vpshufd {{.*#+}} xmm14 = xmm14[3,1,2,3]
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm3 = xmm1[0,1,2,0,4,5,6,7]
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm5 = xmm14[0,1,2,0,4,5,6,7]
; AVX2-FCP-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm5[0],xmm3[0],xmm5[1],xmm3[1]
; AVX2-FCP-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[3,1,2,3]
; AVX2-FCP-NEXT: vpshufd {{.*#+}} xmm13 = xmm13[3,1,2,3]
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm5 = xmm2[2,0,2,3,4,5,6,7]
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm9 = xmm13[2,0,2,3,4,5,6,7]
; AVX2-FCP-NEXT: vpunpckldq {{.*#+}} xmm5 = xmm9[0],xmm5[0],xmm9[1],xmm5[1]
; AVX2-FCP-NEXT: vpblendd {{.*#+}} xmm3 = xmm5[0,1],xmm3[2,3]
; AVX2-FCP-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FCP-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm3 # 32-byte Folded Reload
; AVX2-FCP-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm5 # 32-byte Folded Reload
; AVX2-FCP-NEXT: vpshufb %ymm8, %ymm3, %ymm6
; AVX2-FCP-NEXT: vpshufb %ymm8, %ymm5, %ymm9
; AVX2-FCP-NEXT: vpblendd {{.*#+}} ymm6 = ymm9[0,1,2,3,4,5],ymm6[6,7]
; AVX2-FCP-NEXT: vpshufd {{.*#+}} xmm9 = xmm15[3,1,2,3]
; AVX2-FCP-NEXT: vpshufd {{.*#+}} xmm11 = xmm11[3,1,2,3]
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm15 = xmm9[0,1,2,0,4,5,6,7]
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm0 = xmm11[0,1,2,0,4,5,6,7]
; AVX2-FCP-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm15[0],xmm0[1],xmm15[1]
; AVX2-FCP-NEXT: vpshufd {{.*#+}} xmm12 = xmm12[3,1,2,3]
; AVX2-FCP-NEXT: vpshufd {{.*#+}} xmm10 = xmm10[3,1,2,3]
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm15 = xmm12[2,0,2,3,4,5,6,7]
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm8 = xmm10[2,0,2,3,4,5,6,7]
; AVX2-FCP-NEXT: vpunpckldq {{.*#+}} xmm8 = xmm8[0],xmm15[0],xmm8[1],xmm15[1]
; AVX2-FCP-NEXT: vpblendd {{.*#+}} xmm0 = xmm8[0,1],xmm0[2,3]
; AVX2-FCP-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm6[4,5,6,7]
; AVX2-FCP-NEXT: vmovdqa {{.*#+}} ymm8 = [2,3,6,7,10,11,14,15,14,15,10,11,12,13,14,15,18,19,22,23,26,27,30,31,18,19,22,23,26,27,30,31]
; AVX2-FCP-NEXT: vpshufb %ymm8, %ymm4, %ymm4
; AVX2-FCP-NEXT: vpshufb %ymm8, %ymm7, %ymm6
; AVX2-FCP-NEXT: vpblendd {{.*#+}} ymm4 = ymm6[0,1,2,3,4,5],ymm4[6,7]
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[0,1,3,1,4,5,6,7]
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm6 = xmm14[0,1,3,1,4,5,6,7]
; AVX2-FCP-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm6[0],xmm1[0],xmm6[1],xmm1[1]
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[3,1,2,3,4,5,6,7]
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm6 = xmm13[3,1,2,3,4,5,6,7]
; AVX2-FCP-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm6[0],xmm2[0],xmm6[1],xmm2[1]
; AVX2-FCP-NEXT: vpblendd {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
; AVX2-FCP-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm4[4,5,6,7]
; AVX2-FCP-NEXT: vpshufb %ymm8, %ymm3, %ymm2
; AVX2-FCP-NEXT: vpshufb %ymm8, %ymm5, %ymm3
; AVX2-FCP-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm3 = xmm9[0,1,3,1,4,5,6,7]
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm4 = xmm11[0,1,3,1,4,5,6,7]
; AVX2-FCP-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm4 = xmm12[3,1,2,3,4,5,6,7]
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm5 = xmm10[3,1,2,3,4,5,6,7]
; AVX2-FCP-NEXT: vpunpckldq {{.*#+}} xmm4 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
; AVX2-FCP-NEXT: vpblendd {{.*#+}} xmm3 = xmm4[0,1],xmm3[2,3]
; AVX2-FCP-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
; AVX2-FCP-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-FCP-NEXT: vmovaps %ymm3, 32(%rsi)
; AVX2-FCP-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-FCP-NEXT: vmovaps %ymm3, (%rsi)
; AVX2-FCP-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-FCP-NEXT: vmovaps %ymm3, 32(%rdx)
; AVX2-FCP-NEXT: vmovups (%rsp), %ymm3 # 32-byte Reload
; AVX2-FCP-NEXT: vmovaps %ymm3, (%rdx)
; AVX2-FCP-NEXT: vmovdqa %ymm0, 32(%rcx)
; AVX2-FCP-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FCP-NEXT: vmovaps %ymm0, (%rcx)
; AVX2-FCP-NEXT: vmovdqa %ymm2, 32(%r8)
; AVX2-FCP-NEXT: vmovdqa %ymm1, (%r8)
; AVX2-FCP-NEXT: addq $104, %rsp
; AVX2-FCP-NEXT: vzeroupper
; AVX2-FCP-NEXT: retq
;
; AVX512-LABEL: load_i16_stride4_vf32:
; AVX512: # %bb.0:
; AVX512-NEXT: vmovdqa64 (%rdi), %zmm0
; AVX512-NEXT: vmovdqa64 128(%rdi), %zmm1
; AVX512-NEXT: vmovdqa 192(%rdi), %ymm2
; AVX512-NEXT: vpmovqw %ymm2, %xmm2
; AVX512-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX512-NEXT: vmovdqa 240(%rdi), %xmm6
; AVX512-NEXT: vpshufd {{.*#+}} xmm5 = xmm6[0,2,2,3]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm3 = xmm5[0,1,0,2,4,5,6,7]
; AVX512-NEXT: vmovdqa 224(%rdi), %xmm7
; AVX512-NEXT: vpshufd {{.*#+}} xmm9 = xmm7[0,2,2,3]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm4 = xmm9[0,1,0,2,4,5,6,7]
; AVX512-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
; AVX512-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
; AVX512-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm3[6,7]
; AVX512-NEXT: vpmovqw %zmm1, %xmm3
; AVX512-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
; AVX512-NEXT: vmovdqa 64(%rdi), %ymm3
; AVX512-NEXT: vpmovqw %ymm3, %xmm3
; AVX512-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm10
; AVX512-NEXT: vmovdqa 112(%rdi), %xmm3
; AVX512-NEXT: vpshufd {{.*#+}} xmm8 = xmm3[0,2,2,3]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm11 = xmm8[0,1,0,2,4,5,6,7]
; AVX512-NEXT: vmovdqa 96(%rdi), %xmm4
; AVX512-NEXT: vpshufd {{.*#+}} xmm12 = xmm4[0,2,2,3]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm13 = xmm12[0,1,0,2,4,5,6,7]
; AVX512-NEXT: vpunpckldq {{.*#+}} xmm11 = xmm13[0],xmm11[0],xmm13[1],xmm11[1]
; AVX512-NEXT: vinserti128 $1, %xmm11, %ymm0, %ymm11
; AVX512-NEXT: vpblendd {{.*#+}} ymm10 = ymm10[0,1,2,3,4,5],ymm11[6,7]
; AVX512-NEXT: vpmovqw %zmm0, %xmm11
; AVX512-NEXT: vpblendd {{.*#+}} ymm10 = ymm11[0,1,2,3],ymm10[4,5,6,7]
; AVX512-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm10[0,1,2,3],zmm2[0,1,2,3]
; AVX512-NEXT: vmovdqa 64(%rdi), %xmm10
; AVX512-NEXT: vmovdqa 80(%rdi), %xmm11
; AVX512-NEXT: vmovdqa 192(%rdi), %xmm13
; AVX512-NEXT: vmovdqa 208(%rdi), %xmm14
; AVX512-NEXT: vpshuflw {{.*#+}} xmm5 = xmm5[0,1,1,3,4,5,6,7]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm9 = xmm9[0,1,1,3,4,5,6,7]
; AVX512-NEXT: vpunpckldq {{.*#+}} xmm5 = xmm9[0],xmm5[0],xmm9[1],xmm5[1]
; AVX512-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
; AVX512-NEXT: vpshufd {{.*#+}} xmm9 = xmm14[0,2,2,3]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm9 = xmm9[1,3,2,3,4,5,6,7]
; AVX512-NEXT: vpshufd {{.*#+}} xmm15 = xmm13[0,2,2,3]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm15 = xmm15[1,3,2,3,4,5,6,7]
; AVX512-NEXT: vpunpckldq {{.*#+}} xmm9 = xmm15[0],xmm9[0],xmm15[1],xmm9[1]
; AVX512-NEXT: vinserti128 $1, %xmm9, %ymm0, %ymm9
; AVX512-NEXT: vpblendd {{.*#+}} ymm5 = ymm9[0,1,2,3,4,5],ymm5[6,7]
; AVX512-NEXT: vpsrlq $16, %zmm1, %zmm9
; AVX512-NEXT: vpmovqw %zmm9, %xmm9
; AVX512-NEXT: vpblendd {{.*#+}} ymm5 = ymm9[0,1,2,3],ymm5[4,5,6,7]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm8 = xmm8[0,1,1,3,4,5,6,7]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm9 = xmm12[0,1,1,3,4,5,6,7]
; AVX512-NEXT: vpunpckldq {{.*#+}} xmm8 = xmm9[0],xmm8[0],xmm9[1],xmm8[1]
; AVX512-NEXT: vinserti128 $1, %xmm8, %ymm0, %ymm8
; AVX512-NEXT: vpshufd {{.*#+}} xmm9 = xmm11[0,2,2,3]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm9 = xmm9[1,3,2,3,4,5,6,7]
; AVX512-NEXT: vpshufd {{.*#+}} xmm12 = xmm10[0,2,2,3]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm12 = xmm12[1,3,2,3,4,5,6,7]
; AVX512-NEXT: vpunpckldq {{.*#+}} xmm9 = xmm12[0],xmm9[0],xmm12[1],xmm9[1]
; AVX512-NEXT: vinserti128 $1, %xmm9, %ymm0, %ymm9
; AVX512-NEXT: vpblendd {{.*#+}} ymm8 = ymm9[0,1,2,3,4,5],ymm8[6,7]
; AVX512-NEXT: vpsrlq $16, %zmm0, %zmm9
; AVX512-NEXT: vpmovqw %zmm9, %xmm9
; AVX512-NEXT: vpblendd {{.*#+}} ymm8 = ymm9[0,1,2,3],ymm8[4,5,6,7]
; AVX512-NEXT: vshufi64x2 {{.*#+}} zmm5 = zmm8[0,1,2,3],zmm5[0,1,2,3]
; AVX512-NEXT: vpshufd {{.*#+}} xmm6 = xmm6[3,1,2,3]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm8 = xmm6[0,1,2,0,4,5,6,7]
; AVX512-NEXT: vpshufd {{.*#+}} xmm7 = xmm7[3,1,2,3]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm9 = xmm7[0,1,2,0,4,5,6,7]
; AVX512-NEXT: vpunpckldq {{.*#+}} xmm8 = xmm9[0],xmm8[0],xmm9[1],xmm8[1]
; AVX512-NEXT: vinserti128 $1, %xmm8, %ymm0, %ymm12
; AVX512-NEXT: vpshufd {{.*#+}} xmm8 = xmm14[3,1,2,3]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm14 = xmm8[2,0,2,3,4,5,6,7]
; AVX512-NEXT: vpshufd {{.*#+}} xmm9 = xmm13[3,1,2,3]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm13 = xmm9[2,0,2,3,4,5,6,7]
; AVX512-NEXT: vpunpckldq {{.*#+}} xmm13 = xmm13[0],xmm14[0],xmm13[1],xmm14[1]
; AVX512-NEXT: vinserti128 $1, %xmm13, %ymm0, %ymm13
; AVX512-NEXT: vpblendd {{.*#+}} ymm12 = ymm13[0,1,2,3,4,5],ymm12[6,7]
; AVX512-NEXT: vpsrlq $32, %zmm1, %zmm13
; AVX512-NEXT: vpmovqw %zmm13, %xmm13
; AVX512-NEXT: vpblendd {{.*#+}} ymm12 = ymm13[0,1,2,3],ymm12[4,5,6,7]
; AVX512-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[3,1,2,3]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm13 = xmm3[0,1,2,0,4,5,6,7]
; AVX512-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[3,1,2,3]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm14 = xmm4[0,1,2,0,4,5,6,7]
; AVX512-NEXT: vpunpckldq {{.*#+}} xmm13 = xmm14[0],xmm13[0],xmm14[1],xmm13[1]
; AVX512-NEXT: vinserti128 $1, %xmm13, %ymm0, %ymm13
; AVX512-NEXT: vpshufd {{.*#+}} xmm11 = xmm11[3,1,2,3]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm14 = xmm11[2,0,2,3,4,5,6,7]
; AVX512-NEXT: vpshufd {{.*#+}} xmm10 = xmm10[3,1,2,3]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm15 = xmm10[2,0,2,3,4,5,6,7]
; AVX512-NEXT: vpunpckldq {{.*#+}} xmm14 = xmm15[0],xmm14[0],xmm15[1],xmm14[1]
; AVX512-NEXT: vinserti128 $1, %xmm14, %ymm0, %ymm14
; AVX512-NEXT: vpblendd {{.*#+}} ymm13 = ymm14[0,1,2,3,4,5],ymm13[6,7]
; AVX512-NEXT: vpsrlq $32, %zmm0, %zmm14
; AVX512-NEXT: vpmovqw %zmm14, %xmm14
; AVX512-NEXT: vpblendd {{.*#+}} ymm13 = ymm14[0,1,2,3],ymm13[4,5,6,7]
; AVX512-NEXT: vshufi64x2 {{.*#+}} zmm12 = zmm13[0,1,2,3],zmm12[0,1,2,3]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm6 = xmm6[0,1,3,1,4,5,6,7]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm7 = xmm7[0,1,3,1,4,5,6,7]
; AVX512-NEXT: vpunpckldq {{.*#+}} xmm6 = xmm7[0],xmm6[0],xmm7[1],xmm6[1]
; AVX512-NEXT: vinserti128 $1, %xmm6, %ymm0, %ymm6
; AVX512-NEXT: vpshuflw {{.*#+}} xmm7 = xmm8[3,1,2,3,4,5,6,7]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm8 = xmm9[3,1,2,3,4,5,6,7]
; AVX512-NEXT: vpunpckldq {{.*#+}} xmm7 = xmm8[0],xmm7[0],xmm8[1],xmm7[1]
; AVX512-NEXT: vinserti128 $1, %xmm7, %ymm0, %ymm7
; AVX512-NEXT: vpblendd {{.*#+}} ymm6 = ymm7[0,1,2,3,4,5],ymm6[6,7]
; AVX512-NEXT: vpsrlq $48, %zmm1, %zmm1
; AVX512-NEXT: vpmovqw %zmm1, %xmm1
; AVX512-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm6[4,5,6,7]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[0,1,3,1,4,5,6,7]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[0,1,3,1,4,5,6,7]
; AVX512-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
; AVX512-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
; AVX512-NEXT: vpshuflw {{.*#+}} xmm4 = xmm11[3,1,2,3,4,5,6,7]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm6 = xmm10[3,1,2,3,4,5,6,7]
; AVX512-NEXT: vpunpckldq {{.*#+}} xmm4 = xmm6[0],xmm4[0],xmm6[1],xmm4[1]
; AVX512-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4
; AVX512-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3,4,5],ymm3[6,7]
; AVX512-NEXT: vpsrlq $48, %zmm0, %zmm0
; AVX512-NEXT: vpmovqw %zmm0, %xmm0
; AVX512-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
; AVX512-NEXT: vshufi64x2 {{.*#+}} zmm0 = zmm0[0,1,2,3],zmm1[0,1,2,3]
; AVX512-NEXT: vmovdqa64 %zmm2, (%rsi)
; AVX512-NEXT: vmovdqa64 %zmm5, (%rdx)
; AVX512-NEXT: vmovdqa64 %zmm12, (%rcx)
; AVX512-NEXT: vmovdqa64 %zmm0, (%r8)
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
;
; AVX512-FCP-LABEL: load_i16_stride4_vf32:
; AVX512-FCP: # %bb.0:
; AVX512-FCP-NEXT: vmovdqa64 (%rdi), %zmm0
; AVX512-FCP-NEXT: vmovdqa64 128(%rdi), %zmm1
; AVX512-FCP-NEXT: vbroadcasti128 {{.*#+}} ymm2 = [16,17,20,21,0,0,0,0,24,25,28,29,0,0,0,0,16,17,20,21,0,0,0,0,24,25,28,29,0,0,0,0]
; AVX512-FCP-NEXT: # ymm2 = mem[0,1,0,1]
; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} ymm4 = [0,2,2,3,0,2,4,6]
; AVX512-FCP-NEXT: vmovdqa 224(%rdi), %ymm5
; AVX512-FCP-NEXT: vpermd %ymm5, %ymm4, %ymm6
; AVX512-FCP-NEXT: vpshufb %ymm2, %ymm6, %ymm7
; AVX512-FCP-NEXT: vmovdqa {{.*#+}} ymm3 = [0,1,4,5,u,u,u,u,8,9,12,13,12,13,14,15,16,17,20,21,u,u,u,u,24,25,28,29,u,u,u,u]
; AVX512-FCP-NEXT: vmovdqa 192(%rdi), %ymm8
; AVX512-FCP-NEXT: vpermd %ymm8, %ymm4, %ymm9
; AVX512-FCP-NEXT: vpshufb %ymm3, %ymm9, %ymm10
; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} ymm11 = [0,2,2,3,4,6,12,14]
; AVX512-FCP-NEXT: vpermt2d %ymm7, %ymm11, %ymm10
; AVX512-FCP-NEXT: vpmovqw %zmm1, %xmm7
; AVX512-FCP-NEXT: vpblendd {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm10[4,5,6,7]
; AVX512-FCP-NEXT: vmovdqa 96(%rdi), %ymm10
; AVX512-FCP-NEXT: vpermd %ymm10, %ymm4, %ymm12
; AVX512-FCP-NEXT: vpshufb %ymm2, %ymm12, %ymm13
; AVX512-FCP-NEXT: vmovdqa 64(%rdi), %ymm14
; AVX512-FCP-NEXT: vpermd %ymm14, %ymm4, %ymm15
; AVX512-FCP-NEXT: vpshufb %ymm3, %ymm15, %ymm4
; AVX512-FCP-NEXT: vpermt2d %ymm13, %ymm11, %ymm4
; AVX512-FCP-NEXT: vpmovqw %zmm0, %xmm13
; AVX512-FCP-NEXT: vpblendd {{.*#+}} ymm4 = ymm13[0,1,2,3],ymm4[4,5,6,7]
; AVX512-FCP-NEXT: vshufi64x2 {{.*#+}} zmm7 = zmm4[0,1,2,3],zmm7[0,1,2,3]
; AVX512-FCP-NEXT: vmovdqa {{.*#+}} ymm4 = [2,3,6,7,10,11,14,15,14,15,10,11,12,13,14,15,18,19,22,23,26,27,30,31,18,19,22,23,26,27,30,31]
; AVX512-FCP-NEXT: vpshufb %ymm4, %ymm6, %ymm6
; AVX512-FCP-NEXT: vpshufb %ymm4, %ymm9, %ymm9
; AVX512-FCP-NEXT: vpblendd {{.*#+}} ymm6 = ymm9[0,1,2,3,4,5],ymm6[6,7]
; AVX512-FCP-NEXT: vpsrlq $16, %zmm1, %zmm9
; AVX512-FCP-NEXT: vpmovqw %zmm9, %xmm9
; AVX512-FCP-NEXT: vpblendd {{.*#+}} ymm6 = ymm9[0,1,2,3],ymm6[4,5,6,7]
; AVX512-FCP-NEXT: vpshufb %ymm4, %ymm12, %ymm9
; AVX512-FCP-NEXT: vpshufb %ymm4, %ymm15, %ymm12
; AVX512-FCP-NEXT: vpblendd {{.*#+}} ymm9 = ymm12[0,1,2,3,4,5],ymm9[6,7]
; AVX512-FCP-NEXT: vpsrlq $16, %zmm0, %zmm12
; AVX512-FCP-NEXT: vpmovqw %zmm12, %xmm12
; AVX512-FCP-NEXT: vpblendd {{.*#+}} ymm9 = ymm12[0,1,2,3],ymm9[4,5,6,7]
; AVX512-FCP-NEXT: vshufi64x2 {{.*#+}} zmm6 = zmm9[0,1,2,3],zmm6[0,1,2,3]
; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} ymm9 = [1,3,2,3,1,3,5,7]
; AVX512-FCP-NEXT: vpermd %ymm5, %ymm9, %ymm5
; AVX512-FCP-NEXT: vpshufb %ymm2, %ymm5, %ymm12
; AVX512-FCP-NEXT: vpermd %ymm8, %ymm9, %ymm8
; AVX512-FCP-NEXT: vpshufb %ymm3, %ymm8, %ymm13
; AVX512-FCP-NEXT: vpermt2d %ymm12, %ymm11, %ymm13
; AVX512-FCP-NEXT: vpsrlq $32, %zmm1, %zmm12
; AVX512-FCP-NEXT: vpmovqw %zmm12, %xmm12
; AVX512-FCP-NEXT: vpblendd {{.*#+}} ymm12 = ymm12[0,1,2,3],ymm13[4,5,6,7]
; AVX512-FCP-NEXT: vpermd %ymm10, %ymm9, %ymm10
; AVX512-FCP-NEXT: vpshufb %ymm2, %ymm10, %ymm2
; AVX512-FCP-NEXT: vpermd %ymm14, %ymm9, %ymm9
; AVX512-FCP-NEXT: vpshufb %ymm3, %ymm9, %ymm3
; AVX512-FCP-NEXT: vpermt2d %ymm2, %ymm11, %ymm3
; AVX512-FCP-NEXT: vpsrlq $32, %zmm0, %zmm2
; AVX512-FCP-NEXT: vpmovqw %zmm2, %xmm2
; AVX512-FCP-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
; AVX512-FCP-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm12[0,1,2,3]
; AVX512-FCP-NEXT: vpshufb %ymm4, %ymm5, %ymm3
; AVX512-FCP-NEXT: vpshufb %ymm4, %ymm8, %ymm5
; AVX512-FCP-NEXT: vpblendd {{.*#+}} ymm3 = ymm5[0,1,2,3,4,5],ymm3[6,7]
; AVX512-FCP-NEXT: vpsrlq $48, %zmm1, %zmm1
; AVX512-FCP-NEXT: vpmovqw %zmm1, %xmm1
; AVX512-FCP-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
; AVX512-FCP-NEXT: vpshufb %ymm4, %ymm10, %ymm3
; AVX512-FCP-NEXT: vpshufb %ymm4, %ymm9, %ymm4
; AVX512-FCP-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3,4,5],ymm3[6,7]
; AVX512-FCP-NEXT: vpsrlq $48, %zmm0, %zmm0
; AVX512-FCP-NEXT: vpmovqw %zmm0, %xmm0
; AVX512-FCP-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
; AVX512-FCP-NEXT: vshufi64x2 {{.*#+}} zmm0 = zmm0[0,1,2,3],zmm1[0,1,2,3]
; AVX512-FCP-NEXT: vmovdqa64 %zmm7, (%rsi)
; AVX512-FCP-NEXT: vmovdqa64 %zmm6, (%rdx)
; AVX512-FCP-NEXT: vmovdqa64 %zmm2, (%rcx)
; AVX512-FCP-NEXT: vmovdqa64 %zmm0, (%r8)
; AVX512-FCP-NEXT: vzeroupper
; AVX512-FCP-NEXT: retq
;
; AVX512DQ-LABEL: load_i16_stride4_vf32:
; AVX512DQ: # %bb.0:
; AVX512DQ-NEXT: vmovdqa64 (%rdi), %zmm0
; AVX512DQ-NEXT: vmovdqa64 128(%rdi), %zmm1
; AVX512DQ-NEXT: vmovdqa 192(%rdi), %ymm2
; AVX512DQ-NEXT: vpmovqw %ymm2, %xmm2
; AVX512DQ-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX512DQ-NEXT: vmovdqa 240(%rdi), %xmm6
; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm5 = xmm6[0,2,2,3]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm3 = xmm5[0,1,0,2,4,5,6,7]
; AVX512DQ-NEXT: vmovdqa 224(%rdi), %xmm7
; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm9 = xmm7[0,2,2,3]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm4 = xmm9[0,1,0,2,4,5,6,7]
; AVX512DQ-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
; AVX512DQ-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
; AVX512DQ-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm3[6,7]
; AVX512DQ-NEXT: vpmovqw %zmm1, %xmm3
; AVX512DQ-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
; AVX512DQ-NEXT: vmovdqa 64(%rdi), %ymm3
; AVX512DQ-NEXT: vpmovqw %ymm3, %xmm3
; AVX512DQ-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm10
; AVX512DQ-NEXT: vmovdqa 112(%rdi), %xmm3
; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm8 = xmm3[0,2,2,3]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm11 = xmm8[0,1,0,2,4,5,6,7]
; AVX512DQ-NEXT: vmovdqa 96(%rdi), %xmm4
; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm12 = xmm4[0,2,2,3]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm13 = xmm12[0,1,0,2,4,5,6,7]
; AVX512DQ-NEXT: vpunpckldq {{.*#+}} xmm11 = xmm13[0],xmm11[0],xmm13[1],xmm11[1]
; AVX512DQ-NEXT: vinserti128 $1, %xmm11, %ymm0, %ymm11
; AVX512DQ-NEXT: vpblendd {{.*#+}} ymm10 = ymm10[0,1,2,3,4,5],ymm11[6,7]
; AVX512DQ-NEXT: vpmovqw %zmm0, %xmm11
; AVX512DQ-NEXT: vpblendd {{.*#+}} ymm10 = ymm11[0,1,2,3],ymm10[4,5,6,7]
; AVX512DQ-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm10[0,1,2,3],zmm2[0,1,2,3]
; AVX512DQ-NEXT: vmovdqa 64(%rdi), %xmm10
; AVX512DQ-NEXT: vmovdqa 80(%rdi), %xmm11
; AVX512DQ-NEXT: vmovdqa 192(%rdi), %xmm13
; AVX512DQ-NEXT: vmovdqa 208(%rdi), %xmm14
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm5 = xmm5[0,1,1,3,4,5,6,7]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm9 = xmm9[0,1,1,3,4,5,6,7]
; AVX512DQ-NEXT: vpunpckldq {{.*#+}} xmm5 = xmm9[0],xmm5[0],xmm9[1],xmm5[1]
; AVX512DQ-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm9 = xmm14[0,2,2,3]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm9 = xmm9[1,3,2,3,4,5,6,7]
; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm15 = xmm13[0,2,2,3]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm15 = xmm15[1,3,2,3,4,5,6,7]
; AVX512DQ-NEXT: vpunpckldq {{.*#+}} xmm9 = xmm15[0],xmm9[0],xmm15[1],xmm9[1]
; AVX512DQ-NEXT: vinserti128 $1, %xmm9, %ymm0, %ymm9
; AVX512DQ-NEXT: vpblendd {{.*#+}} ymm5 = ymm9[0,1,2,3,4,5],ymm5[6,7]
; AVX512DQ-NEXT: vpsrlq $16, %zmm1, %zmm9
; AVX512DQ-NEXT: vpmovqw %zmm9, %xmm9
; AVX512DQ-NEXT: vpblendd {{.*#+}} ymm5 = ymm9[0,1,2,3],ymm5[4,5,6,7]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm8 = xmm8[0,1,1,3,4,5,6,7]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm9 = xmm12[0,1,1,3,4,5,6,7]
; AVX512DQ-NEXT: vpunpckldq {{.*#+}} xmm8 = xmm9[0],xmm8[0],xmm9[1],xmm8[1]
; AVX512DQ-NEXT: vinserti128 $1, %xmm8, %ymm0, %ymm8
; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm9 = xmm11[0,2,2,3]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm9 = xmm9[1,3,2,3,4,5,6,7]
; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm12 = xmm10[0,2,2,3]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm12 = xmm12[1,3,2,3,4,5,6,7]
; AVX512DQ-NEXT: vpunpckldq {{.*#+}} xmm9 = xmm12[0],xmm9[0],xmm12[1],xmm9[1]
; AVX512DQ-NEXT: vinserti128 $1, %xmm9, %ymm0, %ymm9
; AVX512DQ-NEXT: vpblendd {{.*#+}} ymm8 = ymm9[0,1,2,3,4,5],ymm8[6,7]
; AVX512DQ-NEXT: vpsrlq $16, %zmm0, %zmm9
; AVX512DQ-NEXT: vpmovqw %zmm9, %xmm9
; AVX512DQ-NEXT: vpblendd {{.*#+}} ymm8 = ymm9[0,1,2,3],ymm8[4,5,6,7]
; AVX512DQ-NEXT: vshufi64x2 {{.*#+}} zmm5 = zmm8[0,1,2,3],zmm5[0,1,2,3]
; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm6 = xmm6[3,1,2,3]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm8 = xmm6[0,1,2,0,4,5,6,7]
; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm7 = xmm7[3,1,2,3]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm9 = xmm7[0,1,2,0,4,5,6,7]
; AVX512DQ-NEXT: vpunpckldq {{.*#+}} xmm8 = xmm9[0],xmm8[0],xmm9[1],xmm8[1]
; AVX512DQ-NEXT: vinserti128 $1, %xmm8, %ymm0, %ymm12
; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm8 = xmm14[3,1,2,3]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm14 = xmm8[2,0,2,3,4,5,6,7]
; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm9 = xmm13[3,1,2,3]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm13 = xmm9[2,0,2,3,4,5,6,7]
; AVX512DQ-NEXT: vpunpckldq {{.*#+}} xmm13 = xmm13[0],xmm14[0],xmm13[1],xmm14[1]
; AVX512DQ-NEXT: vinserti128 $1, %xmm13, %ymm0, %ymm13
; AVX512DQ-NEXT: vpblendd {{.*#+}} ymm12 = ymm13[0,1,2,3,4,5],ymm12[6,7]
; AVX512DQ-NEXT: vpsrlq $32, %zmm1, %zmm13
; AVX512DQ-NEXT: vpmovqw %zmm13, %xmm13
; AVX512DQ-NEXT: vpblendd {{.*#+}} ymm12 = ymm13[0,1,2,3],ymm12[4,5,6,7]
; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[3,1,2,3]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm13 = xmm3[0,1,2,0,4,5,6,7]
; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[3,1,2,3]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm14 = xmm4[0,1,2,0,4,5,6,7]
; AVX512DQ-NEXT: vpunpckldq {{.*#+}} xmm13 = xmm14[0],xmm13[0],xmm14[1],xmm13[1]
; AVX512DQ-NEXT: vinserti128 $1, %xmm13, %ymm0, %ymm13
; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm11 = xmm11[3,1,2,3]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm14 = xmm11[2,0,2,3,4,5,6,7]
; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm10 = xmm10[3,1,2,3]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm15 = xmm10[2,0,2,3,4,5,6,7]
; AVX512DQ-NEXT: vpunpckldq {{.*#+}} xmm14 = xmm15[0],xmm14[0],xmm15[1],xmm14[1]
; AVX512DQ-NEXT: vinserti128 $1, %xmm14, %ymm0, %ymm14
; AVX512DQ-NEXT: vpblendd {{.*#+}} ymm13 = ymm14[0,1,2,3,4,5],ymm13[6,7]
; AVX512DQ-NEXT: vpsrlq $32, %zmm0, %zmm14
; AVX512DQ-NEXT: vpmovqw %zmm14, %xmm14
; AVX512DQ-NEXT: vpblendd {{.*#+}} ymm13 = ymm14[0,1,2,3],ymm13[4,5,6,7]
; AVX512DQ-NEXT: vshufi64x2 {{.*#+}} zmm12 = zmm13[0,1,2,3],zmm12[0,1,2,3]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm6 = xmm6[0,1,3,1,4,5,6,7]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm7 = xmm7[0,1,3,1,4,5,6,7]
; AVX512DQ-NEXT: vpunpckldq {{.*#+}} xmm6 = xmm7[0],xmm6[0],xmm7[1],xmm6[1]
; AVX512DQ-NEXT: vinserti128 $1, %xmm6, %ymm0, %ymm6
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm7 = xmm8[3,1,2,3,4,5,6,7]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm8 = xmm9[3,1,2,3,4,5,6,7]
; AVX512DQ-NEXT: vpunpckldq {{.*#+}} xmm7 = xmm8[0],xmm7[0],xmm8[1],xmm7[1]
; AVX512DQ-NEXT: vinserti128 $1, %xmm7, %ymm0, %ymm7
; AVX512DQ-NEXT: vpblendd {{.*#+}} ymm6 = ymm7[0,1,2,3,4,5],ymm6[6,7]
; AVX512DQ-NEXT: vpsrlq $48, %zmm1, %zmm1
; AVX512DQ-NEXT: vpmovqw %zmm1, %xmm1
; AVX512DQ-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm6[4,5,6,7]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[0,1,3,1,4,5,6,7]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[0,1,3,1,4,5,6,7]
; AVX512DQ-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
; AVX512DQ-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm4 = xmm11[3,1,2,3,4,5,6,7]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm6 = xmm10[3,1,2,3,4,5,6,7]
; AVX512DQ-NEXT: vpunpckldq {{.*#+}} xmm4 = xmm6[0],xmm4[0],xmm6[1],xmm4[1]
; AVX512DQ-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4
; AVX512DQ-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3,4,5],ymm3[6,7]
; AVX512DQ-NEXT: vpsrlq $48, %zmm0, %zmm0
; AVX512DQ-NEXT: vpmovqw %zmm0, %xmm0
; AVX512DQ-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
; AVX512DQ-NEXT: vshufi64x2 {{.*#+}} zmm0 = zmm0[0,1,2,3],zmm1[0,1,2,3]
; AVX512DQ-NEXT: vmovdqa64 %zmm2, (%rsi)
; AVX512DQ-NEXT: vmovdqa64 %zmm5, (%rdx)
; AVX512DQ-NEXT: vmovdqa64 %zmm12, (%rcx)
; AVX512DQ-NEXT: vmovdqa64 %zmm0, (%r8)
; AVX512DQ-NEXT: vzeroupper
; AVX512DQ-NEXT: retq
;
; AVX512DQ-FCP-LABEL: load_i16_stride4_vf32:
; AVX512DQ-FCP: # %bb.0:
; AVX512DQ-FCP-NEXT: vmovdqa64 (%rdi), %zmm0
; AVX512DQ-FCP-NEXT: vmovdqa64 128(%rdi), %zmm1
; AVX512DQ-FCP-NEXT: vbroadcasti128 {{.*#+}} ymm2 = [16,17,20,21,0,0,0,0,24,25,28,29,0,0,0,0,16,17,20,21,0,0,0,0,24,25,28,29,0,0,0,0]
; AVX512DQ-FCP-NEXT: # ymm2 = mem[0,1,0,1]
; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} ymm4 = [0,2,2,3,0,2,4,6]
; AVX512DQ-FCP-NEXT: vmovdqa 224(%rdi), %ymm5
; AVX512DQ-FCP-NEXT: vpermd %ymm5, %ymm4, %ymm6
; AVX512DQ-FCP-NEXT: vpshufb %ymm2, %ymm6, %ymm7
; AVX512DQ-FCP-NEXT: vmovdqa {{.*#+}} ymm3 = [0,1,4,5,u,u,u,u,8,9,12,13,12,13,14,15,16,17,20,21,u,u,u,u,24,25,28,29,u,u,u,u]
; AVX512DQ-FCP-NEXT: vmovdqa 192(%rdi), %ymm8
; AVX512DQ-FCP-NEXT: vpermd %ymm8, %ymm4, %ymm9
; AVX512DQ-FCP-NEXT: vpshufb %ymm3, %ymm9, %ymm10
; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} ymm11 = [0,2,2,3,4,6,12,14]
; AVX512DQ-FCP-NEXT: vpermt2d %ymm7, %ymm11, %ymm10
; AVX512DQ-FCP-NEXT: vpmovqw %zmm1, %xmm7
; AVX512DQ-FCP-NEXT: vpblendd {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm10[4,5,6,7]
; AVX512DQ-FCP-NEXT: vmovdqa 96(%rdi), %ymm10
; AVX512DQ-FCP-NEXT: vpermd %ymm10, %ymm4, %ymm12
; AVX512DQ-FCP-NEXT: vpshufb %ymm2, %ymm12, %ymm13
; AVX512DQ-FCP-NEXT: vmovdqa 64(%rdi), %ymm14
; AVX512DQ-FCP-NEXT: vpermd %ymm14, %ymm4, %ymm15
; AVX512DQ-FCP-NEXT: vpshufb %ymm3, %ymm15, %ymm4
; AVX512DQ-FCP-NEXT: vpermt2d %ymm13, %ymm11, %ymm4
; AVX512DQ-FCP-NEXT: vpmovqw %zmm0, %xmm13
; AVX512DQ-FCP-NEXT: vpblendd {{.*#+}} ymm4 = ymm13[0,1,2,3],ymm4[4,5,6,7]
; AVX512DQ-FCP-NEXT: vshufi64x2 {{.*#+}} zmm7 = zmm4[0,1,2,3],zmm7[0,1,2,3]
; AVX512DQ-FCP-NEXT: vmovdqa {{.*#+}} ymm4 = [2,3,6,7,10,11,14,15,14,15,10,11,12,13,14,15,18,19,22,23,26,27,30,31,18,19,22,23,26,27,30,31]
; AVX512DQ-FCP-NEXT: vpshufb %ymm4, %ymm6, %ymm6
; AVX512DQ-FCP-NEXT: vpshufb %ymm4, %ymm9, %ymm9
; AVX512DQ-FCP-NEXT: vpblendd {{.*#+}} ymm6 = ymm9[0,1,2,3,4,5],ymm6[6,7]
; AVX512DQ-FCP-NEXT: vpsrlq $16, %zmm1, %zmm9
; AVX512DQ-FCP-NEXT: vpmovqw %zmm9, %xmm9
; AVX512DQ-FCP-NEXT: vpblendd {{.*#+}} ymm6 = ymm9[0,1,2,3],ymm6[4,5,6,7]
; AVX512DQ-FCP-NEXT: vpshufb %ymm4, %ymm12, %ymm9
; AVX512DQ-FCP-NEXT: vpshufb %ymm4, %ymm15, %ymm12
; AVX512DQ-FCP-NEXT: vpblendd {{.*#+}} ymm9 = ymm12[0,1,2,3,4,5],ymm9[6,7]
; AVX512DQ-FCP-NEXT: vpsrlq $16, %zmm0, %zmm12
; AVX512DQ-FCP-NEXT: vpmovqw %zmm12, %xmm12
; AVX512DQ-FCP-NEXT: vpblendd {{.*#+}} ymm9 = ymm12[0,1,2,3],ymm9[4,5,6,7]
; AVX512DQ-FCP-NEXT: vshufi64x2 {{.*#+}} zmm6 = zmm9[0,1,2,3],zmm6[0,1,2,3]
; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} ymm9 = [1,3,2,3,1,3,5,7]
; AVX512DQ-FCP-NEXT: vpermd %ymm5, %ymm9, %ymm5
; AVX512DQ-FCP-NEXT: vpshufb %ymm2, %ymm5, %ymm12
; AVX512DQ-FCP-NEXT: vpermd %ymm8, %ymm9, %ymm8
; AVX512DQ-FCP-NEXT: vpshufb %ymm3, %ymm8, %ymm13
; AVX512DQ-FCP-NEXT: vpermt2d %ymm12, %ymm11, %ymm13
; AVX512DQ-FCP-NEXT: vpsrlq $32, %zmm1, %zmm12
; AVX512DQ-FCP-NEXT: vpmovqw %zmm12, %xmm12
; AVX512DQ-FCP-NEXT: vpblendd {{.*#+}} ymm12 = ymm12[0,1,2,3],ymm13[4,5,6,7]
; AVX512DQ-FCP-NEXT: vpermd %ymm10, %ymm9, %ymm10
; AVX512DQ-FCP-NEXT: vpshufb %ymm2, %ymm10, %ymm2
; AVX512DQ-FCP-NEXT: vpermd %ymm14, %ymm9, %ymm9
; AVX512DQ-FCP-NEXT: vpshufb %ymm3, %ymm9, %ymm3
; AVX512DQ-FCP-NEXT: vpermt2d %ymm2, %ymm11, %ymm3
; AVX512DQ-FCP-NEXT: vpsrlq $32, %zmm0, %zmm2
; AVX512DQ-FCP-NEXT: vpmovqw %zmm2, %xmm2
; AVX512DQ-FCP-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
; AVX512DQ-FCP-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm12[0,1,2,3]
; AVX512DQ-FCP-NEXT: vpshufb %ymm4, %ymm5, %ymm3
; AVX512DQ-FCP-NEXT: vpshufb %ymm4, %ymm8, %ymm5
; AVX512DQ-FCP-NEXT: vpblendd {{.*#+}} ymm3 = ymm5[0,1,2,3,4,5],ymm3[6,7]
; AVX512DQ-FCP-NEXT: vpsrlq $48, %zmm1, %zmm1
; AVX512DQ-FCP-NEXT: vpmovqw %zmm1, %xmm1
; AVX512DQ-FCP-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
; AVX512DQ-FCP-NEXT: vpshufb %ymm4, %ymm10, %ymm3
; AVX512DQ-FCP-NEXT: vpshufb %ymm4, %ymm9, %ymm4
; AVX512DQ-FCP-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3,4,5],ymm3[6,7]
; AVX512DQ-FCP-NEXT: vpsrlq $48, %zmm0, %zmm0
; AVX512DQ-FCP-NEXT: vpmovqw %zmm0, %xmm0
; AVX512DQ-FCP-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
; AVX512DQ-FCP-NEXT: vshufi64x2 {{.*#+}} zmm0 = zmm0[0,1,2,3],zmm1[0,1,2,3]
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm7, (%rsi)
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm6, (%rdx)
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm2, (%rcx)
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm0, (%r8)
; AVX512DQ-FCP-NEXT: vzeroupper
; AVX512DQ-FCP-NEXT: retq
;
; AVX512BW-LABEL: load_i16_stride4_vf32:
; AVX512BW: # %bb.0:
; AVX512BW-NEXT: vmovdqa64 (%rdi), %zmm0
; AVX512BW-NEXT: vmovdqa64 64(%rdi), %zmm1
; AVX512BW-NEXT: vmovdqa64 128(%rdi), %zmm2
; AVX512BW-NEXT: vmovdqa64 192(%rdi), %zmm3
; AVX512BW-NEXT: vbroadcasti64x4 {{.*#+}} zmm4 = [0,4,8,12,16,20,24,28,32,36,40,44,48,52,56,60,0,4,8,12,16,20,24,28,32,36,40,44,48,52,56,60]
; AVX512BW-NEXT: # zmm4 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-NEXT: vmovdqa64 %zmm2, %zmm5
; AVX512BW-NEXT: vpermt2w %zmm3, %zmm4, %zmm5
; AVX512BW-NEXT: vpermi2w %zmm1, %zmm0, %zmm4
; AVX512BW-NEXT: vshufi64x2 {{.*#+}} zmm4 = zmm4[0,1,2,3],zmm5[4,5,6,7]
; AVX512BW-NEXT: vbroadcasti64x4 {{.*#+}} zmm5 = [1,5,9,13,17,21,25,29,33,37,41,45,49,53,57,61,1,5,9,13,17,21,25,29,33,37,41,45,49,53,57,61]
; AVX512BW-NEXT: # zmm5 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-NEXT: vmovdqa64 %zmm2, %zmm6
; AVX512BW-NEXT: vpermt2w %zmm3, %zmm5, %zmm6
; AVX512BW-NEXT: vpermi2w %zmm1, %zmm0, %zmm5
; AVX512BW-NEXT: vshufi64x2 {{.*#+}} zmm5 = zmm5[0,1,2,3],zmm6[4,5,6,7]
; AVX512BW-NEXT: vbroadcasti64x4 {{.*#+}} zmm6 = [2,6,10,14,18,22,26,30,34,38,42,46,50,54,58,62,2,6,10,14,18,22,26,30,34,38,42,46,50,54,58,62]
; AVX512BW-NEXT: # zmm6 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-NEXT: vmovdqa64 %zmm2, %zmm7
; AVX512BW-NEXT: vpermt2w %zmm3, %zmm6, %zmm7
; AVX512BW-NEXT: vpermi2w %zmm1, %zmm0, %zmm6
; AVX512BW-NEXT: vshufi64x2 {{.*#+}} zmm6 = zmm6[0,1,2,3],zmm7[4,5,6,7]
; AVX512BW-NEXT: vbroadcasti64x4 {{.*#+}} zmm7 = [3,7,11,15,19,23,27,31,35,39,43,47,51,55,59,63,3,7,11,15,19,23,27,31,35,39,43,47,51,55,59,63]
; AVX512BW-NEXT: # zmm7 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-NEXT: vpermt2w %zmm3, %zmm7, %zmm2
; AVX512BW-NEXT: vpermt2w %zmm1, %zmm7, %zmm0
; AVX512BW-NEXT: vshufi64x2 {{.*#+}} zmm0 = zmm0[0,1,2,3],zmm2[4,5,6,7]
; AVX512BW-NEXT: vmovdqa64 %zmm4, (%rsi)
; AVX512BW-NEXT: vmovdqa64 %zmm5, (%rdx)
; AVX512BW-NEXT: vmovdqa64 %zmm6, (%rcx)
; AVX512BW-NEXT: vmovdqa64 %zmm0, (%r8)
; AVX512BW-NEXT: vzeroupper
; AVX512BW-NEXT: retq
;
; AVX512BW-FCP-LABEL: load_i16_stride4_vf32:
; AVX512BW-FCP: # %bb.0:
; AVX512BW-FCP-NEXT: vmovdqa64 (%rdi), %zmm0
; AVX512BW-FCP-NEXT: vmovdqa64 64(%rdi), %zmm1
; AVX512BW-FCP-NEXT: vmovdqa64 128(%rdi), %zmm2
; AVX512BW-FCP-NEXT: vmovdqa64 192(%rdi), %zmm3
; AVX512BW-FCP-NEXT: vbroadcasti64x4 {{.*#+}} zmm4 = [0,4,8,12,16,20,24,28,32,36,40,44,48,52,56,60,0,4,8,12,16,20,24,28,32,36,40,44,48,52,56,60]
; AVX512BW-FCP-NEXT: # zmm4 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm2, %zmm5
; AVX512BW-FCP-NEXT: vpermt2w %zmm3, %zmm4, %zmm5
; AVX512BW-FCP-NEXT: vpermi2w %zmm1, %zmm0, %zmm4
; AVX512BW-FCP-NEXT: vshufi64x2 {{.*#+}} zmm4 = zmm4[0,1,2,3],zmm5[4,5,6,7]
; AVX512BW-FCP-NEXT: vbroadcasti64x4 {{.*#+}} zmm5 = [1,5,9,13,17,21,25,29,33,37,41,45,49,53,57,61,1,5,9,13,17,21,25,29,33,37,41,45,49,53,57,61]
; AVX512BW-FCP-NEXT: # zmm5 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm2, %zmm6
; AVX512BW-FCP-NEXT: vpermt2w %zmm3, %zmm5, %zmm6
; AVX512BW-FCP-NEXT: vpermi2w %zmm1, %zmm0, %zmm5
; AVX512BW-FCP-NEXT: vshufi64x2 {{.*#+}} zmm5 = zmm5[0,1,2,3],zmm6[4,5,6,7]
; AVX512BW-FCP-NEXT: vbroadcasti64x4 {{.*#+}} zmm6 = [2,6,10,14,18,22,26,30,34,38,42,46,50,54,58,62,2,6,10,14,18,22,26,30,34,38,42,46,50,54,58,62]
; AVX512BW-FCP-NEXT: # zmm6 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm2, %zmm7
; AVX512BW-FCP-NEXT: vpermt2w %zmm3, %zmm6, %zmm7
; AVX512BW-FCP-NEXT: vpermi2w %zmm1, %zmm0, %zmm6
; AVX512BW-FCP-NEXT: vshufi64x2 {{.*#+}} zmm6 = zmm6[0,1,2,3],zmm7[4,5,6,7]
; AVX512BW-FCP-NEXT: vbroadcasti64x4 {{.*#+}} zmm7 = [3,7,11,15,19,23,27,31,35,39,43,47,51,55,59,63,3,7,11,15,19,23,27,31,35,39,43,47,51,55,59,63]
; AVX512BW-FCP-NEXT: # zmm7 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-FCP-NEXT: vpermt2w %zmm3, %zmm7, %zmm2
; AVX512BW-FCP-NEXT: vpermt2w %zmm1, %zmm7, %zmm0
; AVX512BW-FCP-NEXT: vshufi64x2 {{.*#+}} zmm0 = zmm0[0,1,2,3],zmm2[4,5,6,7]
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm4, (%rsi)
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm5, (%rdx)
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm6, (%rcx)
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm0, (%r8)
; AVX512BW-FCP-NEXT: vzeroupper
; AVX512BW-FCP-NEXT: retq
;
; AVX512DQ-BW-LABEL: load_i16_stride4_vf32:
; AVX512DQ-BW: # %bb.0:
; AVX512DQ-BW-NEXT: vmovdqa64 (%rdi), %zmm0
; AVX512DQ-BW-NEXT: vmovdqa64 64(%rdi), %zmm1
; AVX512DQ-BW-NEXT: vmovdqa64 128(%rdi), %zmm2
; AVX512DQ-BW-NEXT: vmovdqa64 192(%rdi), %zmm3
; AVX512DQ-BW-NEXT: vbroadcasti64x4 {{.*#+}} zmm4 = [0,4,8,12,16,20,24,28,32,36,40,44,48,52,56,60,0,4,8,12,16,20,24,28,32,36,40,44,48,52,56,60]
; AVX512DQ-BW-NEXT: # zmm4 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm2, %zmm5
; AVX512DQ-BW-NEXT: vpermt2w %zmm3, %zmm4, %zmm5
; AVX512DQ-BW-NEXT: vpermi2w %zmm1, %zmm0, %zmm4
; AVX512DQ-BW-NEXT: vshufi64x2 {{.*#+}} zmm4 = zmm4[0,1,2,3],zmm5[4,5,6,7]
; AVX512DQ-BW-NEXT: vbroadcasti64x4 {{.*#+}} zmm5 = [1,5,9,13,17,21,25,29,33,37,41,45,49,53,57,61,1,5,9,13,17,21,25,29,33,37,41,45,49,53,57,61]
; AVX512DQ-BW-NEXT: # zmm5 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm2, %zmm6
; AVX512DQ-BW-NEXT: vpermt2w %zmm3, %zmm5, %zmm6
; AVX512DQ-BW-NEXT: vpermi2w %zmm1, %zmm0, %zmm5
; AVX512DQ-BW-NEXT: vshufi64x2 {{.*#+}} zmm5 = zmm5[0,1,2,3],zmm6[4,5,6,7]
; AVX512DQ-BW-NEXT: vbroadcasti64x4 {{.*#+}} zmm6 = [2,6,10,14,18,22,26,30,34,38,42,46,50,54,58,62,2,6,10,14,18,22,26,30,34,38,42,46,50,54,58,62]
; AVX512DQ-BW-NEXT: # zmm6 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm2, %zmm7
; AVX512DQ-BW-NEXT: vpermt2w %zmm3, %zmm6, %zmm7
; AVX512DQ-BW-NEXT: vpermi2w %zmm1, %zmm0, %zmm6
; AVX512DQ-BW-NEXT: vshufi64x2 {{.*#+}} zmm6 = zmm6[0,1,2,3],zmm7[4,5,6,7]
; AVX512DQ-BW-NEXT: vbroadcasti64x4 {{.*#+}} zmm7 = [3,7,11,15,19,23,27,31,35,39,43,47,51,55,59,63,3,7,11,15,19,23,27,31,35,39,43,47,51,55,59,63]
; AVX512DQ-BW-NEXT: # zmm7 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-NEXT: vpermt2w %zmm3, %zmm7, %zmm2
; AVX512DQ-BW-NEXT: vpermt2w %zmm1, %zmm7, %zmm0
; AVX512DQ-BW-NEXT: vshufi64x2 {{.*#+}} zmm0 = zmm0[0,1,2,3],zmm2[4,5,6,7]
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm4, (%rsi)
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm5, (%rdx)
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm6, (%rcx)
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm0, (%r8)
; AVX512DQ-BW-NEXT: vzeroupper
; AVX512DQ-BW-NEXT: retq
;
; AVX512DQ-BW-FCP-LABEL: load_i16_stride4_vf32:
; AVX512DQ-BW-FCP: # %bb.0:
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 (%rdi), %zmm0
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 64(%rdi), %zmm1
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 128(%rdi), %zmm2
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 192(%rdi), %zmm3
; AVX512DQ-BW-FCP-NEXT: vbroadcasti64x4 {{.*#+}} zmm4 = [0,4,8,12,16,20,24,28,32,36,40,44,48,52,56,60,0,4,8,12,16,20,24,28,32,36,40,44,48,52,56,60]
; AVX512DQ-BW-FCP-NEXT: # zmm4 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm2, %zmm5
; AVX512DQ-BW-FCP-NEXT: vpermt2w %zmm3, %zmm4, %zmm5
; AVX512DQ-BW-FCP-NEXT: vpermi2w %zmm1, %zmm0, %zmm4
; AVX512DQ-BW-FCP-NEXT: vshufi64x2 {{.*#+}} zmm4 = zmm4[0,1,2,3],zmm5[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT: vbroadcasti64x4 {{.*#+}} zmm5 = [1,5,9,13,17,21,25,29,33,37,41,45,49,53,57,61,1,5,9,13,17,21,25,29,33,37,41,45,49,53,57,61]
; AVX512DQ-BW-FCP-NEXT: # zmm5 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm2, %zmm6
; AVX512DQ-BW-FCP-NEXT: vpermt2w %zmm3, %zmm5, %zmm6
; AVX512DQ-BW-FCP-NEXT: vpermi2w %zmm1, %zmm0, %zmm5
; AVX512DQ-BW-FCP-NEXT: vshufi64x2 {{.*#+}} zmm5 = zmm5[0,1,2,3],zmm6[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT: vbroadcasti64x4 {{.*#+}} zmm6 = [2,6,10,14,18,22,26,30,34,38,42,46,50,54,58,62,2,6,10,14,18,22,26,30,34,38,42,46,50,54,58,62]
; AVX512DQ-BW-FCP-NEXT: # zmm6 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm2, %zmm7
; AVX512DQ-BW-FCP-NEXT: vpermt2w %zmm3, %zmm6, %zmm7
; AVX512DQ-BW-FCP-NEXT: vpermi2w %zmm1, %zmm0, %zmm6
; AVX512DQ-BW-FCP-NEXT: vshufi64x2 {{.*#+}} zmm6 = zmm6[0,1,2,3],zmm7[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT: vbroadcasti64x4 {{.*#+}} zmm7 = [3,7,11,15,19,23,27,31,35,39,43,47,51,55,59,63,3,7,11,15,19,23,27,31,35,39,43,47,51,55,59,63]
; AVX512DQ-BW-FCP-NEXT: # zmm7 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-FCP-NEXT: vpermt2w %zmm3, %zmm7, %zmm2
; AVX512DQ-BW-FCP-NEXT: vpermt2w %zmm1, %zmm7, %zmm0
; AVX512DQ-BW-FCP-NEXT: vshufi64x2 {{.*#+}} zmm0 = zmm0[0,1,2,3],zmm2[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm4, (%rsi)
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm5, (%rdx)
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm6, (%rcx)
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm0, (%r8)
; AVX512DQ-BW-FCP-NEXT: vzeroupper
; AVX512DQ-BW-FCP-NEXT: retq
%wide.vec = load <128 x i16>, ptr %in.vec, align 64
%strided.vec0 = shufflevector <128 x i16> %wide.vec, <128 x i16> poison, <32 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28, i32 32, i32 36, i32 40, i32 44, i32 48, i32 52, i32 56, i32 60, i32 64, i32 68, i32 72, i32 76, i32 80, i32 84, i32 88, i32 92, i32 96, i32 100, i32 104, i32 108, i32 112, i32 116, i32 120, i32 124>
%strided.vec1 = shufflevector <128 x i16> %wide.vec, <128 x i16> poison, <32 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29, i32 33, i32 37, i32 41, i32 45, i32 49, i32 53, i32 57, i32 61, i32 65, i32 69, i32 73, i32 77, i32 81, i32 85, i32 89, i32 93, i32 97, i32 101, i32 105, i32 109, i32 113, i32 117, i32 121, i32 125>
%strided.vec2 = shufflevector <128 x i16> %wide.vec, <128 x i16> poison, <32 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30, i32 34, i32 38, i32 42, i32 46, i32 50, i32 54, i32 58, i32 62, i32 66, i32 70, i32 74, i32 78, i32 82, i32 86, i32 90, i32 94, i32 98, i32 102, i32 106, i32 110, i32 114, i32 118, i32 122, i32 126>
%strided.vec3 = shufflevector <128 x i16> %wide.vec, <128 x i16> poison, <32 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31, i32 35, i32 39, i32 43, i32 47, i32 51, i32 55, i32 59, i32 63, i32 67, i32 71, i32 75, i32 79, i32 83, i32 87, i32 91, i32 95, i32 99, i32 103, i32 107, i32 111, i32 115, i32 119, i32 123, i32 127>
store <32 x i16> %strided.vec0, ptr %out.vec0, align 64
store <32 x i16> %strided.vec1, ptr %out.vec1, align 64
store <32 x i16> %strided.vec2, ptr %out.vec2, align 64
store <32 x i16> %strided.vec3, ptr %out.vec3, align 64
ret void
}
define void @load_i16_stride4_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3) nounwind {
; SSE-LABEL: load_i16_stride4_vf64:
; SSE: # %bb.0:
; SSE-NEXT: subq $824, %rsp # imm = 0x338
; SSE-NEXT: movdqa 352(%rdi), %xmm3
; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa 320(%rdi), %xmm4
; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa 336(%rdi), %xmm5
; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa 96(%rdi), %xmm2
; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa 112(%rdi), %xmm6
; SSE-NEXT: movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa 64(%rdi), %xmm1
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa 80(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[0,2,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm6[0,2,2,3]
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,1,0,2,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[0,1,0,2,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm5[0,2,2,3]
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm4[0,2,2,3]
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[0,2,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT: movdqa 368(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,1,0,2,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm3[0,2,2,3]
; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[0,1,0,2,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa (%rdi), %xmm1
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa 16(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[0,2,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT: movdqa 32(%rdi), %xmm2
; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa 48(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,1,0,2,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[0,1,0,2,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa 256(%rdi), %xmm1
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa 272(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[0,2,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT: movdqa 288(%rdi), %xmm2
; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa 304(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,1,0,2,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[0,1,0,2,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa 192(%rdi), %xmm1
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa 208(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, (%rsp) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[0,2,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT: movdqa 224(%rdi), %xmm2
; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa 240(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,1,0,2,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm14 = xmm2[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm14[0,1,0,2,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa 448(%rdi), %xmm1
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa 464(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm13 = xmm0[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm13[0,2,2,3,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm12 = xmm1[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm12[0,2,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT: movdqa 480(%rdi), %xmm2
; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa 496(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm11 = xmm0[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm11[0,1,0,2,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm10 = xmm2[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm10[0,1,0,2,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa 128(%rdi), %xmm1
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa 144(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm9 = xmm0[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm9[0,2,2,3,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm1[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm8[0,2,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT: movdqa 160(%rdi), %xmm2
; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa 176(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm0[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm7[0,1,0,2,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm2[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm6[0,1,0,2,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa 384(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa 400(%rdi), %xmm1
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm1[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm5[0,2,2,3,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm0[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm4[0,2,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT: movdqa 416(%rdi), %xmm2
; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa 432(%rdi), %xmm1
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm1[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm3[0,1,0,2,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm15 = xmm2[0,1,0,2,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm15 = xmm15[0],xmm1[0],xmm15[1],xmm1[1]
; SSE-NEXT: movsd {{.*#+}} xmm15 = xmm0[0],xmm15[1]
; SSE-NEXT: movapd %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw $237, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[1,3,2,3,4,5,6,7]
; SSE-NEXT: pshuflw $237, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[1,3,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT: pshuflw $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[0,1,1,3,4,5,6,7]
; SSE-NEXT: pshuflw $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
; SSE-NEXT: # xmm15 = mem[0,1,1,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm15 = xmm15[0],xmm0[0],xmm15[1],xmm0[1]
; SSE-NEXT: movsd {{.*#+}} xmm15 = xmm1[0],xmm15[1]
; SSE-NEXT: movapd %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw $237, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[1,3,2,3,4,5,6,7]
; SSE-NEXT: pshuflw $237, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[1,3,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT: pshuflw $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[0,1,1,3,4,5,6,7]
; SSE-NEXT: pshuflw $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
; SSE-NEXT: # xmm15 = mem[0,1,1,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm15 = xmm15[0],xmm0[0],xmm15[1],xmm0[1]
; SSE-NEXT: movsd {{.*#+}} xmm15 = xmm1[0],xmm15[1]
; SSE-NEXT: movapd %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw $237, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[1,3,2,3,4,5,6,7]
; SSE-NEXT: pshuflw $237, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[1,3,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT: pshuflw $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[0,1,1,3,4,5,6,7]
; SSE-NEXT: pshuflw $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
; SSE-NEXT: # xmm15 = mem[0,1,1,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm15 = xmm15[0],xmm0[0],xmm15[1],xmm0[1]
; SSE-NEXT: movsd {{.*#+}} xmm15 = xmm1[0],xmm15[1]
; SSE-NEXT: movapd %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw $237, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[1,3,2,3,4,5,6,7]
; SSE-NEXT: pshuflw $237, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[1,3,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT: pshuflw $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[0,1,1,3,4,5,6,7]
; SSE-NEXT: pshuflw $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
; SSE-NEXT: # xmm15 = mem[0,1,1,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm15 = xmm15[0],xmm0[0],xmm15[1],xmm0[1]
; SSE-NEXT: movsd {{.*#+}} xmm15 = xmm1[0],xmm15[1]
; SSE-NEXT: movapd %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw $237, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[1,3,2,3,4,5,6,7]
; SSE-NEXT: pshuflw $237, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[1,3,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT: pshuflw $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[0,1,1,3,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm14 = xmm14[0,1,1,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm14 = xmm14[0],xmm0[0],xmm14[1],xmm0[1]
; SSE-NEXT: movsd {{.*#+}} xmm14 = xmm1[0],xmm14[1]
; SSE-NEXT: movapd %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm13[1,3,2,3,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm12[1,3,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm11[0,1,1,3,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm10 = xmm10[0,1,1,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm10 = xmm10[0],xmm0[0],xmm10[1],xmm0[1]
; SSE-NEXT: movsd {{.*#+}} xmm10 = xmm1[0],xmm10[1]
; SSE-NEXT: movapd %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm9[1,3,2,3,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm8[1,3,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm7[0,1,1,3,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm6 = xmm6[0,1,1,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm6 = xmm6[0],xmm0[0],xmm6[1],xmm0[1]
; SSE-NEXT: movsd {{.*#+}} xmm6 = xmm1[0],xmm6[1]
; SSE-NEXT: movapd %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm5[1,3,2,3,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm4[1,3,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm3[0,1,1,3,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[0,1,1,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[3,1,2,3]
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[3,1,2,3]
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[2,0,2,3,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[2,0,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[3,1,2,3]
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; SSE-NEXT: # xmm2 = mem[3,1,2,3]
; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,1,2,0,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[0,1,2,0,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[3,1,2,3]
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[3,1,2,3]
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[2,0,2,3,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[2,0,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[3,1,2,3]
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; SSE-NEXT: # xmm2 = mem[3,1,2,3]
; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,1,2,0,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[0,1,2,0,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[3,1,2,3]
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[3,1,2,3]
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[2,0,2,3,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm1[2,0,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
; SSE-NEXT: # xmm13 = mem[3,1,2,3]
; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
; SSE-NEXT: # xmm8 = mem[3,1,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm13[0,1,2,0,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm8[0,1,2,0,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[3,1,2,3]
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[3,1,2,3]
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[2,0,2,3,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm1[2,0,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[3,1,2,3]
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[3,1,2,3]
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,1,2,0,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm14 = xmm1[0,1,2,0,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm14 = xmm14[0],xmm0[0],xmm14[1],xmm0[1]
; SSE-NEXT: movsd {{.*#+}} xmm14 = xmm2[0],xmm14[1]
; SSE-NEXT: pshufd $231, (%rsp), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[3,1,2,3]
; SSE-NEXT: movdqa %xmm0, (%rsp) # 16-byte Spill
; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[3,1,2,3]
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[2,0,2,3,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm1[2,0,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
; SSE-NEXT: # xmm15 = mem[3,1,2,3]
; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
; SSE-NEXT: # xmm12 = mem[3,1,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm15[0,1,2,0,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm10 = xmm12[0,1,2,0,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm10 = xmm10[0],xmm0[0],xmm10[1],xmm0[1]
; SSE-NEXT: movsd {{.*#+}} xmm10 = xmm2[0],xmm10[1]
; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[3,1,2,3]
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[3,1,2,3]
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[2,0,2,3,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm1[2,0,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[3,1,2,3]
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[3,1,2,3]
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,1,2,0,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm9 = xmm1[0,1,2,0,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm9 = xmm9[0],xmm0[0],xmm9[1],xmm0[1]
; SSE-NEXT: movsd {{.*#+}} xmm9 = xmm2[0],xmm9[1]
; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
; SSE-NEXT: # xmm7 = mem[3,1,2,3]
; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
; SSE-NEXT: # xmm6 = mem[3,1,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm7[2,0,2,3,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm11 = xmm6[2,0,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm11 = xmm11[0],xmm0[0],xmm11[1],xmm0[1]
; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; SSE-NEXT: # xmm4 = mem[3,1,2,3]
; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; SSE-NEXT: # xmm3 = mem[3,1,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm4[0,1,2,0,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm5 = xmm3[0,1,2,0,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1]
; SSE-NEXT: movsd {{.*#+}} xmm5 = xmm11[0],xmm5[1]
; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[3,1,2,3]
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; SSE-NEXT: # xmm2 = mem[3,1,2,3]
; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm0[2,0,2,3,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm2[2,0,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[3,1,2,3]
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
; SSE-NEXT: # xmm11 = mem[3,1,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[0,1,2,0,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm11[0,1,2,0,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm0[0],xmm2[1]
; SSE-NEXT: pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[3,1,2,3,4,5,6,7]
; SSE-NEXT: pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[3,1,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT: pshuflw {{.*#+}} xmm13 = xmm13[0,1,3,1,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm8[0,1,3,1,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm13[0],xmm0[1],xmm13[1]
; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
; SSE-NEXT: pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[3,1,2,3,4,5,6,7]
; SSE-NEXT: pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
; SSE-NEXT: # xmm13 = mem[3,1,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm13 = xmm13[0],xmm1[0],xmm13[1],xmm1[1]
; SSE-NEXT: pshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[0,1,3,1,4,5,6,7]
; SSE-NEXT: pshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
; SSE-NEXT: # xmm8 = mem[0,1,3,1,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm8 = xmm8[0],xmm1[0],xmm8[1],xmm1[1]
; SSE-NEXT: movsd {{.*#+}} xmm8 = xmm13[0],xmm8[1]
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm7[3,1,2,3,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm6 = xmm6[3,1,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm6 = xmm6[0],xmm1[0],xmm6[1],xmm1[1]
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm4[0,1,3,1,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm13 = xmm3[0,1,3,1,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm13 = xmm13[0],xmm1[0],xmm13[1],xmm1[1]
; SSE-NEXT: movsd {{.*#+}} xmm13 = xmm6[0],xmm13[1]
; SSE-NEXT: pshuflw $231, (%rsp), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[3,1,2,3,4,5,6,7]
; SSE-NEXT: pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; SSE-NEXT: # xmm3 = mem[3,1,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm15[0,1,3,1,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm12 = xmm12[0,1,3,1,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm12 = xmm12[0],xmm1[0],xmm12[1],xmm1[1]
; SSE-NEXT: movsd {{.*#+}} xmm12 = xmm3[0],xmm12[1]
; SSE-NEXT: pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[3,1,2,3,4,5,6,7]
; SSE-NEXT: pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; SSE-NEXT: # xmm3 = mem[3,1,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; SSE-NEXT: pshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[0,1,3,1,4,5,6,7]
; SSE-NEXT: pshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
; SSE-NEXT: # xmm15 = mem[0,1,3,1,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm15 = xmm15[0],xmm1[0],xmm15[1],xmm1[1]
; SSE-NEXT: movsd {{.*#+}} xmm15 = xmm3[0],xmm15[1]
; SSE-NEXT: pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[3,1,2,3,4,5,6,7]
; SSE-NEXT: pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; SSE-NEXT: # xmm4 = mem[3,1,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
; SSE-NEXT: pshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[0,1,3,1,4,5,6,7]
; SSE-NEXT: pshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; SSE-NEXT: # xmm3 = mem[0,1,3,1,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; SSE-NEXT: movsd {{.*#+}} xmm3 = xmm4[0],xmm3[1]
; SSE-NEXT: pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[3,1,2,3,4,5,6,7]
; SSE-NEXT: pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
; SSE-NEXT: # xmm6 = mem[3,1,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm6 = xmm6[0],xmm1[0],xmm6[1],xmm1[1]
; SSE-NEXT: pshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[0,1,3,1,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm4 = xmm11[0,1,3,1,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
; SSE-NEXT: movsd {{.*#+}} xmm4 = xmm6[0],xmm4[1]
; SSE-NEXT: pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[3,1,2,3,4,5,6,7]
; SSE-NEXT: pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
; SSE-NEXT: # xmm6 = mem[3,1,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm6 = xmm6[0],xmm1[0],xmm6[1],xmm1[1]
; SSE-NEXT: pshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[0,1,3,1,4,5,6,7]
; SSE-NEXT: pshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
; SSE-NEXT: # xmm7 = mem[0,1,3,1,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm1[0],xmm7[1],xmm1[1]
; SSE-NEXT: movsd {{.*#+}} xmm7 = xmm6[0],xmm7[1]
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: movaps %xmm1, 96(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: movaps %xmm1, 32(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: movaps %xmm1, 112(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: movaps %xmm1, 48(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: movaps %xmm1, 64(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: movaps %xmm1, (%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: movaps %xmm1, 80(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: movaps %xmm1, 16(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: movaps %xmm1, 96(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: movaps %xmm1, 32(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: movaps %xmm1, 112(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: movaps %xmm1, 48(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: movaps %xmm1, 64(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: movaps %xmm1, (%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: movaps %xmm1, 80(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: movaps %xmm1, 16(%rdx)
; SSE-NEXT: movapd %xmm2, 96(%rcx)
; SSE-NEXT: movapd %xmm5, 32(%rcx)
; SSE-NEXT: movapd %xmm9, 112(%rcx)
; SSE-NEXT: movapd %xmm10, 48(%rcx)
; SSE-NEXT: movapd %xmm14, 64(%rcx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: movaps %xmm1, (%rcx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: movaps %xmm1, 80(%rcx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: movaps %xmm1, 16(%rcx)
; SSE-NEXT: movapd %xmm7, 112(%r8)
; SSE-NEXT: movapd %xmm4, 96(%r8)
; SSE-NEXT: movapd %xmm3, 80(%r8)
; SSE-NEXT: movapd %xmm15, 64(%r8)
; SSE-NEXT: movapd %xmm12, 48(%r8)
; SSE-NEXT: movapd %xmm13, 32(%r8)
; SSE-NEXT: movapd %xmm8, 16(%r8)
; SSE-NEXT: movapd %xmm0, (%r8)
; SSE-NEXT: addq $824, %rsp # imm = 0x338
; SSE-NEXT: retq
;
; AVX-LABEL: load_i16_stride4_vf64:
; AVX: # %bb.0:
; AVX-NEXT: subq $776, %rsp # imm = 0x308
; AVX-NEXT: vpxor %xmm10, %xmm10, %xmm10
; AVX-NEXT: vmovdqa 368(%rdi), %xmm8
; AVX-NEXT: vpblendw {{.*#+}} xmm0 = xmm8[0],xmm10[1,2,3],xmm8[4],xmm10[5,6,7]
; AVX-NEXT: vmovdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vmovdqa 352(%rdi), %xmm1
; AVX-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0],xmm10[1,2,3],xmm1[4],xmm10[5,6,7]
; AVX-NEXT: vpackusdw %xmm0, %xmm1, %xmm0
; AVX-NEXT: vmovdqa 336(%rdi), %xmm4
; AVX-NEXT: vpblendw {{.*#+}} xmm1 = xmm4[0],xmm10[1,2,3],xmm4[4],xmm10[5,6,7]
; AVX-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vmovdqa 320(%rdi), %xmm7
; AVX-NEXT: vpblendw {{.*#+}} xmm2 = xmm7[0],xmm10[1,2,3],xmm7[4],xmm10[5,6,7]
; AVX-NEXT: vmovdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vpackusdw %xmm1, %xmm2, %xmm1
; AVX-NEXT: vpackusdw %xmm0, %xmm1, %xmm0
; AVX-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vmovdqa 304(%rdi), %xmm13
; AVX-NEXT: vpblendw {{.*#+}} xmm0 = xmm13[0],xmm10[1,2,3],xmm13[4],xmm10[5,6,7]
; AVX-NEXT: vmovdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vmovdqa 288(%rdi), %xmm1
; AVX-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0],xmm10[1,2,3],xmm1[4],xmm10[5,6,7]
; AVX-NEXT: vpackusdw %xmm0, %xmm1, %xmm0
; AVX-NEXT: vmovdqa 272(%rdi), %xmm14
; AVX-NEXT: vpblendw {{.*#+}} xmm1 = xmm14[0],xmm10[1,2,3],xmm14[4],xmm10[5,6,7]
; AVX-NEXT: vmovdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vmovdqa 256(%rdi), %xmm2
; AVX-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0],xmm10[1,2,3],xmm2[4],xmm10[5,6,7]
; AVX-NEXT: vpackusdw %xmm1, %xmm2, %xmm1
; AVX-NEXT: vpackusdw %xmm0, %xmm1, %xmm0
; AVX-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vmovdqa 112(%rdi), %xmm0
; AVX-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vpblendw {{.*#+}} xmm2 = xmm0[0],xmm10[1,2,3],xmm0[4],xmm10[5,6,7]
; AVX-NEXT: vmovdqa 96(%rdi), %xmm0
; AVX-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vpblendw {{.*#+}} xmm3 = xmm0[0],xmm10[1,2,3],xmm0[4],xmm10[5,6,7]
; AVX-NEXT: vpackusdw %xmm2, %xmm3, %xmm2
; AVX-NEXT: vmovdqa 80(%rdi), %xmm0
; AVX-NEXT: vpblendw {{.*#+}} xmm5 = xmm0[0],xmm10[1,2,3],xmm0[4],xmm10[5,6,7]
; AVX-NEXT: vmovdqa %xmm0, %xmm3
; AVX-NEXT: vmovdqa 64(%rdi), %xmm0
; AVX-NEXT: vmovdqa %xmm0, (%rsp) # 16-byte Spill
; AVX-NEXT: vpblendw {{.*#+}} xmm6 = xmm0[0],xmm10[1,2,3],xmm0[4],xmm10[5,6,7]
; AVX-NEXT: vpackusdw %xmm5, %xmm6, %xmm5
; AVX-NEXT: vpackusdw %xmm2, %xmm5, %xmm2
; AVX-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vmovdqa 32(%rdi), %xmm0
; AVX-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vmovdqa 48(%rdi), %xmm1
; AVX-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vpblendw {{.*#+}} xmm5 = xmm1[0],xmm10[1,2,3],xmm1[4],xmm10[5,6,7]
; AVX-NEXT: vpblendw {{.*#+}} xmm6 = xmm0[0],xmm10[1,2,3],xmm0[4],xmm10[5,6,7]
; AVX-NEXT: vpackusdw %xmm5, %xmm6, %xmm5
; AVX-NEXT: vmovdqa (%rdi), %xmm0
; AVX-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vmovdqa 16(%rdi), %xmm1
; AVX-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vpblendw {{.*#+}} xmm9 = xmm1[0],xmm10[1,2,3],xmm1[4],xmm10[5,6,7]
; AVX-NEXT: vpblendw {{.*#+}} xmm11 = xmm0[0],xmm10[1,2,3],xmm0[4],xmm10[5,6,7]
; AVX-NEXT: vpackusdw %xmm9, %xmm11, %xmm9
; AVX-NEXT: vpackusdw %xmm5, %xmm9, %xmm5
; AVX-NEXT: vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vmovdqa 240(%rdi), %xmm0
; AVX-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vpblendw {{.*#+}} xmm5 = xmm0[0],xmm10[1,2,3],xmm0[4],xmm10[5,6,7]
; AVX-NEXT: vmovdqa 224(%rdi), %xmm15
; AVX-NEXT: vpblendw {{.*#+}} xmm9 = xmm15[0],xmm10[1,2,3],xmm15[4],xmm10[5,6,7]
; AVX-NEXT: vmovdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vpackusdw %xmm5, %xmm9, %xmm5
; AVX-NEXT: vmovdqa 208(%rdi), %xmm0
; AVX-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vpblendw {{.*#+}} xmm9 = xmm0[0],xmm10[1,2,3],xmm0[4],xmm10[5,6,7]
; AVX-NEXT: vmovdqa 192(%rdi), %xmm0
; AVX-NEXT: vpblendw {{.*#+}} xmm11 = xmm0[0],xmm10[1,2,3],xmm0[4],xmm10[5,6,7]
; AVX-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vpackusdw %xmm9, %xmm11, %xmm9
; AVX-NEXT: vpackusdw %xmm5, %xmm9, %xmm5
; AVX-NEXT: vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vmovdqa 176(%rdi), %xmm2
; AVX-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vpblendw {{.*#+}} xmm5 = xmm2[0],xmm10[1,2,3],xmm2[4],xmm10[5,6,7]
; AVX-NEXT: vmovdqa 160(%rdi), %xmm2
; AVX-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vpblendw {{.*#+}} xmm9 = xmm2[0],xmm10[1,2,3],xmm2[4],xmm10[5,6,7]
; AVX-NEXT: vpackusdw %xmm5, %xmm9, %xmm5
; AVX-NEXT: vmovdqa 144(%rdi), %xmm9
; AVX-NEXT: vmovdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vpblendw {{.*#+}} xmm9 = xmm9[0],xmm10[1,2,3],xmm9[4],xmm10[5,6,7]
; AVX-NEXT: vmovdqa 128(%rdi), %xmm11
; AVX-NEXT: vmovdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vpblendw {{.*#+}} xmm11 = xmm11[0],xmm10[1,2,3],xmm11[4],xmm10[5,6,7]
; AVX-NEXT: vpackusdw %xmm9, %xmm11, %xmm9
; AVX-NEXT: vpackusdw %xmm5, %xmm9, %xmm5
; AVX-NEXT: vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vmovdqa 496(%rdi), %xmm1
; AVX-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vpblendw {{.*#+}} xmm9 = xmm1[0],xmm10[1,2,3],xmm1[4],xmm10[5,6,7]
; AVX-NEXT: vmovdqa 480(%rdi), %xmm1
; AVX-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vpblendw {{.*#+}} xmm11 = xmm1[0],xmm10[1,2,3],xmm1[4],xmm10[5,6,7]
; AVX-NEXT: vpackusdw %xmm9, %xmm11, %xmm9
; AVX-NEXT: vmovdqa 464(%rdi), %xmm1
; AVX-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vpblendw {{.*#+}} xmm11 = xmm1[0],xmm10[1,2,3],xmm1[4],xmm10[5,6,7]
; AVX-NEXT: vmovdqa 448(%rdi), %xmm1
; AVX-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vpblendw {{.*#+}} xmm12 = xmm1[0],xmm10[1,2,3],xmm1[4],xmm10[5,6,7]
; AVX-NEXT: vpackusdw %xmm11, %xmm12, %xmm11
; AVX-NEXT: vpackusdw %xmm9, %xmm11, %xmm9
; AVX-NEXT: vmovdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vmovdqa 432(%rdi), %xmm2
; AVX-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vpblendw {{.*#+}} xmm9 = xmm2[0],xmm10[1,2,3],xmm2[4],xmm10[5,6,7]
; AVX-NEXT: vmovdqa 416(%rdi), %xmm1
; AVX-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vpblendw {{.*#+}} xmm11 = xmm1[0],xmm10[1,2,3],xmm1[4],xmm10[5,6,7]
; AVX-NEXT: vpackusdw %xmm9, %xmm11, %xmm9
; AVX-NEXT: vmovdqa 400(%rdi), %xmm5
; AVX-NEXT: vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vpblendw {{.*#+}} xmm11 = xmm5[0],xmm10[1,2,3],xmm5[4],xmm10[5,6,7]
; AVX-NEXT: vmovdqa 384(%rdi), %xmm2
; AVX-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vpblendw {{.*#+}} xmm10 = xmm2[0],xmm10[1,2,3],xmm2[4],xmm10[5,6,7]
; AVX-NEXT: vpackusdw %xmm11, %xmm10, %xmm10
; AVX-NEXT: vpackusdw %xmm9, %xmm10, %xmm9
; AVX-NEXT: vmovdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vpshufd {{.*#+}} xmm9 = xmm8[0,2,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm9 = xmm9[0,1,1,3,4,5,6,7]
; AVX-NEXT: vpshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
; AVX-NEXT: # xmm10 = mem[0,2,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm10 = xmm10[0,1,1,3,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm9 = xmm10[0],xmm9[0],xmm10[1],xmm9[1]
; AVX-NEXT: vpshufd {{.*#+}} xmm10 = xmm4[0,2,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm10 = xmm10[1,3,2,3,4,5,6,7]
; AVX-NEXT: vpshufd {{.*#+}} xmm11 = xmm7[0,2,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm11 = xmm11[1,3,2,3,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm10 = xmm11[0],xmm10[0],xmm11[1],xmm10[1]
; AVX-NEXT: vpblendw {{.*#+}} xmm9 = xmm10[0,1,2,3],xmm9[4,5,6,7]
; AVX-NEXT: vpshufd {{.*#+}} xmm10 = xmm13[0,2,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm10 = xmm10[0,1,1,3,4,5,6,7]
; AVX-NEXT: vpshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
; AVX-NEXT: # xmm11 = mem[0,2,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm11 = xmm11[0,1,1,3,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm10 = xmm11[0],xmm10[0],xmm11[1],xmm10[1]
; AVX-NEXT: vpshufd {{.*#+}} xmm11 = xmm14[0,2,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm11 = xmm11[1,3,2,3,4,5,6,7]
; AVX-NEXT: vpshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
; AVX-NEXT: # xmm12 = mem[0,2,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm12 = xmm12[1,3,2,3,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm11 = xmm12[0],xmm11[0],xmm12[1],xmm11[1]
; AVX-NEXT: vpblendw {{.*#+}} xmm10 = xmm11[0,1,2,3],xmm10[4,5,6,7]
; AVX-NEXT: vinsertf128 $1, %xmm9, %ymm0, %ymm9
; AVX-NEXT: vblendps {{.*#+}} ymm9 = ymm10[0,1,2,3],ymm9[4,5,6,7]
; AVX-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX-NEXT: vpshufd {{.*#+}} xmm9 = xmm1[0,2,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm9 = xmm9[0,1,1,3,4,5,6,7]
; AVX-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; AVX-NEXT: vpshufd {{.*#+}} xmm10 = xmm2[0,2,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm10 = xmm10[0,1,1,3,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm9 = xmm10[0],xmm9[0],xmm10[1],xmm9[1]
; AVX-NEXT: vmovdqa %xmm3, %xmm4
; AVX-NEXT: vpshufd {{.*#+}} xmm10 = xmm3[0,2,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm10 = xmm10[1,3,2,3,4,5,6,7]
; AVX-NEXT: vmovdqa (%rsp), %xmm3 # 16-byte Reload
; AVX-NEXT: vpshufd {{.*#+}} xmm11 = xmm3[0,2,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm11 = xmm11[1,3,2,3,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm10 = xmm11[0],xmm10[0],xmm11[1],xmm10[1]
; AVX-NEXT: vpblendw {{.*#+}} xmm9 = xmm10[0,1,2,3],xmm9[4,5,6,7]
; AVX-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
; AVX-NEXT: vpshufd {{.*#+}} xmm10 = xmm6[0,2,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm10 = xmm10[0,1,1,3,4,5,6,7]
; AVX-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
; AVX-NEXT: vpshufd {{.*#+}} xmm11 = xmm5[0,2,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm11 = xmm11[0,1,1,3,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm10 = xmm11[0],xmm10[0],xmm11[1],xmm10[1]
; AVX-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; AVX-NEXT: vpshufd {{.*#+}} xmm11 = xmm8[0,2,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm11 = xmm11[1,3,2,3,4,5,6,7]
; AVX-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
; AVX-NEXT: vpshufd {{.*#+}} xmm12 = xmm7[0,2,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm12 = xmm12[1,3,2,3,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm11 = xmm12[0],xmm11[0],xmm12[1],xmm11[1]
; AVX-NEXT: vpblendw {{.*#+}} xmm10 = xmm11[0,1,2,3],xmm10[4,5,6,7]
; AVX-NEXT: vinsertf128 $1, %xmm9, %ymm0, %ymm9
; AVX-NEXT: vblendps {{.*#+}} ymm9 = ymm10[0,1,2,3],ymm9[4,5,6,7]
; AVX-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT: vpshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
; AVX-NEXT: # xmm9 = mem[0,2,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm9 = xmm9[0,1,1,3,4,5,6,7]
; AVX-NEXT: vpshufd {{.*#+}} xmm10 = xmm15[0,2,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm10 = xmm10[0,1,1,3,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm9 = xmm10[0],xmm9[0],xmm10[1],xmm9[1]
; AVX-NEXT: vpshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
; AVX-NEXT: # xmm10 = mem[0,2,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm10 = xmm10[1,3,2,3,4,5,6,7]
; AVX-NEXT: vpshufd {{.*#+}} xmm11 = xmm0[0,2,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm11 = xmm11[1,3,2,3,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm10 = xmm11[0],xmm10[0],xmm11[1],xmm10[1]
; AVX-NEXT: vpblendw {{.*#+}} xmm9 = xmm10[0,1,2,3],xmm9[4,5,6,7]
; AVX-NEXT: vpshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
; AVX-NEXT: # xmm10 = mem[0,2,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm10 = xmm10[0,1,1,3,4,5,6,7]
; AVX-NEXT: vpshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
; AVX-NEXT: # xmm11 = mem[0,2,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm11 = xmm11[0,1,1,3,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm10 = xmm11[0],xmm10[0],xmm11[1],xmm10[1]
; AVX-NEXT: vpshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
; AVX-NEXT: # xmm11 = mem[0,2,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm11 = xmm11[1,3,2,3,4,5,6,7]
; AVX-NEXT: vpshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
; AVX-NEXT: # xmm12 = mem[0,2,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm12 = xmm12[1,3,2,3,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm11 = xmm12[0],xmm11[0],xmm12[1],xmm11[1]
; AVX-NEXT: vpblendw {{.*#+}} xmm10 = xmm11[0,1,2,3],xmm10[4,5,6,7]
; AVX-NEXT: vinsertf128 $1, %xmm9, %ymm0, %ymm9
; AVX-NEXT: vblendps {{.*#+}} ymm0 = ymm10[0,1,2,3],ymm9[4,5,6,7]
; AVX-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
; AVX-NEXT: vpshufd {{.*#+}} xmm9 = xmm13[0,2,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm9 = xmm9[0,1,1,3,4,5,6,7]
; AVX-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
; AVX-NEXT: vpshufd {{.*#+}} xmm10 = xmm14[0,2,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm10 = xmm10[0,1,1,3,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm9 = xmm10[0],xmm9[0],xmm10[1],xmm9[1]
; AVX-NEXT: vpshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
; AVX-NEXT: # xmm10 = mem[0,2,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm10 = xmm10[1,3,2,3,4,5,6,7]
; AVX-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; AVX-NEXT: vpshufd {{.*#+}} xmm11 = xmm15[0,2,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm11 = xmm11[1,3,2,3,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm10 = xmm11[0],xmm10[0],xmm11[1],xmm10[1]
; AVX-NEXT: vpblendw {{.*#+}} xmm9 = xmm10[0,1,2,3],xmm9[4,5,6,7]
; AVX-NEXT: vpshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
; AVX-NEXT: # xmm10 = mem[0,2,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm10 = xmm10[0,1,1,3,4,5,6,7]
; AVX-NEXT: vpshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
; AVX-NEXT: # xmm11 = mem[0,2,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm11 = xmm11[0,1,1,3,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm10 = xmm11[0],xmm10[0],xmm11[1],xmm10[1]
; AVX-NEXT: vpshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
; AVX-NEXT: # xmm11 = mem[0,2,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm11 = xmm11[1,3,2,3,4,5,6,7]
; AVX-NEXT: vpshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
; AVX-NEXT: # xmm12 = mem[0,2,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm12 = xmm12[1,3,2,3,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm11 = xmm12[0],xmm11[0],xmm12[1],xmm11[1]
; AVX-NEXT: vpblendw {{.*#+}} xmm10 = xmm11[0,1,2,3],xmm10[4,5,6,7]
; AVX-NEXT: vinsertf128 $1, %xmm9, %ymm0, %ymm9
; AVX-NEXT: vblendps {{.*#+}} ymm0 = ymm10[0,1,2,3],ymm9[4,5,6,7]
; AVX-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT: vpshufd {{.*#+}} xmm0 = xmm1[3,1,2,3]
; AVX-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vpshufd {{.*#+}} xmm1 = xmm2[3,1,2,3]
; AVX-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[0,1,2,0,4,5,6,7]
; AVX-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[0,1,2,0,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX-NEXT: vpshufd {{.*#+}} xmm1 = xmm4[3,1,2,3]
; AVX-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vpshufd {{.*#+}} xmm2 = xmm3[3,1,2,3]
; AVX-NEXT: vmovdqa %xmm2, (%rsp) # 16-byte Spill
; AVX-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[2,0,2,3,4,5,6,7]
; AVX-NEXT: vpshuflw {{.*#+}} xmm3 = xmm2[2,0,2,3,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; AVX-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3],xmm0[4,5,6,7]
; AVX-NEXT: vpshufd {{.*#+}} xmm1 = xmm6[3,1,2,3]
; AVX-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vpshufd {{.*#+}} xmm2 = xmm5[3,1,2,3]
; AVX-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[0,1,2,0,4,5,6,7]
; AVX-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[0,1,2,0,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; AVX-NEXT: vpshufd {{.*#+}} xmm2 = xmm8[3,1,2,3]
; AVX-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vpshufd {{.*#+}} xmm3 = xmm7[3,1,2,3]
; AVX-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[2,0,2,3,4,5,6,7]
; AVX-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[2,0,2,3,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
; AVX-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0,1,2,3],xmm1[4,5,6,7]
; AVX-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
; AVX-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; AVX-NEXT: # xmm0 = mem[3,1,2,3]
; AVX-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; AVX-NEXT: # xmm1 = mem[3,1,2,3]
; AVX-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[0,1,2,0,4,5,6,7]
; AVX-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[0,1,2,0,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; AVX-NEXT: # xmm1 = mem[3,1,2,3]
; AVX-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; AVX-NEXT: # xmm2 = mem[3,1,2,3]
; AVX-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[2,0,2,3,4,5,6,7]
; AVX-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[2,0,2,3,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; AVX-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3],xmm0[4,5,6,7]
; AVX-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; AVX-NEXT: # xmm1 = mem[3,1,2,3]
; AVX-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; AVX-NEXT: # xmm2 = mem[3,1,2,3]
; AVX-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[0,1,2,0,4,5,6,7]
; AVX-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[0,1,2,0,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; AVX-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
; AVX-NEXT: # xmm12 = mem[3,1,2,3]
; AVX-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
; AVX-NEXT: # xmm11 = mem[3,1,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm2 = xmm12[2,0,2,3,4,5,6,7]
; AVX-NEXT: vpshuflw {{.*#+}} xmm3 = xmm11[2,0,2,3,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
; AVX-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0,1,2,3],xmm1[4,5,6,7]
; AVX-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
; AVX-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT: vpshufd {{.*#+}} xmm0 = xmm13[3,1,2,3]
; AVX-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vpshufd {{.*#+}} xmm2 = xmm14[3,1,2,3]
; AVX-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[0,1,2,0,4,5,6,7]
; AVX-NEXT: vpshuflw {{.*#+}} xmm1 = xmm2[0,1,2,0,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; AVX-NEXT: # xmm1 = mem[3,1,2,3]
; AVX-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vpshufd {{.*#+}} xmm2 = xmm15[3,1,2,3]
; AVX-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[2,0,2,3,4,5,6,7]
; AVX-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[2,0,2,3,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; AVX-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3],xmm0[4,5,6,7]
; AVX-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; AVX-NEXT: # xmm1 = mem[3,1,2,3]
; AVX-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; AVX-NEXT: # xmm2 = mem[3,1,2,3]
; AVX-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[0,1,2,0,4,5,6,7]
; AVX-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[0,1,2,0,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; AVX-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
; AVX-NEXT: # xmm13 = mem[3,1,2,3]
; AVX-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
; AVX-NEXT: # xmm14 = mem[3,1,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm2 = xmm13[2,0,2,3,4,5,6,7]
; AVX-NEXT: vpshuflw {{.*#+}} xmm3 = xmm14[2,0,2,3,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
; AVX-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0,1,2,3],xmm1[4,5,6,7]
; AVX-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
; AVX-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
; AVX-NEXT: # xmm10 = mem[3,1,2,3]
; AVX-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
; AVX-NEXT: # xmm9 = mem[3,1,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm0 = xmm10[0,1,2,0,4,5,6,7]
; AVX-NEXT: vpshuflw {{.*#+}} xmm1 = xmm9[0,1,2,0,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
; AVX-NEXT: # xmm8 = mem[3,1,2,3]
; AVX-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
; AVX-NEXT: # xmm7 = mem[3,1,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm1 = xmm8[2,0,2,3,4,5,6,7]
; AVX-NEXT: vpshuflw {{.*#+}} xmm2 = xmm7[2,0,2,3,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; AVX-NEXT: vpblendw {{.*#+}} xmm3 = xmm1[0,1,2,3],xmm0[4,5,6,7]
; AVX-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
; AVX-NEXT: # xmm6 = mem[3,1,2,3]
; AVX-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
; AVX-NEXT: # xmm5 = mem[3,1,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm1 = xmm6[0,1,2,0,4,5,6,7]
; AVX-NEXT: vpshuflw {{.*#+}} xmm2 = xmm5[0,1,2,0,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; AVX-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; AVX-NEXT: # xmm4 = mem[3,1,2,3]
; AVX-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; AVX-NEXT: # xmm2 = mem[3,1,2,3]
; AVX-NEXT: vpshuflw {{.*#+}} xmm0 = xmm4[2,0,2,3,4,5,6,7]
; AVX-NEXT: vpshuflw {{.*#+}} xmm15 = xmm2[2,0,2,3,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm15[0],xmm0[0],xmm15[1],xmm0[1]
; AVX-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm1[4,5,6,7]
; AVX-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm1
; AVX-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT: vpshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; AVX-NEXT: # xmm0 = mem[0,1,3,1,4,5,6,7]
; AVX-NEXT: vpshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; AVX-NEXT: # xmm1 = mem[0,1,3,1,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX-NEXT: vpshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; AVX-NEXT: # xmm1 = mem[3,1,2,3,4,5,6,7]
; AVX-NEXT: vpshuflw $231, (%rsp), %xmm3 # 16-byte Folded Reload
; AVX-NEXT: # xmm3 = mem[3,1,2,3,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; AVX-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3],xmm0[4,5,6,7]
; AVX-NEXT: vpshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; AVX-NEXT: # xmm1 = mem[0,1,3,1,4,5,6,7]
; AVX-NEXT: vpshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; AVX-NEXT: # xmm3 = mem[0,1,3,1,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; AVX-NEXT: vpshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; AVX-NEXT: # xmm3 = mem[3,1,2,3,4,5,6,7]
; AVX-NEXT: vpshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
; AVX-NEXT: # xmm15 = mem[3,1,2,3,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm15[0],xmm3[0],xmm15[1],xmm3[1]
; AVX-NEXT: vpblendw {{.*#+}} xmm1 = xmm3[0,1,2,3],xmm1[4,5,6,7]
; AVX-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
; AVX-NEXT: vblendps {{.*#+}} ymm15 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX-NEXT: vpshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; AVX-NEXT: # xmm0 = mem[0,1,3,1,4,5,6,7]
; AVX-NEXT: vpshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; AVX-NEXT: # xmm1 = mem[0,1,3,1,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX-NEXT: vpshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; AVX-NEXT: # xmm1 = mem[3,1,2,3,4,5,6,7]
; AVX-NEXT: vpshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; AVX-NEXT: # xmm3 = mem[3,1,2,3,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; AVX-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3],xmm0[4,5,6,7]
; AVX-NEXT: vpshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; AVX-NEXT: # xmm1 = mem[0,1,3,1,4,5,6,7]
; AVX-NEXT: vpshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; AVX-NEXT: # xmm3 = mem[0,1,3,1,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; AVX-NEXT: vpshuflw {{.*#+}} xmm3 = xmm12[3,1,2,3,4,5,6,7]
; AVX-NEXT: vpshuflw {{.*#+}} xmm11 = xmm11[3,1,2,3,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm11[0],xmm3[0],xmm11[1],xmm3[1]
; AVX-NEXT: vpblendw {{.*#+}} xmm1 = xmm3[0,1,2,3],xmm1[4,5,6,7]
; AVX-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
; AVX-NEXT: vblendps {{.*#+}} ymm3 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX-NEXT: vpshuflw {{.*#+}} xmm0 = xmm10[0,1,3,1,4,5,6,7]
; AVX-NEXT: vpshuflw {{.*#+}} xmm1 = xmm9[0,1,3,1,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX-NEXT: vpshuflw {{.*#+}} xmm1 = xmm8[3,1,2,3,4,5,6,7]
; AVX-NEXT: vpshuflw {{.*#+}} xmm7 = xmm7[3,1,2,3,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm7[0],xmm1[0],xmm7[1],xmm1[1]
; AVX-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3],xmm0[4,5,6,7]
; AVX-NEXT: vpshuflw {{.*#+}} xmm1 = xmm6[0,1,3,1,4,5,6,7]
; AVX-NEXT: vpshuflw {{.*#+}} xmm5 = xmm5[0,1,3,1,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm5[0],xmm1[0],xmm5[1],xmm1[1]
; AVX-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[3,1,2,3,4,5,6,7]
; AVX-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[3,1,2,3,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1]
; AVX-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0,1,2,3],xmm1[4,5,6,7]
; AVX-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
; AVX-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX-NEXT: vpshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; AVX-NEXT: # xmm1 = mem[0,1,3,1,4,5,6,7]
; AVX-NEXT: vpshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; AVX-NEXT: # xmm2 = mem[0,1,3,1,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; AVX-NEXT: vpshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; AVX-NEXT: # xmm2 = mem[3,1,2,3,4,5,6,7]
; AVX-NEXT: vpshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; AVX-NEXT: # xmm4 = mem[3,1,2,3,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm4[0],xmm2[0],xmm4[1],xmm2[1]
; AVX-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0,1,2,3],xmm1[4,5,6,7]
; AVX-NEXT: vpshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; AVX-NEXT: # xmm2 = mem[0,1,3,1,4,5,6,7]
; AVX-NEXT: vpshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; AVX-NEXT: # xmm4 = mem[0,1,3,1,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm4[0],xmm2[0],xmm4[1],xmm2[1]
; AVX-NEXT: vpshuflw {{.*#+}} xmm4 = xmm13[3,1,2,3,4,5,6,7]
; AVX-NEXT: vpshuflw {{.*#+}} xmm5 = xmm14[3,1,2,3,4,5,6,7]
; AVX-NEXT: vpunpckldq {{.*#+}} xmm4 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
; AVX-NEXT: vpblendw {{.*#+}} xmm2 = xmm4[0,1,2,3],xmm2[4,5,6,7]
; AVX-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; AVX-NEXT: vmovaps %xmm2, 96(%rsi)
; AVX-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; AVX-NEXT: vmovaps %xmm2, 112(%rsi)
; AVX-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; AVX-NEXT: vmovaps %xmm2, 32(%rsi)
; AVX-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; AVX-NEXT: vmovaps %xmm2, 48(%rsi)
; AVX-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; AVX-NEXT: vmovaps %xmm2, (%rsi)
; AVX-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; AVX-NEXT: vmovaps %xmm2, 16(%rsi)
; AVX-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; AVX-NEXT: vmovaps %xmm2, 64(%rsi)
; AVX-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; AVX-NEXT: vmovaps %xmm2, 80(%rsi)
; AVX-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX-NEXT: vmovaps %ymm2, 96(%rdx)
; AVX-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX-NEXT: vmovaps %ymm2, 32(%rdx)
; AVX-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX-NEXT: vmovaps %ymm2, (%rdx)
; AVX-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX-NEXT: vmovaps %ymm2, 64(%rdx)
; AVX-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX-NEXT: vmovaps %ymm2, 32(%rcx)
; AVX-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX-NEXT: vmovaps %ymm2, 96(%rcx)
; AVX-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX-NEXT: vmovaps %ymm2, 64(%rcx)
; AVX-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX-NEXT: vmovaps %ymm2, (%rcx)
; AVX-NEXT: vmovaps %ymm1, 96(%r8)
; AVX-NEXT: vmovaps %ymm0, 32(%r8)
; AVX-NEXT: vmovaps %ymm3, 64(%r8)
; AVX-NEXT: vmovaps %ymm15, (%r8)
; AVX-NEXT: addq $776, %rsp # imm = 0x308
; AVX-NEXT: vzeroupper
; AVX-NEXT: retq
;
; AVX2-LABEL: load_i16_stride4_vf64:
; AVX2: # %bb.0:
; AVX2-NEXT: subq $696, %rsp # imm = 0x2B8
; AVX2-NEXT: vpxor %xmm0, %xmm0, %xmm0
; AVX2-NEXT: vpblendw {{.*#+}} ymm1 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-NEXT: vextracti128 $1, %ymm1, %xmm2
; AVX2-NEXT: vpackusdw %xmm2, %xmm1, %xmm1
; AVX2-NEXT: vpblendw {{.*#+}} ymm2 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-NEXT: vextracti128 $1, %ymm2, %xmm3
; AVX2-NEXT: vpackusdw %xmm3, %xmm2, %xmm2
; AVX2-NEXT: vpackusdw %xmm1, %xmm2, %xmm1
; AVX2-NEXT: vpblendw {{.*#+}} ymm2 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-NEXT: vextracti128 $1, %ymm2, %xmm3
; AVX2-NEXT: vpackusdw %xmm3, %xmm2, %xmm2
; AVX2-NEXT: vpackusdw %xmm2, %xmm2, %xmm2
; AVX2-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX2-NEXT: vpblendw {{.*#+}} ymm3 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-NEXT: vextracti128 $1, %ymm3, %xmm4
; AVX2-NEXT: vpackusdw %xmm4, %xmm3, %xmm3
; AVX2-NEXT: vpackusdw %xmm3, %xmm3, %xmm3
; AVX2-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
; AVX2-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
; AVX2-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX2-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT: vpblendw {{.*#+}} ymm1 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-NEXT: vextracti128 $1, %ymm1, %xmm2
; AVX2-NEXT: vpackusdw %xmm2, %xmm1, %xmm1
; AVX2-NEXT: vpblendw {{.*#+}} ymm2 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-NEXT: vextracti128 $1, %ymm2, %xmm3
; AVX2-NEXT: vpackusdw %xmm3, %xmm2, %xmm2
; AVX2-NEXT: vpackusdw %xmm1, %xmm2, %xmm1
; AVX2-NEXT: vpblendw {{.*#+}} ymm2 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-NEXT: vextracti128 $1, %ymm2, %xmm3
; AVX2-NEXT: vpackusdw %xmm3, %xmm2, %xmm2
; AVX2-NEXT: vpackusdw %xmm2, %xmm2, %xmm2
; AVX2-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX2-NEXT: vpblendw {{.*#+}} ymm3 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-NEXT: vextracti128 $1, %ymm3, %xmm4
; AVX2-NEXT: vpackusdw %xmm4, %xmm3, %xmm3
; AVX2-NEXT: vpackusdw %xmm3, %xmm3, %xmm3
; AVX2-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
; AVX2-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
; AVX2-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX2-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT: vpblendw {{.*#+}} ymm1 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-NEXT: vextracti128 $1, %ymm1, %xmm2
; AVX2-NEXT: vpackusdw %xmm2, %xmm1, %xmm1
; AVX2-NEXT: vpblendw {{.*#+}} ymm2 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-NEXT: vextracti128 $1, %ymm2, %xmm3
; AVX2-NEXT: vpackusdw %xmm3, %xmm2, %xmm2
; AVX2-NEXT: vpackusdw %xmm1, %xmm2, %xmm1
; AVX2-NEXT: vpblendw {{.*#+}} ymm2 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-NEXT: vextracti128 $1, %ymm2, %xmm3
; AVX2-NEXT: vpackusdw %xmm3, %xmm2, %xmm2
; AVX2-NEXT: vpackusdw %xmm2, %xmm2, %xmm2
; AVX2-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX2-NEXT: vpblendw {{.*#+}} ymm3 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-NEXT: vextracti128 $1, %ymm3, %xmm4
; AVX2-NEXT: vpackusdw %xmm4, %xmm3, %xmm3
; AVX2-NEXT: vpackusdw %xmm3, %xmm3, %xmm3
; AVX2-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
; AVX2-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
; AVX2-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX2-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT: vpblendw {{.*#+}} ymm1 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-NEXT: vextracti128 $1, %ymm1, %xmm2
; AVX2-NEXT: vpackusdw %xmm2, %xmm1, %xmm1
; AVX2-NEXT: vpblendw {{.*#+}} ymm2 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-NEXT: vextracti128 $1, %ymm2, %xmm3
; AVX2-NEXT: vpackusdw %xmm3, %xmm2, %xmm2
; AVX2-NEXT: vpackusdw %xmm1, %xmm2, %xmm1
; AVX2-NEXT: vpblendw {{.*#+}} ymm2 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-NEXT: vextracti128 $1, %ymm2, %xmm3
; AVX2-NEXT: vpackusdw %xmm3, %xmm2, %xmm2
; AVX2-NEXT: vpackusdw %xmm2, %xmm2, %xmm2
; AVX2-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX2-NEXT: vpblendw {{.*#+}} ymm0 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm3
; AVX2-NEXT: vpackusdw %xmm3, %xmm0, %xmm0
; AVX2-NEXT: vpackusdw %xmm0, %xmm0, %xmm0
; AVX2-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT: vmovdqa 256(%rdi), %xmm4
; AVX2-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-NEXT: vmovdqa 272(%rdi), %xmm3
; AVX2-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-NEXT: vmovdqa 288(%rdi), %xmm5
; AVX2-NEXT: vmovdqa %xmm5, (%rsp) # 16-byte Spill
; AVX2-NEXT: vmovdqa 304(%rdi), %xmm6
; AVX2-NEXT: vmovdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-NEXT: vmovdqa 368(%rdi), %xmm0
; AVX2-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[0,1,1,3,4,5,6,7]
; AVX2-NEXT: vmovdqa 352(%rdi), %xmm1
; AVX2-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[0,1,1,3,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX2-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-NEXT: vmovdqa 336(%rdi), %xmm1
; AVX2-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[1,3,2,3,4,5,6,7]
; AVX2-NEXT: vmovdqa 320(%rdi), %xmm2
; AVX2-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[1,3,2,3,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm6[0,2,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[0,1,1,3,4,5,6,7]
; AVX2-NEXT: vpshufd {{.*#+}} xmm2 = xmm5[0,2,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[0,1,1,3,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; AVX2-NEXT: vpshufd {{.*#+}} xmm2 = xmm3[0,2,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[1,3,2,3,4,5,6,7]
; AVX2-NEXT: vpshufd {{.*#+}} xmm3 = xmm4[0,2,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[1,3,2,3,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
; AVX2-NEXT: vpblendd {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT: vmovdqa 112(%rdi), %xmm0
; AVX2-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[0,1,1,3,4,5,6,7]
; AVX2-NEXT: vmovdqa 96(%rdi), %xmm1
; AVX2-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[0,1,1,3,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX2-NEXT: vmovdqa 80(%rdi), %xmm1
; AVX2-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[1,3,2,3,4,5,6,7]
; AVX2-NEXT: vmovdqa 64(%rdi), %xmm2
; AVX2-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[1,3,2,3,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; AVX2-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-NEXT: vmovdqa 48(%rdi), %xmm14
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm14[0,2,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[0,1,1,3,4,5,6,7]
; AVX2-NEXT: vmovdqa 32(%rdi), %xmm8
; AVX2-NEXT: vpshufd {{.*#+}} xmm2 = xmm8[0,2,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[0,1,1,3,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; AVX2-NEXT: vmovdqa 16(%rdi), %xmm12
; AVX2-NEXT: vpshufd {{.*#+}} xmm2 = xmm12[0,2,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[1,3,2,3,4,5,6,7]
; AVX2-NEXT: vmovdqa (%rdi), %xmm9
; AVX2-NEXT: vpshufd {{.*#+}} xmm5 = xmm9[0,2,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm5 = xmm5[1,3,2,3,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm5[0],xmm2[0],xmm5[1],xmm2[1]
; AVX2-NEXT: vpblendd {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT: vmovdqa 240(%rdi), %xmm0
; AVX2-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[0,1,1,3,4,5,6,7]
; AVX2-NEXT: vmovdqa 224(%rdi), %xmm1
; AVX2-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[0,1,1,3,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX2-NEXT: vmovdqa 208(%rdi), %xmm1
; AVX2-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[1,3,2,3,4,5,6,7]
; AVX2-NEXT: vmovdqa 192(%rdi), %xmm2
; AVX2-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[1,3,2,3,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; AVX2-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-NEXT: vmovdqa 176(%rdi), %xmm1
; AVX2-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[0,1,1,3,4,5,6,7]
; AVX2-NEXT: vmovdqa 160(%rdi), %xmm2
; AVX2-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[0,1,1,3,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; AVX2-NEXT: vmovdqa 144(%rdi), %xmm2
; AVX2-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[1,3,2,3,4,5,6,7]
; AVX2-NEXT: vmovdqa 128(%rdi), %xmm3
; AVX2-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-NEXT: vpshufd {{.*#+}} xmm5 = xmm3[0,2,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm5 = xmm5[1,3,2,3,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm5[0],xmm2[0],xmm5[1],xmm2[1]
; AVX2-NEXT: vpblendd {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT: vmovdqa 496(%rdi), %xmm11
; AVX2-NEXT: vpshufd {{.*#+}} xmm0 = xmm11[0,2,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[0,1,1,3,4,5,6,7]
; AVX2-NEXT: vmovdqa 480(%rdi), %xmm13
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm13[0,2,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[0,1,1,3,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX2-NEXT: vmovdqa 464(%rdi), %xmm15
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm15[0,2,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[1,3,2,3,4,5,6,7]
; AVX2-NEXT: vmovdqa 448(%rdi), %xmm6
; AVX2-NEXT: vpshufd {{.*#+}} xmm5 = xmm6[0,2,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm5 = xmm5[1,3,2,3,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm5[0],xmm1[0],xmm5[1],xmm1[1]
; AVX2-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX2-NEXT: vpblendd {{.*#+}} ymm3 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-NEXT: vmovdqa 432(%rdi), %xmm5
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm5[0,2,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[0,1,1,3,4,5,6,7]
; AVX2-NEXT: vmovdqa 416(%rdi), %xmm4
; AVX2-NEXT: vpshufd {{.*#+}} xmm7 = xmm4[0,2,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm7 = xmm7[0,1,1,3,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm7[0],xmm1[0],xmm7[1],xmm1[1]
; AVX2-NEXT: vmovdqa 400(%rdi), %xmm7
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm7[0,2,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm0 = xmm1[1,3,2,3,4,5,6,7]
; AVX2-NEXT: vmovdqa 384(%rdi), %xmm1
; AVX2-NEXT: vpshufd {{.*#+}} xmm10 = xmm1[0,2,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm10 = xmm10[1,3,2,3,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm10[0],xmm0[0],xmm10[1],xmm0[1]
; AVX2-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm2[2,3]
; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
; AVX2-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; AVX2-NEXT: # xmm0 = mem[3,1,2,3]
; AVX2-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; AVX2-NEXT: # xmm2 = mem[3,1,2,3]
; AVX2-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[0,1,2,0,4,5,6,7]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[0,1,2,0,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; AVX2-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; AVX2-NEXT: # xmm2 = mem[3,1,2,3]
; AVX2-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; AVX2-NEXT: # xmm3 = mem[3,1,2,3]
; AVX2-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[2,0,2,3,4,5,6,7]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[2,0,2,3,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
; AVX2-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5],ymm0[6,7]
; AVX2-NEXT: vpshufd {{.*#+}} xmm2 = xmm14[3,1,2,3]
; AVX2-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-NEXT: vpshufd {{.*#+}} xmm3 = xmm8[3,1,2,3]
; AVX2-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[0,1,2,0,4,5,6,7]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[0,1,2,0,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
; AVX2-NEXT: vpshufd {{.*#+}} xmm3 = xmm12[3,1,2,3]
; AVX2-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-NEXT: vpshufd {{.*#+}} xmm8 = xmm9[3,1,2,3]
; AVX2-NEXT: vmovdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[2,0,2,3,4,5,6,7]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm8 = xmm8[2,0,2,3,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm8[0],xmm3[0],xmm8[1],xmm3[1]
; AVX2-NEXT: vpblendd {{.*#+}} xmm2 = xmm3[0,1],xmm2[2,3]
; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
; AVX2-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; AVX2-NEXT: # xmm0 = mem[3,1,2,3]
; AVX2-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; AVX2-NEXT: # xmm2 = mem[3,1,2,3]
; AVX2-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[0,1,2,0,4,5,6,7]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[0,1,2,0,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; AVX2-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; AVX2-NEXT: # xmm2 = mem[3,1,2,3]
; AVX2-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; AVX2-NEXT: # xmm3 = mem[3,1,2,3]
; AVX2-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[2,0,2,3,4,5,6,7]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[2,0,2,3,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
; AVX2-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5],ymm0[6,7]
; AVX2-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; AVX2-NEXT: # xmm2 = mem[3,1,2,3]
; AVX2-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-NEXT: vpshufd $231, (%rsp), %xmm3 # 16-byte Folded Reload
; AVX2-NEXT: # xmm3 = mem[3,1,2,3]
; AVX2-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[0,1,2,0,4,5,6,7]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[0,1,2,0,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
; AVX2-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
; AVX2-NEXT: # xmm12 = mem[3,1,2,3]
; AVX2-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
; AVX2-NEXT: # xmm10 = mem[3,1,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm3 = xmm12[2,0,2,3,4,5,6,7]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm14 = xmm10[2,0,2,3,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm14[0],xmm3[0],xmm14[1],xmm3[1]
; AVX2-NEXT: vpblendd {{.*#+}} xmm2 = xmm3[0,1],xmm2[2,3]
; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
; AVX2-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT: vpshufd {{.*#+}} xmm0 = xmm11[3,1,2,3]
; AVX2-NEXT: vmovdqa %xmm0, (%rsp) # 16-byte Spill
; AVX2-NEXT: vpshufd {{.*#+}} xmm2 = xmm13[3,1,2,3]
; AVX2-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[0,1,2,0,4,5,6,7]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[0,1,2,0,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; AVX2-NEXT: vpshufd {{.*#+}} xmm2 = xmm15[3,1,2,3]
; AVX2-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-NEXT: vpshufd {{.*#+}} xmm3 = xmm6[3,1,2,3]
; AVX2-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[2,0,2,3,4,5,6,7]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[2,0,2,3,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
; AVX2-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5],ymm0[6,7]
; AVX2-NEXT: vpshufd {{.*#+}} xmm2 = xmm5[3,1,2,3]
; AVX2-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-NEXT: vpshufd {{.*#+}} xmm3 = xmm4[3,1,2,3]
; AVX2-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[0,1,2,0,4,5,6,7]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[0,1,2,0,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
; AVX2-NEXT: vpshufd {{.*#+}} xmm13 = xmm7[3,1,2,3]
; AVX2-NEXT: vpshufd {{.*#+}} xmm11 = xmm1[3,1,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm1 = xmm13[2,0,2,3,4,5,6,7]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm3 = xmm11[2,0,2,3,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; AVX2-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
; AVX2-NEXT: # xmm9 = mem[3,1,2,3]
; AVX2-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
; AVX2-NEXT: # xmm8 = mem[3,1,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm0 = xmm9[0,1,2,0,4,5,6,7]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm1 = xmm8[0,1,2,0,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX2-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
; AVX2-NEXT: # xmm7 = mem[3,1,2,3]
; AVX2-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
; AVX2-NEXT: # xmm6 = mem[3,1,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm1 = xmm7[2,0,2,3,4,5,6,7]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm2 = xmm6[2,0,2,3,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; AVX2-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX2-NEXT: vpblendd {{.*#+}} ymm14 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
; AVX2-NEXT: # xmm5 = mem[3,1,2,3]
; AVX2-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; AVX2-NEXT: # xmm4 = mem[3,1,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm0 = xmm5[0,1,2,0,4,5,6,7]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm1 = xmm4[0,1,2,0,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX2-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; AVX2-NEXT: # xmm3 = mem[3,1,2,3]
; AVX2-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; AVX2-NEXT: # xmm2 = mem[3,1,2,3]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm0 = xmm3[2,0,2,3,4,5,6,7]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm15 = xmm2[2,0,2,3,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm15[0],xmm0[0],xmm15[1],xmm0[1]
; AVX2-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm14[4,5,6,7]
; AVX2-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT: vpshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; AVX2-NEXT: # xmm0 = mem[0,1,3,1,4,5,6,7]
; AVX2-NEXT: vpshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; AVX2-NEXT: # xmm1 = mem[0,1,3,1,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX2-NEXT: vpshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; AVX2-NEXT: # xmm1 = mem[3,1,2,3,4,5,6,7]
; AVX2-NEXT: vpshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
; AVX2-NEXT: # xmm14 = mem[3,1,2,3,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm14[0],xmm1[0],xmm14[1],xmm1[1]
; AVX2-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-NEXT: vpshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; AVX2-NEXT: # xmm1 = mem[0,1,3,1,4,5,6,7]
; AVX2-NEXT: vpshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
; AVX2-NEXT: # xmm14 = mem[0,1,3,1,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm14[0],xmm1[0],xmm14[1],xmm1[1]
; AVX2-NEXT: vpshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
; AVX2-NEXT: # xmm14 = mem[3,1,2,3,4,5,6,7]
; AVX2-NEXT: vpshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
; AVX2-NEXT: # xmm15 = mem[3,1,2,3,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm14 = xmm15[0],xmm14[0],xmm15[1],xmm14[1]
; AVX2-NEXT: vpblendd {{.*#+}} xmm1 = xmm14[0,1],xmm1[2,3]
; AVX2-NEXT: vpblendd {{.*#+}} ymm15 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-NEXT: vpshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; AVX2-NEXT: # xmm0 = mem[0,1,3,1,4,5,6,7]
; AVX2-NEXT: vpshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; AVX2-NEXT: # xmm1 = mem[0,1,3,1,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX2-NEXT: vpshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; AVX2-NEXT: # xmm1 = mem[3,1,2,3,4,5,6,7]
; AVX2-NEXT: vpshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
; AVX2-NEXT: # xmm14 = mem[3,1,2,3,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm14[0],xmm1[0],xmm14[1],xmm1[1]
; AVX2-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-NEXT: vpshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; AVX2-NEXT: # xmm1 = mem[0,1,3,1,4,5,6,7]
; AVX2-NEXT: vpshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
; AVX2-NEXT: # xmm14 = mem[0,1,3,1,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm14[0],xmm1[0],xmm14[1],xmm1[1]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm12 = xmm12[3,1,2,3,4,5,6,7]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm10 = xmm10[3,1,2,3,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm10 = xmm10[0],xmm12[0],xmm10[1],xmm12[1]
; AVX2-NEXT: vpblendd {{.*#+}} xmm1 = xmm10[0,1],xmm1[2,3]
; AVX2-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm0 = xmm9[0,1,3,1,4,5,6,7]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm8 = xmm8[0,1,3,1,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm8[0],xmm0[0],xmm8[1],xmm0[1]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm7 = xmm7[3,1,2,3,4,5,6,7]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm6 = xmm6[3,1,2,3,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm6 = xmm6[0],xmm7[0],xmm6[1],xmm7[1]
; AVX2-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-NEXT: vinserti128 $1, %xmm6, %ymm0, %ymm6
; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm6[0,1,2,3,4,5],ymm0[6,7]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm5 = xmm5[0,1,3,1,4,5,6,7]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[0,1,3,1,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm4 = xmm4[0],xmm5[0],xmm4[1],xmm5[1]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[3,1,2,3,4,5,6,7]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[3,1,2,3,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
; AVX2-NEXT: vpblendd {{.*#+}} xmm2 = xmm2[0,1],xmm4[2,3]
; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
; AVX2-NEXT: vpshuflw $116, (%rsp), %xmm2 # 16-byte Folded Reload
; AVX2-NEXT: # xmm2 = mem[0,1,3,1,4,5,6,7]
; AVX2-NEXT: vpshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; AVX2-NEXT: # xmm3 = mem[0,1,3,1,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
; AVX2-NEXT: vpshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; AVX2-NEXT: # xmm3 = mem[3,1,2,3,4,5,6,7]
; AVX2-NEXT: vpshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; AVX2-NEXT: # xmm4 = mem[3,1,2,3,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
; AVX2-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX2-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
; AVX2-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
; AVX2-NEXT: vpshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; AVX2-NEXT: # xmm3 = mem[0,1,3,1,4,5,6,7]
; AVX2-NEXT: vpshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; AVX2-NEXT: # xmm4 = mem[0,1,3,1,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm4 = xmm13[3,1,2,3,4,5,6,7]
; AVX2-NEXT: vpshuflw {{.*#+}} xmm5 = xmm11[3,1,2,3,4,5,6,7]
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm4 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
; AVX2-NEXT: vpblendd {{.*#+}} xmm3 = xmm4[0,1],xmm3[2,3]
; AVX2-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
; AVX2-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-NEXT: vmovaps %ymm3, 32(%rsi)
; AVX2-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-NEXT: vmovaps %ymm3, 96(%rsi)
; AVX2-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-NEXT: vmovaps %ymm3, 64(%rsi)
; AVX2-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-NEXT: vmovaps %ymm3, (%rsi)
; AVX2-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-NEXT: vmovaps %ymm3, 96(%rdx)
; AVX2-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-NEXT: vmovaps %ymm3, 32(%rdx)
; AVX2-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-NEXT: vmovaps %ymm3, (%rdx)
; AVX2-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-NEXT: vmovaps %ymm3, 64(%rdx)
; AVX2-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-NEXT: vmovaps %ymm3, 32(%rcx)
; AVX2-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-NEXT: vmovaps %ymm3, 96(%rcx)
; AVX2-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-NEXT: vmovaps %ymm3, 64(%rcx)
; AVX2-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-NEXT: vmovaps %ymm3, (%rcx)
; AVX2-NEXT: vmovdqa %ymm2, 96(%r8)
; AVX2-NEXT: vmovdqa %ymm0, 32(%r8)
; AVX2-NEXT: vmovdqa %ymm1, 64(%r8)
; AVX2-NEXT: vmovdqa %ymm15, (%r8)
; AVX2-NEXT: addq $696, %rsp # imm = 0x2B8
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
; AVX2-FP-LABEL: load_i16_stride4_vf64:
; AVX2-FP: # %bb.0:
; AVX2-FP-NEXT: subq $712, %rsp # imm = 0x2C8
; AVX2-FP-NEXT: vpxor %xmm0, %xmm0, %xmm0
; AVX2-FP-NEXT: vpblendw {{.*#+}} ymm1 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-FP-NEXT: vextracti128 $1, %ymm1, %xmm2
; AVX2-FP-NEXT: vpackusdw %xmm2, %xmm1, %xmm1
; AVX2-FP-NEXT: vpblendw {{.*#+}} ymm2 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-FP-NEXT: vextracti128 $1, %ymm2, %xmm3
; AVX2-FP-NEXT: vpackusdw %xmm3, %xmm2, %xmm2
; AVX2-FP-NEXT: vpackusdw %xmm1, %xmm2, %xmm1
; AVX2-FP-NEXT: vpblendw {{.*#+}} ymm2 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-FP-NEXT: vextracti128 $1, %ymm2, %xmm3
; AVX2-FP-NEXT: vpackusdw %xmm3, %xmm2, %xmm2
; AVX2-FP-NEXT: vpackusdw %xmm2, %xmm2, %xmm2
; AVX2-FP-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX2-FP-NEXT: vpblendw {{.*#+}} ymm3 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-FP-NEXT: vextracti128 $1, %ymm3, %xmm4
; AVX2-FP-NEXT: vpackusdw %xmm4, %xmm3, %xmm3
; AVX2-FP-NEXT: vpackusdw %xmm3, %xmm3, %xmm3
; AVX2-FP-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
; AVX2-FP-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
; AVX2-FP-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX2-FP-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT: vpblendw {{.*#+}} ymm1 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-FP-NEXT: vextracti128 $1, %ymm1, %xmm2
; AVX2-FP-NEXT: vpackusdw %xmm2, %xmm1, %xmm1
; AVX2-FP-NEXT: vpblendw {{.*#+}} ymm2 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-FP-NEXT: vextracti128 $1, %ymm2, %xmm3
; AVX2-FP-NEXT: vpackusdw %xmm3, %xmm2, %xmm2
; AVX2-FP-NEXT: vpackusdw %xmm1, %xmm2, %xmm1
; AVX2-FP-NEXT: vpblendw {{.*#+}} ymm2 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-FP-NEXT: vextracti128 $1, %ymm2, %xmm3
; AVX2-FP-NEXT: vpackusdw %xmm3, %xmm2, %xmm2
; AVX2-FP-NEXT: vpackusdw %xmm2, %xmm2, %xmm2
; AVX2-FP-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX2-FP-NEXT: vpblendw {{.*#+}} ymm3 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-FP-NEXT: vextracti128 $1, %ymm3, %xmm4
; AVX2-FP-NEXT: vpackusdw %xmm4, %xmm3, %xmm3
; AVX2-FP-NEXT: vpackusdw %xmm3, %xmm3, %xmm3
; AVX2-FP-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
; AVX2-FP-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
; AVX2-FP-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX2-FP-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT: vpblendw {{.*#+}} ymm1 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-FP-NEXT: vextracti128 $1, %ymm1, %xmm2
; AVX2-FP-NEXT: vpackusdw %xmm2, %xmm1, %xmm1
; AVX2-FP-NEXT: vpblendw {{.*#+}} ymm2 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-FP-NEXT: vextracti128 $1, %ymm2, %xmm3
; AVX2-FP-NEXT: vpackusdw %xmm3, %xmm2, %xmm2
; AVX2-FP-NEXT: vpackusdw %xmm1, %xmm2, %xmm1
; AVX2-FP-NEXT: vpblendw {{.*#+}} ymm2 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-FP-NEXT: vextracti128 $1, %ymm2, %xmm3
; AVX2-FP-NEXT: vpackusdw %xmm3, %xmm2, %xmm2
; AVX2-FP-NEXT: vpackusdw %xmm2, %xmm2, %xmm2
; AVX2-FP-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX2-FP-NEXT: vpblendw {{.*#+}} ymm3 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-FP-NEXT: vextracti128 $1, %ymm3, %xmm4
; AVX2-FP-NEXT: vpackusdw %xmm4, %xmm3, %xmm3
; AVX2-FP-NEXT: vpackusdw %xmm3, %xmm3, %xmm3
; AVX2-FP-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
; AVX2-FP-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
; AVX2-FP-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX2-FP-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT: vpblendw {{.*#+}} ymm1 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-FP-NEXT: vextracti128 $1, %ymm1, %xmm2
; AVX2-FP-NEXT: vpackusdw %xmm2, %xmm1, %xmm1
; AVX2-FP-NEXT: vpblendw {{.*#+}} ymm2 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-FP-NEXT: vextracti128 $1, %ymm2, %xmm3
; AVX2-FP-NEXT: vpackusdw %xmm3, %xmm2, %xmm2
; AVX2-FP-NEXT: vpackusdw %xmm1, %xmm2, %xmm1
; AVX2-FP-NEXT: vpblendw {{.*#+}} ymm2 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-FP-NEXT: vextracti128 $1, %ymm2, %xmm3
; AVX2-FP-NEXT: vpackusdw %xmm3, %xmm2, %xmm2
; AVX2-FP-NEXT: vpackusdw %xmm2, %xmm2, %xmm2
; AVX2-FP-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX2-FP-NEXT: vpblendw {{.*#+}} ymm0 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-FP-NEXT: vextracti128 $1, %ymm0, %xmm3
; AVX2-FP-NEXT: vpackusdw %xmm3, %xmm0, %xmm0
; AVX2-FP-NEXT: vpackusdw %xmm0, %xmm0, %xmm0
; AVX2-FP-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-FP-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
; AVX2-FP-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FP-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT: vmovdqa 256(%rdi), %xmm5
; AVX2-FP-NEXT: vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FP-NEXT: vmovdqa 288(%rdi), %xmm3
; AVX2-FP-NEXT: vmovdqa %xmm3, (%rsp) # 16-byte Spill
; AVX2-FP-NEXT: vmovdqa 304(%rdi), %xmm7
; AVX2-FP-NEXT: vmovdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FP-NEXT: vmovdqa {{.*#+}} xmm4 = [0,1,2,3,2,3,10,11,8,9,10,11,12,13,14,15]
; AVX2-FP-NEXT: vmovdqa 368(%rdi), %xmm0
; AVX2-FP-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FP-NEXT: vpshufb %xmm4, %xmm0, %xmm0
; AVX2-FP-NEXT: vmovdqa 352(%rdi), %xmm1
; AVX2-FP-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FP-NEXT: vpshufb %xmm4, %xmm1, %xmm1
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX2-FP-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-FP-NEXT: vmovdqa {{.*#+}} xmm6 = [2,3,10,11,8,9,10,11,8,9,10,11,12,13,14,15]
; AVX2-FP-NEXT: vmovdqa 336(%rdi), %xmm1
; AVX2-FP-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FP-NEXT: vpshufb %xmm6, %xmm1, %xmm1
; AVX2-FP-NEXT: vmovdqa 320(%rdi), %xmm2
; AVX2-FP-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FP-NEXT: vpshufb %xmm6, %xmm2, %xmm2
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; AVX2-FP-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX2-FP-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-FP-NEXT: vpshufb %xmm4, %xmm7, %xmm1
; AVX2-FP-NEXT: vpshufb %xmm4, %xmm3, %xmm2
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; AVX2-FP-NEXT: vmovdqa 272(%rdi), %xmm2
; AVX2-FP-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FP-NEXT: vpshufb %xmm6, %xmm2, %xmm2
; AVX2-FP-NEXT: vpshufb %xmm6, %xmm5, %xmm3
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
; AVX2-FP-NEXT: vpblendd {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
; AVX2-FP-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FP-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT: vmovdqa 112(%rdi), %xmm5
; AVX2-FP-NEXT: vpshufb %xmm4, %xmm5, %xmm2
; AVX2-FP-NEXT: vmovdqa 96(%rdi), %xmm11
; AVX2-FP-NEXT: vpshufb %xmm4, %xmm11, %xmm3
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
; AVX2-FP-NEXT: vmovdqa 80(%rdi), %xmm12
; AVX2-FP-NEXT: vpshufb %xmm6, %xmm12, %xmm7
; AVX2-FP-NEXT: vmovdqa 64(%rdi), %xmm15
; AVX2-FP-NEXT: vpshufb %xmm6, %xmm15, %xmm8
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm7 = xmm8[0],xmm7[0],xmm8[1],xmm7[1]
; AVX2-FP-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
; AVX2-FP-NEXT: vinserti128 $1, %xmm7, %ymm0, %ymm7
; AVX2-FP-NEXT: vpblendd {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm3[6,7]
; AVX2-FP-NEXT: vmovdqa 32(%rdi), %xmm0
; AVX2-FP-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FP-NEXT: vmovdqa 48(%rdi), %xmm3
; AVX2-FP-NEXT: vpshufb %xmm4, %xmm3, %xmm8
; AVX2-FP-NEXT: vpshufb %xmm4, %xmm0, %xmm9
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm8 = xmm9[0],xmm8[0],xmm9[1],xmm8[1]
; AVX2-FP-NEXT: vmovdqa (%rdi), %xmm0
; AVX2-FP-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FP-NEXT: vmovdqa 16(%rdi), %xmm1
; AVX2-FP-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FP-NEXT: vpshufb %xmm6, %xmm1, %xmm9
; AVX2-FP-NEXT: vpshufb %xmm6, %xmm0, %xmm10
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm9 = xmm10[0],xmm9[0],xmm10[1],xmm9[1]
; AVX2-FP-NEXT: vpblendd {{.*#+}} xmm8 = xmm9[0,1],xmm8[2,3]
; AVX2-FP-NEXT: vpblendd {{.*#+}} ymm0 = ymm8[0,1,2,3],ymm7[4,5,6,7]
; AVX2-FP-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT: vmovdqa 240(%rdi), %xmm0
; AVX2-FP-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FP-NEXT: vpshufb %xmm4, %xmm0, %xmm7
; AVX2-FP-NEXT: vmovdqa 224(%rdi), %xmm0
; AVX2-FP-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FP-NEXT: vpshufb %xmm4, %xmm0, %xmm8
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm7 = xmm8[0],xmm7[0],xmm8[1],xmm7[1]
; AVX2-FP-NEXT: vmovdqa 208(%rdi), %xmm0
; AVX2-FP-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FP-NEXT: vpshufb %xmm6, %xmm0, %xmm8
; AVX2-FP-NEXT: vmovdqa 192(%rdi), %xmm0
; AVX2-FP-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FP-NEXT: vpshufb %xmm6, %xmm0, %xmm9
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm8 = xmm9[0],xmm8[0],xmm9[1],xmm8[1]
; AVX2-FP-NEXT: vinserti128 $1, %xmm7, %ymm0, %ymm7
; AVX2-FP-NEXT: vinserti128 $1, %xmm8, %ymm0, %ymm8
; AVX2-FP-NEXT: vpblendd {{.*#+}} ymm7 = ymm8[0,1,2,3,4,5],ymm7[6,7]
; AVX2-FP-NEXT: vmovdqa 160(%rdi), %xmm0
; AVX2-FP-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FP-NEXT: vmovdqa 176(%rdi), %xmm1
; AVX2-FP-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FP-NEXT: vpshufb %xmm4, %xmm1, %xmm8
; AVX2-FP-NEXT: vpshufb %xmm4, %xmm0, %xmm9
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm8 = xmm9[0],xmm8[0],xmm9[1],xmm8[1]
; AVX2-FP-NEXT: vmovdqa 128(%rdi), %xmm0
; AVX2-FP-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FP-NEXT: vmovdqa 144(%rdi), %xmm1
; AVX2-FP-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FP-NEXT: vpshufb %xmm6, %xmm1, %xmm9
; AVX2-FP-NEXT: vpshufb %xmm6, %xmm0, %xmm10
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm9 = xmm10[0],xmm9[0],xmm10[1],xmm9[1]
; AVX2-FP-NEXT: vpblendd {{.*#+}} xmm8 = xmm9[0,1],xmm8[2,3]
; AVX2-FP-NEXT: vpblendd {{.*#+}} ymm0 = ymm8[0,1,2,3],ymm7[4,5,6,7]
; AVX2-FP-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT: vmovdqa 496(%rdi), %xmm8
; AVX2-FP-NEXT: vpshufb %xmm4, %xmm8, %xmm9
; AVX2-FP-NEXT: vmovdqa 480(%rdi), %xmm7
; AVX2-FP-NEXT: vpshufb %xmm4, %xmm7, %xmm10
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm13 = xmm10[0],xmm9[0],xmm10[1],xmm9[1]
; AVX2-FP-NEXT: vmovdqa 464(%rdi), %xmm10
; AVX2-FP-NEXT: vpshufb %xmm6, %xmm10, %xmm14
; AVX2-FP-NEXT: vmovdqa 448(%rdi), %xmm0
; AVX2-FP-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FP-NEXT: vpshufb %xmm6, %xmm0, %xmm9
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm9 = xmm9[0],xmm14[0],xmm9[1],xmm14[1]
; AVX2-FP-NEXT: vinserti128 $1, %xmm13, %ymm0, %ymm13
; AVX2-FP-NEXT: vinserti128 $1, %xmm9, %ymm0, %ymm9
; AVX2-FP-NEXT: vpblendd {{.*#+}} ymm9 = ymm9[0,1,2,3,4,5],ymm13[6,7]
; AVX2-FP-NEXT: vmovdqa 416(%rdi), %xmm14
; AVX2-FP-NEXT: vmovdqa 432(%rdi), %xmm13
; AVX2-FP-NEXT: vpshufb %xmm4, %xmm13, %xmm0
; AVX2-FP-NEXT: vpshufb %xmm4, %xmm14, %xmm4
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
; AVX2-FP-NEXT: vmovdqa 384(%rdi), %xmm0
; AVX2-FP-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FP-NEXT: vmovdqa 400(%rdi), %xmm4
; AVX2-FP-NEXT: vpshufb %xmm6, %xmm4, %xmm1
; AVX2-FP-NEXT: vpshufb %xmm6, %xmm0, %xmm6
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm6[0],xmm1[0],xmm6[1],xmm1[1]
; AVX2-FP-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm2[2,3]
; AVX2-FP-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm9[4,5,6,7]
; AVX2-FP-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT: vpshufd {{.*#+}} xmm0 = xmm5[3,1,2,3]
; AVX2-FP-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FP-NEXT: vpshufd {{.*#+}} xmm1 = xmm11[3,1,2,3]
; AVX2-FP-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[0,1,2,0,4,5,6,7]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[0,1,2,0,4,5,6,7]
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX2-FP-NEXT: vpshufd {{.*#+}} xmm1 = xmm12[3,1,2,3]
; AVX2-FP-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FP-NEXT: vpshufd {{.*#+}} xmm2 = xmm15[3,1,2,3]
; AVX2-FP-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[2,0,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[2,0,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; AVX2-FP-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-FP-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX2-FP-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-FP-NEXT: vpshufd {{.*#+}} xmm1 = xmm3[3,1,2,3]
; AVX2-FP-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FP-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; AVX2-FP-NEXT: # xmm2 = mem[3,1,2,3]
; AVX2-FP-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[0,1,2,0,4,5,6,7]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[0,1,2,0,4,5,6,7]
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; AVX2-FP-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; AVX2-FP-NEXT: # xmm2 = mem[3,1,2,3]
; AVX2-FP-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FP-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; AVX2-FP-NEXT: # xmm3 = mem[3,1,2,3]
; AVX2-FP-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[2,0,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[2,0,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
; AVX2-FP-NEXT: vpblendd {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
; AVX2-FP-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FP-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; AVX2-FP-NEXT: # xmm0 = mem[3,1,2,3]
; AVX2-FP-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FP-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; AVX2-FP-NEXT: # xmm1 = mem[3,1,2,3]
; AVX2-FP-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[0,1,2,0,4,5,6,7]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[0,1,2,0,4,5,6,7]
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX2-FP-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; AVX2-FP-NEXT: # xmm1 = mem[3,1,2,3]
; AVX2-FP-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FP-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; AVX2-FP-NEXT: # xmm2 = mem[3,1,2,3]
; AVX2-FP-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[2,0,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[2,0,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; AVX2-FP-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-FP-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX2-FP-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-FP-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; AVX2-FP-NEXT: # xmm1 = mem[3,1,2,3]
; AVX2-FP-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FP-NEXT: vpshufd $231, (%rsp), %xmm15 # 16-byte Folded Reload
; AVX2-FP-NEXT: # xmm15 = mem[3,1,2,3]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[0,1,2,0,4,5,6,7]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm2 = xmm15[0,1,2,0,4,5,6,7]
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; AVX2-FP-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
; AVX2-FP-NEXT: # xmm12 = mem[3,1,2,3]
; AVX2-FP-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
; AVX2-FP-NEXT: # xmm11 = mem[3,1,2,3]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm2 = xmm12[2,0,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm3 = xmm11[2,0,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
; AVX2-FP-NEXT: vpblendd {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
; AVX2-FP-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FP-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT: vpshufd {{.*#+}} xmm0 = xmm8[3,1,2,3]
; AVX2-FP-NEXT: vmovdqa %xmm0, (%rsp) # 16-byte Spill
; AVX2-FP-NEXT: vpshufd {{.*#+}} xmm1 = xmm7[3,1,2,3]
; AVX2-FP-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[0,1,2,0,4,5,6,7]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[0,1,2,0,4,5,6,7]
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX2-FP-NEXT: vpshufd {{.*#+}} xmm1 = xmm10[3,1,2,3]
; AVX2-FP-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FP-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; AVX2-FP-NEXT: # xmm2 = mem[3,1,2,3]
; AVX2-FP-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[2,0,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[2,0,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; AVX2-FP-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-FP-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX2-FP-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-FP-NEXT: vpshufd {{.*#+}} xmm1 = xmm13[3,1,2,3]
; AVX2-FP-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FP-NEXT: vpshufd {{.*#+}} xmm2 = xmm14[3,1,2,3]
; AVX2-FP-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[0,1,2,0,4,5,6,7]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[0,1,2,0,4,5,6,7]
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; AVX2-FP-NEXT: vpshufd {{.*#+}} xmm2 = xmm4[3,1,2,3]
; AVX2-FP-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FP-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
; AVX2-FP-NEXT: # xmm10 = mem[3,1,2,3]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[2,0,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm3 = xmm10[2,0,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
; AVX2-FP-NEXT: vpblendd {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
; AVX2-FP-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FP-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
; AVX2-FP-NEXT: # xmm14 = mem[3,1,2,3]
; AVX2-FP-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
; AVX2-FP-NEXT: # xmm9 = mem[3,1,2,3]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm0 = xmm14[0,1,2,0,4,5,6,7]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm1 = xmm9[0,1,2,0,4,5,6,7]
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX2-FP-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
; AVX2-FP-NEXT: # xmm8 = mem[3,1,2,3]
; AVX2-FP-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
; AVX2-FP-NEXT: # xmm7 = mem[3,1,2,3]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm1 = xmm8[2,0,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm2 = xmm7[2,0,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; AVX2-FP-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-FP-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX2-FP-NEXT: vpblendd {{.*#+}} ymm3 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-FP-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
; AVX2-FP-NEXT: # xmm6 = mem[3,1,2,3]
; AVX2-FP-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
; AVX2-FP-NEXT: # xmm5 = mem[3,1,2,3]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm1 = xmm6[0,1,2,0,4,5,6,7]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm2 = xmm5[0,1,2,0,4,5,6,7]
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; AVX2-FP-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; AVX2-FP-NEXT: # xmm4 = mem[3,1,2,3]
; AVX2-FP-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; AVX2-FP-NEXT: # xmm2 = mem[3,1,2,3]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm0 = xmm4[2,0,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm13 = xmm2[2,0,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm13[0],xmm0[0],xmm13[1],xmm0[1]
; AVX2-FP-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
; AVX2-FP-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
; AVX2-FP-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT: vpshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; AVX2-FP-NEXT: # xmm0 = mem[0,1,3,1,4,5,6,7]
; AVX2-FP-NEXT: vpshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; AVX2-FP-NEXT: # xmm1 = mem[0,1,3,1,4,5,6,7]
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX2-FP-NEXT: vpshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; AVX2-FP-NEXT: # xmm1 = mem[3,1,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; AVX2-FP-NEXT: # xmm3 = mem[3,1,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; AVX2-FP-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-FP-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX2-FP-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-FP-NEXT: vpshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; AVX2-FP-NEXT: # xmm1 = mem[0,1,3,1,4,5,6,7]
; AVX2-FP-NEXT: vpshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; AVX2-FP-NEXT: # xmm3 = mem[0,1,3,1,4,5,6,7]
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; AVX2-FP-NEXT: vpshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; AVX2-FP-NEXT: # xmm3 = mem[3,1,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
; AVX2-FP-NEXT: # xmm13 = mem[3,1,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm13[0],xmm3[0],xmm13[1],xmm3[1]
; AVX2-FP-NEXT: vpblendd {{.*#+}} xmm1 = xmm3[0,1],xmm1[2,3]
; AVX2-FP-NEXT: vpblendd {{.*#+}} ymm13 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FP-NEXT: vpshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; AVX2-FP-NEXT: # xmm0 = mem[0,1,3,1,4,5,6,7]
; AVX2-FP-NEXT: vpshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; AVX2-FP-NEXT: # xmm1 = mem[0,1,3,1,4,5,6,7]
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX2-FP-NEXT: vpshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; AVX2-FP-NEXT: # xmm1 = mem[3,1,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; AVX2-FP-NEXT: # xmm3 = mem[3,1,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; AVX2-FP-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-FP-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX2-FP-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-FP-NEXT: vpshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; AVX2-FP-NEXT: # xmm1 = mem[0,1,3,1,4,5,6,7]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm3 = xmm15[0,1,3,1,4,5,6,7]
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm3 = xmm12[3,1,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm11 = xmm11[3,1,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm11[0],xmm3[0],xmm11[1],xmm3[1]
; AVX2-FP-NEXT: vpblendd {{.*#+}} xmm1 = xmm3[0,1],xmm1[2,3]
; AVX2-FP-NEXT: vpblendd {{.*#+}} ymm11 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm0 = xmm14[0,1,3,1,4,5,6,7]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm1 = xmm9[0,1,3,1,4,5,6,7]
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm1 = xmm8[3,1,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm3 = xmm7[3,1,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; AVX2-FP-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-FP-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX2-FP-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm1 = xmm6[0,1,3,1,4,5,6,7]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm3 = xmm5[0,1,3,1,4,5,6,7]
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm3 = xmm4[3,1,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[3,1,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
; AVX2-FP-NEXT: vpblendd {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
; AVX2-FP-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FP-NEXT: vpshuflw $116, (%rsp), %xmm1 # 16-byte Folded Reload
; AVX2-FP-NEXT: # xmm1 = mem[0,1,3,1,4,5,6,7]
; AVX2-FP-NEXT: vpshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; AVX2-FP-NEXT: # xmm2 = mem[0,1,3,1,4,5,6,7]
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; AVX2-FP-NEXT: vpshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; AVX2-FP-NEXT: # xmm2 = mem[3,1,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; AVX2-FP-NEXT: # xmm3 = mem[3,1,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
; AVX2-FP-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX2-FP-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX2-FP-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FP-NEXT: vpshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; AVX2-FP-NEXT: # xmm2 = mem[0,1,3,1,4,5,6,7]
; AVX2-FP-NEXT: vpshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; AVX2-FP-NEXT: # xmm3 = mem[0,1,3,1,4,5,6,7]
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
; AVX2-FP-NEXT: vpshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; AVX2-FP-NEXT: # xmm3 = mem[3,1,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpshuflw {{.*#+}} xmm4 = xmm10[3,1,2,3,4,5,6,7]
; AVX2-FP-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
; AVX2-FP-NEXT: vpblendd {{.*#+}} xmm2 = xmm3[0,1],xmm2[2,3]
; AVX2-FP-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FP-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FP-NEXT: vmovaps %ymm2, 32(%rsi)
; AVX2-FP-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FP-NEXT: vmovaps %ymm2, 96(%rsi)
; AVX2-FP-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FP-NEXT: vmovaps %ymm2, 64(%rsi)
; AVX2-FP-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FP-NEXT: vmovaps %ymm2, (%rsi)
; AVX2-FP-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FP-NEXT: vmovaps %ymm2, 96(%rdx)
; AVX2-FP-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FP-NEXT: vmovaps %ymm2, 32(%rdx)
; AVX2-FP-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FP-NEXT: vmovaps %ymm2, (%rdx)
; AVX2-FP-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FP-NEXT: vmovaps %ymm2, 64(%rdx)
; AVX2-FP-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FP-NEXT: vmovaps %ymm2, 32(%rcx)
; AVX2-FP-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FP-NEXT: vmovaps %ymm2, 96(%rcx)
; AVX2-FP-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FP-NEXT: vmovaps %ymm2, 64(%rcx)
; AVX2-FP-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FP-NEXT: vmovaps %ymm2, (%rcx)
; AVX2-FP-NEXT: vmovdqa %ymm1, 96(%r8)
; AVX2-FP-NEXT: vmovdqa %ymm0, 32(%r8)
; AVX2-FP-NEXT: vmovdqa %ymm11, 64(%r8)
; AVX2-FP-NEXT: vmovdqa %ymm13, (%r8)
; AVX2-FP-NEXT: addq $712, %rsp # imm = 0x2C8
; AVX2-FP-NEXT: vzeroupper
; AVX2-FP-NEXT: retq
;
; AVX2-FCP-LABEL: load_i16_stride4_vf64:
; AVX2-FCP: # %bb.0:
; AVX2-FCP-NEXT: subq $680, %rsp # imm = 0x2A8
; AVX2-FCP-NEXT: vmovdqa 320(%rdi), %ymm7
; AVX2-FCP-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT: vmovdqa 352(%rdi), %ymm6
; AVX2-FCP-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT: vmovdqa 64(%rdi), %ymm5
; AVX2-FCP-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT: vmovdqa 96(%rdi), %ymm4
; AVX2-FCP-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT: vpxor %xmm1, %xmm1, %xmm1
; AVX2-FCP-NEXT: vpblendw {{.*#+}} ymm0 = mem[0],ymm1[1,2,3],mem[4],ymm1[5,6,7],mem[8],ymm1[9,10,11],mem[12],ymm1[13,14,15]
; AVX2-FCP-NEXT: vextracti128 $1, %ymm0, %xmm2
; AVX2-FCP-NEXT: vpackusdw %xmm2, %xmm0, %xmm0
; AVX2-FCP-NEXT: vpblendw {{.*#+}} ymm2 = mem[0],ymm1[1,2,3],mem[4],ymm1[5,6,7],mem[8],ymm1[9,10,11],mem[12],ymm1[13,14,15]
; AVX2-FCP-NEXT: vextracti128 $1, %ymm2, %xmm3
; AVX2-FCP-NEXT: vpackusdw %xmm3, %xmm2, %xmm2
; AVX2-FCP-NEXT: vpackusdw %xmm0, %xmm2, %xmm2
; AVX2-FCP-NEXT: vpmovsxbd {{.*#+}} ymm0 = [0,2,2,3,0,2,4,6]
; AVX2-FCP-NEXT: vpermd %ymm4, %ymm0, %ymm3
; AVX2-FCP-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT: vmovdqa {{.*#+}} ymm4 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,16,17,20,21,24,25,28,29]
; AVX2-FCP-NEXT: vpshufb %ymm4, %ymm3, %ymm3
; AVX2-FCP-NEXT: vpermd %ymm5, %ymm0, %ymm5
; AVX2-FCP-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT: vpshufb %ymm4, %ymm5, %ymm5
; AVX2-FCP-NEXT: vpblendd {{.*#+}} ymm3 = ymm5[0,1,2,3,4,5],ymm3[6,7]
; AVX2-FCP-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
; AVX2-FCP-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT: vpblendw {{.*#+}} ymm2 = mem[0],ymm1[1,2,3],mem[4],ymm1[5,6,7],mem[8],ymm1[9,10,11],mem[12],ymm1[13,14,15]
; AVX2-FCP-NEXT: vextracti128 $1, %ymm2, %xmm3
; AVX2-FCP-NEXT: vpackusdw %xmm3, %xmm2, %xmm2
; AVX2-FCP-NEXT: vpblendw {{.*#+}} ymm3 = mem[0],ymm1[1,2,3],mem[4],ymm1[5,6,7],mem[8],ymm1[9,10,11],mem[12],ymm1[13,14,15]
; AVX2-FCP-NEXT: vextracti128 $1, %ymm3, %xmm5
; AVX2-FCP-NEXT: vpackusdw %xmm5, %xmm3, %xmm3
; AVX2-FCP-NEXT: vpackusdw %xmm2, %xmm3, %xmm3
; AVX2-FCP-NEXT: vpermd %ymm6, %ymm0, %ymm2
; AVX2-FCP-NEXT: vpshufb %ymm4, %ymm2, %ymm5
; AVX2-FCP-NEXT: vpermd %ymm7, %ymm0, %ymm6
; AVX2-FCP-NEXT: vpshufb %ymm4, %ymm6, %ymm7
; AVX2-FCP-NEXT: vpblendd {{.*#+}} ymm5 = ymm7[0,1,2,3,4,5],ymm5[6,7]
; AVX2-FCP-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm5[4,5,6,7]
; AVX2-FCP-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT: vpblendw {{.*#+}} ymm3 = mem[0],ymm1[1,2,3],mem[4],ymm1[5,6,7],mem[8],ymm1[9,10,11],mem[12],ymm1[13,14,15]
; AVX2-FCP-NEXT: vextracti128 $1, %ymm3, %xmm5
; AVX2-FCP-NEXT: vpackusdw %xmm5, %xmm3, %xmm3
; AVX2-FCP-NEXT: vpblendw {{.*#+}} ymm5 = mem[0],ymm1[1,2,3],mem[4],ymm1[5,6,7],mem[8],ymm1[9,10,11],mem[12],ymm1[13,14,15]
; AVX2-FCP-NEXT: vextracti128 $1, %ymm5, %xmm7
; AVX2-FCP-NEXT: vpackusdw %xmm7, %xmm5, %xmm5
; AVX2-FCP-NEXT: vmovdqa 448(%rdi), %ymm7
; AVX2-FCP-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT: vpackusdw %xmm3, %xmm5, %xmm3
; AVX2-FCP-NEXT: vmovdqa 480(%rdi), %ymm5
; AVX2-FCP-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT: vpermd %ymm5, %ymm0, %ymm5
; AVX2-FCP-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT: vpshufb %ymm4, %ymm5, %ymm5
; AVX2-FCP-NEXT: vpermd %ymm7, %ymm0, %ymm14
; AVX2-FCP-NEXT: vpshufb %ymm4, %ymm14, %ymm7
; AVX2-FCP-NEXT: vpblendd {{.*#+}} ymm5 = ymm7[0,1,2,3,4,5],ymm5[6,7]
; AVX2-FCP-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm5[4,5,6,7]
; AVX2-FCP-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT: vpblendw {{.*#+}} ymm3 = mem[0],ymm1[1,2,3],mem[4],ymm1[5,6,7],mem[8],ymm1[9,10,11],mem[12],ymm1[13,14,15]
; AVX2-FCP-NEXT: vextracti128 $1, %ymm3, %xmm5
; AVX2-FCP-NEXT: vpackusdw %xmm5, %xmm3, %xmm3
; AVX2-FCP-NEXT: vpblendw {{.*#+}} ymm1 = mem[0],ymm1[1,2,3],mem[4],ymm1[5,6,7],mem[8],ymm1[9,10,11],mem[12],ymm1[13,14,15]
; AVX2-FCP-NEXT: vextracti128 $1, %ymm1, %xmm5
; AVX2-FCP-NEXT: vpackusdw %xmm5, %xmm1, %xmm1
; AVX2-FCP-NEXT: vpackusdw %xmm3, %xmm1, %xmm1
; AVX2-FCP-NEXT: vmovdqa 192(%rdi), %ymm5
; AVX2-FCP-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT: vmovdqa 224(%rdi), %ymm3
; AVX2-FCP-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT: vpermd %ymm3, %ymm0, %ymm9
; AVX2-FCP-NEXT: vpermd %ymm5, %ymm0, %ymm3
; AVX2-FCP-NEXT: vpshufb %ymm4, %ymm9, %ymm0
; AVX2-FCP-NEXT: vpshufb %ymm4, %ymm3, %ymm5
; AVX2-FCP-NEXT: vpblendd {{.*#+}} ymm0 = ymm5[0,1,2,3,4,5],ymm0[6,7]
; AVX2-FCP-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FCP-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT: vmovdqa 288(%rdi), %xmm1
; AVX2-FCP-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FCP-NEXT: vmovdqa 304(%rdi), %xmm12
; AVX2-FCP-NEXT: vmovdqa {{.*#+}} xmm5 = [0,1,2,3,2,3,10,11,8,9,10,11,12,13,14,15]
; AVX2-FCP-NEXT: vpshufb %xmm5, %xmm12, %xmm0
; AVX2-FCP-NEXT: vpshufb %xmm5, %xmm1, %xmm1
; AVX2-FCP-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX2-FCP-NEXT: vmovdqa 256(%rdi), %xmm10
; AVX2-FCP-NEXT: vmovdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FCP-NEXT: vmovdqa 272(%rdi), %xmm7
; AVX2-FCP-NEXT: vmovdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FCP-NEXT: vmovdqa {{.*#+}} xmm0 = [2,3,10,11,8,9,10,11,8,9,10,11,12,13,14,15]
; AVX2-FCP-NEXT: vpshufb %xmm0, %xmm7, %xmm7
; AVX2-FCP-NEXT: vpshufb %xmm0, %xmm10, %xmm8
; AVX2-FCP-NEXT: vpunpckldq {{.*#+}} xmm7 = xmm8[0],xmm7[0],xmm8[1],xmm7[1]
; AVX2-FCP-NEXT: vpblendd {{.*#+}} xmm1 = xmm7[0,1],xmm1[2,3]
; AVX2-FCP-NEXT: vmovdqa {{.*#+}} ymm10 = [2,3,6,7,10,11,14,15,14,15,10,11,12,13,14,15,18,19,22,23,26,27,30,31,18,19,22,23,26,27,30,31]
; AVX2-FCP-NEXT: vpshufb %ymm10, %ymm2, %ymm2
; AVX2-FCP-NEXT: vpshufb %ymm10, %ymm6, %ymm6
; AVX2-FCP-NEXT: vpblendd {{.*#+}} ymm2 = ymm6[0,1,2,3,4,5],ymm2[6,7]
; AVX2-FCP-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX2-FCP-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT: vmovdqa 32(%rdi), %xmm2
; AVX2-FCP-NEXT: vmovdqa 48(%rdi), %xmm1
; AVX2-FCP-NEXT: vpshufb %xmm5, %xmm1, %xmm6
; AVX2-FCP-NEXT: vpshufb %xmm5, %xmm2, %xmm7
; AVX2-FCP-NEXT: vpunpckldq {{.*#+}} xmm6 = xmm7[0],xmm6[0],xmm7[1],xmm6[1]
; AVX2-FCP-NEXT: vmovdqa (%rdi), %xmm7
; AVX2-FCP-NEXT: vmovdqa 16(%rdi), %xmm15
; AVX2-FCP-NEXT: vpshufb %xmm0, %xmm15, %xmm11
; AVX2-FCP-NEXT: vpshufb %xmm0, %xmm7, %xmm13
; AVX2-FCP-NEXT: vpunpckldq {{.*#+}} xmm11 = xmm13[0],xmm11[0],xmm13[1],xmm11[1]
; AVX2-FCP-NEXT: vpblendd {{.*#+}} xmm6 = xmm11[0,1],xmm6[2,3]
; AVX2-FCP-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
; AVX2-FCP-NEXT: vpshufb %ymm10, %ymm8, %ymm11
; AVX2-FCP-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
; AVX2-FCP-NEXT: vpshufb %ymm10, %ymm8, %ymm13
; AVX2-FCP-NEXT: vpblendd {{.*#+}} ymm11 = ymm13[0,1,2,3,4,5],ymm11[6,7]
; AVX2-FCP-NEXT: vpblendd {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm11[4,5,6,7]
; AVX2-FCP-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT: vmovdqa 160(%rdi), %xmm8
; AVX2-FCP-NEXT: vmovdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FCP-NEXT: vmovdqa 176(%rdi), %xmm6
; AVX2-FCP-NEXT: vmovdqa %xmm6, (%rsp) # 16-byte Spill
; AVX2-FCP-NEXT: vpshufb %xmm5, %xmm6, %xmm6
; AVX2-FCP-NEXT: vpshufb %xmm5, %xmm8, %xmm11
; AVX2-FCP-NEXT: vpunpckldq {{.*#+}} xmm6 = xmm11[0],xmm6[0],xmm11[1],xmm6[1]
; AVX2-FCP-NEXT: vmovdqa 128(%rdi), %xmm8
; AVX2-FCP-NEXT: vmovdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FCP-NEXT: vmovdqa 144(%rdi), %xmm11
; AVX2-FCP-NEXT: vmovdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FCP-NEXT: vpshufb %xmm0, %xmm11, %xmm11
; AVX2-FCP-NEXT: vpshufb %xmm0, %xmm8, %xmm13
; AVX2-FCP-NEXT: vpunpckldq {{.*#+}} xmm11 = xmm13[0],xmm11[0],xmm13[1],xmm11[1]
; AVX2-FCP-NEXT: vpblendd {{.*#+}} xmm6 = xmm11[0,1],xmm6[2,3]
; AVX2-FCP-NEXT: vpshufb %ymm10, %ymm9, %ymm9
; AVX2-FCP-NEXT: vpshufb %ymm10, %ymm3, %ymm3
; AVX2-FCP-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm9[6,7]
; AVX2-FCP-NEXT: vpblendd {{.*#+}} ymm3 = ymm6[0,1,2,3],ymm3[4,5,6,7]
; AVX2-FCP-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT: vmovdqa 416(%rdi), %xmm6
; AVX2-FCP-NEXT: vmovdqa 432(%rdi), %xmm13
; AVX2-FCP-NEXT: vpshufb %xmm5, %xmm13, %xmm3
; AVX2-FCP-NEXT: vpshufb %xmm5, %xmm6, %xmm5
; AVX2-FCP-NEXT: vpunpckldq {{.*#+}} xmm9 = xmm5[0],xmm3[0],xmm5[1],xmm3[1]
; AVX2-FCP-NEXT: vmovdqa 384(%rdi), %xmm5
; AVX2-FCP-NEXT: vmovdqa 400(%rdi), %xmm3
; AVX2-FCP-NEXT: vpshufb %xmm0, %xmm3, %xmm11
; AVX2-FCP-NEXT: vpshufb %xmm0, %xmm5, %xmm0
; AVX2-FCP-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm11[0],xmm0[1],xmm11[1]
; AVX2-FCP-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm9[2,3]
; AVX2-FCP-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
; AVX2-FCP-NEXT: vpshufb %ymm10, %ymm8, %ymm9
; AVX2-FCP-NEXT: vpshufb %ymm10, %ymm14, %ymm11
; AVX2-FCP-NEXT: vpblendd {{.*#+}} ymm9 = ymm11[0,1,2,3,4,5],ymm9[6,7]
; AVX2-FCP-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm9[4,5,6,7]
; AVX2-FCP-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT: vpmovsxbd {{.*#+}} ymm0 = [1,3,2,3,1,3,5,7]
; AVX2-FCP-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm8 # 32-byte Folded Reload
; AVX2-FCP-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm11 # 32-byte Folded Reload
; AVX2-FCP-NEXT: vmovdqu %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT: vpshufb %ymm4, %ymm8, %ymm9
; AVX2-FCP-NEXT: vpshufb %ymm4, %ymm11, %ymm11
; AVX2-FCP-NEXT: vpblendd {{.*#+}} ymm9 = ymm11[0,1,2,3,4,5],ymm9[6,7]
; AVX2-FCP-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[3,1,2,3]
; AVX2-FCP-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FCP-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[3,1,2,3]
; AVX2-FCP-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[0,1,2,0,4,5,6,7]
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[0,1,2,0,4,5,6,7]
; AVX2-FCP-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; AVX2-FCP-NEXT: vpshufd {{.*#+}} xmm2 = xmm15[3,1,2,3]
; AVX2-FCP-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FCP-NEXT: vpshufd {{.*#+}} xmm7 = xmm7[3,1,2,3]
; AVX2-FCP-NEXT: vmovdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[2,0,2,3,4,5,6,7]
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm7 = xmm7[2,0,2,3,4,5,6,7]
; AVX2-FCP-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm7[0],xmm2[0],xmm7[1],xmm2[1]
; AVX2-FCP-NEXT: vpblendd {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
; AVX2-FCP-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm9[4,5,6,7]
; AVX2-FCP-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
; AVX2-FCP-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
; AVX2-FCP-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT: vpshufb %ymm4, %ymm1, %ymm1
; AVX2-FCP-NEXT: vpshufb %ymm4, %ymm2, %ymm2
; AVX2-FCP-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FCP-NEXT: vpshufd {{.*#+}} xmm2 = xmm12[3,1,2,3]
; AVX2-FCP-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FCP-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
; AVX2-FCP-NEXT: # xmm14 = mem[3,1,2,3]
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[0,1,2,0,4,5,6,7]
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm7 = xmm14[0,1,2,0,4,5,6,7]
; AVX2-FCP-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm7[0],xmm2[0],xmm7[1],xmm2[1]
; AVX2-FCP-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
; AVX2-FCP-NEXT: # xmm11 = mem[3,1,2,3]
; AVX2-FCP-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
; AVX2-FCP-NEXT: # xmm9 = mem[3,1,2,3]
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm12 = xmm11[2,0,2,3,4,5,6,7]
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm15 = xmm9[2,0,2,3,4,5,6,7]
; AVX2-FCP-NEXT: vpunpckldq {{.*#+}} xmm12 = xmm15[0],xmm12[0],xmm15[1],xmm12[1]
; AVX2-FCP-NEXT: vpblendd {{.*#+}} xmm2 = xmm12[0,1],xmm2[2,3]
; AVX2-FCP-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FCP-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
; AVX2-FCP-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
; AVX2-FCP-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT: vpshufb %ymm4, %ymm1, %ymm1
; AVX2-FCP-NEXT: vpshufb %ymm4, %ymm2, %ymm2
; AVX2-FCP-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FCP-NEXT: vpshufd {{.*#+}} xmm2 = xmm13[3,1,2,3]
; AVX2-FCP-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FCP-NEXT: vpshufd {{.*#+}} xmm6 = xmm6[3,1,2,3]
; AVX2-FCP-NEXT: vmovdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[0,1,2,0,4,5,6,7]
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm6 = xmm6[0,1,2,0,4,5,6,7]
; AVX2-FCP-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm6[0],xmm2[0],xmm6[1],xmm2[1]
; AVX2-FCP-NEXT: vpshufd {{.*#+}} xmm13 = xmm3[3,1,2,3]
; AVX2-FCP-NEXT: vpshufd {{.*#+}} xmm12 = xmm5[3,1,2,3]
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm3 = xmm13[2,0,2,3,4,5,6,7]
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm5 = xmm12[2,0,2,3,4,5,6,7]
; AVX2-FCP-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm5[0],xmm3[0],xmm5[1],xmm3[1]
; AVX2-FCP-NEXT: vpblendd {{.*#+}} xmm2 = xmm3[0,1],xmm2[2,3]
; AVX2-FCP-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FCP-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm8 # 32-byte Folded Reload
; AVX2-FCP-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm7 # 32-byte Folded Reload
; AVX2-FCP-NEXT: vpshufb %ymm4, %ymm8, %ymm0
; AVX2-FCP-NEXT: vpshufb %ymm4, %ymm7, %ymm1
; AVX2-FCP-NEXT: vpblendd {{.*#+}} ymm3 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-FCP-NEXT: vpshufd $231, (%rsp), %xmm6 # 16-byte Folded Reload
; AVX2-FCP-NEXT: # xmm6 = mem[3,1,2,3]
; AVX2-FCP-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
; AVX2-FCP-NEXT: # xmm5 = mem[3,1,2,3]
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm1 = xmm6[0,1,2,0,4,5,6,7]
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm2 = xmm5[0,1,2,0,4,5,6,7]
; AVX2-FCP-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; AVX2-FCP-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; AVX2-FCP-NEXT: # xmm4 = mem[3,1,2,3]
; AVX2-FCP-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; AVX2-FCP-NEXT: # xmm2 = mem[3,1,2,3]
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm0 = xmm4[2,0,2,3,4,5,6,7]
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm15 = xmm2[2,0,2,3,4,5,6,7]
; AVX2-FCP-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm15[0],xmm0[0],xmm15[1],xmm0[1]
; AVX2-FCP-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
; AVX2-FCP-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
; AVX2-FCP-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FCP-NEXT: vpshufb %ymm10, %ymm0, %ymm0
; AVX2-FCP-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT: vpshufb %ymm10, %ymm1, %ymm1
; AVX2-FCP-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-FCP-NEXT: vpshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; AVX2-FCP-NEXT: # xmm1 = mem[0,1,3,1,4,5,6,7]
; AVX2-FCP-NEXT: vpshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; AVX2-FCP-NEXT: # xmm3 = mem[0,1,3,1,4,5,6,7]
; AVX2-FCP-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; AVX2-FCP-NEXT: vpshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; AVX2-FCP-NEXT: # xmm3 = mem[3,1,2,3,4,5,6,7]
; AVX2-FCP-NEXT: vpshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
; AVX2-FCP-NEXT: # xmm15 = mem[3,1,2,3,4,5,6,7]
; AVX2-FCP-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm15[0],xmm3[0],xmm15[1],xmm3[1]
; AVX2-FCP-NEXT: vpblendd {{.*#+}} xmm1 = xmm3[0,1],xmm1[2,3]
; AVX2-FCP-NEXT: vpblendd {{.*#+}} ymm15 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FCP-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FCP-NEXT: vpshufb %ymm10, %ymm0, %ymm0
; AVX2-FCP-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT: vpshufb %ymm10, %ymm1, %ymm1
; AVX2-FCP-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-FCP-NEXT: vpshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; AVX2-FCP-NEXT: # xmm1 = mem[0,1,3,1,4,5,6,7]
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm3 = xmm14[0,1,3,1,4,5,6,7]
; AVX2-FCP-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm3 = xmm11[3,1,2,3,4,5,6,7]
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm9 = xmm9[3,1,2,3,4,5,6,7]
; AVX2-FCP-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm9[0],xmm3[0],xmm9[1],xmm3[1]
; AVX2-FCP-NEXT: vpblendd {{.*#+}} xmm1 = xmm3[0,1],xmm1[2,3]
; AVX2-FCP-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FCP-NEXT: vpshufb %ymm10, %ymm8, %ymm1
; AVX2-FCP-NEXT: vpshufb %ymm10, %ymm7, %ymm3
; AVX2-FCP-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm3 = xmm6[0,1,3,1,4,5,6,7]
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm5 = xmm5[0,1,3,1,4,5,6,7]
; AVX2-FCP-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm5[0],xmm3[0],xmm5[1],xmm3[1]
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[3,1,2,3,4,5,6,7]
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[3,1,2,3,4,5,6,7]
; AVX2-FCP-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1]
; AVX2-FCP-NEXT: vpblendd {{.*#+}} xmm2 = xmm2[0,1],xmm3[2,3]
; AVX2-FCP-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FCP-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FCP-NEXT: vpshufb %ymm10, %ymm2, %ymm2
; AVX2-FCP-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-FCP-NEXT: vpshufb %ymm10, %ymm3, %ymm3
; AVX2-FCP-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
; AVX2-FCP-NEXT: vpshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; AVX2-FCP-NEXT: # xmm3 = mem[0,1,3,1,4,5,6,7]
; AVX2-FCP-NEXT: vpshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; AVX2-FCP-NEXT: # xmm4 = mem[0,1,3,1,4,5,6,7]
; AVX2-FCP-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm4 = xmm13[3,1,2,3,4,5,6,7]
; AVX2-FCP-NEXT: vpshuflw {{.*#+}} xmm5 = xmm12[3,1,2,3,4,5,6,7]
; AVX2-FCP-NEXT: vpunpckldq {{.*#+}} xmm4 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
; AVX2-FCP-NEXT: vpblendd {{.*#+}} xmm3 = xmm4[0,1],xmm3[2,3]
; AVX2-FCP-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
; AVX2-FCP-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-FCP-NEXT: vmovaps %ymm3, 32(%rsi)
; AVX2-FCP-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-FCP-NEXT: vmovaps %ymm3, 96(%rsi)
; AVX2-FCP-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-FCP-NEXT: vmovaps %ymm3, 64(%rsi)
; AVX2-FCP-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-FCP-NEXT: vmovaps %ymm3, (%rsi)
; AVX2-FCP-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-FCP-NEXT: vmovaps %ymm3, 96(%rdx)
; AVX2-FCP-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-FCP-NEXT: vmovaps %ymm3, 32(%rdx)
; AVX2-FCP-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-FCP-NEXT: vmovaps %ymm3, (%rdx)
; AVX2-FCP-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-FCP-NEXT: vmovaps %ymm3, 64(%rdx)
; AVX2-FCP-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-FCP-NEXT: vmovaps %ymm3, 32(%rcx)
; AVX2-FCP-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-FCP-NEXT: vmovaps %ymm3, 96(%rcx)
; AVX2-FCP-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-FCP-NEXT: vmovaps %ymm3, 64(%rcx)
; AVX2-FCP-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-FCP-NEXT: vmovaps %ymm3, (%rcx)
; AVX2-FCP-NEXT: vmovdqa %ymm2, 96(%r8)
; AVX2-FCP-NEXT: vmovdqa %ymm1, 32(%r8)
; AVX2-FCP-NEXT: vmovdqa %ymm0, 64(%r8)
; AVX2-FCP-NEXT: vmovdqa %ymm15, (%r8)
; AVX2-FCP-NEXT: addq $680, %rsp # imm = 0x2A8
; AVX2-FCP-NEXT: vzeroupper
; AVX2-FCP-NEXT: retq
;
; AVX512-LABEL: load_i16_stride4_vf64:
; AVX512: # %bb.0:
; AVX512-NEXT: subq $200, %rsp
; AVX512-NEXT: vmovdqa64 256(%rdi), %zmm26
; AVX512-NEXT: vmovdqa64 384(%rdi), %zmm27
; AVX512-NEXT: vmovdqa64 (%rdi), %zmm28
; AVX512-NEXT: vmovdqa64 128(%rdi), %zmm29
; AVX512-NEXT: vmovdqa 192(%rdi), %ymm0
; AVX512-NEXT: vpmovqw %ymm0, %xmm0
; AVX512-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX512-NEXT: vmovdqa 240(%rdi), %xmm14
; AVX512-NEXT: vpshufd {{.*#+}} xmm9 = xmm14[0,2,2,3]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm1 = xmm9[0,1,0,2,4,5,6,7]
; AVX512-NEXT: vmovdqa 224(%rdi), %xmm13
; AVX512-NEXT: vpshufd {{.*#+}} xmm6 = xmm13[0,2,2,3]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm3 = xmm6[0,1,0,2,4,5,6,7]
; AVX512-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; AVX512-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX512-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX512-NEXT: vpmovqw %zmm29, %xmm1
; AVX512-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX512-NEXT: vmovdqa 64(%rdi), %ymm1
; AVX512-NEXT: vpmovqw %ymm1, %xmm1
; AVX512-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX512-NEXT: vmovdqa 112(%rdi), %xmm12
; AVX512-NEXT: vpshufd {{.*#+}} xmm3 = xmm12[0,2,2,3]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm4 = xmm3[0,1,0,2,4,5,6,7]
; AVX512-NEXT: vmovdqa 96(%rdi), %xmm11
; AVX512-NEXT: vpshufd {{.*#+}} xmm7 = xmm11[0,2,2,3]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm5 = xmm7[0,1,0,2,4,5,6,7]
; AVX512-NEXT: vpunpckldq {{.*#+}} xmm4 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
; AVX512-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4
; AVX512-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm4[6,7]
; AVX512-NEXT: vpmovqw %zmm28, %xmm4
; AVX512-NEXT: vpblendd {{.*#+}} ymm1 = ymm4[0,1,2,3],ymm1[4,5,6,7]
; AVX512-NEXT: vshufi64x2 {{.*#+}} zmm0 = zmm1[0,1,2,3],zmm0[0,1,2,3]
; AVX512-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT: vmovdqa 448(%rdi), %ymm0
; AVX512-NEXT: vpmovqw %ymm0, %xmm0
; AVX512-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX512-NEXT: vmovdqa64 496(%rdi), %xmm24
; AVX512-NEXT: vpshufd {{.*#+}} xmm4 = xmm24[0,2,2,3]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm1 = xmm4[0,1,0,2,4,5,6,7]
; AVX512-NEXT: vmovdqa64 480(%rdi), %xmm23
; AVX512-NEXT: vpshufd {{.*#+}} xmm15 = xmm23[0,2,2,3]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm5 = xmm15[0,1,0,2,4,5,6,7]
; AVX512-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm5[0],xmm1[0],xmm5[1],xmm1[1]
; AVX512-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX512-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX512-NEXT: vpmovqw %zmm27, %xmm1
; AVX512-NEXT: vpblendd {{.*#+}} ymm2 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX512-NEXT: vmovdqa 320(%rdi), %ymm1
; AVX512-NEXT: vpmovqw %ymm1, %xmm1
; AVX512-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX512-NEXT: vmovdqa64 368(%rdi), %xmm31
; AVX512-NEXT: vpshufd {{.*#+}} xmm5 = xmm31[0,2,2,3]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm0 = xmm5[0,1,0,2,4,5,6,7]
; AVX512-NEXT: vmovdqa64 352(%rdi), %xmm25
; AVX512-NEXT: vpshufd {{.*#+}} xmm8 = xmm25[0,2,2,3]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm10 = xmm8[0,1,0,2,4,5,6,7]
; AVX512-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm10[0],xmm0[0],xmm10[1],xmm0[1]
; AVX512-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX512-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX512-NEXT: vpmovqw %zmm26, %xmm1
; AVX512-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX512-NEXT: vshufi64x2 {{.*#+}} zmm0 = zmm0[0,1,2,3],zmm2[0,1,2,3]
; AVX512-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT: vmovdqa64 320(%rdi), %xmm30
; AVX512-NEXT: vmovdqa64 336(%rdi), %xmm17
; AVX512-NEXT: vmovdqa64 448(%rdi), %xmm18
; AVX512-NEXT: vmovdqa64 464(%rdi), %xmm19
; AVX512-NEXT: vmovdqa64 64(%rdi), %xmm20
; AVX512-NEXT: vmovdqa64 80(%rdi), %xmm21
; AVX512-NEXT: vmovdqa 192(%rdi), %xmm0
; AVX512-NEXT: vmovdqa 208(%rdi), %xmm1
; AVX512-NEXT: vpshuflw {{.*#+}} xmm2 = xmm9[0,1,1,3,4,5,6,7]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm6 = xmm6[0,1,1,3,4,5,6,7]
; AVX512-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm6[0],xmm2[0],xmm6[1],xmm2[1]
; AVX512-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX512-NEXT: vpshufd {{.*#+}} xmm6 = xmm1[0,2,2,3]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm6 = xmm6[1,3,2,3,4,5,6,7]
; AVX512-NEXT: vpshufd {{.*#+}} xmm9 = xmm0[0,2,2,3]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm9 = xmm9[1,3,2,3,4,5,6,7]
; AVX512-NEXT: vpunpckldq {{.*#+}} xmm6 = xmm9[0],xmm6[0],xmm9[1],xmm6[1]
; AVX512-NEXT: vinserti128 $1, %xmm6, %ymm0, %ymm6
; AVX512-NEXT: vpblendd {{.*#+}} ymm2 = ymm6[0,1,2,3,4,5],ymm2[6,7]
; AVX512-NEXT: vpsrlq $16, %zmm29, %zmm6
; AVX512-NEXT: vpmovqw %zmm6, %xmm6
; AVX512-NEXT: vpblendd {{.*#+}} ymm2 = ymm6[0,1,2,3],ymm2[4,5,6,7]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[0,1,1,3,4,5,6,7]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm6 = xmm7[0,1,1,3,4,5,6,7]
; AVX512-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm6[0],xmm3[0],xmm6[1],xmm3[1]
; AVX512-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
; AVX512-NEXT: vpshufd {{.*#+}} xmm6 = xmm21[0,2,2,3]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm6 = xmm6[1,3,2,3,4,5,6,7]
; AVX512-NEXT: vpshufd {{.*#+}} xmm7 = xmm20[0,2,2,3]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm7 = xmm7[1,3,2,3,4,5,6,7]
; AVX512-NEXT: vpunpckldq {{.*#+}} xmm6 = xmm7[0],xmm6[0],xmm7[1],xmm6[1]
; AVX512-NEXT: vinserti128 $1, %xmm6, %ymm0, %ymm6
; AVX512-NEXT: vpblendd {{.*#+}} ymm3 = ymm6[0,1,2,3,4,5],ymm3[6,7]
; AVX512-NEXT: vpsrlq $16, %zmm28, %zmm6
; AVX512-NEXT: vpmovqw %zmm6, %xmm6
; AVX512-NEXT: vpblendd {{.*#+}} ymm3 = ymm6[0,1,2,3],ymm3[4,5,6,7]
; AVX512-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm3[0,1,2,3],zmm2[0,1,2,3]
; AVX512-NEXT: vmovdqu64 %zmm2, (%rsp) # 64-byte Spill
; AVX512-NEXT: vpshuflw {{.*#+}} xmm2 = xmm4[0,1,1,3,4,5,6,7]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm3 = xmm15[0,1,1,3,4,5,6,7]
; AVX512-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
; AVX512-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX512-NEXT: vpshufd {{.*#+}} xmm3 = xmm19[0,2,2,3]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[1,3,2,3,4,5,6,7]
; AVX512-NEXT: vpshufd {{.*#+}} xmm4 = xmm18[0,2,2,3]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[1,3,2,3,4,5,6,7]
; AVX512-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
; AVX512-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
; AVX512-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
; AVX512-NEXT: vpsrlq $16, %zmm27, %zmm3
; AVX512-NEXT: vpmovqw %zmm3, %xmm3
; AVX512-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm3 = xmm5[0,1,1,3,4,5,6,7]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm4 = xmm8[0,1,1,3,4,5,6,7]
; AVX512-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
; AVX512-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
; AVX512-NEXT: vpshufd {{.*#+}} xmm4 = xmm17[0,2,2,3]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[1,3,2,3,4,5,6,7]
; AVX512-NEXT: vpshufd {{.*#+}} xmm5 = xmm30[0,2,2,3]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm5 = xmm5[1,3,2,3,4,5,6,7]
; AVX512-NEXT: vpunpckldq {{.*#+}} xmm4 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
; AVX512-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4
; AVX512-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3,4,5],ymm3[6,7]
; AVX512-NEXT: vpsrlq $16, %zmm26, %zmm4
; AVX512-NEXT: vpmovqw %zmm4, %xmm4
; AVX512-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
; AVX512-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm3[0,1,2,3],zmm2[0,1,2,3]
; AVX512-NEXT: vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT: vpshufd {{.*#+}} xmm2 = xmm14[3,1,2,3]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm3 = xmm2[0,1,2,0,4,5,6,7]
; AVX512-NEXT: vmovdqa64 %xmm2, %xmm22
; AVX512-NEXT: vpshufd {{.*#+}} xmm13 = xmm13[3,1,2,3]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm4 = xmm13[0,1,2,0,4,5,6,7]
; AVX512-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
; AVX512-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
; AVX512-NEXT: vpshufd {{.*#+}} xmm6 = xmm1[3,1,2,3]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm1 = xmm6[2,0,2,3,4,5,6,7]
; AVX512-NEXT: vpshufd {{.*#+}} xmm7 = xmm0[3,1,2,3]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm0 = xmm7[2,0,2,3,4,5,6,7]
; AVX512-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; AVX512-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX512-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm3[6,7]
; AVX512-NEXT: vpsrlq $32, %zmm29, %zmm1
; AVX512-NEXT: vpmovqw %zmm1, %xmm1
; AVX512-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX512-NEXT: vpshufd {{.*#+}} xmm2 = xmm12[3,1,2,3]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm4 = xmm2[0,1,2,0,4,5,6,7]
; AVX512-NEXT: vpshufd {{.*#+}} xmm3 = xmm11[3,1,2,3]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm5 = xmm3[0,1,2,0,4,5,6,7]
; AVX512-NEXT: vpunpckldq {{.*#+}} xmm4 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
; AVX512-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4
; AVX512-NEXT: vpshufd {{.*#+}} xmm0 = xmm21[3,1,2,3]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm8 = xmm0[2,0,2,3,4,5,6,7]
; AVX512-NEXT: vmovdqa64 %xmm0, %xmm21
; AVX512-NEXT: vpshufd {{.*#+}} xmm0 = xmm20[3,1,2,3]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm9 = xmm0[2,0,2,3,4,5,6,7]
; AVX512-NEXT: vmovdqa64 %xmm0, %xmm16
; AVX512-NEXT: vpunpckldq {{.*#+}} xmm8 = xmm9[0],xmm8[0],xmm9[1],xmm8[1]
; AVX512-NEXT: vinserti128 $1, %xmm8, %ymm0, %ymm8
; AVX512-NEXT: vpblendd {{.*#+}} ymm4 = ymm8[0,1,2,3,4,5],ymm4[6,7]
; AVX512-NEXT: vpsrlq $32, %zmm28, %zmm8
; AVX512-NEXT: vpmovqw %zmm8, %xmm8
; AVX512-NEXT: vpblendd {{.*#+}} ymm4 = ymm8[0,1,2,3],ymm4[4,5,6,7]
; AVX512-NEXT: vshufi64x2 {{.*#+}} zmm0 = zmm4[0,1,2,3],zmm1[0,1,2,3]
; AVX512-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT: vpshufd {{.*#+}} xmm10 = xmm24[3,1,2,3]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm1 = xmm10[0,1,2,0,4,5,6,7]
; AVX512-NEXT: vpshufd {{.*#+}} xmm11 = xmm23[3,1,2,3]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm4 = xmm11[0,1,2,0,4,5,6,7]
; AVX512-NEXT: vpunpckldq {{.*#+}} xmm20 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
; AVX512-NEXT: vpshufd {{.*#+}} xmm12 = xmm19[3,1,2,3]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm4 = xmm12[2,0,2,3,4,5,6,7]
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm18[3,1,2,3]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm8 = xmm1[2,0,2,3,4,5,6,7]
; AVX512-NEXT: vpunpckldq {{.*#+}} xmm18 = xmm8[0],xmm4[0],xmm8[1],xmm4[1]
; AVX512-NEXT: vpshufd {{.*#+}} xmm4 = xmm31[3,1,2,3]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm8 = xmm4[0,1,2,0,4,5,6,7]
; AVX512-NEXT: vpshufd {{.*#+}} xmm15 = xmm25[3,1,2,3]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm9 = xmm15[0,1,2,0,4,5,6,7]
; AVX512-NEXT: vpunpckldq {{.*#+}} xmm19 = xmm9[0],xmm8[0],xmm9[1],xmm8[1]
; AVX512-NEXT: vpshufd {{.*#+}} xmm8 = xmm17[3,1,2,3]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm0 = xmm8[2,0,2,3,4,5,6,7]
; AVX512-NEXT: vpshufd {{.*#+}} xmm9 = xmm30[3,1,2,3]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm14 = xmm9[2,0,2,3,4,5,6,7]
; AVX512-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm14[0],xmm0[0],xmm14[1],xmm0[1]
; AVX512-NEXT: vinserti32x4 $1, %xmm20, %ymm0, %ymm14
; AVX512-NEXT: vinserti32x4 $1, %xmm18, %ymm0, %ymm5
; AVX512-NEXT: vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],ymm14[6,7]
; AVX512-NEXT: vpsrlq $32, %zmm27, %zmm14
; AVX512-NEXT: vpmovqw %zmm14, %xmm14
; AVX512-NEXT: vpblendd {{.*#+}} ymm5 = ymm14[0,1,2,3],ymm5[4,5,6,7]
; AVX512-NEXT: vinserti32x4 $1, %xmm19, %ymm0, %ymm14
; AVX512-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX512-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm14[6,7]
; AVX512-NEXT: vpsrlq $32, %zmm26, %zmm14
; AVX512-NEXT: vpmovqw %zmm14, %xmm14
; AVX512-NEXT: vpblendd {{.*#+}} ymm0 = ymm14[0,1,2,3],ymm0[4,5,6,7]
; AVX512-NEXT: vshufi64x2 {{.*#+}} zmm17 = zmm0[0,1,2,3],zmm5[0,1,2,3]
; AVX512-NEXT: vmovdqa64 %xmm22, %xmm0
; AVX512-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[0,1,3,1,4,5,6,7]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm5 = xmm13[0,1,3,1,4,5,6,7]
; AVX512-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm5[0],xmm0[0],xmm5[1],xmm0[1]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm5 = xmm6[3,1,2,3,4,5,6,7]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm6 = xmm7[3,1,2,3,4,5,6,7]
; AVX512-NEXT: vpunpckldq {{.*#+}} xmm5 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[0,1,3,1,4,5,6,7]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[0,1,3,1,4,5,6,7]
; AVX512-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
; AVX512-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX512-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm3
; AVX512-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
; AVX512-NEXT: vpsrlq $48, %zmm29, %zmm3
; AVX512-NEXT: vpmovqw %zmm3, %xmm3
; AVX512-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7]
; AVX512-NEXT: vmovdqa64 %xmm21, %xmm3
; AVX512-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[3,1,2,3,4,5,6,7]
; AVX512-NEXT: vmovdqa64 %xmm16, %xmm5
; AVX512-NEXT: vpshuflw {{.*#+}} xmm5 = xmm5[3,1,2,3,4,5,6,7]
; AVX512-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm5[0],xmm3[0],xmm5[1],xmm3[1]
; AVX512-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX512-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
; AVX512-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
; AVX512-NEXT: vpsrlq $48, %zmm28, %zmm3
; AVX512-NEXT: vpmovqw %zmm3, %xmm3
; AVX512-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
; AVX512-NEXT: vshufi64x2 {{.*#+}} zmm0 = zmm2[0,1,2,3],zmm0[0,1,2,3]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm2 = xmm10[0,1,3,1,4,5,6,7]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm3 = xmm11[0,1,3,1,4,5,6,7]
; AVX512-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm3 = xmm12[3,1,2,3,4,5,6,7]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[3,1,2,3,4,5,6,7]
; AVX512-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm3 = xmm4[0,1,3,1,4,5,6,7]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm4 = xmm15[0,1,3,1,4,5,6,7]
; AVX512-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
; AVX512-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX512-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX512-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
; AVX512-NEXT: vpsrlq $48, %zmm27, %zmm2
; AVX512-NEXT: vpmovqw %zmm2, %xmm2
; AVX512-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm2 = xmm8[3,1,2,3,4,5,6,7]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm4 = xmm9[3,1,2,3,4,5,6,7]
; AVX512-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm4[0],xmm2[0],xmm4[1],xmm2[1]
; AVX512-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
; AVX512-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX512-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm3[6,7]
; AVX512-NEXT: vpsrlq $48, %zmm26, %zmm3
; AVX512-NEXT: vpmovqw %zmm3, %xmm3
; AVX512-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
; AVX512-NEXT: vshufi64x2 {{.*#+}} zmm1 = zmm2[0,1,2,3],zmm1[0,1,2,3]
; AVX512-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512-NEXT: vmovaps %zmm2, 64(%rsi)
; AVX512-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512-NEXT: vmovaps %zmm2, (%rsi)
; AVX512-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512-NEXT: vmovaps %zmm2, 64(%rdx)
; AVX512-NEXT: vmovups (%rsp), %zmm2 # 64-byte Reload
; AVX512-NEXT: vmovaps %zmm2, (%rdx)
; AVX512-NEXT: vmovdqa64 %zmm17, 64(%rcx)
; AVX512-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512-NEXT: vmovaps %zmm2, (%rcx)
; AVX512-NEXT: vmovdqa64 %zmm1, 64(%r8)
; AVX512-NEXT: vmovdqa64 %zmm0, (%r8)
; AVX512-NEXT: addq $200, %rsp
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
;
; AVX512-FCP-LABEL: load_i16_stride4_vf64:
; AVX512-FCP: # %bb.0:
; AVX512-FCP-NEXT: vmovdqa64 256(%rdi), %zmm22
; AVX512-FCP-NEXT: vmovdqa64 384(%rdi), %zmm25
; AVX512-FCP-NEXT: vmovdqa64 (%rdi), %zmm2
; AVX512-FCP-NEXT: vmovdqa64 128(%rdi), %zmm4
; AVX512-FCP-NEXT: vbroadcasti128 {{.*#+}} ymm5 = [16,17,20,21,0,0,0,0,24,25,28,29,0,0,0,0,16,17,20,21,0,0,0,0,24,25,28,29,0,0,0,0]
; AVX512-FCP-NEXT: # ymm5 = mem[0,1,0,1]
; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} ymm9 = [0,2,2,3,0,2,4,6]
; AVX512-FCP-NEXT: vmovdqa64 224(%rdi), %ymm23
; AVX512-FCP-NEXT: vpermd %ymm23, %ymm9, %ymm10
; AVX512-FCP-NEXT: vpshufb %ymm5, %ymm10, %ymm0
; AVX512-FCP-NEXT: vmovdqa {{.*#+}} ymm6 = [0,1,4,5,u,u,u,u,8,9,12,13,12,13,14,15,16,17,20,21,u,u,u,u,24,25,28,29,u,u,u,u]
; AVX512-FCP-NEXT: vmovdqa64 192(%rdi), %ymm24
; AVX512-FCP-NEXT: vpermd %ymm24, %ymm9, %ymm3
; AVX512-FCP-NEXT: vpshufb %ymm6, %ymm3, %ymm1
; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} ymm7 = [0,2,2,3,4,6,12,14]
; AVX512-FCP-NEXT: vpermt2d %ymm0, %ymm7, %ymm1
; AVX512-FCP-NEXT: vpmovqw %zmm4, %xmm0
; AVX512-FCP-NEXT: vpblendd {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX512-FCP-NEXT: vmovdqa64 96(%rdi), %ymm26
; AVX512-FCP-NEXT: vpermd %ymm26, %ymm9, %ymm8
; AVX512-FCP-NEXT: vpshufb %ymm5, %ymm8, %ymm11
; AVX512-FCP-NEXT: vmovdqa64 64(%rdi), %ymm27
; AVX512-FCP-NEXT: vpermd %ymm27, %ymm9, %ymm0
; AVX512-FCP-NEXT: vpshufb %ymm6, %ymm0, %ymm12
; AVX512-FCP-NEXT: vpermt2d %ymm11, %ymm7, %ymm12
; AVX512-FCP-NEXT: vpmovqw %zmm2, %xmm11
; AVX512-FCP-NEXT: vpblendd {{.*#+}} ymm11 = ymm11[0,1,2,3],ymm12[4,5,6,7]
; AVX512-FCP-NEXT: vshufi64x2 {{.*#+}} zmm20 = zmm11[0,1,2,3],zmm1[0,1,2,3]
; AVX512-FCP-NEXT: vmovdqa64 480(%rdi), %ymm28
; AVX512-FCP-NEXT: vpermd %ymm28, %ymm9, %ymm11
; AVX512-FCP-NEXT: vpshufb %ymm5, %ymm11, %ymm1
; AVX512-FCP-NEXT: vmovdqa64 448(%rdi), %ymm17
; AVX512-FCP-NEXT: vpermd %ymm17, %ymm9, %ymm12
; AVX512-FCP-NEXT: vpshufb %ymm6, %ymm12, %ymm13
; AVX512-FCP-NEXT: vpermt2d %ymm1, %ymm7, %ymm13
; AVX512-FCP-NEXT: vpmovqw %zmm25, %xmm1
; AVX512-FCP-NEXT: vpblendd {{.*#+}} ymm14 = ymm1[0,1,2,3],ymm13[4,5,6,7]
; AVX512-FCP-NEXT: vmovdqa64 352(%rdi), %ymm18
; AVX512-FCP-NEXT: vpermd %ymm18, %ymm9, %ymm1
; AVX512-FCP-NEXT: vpshufb %ymm5, %ymm1, %ymm15
; AVX512-FCP-NEXT: vmovdqa64 320(%rdi), %ymm19
; AVX512-FCP-NEXT: vpermd %ymm19, %ymm9, %ymm13
; AVX512-FCP-NEXT: vpshufb %ymm6, %ymm13, %ymm9
; AVX512-FCP-NEXT: vpermt2d %ymm15, %ymm7, %ymm9
; AVX512-FCP-NEXT: vpmovqw %zmm22, %xmm15
; AVX512-FCP-NEXT: vpblendd {{.*#+}} ymm9 = ymm15[0,1,2,3],ymm9[4,5,6,7]
; AVX512-FCP-NEXT: vshufi64x2 {{.*#+}} zmm21 = zmm9[0,1,2,3],zmm14[0,1,2,3]
; AVX512-FCP-NEXT: vmovdqa {{.*#+}} ymm9 = [2,3,6,7,10,11,14,15,14,15,10,11,12,13,14,15,18,19,22,23,26,27,30,31,18,19,22,23,26,27,30,31]
; AVX512-FCP-NEXT: vpshufb %ymm9, %ymm10, %ymm10
; AVX512-FCP-NEXT: vpshufb %ymm9, %ymm3, %ymm3
; AVX512-FCP-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm10[6,7]
; AVX512-FCP-NEXT: vpsrlq $16, %zmm4, %zmm10
; AVX512-FCP-NEXT: vpmovqw %zmm10, %xmm10
; AVX512-FCP-NEXT: vpblendd {{.*#+}} ymm3 = ymm10[0,1,2,3],ymm3[4,5,6,7]
; AVX512-FCP-NEXT: vpshufb %ymm9, %ymm8, %ymm8
; AVX512-FCP-NEXT: vpshufb %ymm9, %ymm0, %ymm0
; AVX512-FCP-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm8[6,7]
; AVX512-FCP-NEXT: vpsrlq $16, %zmm2, %zmm8
; AVX512-FCP-NEXT: vpmovqw %zmm8, %xmm8
; AVX512-FCP-NEXT: vpblendd {{.*#+}} ymm0 = ymm8[0,1,2,3],ymm0[4,5,6,7]
; AVX512-FCP-NEXT: vshufi64x2 {{.*#+}} zmm29 = zmm0[0,1,2,3],zmm3[0,1,2,3]
; AVX512-FCP-NEXT: vpshufb %ymm9, %ymm11, %ymm0
; AVX512-FCP-NEXT: vpshufb %ymm9, %ymm12, %ymm3
; AVX512-FCP-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
; AVX512-FCP-NEXT: vpsrlq $16, %zmm25, %zmm3
; AVX512-FCP-NEXT: vpmovqw %zmm3, %xmm3
; AVX512-FCP-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7]
; AVX512-FCP-NEXT: vpshufb %ymm9, %ymm1, %ymm1
; AVX512-FCP-NEXT: vpshufb %ymm9, %ymm13, %ymm3
; AVX512-FCP-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm1[6,7]
; AVX512-FCP-NEXT: vpsrlq $16, %zmm22, %zmm3
; AVX512-FCP-NEXT: vpmovqw %zmm3, %xmm3
; AVX512-FCP-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm1[4,5,6,7]
; AVX512-FCP-NEXT: vshufi64x2 {{.*#+}} zmm16 = zmm1[0,1,2,3],zmm0[0,1,2,3]
; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} ymm14 = [1,3,2,3,1,3,5,7]
; AVX512-FCP-NEXT: vpermd %ymm23, %ymm14, %ymm3
; AVX512-FCP-NEXT: vpshufb %ymm5, %ymm3, %ymm0
; AVX512-FCP-NEXT: vpermd %ymm24, %ymm14, %ymm8
; AVX512-FCP-NEXT: vpshufb %ymm6, %ymm8, %ymm1
; AVX512-FCP-NEXT: vpermt2d %ymm0, %ymm7, %ymm1
; AVX512-FCP-NEXT: vpsrlq $32, %zmm4, %zmm0
; AVX512-FCP-NEXT: vpmovqw %zmm0, %xmm0
; AVX512-FCP-NEXT: vpblendd {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX512-FCP-NEXT: vpermd %ymm26, %ymm14, %ymm0
; AVX512-FCP-NEXT: vpshufb %ymm5, %ymm0, %ymm12
; AVX512-FCP-NEXT: vpermd %ymm27, %ymm14, %ymm11
; AVX512-FCP-NEXT: vpshufb %ymm6, %ymm11, %ymm13
; AVX512-FCP-NEXT: vpermt2d %ymm12, %ymm7, %ymm13
; AVX512-FCP-NEXT: vpsrlq $32, %zmm2, %zmm12
; AVX512-FCP-NEXT: vpmovqw %zmm12, %xmm12
; AVX512-FCP-NEXT: vpblendd {{.*#+}} ymm12 = ymm12[0,1,2,3],ymm13[4,5,6,7]
; AVX512-FCP-NEXT: vshufi64x2 {{.*#+}} zmm23 = zmm12[0,1,2,3],zmm1[0,1,2,3]
; AVX512-FCP-NEXT: vpermd %ymm28, %ymm14, %ymm12
; AVX512-FCP-NEXT: vpshufb %ymm5, %ymm12, %ymm1
; AVX512-FCP-NEXT: vpermd %ymm17, %ymm14, %ymm13
; AVX512-FCP-NEXT: vpshufb %ymm6, %ymm13, %ymm10
; AVX512-FCP-NEXT: vpermt2d %ymm1, %ymm7, %ymm10
; AVX512-FCP-NEXT: vpsrlq $32, %zmm25, %zmm1
; AVX512-FCP-NEXT: vpmovqw %zmm1, %xmm1
; AVX512-FCP-NEXT: vpblendd {{.*#+}} ymm10 = ymm1[0,1,2,3],ymm10[4,5,6,7]
; AVX512-FCP-NEXT: vpermd %ymm18, %ymm14, %ymm1
; AVX512-FCP-NEXT: vpshufb %ymm5, %ymm1, %ymm15
; AVX512-FCP-NEXT: vpermd %ymm19, %ymm14, %ymm5
; AVX512-FCP-NEXT: vpshufb %ymm6, %ymm5, %ymm6
; AVX512-FCP-NEXT: vpermt2d %ymm15, %ymm7, %ymm6
; AVX512-FCP-NEXT: vpsrlq $32, %zmm22, %zmm7
; AVX512-FCP-NEXT: vpmovqw %zmm7, %xmm7
; AVX512-FCP-NEXT: vpblendd {{.*#+}} ymm6 = ymm7[0,1,2,3],ymm6[4,5,6,7]
; AVX512-FCP-NEXT: vshufi64x2 {{.*#+}} zmm6 = zmm6[0,1,2,3],zmm10[0,1,2,3]
; AVX512-FCP-NEXT: vpshufb %ymm9, %ymm3, %ymm3
; AVX512-FCP-NEXT: vpshufb %ymm9, %ymm8, %ymm7
; AVX512-FCP-NEXT: vpblendd {{.*#+}} ymm3 = ymm7[0,1,2,3,4,5],ymm3[6,7]
; AVX512-FCP-NEXT: vpsrlq $48, %zmm4, %zmm4
; AVX512-FCP-NEXT: vpmovqw %zmm4, %xmm4
; AVX512-FCP-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
; AVX512-FCP-NEXT: vpshufb %ymm9, %ymm0, %ymm0
; AVX512-FCP-NEXT: vpshufb %ymm9, %ymm11, %ymm4
; AVX512-FCP-NEXT: vpblendd {{.*#+}} ymm0 = ymm4[0,1,2,3,4,5],ymm0[6,7]
; AVX512-FCP-NEXT: vpsrlq $48, %zmm2, %zmm2
; AVX512-FCP-NEXT: vpmovqw %zmm2, %xmm2
; AVX512-FCP-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
; AVX512-FCP-NEXT: vshufi64x2 {{.*#+}} zmm0 = zmm0[0,1,2,3],zmm3[0,1,2,3]
; AVX512-FCP-NEXT: vpshufb %ymm9, %ymm12, %ymm2
; AVX512-FCP-NEXT: vpshufb %ymm9, %ymm13, %ymm3
; AVX512-FCP-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
; AVX512-FCP-NEXT: vpsrlq $48, %zmm25, %zmm3
; AVX512-FCP-NEXT: vpmovqw %zmm3, %xmm3
; AVX512-FCP-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
; AVX512-FCP-NEXT: vpshufb %ymm9, %ymm1, %ymm1
; AVX512-FCP-NEXT: vpshufb %ymm9, %ymm5, %ymm3
; AVX512-FCP-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm1[6,7]
; AVX512-FCP-NEXT: vpsrlq $48, %zmm22, %zmm3
; AVX512-FCP-NEXT: vpmovqw %zmm3, %xmm3
; AVX512-FCP-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm1[4,5,6,7]
; AVX512-FCP-NEXT: vshufi64x2 {{.*#+}} zmm1 = zmm1[0,1,2,3],zmm2[0,1,2,3]
; AVX512-FCP-NEXT: vmovdqa64 %zmm21, 64(%rsi)
; AVX512-FCP-NEXT: vmovdqa64 %zmm20, (%rsi)
; AVX512-FCP-NEXT: vmovdqa64 %zmm16, 64(%rdx)
; AVX512-FCP-NEXT: vmovdqa64 %zmm29, (%rdx)
; AVX512-FCP-NEXT: vmovdqa64 %zmm6, 64(%rcx)
; AVX512-FCP-NEXT: vmovdqa64 %zmm23, (%rcx)
; AVX512-FCP-NEXT: vmovdqa64 %zmm1, 64(%r8)
; AVX512-FCP-NEXT: vmovdqa64 %zmm0, (%r8)
; AVX512-FCP-NEXT: vzeroupper
; AVX512-FCP-NEXT: retq
;
; AVX512DQ-LABEL: load_i16_stride4_vf64:
; AVX512DQ: # %bb.0:
; AVX512DQ-NEXT: subq $200, %rsp
; AVX512DQ-NEXT: vmovdqa64 256(%rdi), %zmm26
; AVX512DQ-NEXT: vmovdqa64 384(%rdi), %zmm27
; AVX512DQ-NEXT: vmovdqa64 (%rdi), %zmm28
; AVX512DQ-NEXT: vmovdqa64 128(%rdi), %zmm29
; AVX512DQ-NEXT: vmovdqa 192(%rdi), %ymm0
; AVX512DQ-NEXT: vpmovqw %ymm0, %xmm0
; AVX512DQ-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX512DQ-NEXT: vmovdqa 240(%rdi), %xmm14
; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm9 = xmm14[0,2,2,3]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm1 = xmm9[0,1,0,2,4,5,6,7]
; AVX512DQ-NEXT: vmovdqa 224(%rdi), %xmm13
; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm6 = xmm13[0,2,2,3]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm3 = xmm6[0,1,0,2,4,5,6,7]
; AVX512DQ-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; AVX512DQ-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX512DQ-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX512DQ-NEXT: vpmovqw %zmm29, %xmm1
; AVX512DQ-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX512DQ-NEXT: vmovdqa 64(%rdi), %ymm1
; AVX512DQ-NEXT: vpmovqw %ymm1, %xmm1
; AVX512DQ-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX512DQ-NEXT: vmovdqa 112(%rdi), %xmm12
; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm3 = xmm12[0,2,2,3]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm4 = xmm3[0,1,0,2,4,5,6,7]
; AVX512DQ-NEXT: vmovdqa 96(%rdi), %xmm11
; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm7 = xmm11[0,2,2,3]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm5 = xmm7[0,1,0,2,4,5,6,7]
; AVX512DQ-NEXT: vpunpckldq {{.*#+}} xmm4 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
; AVX512DQ-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4
; AVX512DQ-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm4[6,7]
; AVX512DQ-NEXT: vpmovqw %zmm28, %xmm4
; AVX512DQ-NEXT: vpblendd {{.*#+}} ymm1 = ymm4[0,1,2,3],ymm1[4,5,6,7]
; AVX512DQ-NEXT: vshufi64x2 {{.*#+}} zmm0 = zmm1[0,1,2,3],zmm0[0,1,2,3]
; AVX512DQ-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT: vmovdqa 448(%rdi), %ymm0
; AVX512DQ-NEXT: vpmovqw %ymm0, %xmm0
; AVX512DQ-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX512DQ-NEXT: vmovdqa64 496(%rdi), %xmm24
; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm4 = xmm24[0,2,2,3]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm1 = xmm4[0,1,0,2,4,5,6,7]
; AVX512DQ-NEXT: vmovdqa64 480(%rdi), %xmm23
; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm15 = xmm23[0,2,2,3]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm5 = xmm15[0,1,0,2,4,5,6,7]
; AVX512DQ-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm5[0],xmm1[0],xmm5[1],xmm1[1]
; AVX512DQ-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX512DQ-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX512DQ-NEXT: vpmovqw %zmm27, %xmm1
; AVX512DQ-NEXT: vpblendd {{.*#+}} ymm2 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX512DQ-NEXT: vmovdqa 320(%rdi), %ymm1
; AVX512DQ-NEXT: vpmovqw %ymm1, %xmm1
; AVX512DQ-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX512DQ-NEXT: vmovdqa64 368(%rdi), %xmm31
; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm5 = xmm31[0,2,2,3]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm0 = xmm5[0,1,0,2,4,5,6,7]
; AVX512DQ-NEXT: vmovdqa64 352(%rdi), %xmm25
; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm8 = xmm25[0,2,2,3]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm10 = xmm8[0,1,0,2,4,5,6,7]
; AVX512DQ-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm10[0],xmm0[0],xmm10[1],xmm0[1]
; AVX512DQ-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX512DQ-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX512DQ-NEXT: vpmovqw %zmm26, %xmm1
; AVX512DQ-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX512DQ-NEXT: vshufi64x2 {{.*#+}} zmm0 = zmm0[0,1,2,3],zmm2[0,1,2,3]
; AVX512DQ-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT: vmovdqa64 320(%rdi), %xmm30
; AVX512DQ-NEXT: vmovdqa64 336(%rdi), %xmm17
; AVX512DQ-NEXT: vmovdqa64 448(%rdi), %xmm18
; AVX512DQ-NEXT: vmovdqa64 464(%rdi), %xmm19
; AVX512DQ-NEXT: vmovdqa64 64(%rdi), %xmm20
; AVX512DQ-NEXT: vmovdqa64 80(%rdi), %xmm21
; AVX512DQ-NEXT: vmovdqa 192(%rdi), %xmm0
; AVX512DQ-NEXT: vmovdqa 208(%rdi), %xmm1
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm2 = xmm9[0,1,1,3,4,5,6,7]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm6 = xmm6[0,1,1,3,4,5,6,7]
; AVX512DQ-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm6[0],xmm2[0],xmm6[1],xmm2[1]
; AVX512DQ-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm6 = xmm1[0,2,2,3]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm6 = xmm6[1,3,2,3,4,5,6,7]
; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm9 = xmm0[0,2,2,3]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm9 = xmm9[1,3,2,3,4,5,6,7]
; AVX512DQ-NEXT: vpunpckldq {{.*#+}} xmm6 = xmm9[0],xmm6[0],xmm9[1],xmm6[1]
; AVX512DQ-NEXT: vinserti128 $1, %xmm6, %ymm0, %ymm6
; AVX512DQ-NEXT: vpblendd {{.*#+}} ymm2 = ymm6[0,1,2,3,4,5],ymm2[6,7]
; AVX512DQ-NEXT: vpsrlq $16, %zmm29, %zmm6
; AVX512DQ-NEXT: vpmovqw %zmm6, %xmm6
; AVX512DQ-NEXT: vpblendd {{.*#+}} ymm2 = ymm6[0,1,2,3],ymm2[4,5,6,7]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[0,1,1,3,4,5,6,7]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm6 = xmm7[0,1,1,3,4,5,6,7]
; AVX512DQ-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm6[0],xmm3[0],xmm6[1],xmm3[1]
; AVX512DQ-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm6 = xmm21[0,2,2,3]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm6 = xmm6[1,3,2,3,4,5,6,7]
; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm7 = xmm20[0,2,2,3]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm7 = xmm7[1,3,2,3,4,5,6,7]
; AVX512DQ-NEXT: vpunpckldq {{.*#+}} xmm6 = xmm7[0],xmm6[0],xmm7[1],xmm6[1]
; AVX512DQ-NEXT: vinserti128 $1, %xmm6, %ymm0, %ymm6
; AVX512DQ-NEXT: vpblendd {{.*#+}} ymm3 = ymm6[0,1,2,3,4,5],ymm3[6,7]
; AVX512DQ-NEXT: vpsrlq $16, %zmm28, %zmm6
; AVX512DQ-NEXT: vpmovqw %zmm6, %xmm6
; AVX512DQ-NEXT: vpblendd {{.*#+}} ymm3 = ymm6[0,1,2,3],ymm3[4,5,6,7]
; AVX512DQ-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm3[0,1,2,3],zmm2[0,1,2,3]
; AVX512DQ-NEXT: vmovdqu64 %zmm2, (%rsp) # 64-byte Spill
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm2 = xmm4[0,1,1,3,4,5,6,7]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm3 = xmm15[0,1,1,3,4,5,6,7]
; AVX512DQ-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
; AVX512DQ-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm3 = xmm19[0,2,2,3]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[1,3,2,3,4,5,6,7]
; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm4 = xmm18[0,2,2,3]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[1,3,2,3,4,5,6,7]
; AVX512DQ-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
; AVX512DQ-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
; AVX512DQ-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
; AVX512DQ-NEXT: vpsrlq $16, %zmm27, %zmm3
; AVX512DQ-NEXT: vpmovqw %zmm3, %xmm3
; AVX512DQ-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm3 = xmm5[0,1,1,3,4,5,6,7]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm4 = xmm8[0,1,1,3,4,5,6,7]
; AVX512DQ-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
; AVX512DQ-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm4 = xmm17[0,2,2,3]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[1,3,2,3,4,5,6,7]
; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm5 = xmm30[0,2,2,3]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm5 = xmm5[1,3,2,3,4,5,6,7]
; AVX512DQ-NEXT: vpunpckldq {{.*#+}} xmm4 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
; AVX512DQ-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4
; AVX512DQ-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3,4,5],ymm3[6,7]
; AVX512DQ-NEXT: vpsrlq $16, %zmm26, %zmm4
; AVX512DQ-NEXT: vpmovqw %zmm4, %xmm4
; AVX512DQ-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
; AVX512DQ-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm3[0,1,2,3],zmm2[0,1,2,3]
; AVX512DQ-NEXT: vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm2 = xmm14[3,1,2,3]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm3 = xmm2[0,1,2,0,4,5,6,7]
; AVX512DQ-NEXT: vmovdqa64 %xmm2, %xmm22
; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm13 = xmm13[3,1,2,3]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm4 = xmm13[0,1,2,0,4,5,6,7]
; AVX512DQ-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
; AVX512DQ-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm6 = xmm1[3,1,2,3]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm1 = xmm6[2,0,2,3,4,5,6,7]
; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm7 = xmm0[3,1,2,3]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm0 = xmm7[2,0,2,3,4,5,6,7]
; AVX512DQ-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; AVX512DQ-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX512DQ-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm3[6,7]
; AVX512DQ-NEXT: vpsrlq $32, %zmm29, %zmm1
; AVX512DQ-NEXT: vpmovqw %zmm1, %xmm1
; AVX512DQ-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm2 = xmm12[3,1,2,3]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm4 = xmm2[0,1,2,0,4,5,6,7]
; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm3 = xmm11[3,1,2,3]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm5 = xmm3[0,1,2,0,4,5,6,7]
; AVX512DQ-NEXT: vpunpckldq {{.*#+}} xmm4 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
; AVX512DQ-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4
; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm0 = xmm21[3,1,2,3]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm8 = xmm0[2,0,2,3,4,5,6,7]
; AVX512DQ-NEXT: vmovdqa64 %xmm0, %xmm21
; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm0 = xmm20[3,1,2,3]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm9 = xmm0[2,0,2,3,4,5,6,7]
; AVX512DQ-NEXT: vmovdqa64 %xmm0, %xmm16
; AVX512DQ-NEXT: vpunpckldq {{.*#+}} xmm8 = xmm9[0],xmm8[0],xmm9[1],xmm8[1]
; AVX512DQ-NEXT: vinserti128 $1, %xmm8, %ymm0, %ymm8
; AVX512DQ-NEXT: vpblendd {{.*#+}} ymm4 = ymm8[0,1,2,3,4,5],ymm4[6,7]
; AVX512DQ-NEXT: vpsrlq $32, %zmm28, %zmm8
; AVX512DQ-NEXT: vpmovqw %zmm8, %xmm8
; AVX512DQ-NEXT: vpblendd {{.*#+}} ymm4 = ymm8[0,1,2,3],ymm4[4,5,6,7]
; AVX512DQ-NEXT: vshufi64x2 {{.*#+}} zmm0 = zmm4[0,1,2,3],zmm1[0,1,2,3]
; AVX512DQ-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm10 = xmm24[3,1,2,3]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm1 = xmm10[0,1,2,0,4,5,6,7]
; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm11 = xmm23[3,1,2,3]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm4 = xmm11[0,1,2,0,4,5,6,7]
; AVX512DQ-NEXT: vpunpckldq {{.*#+}} xmm20 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm12 = xmm19[3,1,2,3]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm4 = xmm12[2,0,2,3,4,5,6,7]
; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm1 = xmm18[3,1,2,3]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm8 = xmm1[2,0,2,3,4,5,6,7]
; AVX512DQ-NEXT: vpunpckldq {{.*#+}} xmm18 = xmm8[0],xmm4[0],xmm8[1],xmm4[1]
; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm4 = xmm31[3,1,2,3]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm8 = xmm4[0,1,2,0,4,5,6,7]
; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm15 = xmm25[3,1,2,3]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm9 = xmm15[0,1,2,0,4,5,6,7]
; AVX512DQ-NEXT: vpunpckldq {{.*#+}} xmm19 = xmm9[0],xmm8[0],xmm9[1],xmm8[1]
; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm8 = xmm17[3,1,2,3]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm0 = xmm8[2,0,2,3,4,5,6,7]
; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm9 = xmm30[3,1,2,3]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm14 = xmm9[2,0,2,3,4,5,6,7]
; AVX512DQ-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm14[0],xmm0[0],xmm14[1],xmm0[1]
; AVX512DQ-NEXT: vinserti32x4 $1, %xmm20, %ymm0, %ymm14
; AVX512DQ-NEXT: vinserti32x4 $1, %xmm18, %ymm0, %ymm5
; AVX512DQ-NEXT: vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],ymm14[6,7]
; AVX512DQ-NEXT: vpsrlq $32, %zmm27, %zmm14
; AVX512DQ-NEXT: vpmovqw %zmm14, %xmm14
; AVX512DQ-NEXT: vpblendd {{.*#+}} ymm5 = ymm14[0,1,2,3],ymm5[4,5,6,7]
; AVX512DQ-NEXT: vinserti32x4 $1, %xmm19, %ymm0, %ymm14
; AVX512DQ-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX512DQ-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm14[6,7]
; AVX512DQ-NEXT: vpsrlq $32, %zmm26, %zmm14
; AVX512DQ-NEXT: vpmovqw %zmm14, %xmm14
; AVX512DQ-NEXT: vpblendd {{.*#+}} ymm0 = ymm14[0,1,2,3],ymm0[4,5,6,7]
; AVX512DQ-NEXT: vshufi64x2 {{.*#+}} zmm17 = zmm0[0,1,2,3],zmm5[0,1,2,3]
; AVX512DQ-NEXT: vmovdqa64 %xmm22, %xmm0
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[0,1,3,1,4,5,6,7]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm5 = xmm13[0,1,3,1,4,5,6,7]
; AVX512DQ-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm5[0],xmm0[0],xmm5[1],xmm0[1]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm5 = xmm6[3,1,2,3,4,5,6,7]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm6 = xmm7[3,1,2,3,4,5,6,7]
; AVX512DQ-NEXT: vpunpckldq {{.*#+}} xmm5 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[0,1,3,1,4,5,6,7]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[0,1,3,1,4,5,6,7]
; AVX512DQ-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
; AVX512DQ-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX512DQ-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm3
; AVX512DQ-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
; AVX512DQ-NEXT: vpsrlq $48, %zmm29, %zmm3
; AVX512DQ-NEXT: vpmovqw %zmm3, %xmm3
; AVX512DQ-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7]
; AVX512DQ-NEXT: vmovdqa64 %xmm21, %xmm3
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[3,1,2,3,4,5,6,7]
; AVX512DQ-NEXT: vmovdqa64 %xmm16, %xmm5
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm5 = xmm5[3,1,2,3,4,5,6,7]
; AVX512DQ-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm5[0],xmm3[0],xmm5[1],xmm3[1]
; AVX512DQ-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX512DQ-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
; AVX512DQ-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
; AVX512DQ-NEXT: vpsrlq $48, %zmm28, %zmm3
; AVX512DQ-NEXT: vpmovqw %zmm3, %xmm3
; AVX512DQ-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
; AVX512DQ-NEXT: vshufi64x2 {{.*#+}} zmm0 = zmm2[0,1,2,3],zmm0[0,1,2,3]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm2 = xmm10[0,1,3,1,4,5,6,7]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm3 = xmm11[0,1,3,1,4,5,6,7]
; AVX512DQ-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm3 = xmm12[3,1,2,3,4,5,6,7]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[3,1,2,3,4,5,6,7]
; AVX512DQ-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm3 = xmm4[0,1,3,1,4,5,6,7]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm4 = xmm15[0,1,3,1,4,5,6,7]
; AVX512DQ-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
; AVX512DQ-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX512DQ-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX512DQ-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
; AVX512DQ-NEXT: vpsrlq $48, %zmm27, %zmm2
; AVX512DQ-NEXT: vpmovqw %zmm2, %xmm2
; AVX512DQ-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm2 = xmm8[3,1,2,3,4,5,6,7]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm4 = xmm9[3,1,2,3,4,5,6,7]
; AVX512DQ-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm4[0],xmm2[0],xmm4[1],xmm2[1]
; AVX512DQ-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
; AVX512DQ-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX512DQ-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm3[6,7]
; AVX512DQ-NEXT: vpsrlq $48, %zmm26, %zmm3
; AVX512DQ-NEXT: vpmovqw %zmm3, %xmm3
; AVX512DQ-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
; AVX512DQ-NEXT: vshufi64x2 {{.*#+}} zmm1 = zmm2[0,1,2,3],zmm1[0,1,2,3]
; AVX512DQ-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512DQ-NEXT: vmovaps %zmm2, 64(%rsi)
; AVX512DQ-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512DQ-NEXT: vmovaps %zmm2, (%rsi)
; AVX512DQ-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512DQ-NEXT: vmovaps %zmm2, 64(%rdx)
; AVX512DQ-NEXT: vmovups (%rsp), %zmm2 # 64-byte Reload
; AVX512DQ-NEXT: vmovaps %zmm2, (%rdx)
; AVX512DQ-NEXT: vmovdqa64 %zmm17, 64(%rcx)
; AVX512DQ-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512DQ-NEXT: vmovaps %zmm2, (%rcx)
; AVX512DQ-NEXT: vmovdqa64 %zmm1, 64(%r8)
; AVX512DQ-NEXT: vmovdqa64 %zmm0, (%r8)
; AVX512DQ-NEXT: addq $200, %rsp
; AVX512DQ-NEXT: vzeroupper
; AVX512DQ-NEXT: retq
;
; AVX512DQ-FCP-LABEL: load_i16_stride4_vf64:
; AVX512DQ-FCP: # %bb.0:
; AVX512DQ-FCP-NEXT: vmovdqa64 256(%rdi), %zmm22
; AVX512DQ-FCP-NEXT: vmovdqa64 384(%rdi), %zmm25
; AVX512DQ-FCP-NEXT: vmovdqa64 (%rdi), %zmm2
; AVX512DQ-FCP-NEXT: vmovdqa64 128(%rdi), %zmm4
; AVX512DQ-FCP-NEXT: vbroadcasti128 {{.*#+}} ymm5 = [16,17,20,21,0,0,0,0,24,25,28,29,0,0,0,0,16,17,20,21,0,0,0,0,24,25,28,29,0,0,0,0]
; AVX512DQ-FCP-NEXT: # ymm5 = mem[0,1,0,1]
; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} ymm9 = [0,2,2,3,0,2,4,6]
; AVX512DQ-FCP-NEXT: vmovdqa64 224(%rdi), %ymm23
; AVX512DQ-FCP-NEXT: vpermd %ymm23, %ymm9, %ymm10
; AVX512DQ-FCP-NEXT: vpshufb %ymm5, %ymm10, %ymm0
; AVX512DQ-FCP-NEXT: vmovdqa {{.*#+}} ymm6 = [0,1,4,5,u,u,u,u,8,9,12,13,12,13,14,15,16,17,20,21,u,u,u,u,24,25,28,29,u,u,u,u]
; AVX512DQ-FCP-NEXT: vmovdqa64 192(%rdi), %ymm24
; AVX512DQ-FCP-NEXT: vpermd %ymm24, %ymm9, %ymm3
; AVX512DQ-FCP-NEXT: vpshufb %ymm6, %ymm3, %ymm1
; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} ymm7 = [0,2,2,3,4,6,12,14]
; AVX512DQ-FCP-NEXT: vpermt2d %ymm0, %ymm7, %ymm1
; AVX512DQ-FCP-NEXT: vpmovqw %zmm4, %xmm0
; AVX512DQ-FCP-NEXT: vpblendd {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX512DQ-FCP-NEXT: vmovdqa64 96(%rdi), %ymm26
; AVX512DQ-FCP-NEXT: vpermd %ymm26, %ymm9, %ymm8
; AVX512DQ-FCP-NEXT: vpshufb %ymm5, %ymm8, %ymm11
; AVX512DQ-FCP-NEXT: vmovdqa64 64(%rdi), %ymm27
; AVX512DQ-FCP-NEXT: vpermd %ymm27, %ymm9, %ymm0
; AVX512DQ-FCP-NEXT: vpshufb %ymm6, %ymm0, %ymm12
; AVX512DQ-FCP-NEXT: vpermt2d %ymm11, %ymm7, %ymm12
; AVX512DQ-FCP-NEXT: vpmovqw %zmm2, %xmm11
; AVX512DQ-FCP-NEXT: vpblendd {{.*#+}} ymm11 = ymm11[0,1,2,3],ymm12[4,5,6,7]
; AVX512DQ-FCP-NEXT: vshufi64x2 {{.*#+}} zmm20 = zmm11[0,1,2,3],zmm1[0,1,2,3]
; AVX512DQ-FCP-NEXT: vmovdqa64 480(%rdi), %ymm28
; AVX512DQ-FCP-NEXT: vpermd %ymm28, %ymm9, %ymm11
; AVX512DQ-FCP-NEXT: vpshufb %ymm5, %ymm11, %ymm1
; AVX512DQ-FCP-NEXT: vmovdqa64 448(%rdi), %ymm17
; AVX512DQ-FCP-NEXT: vpermd %ymm17, %ymm9, %ymm12
; AVX512DQ-FCP-NEXT: vpshufb %ymm6, %ymm12, %ymm13
; AVX512DQ-FCP-NEXT: vpermt2d %ymm1, %ymm7, %ymm13
; AVX512DQ-FCP-NEXT: vpmovqw %zmm25, %xmm1
; AVX512DQ-FCP-NEXT: vpblendd {{.*#+}} ymm14 = ymm1[0,1,2,3],ymm13[4,5,6,7]
; AVX512DQ-FCP-NEXT: vmovdqa64 352(%rdi), %ymm18
; AVX512DQ-FCP-NEXT: vpermd %ymm18, %ymm9, %ymm1
; AVX512DQ-FCP-NEXT: vpshufb %ymm5, %ymm1, %ymm15
; AVX512DQ-FCP-NEXT: vmovdqa64 320(%rdi), %ymm19
; AVX512DQ-FCP-NEXT: vpermd %ymm19, %ymm9, %ymm13
; AVX512DQ-FCP-NEXT: vpshufb %ymm6, %ymm13, %ymm9
; AVX512DQ-FCP-NEXT: vpermt2d %ymm15, %ymm7, %ymm9
; AVX512DQ-FCP-NEXT: vpmovqw %zmm22, %xmm15
; AVX512DQ-FCP-NEXT: vpblendd {{.*#+}} ymm9 = ymm15[0,1,2,3],ymm9[4,5,6,7]
; AVX512DQ-FCP-NEXT: vshufi64x2 {{.*#+}} zmm21 = zmm9[0,1,2,3],zmm14[0,1,2,3]
; AVX512DQ-FCP-NEXT: vmovdqa {{.*#+}} ymm9 = [2,3,6,7,10,11,14,15,14,15,10,11,12,13,14,15,18,19,22,23,26,27,30,31,18,19,22,23,26,27,30,31]
; AVX512DQ-FCP-NEXT: vpshufb %ymm9, %ymm10, %ymm10
; AVX512DQ-FCP-NEXT: vpshufb %ymm9, %ymm3, %ymm3
; AVX512DQ-FCP-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm10[6,7]
; AVX512DQ-FCP-NEXT: vpsrlq $16, %zmm4, %zmm10
; AVX512DQ-FCP-NEXT: vpmovqw %zmm10, %xmm10
; AVX512DQ-FCP-NEXT: vpblendd {{.*#+}} ymm3 = ymm10[0,1,2,3],ymm3[4,5,6,7]
; AVX512DQ-FCP-NEXT: vpshufb %ymm9, %ymm8, %ymm8
; AVX512DQ-FCP-NEXT: vpshufb %ymm9, %ymm0, %ymm0
; AVX512DQ-FCP-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm8[6,7]
; AVX512DQ-FCP-NEXT: vpsrlq $16, %zmm2, %zmm8
; AVX512DQ-FCP-NEXT: vpmovqw %zmm8, %xmm8
; AVX512DQ-FCP-NEXT: vpblendd {{.*#+}} ymm0 = ymm8[0,1,2,3],ymm0[4,5,6,7]
; AVX512DQ-FCP-NEXT: vshufi64x2 {{.*#+}} zmm29 = zmm0[0,1,2,3],zmm3[0,1,2,3]
; AVX512DQ-FCP-NEXT: vpshufb %ymm9, %ymm11, %ymm0
; AVX512DQ-FCP-NEXT: vpshufb %ymm9, %ymm12, %ymm3
; AVX512DQ-FCP-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
; AVX512DQ-FCP-NEXT: vpsrlq $16, %zmm25, %zmm3
; AVX512DQ-FCP-NEXT: vpmovqw %zmm3, %xmm3
; AVX512DQ-FCP-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7]
; AVX512DQ-FCP-NEXT: vpshufb %ymm9, %ymm1, %ymm1
; AVX512DQ-FCP-NEXT: vpshufb %ymm9, %ymm13, %ymm3
; AVX512DQ-FCP-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm1[6,7]
; AVX512DQ-FCP-NEXT: vpsrlq $16, %zmm22, %zmm3
; AVX512DQ-FCP-NEXT: vpmovqw %zmm3, %xmm3
; AVX512DQ-FCP-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm1[4,5,6,7]
; AVX512DQ-FCP-NEXT: vshufi64x2 {{.*#+}} zmm16 = zmm1[0,1,2,3],zmm0[0,1,2,3]
; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} ymm14 = [1,3,2,3,1,3,5,7]
; AVX512DQ-FCP-NEXT: vpermd %ymm23, %ymm14, %ymm3
; AVX512DQ-FCP-NEXT: vpshufb %ymm5, %ymm3, %ymm0
; AVX512DQ-FCP-NEXT: vpermd %ymm24, %ymm14, %ymm8
; AVX512DQ-FCP-NEXT: vpshufb %ymm6, %ymm8, %ymm1
; AVX512DQ-FCP-NEXT: vpermt2d %ymm0, %ymm7, %ymm1
; AVX512DQ-FCP-NEXT: vpsrlq $32, %zmm4, %zmm0
; AVX512DQ-FCP-NEXT: vpmovqw %zmm0, %xmm0
; AVX512DQ-FCP-NEXT: vpblendd {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX512DQ-FCP-NEXT: vpermd %ymm26, %ymm14, %ymm0
; AVX512DQ-FCP-NEXT: vpshufb %ymm5, %ymm0, %ymm12
; AVX512DQ-FCP-NEXT: vpermd %ymm27, %ymm14, %ymm11
; AVX512DQ-FCP-NEXT: vpshufb %ymm6, %ymm11, %ymm13
; AVX512DQ-FCP-NEXT: vpermt2d %ymm12, %ymm7, %ymm13
; AVX512DQ-FCP-NEXT: vpsrlq $32, %zmm2, %zmm12
; AVX512DQ-FCP-NEXT: vpmovqw %zmm12, %xmm12
; AVX512DQ-FCP-NEXT: vpblendd {{.*#+}} ymm12 = ymm12[0,1,2,3],ymm13[4,5,6,7]
; AVX512DQ-FCP-NEXT: vshufi64x2 {{.*#+}} zmm23 = zmm12[0,1,2,3],zmm1[0,1,2,3]
; AVX512DQ-FCP-NEXT: vpermd %ymm28, %ymm14, %ymm12
; AVX512DQ-FCP-NEXT: vpshufb %ymm5, %ymm12, %ymm1
; AVX512DQ-FCP-NEXT: vpermd %ymm17, %ymm14, %ymm13
; AVX512DQ-FCP-NEXT: vpshufb %ymm6, %ymm13, %ymm10
; AVX512DQ-FCP-NEXT: vpermt2d %ymm1, %ymm7, %ymm10
; AVX512DQ-FCP-NEXT: vpsrlq $32, %zmm25, %zmm1
; AVX512DQ-FCP-NEXT: vpmovqw %zmm1, %xmm1
; AVX512DQ-FCP-NEXT: vpblendd {{.*#+}} ymm10 = ymm1[0,1,2,3],ymm10[4,5,6,7]
; AVX512DQ-FCP-NEXT: vpermd %ymm18, %ymm14, %ymm1
; AVX512DQ-FCP-NEXT: vpshufb %ymm5, %ymm1, %ymm15
; AVX512DQ-FCP-NEXT: vpermd %ymm19, %ymm14, %ymm5
; AVX512DQ-FCP-NEXT: vpshufb %ymm6, %ymm5, %ymm6
; AVX512DQ-FCP-NEXT: vpermt2d %ymm15, %ymm7, %ymm6
; AVX512DQ-FCP-NEXT: vpsrlq $32, %zmm22, %zmm7
; AVX512DQ-FCP-NEXT: vpmovqw %zmm7, %xmm7
; AVX512DQ-FCP-NEXT: vpblendd {{.*#+}} ymm6 = ymm7[0,1,2,3],ymm6[4,5,6,7]
; AVX512DQ-FCP-NEXT: vshufi64x2 {{.*#+}} zmm6 = zmm6[0,1,2,3],zmm10[0,1,2,3]
; AVX512DQ-FCP-NEXT: vpshufb %ymm9, %ymm3, %ymm3
; AVX512DQ-FCP-NEXT: vpshufb %ymm9, %ymm8, %ymm7
; AVX512DQ-FCP-NEXT: vpblendd {{.*#+}} ymm3 = ymm7[0,1,2,3,4,5],ymm3[6,7]
; AVX512DQ-FCP-NEXT: vpsrlq $48, %zmm4, %zmm4
; AVX512DQ-FCP-NEXT: vpmovqw %zmm4, %xmm4
; AVX512DQ-FCP-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
; AVX512DQ-FCP-NEXT: vpshufb %ymm9, %ymm0, %ymm0
; AVX512DQ-FCP-NEXT: vpshufb %ymm9, %ymm11, %ymm4
; AVX512DQ-FCP-NEXT: vpblendd {{.*#+}} ymm0 = ymm4[0,1,2,3,4,5],ymm0[6,7]
; AVX512DQ-FCP-NEXT: vpsrlq $48, %zmm2, %zmm2
; AVX512DQ-FCP-NEXT: vpmovqw %zmm2, %xmm2
; AVX512DQ-FCP-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
; AVX512DQ-FCP-NEXT: vshufi64x2 {{.*#+}} zmm0 = zmm0[0,1,2,3],zmm3[0,1,2,3]
; AVX512DQ-FCP-NEXT: vpshufb %ymm9, %ymm12, %ymm2
; AVX512DQ-FCP-NEXT: vpshufb %ymm9, %ymm13, %ymm3
; AVX512DQ-FCP-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
; AVX512DQ-FCP-NEXT: vpsrlq $48, %zmm25, %zmm3
; AVX512DQ-FCP-NEXT: vpmovqw %zmm3, %xmm3
; AVX512DQ-FCP-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
; AVX512DQ-FCP-NEXT: vpshufb %ymm9, %ymm1, %ymm1
; AVX512DQ-FCP-NEXT: vpshufb %ymm9, %ymm5, %ymm3
; AVX512DQ-FCP-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm1[6,7]
; AVX512DQ-FCP-NEXT: vpsrlq $48, %zmm22, %zmm3
; AVX512DQ-FCP-NEXT: vpmovqw %zmm3, %xmm3
; AVX512DQ-FCP-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm1[4,5,6,7]
; AVX512DQ-FCP-NEXT: vshufi64x2 {{.*#+}} zmm1 = zmm1[0,1,2,3],zmm2[0,1,2,3]
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm21, 64(%rsi)
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm20, (%rsi)
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm16, 64(%rdx)
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm29, (%rdx)
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm6, 64(%rcx)
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm23, (%rcx)
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm1, 64(%r8)
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm0, (%r8)
; AVX512DQ-FCP-NEXT: vzeroupper
; AVX512DQ-FCP-NEXT: retq
;
; AVX512BW-LABEL: load_i16_stride4_vf64:
; AVX512BW: # %bb.0:
; AVX512BW-NEXT: vmovdqa64 (%rdi), %zmm0
; AVX512BW-NEXT: vmovdqa64 64(%rdi), %zmm1
; AVX512BW-NEXT: vmovdqa64 128(%rdi), %zmm2
; AVX512BW-NEXT: vmovdqa64 192(%rdi), %zmm3
; AVX512BW-NEXT: vmovdqa64 320(%rdi), %zmm4
; AVX512BW-NEXT: vmovdqa64 256(%rdi), %zmm5
; AVX512BW-NEXT: vmovdqa64 448(%rdi), %zmm6
; AVX512BW-NEXT: vmovdqa64 384(%rdi), %zmm7
; AVX512BW-NEXT: vbroadcasti64x4 {{.*#+}} zmm8 = [0,4,8,12,16,20,24,28,32,36,40,44,48,52,56,60,0,4,8,12,16,20,24,28,32,36,40,44,48,52,56,60]
; AVX512BW-NEXT: # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-NEXT: vmovdqa64 %zmm7, %zmm9
; AVX512BW-NEXT: vpermt2w %zmm6, %zmm8, %zmm9
; AVX512BW-NEXT: vmovdqa64 %zmm5, %zmm10
; AVX512BW-NEXT: vpermt2w %zmm4, %zmm8, %zmm10
; AVX512BW-NEXT: vshufi64x2 {{.*#+}} zmm9 = zmm10[0,1,2,3],zmm9[4,5,6,7]
; AVX512BW-NEXT: vmovdqa64 %zmm2, %zmm10
; AVX512BW-NEXT: vpermt2w %zmm3, %zmm8, %zmm10
; AVX512BW-NEXT: vpermi2w %zmm1, %zmm0, %zmm8
; AVX512BW-NEXT: vshufi64x2 {{.*#+}} zmm8 = zmm8[0,1,2,3],zmm10[4,5,6,7]
; AVX512BW-NEXT: vbroadcasti64x4 {{.*#+}} zmm10 = [1,5,9,13,17,21,25,29,33,37,41,45,49,53,57,61,1,5,9,13,17,21,25,29,33,37,41,45,49,53,57,61]
; AVX512BW-NEXT: # zmm10 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-NEXT: vmovdqa64 %zmm7, %zmm11
; AVX512BW-NEXT: vpermt2w %zmm6, %zmm10, %zmm11
; AVX512BW-NEXT: vmovdqa64 %zmm5, %zmm12
; AVX512BW-NEXT: vpermt2w %zmm4, %zmm10, %zmm12
; AVX512BW-NEXT: vshufi64x2 {{.*#+}} zmm11 = zmm12[0,1,2,3],zmm11[4,5,6,7]
; AVX512BW-NEXT: vmovdqa64 %zmm2, %zmm12
; AVX512BW-NEXT: vpermt2w %zmm3, %zmm10, %zmm12
; AVX512BW-NEXT: vpermi2w %zmm1, %zmm0, %zmm10
; AVX512BW-NEXT: vshufi64x2 {{.*#+}} zmm10 = zmm10[0,1,2,3],zmm12[4,5,6,7]
; AVX512BW-NEXT: vbroadcasti64x4 {{.*#+}} zmm12 = [2,6,10,14,18,22,26,30,34,38,42,46,50,54,58,62,2,6,10,14,18,22,26,30,34,38,42,46,50,54,58,62]
; AVX512BW-NEXT: # zmm12 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-NEXT: vmovdqa64 %zmm7, %zmm13
; AVX512BW-NEXT: vpermt2w %zmm6, %zmm12, %zmm13
; AVX512BW-NEXT: vmovdqa64 %zmm5, %zmm14
; AVX512BW-NEXT: vpermt2w %zmm4, %zmm12, %zmm14
; AVX512BW-NEXT: vshufi64x2 {{.*#+}} zmm13 = zmm14[0,1,2,3],zmm13[4,5,6,7]
; AVX512BW-NEXT: vmovdqa64 %zmm2, %zmm14
; AVX512BW-NEXT: vpermt2w %zmm3, %zmm12, %zmm14
; AVX512BW-NEXT: vpermi2w %zmm1, %zmm0, %zmm12
; AVX512BW-NEXT: vshufi64x2 {{.*#+}} zmm12 = zmm12[0,1,2,3],zmm14[4,5,6,7]
; AVX512BW-NEXT: vbroadcasti64x4 {{.*#+}} zmm14 = [3,7,11,15,19,23,27,31,35,39,43,47,51,55,59,63,3,7,11,15,19,23,27,31,35,39,43,47,51,55,59,63]
; AVX512BW-NEXT: # zmm14 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-NEXT: vpermt2w %zmm6, %zmm14, %zmm7
; AVX512BW-NEXT: vpermt2w %zmm4, %zmm14, %zmm5
; AVX512BW-NEXT: vshufi64x2 {{.*#+}} zmm4 = zmm5[0,1,2,3],zmm7[4,5,6,7]
; AVX512BW-NEXT: vpermt2w %zmm3, %zmm14, %zmm2
; AVX512BW-NEXT: vpermt2w %zmm1, %zmm14, %zmm0
; AVX512BW-NEXT: vshufi64x2 {{.*#+}} zmm0 = zmm0[0,1,2,3],zmm2[4,5,6,7]
; AVX512BW-NEXT: vmovdqa64 %zmm9, 64(%rsi)
; AVX512BW-NEXT: vmovdqa64 %zmm8, (%rsi)
; AVX512BW-NEXT: vmovdqa64 %zmm11, 64(%rdx)
; AVX512BW-NEXT: vmovdqa64 %zmm10, (%rdx)
; AVX512BW-NEXT: vmovdqa64 %zmm13, 64(%rcx)
; AVX512BW-NEXT: vmovdqa64 %zmm12, (%rcx)
; AVX512BW-NEXT: vmovdqa64 %zmm4, 64(%r8)
; AVX512BW-NEXT: vmovdqa64 %zmm0, (%r8)
; AVX512BW-NEXT: vzeroupper
; AVX512BW-NEXT: retq
;
; AVX512BW-FCP-LABEL: load_i16_stride4_vf64:
; AVX512BW-FCP: # %bb.0:
; AVX512BW-FCP-NEXT: vmovdqa64 (%rdi), %zmm0
; AVX512BW-FCP-NEXT: vmovdqa64 64(%rdi), %zmm1
; AVX512BW-FCP-NEXT: vmovdqa64 128(%rdi), %zmm2
; AVX512BW-FCP-NEXT: vmovdqa64 192(%rdi), %zmm3
; AVX512BW-FCP-NEXT: vmovdqa64 320(%rdi), %zmm4
; AVX512BW-FCP-NEXT: vmovdqa64 256(%rdi), %zmm5
; AVX512BW-FCP-NEXT: vmovdqa64 448(%rdi), %zmm6
; AVX512BW-FCP-NEXT: vmovdqa64 384(%rdi), %zmm7
; AVX512BW-FCP-NEXT: vbroadcasti64x4 {{.*#+}} zmm8 = [0,4,8,12,16,20,24,28,32,36,40,44,48,52,56,60,0,4,8,12,16,20,24,28,32,36,40,44,48,52,56,60]
; AVX512BW-FCP-NEXT: # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm7, %zmm9
; AVX512BW-FCP-NEXT: vpermt2w %zmm6, %zmm8, %zmm9
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm5, %zmm10
; AVX512BW-FCP-NEXT: vpermt2w %zmm4, %zmm8, %zmm10
; AVX512BW-FCP-NEXT: vshufi64x2 {{.*#+}} zmm9 = zmm10[0,1,2,3],zmm9[4,5,6,7]
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm2, %zmm10
; AVX512BW-FCP-NEXT: vpermt2w %zmm3, %zmm8, %zmm10
; AVX512BW-FCP-NEXT: vpermi2w %zmm1, %zmm0, %zmm8
; AVX512BW-FCP-NEXT: vshufi64x2 {{.*#+}} zmm8 = zmm8[0,1,2,3],zmm10[4,5,6,7]
; AVX512BW-FCP-NEXT: vbroadcasti64x4 {{.*#+}} zmm10 = [1,5,9,13,17,21,25,29,33,37,41,45,49,53,57,61,1,5,9,13,17,21,25,29,33,37,41,45,49,53,57,61]
; AVX512BW-FCP-NEXT: # zmm10 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm7, %zmm11
; AVX512BW-FCP-NEXT: vpermt2w %zmm6, %zmm10, %zmm11
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm5, %zmm12
; AVX512BW-FCP-NEXT: vpermt2w %zmm4, %zmm10, %zmm12
; AVX512BW-FCP-NEXT: vshufi64x2 {{.*#+}} zmm11 = zmm12[0,1,2,3],zmm11[4,5,6,7]
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm2, %zmm12
; AVX512BW-FCP-NEXT: vpermt2w %zmm3, %zmm10, %zmm12
; AVX512BW-FCP-NEXT: vpermi2w %zmm1, %zmm0, %zmm10
; AVX512BW-FCP-NEXT: vshufi64x2 {{.*#+}} zmm10 = zmm10[0,1,2,3],zmm12[4,5,6,7]
; AVX512BW-FCP-NEXT: vbroadcasti64x4 {{.*#+}} zmm12 = [2,6,10,14,18,22,26,30,34,38,42,46,50,54,58,62,2,6,10,14,18,22,26,30,34,38,42,46,50,54,58,62]
; AVX512BW-FCP-NEXT: # zmm12 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm7, %zmm13
; AVX512BW-FCP-NEXT: vpermt2w %zmm6, %zmm12, %zmm13
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm5, %zmm14
; AVX512BW-FCP-NEXT: vpermt2w %zmm4, %zmm12, %zmm14
; AVX512BW-FCP-NEXT: vshufi64x2 {{.*#+}} zmm13 = zmm14[0,1,2,3],zmm13[4,5,6,7]
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm2, %zmm14
; AVX512BW-FCP-NEXT: vpermt2w %zmm3, %zmm12, %zmm14
; AVX512BW-FCP-NEXT: vpermi2w %zmm1, %zmm0, %zmm12
; AVX512BW-FCP-NEXT: vshufi64x2 {{.*#+}} zmm12 = zmm12[0,1,2,3],zmm14[4,5,6,7]
; AVX512BW-FCP-NEXT: vbroadcasti64x4 {{.*#+}} zmm14 = [3,7,11,15,19,23,27,31,35,39,43,47,51,55,59,63,3,7,11,15,19,23,27,31,35,39,43,47,51,55,59,63]
; AVX512BW-FCP-NEXT: # zmm14 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-FCP-NEXT: vpermt2w %zmm6, %zmm14, %zmm7
; AVX512BW-FCP-NEXT: vpermt2w %zmm4, %zmm14, %zmm5
; AVX512BW-FCP-NEXT: vshufi64x2 {{.*#+}} zmm4 = zmm5[0,1,2,3],zmm7[4,5,6,7]
; AVX512BW-FCP-NEXT: vpermt2w %zmm3, %zmm14, %zmm2
; AVX512BW-FCP-NEXT: vpermt2w %zmm1, %zmm14, %zmm0
; AVX512BW-FCP-NEXT: vshufi64x2 {{.*#+}} zmm0 = zmm0[0,1,2,3],zmm2[4,5,6,7]
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm9, 64(%rsi)
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm8, (%rsi)
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm11, 64(%rdx)
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm10, (%rdx)
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm13, 64(%rcx)
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm12, (%rcx)
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm4, 64(%r8)
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm0, (%r8)
; AVX512BW-FCP-NEXT: vzeroupper
; AVX512BW-FCP-NEXT: retq
;
; AVX512DQ-BW-LABEL: load_i16_stride4_vf64:
; AVX512DQ-BW: # %bb.0:
; AVX512DQ-BW-NEXT: vmovdqa64 (%rdi), %zmm0
; AVX512DQ-BW-NEXT: vmovdqa64 64(%rdi), %zmm1
; AVX512DQ-BW-NEXT: vmovdqa64 128(%rdi), %zmm2
; AVX512DQ-BW-NEXT: vmovdqa64 192(%rdi), %zmm3
; AVX512DQ-BW-NEXT: vmovdqa64 320(%rdi), %zmm4
; AVX512DQ-BW-NEXT: vmovdqa64 256(%rdi), %zmm5
; AVX512DQ-BW-NEXT: vmovdqa64 448(%rdi), %zmm6
; AVX512DQ-BW-NEXT: vmovdqa64 384(%rdi), %zmm7
; AVX512DQ-BW-NEXT: vbroadcasti64x4 {{.*#+}} zmm8 = [0,4,8,12,16,20,24,28,32,36,40,44,48,52,56,60,0,4,8,12,16,20,24,28,32,36,40,44,48,52,56,60]
; AVX512DQ-BW-NEXT: # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm7, %zmm9
; AVX512DQ-BW-NEXT: vpermt2w %zmm6, %zmm8, %zmm9
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm5, %zmm10
; AVX512DQ-BW-NEXT: vpermt2w %zmm4, %zmm8, %zmm10
; AVX512DQ-BW-NEXT: vshufi64x2 {{.*#+}} zmm9 = zmm10[0,1,2,3],zmm9[4,5,6,7]
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm2, %zmm10
; AVX512DQ-BW-NEXT: vpermt2w %zmm3, %zmm8, %zmm10
; AVX512DQ-BW-NEXT: vpermi2w %zmm1, %zmm0, %zmm8
; AVX512DQ-BW-NEXT: vshufi64x2 {{.*#+}} zmm8 = zmm8[0,1,2,3],zmm10[4,5,6,7]
; AVX512DQ-BW-NEXT: vbroadcasti64x4 {{.*#+}} zmm10 = [1,5,9,13,17,21,25,29,33,37,41,45,49,53,57,61,1,5,9,13,17,21,25,29,33,37,41,45,49,53,57,61]
; AVX512DQ-BW-NEXT: # zmm10 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm7, %zmm11
; AVX512DQ-BW-NEXT: vpermt2w %zmm6, %zmm10, %zmm11
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm5, %zmm12
; AVX512DQ-BW-NEXT: vpermt2w %zmm4, %zmm10, %zmm12
; AVX512DQ-BW-NEXT: vshufi64x2 {{.*#+}} zmm11 = zmm12[0,1,2,3],zmm11[4,5,6,7]
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm2, %zmm12
; AVX512DQ-BW-NEXT: vpermt2w %zmm3, %zmm10, %zmm12
; AVX512DQ-BW-NEXT: vpermi2w %zmm1, %zmm0, %zmm10
; AVX512DQ-BW-NEXT: vshufi64x2 {{.*#+}} zmm10 = zmm10[0,1,2,3],zmm12[4,5,6,7]
; AVX512DQ-BW-NEXT: vbroadcasti64x4 {{.*#+}} zmm12 = [2,6,10,14,18,22,26,30,34,38,42,46,50,54,58,62,2,6,10,14,18,22,26,30,34,38,42,46,50,54,58,62]
; AVX512DQ-BW-NEXT: # zmm12 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm7, %zmm13
; AVX512DQ-BW-NEXT: vpermt2w %zmm6, %zmm12, %zmm13
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm5, %zmm14
; AVX512DQ-BW-NEXT: vpermt2w %zmm4, %zmm12, %zmm14
; AVX512DQ-BW-NEXT: vshufi64x2 {{.*#+}} zmm13 = zmm14[0,1,2,3],zmm13[4,5,6,7]
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm2, %zmm14
; AVX512DQ-BW-NEXT: vpermt2w %zmm3, %zmm12, %zmm14
; AVX512DQ-BW-NEXT: vpermi2w %zmm1, %zmm0, %zmm12
; AVX512DQ-BW-NEXT: vshufi64x2 {{.*#+}} zmm12 = zmm12[0,1,2,3],zmm14[4,5,6,7]
; AVX512DQ-BW-NEXT: vbroadcasti64x4 {{.*#+}} zmm14 = [3,7,11,15,19,23,27,31,35,39,43,47,51,55,59,63,3,7,11,15,19,23,27,31,35,39,43,47,51,55,59,63]
; AVX512DQ-BW-NEXT: # zmm14 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-NEXT: vpermt2w %zmm6, %zmm14, %zmm7
; AVX512DQ-BW-NEXT: vpermt2w %zmm4, %zmm14, %zmm5
; AVX512DQ-BW-NEXT: vshufi64x2 {{.*#+}} zmm4 = zmm5[0,1,2,3],zmm7[4,5,6,7]
; AVX512DQ-BW-NEXT: vpermt2w %zmm3, %zmm14, %zmm2
; AVX512DQ-BW-NEXT: vpermt2w %zmm1, %zmm14, %zmm0
; AVX512DQ-BW-NEXT: vshufi64x2 {{.*#+}} zmm0 = zmm0[0,1,2,3],zmm2[4,5,6,7]
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm9, 64(%rsi)
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm8, (%rsi)
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm11, 64(%rdx)
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm10, (%rdx)
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm13, 64(%rcx)
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm12, (%rcx)
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm4, 64(%r8)
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm0, (%r8)
; AVX512DQ-BW-NEXT: vzeroupper
; AVX512DQ-BW-NEXT: retq
;
; AVX512DQ-BW-FCP-LABEL: load_i16_stride4_vf64:
; AVX512DQ-BW-FCP: # %bb.0:
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 (%rdi), %zmm0
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 64(%rdi), %zmm1
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 128(%rdi), %zmm2
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 192(%rdi), %zmm3
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 320(%rdi), %zmm4
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 256(%rdi), %zmm5
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 448(%rdi), %zmm6
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 384(%rdi), %zmm7
; AVX512DQ-BW-FCP-NEXT: vbroadcasti64x4 {{.*#+}} zmm8 = [0,4,8,12,16,20,24,28,32,36,40,44,48,52,56,60,0,4,8,12,16,20,24,28,32,36,40,44,48,52,56,60]
; AVX512DQ-BW-FCP-NEXT: # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm7, %zmm9
; AVX512DQ-BW-FCP-NEXT: vpermt2w %zmm6, %zmm8, %zmm9
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm5, %zmm10
; AVX512DQ-BW-FCP-NEXT: vpermt2w %zmm4, %zmm8, %zmm10
; AVX512DQ-BW-FCP-NEXT: vshufi64x2 {{.*#+}} zmm9 = zmm10[0,1,2,3],zmm9[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm2, %zmm10
; AVX512DQ-BW-FCP-NEXT: vpermt2w %zmm3, %zmm8, %zmm10
; AVX512DQ-BW-FCP-NEXT: vpermi2w %zmm1, %zmm0, %zmm8
; AVX512DQ-BW-FCP-NEXT: vshufi64x2 {{.*#+}} zmm8 = zmm8[0,1,2,3],zmm10[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT: vbroadcasti64x4 {{.*#+}} zmm10 = [1,5,9,13,17,21,25,29,33,37,41,45,49,53,57,61,1,5,9,13,17,21,25,29,33,37,41,45,49,53,57,61]
; AVX512DQ-BW-FCP-NEXT: # zmm10 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm7, %zmm11
; AVX512DQ-BW-FCP-NEXT: vpermt2w %zmm6, %zmm10, %zmm11
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm5, %zmm12
; AVX512DQ-BW-FCP-NEXT: vpermt2w %zmm4, %zmm10, %zmm12
; AVX512DQ-BW-FCP-NEXT: vshufi64x2 {{.*#+}} zmm11 = zmm12[0,1,2,3],zmm11[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm2, %zmm12
; AVX512DQ-BW-FCP-NEXT: vpermt2w %zmm3, %zmm10, %zmm12
; AVX512DQ-BW-FCP-NEXT: vpermi2w %zmm1, %zmm0, %zmm10
; AVX512DQ-BW-FCP-NEXT: vshufi64x2 {{.*#+}} zmm10 = zmm10[0,1,2,3],zmm12[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT: vbroadcasti64x4 {{.*#+}} zmm12 = [2,6,10,14,18,22,26,30,34,38,42,46,50,54,58,62,2,6,10,14,18,22,26,30,34,38,42,46,50,54,58,62]
; AVX512DQ-BW-FCP-NEXT: # zmm12 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm7, %zmm13
; AVX512DQ-BW-FCP-NEXT: vpermt2w %zmm6, %zmm12, %zmm13
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm5, %zmm14
; AVX512DQ-BW-FCP-NEXT: vpermt2w %zmm4, %zmm12, %zmm14
; AVX512DQ-BW-FCP-NEXT: vshufi64x2 {{.*#+}} zmm13 = zmm14[0,1,2,3],zmm13[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm2, %zmm14
; AVX512DQ-BW-FCP-NEXT: vpermt2w %zmm3, %zmm12, %zmm14
; AVX512DQ-BW-FCP-NEXT: vpermi2w %zmm1, %zmm0, %zmm12
; AVX512DQ-BW-FCP-NEXT: vshufi64x2 {{.*#+}} zmm12 = zmm12[0,1,2,3],zmm14[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT: vbroadcasti64x4 {{.*#+}} zmm14 = [3,7,11,15,19,23,27,31,35,39,43,47,51,55,59,63,3,7,11,15,19,23,27,31,35,39,43,47,51,55,59,63]
; AVX512DQ-BW-FCP-NEXT: # zmm14 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-FCP-NEXT: vpermt2w %zmm6, %zmm14, %zmm7
; AVX512DQ-BW-FCP-NEXT: vpermt2w %zmm4, %zmm14, %zmm5
; AVX512DQ-BW-FCP-NEXT: vshufi64x2 {{.*#+}} zmm4 = zmm5[0,1,2,3],zmm7[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT: vpermt2w %zmm3, %zmm14, %zmm2
; AVX512DQ-BW-FCP-NEXT: vpermt2w %zmm1, %zmm14, %zmm0
; AVX512DQ-BW-FCP-NEXT: vshufi64x2 {{.*#+}} zmm0 = zmm0[0,1,2,3],zmm2[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm9, 64(%rsi)
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm8, (%rsi)
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm11, 64(%rdx)
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm10, (%rdx)
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm13, 64(%rcx)
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm12, (%rcx)
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm4, 64(%r8)
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm0, (%r8)
; AVX512DQ-BW-FCP-NEXT: vzeroupper
; AVX512DQ-BW-FCP-NEXT: retq
%wide.vec = load <256 x i16>, ptr %in.vec, align 64
%strided.vec0 = shufflevector <256 x i16> %wide.vec, <256 x i16> poison, <64 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28, i32 32, i32 36, i32 40, i32 44, i32 48, i32 52, i32 56, i32 60, i32 64, i32 68, i32 72, i32 76, i32 80, i32 84, i32 88, i32 92, i32 96, i32 100, i32 104, i32 108, i32 112, i32 116, i32 120, i32 124, i32 128, i32 132, i32 136, i32 140, i32 144, i32 148, i32 152, i32 156, i32 160, i32 164, i32 168, i32 172, i32 176, i32 180, i32 184, i32 188, i32 192, i32 196, i32 200, i32 204, i32 208, i32 212, i32 216, i32 220, i32 224, i32 228, i32 232, i32 236, i32 240, i32 244, i32 248, i32 252>
%strided.vec1 = shufflevector <256 x i16> %wide.vec, <256 x i16> poison, <64 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29, i32 33, i32 37, i32 41, i32 45, i32 49, i32 53, i32 57, i32 61, i32 65, i32 69, i32 73, i32 77, i32 81, i32 85, i32 89, i32 93, i32 97, i32 101, i32 105, i32 109, i32 113, i32 117, i32 121, i32 125, i32 129, i32 133, i32 137, i32 141, i32 145, i32 149, i32 153, i32 157, i32 161, i32 165, i32 169, i32 173, i32 177, i32 181, i32 185, i32 189, i32 193, i32 197, i32 201, i32 205, i32 209, i32 213, i32 217, i32 221, i32 225, i32 229, i32 233, i32 237, i32 241, i32 245, i32 249, i32 253>
%strided.vec2 = shufflevector <256 x i16> %wide.vec, <256 x i16> poison, <64 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30, i32 34, i32 38, i32 42, i32 46, i32 50, i32 54, i32 58, i32 62, i32 66, i32 70, i32 74, i32 78, i32 82, i32 86, i32 90, i32 94, i32 98, i32 102, i32 106, i32 110, i32 114, i32 118, i32 122, i32 126, i32 130, i32 134, i32 138, i32 142, i32 146, i32 150, i32 154, i32 158, i32 162, i32 166, i32 170, i32 174, i32 178, i32 182, i32 186, i32 190, i32 194, i32 198, i32 202, i32 206, i32 210, i32 214, i32 218, i32 222, i32 226, i32 230, i32 234, i32 238, i32 242, i32 246, i32 250, i32 254>
%strided.vec3 = shufflevector <256 x i16> %wide.vec, <256 x i16> poison, <64 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31, i32 35, i32 39, i32 43, i32 47, i32 51, i32 55, i32 59, i32 63, i32 67, i32 71, i32 75, i32 79, i32 83, i32 87, i32 91, i32 95, i32 99, i32 103, i32 107, i32 111, i32 115, i32 119, i32 123, i32 127, i32 131, i32 135, i32 139, i32 143, i32 147, i32 151, i32 155, i32 159, i32 163, i32 167, i32 171, i32 175, i32 179, i32 183, i32 187, i32 191, i32 195, i32 199, i32 203, i32 207, i32 211, i32 215, i32 219, i32 223, i32 227, i32 231, i32 235, i32 239, i32 243, i32 247, i32 251, i32 255>
store <64 x i16> %strided.vec0, ptr %out.vec0, align 64
store <64 x i16> %strided.vec1, ptr %out.vec1, align 64
store <64 x i16> %strided.vec2, ptr %out.vec2, align 64
store <64 x i16> %strided.vec3, ptr %out.vec3, align 64
ret void
}
Codebase Browser
llvm