; 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_i64_stride2_vf2(ptr %in.vec, ptr %out.vec0, ptr %out.vec1) nounwind {
;
; SSE-LABEL: load_i64_stride2_vf2:
; SSE: # %bb.0:
; SSE-NEXT: movaps (%rdi), %xmm0
; SSE-NEXT: movaps 16(%rdi), %xmm1
; SSE-NEXT: movaps %xmm0, %xmm2
; SSE-NEXT: movlhps {{.*#+}} xmm2 = xmm2[0],xmm1[0]
; SSE-NEXT: unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
; SSE-NEXT: movaps %xmm2, (%rsi)
; SSE-NEXT: movaps %xmm0, (%rdx)
; SSE-NEXT: retq
;
; AVX-LABEL: load_i64_stride2_vf2:
; AVX: # %bb.0:
; AVX-NEXT: vmovaps (%rdi), %xmm0
; AVX-NEXT: vmovaps 16(%rdi), %xmm1
; AVX-NEXT: vmovlhps {{.*#+}} xmm2 = xmm0[0],xmm1[0]
; AVX-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
; AVX-NEXT: vmovaps %xmm2, (%rsi)
; AVX-NEXT: vmovaps %xmm0, (%rdx)
; AVX-NEXT: retq
;
; AVX2-LABEL: load_i64_stride2_vf2:
; AVX2: # %bb.0:
; AVX2-NEXT: vmovaps (%rdi), %xmm0
; AVX2-NEXT: vmovaps 16(%rdi), %xmm1
; AVX2-NEXT: vmovlhps {{.*#+}} xmm2 = xmm0[0],xmm1[0]
; AVX2-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
; AVX2-NEXT: vmovaps %xmm2, (%rsi)
; AVX2-NEXT: vmovaps %xmm0, (%rdx)
; AVX2-NEXT: retq
;
; AVX2-FP-LABEL: load_i64_stride2_vf2:
; AVX2-FP: # %bb.0:
; AVX2-FP-NEXT: vmovaps (%rdi), %xmm0
; AVX2-FP-NEXT: vmovaps 16(%rdi), %xmm1
; AVX2-FP-NEXT: vmovlhps {{.*#+}} xmm2 = xmm0[0],xmm1[0]
; AVX2-FP-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
; AVX2-FP-NEXT: vmovaps %xmm2, (%rsi)
; AVX2-FP-NEXT: vmovaps %xmm0, (%rdx)
; AVX2-FP-NEXT: retq
;
; AVX2-FCP-LABEL: load_i64_stride2_vf2:
; AVX2-FCP: # %bb.0:
; AVX2-FCP-NEXT: vmovaps (%rdi), %xmm0
; AVX2-FCP-NEXT: vmovaps 16(%rdi), %xmm1
; AVX2-FCP-NEXT: vmovlhps {{.*#+}} xmm2 = xmm0[0],xmm1[0]
; AVX2-FCP-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
; AVX2-FCP-NEXT: vmovaps %xmm2, (%rsi)
; AVX2-FCP-NEXT: vmovaps %xmm0, (%rdx)
; AVX2-FCP-NEXT: retq
;
; AVX512-LABEL: load_i64_stride2_vf2:
; AVX512: # %bb.0:
; AVX512-NEXT: vmovaps (%rdi), %xmm0
; AVX512-NEXT: vmovaps 16(%rdi), %xmm1
; AVX512-NEXT: vmovlhps {{.*#+}} xmm2 = xmm0[0],xmm1[0]
; AVX512-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
; AVX512-NEXT: vmovaps %xmm2, (%rsi)
; AVX512-NEXT: vmovaps %xmm0, (%rdx)
; AVX512-NEXT: retq
;
; AVX512-FCP-LABEL: load_i64_stride2_vf2:
; AVX512-FCP: # %bb.0:
; AVX512-FCP-NEXT: vmovaps (%rdi), %xmm0
; AVX512-FCP-NEXT: vmovaps 16(%rdi), %xmm1
; AVX512-FCP-NEXT: vmovlhps {{.*#+}} xmm2 = xmm0[0],xmm1[0]
; AVX512-FCP-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
; AVX512-FCP-NEXT: vmovaps %xmm2, (%rsi)
; AVX512-FCP-NEXT: vmovaps %xmm0, (%rdx)
; AVX512-FCP-NEXT: retq
;
; AVX512DQ-LABEL: load_i64_stride2_vf2:
; AVX512DQ: # %bb.0:
; AVX512DQ-NEXT: vmovaps (%rdi), %xmm0
; AVX512DQ-NEXT: vmovaps 16(%rdi), %xmm1
; AVX512DQ-NEXT: vmovlhps {{.*#+}} xmm2 = xmm0[0],xmm1[0]
; AVX512DQ-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
; AVX512DQ-NEXT: vmovaps %xmm2, (%rsi)
; AVX512DQ-NEXT: vmovaps %xmm0, (%rdx)
; AVX512DQ-NEXT: retq
;
; AVX512DQ-FCP-LABEL: load_i64_stride2_vf2:
; AVX512DQ-FCP: # %bb.0:
; AVX512DQ-FCP-NEXT: vmovaps (%rdi), %xmm0
; AVX512DQ-FCP-NEXT: vmovaps 16(%rdi), %xmm1
; AVX512DQ-FCP-NEXT: vmovlhps {{.*#+}} xmm2 = xmm0[0],xmm1[0]
; AVX512DQ-FCP-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
; AVX512DQ-FCP-NEXT: vmovaps %xmm2, (%rsi)
; AVX512DQ-FCP-NEXT: vmovaps %xmm0, (%rdx)
; AVX512DQ-FCP-NEXT: retq
;
; AVX512BW-LABEL: load_i64_stride2_vf2:
; AVX512BW: # %bb.0:
; AVX512BW-NEXT: vmovaps (%rdi), %xmm0
; AVX512BW-NEXT: vmovaps 16(%rdi), %xmm1
; AVX512BW-NEXT: vmovlhps {{.*#+}} xmm2 = xmm0[0],xmm1[0]
; AVX512BW-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
; AVX512BW-NEXT: vmovaps %xmm2, (%rsi)
; AVX512BW-NEXT: vmovaps %xmm0, (%rdx)
; AVX512BW-NEXT: retq
;
; AVX512BW-FCP-LABEL: load_i64_stride2_vf2:
; AVX512BW-FCP: # %bb.0:
; AVX512BW-FCP-NEXT: vmovaps (%rdi), %xmm0
; AVX512BW-FCP-NEXT: vmovaps 16(%rdi), %xmm1
; AVX512BW-FCP-NEXT: vmovlhps {{.*#+}} xmm2 = xmm0[0],xmm1[0]
; AVX512BW-FCP-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
; AVX512BW-FCP-NEXT: vmovaps %xmm2, (%rsi)
; AVX512BW-FCP-NEXT: vmovaps %xmm0, (%rdx)
; AVX512BW-FCP-NEXT: retq
;
; AVX512DQ-BW-LABEL: load_i64_stride2_vf2:
; AVX512DQ-BW: # %bb.0:
; AVX512DQ-BW-NEXT: vmovaps (%rdi), %xmm0
; AVX512DQ-BW-NEXT: vmovaps 16(%rdi), %xmm1
; AVX512DQ-BW-NEXT: vmovlhps {{.*#+}} xmm2 = xmm0[0],xmm1[0]
; AVX512DQ-BW-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
; AVX512DQ-BW-NEXT: vmovaps %xmm2, (%rsi)
; AVX512DQ-BW-NEXT: vmovaps %xmm0, (%rdx)
; AVX512DQ-BW-NEXT: retq
;
; AVX512DQ-BW-FCP-LABEL: load_i64_stride2_vf2:
; AVX512DQ-BW-FCP: # %bb.0:
; AVX512DQ-BW-FCP-NEXT: vmovaps (%rdi), %xmm0
; AVX512DQ-BW-FCP-NEXT: vmovaps 16(%rdi), %xmm1
; AVX512DQ-BW-FCP-NEXT: vmovlhps {{.*#+}} xmm2 = xmm0[0],xmm1[0]
; AVX512DQ-BW-FCP-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
; AVX512DQ-BW-FCP-NEXT: vmovaps %xmm2, (%rsi)
; AVX512DQ-BW-FCP-NEXT: vmovaps %xmm0, (%rdx)
; AVX512DQ-BW-FCP-NEXT: retq
%wide.vec = load <4 x i64>, ptr %in.vec, align 64
%strided.vec0 = shufflevector <4 x i64> %wide.vec, <4 x i64> poison, <2 x i32> <i32 0, i32 2>
%strided.vec1 = shufflevector <4 x i64> %wide.vec, <4 x i64> poison, <2 x i32> <i32 1, i32 3>
store <2 x i64> %strided.vec0, ptr %out.vec0, align 64
store <2 x i64> %strided.vec1, ptr %out.vec1, align 64
ret void
}
define void @load_i64_stride2_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1) nounwind {
;
;
;
;
; SSE-LABEL: load_i64_stride2_vf4:
; SSE: # %bb.0:
; SSE-NEXT: movaps (%rdi), %xmm0
; SSE-NEXT: movaps 16(%rdi), %xmm1
; SSE-NEXT: movaps 32(%rdi), %xmm2
; SSE-NEXT: movaps 48(%rdi), %xmm3
; SSE-NEXT: movaps %xmm2, %xmm4
; SSE-NEXT: movlhps {{.*#+}} xmm4 = xmm4[0],xmm3[0]
; SSE-NEXT: movaps %xmm0, %xmm5
; SSE-NEXT: movlhps {{.*#+}} xmm5 = xmm5[0],xmm1[0]
; SSE-NEXT: unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm3[1]
; SSE-NEXT: unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
; SSE-NEXT: movaps %xmm5, (%rsi)
; SSE-NEXT: movaps %xmm4, 16(%rsi)
; SSE-NEXT: movaps %xmm0, (%rdx)
; SSE-NEXT: movaps %xmm2, 16(%rdx)
; SSE-NEXT: retq
;
; AVX-LABEL: load_i64_stride2_vf4:
; AVX: # %bb.0:
; AVX-NEXT: vmovaps (%rdi), %ymm0
; AVX-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3],mem[2,3]
; AVX-NEXT: vinsertf128 $1, 32(%rdi), %ymm0, %ymm0
; AVX-NEXT: vunpcklpd {{.*#+}} ymm2 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
; AVX-NEXT: vmovaps %ymm2, (%rsi)
; AVX-NEXT: vmovaps %ymm0, (%rdx)
; AVX-NEXT: vzeroupper
; AVX-NEXT: retq
;
; AVX2-LABEL: load_i64_stride2_vf4:
; AVX2: # %bb.0:
; AVX2-NEXT: vmovaps (%rdi), %ymm0
; AVX2-NEXT: vmovaps 32(%rdi), %ymm1
; AVX2-NEXT: vunpcklpd {{.*#+}} ymm2 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX2-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,2,1,3]
; AVX2-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,2,1,3]
; AVX2-NEXT: vmovaps %ymm2, (%rsi)
; AVX2-NEXT: vmovaps %ymm0, (%rdx)
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
; AVX2-FP-LABEL: load_i64_stride2_vf4:
; AVX2-FP: # %bb.0:
; AVX2-FP-NEXT: vmovaps (%rdi), %ymm0
; AVX2-FP-NEXT: vmovaps 32(%rdi), %ymm1
; AVX2-FP-NEXT: vunpcklpd {{.*#+}} ymm2 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX2-FP-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,2,1,3]
; AVX2-FP-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
; AVX2-FP-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,2,1,3]
; AVX2-FP-NEXT: vmovaps %ymm2, (%rsi)
; AVX2-FP-NEXT: vmovaps %ymm0, (%rdx)
; AVX2-FP-NEXT: vzeroupper
; AVX2-FP-NEXT: retq
;
; AVX2-FCP-LABEL: load_i64_stride2_vf4:
; AVX2-FCP: # %bb.0:
; AVX2-FCP-NEXT: vmovaps (%rdi), %ymm0
; AVX2-FCP-NEXT: vmovaps 32(%rdi), %ymm1
; AVX2-FCP-NEXT: vunpcklpd {{.*#+}} ymm2 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX2-FCP-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,2,1,3]
; AVX2-FCP-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
; AVX2-FCP-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,2,1,3]
; AVX2-FCP-NEXT: vmovaps %ymm2, (%rsi)
; AVX2-FCP-NEXT: vmovaps %ymm0, (%rdx)
; AVX2-FCP-NEXT: vzeroupper
; AVX2-FCP-NEXT: retq
;
; AVX512-LABEL: load_i64_stride2_vf4:
; AVX512: # %bb.0:
; AVX512-NEXT: vmovaps (%rdi), %ymm0
; AVX512-NEXT: vmovaps 32(%rdi), %ymm1
; AVX512-NEXT: vunpcklpd {{.*#+}} ymm2 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX512-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,2,1,3]
; AVX512-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
; AVX512-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,2,1,3]
; AVX512-NEXT: vmovaps %ymm2, (%rsi)
; AVX512-NEXT: vmovaps %ymm0, (%rdx)
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
;
; AVX512-FCP-LABEL: load_i64_stride2_vf4:
; AVX512-FCP: # %bb.0:
; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} ymm0 = [0,2,4,6]
; AVX512-FCP-NEXT: vmovdqa (%rdi), %ymm1
; AVX512-FCP-NEXT: vmovdqa 32(%rdi), %ymm2
; AVX512-FCP-NEXT: vpermi2q %ymm2, %ymm1, %ymm0
; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} ymm3 = [1,3,5,7]
; AVX512-FCP-NEXT: vpermi2q %ymm2, %ymm1, %ymm3
; AVX512-FCP-NEXT: vmovdqa %ymm0, (%rsi)
; AVX512-FCP-NEXT: vmovdqa %ymm3, (%rdx)
; AVX512-FCP-NEXT: vzeroupper
; AVX512-FCP-NEXT: retq
;
; AVX512DQ-LABEL: load_i64_stride2_vf4:
; AVX512DQ: # %bb.0:
; AVX512DQ-NEXT: vmovaps (%rdi), %ymm0
; AVX512DQ-NEXT: vmovaps 32(%rdi), %ymm1
; AVX512DQ-NEXT: vunpcklpd {{.*#+}} ymm2 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX512DQ-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,2,1,3]
; AVX512DQ-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
; AVX512DQ-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,2,1,3]
; AVX512DQ-NEXT: vmovaps %ymm2, (%rsi)
; AVX512DQ-NEXT: vmovaps %ymm0, (%rdx)
; AVX512DQ-NEXT: vzeroupper
; AVX512DQ-NEXT: retq
;
; AVX512DQ-FCP-LABEL: load_i64_stride2_vf4:
; AVX512DQ-FCP: # %bb.0:
; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} ymm0 = [0,2,4,6]
; AVX512DQ-FCP-NEXT: vmovdqa (%rdi), %ymm1
; AVX512DQ-FCP-NEXT: vmovdqa 32(%rdi), %ymm2
; AVX512DQ-FCP-NEXT: vpermi2q %ymm2, %ymm1, %ymm0
; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} ymm3 = [1,3,5,7]
; AVX512DQ-FCP-NEXT: vpermi2q %ymm2, %ymm1, %ymm3
; AVX512DQ-FCP-NEXT: vmovdqa %ymm0, (%rsi)
; AVX512DQ-FCP-NEXT: vmovdqa %ymm3, (%rdx)
; AVX512DQ-FCP-NEXT: vzeroupper
; AVX512DQ-FCP-NEXT: retq
;
; AVX512BW-LABEL: load_i64_stride2_vf4:
; AVX512BW: # %bb.0:
; AVX512BW-NEXT: vmovaps (%rdi), %ymm0
; AVX512BW-NEXT: vmovaps 32(%rdi), %ymm1
; AVX512BW-NEXT: vunpcklpd {{.*#+}} ymm2 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX512BW-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,2,1,3]
; AVX512BW-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
; AVX512BW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,2,1,3]
; AVX512BW-NEXT: vmovaps %ymm2, (%rsi)
; AVX512BW-NEXT: vmovaps %ymm0, (%rdx)
; AVX512BW-NEXT: vzeroupper
; AVX512BW-NEXT: retq
;
; AVX512BW-FCP-LABEL: load_i64_stride2_vf4:
; AVX512BW-FCP: # %bb.0:
; AVX512BW-FCP-NEXT: vpmovsxbq {{.*#+}} ymm0 = [0,2,4,6]
; AVX512BW-FCP-NEXT: vmovdqa (%rdi), %ymm1
; AVX512BW-FCP-NEXT: vmovdqa 32(%rdi), %ymm2
; AVX512BW-FCP-NEXT: vpermi2q %ymm2, %ymm1, %ymm0
; AVX512BW-FCP-NEXT: vpmovsxbq {{.*#+}} ymm3 = [1,3,5,7]
; AVX512BW-FCP-NEXT: vpermi2q %ymm2, %ymm1, %ymm3
; AVX512BW-FCP-NEXT: vmovdqa %ymm0, (%rsi)
; AVX512BW-FCP-NEXT: vmovdqa %ymm3, (%rdx)
; AVX512BW-FCP-NEXT: vzeroupper
; AVX512BW-FCP-NEXT: retq
;
; AVX512DQ-BW-LABEL: load_i64_stride2_vf4:
; AVX512DQ-BW: # %bb.0:
; AVX512DQ-BW-NEXT: vmovaps (%rdi), %ymm0
; AVX512DQ-BW-NEXT: vmovaps 32(%rdi), %ymm1
; AVX512DQ-BW-NEXT: vunpcklpd {{.*#+}} ymm2 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX512DQ-BW-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,2,1,3]
; AVX512DQ-BW-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
; AVX512DQ-BW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,2,1,3]
; AVX512DQ-BW-NEXT: vmovaps %ymm2, (%rsi)
; AVX512DQ-BW-NEXT: vmovaps %ymm0, (%rdx)
; AVX512DQ-BW-NEXT: vzeroupper
; AVX512DQ-BW-NEXT: retq
;
; AVX512DQ-BW-FCP-LABEL: load_i64_stride2_vf4:
; AVX512DQ-BW-FCP: # %bb.0:
; AVX512DQ-BW-FCP-NEXT: vpmovsxbq {{.*#+}} ymm0 = [0,2,4,6]
; AVX512DQ-BW-FCP-NEXT: vmovdqa (%rdi), %ymm1
; AVX512DQ-BW-FCP-NEXT: vmovdqa 32(%rdi), %ymm2
; AVX512DQ-BW-FCP-NEXT: vpermi2q %ymm2, %ymm1, %ymm0
; AVX512DQ-BW-FCP-NEXT: vpmovsxbq {{.*#+}} ymm3 = [1,3,5,7]
; AVX512DQ-BW-FCP-NEXT: vpermi2q %ymm2, %ymm1, %ymm3
; AVX512DQ-BW-FCP-NEXT: vmovdqa %ymm0, (%rsi)
; AVX512DQ-BW-FCP-NEXT: vmovdqa %ymm3, (%rdx)
; AVX512DQ-BW-FCP-NEXT: vzeroupper
; AVX512DQ-BW-FCP-NEXT: retq
%wide.vec = load <8 x i64>, ptr %in.vec, align 64
%strided.vec0 = shufflevector <8 x i64> %wide.vec, <8 x i64> poison, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
%strided.vec1 = shufflevector <8 x i64> %wide.vec, <8 x i64> poison, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
store <4 x i64> %strided.vec0, ptr %out.vec0, align 64
store <4 x i64> %strided.vec1, ptr %out.vec1, align 64
ret void
}
define void @load_i64_stride2_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1) nounwind {
;
;
;
; SSE-LABEL: load_i64_stride2_vf8:
; SSE: # %bb.0:
; SSE-NEXT: movaps (%rdi), %xmm0
; SSE-NEXT: movaps 16(%rdi), %xmm1
; SSE-NEXT: movaps 32(%rdi), %xmm2
; SSE-NEXT: movaps 48(%rdi), %xmm3
; SSE-NEXT: movaps 80(%rdi), %xmm4
; SSE-NEXT: movaps 64(%rdi), %xmm5
; SSE-NEXT: movaps 112(%rdi), %xmm6
; SSE-NEXT: movaps 96(%rdi), %xmm7
; SSE-NEXT: movaps %xmm7, %xmm8
; SSE-NEXT: movlhps {{.*#+}} xmm8 = xmm8[0],xmm6[0]
; SSE-NEXT: movaps %xmm5, %xmm9
; SSE-NEXT: movlhps {{.*#+}} xmm9 = xmm9[0],xmm4[0]
; SSE-NEXT: movaps %xmm2, %xmm10
; SSE-NEXT: movlhps {{.*#+}} xmm10 = xmm10[0],xmm3[0]
; SSE-NEXT: movaps %xmm0, %xmm11
; SSE-NEXT: movlhps {{.*#+}} xmm11 = xmm11[0],xmm1[0]
; SSE-NEXT: unpckhpd {{.*#+}} xmm7 = xmm7[1],xmm6[1]
; SSE-NEXT: unpckhpd {{.*#+}} xmm5 = xmm5[1],xmm4[1]
; SSE-NEXT: unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm3[1]
; SSE-NEXT: unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
; SSE-NEXT: movaps %xmm9, 32(%rsi)
; SSE-NEXT: movaps %xmm8, 48(%rsi)
; SSE-NEXT: movaps %xmm11, (%rsi)
; SSE-NEXT: movaps %xmm10, 16(%rsi)
; SSE-NEXT: movaps %xmm5, 32(%rdx)
; SSE-NEXT: movaps %xmm7, 48(%rdx)
; SSE-NEXT: movaps %xmm0, (%rdx)
; SSE-NEXT: movaps %xmm2, 16(%rdx)
; SSE-NEXT: retq
;
; AVX-LABEL: load_i64_stride2_vf8:
; AVX: # %bb.0:
; AVX-NEXT: vmovaps (%rdi), %ymm0
; AVX-NEXT: vmovaps 64(%rdi), %ymm1
; AVX-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm1[2,3],mem[2,3]
; AVX-NEXT: vinsertf128 $1, 96(%rdi), %ymm1, %ymm1
; AVX-NEXT: vunpcklpd {{.*#+}} ymm3 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
; AVX-NEXT: vperm2f128 {{.*#+}} ymm4 = ymm0[2,3],mem[2,3]
; AVX-NEXT: vinsertf128 $1, 32(%rdi), %ymm0, %ymm0
; AVX-NEXT: vunpcklpd {{.*#+}} ymm5 = ymm0[0],ymm4[0],ymm0[2],ymm4[2]
; AVX-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm2[1],ymm1[3],ymm2[3]
; AVX-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm4[1],ymm0[3],ymm4[3]
; AVX-NEXT: vmovaps %ymm5, (%rsi)
; AVX-NEXT: vmovaps %ymm3, 32(%rsi)
; AVX-NEXT: vmovaps %ymm0, (%rdx)
; AVX-NEXT: vmovaps %ymm1, 32(%rdx)
; AVX-NEXT: vzeroupper
; AVX-NEXT: retq
;
; AVX2-LABEL: load_i64_stride2_vf8:
; AVX2: # %bb.0:
; AVX2-NEXT: vmovaps (%rdi), %ymm0
; AVX2-NEXT: vmovaps 32(%rdi), %ymm1
; AVX2-NEXT: vmovaps 64(%rdi), %ymm2
; AVX2-NEXT: vmovaps 96(%rdi), %ymm3
; AVX2-NEXT: vunpcklpd {{.*#+}} ymm4 = ymm2[0],ymm3[0],ymm2[2],ymm3[2]
; AVX2-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,2,1,3]
; AVX2-NEXT: vunpcklpd {{.*#+}} ymm5 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX2-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,2,1,3]
; AVX2-NEXT: vunpckhpd {{.*#+}} ymm2 = ymm2[1],ymm3[1],ymm2[3],ymm3[3]
; AVX2-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,2,1,3]
; AVX2-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,2,1,3]
; AVX2-NEXT: vmovaps %ymm5, (%rsi)
; AVX2-NEXT: vmovaps %ymm4, 32(%rsi)
; AVX2-NEXT: vmovaps %ymm0, (%rdx)
; AVX2-NEXT: vmovaps %ymm2, 32(%rdx)
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
; AVX2-FP-LABEL: load_i64_stride2_vf8:
; AVX2-FP: # %bb.0:
; AVX2-FP-NEXT: vmovaps (%rdi), %ymm0
; AVX2-FP-NEXT: vmovaps 32(%rdi), %ymm1
; AVX2-FP-NEXT: vmovaps 64(%rdi), %ymm2
; AVX2-FP-NEXT: vmovaps 96(%rdi), %ymm3
; AVX2-FP-NEXT: vunpcklpd {{.*#+}} ymm4 = ymm2[0],ymm3[0],ymm2[2],ymm3[2]
; AVX2-FP-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,2,1,3]
; AVX2-FP-NEXT: vunpcklpd {{.*#+}} ymm5 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX2-FP-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,2,1,3]
; AVX2-FP-NEXT: vunpckhpd {{.*#+}} ymm2 = ymm2[1],ymm3[1],ymm2[3],ymm3[3]
; AVX2-FP-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,2,1,3]
; AVX2-FP-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
; AVX2-FP-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,2,1,3]
; AVX2-FP-NEXT: vmovaps %ymm5, (%rsi)
; AVX2-FP-NEXT: vmovaps %ymm4, 32(%rsi)
; AVX2-FP-NEXT: vmovaps %ymm0, (%rdx)
; AVX2-FP-NEXT: vmovaps %ymm2, 32(%rdx)
; AVX2-FP-NEXT: vzeroupper
; AVX2-FP-NEXT: retq
;
; AVX2-FCP-LABEL: load_i64_stride2_vf8:
; AVX2-FCP: # %bb.0:
; AVX2-FCP-NEXT: vmovaps (%rdi), %ymm0
; AVX2-FCP-NEXT: vmovaps 32(%rdi), %ymm1
; AVX2-FCP-NEXT: vmovaps 64(%rdi), %ymm2
; AVX2-FCP-NEXT: vmovaps 96(%rdi), %ymm3
; AVX2-FCP-NEXT: vunpcklpd {{.*#+}} ymm4 = ymm2[0],ymm3[0],ymm2[2],ymm3[2]
; AVX2-FCP-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,2,1,3]
; AVX2-FCP-NEXT: vunpcklpd {{.*#+}} ymm5 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX2-FCP-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,2,1,3]
; AVX2-FCP-NEXT: vunpckhpd {{.*#+}} ymm2 = ymm2[1],ymm3[1],ymm2[3],ymm3[3]
; AVX2-FCP-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,2,1,3]
; AVX2-FCP-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
; AVX2-FCP-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,2,1,3]
; AVX2-FCP-NEXT: vmovaps %ymm5, (%rsi)
; AVX2-FCP-NEXT: vmovaps %ymm4, 32(%rsi)
; AVX2-FCP-NEXT: vmovaps %ymm0, (%rdx)
; AVX2-FCP-NEXT: vmovaps %ymm2, 32(%rdx)
; AVX2-FCP-NEXT: vzeroupper
; AVX2-FCP-NEXT: retq
;
; AVX512-LABEL: load_i64_stride2_vf8:
; AVX512: # %bb.0:
; AVX512-NEXT: vmovdqa64 (%rdi), %zmm0
; AVX512-NEXT: vmovdqa64 64(%rdi), %zmm1
; AVX512-NEXT: vpmovsxbq {{.*#+}} zmm2 = [0,2,4,6,8,10,12,14]
; AVX512-NEXT: vpermi2q %zmm1, %zmm0, %zmm2
; AVX512-NEXT: vpmovsxbq {{.*#+}} zmm3 = [1,3,5,7,9,11,13,15]
; AVX512-NEXT: vpermi2q %zmm1, %zmm0, %zmm3
; AVX512-NEXT: vmovdqa64 %zmm2, (%rsi)
; AVX512-NEXT: vmovdqa64 %zmm3, (%rdx)
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
;
; AVX512-FCP-LABEL: load_i64_stride2_vf8:
; AVX512-FCP: # %bb.0:
; AVX512-FCP-NEXT: vmovdqa64 (%rdi), %zmm0
; AVX512-FCP-NEXT: vmovdqa64 64(%rdi), %zmm1
; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm2 = [0,2,4,6,8,10,12,14]
; AVX512-FCP-NEXT: vpermi2q %zmm1, %zmm0, %zmm2
; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm3 = [1,3,5,7,9,11,13,15]
; AVX512-FCP-NEXT: vpermi2q %zmm1, %zmm0, %zmm3
; AVX512-FCP-NEXT: vmovdqa64 %zmm2, (%rsi)
; AVX512-FCP-NEXT: vmovdqa64 %zmm3, (%rdx)
; AVX512-FCP-NEXT: vzeroupper
; AVX512-FCP-NEXT: retq
;
; AVX512DQ-LABEL: load_i64_stride2_vf8:
; AVX512DQ: # %bb.0:
; AVX512DQ-NEXT: vmovdqa64 (%rdi), %zmm0
; AVX512DQ-NEXT: vmovdqa64 64(%rdi), %zmm1
; AVX512DQ-NEXT: vpmovsxbq {{.*#+}} zmm2 = [0,2,4,6,8,10,12,14]
; AVX512DQ-NEXT: vpermi2q %zmm1, %zmm0, %zmm2
; AVX512DQ-NEXT: vpmovsxbq {{.*#+}} zmm3 = [1,3,5,7,9,11,13,15]
; AVX512DQ-NEXT: vpermi2q %zmm1, %zmm0, %zmm3
; AVX512DQ-NEXT: vmovdqa64 %zmm2, (%rsi)
; AVX512DQ-NEXT: vmovdqa64 %zmm3, (%rdx)
; AVX512DQ-NEXT: vzeroupper
; AVX512DQ-NEXT: retq
;
; AVX512DQ-FCP-LABEL: load_i64_stride2_vf8:
; AVX512DQ-FCP: # %bb.0:
; AVX512DQ-FCP-NEXT: vmovdqa64 (%rdi), %zmm0
; AVX512DQ-FCP-NEXT: vmovdqa64 64(%rdi), %zmm1
; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm2 = [0,2,4,6,8,10,12,14]
; AVX512DQ-FCP-NEXT: vpermi2q %zmm1, %zmm0, %zmm2
; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm3 = [1,3,5,7,9,11,13,15]
; AVX512DQ-FCP-NEXT: vpermi2q %zmm1, %zmm0, %zmm3
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm2, (%rsi)
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm3, (%rdx)
; AVX512DQ-FCP-NEXT: vzeroupper
; AVX512DQ-FCP-NEXT: retq
;
; AVX512BW-LABEL: load_i64_stride2_vf8:
; AVX512BW: # %bb.0:
; AVX512BW-NEXT: vmovdqa64 (%rdi), %zmm0
; AVX512BW-NEXT: vmovdqa64 64(%rdi), %zmm1
; AVX512BW-NEXT: vpmovsxbq {{.*#+}} zmm2 = [0,2,4,6,8,10,12,14]
; AVX512BW-NEXT: vpermi2q %zmm1, %zmm0, %zmm2
; AVX512BW-NEXT: vpmovsxbq {{.*#+}} zmm3 = [1,3,5,7,9,11,13,15]
; AVX512BW-NEXT: vpermi2q %zmm1, %zmm0, %zmm3
; AVX512BW-NEXT: vmovdqa64 %zmm2, (%rsi)
; AVX512BW-NEXT: vmovdqa64 %zmm3, (%rdx)
; AVX512BW-NEXT: vzeroupper
; AVX512BW-NEXT: retq
;
; AVX512BW-FCP-LABEL: load_i64_stride2_vf8:
; AVX512BW-FCP: # %bb.0:
; AVX512BW-FCP-NEXT: vmovdqa64 (%rdi), %zmm0
; AVX512BW-FCP-NEXT: vmovdqa64 64(%rdi), %zmm1
; AVX512BW-FCP-NEXT: vpmovsxbq {{.*#+}} zmm2 = [0,2,4,6,8,10,12,14]
; AVX512BW-FCP-NEXT: vpermi2q %zmm1, %zmm0, %zmm2
; AVX512BW-FCP-NEXT: vpmovsxbq {{.*#+}} zmm3 = [1,3,5,7,9,11,13,15]
; AVX512BW-FCP-NEXT: vpermi2q %zmm1, %zmm0, %zmm3
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm2, (%rsi)
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm3, (%rdx)
; AVX512BW-FCP-NEXT: vzeroupper
; AVX512BW-FCP-NEXT: retq
;
; AVX512DQ-BW-LABEL: load_i64_stride2_vf8:
; AVX512DQ-BW: # %bb.0:
; AVX512DQ-BW-NEXT: vmovdqa64 (%rdi), %zmm0
; AVX512DQ-BW-NEXT: vmovdqa64 64(%rdi), %zmm1
; AVX512DQ-BW-NEXT: vpmovsxbq {{.*#+}} zmm2 = [0,2,4,6,8,10,12,14]
; AVX512DQ-BW-NEXT: vpermi2q %zmm1, %zmm0, %zmm2
; AVX512DQ-BW-NEXT: vpmovsxbq {{.*#+}} zmm3 = [1,3,5,7,9,11,13,15]
; AVX512DQ-BW-NEXT: vpermi2q %zmm1, %zmm0, %zmm3
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm2, (%rsi)
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm3, (%rdx)
; AVX512DQ-BW-NEXT: vzeroupper
; AVX512DQ-BW-NEXT: retq
;
; AVX512DQ-BW-FCP-LABEL: load_i64_stride2_vf8:
; AVX512DQ-BW-FCP: # %bb.0:
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 (%rdi), %zmm0
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 64(%rdi), %zmm1
; AVX512DQ-BW-FCP-NEXT: vpmovsxbq {{.*#+}} zmm2 = [0,2,4,6,8,10,12,14]
; AVX512DQ-BW-FCP-NEXT: vpermi2q %zmm1, %zmm0, %zmm2
; AVX512DQ-BW-FCP-NEXT: vpmovsxbq {{.*#+}} zmm3 = [1,3,5,7,9,11,13,15]
; AVX512DQ-BW-FCP-NEXT: vpermi2q %zmm1, %zmm0, %zmm3
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm2, (%rsi)
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm3, (%rdx)
; AVX512DQ-BW-FCP-NEXT: vzeroupper
; AVX512DQ-BW-FCP-NEXT: retq
%wide.vec = load <16 x i64>, ptr %in.vec, align 64
%strided.vec0 = shufflevector <16 x i64> %wide.vec, <16 x i64> poison, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
%strided.vec1 = shufflevector <16 x i64> %wide.vec, <16 x i64> poison, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
store <8 x i64> %strided.vec0, ptr %out.vec0, align 64
store <8 x i64> %strided.vec1, ptr %out.vec1, align 64
ret void
}
define void @load_i64_stride2_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1) nounwind {
;
;
;
; SSE-LABEL: load_i64_stride2_vf16:
; SSE: # %bb.0:
; SSE-NEXT: movaps (%rdi), %xmm1
; SSE-NEXT: movaps 16(%rdi), %xmm8
; SSE-NEXT: movaps 32(%rdi), %xmm0
; SSE-NEXT: movaps 208(%rdi), %xmm11
; SSE-NEXT: movaps 192(%rdi), %xmm2
; SSE-NEXT: movaps 240(%rdi), %xmm10
; SSE-NEXT: movaps 224(%rdi), %xmm4
; SSE-NEXT: movaps 144(%rdi), %xmm14
; SSE-NEXT: movaps 128(%rdi), %xmm3
; SSE-NEXT: movaps 176(%rdi), %xmm12
; SSE-NEXT: movaps 160(%rdi), %xmm6
; SSE-NEXT: movaps 80(%rdi), %xmm13
; SSE-NEXT: movaps 64(%rdi), %xmm5
; SSE-NEXT: movaps 112(%rdi), %xmm15
; SSE-NEXT: movaps 96(%rdi), %xmm7
; SSE-NEXT: movaps %xmm5, %xmm9
; SSE-NEXT: movlhps {{.*#+}} xmm9 = xmm9[0],xmm13[0]
; SSE-NEXT: unpckhpd {{.*#+}} xmm5 = xmm5[1],xmm13[1]
; SSE-NEXT: movaps %xmm7, %xmm13
; SSE-NEXT: movlhps {{.*#+}} xmm13 = xmm13[0],xmm15[0]
; SSE-NEXT: unpckhpd {{.*#+}} xmm7 = xmm7[1],xmm15[1]
; SSE-NEXT: movaps %xmm3, %xmm15
; SSE-NEXT: movlhps {{.*#+}} xmm15 = xmm15[0],xmm14[0]
; SSE-NEXT: unpckhpd {{.*#+}} xmm3 = xmm3[1],xmm14[1]
; SSE-NEXT: movaps %xmm6, %xmm14
; SSE-NEXT: movlhps {{.*#+}} xmm14 = xmm14[0],xmm12[0]
; SSE-NEXT: unpckhpd {{.*#+}} xmm6 = xmm6[1],xmm12[1]
; SSE-NEXT: movaps %xmm2, %xmm12
; SSE-NEXT: movlhps {{.*#+}} xmm12 = xmm12[0],xmm11[0]
; SSE-NEXT: unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm11[1]
; SSE-NEXT: movaps %xmm4, %xmm11
; SSE-NEXT: movlhps {{.*#+}} xmm11 = xmm11[0],xmm10[0]
; SSE-NEXT: unpckhpd {{.*#+}} xmm4 = xmm4[1],xmm10[1]
; SSE-NEXT: movaps %xmm1, %xmm10
; SSE-NEXT: movlhps {{.*#+}} xmm10 = xmm10[0],xmm8[0]
; SSE-NEXT: unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm8[1]
; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movaps 48(%rdi), %xmm8
; SSE-NEXT: movaps %xmm0, %xmm1
; SSE-NEXT: movlhps {{.*#+}} xmm1 = xmm1[0],xmm8[0]
; SSE-NEXT: unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm8[1]
; SSE-NEXT: movaps %xmm12, 96(%rsi)
; SSE-NEXT: movaps %xmm11, 112(%rsi)
; SSE-NEXT: movaps %xmm15, 64(%rsi)
; SSE-NEXT: movaps %xmm14, 80(%rsi)
; SSE-NEXT: movaps %xmm9, 32(%rsi)
; SSE-NEXT: movaps %xmm13, 48(%rsi)
; SSE-NEXT: movaps %xmm10, (%rsi)
; SSE-NEXT: movaps %xmm1, 16(%rsi)
; SSE-NEXT: movaps %xmm4, 112(%rdx)
; SSE-NEXT: movaps %xmm2, 96(%rdx)
; SSE-NEXT: movaps %xmm6, 80(%rdx)
; SSE-NEXT: movaps %xmm3, 64(%rdx)
; SSE-NEXT: movaps %xmm7, 48(%rdx)
; SSE-NEXT: movaps %xmm5, 32(%rdx)
; SSE-NEXT: movaps %xmm0, 16(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, (%rdx)
; SSE-NEXT: retq
;
; AVX-LABEL: load_i64_stride2_vf16:
; AVX: # %bb.0:
; AVX-NEXT: vmovaps (%rdi), %ymm0
; AVX-NEXT: vmovaps 64(%rdi), %ymm1
; AVX-NEXT: vmovaps 128(%rdi), %ymm2
; AVX-NEXT: vmovaps 192(%rdi), %ymm3
; AVX-NEXT: vperm2f128 {{.*#+}} ymm4 = ymm3[2,3],mem[2,3]
; AVX-NEXT: vinsertf128 $1, 224(%rdi), %ymm3, %ymm3
; AVX-NEXT: vunpcklpd {{.*#+}} ymm5 = ymm3[0],ymm4[0],ymm3[2],ymm4[2]
; AVX-NEXT: vperm2f128 {{.*#+}} ymm6 = ymm1[2,3],mem[2,3]
; AVX-NEXT: vinsertf128 $1, 96(%rdi), %ymm1, %ymm1
; AVX-NEXT: vunpcklpd {{.*#+}} ymm7 = ymm1[0],ymm6[0],ymm1[2],ymm6[2]
; AVX-NEXT: vperm2f128 {{.*#+}} ymm8 = ymm0[2,3],mem[2,3]
; AVX-NEXT: vinsertf128 $1, 32(%rdi), %ymm0, %ymm0
; AVX-NEXT: vunpcklpd {{.*#+}} ymm9 = ymm0[0],ymm8[0],ymm0[2],ymm8[2]
; AVX-NEXT: vperm2f128 {{.*#+}} ymm10 = ymm2[2,3],mem[2,3]
; AVX-NEXT: vinsertf128 $1, 160(%rdi), %ymm2, %ymm2
; AVX-NEXT: vunpcklpd {{.*#+}} ymm11 = ymm2[0],ymm10[0],ymm2[2],ymm10[2]
; AVX-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm6[1],ymm1[3],ymm6[3]
; AVX-NEXT: vunpckhpd {{.*#+}} ymm3 = ymm3[1],ymm4[1],ymm3[3],ymm4[3]
; AVX-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm8[1],ymm0[3],ymm8[3]
; AVX-NEXT: vunpckhpd {{.*#+}} ymm2 = ymm2[1],ymm10[1],ymm2[3],ymm10[3]
; AVX-NEXT: vmovaps %ymm11, 64(%rsi)
; AVX-NEXT: vmovaps %ymm9, (%rsi)
; AVX-NEXT: vmovaps %ymm7, 32(%rsi)
; AVX-NEXT: vmovaps %ymm5, 96(%rsi)
; AVX-NEXT: vmovaps %ymm2, 64(%rdx)
; AVX-NEXT: vmovaps %ymm0, (%rdx)
; AVX-NEXT: vmovaps %ymm3, 96(%rdx)
; AVX-NEXT: vmovaps %ymm1, 32(%rdx)
; AVX-NEXT: vzeroupper
; AVX-NEXT: retq
;
; AVX2-LABEL: load_i64_stride2_vf16:
; AVX2: # %bb.0:
; AVX2-NEXT: vmovaps (%rdi), %ymm0
; AVX2-NEXT: vmovaps 32(%rdi), %ymm1
; AVX2-NEXT: vmovaps 64(%rdi), %ymm2
; AVX2-NEXT: vmovaps 96(%rdi), %ymm3
; AVX2-NEXT: vmovaps 160(%rdi), %ymm4
; AVX2-NEXT: vmovaps 128(%rdi), %ymm5
; AVX2-NEXT: vmovaps 224(%rdi), %ymm6
; AVX2-NEXT: vmovaps 192(%rdi), %ymm7
; AVX2-NEXT: vunpcklpd {{.*#+}} ymm8 = ymm7[0],ymm6[0],ymm7[2],ymm6[2]
; AVX2-NEXT: vpermpd {{.*#+}} ymm8 = ymm8[0,2,1,3]
; AVX2-NEXT: vunpcklpd {{.*#+}} ymm9 = ymm5[0],ymm4[0],ymm5[2],ymm4[2]
; AVX2-NEXT: vpermpd {{.*#+}} ymm9 = ymm9[0,2,1,3]
; AVX2-NEXT: vunpcklpd {{.*#+}} ymm10 = ymm2[0],ymm3[0],ymm2[2],ymm3[2]
; AVX2-NEXT: vpermpd {{.*#+}} ymm10 = ymm10[0,2,1,3]
; AVX2-NEXT: vunpcklpd {{.*#+}} ymm11 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX2-NEXT: vpermpd {{.*#+}} ymm11 = ymm11[0,2,1,3]
; AVX2-NEXT: vunpckhpd {{.*#+}} ymm6 = ymm7[1],ymm6[1],ymm7[3],ymm6[3]
; AVX2-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,2,1,3]
; AVX2-NEXT: vunpckhpd {{.*#+}} ymm4 = ymm5[1],ymm4[1],ymm5[3],ymm4[3]
; AVX2-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,2,1,3]
; AVX2-NEXT: vunpckhpd {{.*#+}} ymm2 = ymm2[1],ymm3[1],ymm2[3],ymm3[3]
; AVX2-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,2,1,3]
; AVX2-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,2,1,3]
; AVX2-NEXT: vmovaps %ymm9, 64(%rsi)
; AVX2-NEXT: vmovaps %ymm11, (%rsi)
; AVX2-NEXT: vmovaps %ymm8, 96(%rsi)
; AVX2-NEXT: vmovaps %ymm10, 32(%rsi)
; AVX2-NEXT: vmovaps %ymm4, 64(%rdx)
; AVX2-NEXT: vmovaps %ymm0, (%rdx)
; AVX2-NEXT: vmovaps %ymm6, 96(%rdx)
; AVX2-NEXT: vmovaps %ymm2, 32(%rdx)
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
; AVX2-FP-LABEL: load_i64_stride2_vf16:
; AVX2-FP: # %bb.0:
; AVX2-FP-NEXT: vmovaps (%rdi), %ymm0
; AVX2-FP-NEXT: vmovaps 32(%rdi), %ymm1
; AVX2-FP-NEXT: vmovaps 64(%rdi), %ymm2
; AVX2-FP-NEXT: vmovaps 96(%rdi), %ymm3
; AVX2-FP-NEXT: vmovaps 160(%rdi), %ymm4
; AVX2-FP-NEXT: vmovaps 128(%rdi), %ymm5
; AVX2-FP-NEXT: vmovaps 224(%rdi), %ymm6
; AVX2-FP-NEXT: vmovaps 192(%rdi), %ymm7
; AVX2-FP-NEXT: vunpcklpd {{.*#+}} ymm8 = ymm7[0],ymm6[0],ymm7[2],ymm6[2]
; AVX2-FP-NEXT: vpermpd {{.*#+}} ymm8 = ymm8[0,2,1,3]
; AVX2-FP-NEXT: vunpcklpd {{.*#+}} ymm9 = ymm5[0],ymm4[0],ymm5[2],ymm4[2]
; AVX2-FP-NEXT: vpermpd {{.*#+}} ymm9 = ymm9[0,2,1,3]
; AVX2-FP-NEXT: vunpcklpd {{.*#+}} ymm10 = ymm2[0],ymm3[0],ymm2[2],ymm3[2]
; AVX2-FP-NEXT: vpermpd {{.*#+}} ymm10 = ymm10[0,2,1,3]
; AVX2-FP-NEXT: vunpcklpd {{.*#+}} ymm11 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX2-FP-NEXT: vpermpd {{.*#+}} ymm11 = ymm11[0,2,1,3]
; AVX2-FP-NEXT: vunpckhpd {{.*#+}} ymm6 = ymm7[1],ymm6[1],ymm7[3],ymm6[3]
; AVX2-FP-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,2,1,3]
; AVX2-FP-NEXT: vunpckhpd {{.*#+}} ymm4 = ymm5[1],ymm4[1],ymm5[3],ymm4[3]
; AVX2-FP-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,2,1,3]
; AVX2-FP-NEXT: vunpckhpd {{.*#+}} ymm2 = ymm2[1],ymm3[1],ymm2[3],ymm3[3]
; AVX2-FP-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,2,1,3]
; AVX2-FP-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
; AVX2-FP-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,2,1,3]
; AVX2-FP-NEXT: vmovaps %ymm9, 64(%rsi)
; AVX2-FP-NEXT: vmovaps %ymm11, (%rsi)
; AVX2-FP-NEXT: vmovaps %ymm8, 96(%rsi)
; AVX2-FP-NEXT: vmovaps %ymm10, 32(%rsi)
; AVX2-FP-NEXT: vmovaps %ymm4, 64(%rdx)
; AVX2-FP-NEXT: vmovaps %ymm0, (%rdx)
; AVX2-FP-NEXT: vmovaps %ymm6, 96(%rdx)
; AVX2-FP-NEXT: vmovaps %ymm2, 32(%rdx)
; AVX2-FP-NEXT: vzeroupper
; AVX2-FP-NEXT: retq
;
; AVX2-FCP-LABEL: load_i64_stride2_vf16:
; AVX2-FCP: # %bb.0:
; AVX2-FCP-NEXT: vmovaps (%rdi), %ymm0
; AVX2-FCP-NEXT: vmovaps 32(%rdi), %ymm1
; AVX2-FCP-NEXT: vmovaps 64(%rdi), %ymm2
; AVX2-FCP-NEXT: vmovaps 96(%rdi), %ymm3
; AVX2-FCP-NEXT: vmovaps 160(%rdi), %ymm4
; AVX2-FCP-NEXT: vmovaps 128(%rdi), %ymm5
; AVX2-FCP-NEXT: vmovaps 224(%rdi), %ymm6
; AVX2-FCP-NEXT: vmovaps 192(%rdi), %ymm7
; AVX2-FCP-NEXT: vunpcklpd {{.*#+}} ymm8 = ymm7[0],ymm6[0],ymm7[2],ymm6[2]
; AVX2-FCP-NEXT: vpermpd {{.*#+}} ymm8 = ymm8[0,2,1,3]
; AVX2-FCP-NEXT: vunpcklpd {{.*#+}} ymm9 = ymm5[0],ymm4[0],ymm5[2],ymm4[2]
; AVX2-FCP-NEXT: vpermpd {{.*#+}} ymm9 = ymm9[0,2,1,3]
; AVX2-FCP-NEXT: vunpcklpd {{.*#+}} ymm10 = ymm2[0],ymm3[0],ymm2[2],ymm3[2]
; AVX2-FCP-NEXT: vpermpd {{.*#+}} ymm10 = ymm10[0,2,1,3]
; AVX2-FCP-NEXT: vunpcklpd {{.*#+}} ymm11 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX2-FCP-NEXT: vpermpd {{.*#+}} ymm11 = ymm11[0,2,1,3]
; AVX2-FCP-NEXT: vunpckhpd {{.*#+}} ymm6 = ymm7[1],ymm6[1],ymm7[3],ymm6[3]
; AVX2-FCP-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,2,1,3]
; AVX2-FCP-NEXT: vunpckhpd {{.*#+}} ymm4 = ymm5[1],ymm4[1],ymm5[3],ymm4[3]
; AVX2-FCP-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,2,1,3]
; AVX2-FCP-NEXT: vunpckhpd {{.*#+}} ymm2 = ymm2[1],ymm3[1],ymm2[3],ymm3[3]
; AVX2-FCP-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,2,1,3]
; AVX2-FCP-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
; AVX2-FCP-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,2,1,3]
; AVX2-FCP-NEXT: vmovaps %ymm9, 64(%rsi)
; AVX2-FCP-NEXT: vmovaps %ymm11, (%rsi)
; AVX2-FCP-NEXT: vmovaps %ymm8, 96(%rsi)
; AVX2-FCP-NEXT: vmovaps %ymm10, 32(%rsi)
; AVX2-FCP-NEXT: vmovaps %ymm4, 64(%rdx)
; AVX2-FCP-NEXT: vmovaps %ymm0, (%rdx)
; AVX2-FCP-NEXT: vmovaps %ymm6, 96(%rdx)
; AVX2-FCP-NEXT: vmovaps %ymm2, 32(%rdx)
; AVX2-FCP-NEXT: vzeroupper
; AVX2-FCP-NEXT: retq
;
; AVX512-LABEL: load_i64_stride2_vf16:
; AVX512: # %bb.0:
; AVX512-NEXT: vmovdqa64 (%rdi), %zmm0
; AVX512-NEXT: vmovdqa64 64(%rdi), %zmm1
; AVX512-NEXT: vmovdqa64 128(%rdi), %zmm2
; AVX512-NEXT: vmovdqa64 192(%rdi), %zmm3
; AVX512-NEXT: vpmovsxbq {{.*#+}} zmm4 = [0,2,4,6,8,10,12,14]
; AVX512-NEXT: vmovdqa64 %zmm0, %zmm5
; AVX512-NEXT: vpermt2q %zmm1, %zmm4, %zmm5
; AVX512-NEXT: vpermi2q %zmm3, %zmm2, %zmm4
; AVX512-NEXT: vpmovsxbq {{.*#+}} zmm6 = [1,3,5,7,9,11,13,15]
; AVX512-NEXT: vpermt2q %zmm1, %zmm6, %zmm0
; AVX512-NEXT: vpermt2q %zmm3, %zmm6, %zmm2
; AVX512-NEXT: vmovdqa64 %zmm4, 64(%rsi)
; AVX512-NEXT: vmovdqa64 %zmm5, (%rsi)
; AVX512-NEXT: vmovdqa64 %zmm2, 64(%rdx)
; AVX512-NEXT: vmovdqa64 %zmm0, (%rdx)
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
;
; AVX512-FCP-LABEL: load_i64_stride2_vf16:
; AVX512-FCP: # %bb.0:
; AVX512-FCP-NEXT: vmovdqa64 (%rdi), %zmm0
; AVX512-FCP-NEXT: vmovdqa64 64(%rdi), %zmm1
; AVX512-FCP-NEXT: vmovdqa64 128(%rdi), %zmm2
; AVX512-FCP-NEXT: vmovdqa64 192(%rdi), %zmm3
; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm4 = [0,2,4,6,8,10,12,14]
; AVX512-FCP-NEXT: vmovdqa64 %zmm0, %zmm5
; AVX512-FCP-NEXT: vpermt2q %zmm1, %zmm4, %zmm5
; AVX512-FCP-NEXT: vpermi2q %zmm3, %zmm2, %zmm4
; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm6 = [1,3,5,7,9,11,13,15]
; AVX512-FCP-NEXT: vpermt2q %zmm1, %zmm6, %zmm0
; AVX512-FCP-NEXT: vpermt2q %zmm3, %zmm6, %zmm2
; AVX512-FCP-NEXT: vmovdqa64 %zmm4, 64(%rsi)
; AVX512-FCP-NEXT: vmovdqa64 %zmm5, (%rsi)
; AVX512-FCP-NEXT: vmovdqa64 %zmm2, 64(%rdx)
; AVX512-FCP-NEXT: vmovdqa64 %zmm0, (%rdx)
; AVX512-FCP-NEXT: vzeroupper
; AVX512-FCP-NEXT: retq
;
; AVX512DQ-LABEL: load_i64_stride2_vf16:
; AVX512DQ: # %bb.0:
; AVX512DQ-NEXT: vmovdqa64 (%rdi), %zmm0
; AVX512DQ-NEXT: vmovdqa64 64(%rdi), %zmm1
; AVX512DQ-NEXT: vmovdqa64 128(%rdi), %zmm2
; AVX512DQ-NEXT: vmovdqa64 192(%rdi), %zmm3
; AVX512DQ-NEXT: vpmovsxbq {{.*#+}} zmm4 = [0,2,4,6,8,10,12,14]
; AVX512DQ-NEXT: vmovdqa64 %zmm0, %zmm5
; AVX512DQ-NEXT: vpermt2q %zmm1, %zmm4, %zmm5
; AVX512DQ-NEXT: vpermi2q %zmm3, %zmm2, %zmm4
; AVX512DQ-NEXT: vpmovsxbq {{.*#+}} zmm6 = [1,3,5,7,9,11,13,15]
; AVX512DQ-NEXT: vpermt2q %zmm1, %zmm6, %zmm0
; AVX512DQ-NEXT: vpermt2q %zmm3, %zmm6, %zmm2
; AVX512DQ-NEXT: vmovdqa64 %zmm4, 64(%rsi)
; AVX512DQ-NEXT: vmovdqa64 %zmm5, (%rsi)
; AVX512DQ-NEXT: vmovdqa64 %zmm2, 64(%rdx)
; AVX512DQ-NEXT: vmovdqa64 %zmm0, (%rdx)
; AVX512DQ-NEXT: vzeroupper
; AVX512DQ-NEXT: retq
;
; AVX512DQ-FCP-LABEL: load_i64_stride2_vf16:
; AVX512DQ-FCP: # %bb.0:
; AVX512DQ-FCP-NEXT: vmovdqa64 (%rdi), %zmm0
; AVX512DQ-FCP-NEXT: vmovdqa64 64(%rdi), %zmm1
; AVX512DQ-FCP-NEXT: vmovdqa64 128(%rdi), %zmm2
; AVX512DQ-FCP-NEXT: vmovdqa64 192(%rdi), %zmm3
; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm4 = [0,2,4,6,8,10,12,14]
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm0, %zmm5
; AVX512DQ-FCP-NEXT: vpermt2q %zmm1, %zmm4, %zmm5
; AVX512DQ-FCP-NEXT: vpermi2q %zmm3, %zmm2, %zmm4
; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm6 = [1,3,5,7,9,11,13,15]
; AVX512DQ-FCP-NEXT: vpermt2q %zmm1, %zmm6, %zmm0
; AVX512DQ-FCP-NEXT: vpermt2q %zmm3, %zmm6, %zmm2
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm4, 64(%rsi)
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm5, (%rsi)
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm2, 64(%rdx)
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm0, (%rdx)
; AVX512DQ-FCP-NEXT: vzeroupper
; AVX512DQ-FCP-NEXT: retq
;
; AVX512BW-LABEL: load_i64_stride2_vf16:
; 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: vpmovsxbq {{.*#+}} zmm4 = [0,2,4,6,8,10,12,14]
; AVX512BW-NEXT: vmovdqa64 %zmm0, %zmm5
; AVX512BW-NEXT: vpermt2q %zmm1, %zmm4, %zmm5
; AVX512BW-NEXT: vpermi2q %zmm3, %zmm2, %zmm4
; AVX512BW-NEXT: vpmovsxbq {{.*#+}} zmm6 = [1,3,5,7,9,11,13,15]
; AVX512BW-NEXT: vpermt2q %zmm1, %zmm6, %zmm0
; AVX512BW-NEXT: vpermt2q %zmm3, %zmm6, %zmm2
; AVX512BW-NEXT: vmovdqa64 %zmm4, 64(%rsi)
; AVX512BW-NEXT: vmovdqa64 %zmm5, (%rsi)
; AVX512BW-NEXT: vmovdqa64 %zmm2, 64(%rdx)
; AVX512BW-NEXT: vmovdqa64 %zmm0, (%rdx)
; AVX512BW-NEXT: vzeroupper
; AVX512BW-NEXT: retq
;
; AVX512BW-FCP-LABEL: load_i64_stride2_vf16:
; 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: vpmovsxbq {{.*#+}} zmm4 = [0,2,4,6,8,10,12,14]
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm0, %zmm5
; AVX512BW-FCP-NEXT: vpermt2q %zmm1, %zmm4, %zmm5
; AVX512BW-FCP-NEXT: vpermi2q %zmm3, %zmm2, %zmm4
; AVX512BW-FCP-NEXT: vpmovsxbq {{.*#+}} zmm6 = [1,3,5,7,9,11,13,15]
; AVX512BW-FCP-NEXT: vpermt2q %zmm1, %zmm6, %zmm0
; AVX512BW-FCP-NEXT: vpermt2q %zmm3, %zmm6, %zmm2
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm4, 64(%rsi)
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm5, (%rsi)
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm2, 64(%rdx)
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm0, (%rdx)
; AVX512BW-FCP-NEXT: vzeroupper
; AVX512BW-FCP-NEXT: retq
;
; AVX512DQ-BW-LABEL: load_i64_stride2_vf16:
; 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: vpmovsxbq {{.*#+}} zmm4 = [0,2,4,6,8,10,12,14]
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm0, %zmm5
; AVX512DQ-BW-NEXT: vpermt2q %zmm1, %zmm4, %zmm5
; AVX512DQ-BW-NEXT: vpermi2q %zmm3, %zmm2, %zmm4
; AVX512DQ-BW-NEXT: vpmovsxbq {{.*#+}} zmm6 = [1,3,5,7,9,11,13,15]
; AVX512DQ-BW-NEXT: vpermt2q %zmm1, %zmm6, %zmm0
; AVX512DQ-BW-NEXT: vpermt2q %zmm3, %zmm6, %zmm2
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm4, 64(%rsi)
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm5, (%rsi)
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm2, 64(%rdx)
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm0, (%rdx)
; AVX512DQ-BW-NEXT: vzeroupper
; AVX512DQ-BW-NEXT: retq
;
; AVX512DQ-BW-FCP-LABEL: load_i64_stride2_vf16:
; 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: vpmovsxbq {{.*#+}} zmm4 = [0,2,4,6,8,10,12,14]
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm0, %zmm5
; AVX512DQ-BW-FCP-NEXT: vpermt2q %zmm1, %zmm4, %zmm5
; AVX512DQ-BW-FCP-NEXT: vpermi2q %zmm3, %zmm2, %zmm4
; AVX512DQ-BW-FCP-NEXT: vpmovsxbq {{.*#+}} zmm6 = [1,3,5,7,9,11,13,15]
; AVX512DQ-BW-FCP-NEXT: vpermt2q %zmm1, %zmm6, %zmm0
; AVX512DQ-BW-FCP-NEXT: vpermt2q %zmm3, %zmm6, %zmm2
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm4, 64(%rsi)
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm5, (%rsi)
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm2, 64(%rdx)
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm0, (%rdx)
; AVX512DQ-BW-FCP-NEXT: vzeroupper
; AVX512DQ-BW-FCP-NEXT: retq
%wide.vec = load <32 x i64>, ptr %in.vec, align 64
%strided.vec0 = shufflevector <32 x i64> %wide.vec, <32 x i64> poison, <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30>
%strided.vec1 = shufflevector <32 x i64> %wide.vec, <32 x i64> poison, <16 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31>
store <16 x i64> %strided.vec0, ptr %out.vec0, align 64
store <16 x i64> %strided.vec1, ptr %out.vec1, align 64
ret void
}
define void @load_i64_stride2_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1) nounwind {
;
;
;
; SSE-LABEL: load_i64_stride2_vf32:
; SSE: # %bb.0:
; SSE-NEXT: subq $152, %rsp
; SSE-NEXT: movaps 208(%rdi), %xmm11
; SSE-NEXT: movaps 192(%rdi), %xmm6
; SSE-NEXT: movaps 80(%rdi), %xmm1
; SSE-NEXT: movaps 64(%rdi), %xmm5
; SSE-NEXT: movaps 240(%rdi), %xmm14
; SSE-NEXT: movaps 224(%rdi), %xmm8
; SSE-NEXT: movaps 112(%rdi), %xmm3
; SSE-NEXT: movaps 96(%rdi), %xmm7
; SSE-NEXT: movaps 272(%rdi), %xmm12
; SSE-NEXT: movaps 144(%rdi), %xmm2
; SSE-NEXT: movaps 128(%rdi), %xmm9
; SSE-NEXT: movaps 304(%rdi), %xmm0
; SSE-NEXT: movaps 288(%rdi), %xmm13
; SSE-NEXT: movaps 176(%rdi), %xmm4
; SSE-NEXT: movaps 160(%rdi), %xmm10
; SSE-NEXT: movaps %xmm7, %xmm15
; SSE-NEXT: movlhps {{.*#+}} xmm15 = xmm15[0],xmm3[0]
; SSE-NEXT: movaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: unpckhpd {{.*#+}} xmm7 = xmm7[1],xmm3[1]
; SSE-NEXT: movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movaps %xmm5, %xmm3
; SSE-NEXT: movlhps {{.*#+}} xmm3 = xmm3[0],xmm1[0]
; SSE-NEXT: movaps %xmm3, (%rsp) # 16-byte Spill
; SSE-NEXT: unpckhpd {{.*#+}} xmm5 = xmm5[1],xmm1[1]
; SSE-NEXT: movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movaps %xmm10, %xmm1
; SSE-NEXT: movlhps {{.*#+}} xmm1 = xmm1[0],xmm4[0]
; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: unpckhpd {{.*#+}} xmm10 = xmm10[1],xmm4[1]
; SSE-NEXT: movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movaps %xmm9, %xmm1
; SSE-NEXT: movlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: unpckhpd {{.*#+}} xmm9 = xmm9[1],xmm2[1]
; SSE-NEXT: movaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movaps %xmm8, %xmm1
; SSE-NEXT: movlhps {{.*#+}} xmm1 = xmm1[0],xmm14[0]
; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: unpckhpd {{.*#+}} xmm8 = xmm8[1],xmm14[1]
; SSE-NEXT: movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movaps %xmm6, %xmm1
; SSE-NEXT: movlhps {{.*#+}} xmm1 = xmm1[0],xmm11[0]
; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: unpckhpd {{.*#+}} xmm6 = xmm6[1],xmm11[1]
; SSE-NEXT: movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movaps %xmm13, %xmm1
; SSE-NEXT: movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: unpckhpd {{.*#+}} xmm13 = xmm13[1],xmm0[1]
; SSE-NEXT: movaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movaps 256(%rdi), %xmm0
; SSE-NEXT: movaps %xmm0, %xmm1
; SSE-NEXT: movlhps {{.*#+}} xmm1 = xmm1[0],xmm12[0]
; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm12[1]
; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movaps 368(%rdi), %xmm0
; SSE-NEXT: movaps 352(%rdi), %xmm15
; SSE-NEXT: movaps %xmm15, %xmm1
; SSE-NEXT: movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: unpckhpd {{.*#+}} xmm15 = xmm15[1],xmm0[1]
; SSE-NEXT: movaps 336(%rdi), %xmm0
; SSE-NEXT: movaps 320(%rdi), %xmm13
; SSE-NEXT: movaps %xmm13, %xmm11
; SSE-NEXT: movlhps {{.*#+}} xmm11 = xmm11[0],xmm0[0]
; SSE-NEXT: unpckhpd {{.*#+}} xmm13 = xmm13[1],xmm0[1]
; SSE-NEXT: movaps 432(%rdi), %xmm0
; SSE-NEXT: movaps 416(%rdi), %xmm12
; SSE-NEXT: movaps %xmm12, %xmm14
; SSE-NEXT: movlhps {{.*#+}} xmm14 = xmm14[0],xmm0[0]
; SSE-NEXT: unpckhpd {{.*#+}} xmm12 = xmm12[1],xmm0[1]
; SSE-NEXT: movaps 400(%rdi), %xmm0
; SSE-NEXT: movaps 384(%rdi), %xmm9
; SSE-NEXT: movaps %xmm9, %xmm10
; SSE-NEXT: movlhps {{.*#+}} xmm10 = xmm10[0],xmm0[0]
; SSE-NEXT: unpckhpd {{.*#+}} xmm9 = xmm9[1],xmm0[1]
; SSE-NEXT: movaps 496(%rdi), %xmm0
; SSE-NEXT: movaps 480(%rdi), %xmm7
; SSE-NEXT: movaps %xmm7, %xmm6
; SSE-NEXT: movlhps {{.*#+}} xmm6 = xmm6[0],xmm0[0]
; SSE-NEXT: unpckhpd {{.*#+}} xmm7 = xmm7[1],xmm0[1]
; SSE-NEXT: movaps 464(%rdi), %xmm1
; SSE-NEXT: movaps 448(%rdi), %xmm3
; SSE-NEXT: movaps %xmm3, %xmm2
; SSE-NEXT: movlhps {{.*#+}} xmm2 = xmm2[0],xmm1[0]
; SSE-NEXT: unpckhpd {{.*#+}} xmm3 = xmm3[1],xmm1[1]
; SSE-NEXT: movaps 32(%rdi), %xmm8
; SSE-NEXT: movaps 48(%rdi), %xmm1
; SSE-NEXT: movaps %xmm8, %xmm5
; SSE-NEXT: movlhps {{.*#+}} xmm5 = xmm5[0],xmm1[0]
; SSE-NEXT: unpckhpd {{.*#+}} xmm8 = xmm8[1],xmm1[1]
; SSE-NEXT: movaps (%rdi), %xmm4
; SSE-NEXT: movaps 16(%rdi), %xmm0
; SSE-NEXT: movaps %xmm4, %xmm1
; SSE-NEXT: movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
; SSE-NEXT: unpckhpd {{.*#+}} xmm4 = xmm4[1],xmm0[1]
; SSE-NEXT: movaps %xmm2, 224(%rsi)
; SSE-NEXT: movaps %xmm11, 160(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 96(%rsi)
; SSE-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 32(%rsi)
; SSE-NEXT: movaps %xmm6, 240(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 176(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 112(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 48(%rsi)
; SSE-NEXT: movaps %xmm10, 192(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 128(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 64(%rsi)
; SSE-NEXT: movaps %xmm1, (%rsi)
; SSE-NEXT: movaps %xmm14, 208(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 144(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 80(%rsi)
; SSE-NEXT: movaps %xmm5, 16(%rsi)
; SSE-NEXT: movaps %xmm3, 224(%rdx)
; SSE-NEXT: movaps %xmm7, 240(%rdx)
; SSE-NEXT: movaps %xmm9, 192(%rdx)
; SSE-NEXT: movaps %xmm12, 208(%rdx)
; SSE-NEXT: movaps %xmm13, 160(%rdx)
; SSE-NEXT: movaps %xmm15, 176(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 128(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 144(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 96(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 112(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 64(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 80(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %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, 48(%rdx)
; SSE-NEXT: movaps %xmm4, (%rdx)
; SSE-NEXT: movaps %xmm8, 16(%rdx)
; SSE-NEXT: addq $152, %rsp
; SSE-NEXT: retq
;
; AVX-LABEL: load_i64_stride2_vf32:
; AVX: # %bb.0:
; AVX-NEXT: vmovaps 384(%rdi), %ymm4
; AVX-NEXT: vmovaps 256(%rdi), %ymm6
; AVX-NEXT: vmovaps 320(%rdi), %ymm5
; AVX-NEXT: vmovaps (%rdi), %ymm1
; AVX-NEXT: vmovaps 64(%rdi), %ymm2
; AVX-NEXT: vmovaps 128(%rdi), %ymm9
; AVX-NEXT: vmovaps 192(%rdi), %ymm3
; AVX-NEXT: vmovaps 448(%rdi), %ymm0
; AVX-NEXT: vperm2f128 {{.*#+}} ymm8 = ymm0[2,3],mem[2,3]
; AVX-NEXT: vinsertf128 $1, 480(%rdi), %ymm0, %ymm10
; AVX-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm10[0],ymm8[0],ymm10[2],ymm8[2]
; AVX-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT: vperm2f128 {{.*#+}} ymm7 = ymm2[2,3],mem[2,3]
; AVX-NEXT: vinsertf128 $1, 96(%rdi), %ymm2, %ymm11
; AVX-NEXT: vunpcklpd {{.*#+}} ymm2 = ymm11[0],ymm7[0],ymm11[2],ymm7[2]
; AVX-NEXT: vperm2f128 {{.*#+}} ymm12 = ymm3[2,3],mem[2,3]
; AVX-NEXT: vinsertf128 $1, 224(%rdi), %ymm3, %ymm13
; AVX-NEXT: vunpcklpd {{.*#+}} ymm3 = ymm13[0],ymm12[0],ymm13[2],ymm12[2]
; AVX-NEXT: vperm2f128 {{.*#+}} ymm14 = ymm5[2,3],mem[2,3]
; AVX-NEXT: vinsertf128 $1, 352(%rdi), %ymm5, %ymm15
; AVX-NEXT: vunpcklpd {{.*#+}} ymm5 = ymm15[0],ymm14[0],ymm15[2],ymm14[2]
; AVX-NEXT: vunpckhpd {{.*#+}} ymm7 = ymm11[1],ymm7[1],ymm11[3],ymm7[3]
; AVX-NEXT: vperm2f128 {{.*#+}} ymm11 = ymm9[2,3],mem[2,3]
; AVX-NEXT: vinsertf128 $1, 160(%rdi), %ymm9, %ymm9
; AVX-NEXT: vunpckhpd {{.*#+}} ymm12 = ymm13[1],ymm12[1],ymm13[3],ymm12[3]
; AVX-NEXT: vperm2f128 {{.*#+}} ymm13 = ymm6[2,3],mem[2,3]
; AVX-NEXT: vinsertf128 $1, 288(%rdi), %ymm6, %ymm6
; AVX-NEXT: vunpckhpd {{.*#+}} ymm14 = ymm15[1],ymm14[1],ymm15[3],ymm14[3]
; AVX-NEXT: vperm2f128 {{.*#+}} ymm15 = ymm4[2,3],mem[2,3]
; AVX-NEXT: vinsertf128 $1, 416(%rdi), %ymm4, %ymm4
; AVX-NEXT: vunpckhpd {{.*#+}} ymm8 = ymm10[1],ymm8[1],ymm10[3],ymm8[3]
; AVX-NEXT: vunpcklpd {{.*#+}} ymm10 = ymm4[0],ymm15[0],ymm4[2],ymm15[2]
; AVX-NEXT: vunpckhpd {{.*#+}} ymm4 = ymm4[1],ymm15[1],ymm4[3],ymm15[3]
; AVX-NEXT: vunpcklpd {{.*#+}} ymm15 = ymm6[0],ymm13[0],ymm6[2],ymm13[2]
; AVX-NEXT: vunpckhpd {{.*#+}} ymm6 = ymm6[1],ymm13[1],ymm6[3],ymm13[3]
; AVX-NEXT: vunpcklpd {{.*#+}} ymm13 = ymm9[0],ymm11[0],ymm9[2],ymm11[2]
; AVX-NEXT: vunpckhpd {{.*#+}} ymm9 = ymm9[1],ymm11[1],ymm9[3],ymm11[3]
; AVX-NEXT: vperm2f128 {{.*#+}} ymm11 = ymm1[2,3],mem[2,3]
; AVX-NEXT: vinsertf128 $1, 32(%rdi), %ymm1, %ymm1
; AVX-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm11[0],ymm1[2],ymm11[2]
; AVX-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm11[1],ymm1[3],ymm11[3]
; AVX-NEXT: vmovaps %ymm10, 192(%rsi)
; AVX-NEXT: vmovaps %ymm15, 128(%rsi)
; AVX-NEXT: vmovaps %ymm13, 64(%rsi)
; AVX-NEXT: vmovaps %ymm0, (%rsi)
; AVX-NEXT: vmovaps %ymm5, 160(%rsi)
; AVX-NEXT: vmovaps %ymm3, 96(%rsi)
; AVX-NEXT: vmovaps %ymm2, 32(%rsi)
; AVX-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT: vmovaps %ymm0, 224(%rsi)
; AVX-NEXT: vmovaps %ymm1, (%rdx)
; AVX-NEXT: vmovaps %ymm9, 64(%rdx)
; AVX-NEXT: vmovaps %ymm6, 128(%rdx)
; AVX-NEXT: vmovaps %ymm4, 192(%rdx)
; AVX-NEXT: vmovaps %ymm8, 224(%rdx)
; AVX-NEXT: vmovaps %ymm14, 160(%rdx)
; AVX-NEXT: vmovaps %ymm12, 96(%rdx)
; AVX-NEXT: vmovaps %ymm7, 32(%rdx)
; AVX-NEXT: vzeroupper
; AVX-NEXT: retq
;
; AVX2-LABEL: load_i64_stride2_vf32:
; AVX2: # %bb.0:
; AVX2-NEXT: vmovaps (%rdi), %ymm1
; AVX2-NEXT: vmovaps 64(%rdi), %ymm3
; AVX2-NEXT: vmovaps 96(%rdi), %ymm4
; AVX2-NEXT: vmovaps 416(%rdi), %ymm10
; AVX2-NEXT: vmovaps 384(%rdi), %ymm12
; AVX2-NEXT: vmovaps 288(%rdi), %ymm8
; AVX2-NEXT: vmovaps 256(%rdi), %ymm11
; AVX2-NEXT: vmovaps 160(%rdi), %ymm7
; AVX2-NEXT: vmovaps 128(%rdi), %ymm9
; AVX2-NEXT: vmovaps 480(%rdi), %ymm6
; AVX2-NEXT: vmovaps 448(%rdi), %ymm14
; AVX2-NEXT: vmovaps 352(%rdi), %ymm5
; AVX2-NEXT: vmovaps 320(%rdi), %ymm15
; AVX2-NEXT: vmovaps 224(%rdi), %ymm2
; AVX2-NEXT: vmovaps 192(%rdi), %ymm13
; AVX2-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm13[0],ymm2[0],ymm13[2],ymm2[2]
; AVX2-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT: vunpckhpd {{.*#+}} ymm2 = ymm13[1],ymm2[1],ymm13[3],ymm2[3]
; AVX2-NEXT: vunpcklpd {{.*#+}} ymm13 = ymm15[0],ymm5[0],ymm15[2],ymm5[2]
; AVX2-NEXT: vunpckhpd {{.*#+}} ymm5 = ymm15[1],ymm5[1],ymm15[3],ymm5[3]
; AVX2-NEXT: vunpcklpd {{.*#+}} ymm15 = ymm14[0],ymm6[0],ymm14[2],ymm6[2]
; AVX2-NEXT: vunpckhpd {{.*#+}} ymm6 = ymm14[1],ymm6[1],ymm14[3],ymm6[3]
; AVX2-NEXT: vunpcklpd {{.*#+}} ymm14 = ymm12[0],ymm10[0],ymm12[2],ymm10[2]
; AVX2-NEXT: vunpckhpd {{.*#+}} ymm10 = ymm12[1],ymm10[1],ymm12[3],ymm10[3]
; AVX2-NEXT: vunpcklpd {{.*#+}} ymm12 = ymm11[0],ymm8[0],ymm11[2],ymm8[2]
; AVX2-NEXT: vunpckhpd {{.*#+}} ymm8 = ymm11[1],ymm8[1],ymm11[3],ymm8[3]
; AVX2-NEXT: vunpcklpd {{.*#+}} ymm11 = ymm9[0],ymm7[0],ymm9[2],ymm7[2]
; AVX2-NEXT: vunpckhpd {{.*#+}} ymm7 = ymm9[1],ymm7[1],ymm9[3],ymm7[3]
; AVX2-NEXT: vunpcklpd {{.*#+}} ymm9 = ymm3[0],ymm4[0],ymm3[2],ymm4[2]
; AVX2-NEXT: vunpckhpd {{.*#+}} ymm3 = ymm3[1],ymm4[1],ymm3[3],ymm4[3]
; AVX2-NEXT: vmovaps 32(%rdi), %ymm4
; AVX2-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm4[0],ymm1[2],ymm4[2]
; AVX2-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm4[1],ymm1[3],ymm4[3]
; AVX2-NEXT: vpermpd {{.*#+}} ymm4 = ymm14[0,2,1,3]
; AVX2-NEXT: vmovaps %ymm4, 192(%rsi)
; AVX2-NEXT: vpermpd {{.*#+}} ymm4 = ymm12[0,2,1,3]
; AVX2-NEXT: vmovaps %ymm4, 128(%rsi)
; AVX2-NEXT: vpermpd {{.*#+}} ymm4 = ymm11[0,2,1,3]
; AVX2-NEXT: vmovaps %ymm4, 64(%rsi)
; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,2,1,3]
; AVX2-NEXT: vmovaps %ymm0, (%rsi)
; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm15[0,2,1,3]
; AVX2-NEXT: vmovaps %ymm0, 224(%rsi)
; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm13[0,2,1,3]
; AVX2-NEXT: vmovaps %ymm0, 160(%rsi)
; AVX2-NEXT: vpermpd $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX2-NEXT: # ymm0 = mem[0,2,1,3]
; AVX2-NEXT: vmovaps %ymm0, 96(%rsi)
; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm9[0,2,1,3]
; AVX2-NEXT: vmovaps %ymm0, 32(%rsi)
; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm1[0,2,1,3]
; AVX2-NEXT: vmovaps %ymm0, (%rdx)
; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm7[0,2,1,3]
; AVX2-NEXT: vmovaps %ymm0, 64(%rdx)
; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm8[0,2,1,3]
; AVX2-NEXT: vmovaps %ymm0, 128(%rdx)
; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm10[0,2,1,3]
; AVX2-NEXT: vmovaps %ymm0, 192(%rdx)
; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm6[0,2,1,3]
; AVX2-NEXT: vmovaps %ymm0, 224(%rdx)
; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm5[0,2,1,3]
; AVX2-NEXT: vmovaps %ymm0, 160(%rdx)
; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm2[0,2,1,3]
; AVX2-NEXT: vmovaps %ymm0, 96(%rdx)
; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm3[0,2,1,3]
; AVX2-NEXT: vmovaps %ymm0, 32(%rdx)
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
; AVX2-FP-LABEL: load_i64_stride2_vf32:
; AVX2-FP: # %bb.0:
; AVX2-FP-NEXT: vmovaps (%rdi), %ymm1
; AVX2-FP-NEXT: vmovaps 64(%rdi), %ymm3
; AVX2-FP-NEXT: vmovaps 96(%rdi), %ymm4
; AVX2-FP-NEXT: vmovaps 416(%rdi), %ymm10
; AVX2-FP-NEXT: vmovaps 384(%rdi), %ymm12
; AVX2-FP-NEXT: vmovaps 288(%rdi), %ymm8
; AVX2-FP-NEXT: vmovaps 256(%rdi), %ymm11
; AVX2-FP-NEXT: vmovaps 160(%rdi), %ymm7
; AVX2-FP-NEXT: vmovaps 128(%rdi), %ymm9
; AVX2-FP-NEXT: vmovaps 480(%rdi), %ymm6
; AVX2-FP-NEXT: vmovaps 448(%rdi), %ymm14
; AVX2-FP-NEXT: vmovaps 352(%rdi), %ymm5
; AVX2-FP-NEXT: vmovaps 320(%rdi), %ymm15
; AVX2-FP-NEXT: vmovaps 224(%rdi), %ymm2
; AVX2-FP-NEXT: vmovaps 192(%rdi), %ymm13
; AVX2-FP-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm13[0],ymm2[0],ymm13[2],ymm2[2]
; AVX2-FP-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT: vunpckhpd {{.*#+}} ymm2 = ymm13[1],ymm2[1],ymm13[3],ymm2[3]
; AVX2-FP-NEXT: vunpcklpd {{.*#+}} ymm13 = ymm15[0],ymm5[0],ymm15[2],ymm5[2]
; AVX2-FP-NEXT: vunpckhpd {{.*#+}} ymm5 = ymm15[1],ymm5[1],ymm15[3],ymm5[3]
; AVX2-FP-NEXT: vunpcklpd {{.*#+}} ymm15 = ymm14[0],ymm6[0],ymm14[2],ymm6[2]
; AVX2-FP-NEXT: vunpckhpd {{.*#+}} ymm6 = ymm14[1],ymm6[1],ymm14[3],ymm6[3]
; AVX2-FP-NEXT: vunpcklpd {{.*#+}} ymm14 = ymm12[0],ymm10[0],ymm12[2],ymm10[2]
; AVX2-FP-NEXT: vunpckhpd {{.*#+}} ymm10 = ymm12[1],ymm10[1],ymm12[3],ymm10[3]
; AVX2-FP-NEXT: vunpcklpd {{.*#+}} ymm12 = ymm11[0],ymm8[0],ymm11[2],ymm8[2]
; AVX2-FP-NEXT: vunpckhpd {{.*#+}} ymm8 = ymm11[1],ymm8[1],ymm11[3],ymm8[3]
; AVX2-FP-NEXT: vunpcklpd {{.*#+}} ymm11 = ymm9[0],ymm7[0],ymm9[2],ymm7[2]
; AVX2-FP-NEXT: vunpckhpd {{.*#+}} ymm7 = ymm9[1],ymm7[1],ymm9[3],ymm7[3]
; AVX2-FP-NEXT: vunpcklpd {{.*#+}} ymm9 = ymm3[0],ymm4[0],ymm3[2],ymm4[2]
; AVX2-FP-NEXT: vunpckhpd {{.*#+}} ymm3 = ymm3[1],ymm4[1],ymm3[3],ymm4[3]
; AVX2-FP-NEXT: vmovaps 32(%rdi), %ymm4
; AVX2-FP-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm4[0],ymm1[2],ymm4[2]
; AVX2-FP-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm4[1],ymm1[3],ymm4[3]
; AVX2-FP-NEXT: vpermpd {{.*#+}} ymm4 = ymm14[0,2,1,3]
; AVX2-FP-NEXT: vmovaps %ymm4, 192(%rsi)
; AVX2-FP-NEXT: vpermpd {{.*#+}} ymm4 = ymm12[0,2,1,3]
; AVX2-FP-NEXT: vmovaps %ymm4, 128(%rsi)
; AVX2-FP-NEXT: vpermpd {{.*#+}} ymm4 = ymm11[0,2,1,3]
; AVX2-FP-NEXT: vmovaps %ymm4, 64(%rsi)
; AVX2-FP-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,2,1,3]
; AVX2-FP-NEXT: vmovaps %ymm0, (%rsi)
; AVX2-FP-NEXT: vpermpd {{.*#+}} ymm0 = ymm15[0,2,1,3]
; AVX2-FP-NEXT: vmovaps %ymm0, 224(%rsi)
; AVX2-FP-NEXT: vpermpd {{.*#+}} ymm0 = ymm13[0,2,1,3]
; AVX2-FP-NEXT: vmovaps %ymm0, 160(%rsi)
; AVX2-FP-NEXT: vpermpd $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX2-FP-NEXT: # ymm0 = mem[0,2,1,3]
; AVX2-FP-NEXT: vmovaps %ymm0, 96(%rsi)
; AVX2-FP-NEXT: vpermpd {{.*#+}} ymm0 = ymm9[0,2,1,3]
; AVX2-FP-NEXT: vmovaps %ymm0, 32(%rsi)
; AVX2-FP-NEXT: vpermpd {{.*#+}} ymm0 = ymm1[0,2,1,3]
; AVX2-FP-NEXT: vmovaps %ymm0, (%rdx)
; AVX2-FP-NEXT: vpermpd {{.*#+}} ymm0 = ymm7[0,2,1,3]
; AVX2-FP-NEXT: vmovaps %ymm0, 64(%rdx)
; AVX2-FP-NEXT: vpermpd {{.*#+}} ymm0 = ymm8[0,2,1,3]
; AVX2-FP-NEXT: vmovaps %ymm0, 128(%rdx)
; AVX2-FP-NEXT: vpermpd {{.*#+}} ymm0 = ymm10[0,2,1,3]
; AVX2-FP-NEXT: vmovaps %ymm0, 192(%rdx)
; AVX2-FP-NEXT: vpermpd {{.*#+}} ymm0 = ymm6[0,2,1,3]
; AVX2-FP-NEXT: vmovaps %ymm0, 224(%rdx)
; AVX2-FP-NEXT: vpermpd {{.*#+}} ymm0 = ymm5[0,2,1,3]
; AVX2-FP-NEXT: vmovaps %ymm0, 160(%rdx)
; AVX2-FP-NEXT: vpermpd {{.*#+}} ymm0 = ymm2[0,2,1,3]
; AVX2-FP-NEXT: vmovaps %ymm0, 96(%rdx)
; AVX2-FP-NEXT: vpermpd {{.*#+}} ymm0 = ymm3[0,2,1,3]
; AVX2-FP-NEXT: vmovaps %ymm0, 32(%rdx)
; AVX2-FP-NEXT: vzeroupper
; AVX2-FP-NEXT: retq
;
; AVX2-FCP-LABEL: load_i64_stride2_vf32:
; AVX2-FCP: # %bb.0:
; AVX2-FCP-NEXT: vmovaps (%rdi), %ymm1
; AVX2-FCP-NEXT: vmovaps 64(%rdi), %ymm3
; AVX2-FCP-NEXT: vmovaps 96(%rdi), %ymm4
; AVX2-FCP-NEXT: vmovaps 416(%rdi), %ymm10
; AVX2-FCP-NEXT: vmovaps 384(%rdi), %ymm12
; AVX2-FCP-NEXT: vmovaps 288(%rdi), %ymm8
; AVX2-FCP-NEXT: vmovaps 256(%rdi), %ymm11
; AVX2-FCP-NEXT: vmovaps 160(%rdi), %ymm7
; AVX2-FCP-NEXT: vmovaps 128(%rdi), %ymm9
; AVX2-FCP-NEXT: vmovaps 480(%rdi), %ymm6
; AVX2-FCP-NEXT: vmovaps 448(%rdi), %ymm14
; AVX2-FCP-NEXT: vmovaps 352(%rdi), %ymm5
; AVX2-FCP-NEXT: vmovaps 320(%rdi), %ymm15
; AVX2-FCP-NEXT: vmovaps 224(%rdi), %ymm2
; AVX2-FCP-NEXT: vmovaps 192(%rdi), %ymm13
; AVX2-FCP-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm13[0],ymm2[0],ymm13[2],ymm2[2]
; AVX2-FCP-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT: vunpckhpd {{.*#+}} ymm2 = ymm13[1],ymm2[1],ymm13[3],ymm2[3]
; AVX2-FCP-NEXT: vunpcklpd {{.*#+}} ymm13 = ymm15[0],ymm5[0],ymm15[2],ymm5[2]
; AVX2-FCP-NEXT: vunpckhpd {{.*#+}} ymm5 = ymm15[1],ymm5[1],ymm15[3],ymm5[3]
; AVX2-FCP-NEXT: vunpcklpd {{.*#+}} ymm15 = ymm14[0],ymm6[0],ymm14[2],ymm6[2]
; AVX2-FCP-NEXT: vunpckhpd {{.*#+}} ymm6 = ymm14[1],ymm6[1],ymm14[3],ymm6[3]
; AVX2-FCP-NEXT: vunpcklpd {{.*#+}} ymm14 = ymm12[0],ymm10[0],ymm12[2],ymm10[2]
; AVX2-FCP-NEXT: vunpckhpd {{.*#+}} ymm10 = ymm12[1],ymm10[1],ymm12[3],ymm10[3]
; AVX2-FCP-NEXT: vunpcklpd {{.*#+}} ymm12 = ymm11[0],ymm8[0],ymm11[2],ymm8[2]
; AVX2-FCP-NEXT: vunpckhpd {{.*#+}} ymm8 = ymm11[1],ymm8[1],ymm11[3],ymm8[3]
; AVX2-FCP-NEXT: vunpcklpd {{.*#+}} ymm11 = ymm9[0],ymm7[0],ymm9[2],ymm7[2]
; AVX2-FCP-NEXT: vunpckhpd {{.*#+}} ymm7 = ymm9[1],ymm7[1],ymm9[3],ymm7[3]
; AVX2-FCP-NEXT: vunpcklpd {{.*#+}} ymm9 = ymm3[0],ymm4[0],ymm3[2],ymm4[2]
; AVX2-FCP-NEXT: vunpckhpd {{.*#+}} ymm3 = ymm3[1],ymm4[1],ymm3[3],ymm4[3]
; AVX2-FCP-NEXT: vmovaps 32(%rdi), %ymm4
; AVX2-FCP-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm4[0],ymm1[2],ymm4[2]
; AVX2-FCP-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm4[1],ymm1[3],ymm4[3]
; AVX2-FCP-NEXT: vpermpd {{.*#+}} ymm4 = ymm14[0,2,1,3]
; AVX2-FCP-NEXT: vmovaps %ymm4, 192(%rsi)
; AVX2-FCP-NEXT: vpermpd {{.*#+}} ymm4 = ymm12[0,2,1,3]
; AVX2-FCP-NEXT: vmovaps %ymm4, 128(%rsi)
; AVX2-FCP-NEXT: vpermpd {{.*#+}} ymm4 = ymm11[0,2,1,3]
; AVX2-FCP-NEXT: vmovaps %ymm4, 64(%rsi)
; AVX2-FCP-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,2,1,3]
; AVX2-FCP-NEXT: vmovaps %ymm0, (%rsi)
; AVX2-FCP-NEXT: vpermpd {{.*#+}} ymm0 = ymm15[0,2,1,3]
; AVX2-FCP-NEXT: vmovaps %ymm0, 224(%rsi)
; AVX2-FCP-NEXT: vpermpd {{.*#+}} ymm0 = ymm13[0,2,1,3]
; AVX2-FCP-NEXT: vmovaps %ymm0, 160(%rsi)
; AVX2-FCP-NEXT: vpermpd $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX2-FCP-NEXT: # ymm0 = mem[0,2,1,3]
; AVX2-FCP-NEXT: vmovaps %ymm0, 96(%rsi)
; AVX2-FCP-NEXT: vpermpd {{.*#+}} ymm0 = ymm9[0,2,1,3]
; AVX2-FCP-NEXT: vmovaps %ymm0, 32(%rsi)
; AVX2-FCP-NEXT: vpermpd {{.*#+}} ymm0 = ymm1[0,2,1,3]
; AVX2-FCP-NEXT: vmovaps %ymm0, (%rdx)
; AVX2-FCP-NEXT: vpermpd {{.*#+}} ymm0 = ymm7[0,2,1,3]
; AVX2-FCP-NEXT: vmovaps %ymm0, 64(%rdx)
; AVX2-FCP-NEXT: vpermpd {{.*#+}} ymm0 = ymm8[0,2,1,3]
; AVX2-FCP-NEXT: vmovaps %ymm0, 128(%rdx)
; AVX2-FCP-NEXT: vpermpd {{.*#+}} ymm0 = ymm10[0,2,1,3]
; AVX2-FCP-NEXT: vmovaps %ymm0, 192(%rdx)
; AVX2-FCP-NEXT: vpermpd {{.*#+}} ymm0 = ymm6[0,2,1,3]
; AVX2-FCP-NEXT: vmovaps %ymm0, 224(%rdx)
; AVX2-FCP-NEXT: vpermpd {{.*#+}} ymm0 = ymm5[0,2,1,3]
; AVX2-FCP-NEXT: vmovaps %ymm0, 160(%rdx)
; AVX2-FCP-NEXT: vpermpd {{.*#+}} ymm0 = ymm2[0,2,1,3]
; AVX2-FCP-NEXT: vmovaps %ymm0, 96(%rdx)
; AVX2-FCP-NEXT: vpermpd {{.*#+}} ymm0 = ymm3[0,2,1,3]
; AVX2-FCP-NEXT: vmovaps %ymm0, 32(%rdx)
; AVX2-FCP-NEXT: vzeroupper
; AVX2-FCP-NEXT: retq
;
; AVX512-LABEL: load_i64_stride2_vf32:
; AVX512: # %bb.0:
; AVX512-NEXT: vmovdqa64 (%rdi), %zmm0
; AVX512-NEXT: vmovdqa64 64(%rdi), %zmm1
; AVX512-NEXT: vmovdqa64 128(%rdi), %zmm2
; AVX512-NEXT: vmovdqa64 192(%rdi), %zmm3
; AVX512-NEXT: vmovdqa64 448(%rdi), %zmm4
; AVX512-NEXT: vmovdqa64 384(%rdi), %zmm5
; AVX512-NEXT: vmovdqa64 320(%rdi), %zmm6
; AVX512-NEXT: vmovdqa64 256(%rdi), %zmm7
; AVX512-NEXT: vpmovsxbq {{.*#+}} zmm8 = [0,2,4,6,8,10,12,14]
; AVX512-NEXT: vmovdqa64 %zmm7, %zmm9
; AVX512-NEXT: vpermt2q %zmm6, %zmm8, %zmm9
; AVX512-NEXT: vmovdqa64 %zmm5, %zmm10
; AVX512-NEXT: vpermt2q %zmm4, %zmm8, %zmm10
; AVX512-NEXT: vmovdqa64 %zmm2, %zmm11
; AVX512-NEXT: vpermt2q %zmm3, %zmm8, %zmm11
; AVX512-NEXT: vpermi2q %zmm1, %zmm0, %zmm8
; AVX512-NEXT: vpmovsxbq {{.*#+}} zmm12 = [1,3,5,7,9,11,13,15]
; AVX512-NEXT: vpermt2q %zmm4, %zmm12, %zmm5
; AVX512-NEXT: vpermt2q %zmm6, %zmm12, %zmm7
; AVX512-NEXT: vpermt2q %zmm3, %zmm12, %zmm2
; AVX512-NEXT: vpermt2q %zmm1, %zmm12, %zmm0
; AVX512-NEXT: vmovdqa64 %zmm10, 192(%rsi)
; AVX512-NEXT: vmovdqa64 %zmm8, (%rsi)
; AVX512-NEXT: vmovdqa64 %zmm11, 64(%rsi)
; AVX512-NEXT: vmovdqa64 %zmm9, 128(%rsi)
; AVX512-NEXT: vmovdqa64 %zmm7, 128(%rdx)
; AVX512-NEXT: vmovdqa64 %zmm5, 192(%rdx)
; AVX512-NEXT: vmovdqa64 %zmm0, (%rdx)
; AVX512-NEXT: vmovdqa64 %zmm2, 64(%rdx)
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
;
; AVX512-FCP-LABEL: load_i64_stride2_vf32:
; AVX512-FCP: # %bb.0:
; AVX512-FCP-NEXT: vmovdqa64 (%rdi), %zmm0
; AVX512-FCP-NEXT: vmovdqa64 64(%rdi), %zmm1
; AVX512-FCP-NEXT: vmovdqa64 128(%rdi), %zmm2
; AVX512-FCP-NEXT: vmovdqa64 192(%rdi), %zmm3
; AVX512-FCP-NEXT: vmovdqa64 448(%rdi), %zmm4
; AVX512-FCP-NEXT: vmovdqa64 384(%rdi), %zmm5
; AVX512-FCP-NEXT: vmovdqa64 320(%rdi), %zmm6
; AVX512-FCP-NEXT: vmovdqa64 256(%rdi), %zmm7
; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm8 = [0,2,4,6,8,10,12,14]
; AVX512-FCP-NEXT: vmovdqa64 %zmm7, %zmm9
; AVX512-FCP-NEXT: vpermt2q %zmm6, %zmm8, %zmm9
; AVX512-FCP-NEXT: vmovdqa64 %zmm5, %zmm10
; AVX512-FCP-NEXT: vpermt2q %zmm4, %zmm8, %zmm10
; AVX512-FCP-NEXT: vmovdqa64 %zmm2, %zmm11
; AVX512-FCP-NEXT: vpermt2q %zmm3, %zmm8, %zmm11
; AVX512-FCP-NEXT: vpermi2q %zmm1, %zmm0, %zmm8
; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm12 = [1,3,5,7,9,11,13,15]
; AVX512-FCP-NEXT: vpermt2q %zmm4, %zmm12, %zmm5
; AVX512-FCP-NEXT: vpermt2q %zmm6, %zmm12, %zmm7
; AVX512-FCP-NEXT: vpermt2q %zmm3, %zmm12, %zmm2
; AVX512-FCP-NEXT: vpermt2q %zmm1, %zmm12, %zmm0
; AVX512-FCP-NEXT: vmovdqa64 %zmm10, 192(%rsi)
; AVX512-FCP-NEXT: vmovdqa64 %zmm8, (%rsi)
; AVX512-FCP-NEXT: vmovdqa64 %zmm11, 64(%rsi)
; AVX512-FCP-NEXT: vmovdqa64 %zmm9, 128(%rsi)
; AVX512-FCP-NEXT: vmovdqa64 %zmm7, 128(%rdx)
; AVX512-FCP-NEXT: vmovdqa64 %zmm5, 192(%rdx)
; AVX512-FCP-NEXT: vmovdqa64 %zmm0, (%rdx)
; AVX512-FCP-NEXT: vmovdqa64 %zmm2, 64(%rdx)
; AVX512-FCP-NEXT: vzeroupper
; AVX512-FCP-NEXT: retq
;
; AVX512DQ-LABEL: load_i64_stride2_vf32:
; AVX512DQ: # %bb.0:
; AVX512DQ-NEXT: vmovdqa64 (%rdi), %zmm0
; AVX512DQ-NEXT: vmovdqa64 64(%rdi), %zmm1
; AVX512DQ-NEXT: vmovdqa64 128(%rdi), %zmm2
; AVX512DQ-NEXT: vmovdqa64 192(%rdi), %zmm3
; AVX512DQ-NEXT: vmovdqa64 448(%rdi), %zmm4
; AVX512DQ-NEXT: vmovdqa64 384(%rdi), %zmm5
; AVX512DQ-NEXT: vmovdqa64 320(%rdi), %zmm6
; AVX512DQ-NEXT: vmovdqa64 256(%rdi), %zmm7
; AVX512DQ-NEXT: vpmovsxbq {{.*#+}} zmm8 = [0,2,4,6,8,10,12,14]
; AVX512DQ-NEXT: vmovdqa64 %zmm7, %zmm9
; AVX512DQ-NEXT: vpermt2q %zmm6, %zmm8, %zmm9
; AVX512DQ-NEXT: vmovdqa64 %zmm5, %zmm10
; AVX512DQ-NEXT: vpermt2q %zmm4, %zmm8, %zmm10
; AVX512DQ-NEXT: vmovdqa64 %zmm2, %zmm11
; AVX512DQ-NEXT: vpermt2q %zmm3, %zmm8, %zmm11
; AVX512DQ-NEXT: vpermi2q %zmm1, %zmm0, %zmm8
; AVX512DQ-NEXT: vpmovsxbq {{.*#+}} zmm12 = [1,3,5,7,9,11,13,15]
; AVX512DQ-NEXT: vpermt2q %zmm4, %zmm12, %zmm5
; AVX512DQ-NEXT: vpermt2q %zmm6, %zmm12, %zmm7
; AVX512DQ-NEXT: vpermt2q %zmm3, %zmm12, %zmm2
; AVX512DQ-NEXT: vpermt2q %zmm1, %zmm12, %zmm0
; AVX512DQ-NEXT: vmovdqa64 %zmm10, 192(%rsi)
; AVX512DQ-NEXT: vmovdqa64 %zmm8, (%rsi)
; AVX512DQ-NEXT: vmovdqa64 %zmm11, 64(%rsi)
; AVX512DQ-NEXT: vmovdqa64 %zmm9, 128(%rsi)
; AVX512DQ-NEXT: vmovdqa64 %zmm7, 128(%rdx)
; AVX512DQ-NEXT: vmovdqa64 %zmm5, 192(%rdx)
; AVX512DQ-NEXT: vmovdqa64 %zmm0, (%rdx)
; AVX512DQ-NEXT: vmovdqa64 %zmm2, 64(%rdx)
; AVX512DQ-NEXT: vzeroupper
; AVX512DQ-NEXT: retq
;
; AVX512DQ-FCP-LABEL: load_i64_stride2_vf32:
; AVX512DQ-FCP: # %bb.0:
; AVX512DQ-FCP-NEXT: vmovdqa64 (%rdi), %zmm0
; AVX512DQ-FCP-NEXT: vmovdqa64 64(%rdi), %zmm1
; AVX512DQ-FCP-NEXT: vmovdqa64 128(%rdi), %zmm2
; AVX512DQ-FCP-NEXT: vmovdqa64 192(%rdi), %zmm3
; AVX512DQ-FCP-NEXT: vmovdqa64 448(%rdi), %zmm4
; AVX512DQ-FCP-NEXT: vmovdqa64 384(%rdi), %zmm5
; AVX512DQ-FCP-NEXT: vmovdqa64 320(%rdi), %zmm6
; AVX512DQ-FCP-NEXT: vmovdqa64 256(%rdi), %zmm7
; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm8 = [0,2,4,6,8,10,12,14]
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm7, %zmm9
; AVX512DQ-FCP-NEXT: vpermt2q %zmm6, %zmm8, %zmm9
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm5, %zmm10
; AVX512DQ-FCP-NEXT: vpermt2q %zmm4, %zmm8, %zmm10
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm2, %zmm11
; AVX512DQ-FCP-NEXT: vpermt2q %zmm3, %zmm8, %zmm11
; AVX512DQ-FCP-NEXT: vpermi2q %zmm1, %zmm0, %zmm8
; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm12 = [1,3,5,7,9,11,13,15]
; AVX512DQ-FCP-NEXT: vpermt2q %zmm4, %zmm12, %zmm5
; AVX512DQ-FCP-NEXT: vpermt2q %zmm6, %zmm12, %zmm7
; AVX512DQ-FCP-NEXT: vpermt2q %zmm3, %zmm12, %zmm2
; AVX512DQ-FCP-NEXT: vpermt2q %zmm1, %zmm12, %zmm0
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm10, 192(%rsi)
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm8, (%rsi)
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm11, 64(%rsi)
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm9, 128(%rsi)
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm7, 128(%rdx)
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm5, 192(%rdx)
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm0, (%rdx)
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm2, 64(%rdx)
; AVX512DQ-FCP-NEXT: vzeroupper
; AVX512DQ-FCP-NEXT: retq
;
; AVX512BW-LABEL: load_i64_stride2_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: vmovdqa64 448(%rdi), %zmm4
; AVX512BW-NEXT: vmovdqa64 384(%rdi), %zmm5
; AVX512BW-NEXT: vmovdqa64 320(%rdi), %zmm6
; AVX512BW-NEXT: vmovdqa64 256(%rdi), %zmm7
; AVX512BW-NEXT: vpmovsxbq {{.*#+}} zmm8 = [0,2,4,6,8,10,12,14]
; AVX512BW-NEXT: vmovdqa64 %zmm7, %zmm9
; AVX512BW-NEXT: vpermt2q %zmm6, %zmm8, %zmm9
; AVX512BW-NEXT: vmovdqa64 %zmm5, %zmm10
; AVX512BW-NEXT: vpermt2q %zmm4, %zmm8, %zmm10
; AVX512BW-NEXT: vmovdqa64 %zmm2, %zmm11
; AVX512BW-NEXT: vpermt2q %zmm3, %zmm8, %zmm11
; AVX512BW-NEXT: vpermi2q %zmm1, %zmm0, %zmm8
; AVX512BW-NEXT: vpmovsxbq {{.*#+}} zmm12 = [1,3,5,7,9,11,13,15]
; AVX512BW-NEXT: vpermt2q %zmm4, %zmm12, %zmm5
; AVX512BW-NEXT: vpermt2q %zmm6, %zmm12, %zmm7
; AVX512BW-NEXT: vpermt2q %zmm3, %zmm12, %zmm2
; AVX512BW-NEXT: vpermt2q %zmm1, %zmm12, %zmm0
; AVX512BW-NEXT: vmovdqa64 %zmm10, 192(%rsi)
; AVX512BW-NEXT: vmovdqa64 %zmm8, (%rsi)
; AVX512BW-NEXT: vmovdqa64 %zmm11, 64(%rsi)
; AVX512BW-NEXT: vmovdqa64 %zmm9, 128(%rsi)
; AVX512BW-NEXT: vmovdqa64 %zmm7, 128(%rdx)
; AVX512BW-NEXT: vmovdqa64 %zmm5, 192(%rdx)
; AVX512BW-NEXT: vmovdqa64 %zmm0, (%rdx)
; AVX512BW-NEXT: vmovdqa64 %zmm2, 64(%rdx)
; AVX512BW-NEXT: vzeroupper
; AVX512BW-NEXT: retq
;
; AVX512BW-FCP-LABEL: load_i64_stride2_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: vmovdqa64 448(%rdi), %zmm4
; AVX512BW-FCP-NEXT: vmovdqa64 384(%rdi), %zmm5
; AVX512BW-FCP-NEXT: vmovdqa64 320(%rdi), %zmm6
; AVX512BW-FCP-NEXT: vmovdqa64 256(%rdi), %zmm7
; AVX512BW-FCP-NEXT: vpmovsxbq {{.*#+}} zmm8 = [0,2,4,6,8,10,12,14]
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm7, %zmm9
; AVX512BW-FCP-NEXT: vpermt2q %zmm6, %zmm8, %zmm9
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm5, %zmm10
; AVX512BW-FCP-NEXT: vpermt2q %zmm4, %zmm8, %zmm10
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm2, %zmm11
; AVX512BW-FCP-NEXT: vpermt2q %zmm3, %zmm8, %zmm11
; AVX512BW-FCP-NEXT: vpermi2q %zmm1, %zmm0, %zmm8
; AVX512BW-FCP-NEXT: vpmovsxbq {{.*#+}} zmm12 = [1,3,5,7,9,11,13,15]
; AVX512BW-FCP-NEXT: vpermt2q %zmm4, %zmm12, %zmm5
; AVX512BW-FCP-NEXT: vpermt2q %zmm6, %zmm12, %zmm7
; AVX512BW-FCP-NEXT: vpermt2q %zmm3, %zmm12, %zmm2
; AVX512BW-FCP-NEXT: vpermt2q %zmm1, %zmm12, %zmm0
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm10, 192(%rsi)
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm8, (%rsi)
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm11, 64(%rsi)
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm9, 128(%rsi)
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm7, 128(%rdx)
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm5, 192(%rdx)
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm0, (%rdx)
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm2, 64(%rdx)
; AVX512BW-FCP-NEXT: vzeroupper
; AVX512BW-FCP-NEXT: retq
;
; AVX512DQ-BW-LABEL: load_i64_stride2_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: vmovdqa64 448(%rdi), %zmm4
; AVX512DQ-BW-NEXT: vmovdqa64 384(%rdi), %zmm5
; AVX512DQ-BW-NEXT: vmovdqa64 320(%rdi), %zmm6
; AVX512DQ-BW-NEXT: vmovdqa64 256(%rdi), %zmm7
; AVX512DQ-BW-NEXT: vpmovsxbq {{.*#+}} zmm8 = [0,2,4,6,8,10,12,14]
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm7, %zmm9
; AVX512DQ-BW-NEXT: vpermt2q %zmm6, %zmm8, %zmm9
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm5, %zmm10
; AVX512DQ-BW-NEXT: vpermt2q %zmm4, %zmm8, %zmm10
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm2, %zmm11
; AVX512DQ-BW-NEXT: vpermt2q %zmm3, %zmm8, %zmm11
; AVX512DQ-BW-NEXT: vpermi2q %zmm1, %zmm0, %zmm8
; AVX512DQ-BW-NEXT: vpmovsxbq {{.*#+}} zmm12 = [1,3,5,7,9,11,13,15]
; AVX512DQ-BW-NEXT: vpermt2q %zmm4, %zmm12, %zmm5
; AVX512DQ-BW-NEXT: vpermt2q %zmm6, %zmm12, %zmm7
; AVX512DQ-BW-NEXT: vpermt2q %zmm3, %zmm12, %zmm2
; AVX512DQ-BW-NEXT: vpermt2q %zmm1, %zmm12, %zmm0
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm10, 192(%rsi)
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm8, (%rsi)
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm11, 64(%rsi)
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm9, 128(%rsi)
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm7, 128(%rdx)
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm5, 192(%rdx)
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm0, (%rdx)
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm2, 64(%rdx)
; AVX512DQ-BW-NEXT: vzeroupper
; AVX512DQ-BW-NEXT: retq
;
; AVX512DQ-BW-FCP-LABEL: load_i64_stride2_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: vmovdqa64 448(%rdi), %zmm4
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 384(%rdi), %zmm5
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 320(%rdi), %zmm6
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 256(%rdi), %zmm7
; AVX512DQ-BW-FCP-NEXT: vpmovsxbq {{.*#+}} zmm8 = [0,2,4,6,8,10,12,14]
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm7, %zmm9
; AVX512DQ-BW-FCP-NEXT: vpermt2q %zmm6, %zmm8, %zmm9
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm5, %zmm10
; AVX512DQ-BW-FCP-NEXT: vpermt2q %zmm4, %zmm8, %zmm10
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm2, %zmm11
; AVX512DQ-BW-FCP-NEXT: vpermt2q %zmm3, %zmm8, %zmm11
; AVX512DQ-BW-FCP-NEXT: vpermi2q %zmm1, %zmm0, %zmm8
; AVX512DQ-BW-FCP-NEXT: vpmovsxbq {{.*#+}} zmm12 = [1,3,5,7,9,11,13,15]
; AVX512DQ-BW-FCP-NEXT: vpermt2q %zmm4, %zmm12, %zmm5
; AVX512DQ-BW-FCP-NEXT: vpermt2q %zmm6, %zmm12, %zmm7
; AVX512DQ-BW-FCP-NEXT: vpermt2q %zmm3, %zmm12, %zmm2
; AVX512DQ-BW-FCP-NEXT: vpermt2q %zmm1, %zmm12, %zmm0
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm10, 192(%rsi)
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm8, (%rsi)
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm11, 64(%rsi)
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm9, 128(%rsi)
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm7, 128(%rdx)
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm5, 192(%rdx)
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm0, (%rdx)
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm2, 64(%rdx)
; AVX512DQ-BW-FCP-NEXT: vzeroupper
; AVX512DQ-BW-FCP-NEXT: retq
%wide.vec = load <64 x i64>, ptr %in.vec, align 64
%strided.vec0 = shufflevector <64 x i64> %wide.vec, <64 x i64> poison, <32 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30, i32 32, i32 34, i32 36, i32 38, i32 40, i32 42, i32 44, i32 46, i32 48, i32 50, i32 52, i32 54, i32 56, i32 58, i32 60, i32 62>
%strided.vec1 = shufflevector <64 x i64> %wide.vec, <64 x i64> poison, <32 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31, i32 33, i32 35, i32 37, i32 39, i32 41, i32 43, i32 45, i32 47, i32 49, i32 51, i32 53, i32 55, i32 57, i32 59, i32 61, i32 63>
store <32 x i64> %strided.vec0, ptr %out.vec0, align 64
store <32 x i64> %strided.vec1, ptr %out.vec1, align 64
ret void
}
define void @load_i64_stride2_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1) nounwind {
;
;
;
; SSE-LABEL: load_i64_stride2_vf64:
; SSE: # %bb.0:
; SSE-NEXT: subq $664, %rsp # imm = 0x298
; SSE-NEXT: movaps 304(%rdi), %xmm0
; SSE-NEXT: movaps 272(%rdi), %xmm1
; SSE-NEXT: movaps 256(%rdi), %xmm8
; SSE-NEXT: movaps 240(%rdi), %xmm2
; SSE-NEXT: movaps 224(%rdi), %xmm9
; SSE-NEXT: movaps 208(%rdi), %xmm3
; SSE-NEXT: movaps 192(%rdi), %xmm10
; SSE-NEXT: movaps 176(%rdi), %xmm4
; SSE-NEXT: movaps 160(%rdi), %xmm11
; SSE-NEXT: movaps 144(%rdi), %xmm5
; SSE-NEXT: movaps 128(%rdi), %xmm12
; SSE-NEXT: movaps 112(%rdi), %xmm6
; SSE-NEXT: movaps 96(%rdi), %xmm13
; SSE-NEXT: movaps 80(%rdi), %xmm7
; SSE-NEXT: movaps 64(%rdi), %xmm14
; SSE-NEXT: movaps %xmm14, %xmm15
; SSE-NEXT: movlhps {{.*#+}} xmm15 = xmm15[0],xmm7[0]
; SSE-NEXT: movaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: unpckhpd {{.*#+}} xmm14 = xmm14[1],xmm7[1]
; SSE-NEXT: movaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movaps %xmm13, %xmm7
; SSE-NEXT: movlhps {{.*#+}} xmm7 = xmm7[0],xmm6[0]
; SSE-NEXT: movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: unpckhpd {{.*#+}} xmm13 = xmm13[1],xmm6[1]
; SSE-NEXT: movaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movaps %xmm12, %xmm6
; SSE-NEXT: movlhps {{.*#+}} xmm6 = xmm6[0],xmm5[0]
; SSE-NEXT: movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: unpckhpd {{.*#+}} xmm12 = xmm12[1],xmm5[1]
; SSE-NEXT: movaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movaps %xmm11, %xmm5
; SSE-NEXT: movlhps {{.*#+}} xmm5 = xmm5[0],xmm4[0]
; SSE-NEXT: movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: unpckhpd {{.*#+}} xmm11 = xmm11[1],xmm4[1]
; SSE-NEXT: movaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movaps %xmm10, %xmm4
; SSE-NEXT: movlhps {{.*#+}} xmm4 = xmm4[0],xmm3[0]
; SSE-NEXT: movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: unpckhpd {{.*#+}} xmm10 = xmm10[1],xmm3[1]
; SSE-NEXT: movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movaps %xmm9, %xmm3
; SSE-NEXT: movlhps {{.*#+}} xmm3 = xmm3[0],xmm2[0]
; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: unpckhpd {{.*#+}} xmm9 = xmm9[1],xmm2[1]
; SSE-NEXT: movaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movaps %xmm8, %xmm2
; SSE-NEXT: movlhps {{.*#+}} xmm2 = xmm2[0],xmm1[0]
; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: unpckhpd {{.*#+}} xmm8 = xmm8[1],xmm1[1]
; SSE-NEXT: movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movaps 288(%rdi), %xmm1
; SSE-NEXT: movaps %xmm1, %xmm2
; SSE-NEXT: movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movaps 336(%rdi), %xmm0
; SSE-NEXT: movaps 320(%rdi), %xmm1
; SSE-NEXT: movaps %xmm1, %xmm2
; SSE-NEXT: movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movaps 368(%rdi), %xmm0
; SSE-NEXT: movaps 352(%rdi), %xmm1
; SSE-NEXT: movaps %xmm1, %xmm2
; SSE-NEXT: movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movaps 400(%rdi), %xmm0
; SSE-NEXT: movaps 384(%rdi), %xmm1
; SSE-NEXT: movaps %xmm1, %xmm2
; SSE-NEXT: movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movaps 432(%rdi), %xmm0
; SSE-NEXT: movaps 416(%rdi), %xmm1
; SSE-NEXT: movaps %xmm1, %xmm2
; SSE-NEXT: movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movaps 464(%rdi), %xmm0
; SSE-NEXT: movaps 448(%rdi), %xmm1
; SSE-NEXT: movaps %xmm1, %xmm2
; SSE-NEXT: movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movaps 496(%rdi), %xmm0
; SSE-NEXT: movaps 480(%rdi), %xmm1
; SSE-NEXT: movaps %xmm1, %xmm2
; SSE-NEXT: movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movaps 528(%rdi), %xmm0
; SSE-NEXT: movaps 512(%rdi), %xmm1
; SSE-NEXT: movaps %xmm1, %xmm2
; SSE-NEXT: movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movaps 560(%rdi), %xmm0
; SSE-NEXT: movaps 544(%rdi), %xmm1
; SSE-NEXT: movaps %xmm1, %xmm2
; SSE-NEXT: movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movaps 592(%rdi), %xmm0
; SSE-NEXT: movaps 576(%rdi), %xmm1
; SSE-NEXT: movaps %xmm1, %xmm2
; SSE-NEXT: movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movaps 624(%rdi), %xmm0
; SSE-NEXT: movaps 608(%rdi), %xmm1
; SSE-NEXT: movaps %xmm1, %xmm2
; SSE-NEXT: movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movaps 656(%rdi), %xmm0
; SSE-NEXT: movaps 640(%rdi), %xmm1
; SSE-NEXT: movaps %xmm1, %xmm2
; SSE-NEXT: movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movaps 688(%rdi), %xmm0
; SSE-NEXT: movaps 672(%rdi), %xmm1
; SSE-NEXT: movaps %xmm1, %xmm2
; SSE-NEXT: movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movaps 720(%rdi), %xmm0
; SSE-NEXT: movaps 704(%rdi), %xmm1
; SSE-NEXT: movaps %xmm1, %xmm2
; SSE-NEXT: movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movaps 752(%rdi), %xmm0
; SSE-NEXT: movaps 736(%rdi), %xmm1
; SSE-NEXT: movaps %xmm1, %xmm2
; SSE-NEXT: movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movaps 784(%rdi), %xmm0
; SSE-NEXT: movaps 768(%rdi), %xmm1
; SSE-NEXT: movaps %xmm1, %xmm2
; SSE-NEXT: movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT: movaps %xmm1, (%rsp) # 16-byte Spill
; SSE-NEXT: movaps 816(%rdi), %xmm0
; SSE-NEXT: movaps 800(%rdi), %xmm1
; SSE-NEXT: movaps %xmm1, %xmm2
; SSE-NEXT: movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movaps 848(%rdi), %xmm0
; SSE-NEXT: movaps 832(%rdi), %xmm1
; SSE-NEXT: movaps %xmm1, %xmm12
; SSE-NEXT: movlhps {{.*#+}} xmm12 = xmm12[0],xmm0[0]
; SSE-NEXT: unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movaps 880(%rdi), %xmm0
; SSE-NEXT: movaps 864(%rdi), %xmm15
; SSE-NEXT: movaps %xmm15, %xmm7
; SSE-NEXT: movlhps {{.*#+}} xmm7 = xmm7[0],xmm0[0]
; SSE-NEXT: unpckhpd {{.*#+}} xmm15 = xmm15[1],xmm0[1]
; SSE-NEXT: movaps 912(%rdi), %xmm0
; SSE-NEXT: movaps 896(%rdi), %xmm13
; SSE-NEXT: movaps %xmm13, %xmm9
; SSE-NEXT: movlhps {{.*#+}} xmm9 = xmm9[0],xmm0[0]
; SSE-NEXT: unpckhpd {{.*#+}} xmm13 = xmm13[1],xmm0[1]
; SSE-NEXT: movaps 944(%rdi), %xmm0
; SSE-NEXT: movaps 928(%rdi), %xmm8
; SSE-NEXT: movaps %xmm8, %xmm2
; SSE-NEXT: movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT: unpckhpd {{.*#+}} xmm8 = xmm8[1],xmm0[1]
; SSE-NEXT: movaps 976(%rdi), %xmm0
; SSE-NEXT: movaps 960(%rdi), %xmm6
; SSE-NEXT: movaps %xmm6, %xmm1
; SSE-NEXT: movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
; SSE-NEXT: unpckhpd {{.*#+}} xmm6 = xmm6[1],xmm0[1]
; SSE-NEXT: movaps 1008(%rdi), %xmm3
; SSE-NEXT: movaps 992(%rdi), %xmm5
; SSE-NEXT: movaps %xmm5, %xmm4
; SSE-NEXT: movlhps {{.*#+}} xmm4 = xmm4[0],xmm3[0]
; SSE-NEXT: unpckhpd {{.*#+}} xmm5 = xmm5[1],xmm3[1]
; SSE-NEXT: movaps (%rdi), %xmm14
; SSE-NEXT: movaps 16(%rdi), %xmm3
; SSE-NEXT: movaps %xmm14, %xmm10
; SSE-NEXT: movlhps {{.*#+}} xmm10 = xmm10[0],xmm3[0]
; SSE-NEXT: unpckhpd {{.*#+}} xmm14 = xmm14[1],xmm3[1]
; SSE-NEXT: movaps 32(%rdi), %xmm11
; SSE-NEXT: movaps 48(%rdi), %xmm0
; SSE-NEXT: movaps %xmm11, %xmm3
; SSE-NEXT: movlhps {{.*#+}} xmm3 = xmm3[0],xmm0[0]
; SSE-NEXT: unpckhpd {{.*#+}} xmm11 = xmm11[1],xmm0[1]
; SSE-NEXT: movaps %xmm4, 496(%rsi)
; SSE-NEXT: movaps %xmm1, 480(%rsi)
; SSE-NEXT: movaps %xmm2, 464(%rsi)
; SSE-NEXT: movaps %xmm9, 448(%rsi)
; SSE-NEXT: movaps %xmm7, 432(%rsi)
; SSE-NEXT: movaps %xmm12, 416(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 400(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 384(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 368(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 352(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 336(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 320(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 304(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 288(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 272(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 256(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 240(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 224(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 208(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 192(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 176(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 160(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 144(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 128(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 112(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 96(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 80(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 64(%rsi)
; 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, 32(%rsi)
; SSE-NEXT: movaps %xmm3, 16(%rsi)
; SSE-NEXT: movaps %xmm10, (%rsi)
; SSE-NEXT: movaps %xmm5, 496(%rdx)
; SSE-NEXT: movaps %xmm6, 480(%rdx)
; SSE-NEXT: movaps %xmm8, 464(%rdx)
; SSE-NEXT: movaps %xmm13, 448(%rdx)
; SSE-NEXT: movaps %xmm15, 432(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 416(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 400(%rdx)
; SSE-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 384(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 368(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 352(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 336(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 320(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 304(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 288(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 272(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 256(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 240(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 224(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 208(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 192(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 176(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 160(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 144(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 128(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 112(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 96(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 80(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 64(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 48(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 32(%rdx)
; SSE-NEXT: movaps %xmm11, 16(%rdx)
; SSE-NEXT: movaps %xmm14, (%rdx)
; SSE-NEXT: addq $664, %rsp # imm = 0x298
; SSE-NEXT: retq
;
; AVX-LABEL: load_i64_stride2_vf64:
; AVX: # %bb.0:
; AVX-NEXT: subq $424, %rsp # imm = 0x1A8
; AVX-NEXT: vmovaps 384(%rdi), %ymm7
; AVX-NEXT: vmovaps 256(%rdi), %ymm8
; AVX-NEXT: vmovaps 448(%rdi), %ymm2
; AVX-NEXT: vmovaps 320(%rdi), %ymm4
; AVX-NEXT: vmovaps (%rdi), %ymm11
; AVX-NEXT: vmovaps 64(%rdi), %ymm0
; AVX-NEXT: vmovaps 128(%rdi), %ymm5
; AVX-NEXT: vmovaps 192(%rdi), %ymm9
; AVX-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm0[2,3],mem[2,3]
; AVX-NEXT: vinsertf128 $1, 96(%rdi), %ymm0, %ymm1
; AVX-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm3[0],ymm1[2],ymm3[2]
; AVX-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT: vperm2f128 {{.*#+}} ymm6 = ymm9[2,3],mem[2,3]
; AVX-NEXT: vinsertf128 $1, 224(%rdi), %ymm9, %ymm12
; AVX-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm12[0],ymm6[0],ymm12[2],ymm6[2]
; AVX-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm4[2,3],mem[2,3]
; AVX-NEXT: vinsertf128 $1, 352(%rdi), %ymm4, %ymm9
; AVX-NEXT: vunpcklpd {{.*#+}} ymm4 = ymm9[0],ymm0[0],ymm9[2],ymm0[2]
; AVX-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT: vperm2f128 {{.*#+}} ymm10 = ymm2[2,3],mem[2,3]
; AVX-NEXT: vinsertf128 $1, 480(%rdi), %ymm2, %ymm13
; AVX-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm11[2,3],mem[2,3]
; AVX-NEXT: vinsertf128 $1, 32(%rdi), %ymm11, %ymm4
; AVX-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm3[1],ymm1[3],ymm3[3]
; AVX-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm4[0],ymm2[0],ymm4[2],ymm2[2]
; AVX-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm4[1],ymm2[1],ymm4[3],ymm2[3]
; AVX-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm5[2,3],mem[2,3]
; AVX-NEXT: vinsertf128 $1, 160(%rdi), %ymm5, %ymm2
; AVX-NEXT: vunpckhpd {{.*#+}} ymm3 = ymm12[1],ymm6[1],ymm12[3],ymm6[3]
; AVX-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT: vunpcklpd {{.*#+}} ymm3 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
; AVX-NEXT: vmovups %ymm3, (%rsp) # 32-byte Spill
; AVX-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm2[1],ymm1[1],ymm2[3],ymm1[3]
; AVX-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm8[2,3],mem[2,3]
; AVX-NEXT: vinsertf128 $1, 288(%rdi), %ymm8, %ymm2
; AVX-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm9[1],ymm0[1],ymm9[3],ymm0[3]
; AVX-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
; AVX-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm2[1],ymm1[1],ymm2[3],ymm1[3]
; AVX-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm13[0],ymm10[0],ymm13[2],ymm10[2]
; AVX-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm13[1],ymm10[1],ymm13[3],ymm10[3]
; AVX-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm7[2,3],mem[2,3]
; AVX-NEXT: vinsertf128 $1, 416(%rdi), %ymm7, %ymm1
; AVX-NEXT: vunpcklpd {{.*#+}} ymm13 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
; AVX-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
; AVX-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT: vmovaps 576(%rdi), %ymm0
; AVX-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3],mem[2,3]
; AVX-NEXT: vinsertf128 $1, 608(%rdi), %ymm0, %ymm0
; AVX-NEXT: vunpcklpd {{.*#+}} ymm15 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
; AVX-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT: vmovaps 512(%rdi), %ymm0
; AVX-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3],mem[2,3]
; AVX-NEXT: vinsertf128 $1, 544(%rdi), %ymm0, %ymm2
; AVX-NEXT: vunpcklpd {{.*#+}} ymm10 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
; AVX-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm2[1],ymm1[1],ymm2[3],ymm1[3]
; AVX-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT: vmovaps 704(%rdi), %ymm1
; AVX-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm1[2,3],mem[2,3]
; AVX-NEXT: vinsertf128 $1, 736(%rdi), %ymm1, %ymm1
; AVX-NEXT: vunpcklpd {{.*#+}} ymm11 = ymm1[0],ymm3[0],ymm1[2],ymm3[2]
; AVX-NEXT: vunpckhpd {{.*#+}} ymm14 = ymm1[1],ymm3[1],ymm1[3],ymm3[3]
; AVX-NEXT: vmovaps 640(%rdi), %ymm1
; AVX-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm1[2,3],mem[2,3]
; AVX-NEXT: vinsertf128 $1, 672(%rdi), %ymm1, %ymm1
; AVX-NEXT: vunpcklpd {{.*#+}} ymm6 = ymm1[0],ymm3[0],ymm1[2],ymm3[2]
; AVX-NEXT: vunpckhpd {{.*#+}} ymm12 = ymm1[1],ymm3[1],ymm1[3],ymm3[3]
; AVX-NEXT: vmovaps 832(%rdi), %ymm1
; AVX-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],mem[2,3]
; AVX-NEXT: vinsertf128 $1, 864(%rdi), %ymm1, %ymm1
; AVX-NEXT: vunpcklpd {{.*#+}} ymm7 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
; AVX-NEXT: vunpckhpd {{.*#+}} ymm9 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
; AVX-NEXT: vmovaps 768(%rdi), %ymm0
; AVX-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3],mem[2,3]
; AVX-NEXT: vinsertf128 $1, 800(%rdi), %ymm0, %ymm0
; AVX-NEXT: vunpcklpd {{.*#+}} ymm3 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX-NEXT: vunpckhpd {{.*#+}} ymm8 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
; AVX-NEXT: vmovaps 960(%rdi), %ymm1
; AVX-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],mem[2,3]
; AVX-NEXT: vinsertf128 $1, 992(%rdi), %ymm1, %ymm1
; AVX-NEXT: vunpcklpd {{.*#+}} ymm4 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
; AVX-NEXT: vunpckhpd {{.*#+}} ymm5 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
; AVX-NEXT: vmovaps 896(%rdi), %ymm1
; AVX-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],mem[2,3]
; AVX-NEXT: vinsertf128 $1, 928(%rdi), %ymm1, %ymm1
; AVX-NEXT: vunpcklpd {{.*#+}} ymm2 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
; AVX-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
; AVX-NEXT: vmovaps %ymm2, 448(%rsi)
; AVX-NEXT: vmovaps %ymm3, 384(%rsi)
; AVX-NEXT: vmovaps %ymm6, 320(%rsi)
; AVX-NEXT: vmovaps %ymm10, 256(%rsi)
; AVX-NEXT: vmovaps %ymm13, 192(%rsi)
; AVX-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX-NEXT: vmovaps %ymm1, 128(%rsi)
; AVX-NEXT: vmovups (%rsp), %ymm1 # 32-byte Reload
; AVX-NEXT: vmovaps %ymm1, 64(%rsi)
; AVX-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX-NEXT: vmovaps %ymm1, (%rsi)
; AVX-NEXT: vmovaps %ymm4, 480(%rsi)
; AVX-NEXT: vmovaps %ymm7, 416(%rsi)
; AVX-NEXT: vmovaps %ymm11, 352(%rsi)
; AVX-NEXT: vmovaps %ymm15, 288(%rsi)
; AVX-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX-NEXT: vmovaps %ymm1, 224(%rsi)
; AVX-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX-NEXT: vmovaps %ymm1, 160(%rsi)
; AVX-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX-NEXT: vmovaps %ymm1, 96(%rsi)
; AVX-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX-NEXT: vmovaps %ymm1, 32(%rsi)
; AVX-NEXT: vmovaps %ymm0, 448(%rdx)
; AVX-NEXT: vmovaps %ymm5, 480(%rdx)
; AVX-NEXT: vmovaps %ymm8, 384(%rdx)
; AVX-NEXT: vmovaps %ymm9, 416(%rdx)
; AVX-NEXT: vmovaps %ymm12, 320(%rdx)
; AVX-NEXT: vmovaps %ymm14, 352(%rdx)
; AVX-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT: vmovaps %ymm0, 256(%rdx)
; AVX-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT: vmovaps %ymm0, 288(%rdx)
; AVX-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT: vmovaps %ymm0, 192(%rdx)
; AVX-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT: vmovaps %ymm0, 224(%rdx)
; AVX-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT: vmovaps %ymm0, 128(%rdx)
; AVX-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT: vmovaps %ymm0, 160(%rdx)
; AVX-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT: vmovaps %ymm0, 64(%rdx)
; AVX-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT: vmovaps %ymm0, 96(%rdx)
; AVX-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT: vmovaps %ymm0, (%rdx)
; AVX-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT: vmovaps %ymm0, 32(%rdx)
; AVX-NEXT: addq $424, %rsp # imm = 0x1A8
; AVX-NEXT: vzeroupper
; AVX-NEXT: retq
;
; AVX2-LABEL: load_i64_stride2_vf64:
; AVX2: # %bb.0:
; AVX2-NEXT: subq $424, %rsp # imm = 0x1A8
; AVX2-NEXT: vmovaps 544(%rdi), %ymm9
; AVX2-NEXT: vmovaps 416(%rdi), %ymm6
; AVX2-NEXT: vmovaps 384(%rdi), %ymm7
; AVX2-NEXT: vmovaps 288(%rdi), %ymm0
; AVX2-NEXT: vmovaps 256(%rdi), %ymm1
; AVX2-NEXT: vmovaps 160(%rdi), %ymm4
; AVX2-NEXT: vmovaps 128(%rdi), %ymm5
; AVX2-NEXT: vmovaps 608(%rdi), %ymm2
; AVX2-NEXT: vmovaps 576(%rdi), %ymm3
; AVX2-NEXT: vmovaps 480(%rdi), %ymm10
; AVX2-NEXT: vmovaps 448(%rdi), %ymm13
; AVX2-NEXT: vmovaps 352(%rdi), %ymm8
; AVX2-NEXT: vmovaps 320(%rdi), %ymm11
; AVX2-NEXT: vmovaps 224(%rdi), %ymm12
; AVX2-NEXT: vmovaps 192(%rdi), %ymm14
; AVX2-NEXT: vunpcklpd {{.*#+}} ymm15 = ymm14[0],ymm12[0],ymm14[2],ymm12[2]
; AVX2-NEXT: vmovups %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT: vunpckhpd {{.*#+}} ymm12 = ymm14[1],ymm12[1],ymm14[3],ymm12[3]
; AVX2-NEXT: vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT: vunpcklpd {{.*#+}} ymm12 = ymm5[0],ymm4[0],ymm5[2],ymm4[2]
; AVX2-NEXT: vmovups %ymm12, (%rsp) # 32-byte Spill
; AVX2-NEXT: vunpckhpd {{.*#+}} ymm4 = ymm5[1],ymm4[1],ymm5[3],ymm4[3]
; AVX2-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT: vunpcklpd {{.*#+}} ymm4 = ymm11[0],ymm8[0],ymm11[2],ymm8[2]
; AVX2-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT: vunpckhpd {{.*#+}} ymm4 = ymm11[1],ymm8[1],ymm11[3],ymm8[3]
; AVX2-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT: vunpcklpd {{.*#+}} ymm4 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
; AVX2-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
; AVX2-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm13[0],ymm10[0],ymm13[2],ymm10[2]
; AVX2-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm13[1],ymm10[1],ymm13[3],ymm10[3]
; AVX2-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT: vunpcklpd {{.*#+}} ymm13 = ymm7[0],ymm6[0],ymm7[2],ymm6[2]
; AVX2-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm7[1],ymm6[1],ymm7[3],ymm6[3]
; AVX2-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
; AVX2-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm3[1],ymm2[1],ymm3[3],ymm2[3]
; AVX2-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT: vmovaps 512(%rdi), %ymm0
; AVX2-NEXT: vunpcklpd {{.*#+}} ymm10 = ymm0[0],ymm9[0],ymm0[2],ymm9[2]
; AVX2-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm9[1],ymm0[3],ymm9[3]
; AVX2-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT: vmovaps 736(%rdi), %ymm0
; AVX2-NEXT: vmovaps 704(%rdi), %ymm1
; AVX2-NEXT: vunpcklpd {{.*#+}} ymm14 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
; AVX2-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
; AVX2-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT: vmovaps 672(%rdi), %ymm0
; AVX2-NEXT: vmovaps 640(%rdi), %ymm2
; AVX2-NEXT: vunpcklpd {{.*#+}} ymm6 = ymm2[0],ymm0[0],ymm2[2],ymm0[2]
; AVX2-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm2[1],ymm0[1],ymm2[3],ymm0[3]
; AVX2-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT: vmovaps 864(%rdi), %ymm0
; AVX2-NEXT: vmovaps 832(%rdi), %ymm2
; AVX2-NEXT: vunpcklpd {{.*#+}} ymm8 = ymm2[0],ymm0[0],ymm2[2],ymm0[2]
; AVX2-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm2[1],ymm0[1],ymm2[3],ymm0[3]
; AVX2-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT: vmovaps 800(%rdi), %ymm0
; AVX2-NEXT: vmovaps 768(%rdi), %ymm2
; AVX2-NEXT: vunpcklpd {{.*#+}} ymm3 = ymm2[0],ymm0[0],ymm2[2],ymm0[2]
; AVX2-NEXT: vunpckhpd {{.*#+}} ymm11 = ymm2[1],ymm0[1],ymm2[3],ymm0[3]
; AVX2-NEXT: vmovaps 992(%rdi), %ymm2
; AVX2-NEXT: vmovaps 960(%rdi), %ymm4
; AVX2-NEXT: vunpcklpd {{.*#+}} ymm5 = ymm4[0],ymm2[0],ymm4[2],ymm2[2]
; AVX2-NEXT: vunpckhpd {{.*#+}} ymm9 = ymm4[1],ymm2[1],ymm4[3],ymm2[3]
; AVX2-NEXT: vmovaps 928(%rdi), %ymm2
; AVX2-NEXT: vmovaps 896(%rdi), %ymm4
; AVX2-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm4[0],ymm2[0],ymm4[2],ymm2[2]
; AVX2-NEXT: vunpckhpd {{.*#+}} ymm7 = ymm4[1],ymm2[1],ymm4[3],ymm2[3]
; AVX2-NEXT: vmovaps 64(%rdi), %ymm12
; AVX2-NEXT: vmovaps 96(%rdi), %ymm15
; AVX2-NEXT: vunpcklpd {{.*#+}} ymm4 = ymm12[0],ymm15[0],ymm12[2],ymm15[2]
; AVX2-NEXT: vunpckhpd {{.*#+}} ymm15 = ymm12[1],ymm15[1],ymm12[3],ymm15[3]
; AVX2-NEXT: vmovaps (%rdi), %ymm12
; AVX2-NEXT: vmovaps 32(%rdi), %ymm0
; AVX2-NEXT: vunpcklpd {{.*#+}} ymm2 = ymm12[0],ymm0[0],ymm12[2],ymm0[2]
; AVX2-NEXT: vunpckhpd {{.*#+}} ymm12 = ymm12[1],ymm0[1],ymm12[3],ymm0[3]
; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm1[0,2,1,3]
; AVX2-NEXT: vmovaps %ymm0, 448(%rsi)
; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm3[0,2,1,3]
; AVX2-NEXT: vmovaps %ymm0, 384(%rsi)
; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm6[0,2,1,3]
; AVX2-NEXT: vmovaps %ymm0, 320(%rsi)
; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm10[0,2,1,3]
; AVX2-NEXT: vmovaps %ymm0, 256(%rsi)
; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm13[0,2,1,3]
; AVX2-NEXT: vmovaps %ymm0, 192(%rsi)
; AVX2-NEXT: vpermpd $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX2-NEXT: # ymm0 = mem[0,2,1,3]
; AVX2-NEXT: vmovaps %ymm0, 128(%rsi)
; AVX2-NEXT: vpermpd $216, (%rsp), %ymm0 # 32-byte Folded Reload
; AVX2-NEXT: # ymm0 = mem[0,2,1,3]
; AVX2-NEXT: vmovaps %ymm0, 64(%rsi)
; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm2[0,2,1,3]
; AVX2-NEXT: vmovaps %ymm0, (%rsi)
; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm5[0,2,1,3]
; AVX2-NEXT: vmovaps %ymm0, 480(%rsi)
; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm8[0,2,1,3]
; AVX2-NEXT: vmovaps %ymm0, 416(%rsi)
; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm14[0,2,1,3]
; AVX2-NEXT: vmovaps %ymm0, 352(%rsi)
; AVX2-NEXT: vpermpd $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX2-NEXT: # ymm0 = mem[0,2,1,3]
; AVX2-NEXT: vmovaps %ymm0, 288(%rsi)
; AVX2-NEXT: vpermpd $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX2-NEXT: # ymm0 = mem[0,2,1,3]
; AVX2-NEXT: vmovaps %ymm0, 224(%rsi)
; AVX2-NEXT: vpermpd $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX2-NEXT: # ymm0 = mem[0,2,1,3]
; AVX2-NEXT: vmovaps %ymm0, 160(%rsi)
; AVX2-NEXT: vpermpd $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX2-NEXT: # ymm0 = mem[0,2,1,3]
; AVX2-NEXT: vmovaps %ymm0, 96(%rsi)
; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm4[0,2,1,3]
; AVX2-NEXT: vmovaps %ymm0, 32(%rsi)
; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm7[0,2,1,3]
; AVX2-NEXT: vmovaps %ymm0, 448(%rdx)
; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm9[0,2,1,3]
; AVX2-NEXT: vmovaps %ymm0, 480(%rdx)
; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm11[0,2,1,3]
; AVX2-NEXT: vmovaps %ymm0, 384(%rdx)
; AVX2-NEXT: vpermpd $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX2-NEXT: # ymm0 = mem[0,2,1,3]
; AVX2-NEXT: vmovaps %ymm0, 416(%rdx)
; AVX2-NEXT: vpermpd $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX2-NEXT: # ymm0 = mem[0,2,1,3]
; AVX2-NEXT: vmovaps %ymm0, 320(%rdx)
; AVX2-NEXT: vpermpd $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX2-NEXT: # ymm0 = mem[0,2,1,3]
; AVX2-NEXT: vmovaps %ymm0, 352(%rdx)
; AVX2-NEXT: vpermpd $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX2-NEXT: # ymm0 = mem[0,2,1,3]
; AVX2-NEXT: vmovaps %ymm0, 256(%rdx)
; AVX2-NEXT: vpermpd $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX2-NEXT: # ymm0 = mem[0,2,1,3]
; AVX2-NEXT: vmovaps %ymm0, 288(%rdx)
; AVX2-NEXT: vpermpd $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX2-NEXT: # ymm0 = mem[0,2,1,3]
; AVX2-NEXT: vmovaps %ymm0, 192(%rdx)
; AVX2-NEXT: vpermpd $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX2-NEXT: # ymm0 = mem[0,2,1,3]
; AVX2-NEXT: vmovaps %ymm0, 224(%rdx)
; AVX2-NEXT: vpermpd $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX2-NEXT: # ymm0 = mem[0,2,1,3]
; AVX2-NEXT: vmovaps %ymm0, 128(%rdx)
; AVX2-NEXT: vpermpd $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX2-NEXT: # ymm0 = mem[0,2,1,3]
; AVX2-NEXT: vmovaps %ymm0, 160(%rdx)
; AVX2-NEXT: vpermpd $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX2-NEXT: # ymm0 = mem[0,2,1,3]
; AVX2-NEXT: vmovaps %ymm0, 64(%rdx)
; AVX2-NEXT: vpermpd $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX2-NEXT: # ymm0 = mem[0,2,1,3]
; AVX2-NEXT: vmovaps %ymm0, 96(%rdx)
; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm12[0,2,1,3]
; AVX2-NEXT: vmovaps %ymm0, (%rdx)
; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm15[0,2,1,3]
; AVX2-NEXT: vmovaps %ymm0, 32(%rdx)
; AVX2-NEXT: addq $424, %rsp # imm = 0x1A8
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
; AVX2-FP-LABEL: load_i64_stride2_vf64:
; AVX2-FP: # %bb.0:
; AVX2-FP-NEXT: subq $424, %rsp # imm = 0x1A8
; AVX2-FP-NEXT: vmovaps 544(%rdi), %ymm9
; AVX2-FP-NEXT: vmovaps 416(%rdi), %ymm6
; AVX2-FP-NEXT: vmovaps 384(%rdi), %ymm7
; AVX2-FP-NEXT: vmovaps 288(%rdi), %ymm0
; AVX2-FP-NEXT: vmovaps 256(%rdi), %ymm1
; AVX2-FP-NEXT: vmovaps 160(%rdi), %ymm4
; AVX2-FP-NEXT: vmovaps 128(%rdi), %ymm5
; AVX2-FP-NEXT: vmovaps 608(%rdi), %ymm2
; AVX2-FP-NEXT: vmovaps 576(%rdi), %ymm3
; AVX2-FP-NEXT: vmovaps 480(%rdi), %ymm10
; AVX2-FP-NEXT: vmovaps 448(%rdi), %ymm13
; AVX2-FP-NEXT: vmovaps 352(%rdi), %ymm8
; AVX2-FP-NEXT: vmovaps 320(%rdi), %ymm11
; AVX2-FP-NEXT: vmovaps 224(%rdi), %ymm12
; AVX2-FP-NEXT: vmovaps 192(%rdi), %ymm14
; AVX2-FP-NEXT: vunpcklpd {{.*#+}} ymm15 = ymm14[0],ymm12[0],ymm14[2],ymm12[2]
; AVX2-FP-NEXT: vmovups %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT: vunpckhpd {{.*#+}} ymm12 = ymm14[1],ymm12[1],ymm14[3],ymm12[3]
; AVX2-FP-NEXT: vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT: vunpcklpd {{.*#+}} ymm12 = ymm5[0],ymm4[0],ymm5[2],ymm4[2]
; AVX2-FP-NEXT: vmovups %ymm12, (%rsp) # 32-byte Spill
; AVX2-FP-NEXT: vunpckhpd {{.*#+}} ymm4 = ymm5[1],ymm4[1],ymm5[3],ymm4[3]
; AVX2-FP-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT: vunpcklpd {{.*#+}} ymm4 = ymm11[0],ymm8[0],ymm11[2],ymm8[2]
; AVX2-FP-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT: vunpckhpd {{.*#+}} ymm4 = ymm11[1],ymm8[1],ymm11[3],ymm8[3]
; AVX2-FP-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT: vunpcklpd {{.*#+}} ymm4 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
; AVX2-FP-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
; AVX2-FP-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm13[0],ymm10[0],ymm13[2],ymm10[2]
; AVX2-FP-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm13[1],ymm10[1],ymm13[3],ymm10[3]
; AVX2-FP-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT: vunpcklpd {{.*#+}} ymm13 = ymm7[0],ymm6[0],ymm7[2],ymm6[2]
; AVX2-FP-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm7[1],ymm6[1],ymm7[3],ymm6[3]
; AVX2-FP-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
; AVX2-FP-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm3[1],ymm2[1],ymm3[3],ymm2[3]
; AVX2-FP-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT: vmovaps 512(%rdi), %ymm0
; AVX2-FP-NEXT: vunpcklpd {{.*#+}} ymm10 = ymm0[0],ymm9[0],ymm0[2],ymm9[2]
; AVX2-FP-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm9[1],ymm0[3],ymm9[3]
; AVX2-FP-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT: vmovaps 736(%rdi), %ymm0
; AVX2-FP-NEXT: vmovaps 704(%rdi), %ymm1
; AVX2-FP-NEXT: vunpcklpd {{.*#+}} ymm14 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
; AVX2-FP-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
; AVX2-FP-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT: vmovaps 672(%rdi), %ymm0
; AVX2-FP-NEXT: vmovaps 640(%rdi), %ymm2
; AVX2-FP-NEXT: vunpcklpd {{.*#+}} ymm6 = ymm2[0],ymm0[0],ymm2[2],ymm0[2]
; AVX2-FP-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm2[1],ymm0[1],ymm2[3],ymm0[3]
; AVX2-FP-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT: vmovaps 864(%rdi), %ymm0
; AVX2-FP-NEXT: vmovaps 832(%rdi), %ymm2
; AVX2-FP-NEXT: vunpcklpd {{.*#+}} ymm8 = ymm2[0],ymm0[0],ymm2[2],ymm0[2]
; AVX2-FP-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm2[1],ymm0[1],ymm2[3],ymm0[3]
; AVX2-FP-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT: vmovaps 800(%rdi), %ymm0
; AVX2-FP-NEXT: vmovaps 768(%rdi), %ymm2
; AVX2-FP-NEXT: vunpcklpd {{.*#+}} ymm3 = ymm2[0],ymm0[0],ymm2[2],ymm0[2]
; AVX2-FP-NEXT: vunpckhpd {{.*#+}} ymm11 = ymm2[1],ymm0[1],ymm2[3],ymm0[3]
; AVX2-FP-NEXT: vmovaps 992(%rdi), %ymm2
; AVX2-FP-NEXT: vmovaps 960(%rdi), %ymm4
; AVX2-FP-NEXT: vunpcklpd {{.*#+}} ymm5 = ymm4[0],ymm2[0],ymm4[2],ymm2[2]
; AVX2-FP-NEXT: vunpckhpd {{.*#+}} ymm9 = ymm4[1],ymm2[1],ymm4[3],ymm2[3]
; AVX2-FP-NEXT: vmovaps 928(%rdi), %ymm2
; AVX2-FP-NEXT: vmovaps 896(%rdi), %ymm4
; AVX2-FP-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm4[0],ymm2[0],ymm4[2],ymm2[2]
; AVX2-FP-NEXT: vunpckhpd {{.*#+}} ymm7 = ymm4[1],ymm2[1],ymm4[3],ymm2[3]
; AVX2-FP-NEXT: vmovaps 64(%rdi), %ymm12
; AVX2-FP-NEXT: vmovaps 96(%rdi), %ymm15
; AVX2-FP-NEXT: vunpcklpd {{.*#+}} ymm4 = ymm12[0],ymm15[0],ymm12[2],ymm15[2]
; AVX2-FP-NEXT: vunpckhpd {{.*#+}} ymm15 = ymm12[1],ymm15[1],ymm12[3],ymm15[3]
; AVX2-FP-NEXT: vmovaps (%rdi), %ymm12
; AVX2-FP-NEXT: vmovaps 32(%rdi), %ymm0
; AVX2-FP-NEXT: vunpcklpd {{.*#+}} ymm2 = ymm12[0],ymm0[0],ymm12[2],ymm0[2]
; AVX2-FP-NEXT: vunpckhpd {{.*#+}} ymm12 = ymm12[1],ymm0[1],ymm12[3],ymm0[3]
; AVX2-FP-NEXT: vpermpd {{.*#+}} ymm0 = ymm1[0,2,1,3]
; AVX2-FP-NEXT: vmovaps %ymm0, 448(%rsi)
; AVX2-FP-NEXT: vpermpd {{.*#+}} ymm0 = ymm3[0,2,1,3]
; AVX2-FP-NEXT: vmovaps %ymm0, 384(%rsi)
; AVX2-FP-NEXT: vpermpd {{.*#+}} ymm0 = ymm6[0,2,1,3]
; AVX2-FP-NEXT: vmovaps %ymm0, 320(%rsi)
; AVX2-FP-NEXT: vpermpd {{.*#+}} ymm0 = ymm10[0,2,1,3]
; AVX2-FP-NEXT: vmovaps %ymm0, 256(%rsi)
; AVX2-FP-NEXT: vpermpd {{.*#+}} ymm0 = ymm13[0,2,1,3]
; AVX2-FP-NEXT: vmovaps %ymm0, 192(%rsi)
; AVX2-FP-NEXT: vpermpd $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX2-FP-NEXT: # ymm0 = mem[0,2,1,3]
; AVX2-FP-NEXT: vmovaps %ymm0, 128(%rsi)
; AVX2-FP-NEXT: vpermpd $216, (%rsp), %ymm0 # 32-byte Folded Reload
; AVX2-FP-NEXT: # ymm0 = mem[0,2,1,3]
; AVX2-FP-NEXT: vmovaps %ymm0, 64(%rsi)
; AVX2-FP-NEXT: vpermpd {{.*#+}} ymm0 = ymm2[0,2,1,3]
; AVX2-FP-NEXT: vmovaps %ymm0, (%rsi)
; AVX2-FP-NEXT: vpermpd {{.*#+}} ymm0 = ymm5[0,2,1,3]
; AVX2-FP-NEXT: vmovaps %ymm0, 480(%rsi)
; AVX2-FP-NEXT: vpermpd {{.*#+}} ymm0 = ymm8[0,2,1,3]
; AVX2-FP-NEXT: vmovaps %ymm0, 416(%rsi)
; AVX2-FP-NEXT: vpermpd {{.*#+}} ymm0 = ymm14[0,2,1,3]
; AVX2-FP-NEXT: vmovaps %ymm0, 352(%rsi)
; AVX2-FP-NEXT: vpermpd $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX2-FP-NEXT: # ymm0 = mem[0,2,1,3]
; AVX2-FP-NEXT: vmovaps %ymm0, 288(%rsi)
; AVX2-FP-NEXT: vpermpd $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX2-FP-NEXT: # ymm0 = mem[0,2,1,3]
; AVX2-FP-NEXT: vmovaps %ymm0, 224(%rsi)
; AVX2-FP-NEXT: vpermpd $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX2-FP-NEXT: # ymm0 = mem[0,2,1,3]
; AVX2-FP-NEXT: vmovaps %ymm0, 160(%rsi)
; AVX2-FP-NEXT: vpermpd $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX2-FP-NEXT: # ymm0 = mem[0,2,1,3]
; AVX2-FP-NEXT: vmovaps %ymm0, 96(%rsi)
; AVX2-FP-NEXT: vpermpd {{.*#+}} ymm0 = ymm4[0,2,1,3]
; AVX2-FP-NEXT: vmovaps %ymm0, 32(%rsi)
; AVX2-FP-NEXT: vpermpd {{.*#+}} ymm0 = ymm7[0,2,1,3]
; AVX2-FP-NEXT: vmovaps %ymm0, 448(%rdx)
; AVX2-FP-NEXT: vpermpd {{.*#+}} ymm0 = ymm9[0,2,1,3]
; AVX2-FP-NEXT: vmovaps %ymm0, 480(%rdx)
; AVX2-FP-NEXT: vpermpd {{.*#+}} ymm0 = ymm11[0,2,1,3]
; AVX2-FP-NEXT: vmovaps %ymm0, 384(%rdx)
; AVX2-FP-NEXT: vpermpd $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX2-FP-NEXT: # ymm0 = mem[0,2,1,3]
; AVX2-FP-NEXT: vmovaps %ymm0, 416(%rdx)
; AVX2-FP-NEXT: vpermpd $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX2-FP-NEXT: # ymm0 = mem[0,2,1,3]
; AVX2-FP-NEXT: vmovaps %ymm0, 320(%rdx)
; AVX2-FP-NEXT: vpermpd $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX2-FP-NEXT: # ymm0 = mem[0,2,1,3]
; AVX2-FP-NEXT: vmovaps %ymm0, 352(%rdx)
; AVX2-FP-NEXT: vpermpd $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX2-FP-NEXT: # ymm0 = mem[0,2,1,3]
; AVX2-FP-NEXT: vmovaps %ymm0, 256(%rdx)
; AVX2-FP-NEXT: vpermpd $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX2-FP-NEXT: # ymm0 = mem[0,2,1,3]
; AVX2-FP-NEXT: vmovaps %ymm0, 288(%rdx)
; AVX2-FP-NEXT: vpermpd $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX2-FP-NEXT: # ymm0 = mem[0,2,1,3]
; AVX2-FP-NEXT: vmovaps %ymm0, 192(%rdx)
; AVX2-FP-NEXT: vpermpd $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX2-FP-NEXT: # ymm0 = mem[0,2,1,3]
; AVX2-FP-NEXT: vmovaps %ymm0, 224(%rdx)
; AVX2-FP-NEXT: vpermpd $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX2-FP-NEXT: # ymm0 = mem[0,2,1,3]
; AVX2-FP-NEXT: vmovaps %ymm0, 128(%rdx)
; AVX2-FP-NEXT: vpermpd $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX2-FP-NEXT: # ymm0 = mem[0,2,1,3]
; AVX2-FP-NEXT: vmovaps %ymm0, 160(%rdx)
; AVX2-FP-NEXT: vpermpd $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX2-FP-NEXT: # ymm0 = mem[0,2,1,3]
; AVX2-FP-NEXT: vmovaps %ymm0, 64(%rdx)
; AVX2-FP-NEXT: vpermpd $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX2-FP-NEXT: # ymm0 = mem[0,2,1,3]
; AVX2-FP-NEXT: vmovaps %ymm0, 96(%rdx)
; AVX2-FP-NEXT: vpermpd {{.*#+}} ymm0 = ymm12[0,2,1,3]
; AVX2-FP-NEXT: vmovaps %ymm0, (%rdx)
; AVX2-FP-NEXT: vpermpd {{.*#+}} ymm0 = ymm15[0,2,1,3]
; AVX2-FP-NEXT: vmovaps %ymm0, 32(%rdx)
; AVX2-FP-NEXT: addq $424, %rsp # imm = 0x1A8
; AVX2-FP-NEXT: vzeroupper
; AVX2-FP-NEXT: retq
;
; AVX2-FCP-LABEL: load_i64_stride2_vf64:
; AVX2-FCP: # %bb.0:
; AVX2-FCP-NEXT: subq $424, %rsp # imm = 0x1A8
; AVX2-FCP-NEXT: vmovaps 544(%rdi), %ymm9
; AVX2-FCP-NEXT: vmovaps 416(%rdi), %ymm6
; AVX2-FCP-NEXT: vmovaps 384(%rdi), %ymm7
; AVX2-FCP-NEXT: vmovaps 288(%rdi), %ymm0
; AVX2-FCP-NEXT: vmovaps 256(%rdi), %ymm1
; AVX2-FCP-NEXT: vmovaps 160(%rdi), %ymm4
; AVX2-FCP-NEXT: vmovaps 128(%rdi), %ymm5
; AVX2-FCP-NEXT: vmovaps 608(%rdi), %ymm2
; AVX2-FCP-NEXT: vmovaps 576(%rdi), %ymm3
; AVX2-FCP-NEXT: vmovaps 480(%rdi), %ymm10
; AVX2-FCP-NEXT: vmovaps 448(%rdi), %ymm13
; AVX2-FCP-NEXT: vmovaps 352(%rdi), %ymm8
; AVX2-FCP-NEXT: vmovaps 320(%rdi), %ymm11
; AVX2-FCP-NEXT: vmovaps 224(%rdi), %ymm12
; AVX2-FCP-NEXT: vmovaps 192(%rdi), %ymm14
; AVX2-FCP-NEXT: vunpcklpd {{.*#+}} ymm15 = ymm14[0],ymm12[0],ymm14[2],ymm12[2]
; AVX2-FCP-NEXT: vmovups %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT: vunpckhpd {{.*#+}} ymm12 = ymm14[1],ymm12[1],ymm14[3],ymm12[3]
; AVX2-FCP-NEXT: vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT: vunpcklpd {{.*#+}} ymm12 = ymm5[0],ymm4[0],ymm5[2],ymm4[2]
; AVX2-FCP-NEXT: vmovups %ymm12, (%rsp) # 32-byte Spill
; AVX2-FCP-NEXT: vunpckhpd {{.*#+}} ymm4 = ymm5[1],ymm4[1],ymm5[3],ymm4[3]
; AVX2-FCP-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT: vunpcklpd {{.*#+}} ymm4 = ymm11[0],ymm8[0],ymm11[2],ymm8[2]
; AVX2-FCP-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT: vunpckhpd {{.*#+}} ymm4 = ymm11[1],ymm8[1],ymm11[3],ymm8[3]
; AVX2-FCP-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT: vunpcklpd {{.*#+}} ymm4 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
; AVX2-FCP-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
; AVX2-FCP-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm13[0],ymm10[0],ymm13[2],ymm10[2]
; AVX2-FCP-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm13[1],ymm10[1],ymm13[3],ymm10[3]
; AVX2-FCP-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT: vunpcklpd {{.*#+}} ymm13 = ymm7[0],ymm6[0],ymm7[2],ymm6[2]
; AVX2-FCP-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm7[1],ymm6[1],ymm7[3],ymm6[3]
; AVX2-FCP-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
; AVX2-FCP-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm3[1],ymm2[1],ymm3[3],ymm2[3]
; AVX2-FCP-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT: vmovaps 512(%rdi), %ymm0
; AVX2-FCP-NEXT: vunpcklpd {{.*#+}} ymm10 = ymm0[0],ymm9[0],ymm0[2],ymm9[2]
; AVX2-FCP-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm9[1],ymm0[3],ymm9[3]
; AVX2-FCP-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT: vmovaps 736(%rdi), %ymm0
; AVX2-FCP-NEXT: vmovaps 704(%rdi), %ymm1
; AVX2-FCP-NEXT: vunpcklpd {{.*#+}} ymm14 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
; AVX2-FCP-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
; AVX2-FCP-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT: vmovaps 672(%rdi), %ymm0
; AVX2-FCP-NEXT: vmovaps 640(%rdi), %ymm2
; AVX2-FCP-NEXT: vunpcklpd {{.*#+}} ymm6 = ymm2[0],ymm0[0],ymm2[2],ymm0[2]
; AVX2-FCP-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm2[1],ymm0[1],ymm2[3],ymm0[3]
; AVX2-FCP-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT: vmovaps 864(%rdi), %ymm0
; AVX2-FCP-NEXT: vmovaps 832(%rdi), %ymm2
; AVX2-FCP-NEXT: vunpcklpd {{.*#+}} ymm8 = ymm2[0],ymm0[0],ymm2[2],ymm0[2]
; AVX2-FCP-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm2[1],ymm0[1],ymm2[3],ymm0[3]
; AVX2-FCP-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT: vmovaps 800(%rdi), %ymm0
; AVX2-FCP-NEXT: vmovaps 768(%rdi), %ymm2
; AVX2-FCP-NEXT: vunpcklpd {{.*#+}} ymm3 = ymm2[0],ymm0[0],ymm2[2],ymm0[2]
; AVX2-FCP-NEXT: vunpckhpd {{.*#+}} ymm11 = ymm2[1],ymm0[1],ymm2[3],ymm0[3]
; AVX2-FCP-NEXT: vmovaps 992(%rdi), %ymm2
; AVX2-FCP-NEXT: vmovaps 960(%rdi), %ymm4
; AVX2-FCP-NEXT: vunpcklpd {{.*#+}} ymm5 = ymm4[0],ymm2[0],ymm4[2],ymm2[2]
; AVX2-FCP-NEXT: vunpckhpd {{.*#+}} ymm9 = ymm4[1],ymm2[1],ymm4[3],ymm2[3]
; AVX2-FCP-NEXT: vmovaps 928(%rdi), %ymm2
; AVX2-FCP-NEXT: vmovaps 896(%rdi), %ymm4
; AVX2-FCP-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm4[0],ymm2[0],ymm4[2],ymm2[2]
; AVX2-FCP-NEXT: vunpckhpd {{.*#+}} ymm7 = ymm4[1],ymm2[1],ymm4[3],ymm2[3]
; AVX2-FCP-NEXT: vmovaps 64(%rdi), %ymm12
; AVX2-FCP-NEXT: vmovaps 96(%rdi), %ymm15
; AVX2-FCP-NEXT: vunpcklpd {{.*#+}} ymm4 = ymm12[0],ymm15[0],ymm12[2],ymm15[2]
; AVX2-FCP-NEXT: vunpckhpd {{.*#+}} ymm15 = ymm12[1],ymm15[1],ymm12[3],ymm15[3]
; AVX2-FCP-NEXT: vmovaps (%rdi), %ymm12
; AVX2-FCP-NEXT: vmovaps 32(%rdi), %ymm0
; AVX2-FCP-NEXT: vunpcklpd {{.*#+}} ymm2 = ymm12[0],ymm0[0],ymm12[2],ymm0[2]
; AVX2-FCP-NEXT: vunpckhpd {{.*#+}} ymm12 = ymm12[1],ymm0[1],ymm12[3],ymm0[3]
; AVX2-FCP-NEXT: vpermpd {{.*#+}} ymm0 = ymm1[0,2,1,3]
; AVX2-FCP-NEXT: vmovaps %ymm0, 448(%rsi)
; AVX2-FCP-NEXT: vpermpd {{.*#+}} ymm0 = ymm3[0,2,1,3]
; AVX2-FCP-NEXT: vmovaps %ymm0, 384(%rsi)
; AVX2-FCP-NEXT: vpermpd {{.*#+}} ymm0 = ymm6[0,2,1,3]
; AVX2-FCP-NEXT: vmovaps %ymm0, 320(%rsi)
; AVX2-FCP-NEXT: vpermpd {{.*#+}} ymm0 = ymm10[0,2,1,3]
; AVX2-FCP-NEXT: vmovaps %ymm0, 256(%rsi)
; AVX2-FCP-NEXT: vpermpd {{.*#+}} ymm0 = ymm13[0,2,1,3]
; AVX2-FCP-NEXT: vmovaps %ymm0, 192(%rsi)
; AVX2-FCP-NEXT: vpermpd $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX2-FCP-NEXT: # ymm0 = mem[0,2,1,3]
; AVX2-FCP-NEXT: vmovaps %ymm0, 128(%rsi)
; AVX2-FCP-NEXT: vpermpd $216, (%rsp), %ymm0 # 32-byte Folded Reload
; AVX2-FCP-NEXT: # ymm0 = mem[0,2,1,3]
; AVX2-FCP-NEXT: vmovaps %ymm0, 64(%rsi)
; AVX2-FCP-NEXT: vpermpd {{.*#+}} ymm0 = ymm2[0,2,1,3]
; AVX2-FCP-NEXT: vmovaps %ymm0, (%rsi)
; AVX2-FCP-NEXT: vpermpd {{.*#+}} ymm0 = ymm5[0,2,1,3]
; AVX2-FCP-NEXT: vmovaps %ymm0, 480(%rsi)
; AVX2-FCP-NEXT: vpermpd {{.*#+}} ymm0 = ymm8[0,2,1,3]
; AVX2-FCP-NEXT: vmovaps %ymm0, 416(%rsi)
; AVX2-FCP-NEXT: vpermpd {{.*#+}} ymm0 = ymm14[0,2,1,3]
; AVX2-FCP-NEXT: vmovaps %ymm0, 352(%rsi)
; AVX2-FCP-NEXT: vpermpd $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX2-FCP-NEXT: # ymm0 = mem[0,2,1,3]
; AVX2-FCP-NEXT: vmovaps %ymm0, 288(%rsi)
; AVX2-FCP-NEXT: vpermpd $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX2-FCP-NEXT: # ymm0 = mem[0,2,1,3]
; AVX2-FCP-NEXT: vmovaps %ymm0, 224(%rsi)
; AVX2-FCP-NEXT: vpermpd $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX2-FCP-NEXT: # ymm0 = mem[0,2,1,3]
; AVX2-FCP-NEXT: vmovaps %ymm0, 160(%rsi)
; AVX2-FCP-NEXT: vpermpd $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX2-FCP-NEXT: # ymm0 = mem[0,2,1,3]
; AVX2-FCP-NEXT: vmovaps %ymm0, 96(%rsi)
; AVX2-FCP-NEXT: vpermpd {{.*#+}} ymm0 = ymm4[0,2,1,3]
; AVX2-FCP-NEXT: vmovaps %ymm0, 32(%rsi)
; AVX2-FCP-NEXT: vpermpd {{.*#+}} ymm0 = ymm7[0,2,1,3]
; AVX2-FCP-NEXT: vmovaps %ymm0, 448(%rdx)
; AVX2-FCP-NEXT: vpermpd {{.*#+}} ymm0 = ymm9[0,2,1,3]
; AVX2-FCP-NEXT: vmovaps %ymm0, 480(%rdx)
; AVX2-FCP-NEXT: vpermpd {{.*#+}} ymm0 = ymm11[0,2,1,3]
; AVX2-FCP-NEXT: vmovaps %ymm0, 384(%rdx)
; AVX2-FCP-NEXT: vpermpd $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX2-FCP-NEXT: # ymm0 = mem[0,2,1,3]
; AVX2-FCP-NEXT: vmovaps %ymm0, 416(%rdx)
; AVX2-FCP-NEXT: vpermpd $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX2-FCP-NEXT: # ymm0 = mem[0,2,1,3]
; AVX2-FCP-NEXT: vmovaps %ymm0, 320(%rdx)
; AVX2-FCP-NEXT: vpermpd $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX2-FCP-NEXT: # ymm0 = mem[0,2,1,3]
; AVX2-FCP-NEXT: vmovaps %ymm0, 352(%rdx)
; AVX2-FCP-NEXT: vpermpd $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX2-FCP-NEXT: # ymm0 = mem[0,2,1,3]
; AVX2-FCP-NEXT: vmovaps %ymm0, 256(%rdx)
; AVX2-FCP-NEXT: vpermpd $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX2-FCP-NEXT: # ymm0 = mem[0,2,1,3]
; AVX2-FCP-NEXT: vmovaps %ymm0, 288(%rdx)
; AVX2-FCP-NEXT: vpermpd $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX2-FCP-NEXT: # ymm0 = mem[0,2,1,3]
; AVX2-FCP-NEXT: vmovaps %ymm0, 192(%rdx)
; AVX2-FCP-NEXT: vpermpd $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX2-FCP-NEXT: # ymm0 = mem[0,2,1,3]
; AVX2-FCP-NEXT: vmovaps %ymm0, 224(%rdx)
; AVX2-FCP-NEXT: vpermpd $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX2-FCP-NEXT: # ymm0 = mem[0,2,1,3]
; AVX2-FCP-NEXT: vmovaps %ymm0, 128(%rdx)
; AVX2-FCP-NEXT: vpermpd $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX2-FCP-NEXT: # ymm0 = mem[0,2,1,3]
; AVX2-FCP-NEXT: vmovaps %ymm0, 160(%rdx)
; AVX2-FCP-NEXT: vpermpd $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX2-FCP-NEXT: # ymm0 = mem[0,2,1,3]
; AVX2-FCP-NEXT: vmovaps %ymm0, 64(%rdx)
; AVX2-FCP-NEXT: vpermpd $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX2-FCP-NEXT: # ymm0 = mem[0,2,1,3]
; AVX2-FCP-NEXT: vmovaps %ymm0, 96(%rdx)
; AVX2-FCP-NEXT: vpermpd {{.*#+}} ymm0 = ymm12[0,2,1,3]
; AVX2-FCP-NEXT: vmovaps %ymm0, (%rdx)
; AVX2-FCP-NEXT: vpermpd {{.*#+}} ymm0 = ymm15[0,2,1,3]
; AVX2-FCP-NEXT: vmovaps %ymm0, 32(%rdx)
; AVX2-FCP-NEXT: addq $424, %rsp # imm = 0x1A8
; AVX2-FCP-NEXT: vzeroupper
; AVX2-FCP-NEXT: retq
;
; AVX512-LABEL: load_i64_stride2_vf64:
; AVX512: # %bb.0:
; AVX512-NEXT: vmovdqa64 (%rdi), %zmm0
; AVX512-NEXT: vmovdqa64 64(%rdi), %zmm1
; AVX512-NEXT: vmovdqa64 128(%rdi), %zmm2
; AVX512-NEXT: vmovdqa64 192(%rdi), %zmm3
; AVX512-NEXT: vmovdqa64 960(%rdi), %zmm4
; AVX512-NEXT: vmovdqa64 896(%rdi), %zmm5
; AVX512-NEXT: vmovdqa64 576(%rdi), %zmm6
; AVX512-NEXT: vmovdqa64 512(%rdi), %zmm7
; AVX512-NEXT: vmovdqa64 704(%rdi), %zmm8
; AVX512-NEXT: vmovdqa64 640(%rdi), %zmm9
; AVX512-NEXT: vmovdqa64 320(%rdi), %zmm10
; AVX512-NEXT: vmovdqa64 256(%rdi), %zmm11
; AVX512-NEXT: vmovdqa64 448(%rdi), %zmm12
; AVX512-NEXT: vmovdqa64 384(%rdi), %zmm13
; AVX512-NEXT: vmovdqa64 832(%rdi), %zmm14
; AVX512-NEXT: vmovdqa64 768(%rdi), %zmm15
; AVX512-NEXT: vpmovsxbq {{.*#+}} zmm16 = [0,2,4,6,8,10,12,14]
; AVX512-NEXT: vmovdqa64 %zmm15, %zmm17
; AVX512-NEXT: vpermt2q %zmm14, %zmm16, %zmm17
; AVX512-NEXT: vmovdqa64 %zmm13, %zmm18
; AVX512-NEXT: vpermt2q %zmm12, %zmm16, %zmm18
; AVX512-NEXT: vmovdqa64 %zmm11, %zmm19
; AVX512-NEXT: vpermt2q %zmm10, %zmm16, %zmm19
; AVX512-NEXT: vmovdqa64 %zmm9, %zmm20
; AVX512-NEXT: vpermt2q %zmm8, %zmm16, %zmm20
; AVX512-NEXT: vmovdqa64 %zmm7, %zmm21
; AVX512-NEXT: vpermt2q %zmm6, %zmm16, %zmm21
; AVX512-NEXT: vmovdqa64 %zmm5, %zmm22
; AVX512-NEXT: vpermt2q %zmm4, %zmm16, %zmm22
; AVX512-NEXT: vmovdqa64 %zmm2, %zmm23
; AVX512-NEXT: vpermt2q %zmm3, %zmm16, %zmm23
; AVX512-NEXT: vpermi2q %zmm1, %zmm0, %zmm16
; AVX512-NEXT: vpmovsxbq {{.*#+}} zmm24 = [1,3,5,7,9,11,13,15]
; AVX512-NEXT: vpermt2q %zmm12, %zmm24, %zmm13
; AVX512-NEXT: vpermt2q %zmm10, %zmm24, %zmm11
; AVX512-NEXT: vpermt2q %zmm8, %zmm24, %zmm9
; AVX512-NEXT: vpermt2q %zmm6, %zmm24, %zmm7
; AVX512-NEXT: vpermt2q %zmm4, %zmm24, %zmm5
; AVX512-NEXT: vpermt2q %zmm14, %zmm24, %zmm15
; AVX512-NEXT: vpermt2q %zmm3, %zmm24, %zmm2
; AVX512-NEXT: vpermt2q %zmm1, %zmm24, %zmm0
; AVX512-NEXT: vmovdqa64 %zmm22, 448(%rsi)
; AVX512-NEXT: vmovdqa64 %zmm21, 256(%rsi)
; AVX512-NEXT: vmovdqa64 %zmm20, 320(%rsi)
; AVX512-NEXT: vmovdqa64 %zmm19, 128(%rsi)
; AVX512-NEXT: vmovdqa64 %zmm18, 192(%rsi)
; AVX512-NEXT: vmovdqa64 %zmm16, (%rsi)
; AVX512-NEXT: vmovdqa64 %zmm23, 64(%rsi)
; AVX512-NEXT: vmovdqa64 %zmm17, 384(%rsi)
; AVX512-NEXT: vmovdqa64 %zmm15, 384(%rdx)
; AVX512-NEXT: vmovdqa64 %zmm5, 448(%rdx)
; AVX512-NEXT: vmovdqa64 %zmm7, 256(%rdx)
; AVX512-NEXT: vmovdqa64 %zmm9, 320(%rdx)
; AVX512-NEXT: vmovdqa64 %zmm11, 128(%rdx)
; AVX512-NEXT: vmovdqa64 %zmm13, 192(%rdx)
; AVX512-NEXT: vmovdqa64 %zmm0, (%rdx)
; AVX512-NEXT: vmovdqa64 %zmm2, 64(%rdx)
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
;
; AVX512-FCP-LABEL: load_i64_stride2_vf64:
; AVX512-FCP: # %bb.0:
; AVX512-FCP-NEXT: vmovdqa64 (%rdi), %zmm0
; AVX512-FCP-NEXT: vmovdqa64 64(%rdi), %zmm1
; AVX512-FCP-NEXT: vmovdqa64 128(%rdi), %zmm2
; AVX512-FCP-NEXT: vmovdqa64 192(%rdi), %zmm3
; AVX512-FCP-NEXT: vmovdqa64 960(%rdi), %zmm4
; AVX512-FCP-NEXT: vmovdqa64 896(%rdi), %zmm5
; AVX512-FCP-NEXT: vmovdqa64 576(%rdi), %zmm6
; AVX512-FCP-NEXT: vmovdqa64 512(%rdi), %zmm7
; AVX512-FCP-NEXT: vmovdqa64 704(%rdi), %zmm8
; AVX512-FCP-NEXT: vmovdqa64 640(%rdi), %zmm9
; AVX512-FCP-NEXT: vmovdqa64 320(%rdi), %zmm10
; AVX512-FCP-NEXT: vmovdqa64 256(%rdi), %zmm11
; AVX512-FCP-NEXT: vmovdqa64 448(%rdi), %zmm12
; AVX512-FCP-NEXT: vmovdqa64 384(%rdi), %zmm13
; AVX512-FCP-NEXT: vmovdqa64 832(%rdi), %zmm14
; AVX512-FCP-NEXT: vmovdqa64 768(%rdi), %zmm15
; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm16 = [0,2,4,6,8,10,12,14]
; AVX512-FCP-NEXT: vmovdqa64 %zmm15, %zmm17
; AVX512-FCP-NEXT: vpermt2q %zmm14, %zmm16, %zmm17
; AVX512-FCP-NEXT: vmovdqa64 %zmm13, %zmm18
; AVX512-FCP-NEXT: vpermt2q %zmm12, %zmm16, %zmm18
; AVX512-FCP-NEXT: vmovdqa64 %zmm11, %zmm19
; AVX512-FCP-NEXT: vpermt2q %zmm10, %zmm16, %zmm19
; AVX512-FCP-NEXT: vmovdqa64 %zmm9, %zmm20
; AVX512-FCP-NEXT: vpermt2q %zmm8, %zmm16, %zmm20
; AVX512-FCP-NEXT: vmovdqa64 %zmm7, %zmm21
; AVX512-FCP-NEXT: vpermt2q %zmm6, %zmm16, %zmm21
; AVX512-FCP-NEXT: vmovdqa64 %zmm5, %zmm22
; AVX512-FCP-NEXT: vpermt2q %zmm4, %zmm16, %zmm22
; AVX512-FCP-NEXT: vmovdqa64 %zmm2, %zmm23
; AVX512-FCP-NEXT: vpermt2q %zmm3, %zmm16, %zmm23
; AVX512-FCP-NEXT: vpermi2q %zmm1, %zmm0, %zmm16
; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm24 = [1,3,5,7,9,11,13,15]
; AVX512-FCP-NEXT: vpermt2q %zmm12, %zmm24, %zmm13
; AVX512-FCP-NEXT: vpermt2q %zmm10, %zmm24, %zmm11
; AVX512-FCP-NEXT: vpermt2q %zmm8, %zmm24, %zmm9
; AVX512-FCP-NEXT: vpermt2q %zmm6, %zmm24, %zmm7
; AVX512-FCP-NEXT: vpermt2q %zmm4, %zmm24, %zmm5
; AVX512-FCP-NEXT: vpermt2q %zmm14, %zmm24, %zmm15
; AVX512-FCP-NEXT: vpermt2q %zmm3, %zmm24, %zmm2
; AVX512-FCP-NEXT: vpermt2q %zmm1, %zmm24, %zmm0
; AVX512-FCP-NEXT: vmovdqa64 %zmm22, 448(%rsi)
; AVX512-FCP-NEXT: vmovdqa64 %zmm21, 256(%rsi)
; AVX512-FCP-NEXT: vmovdqa64 %zmm20, 320(%rsi)
; AVX512-FCP-NEXT: vmovdqa64 %zmm19, 128(%rsi)
; AVX512-FCP-NEXT: vmovdqa64 %zmm18, 192(%rsi)
; AVX512-FCP-NEXT: vmovdqa64 %zmm16, (%rsi)
; AVX512-FCP-NEXT: vmovdqa64 %zmm23, 64(%rsi)
; AVX512-FCP-NEXT: vmovdqa64 %zmm17, 384(%rsi)
; AVX512-FCP-NEXT: vmovdqa64 %zmm15, 384(%rdx)
; AVX512-FCP-NEXT: vmovdqa64 %zmm5, 448(%rdx)
; AVX512-FCP-NEXT: vmovdqa64 %zmm7, 256(%rdx)
; AVX512-FCP-NEXT: vmovdqa64 %zmm9, 320(%rdx)
; AVX512-FCP-NEXT: vmovdqa64 %zmm11, 128(%rdx)
; AVX512-FCP-NEXT: vmovdqa64 %zmm13, 192(%rdx)
; AVX512-FCP-NEXT: vmovdqa64 %zmm0, (%rdx)
; AVX512-FCP-NEXT: vmovdqa64 %zmm2, 64(%rdx)
; AVX512-FCP-NEXT: vzeroupper
; AVX512-FCP-NEXT: retq
;
; AVX512DQ-LABEL: load_i64_stride2_vf64:
; AVX512DQ: # %bb.0:
; AVX512DQ-NEXT: vmovdqa64 (%rdi), %zmm0
; AVX512DQ-NEXT: vmovdqa64 64(%rdi), %zmm1
; AVX512DQ-NEXT: vmovdqa64 128(%rdi), %zmm2
; AVX512DQ-NEXT: vmovdqa64 192(%rdi), %zmm3
; AVX512DQ-NEXT: vmovdqa64 960(%rdi), %zmm4
; AVX512DQ-NEXT: vmovdqa64 896(%rdi), %zmm5
; AVX512DQ-NEXT: vmovdqa64 576(%rdi), %zmm6
; AVX512DQ-NEXT: vmovdqa64 512(%rdi), %zmm7
; AVX512DQ-NEXT: vmovdqa64 704(%rdi), %zmm8
; AVX512DQ-NEXT: vmovdqa64 640(%rdi), %zmm9
; AVX512DQ-NEXT: vmovdqa64 320(%rdi), %zmm10
; AVX512DQ-NEXT: vmovdqa64 256(%rdi), %zmm11
; AVX512DQ-NEXT: vmovdqa64 448(%rdi), %zmm12
; AVX512DQ-NEXT: vmovdqa64 384(%rdi), %zmm13
; AVX512DQ-NEXT: vmovdqa64 832(%rdi), %zmm14
; AVX512DQ-NEXT: vmovdqa64 768(%rdi), %zmm15
; AVX512DQ-NEXT: vpmovsxbq {{.*#+}} zmm16 = [0,2,4,6,8,10,12,14]
; AVX512DQ-NEXT: vmovdqa64 %zmm15, %zmm17
; AVX512DQ-NEXT: vpermt2q %zmm14, %zmm16, %zmm17
; AVX512DQ-NEXT: vmovdqa64 %zmm13, %zmm18
; AVX512DQ-NEXT: vpermt2q %zmm12, %zmm16, %zmm18
; AVX512DQ-NEXT: vmovdqa64 %zmm11, %zmm19
; AVX512DQ-NEXT: vpermt2q %zmm10, %zmm16, %zmm19
; AVX512DQ-NEXT: vmovdqa64 %zmm9, %zmm20
; AVX512DQ-NEXT: vpermt2q %zmm8, %zmm16, %zmm20
; AVX512DQ-NEXT: vmovdqa64 %zmm7, %zmm21
; AVX512DQ-NEXT: vpermt2q %zmm6, %zmm16, %zmm21
; AVX512DQ-NEXT: vmovdqa64 %zmm5, %zmm22
; AVX512DQ-NEXT: vpermt2q %zmm4, %zmm16, %zmm22
; AVX512DQ-NEXT: vmovdqa64 %zmm2, %zmm23
; AVX512DQ-NEXT: vpermt2q %zmm3, %zmm16, %zmm23
; AVX512DQ-NEXT: vpermi2q %zmm1, %zmm0, %zmm16
; AVX512DQ-NEXT: vpmovsxbq {{.*#+}} zmm24 = [1,3,5,7,9,11,13,15]
; AVX512DQ-NEXT: vpermt2q %zmm12, %zmm24, %zmm13
; AVX512DQ-NEXT: vpermt2q %zmm10, %zmm24, %zmm11
; AVX512DQ-NEXT: vpermt2q %zmm8, %zmm24, %zmm9
; AVX512DQ-NEXT: vpermt2q %zmm6, %zmm24, %zmm7
; AVX512DQ-NEXT: vpermt2q %zmm4, %zmm24, %zmm5
; AVX512DQ-NEXT: vpermt2q %zmm14, %zmm24, %zmm15
; AVX512DQ-NEXT: vpermt2q %zmm3, %zmm24, %zmm2
; AVX512DQ-NEXT: vpermt2q %zmm1, %zmm24, %zmm0
; AVX512DQ-NEXT: vmovdqa64 %zmm22, 448(%rsi)
; AVX512DQ-NEXT: vmovdqa64 %zmm21, 256(%rsi)
; AVX512DQ-NEXT: vmovdqa64 %zmm20, 320(%rsi)
; AVX512DQ-NEXT: vmovdqa64 %zmm19, 128(%rsi)
; AVX512DQ-NEXT: vmovdqa64 %zmm18, 192(%rsi)
; AVX512DQ-NEXT: vmovdqa64 %zmm16, (%rsi)
; AVX512DQ-NEXT: vmovdqa64 %zmm23, 64(%rsi)
; AVX512DQ-NEXT: vmovdqa64 %zmm17, 384(%rsi)
; AVX512DQ-NEXT: vmovdqa64 %zmm15, 384(%rdx)
; AVX512DQ-NEXT: vmovdqa64 %zmm5, 448(%rdx)
; AVX512DQ-NEXT: vmovdqa64 %zmm7, 256(%rdx)
; AVX512DQ-NEXT: vmovdqa64 %zmm9, 320(%rdx)
; AVX512DQ-NEXT: vmovdqa64 %zmm11, 128(%rdx)
; AVX512DQ-NEXT: vmovdqa64 %zmm13, 192(%rdx)
; AVX512DQ-NEXT: vmovdqa64 %zmm0, (%rdx)
; AVX512DQ-NEXT: vmovdqa64 %zmm2, 64(%rdx)
; AVX512DQ-NEXT: vzeroupper
; AVX512DQ-NEXT: retq
;
; AVX512DQ-FCP-LABEL: load_i64_stride2_vf64:
; AVX512DQ-FCP: # %bb.0:
; AVX512DQ-FCP-NEXT: vmovdqa64 (%rdi), %zmm0
; AVX512DQ-FCP-NEXT: vmovdqa64 64(%rdi), %zmm1
; AVX512DQ-FCP-NEXT: vmovdqa64 128(%rdi), %zmm2
; AVX512DQ-FCP-NEXT: vmovdqa64 192(%rdi), %zmm3
; AVX512DQ-FCP-NEXT: vmovdqa64 960(%rdi), %zmm4
; AVX512DQ-FCP-NEXT: vmovdqa64 896(%rdi), %zmm5
; AVX512DQ-FCP-NEXT: vmovdqa64 576(%rdi), %zmm6
; AVX512DQ-FCP-NEXT: vmovdqa64 512(%rdi), %zmm7
; AVX512DQ-FCP-NEXT: vmovdqa64 704(%rdi), %zmm8
; AVX512DQ-FCP-NEXT: vmovdqa64 640(%rdi), %zmm9
; AVX512DQ-FCP-NEXT: vmovdqa64 320(%rdi), %zmm10
; AVX512DQ-FCP-NEXT: vmovdqa64 256(%rdi), %zmm11
; AVX512DQ-FCP-NEXT: vmovdqa64 448(%rdi), %zmm12
; AVX512DQ-FCP-NEXT: vmovdqa64 384(%rdi), %zmm13
; AVX512DQ-FCP-NEXT: vmovdqa64 832(%rdi), %zmm14
; AVX512DQ-FCP-NEXT: vmovdqa64 768(%rdi), %zmm15
; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm16 = [0,2,4,6,8,10,12,14]
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm15, %zmm17
; AVX512DQ-FCP-NEXT: vpermt2q %zmm14, %zmm16, %zmm17
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm13, %zmm18
; AVX512DQ-FCP-NEXT: vpermt2q %zmm12, %zmm16, %zmm18
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm11, %zmm19
; AVX512DQ-FCP-NEXT: vpermt2q %zmm10, %zmm16, %zmm19
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm9, %zmm20
; AVX512DQ-FCP-NEXT: vpermt2q %zmm8, %zmm16, %zmm20
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm7, %zmm21
; AVX512DQ-FCP-NEXT: vpermt2q %zmm6, %zmm16, %zmm21
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm5, %zmm22
; AVX512DQ-FCP-NEXT: vpermt2q %zmm4, %zmm16, %zmm22
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm2, %zmm23
; AVX512DQ-FCP-NEXT: vpermt2q %zmm3, %zmm16, %zmm23
; AVX512DQ-FCP-NEXT: vpermi2q %zmm1, %zmm0, %zmm16
; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm24 = [1,3,5,7,9,11,13,15]
; AVX512DQ-FCP-NEXT: vpermt2q %zmm12, %zmm24, %zmm13
; AVX512DQ-FCP-NEXT: vpermt2q %zmm10, %zmm24, %zmm11
; AVX512DQ-FCP-NEXT: vpermt2q %zmm8, %zmm24, %zmm9
; AVX512DQ-FCP-NEXT: vpermt2q %zmm6, %zmm24, %zmm7
; AVX512DQ-FCP-NEXT: vpermt2q %zmm4, %zmm24, %zmm5
; AVX512DQ-FCP-NEXT: vpermt2q %zmm14, %zmm24, %zmm15
; AVX512DQ-FCP-NEXT: vpermt2q %zmm3, %zmm24, %zmm2
; AVX512DQ-FCP-NEXT: vpermt2q %zmm1, %zmm24, %zmm0
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm22, 448(%rsi)
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm21, 256(%rsi)
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm20, 320(%rsi)
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm19, 128(%rsi)
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm18, 192(%rsi)
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm16, (%rsi)
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm23, 64(%rsi)
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm17, 384(%rsi)
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm15, 384(%rdx)
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm5, 448(%rdx)
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm7, 256(%rdx)
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm9, 320(%rdx)
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm11, 128(%rdx)
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm13, 192(%rdx)
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm0, (%rdx)
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm2, 64(%rdx)
; AVX512DQ-FCP-NEXT: vzeroupper
; AVX512DQ-FCP-NEXT: retq
;
; AVX512BW-LABEL: load_i64_stride2_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 960(%rdi), %zmm4
; AVX512BW-NEXT: vmovdqa64 896(%rdi), %zmm5
; AVX512BW-NEXT: vmovdqa64 576(%rdi), %zmm6
; AVX512BW-NEXT: vmovdqa64 512(%rdi), %zmm7
; AVX512BW-NEXT: vmovdqa64 704(%rdi), %zmm8
; AVX512BW-NEXT: vmovdqa64 640(%rdi), %zmm9
; AVX512BW-NEXT: vmovdqa64 320(%rdi), %zmm10
; AVX512BW-NEXT: vmovdqa64 256(%rdi), %zmm11
; AVX512BW-NEXT: vmovdqa64 448(%rdi), %zmm12
; AVX512BW-NEXT: vmovdqa64 384(%rdi), %zmm13
; AVX512BW-NEXT: vmovdqa64 832(%rdi), %zmm14
; AVX512BW-NEXT: vmovdqa64 768(%rdi), %zmm15
; AVX512BW-NEXT: vpmovsxbq {{.*#+}} zmm16 = [0,2,4,6,8,10,12,14]
; AVX512BW-NEXT: vmovdqa64 %zmm15, %zmm17
; AVX512BW-NEXT: vpermt2q %zmm14, %zmm16, %zmm17
; AVX512BW-NEXT: vmovdqa64 %zmm13, %zmm18
; AVX512BW-NEXT: vpermt2q %zmm12, %zmm16, %zmm18
; AVX512BW-NEXT: vmovdqa64 %zmm11, %zmm19
; AVX512BW-NEXT: vpermt2q %zmm10, %zmm16, %zmm19
; AVX512BW-NEXT: vmovdqa64 %zmm9, %zmm20
; AVX512BW-NEXT: vpermt2q %zmm8, %zmm16, %zmm20
; AVX512BW-NEXT: vmovdqa64 %zmm7, %zmm21
; AVX512BW-NEXT: vpermt2q %zmm6, %zmm16, %zmm21
; AVX512BW-NEXT: vmovdqa64 %zmm5, %zmm22
; AVX512BW-NEXT: vpermt2q %zmm4, %zmm16, %zmm22
; AVX512BW-NEXT: vmovdqa64 %zmm2, %zmm23
; AVX512BW-NEXT: vpermt2q %zmm3, %zmm16, %zmm23
; AVX512BW-NEXT: vpermi2q %zmm1, %zmm0, %zmm16
; AVX512BW-NEXT: vpmovsxbq {{.*#+}} zmm24 = [1,3,5,7,9,11,13,15]
; AVX512BW-NEXT: vpermt2q %zmm12, %zmm24, %zmm13
; AVX512BW-NEXT: vpermt2q %zmm10, %zmm24, %zmm11
; AVX512BW-NEXT: vpermt2q %zmm8, %zmm24, %zmm9
; AVX512BW-NEXT: vpermt2q %zmm6, %zmm24, %zmm7
; AVX512BW-NEXT: vpermt2q %zmm4, %zmm24, %zmm5
; AVX512BW-NEXT: vpermt2q %zmm14, %zmm24, %zmm15
; AVX512BW-NEXT: vpermt2q %zmm3, %zmm24, %zmm2
; AVX512BW-NEXT: vpermt2q %zmm1, %zmm24, %zmm0
; AVX512BW-NEXT: vmovdqa64 %zmm22, 448(%rsi)
; AVX512BW-NEXT: vmovdqa64 %zmm21, 256(%rsi)
; AVX512BW-NEXT: vmovdqa64 %zmm20, 320(%rsi)
; AVX512BW-NEXT: vmovdqa64 %zmm19, 128(%rsi)
; AVX512BW-NEXT: vmovdqa64 %zmm18, 192(%rsi)
; AVX512BW-NEXT: vmovdqa64 %zmm16, (%rsi)
; AVX512BW-NEXT: vmovdqa64 %zmm23, 64(%rsi)
; AVX512BW-NEXT: vmovdqa64 %zmm17, 384(%rsi)
; AVX512BW-NEXT: vmovdqa64 %zmm15, 384(%rdx)
; AVX512BW-NEXT: vmovdqa64 %zmm5, 448(%rdx)
; AVX512BW-NEXT: vmovdqa64 %zmm7, 256(%rdx)
; AVX512BW-NEXT: vmovdqa64 %zmm9, 320(%rdx)
; AVX512BW-NEXT: vmovdqa64 %zmm11, 128(%rdx)
; AVX512BW-NEXT: vmovdqa64 %zmm13, 192(%rdx)
; AVX512BW-NEXT: vmovdqa64 %zmm0, (%rdx)
; AVX512BW-NEXT: vmovdqa64 %zmm2, 64(%rdx)
; AVX512BW-NEXT: vzeroupper
; AVX512BW-NEXT: retq
;
; AVX512BW-FCP-LABEL: load_i64_stride2_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 960(%rdi), %zmm4
; AVX512BW-FCP-NEXT: vmovdqa64 896(%rdi), %zmm5
; AVX512BW-FCP-NEXT: vmovdqa64 576(%rdi), %zmm6
; AVX512BW-FCP-NEXT: vmovdqa64 512(%rdi), %zmm7
; AVX512BW-FCP-NEXT: vmovdqa64 704(%rdi), %zmm8
; AVX512BW-FCP-NEXT: vmovdqa64 640(%rdi), %zmm9
; AVX512BW-FCP-NEXT: vmovdqa64 320(%rdi), %zmm10
; AVX512BW-FCP-NEXT: vmovdqa64 256(%rdi), %zmm11
; AVX512BW-FCP-NEXT: vmovdqa64 448(%rdi), %zmm12
; AVX512BW-FCP-NEXT: vmovdqa64 384(%rdi), %zmm13
; AVX512BW-FCP-NEXT: vmovdqa64 832(%rdi), %zmm14
; AVX512BW-FCP-NEXT: vmovdqa64 768(%rdi), %zmm15
; AVX512BW-FCP-NEXT: vpmovsxbq {{.*#+}} zmm16 = [0,2,4,6,8,10,12,14]
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm15, %zmm17
; AVX512BW-FCP-NEXT: vpermt2q %zmm14, %zmm16, %zmm17
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm13, %zmm18
; AVX512BW-FCP-NEXT: vpermt2q %zmm12, %zmm16, %zmm18
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm11, %zmm19
; AVX512BW-FCP-NEXT: vpermt2q %zmm10, %zmm16, %zmm19
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm9, %zmm20
; AVX512BW-FCP-NEXT: vpermt2q %zmm8, %zmm16, %zmm20
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm7, %zmm21
; AVX512BW-FCP-NEXT: vpermt2q %zmm6, %zmm16, %zmm21
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm5, %zmm22
; AVX512BW-FCP-NEXT: vpermt2q %zmm4, %zmm16, %zmm22
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm2, %zmm23
; AVX512BW-FCP-NEXT: vpermt2q %zmm3, %zmm16, %zmm23
; AVX512BW-FCP-NEXT: vpermi2q %zmm1, %zmm0, %zmm16
; AVX512BW-FCP-NEXT: vpmovsxbq {{.*#+}} zmm24 = [1,3,5,7,9,11,13,15]
; AVX512BW-FCP-NEXT: vpermt2q %zmm12, %zmm24, %zmm13
; AVX512BW-FCP-NEXT: vpermt2q %zmm10, %zmm24, %zmm11
; AVX512BW-FCP-NEXT: vpermt2q %zmm8, %zmm24, %zmm9
; AVX512BW-FCP-NEXT: vpermt2q %zmm6, %zmm24, %zmm7
; AVX512BW-FCP-NEXT: vpermt2q %zmm4, %zmm24, %zmm5
; AVX512BW-FCP-NEXT: vpermt2q %zmm14, %zmm24, %zmm15
; AVX512BW-FCP-NEXT: vpermt2q %zmm3, %zmm24, %zmm2
; AVX512BW-FCP-NEXT: vpermt2q %zmm1, %zmm24, %zmm0
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm22, 448(%rsi)
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm21, 256(%rsi)
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm20, 320(%rsi)
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm19, 128(%rsi)
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm18, 192(%rsi)
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm16, (%rsi)
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm23, 64(%rsi)
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm17, 384(%rsi)
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm15, 384(%rdx)
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm5, 448(%rdx)
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm7, 256(%rdx)
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm9, 320(%rdx)
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm11, 128(%rdx)
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm13, 192(%rdx)
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm0, (%rdx)
; AVX512BW-FCP-NEXT: vmovdqa64 %zmm2, 64(%rdx)
; AVX512BW-FCP-NEXT: vzeroupper
; AVX512BW-FCP-NEXT: retq
;
; AVX512DQ-BW-LABEL: load_i64_stride2_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 960(%rdi), %zmm4
; AVX512DQ-BW-NEXT: vmovdqa64 896(%rdi), %zmm5
; AVX512DQ-BW-NEXT: vmovdqa64 576(%rdi), %zmm6
; AVX512DQ-BW-NEXT: vmovdqa64 512(%rdi), %zmm7
; AVX512DQ-BW-NEXT: vmovdqa64 704(%rdi), %zmm8
; AVX512DQ-BW-NEXT: vmovdqa64 640(%rdi), %zmm9
; AVX512DQ-BW-NEXT: vmovdqa64 320(%rdi), %zmm10
; AVX512DQ-BW-NEXT: vmovdqa64 256(%rdi), %zmm11
; AVX512DQ-BW-NEXT: vmovdqa64 448(%rdi), %zmm12
; AVX512DQ-BW-NEXT: vmovdqa64 384(%rdi), %zmm13
; AVX512DQ-BW-NEXT: vmovdqa64 832(%rdi), %zmm14
; AVX512DQ-BW-NEXT: vmovdqa64 768(%rdi), %zmm15
; AVX512DQ-BW-NEXT: vpmovsxbq {{.*#+}} zmm16 = [0,2,4,6,8,10,12,14]
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm15, %zmm17
; AVX512DQ-BW-NEXT: vpermt2q %zmm14, %zmm16, %zmm17
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm13, %zmm18
; AVX512DQ-BW-NEXT: vpermt2q %zmm12, %zmm16, %zmm18
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm11, %zmm19
; AVX512DQ-BW-NEXT: vpermt2q %zmm10, %zmm16, %zmm19
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm9, %zmm20
; AVX512DQ-BW-NEXT: vpermt2q %zmm8, %zmm16, %zmm20
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm7, %zmm21
; AVX512DQ-BW-NEXT: vpermt2q %zmm6, %zmm16, %zmm21
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm5, %zmm22
; AVX512DQ-BW-NEXT: vpermt2q %zmm4, %zmm16, %zmm22
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm2, %zmm23
; AVX512DQ-BW-NEXT: vpermt2q %zmm3, %zmm16, %zmm23
; AVX512DQ-BW-NEXT: vpermi2q %zmm1, %zmm0, %zmm16
; AVX512DQ-BW-NEXT: vpmovsxbq {{.*#+}} zmm24 = [1,3,5,7,9,11,13,15]
; AVX512DQ-BW-NEXT: vpermt2q %zmm12, %zmm24, %zmm13
; AVX512DQ-BW-NEXT: vpermt2q %zmm10, %zmm24, %zmm11
; AVX512DQ-BW-NEXT: vpermt2q %zmm8, %zmm24, %zmm9
; AVX512DQ-BW-NEXT: vpermt2q %zmm6, %zmm24, %zmm7
; AVX512DQ-BW-NEXT: vpermt2q %zmm4, %zmm24, %zmm5
; AVX512DQ-BW-NEXT: vpermt2q %zmm14, %zmm24, %zmm15
; AVX512DQ-BW-NEXT: vpermt2q %zmm3, %zmm24, %zmm2
; AVX512DQ-BW-NEXT: vpermt2q %zmm1, %zmm24, %zmm0
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm22, 448(%rsi)
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm21, 256(%rsi)
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm20, 320(%rsi)
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm19, 128(%rsi)
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm18, 192(%rsi)
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm16, (%rsi)
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm23, 64(%rsi)
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm17, 384(%rsi)
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm15, 384(%rdx)
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm5, 448(%rdx)
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm7, 256(%rdx)
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm9, 320(%rdx)
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm11, 128(%rdx)
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm13, 192(%rdx)
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm0, (%rdx)
; AVX512DQ-BW-NEXT: vmovdqa64 %zmm2, 64(%rdx)
; AVX512DQ-BW-NEXT: vzeroupper
; AVX512DQ-BW-NEXT: retq
;
; AVX512DQ-BW-FCP-LABEL: load_i64_stride2_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 960(%rdi), %zmm4
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 896(%rdi), %zmm5
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 576(%rdi), %zmm6
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 512(%rdi), %zmm7
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 704(%rdi), %zmm8
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 640(%rdi), %zmm9
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 320(%rdi), %zmm10
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 256(%rdi), %zmm11
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 448(%rdi), %zmm12
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 384(%rdi), %zmm13
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 832(%rdi), %zmm14
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 768(%rdi), %zmm15
; AVX512DQ-BW-FCP-NEXT: vpmovsxbq {{.*#+}} zmm16 = [0,2,4,6,8,10,12,14]
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm15, %zmm17
; AVX512DQ-BW-FCP-NEXT: vpermt2q %zmm14, %zmm16, %zmm17
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm13, %zmm18
; AVX512DQ-BW-FCP-NEXT: vpermt2q %zmm12, %zmm16, %zmm18
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm11, %zmm19
; AVX512DQ-BW-FCP-NEXT: vpermt2q %zmm10, %zmm16, %zmm19
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm9, %zmm20
; AVX512DQ-BW-FCP-NEXT: vpermt2q %zmm8, %zmm16, %zmm20
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm7, %zmm21
; AVX512DQ-BW-FCP-NEXT: vpermt2q %zmm6, %zmm16, %zmm21
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm5, %zmm22
; AVX512DQ-BW-FCP-NEXT: vpermt2q %zmm4, %zmm16, %zmm22
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm2, %zmm23
; AVX512DQ-BW-FCP-NEXT: vpermt2q %zmm3, %zmm16, %zmm23
; AVX512DQ-BW-FCP-NEXT: vpermi2q %zmm1, %zmm0, %zmm16
; AVX512DQ-BW-FCP-NEXT: vpmovsxbq {{.*#+}} zmm24 = [1,3,5,7,9,11,13,15]
; AVX512DQ-BW-FCP-NEXT: vpermt2q %zmm12, %zmm24, %zmm13
; AVX512DQ-BW-FCP-NEXT: vpermt2q %zmm10, %zmm24, %zmm11
; AVX512DQ-BW-FCP-NEXT: vpermt2q %zmm8, %zmm24, %zmm9
; AVX512DQ-BW-FCP-NEXT: vpermt2q %zmm6, %zmm24, %zmm7
; AVX512DQ-BW-FCP-NEXT: vpermt2q %zmm4, %zmm24, %zmm5
; AVX512DQ-BW-FCP-NEXT: vpermt2q %zmm14, %zmm24, %zmm15
; AVX512DQ-BW-FCP-NEXT: vpermt2q %zmm3, %zmm24, %zmm2
; AVX512DQ-BW-FCP-NEXT: vpermt2q %zmm1, %zmm24, %zmm0
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm22, 448(%rsi)
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm21, 256(%rsi)
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm20, 320(%rsi)
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm19, 128(%rsi)
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm18, 192(%rsi)
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm16, (%rsi)
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm23, 64(%rsi)
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm17, 384(%rsi)
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm15, 384(%rdx)
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm5, 448(%rdx)
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm7, 256(%rdx)
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm9, 320(%rdx)
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm11, 128(%rdx)
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm13, 192(%rdx)
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm0, (%rdx)
; AVX512DQ-BW-FCP-NEXT: vmovdqa64 %zmm2, 64(%rdx)
; AVX512DQ-BW-FCP-NEXT: vzeroupper
; AVX512DQ-BW-FCP-NEXT: retq
%wide.vec = load <128 x i64>, ptr %in.vec, align 64
%strided.vec0 = shufflevector <128 x i64> %wide.vec, <128 x i64> poison, <64 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30, i32 32, i32 34, i32 36, i32 38, i32 40, i32 42, i32 44, i32 46, i32 48, i32 50, i32 52, i32 54, i32 56, i32 58, i32 60, i32 62, i32 64, i32 66, i32 68, i32 70, i32 72, i32 74, i32 76, i32 78, i32 80, i32 82, i32 84, i32 86, i32 88, i32 90, i32 92, i32 94, i32 96, i32 98, i32 100, i32 102, i32 104, i32 106, i32 108, i32 110, i32 112, i32 114, i32 116, i32 118, i32 120, i32 122, i32 124, i32 126>
%strided.vec1 = shufflevector <128 x i64> %wide.vec, <128 x i64> poison, <64 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31, i32 33, i32 35, i32 37, i32 39, i32 41, i32 43, i32 45, i32 47, i32 49, i32 51, i32 53, i32 55, i32 57, i32 59, i32 61, i32 63, i32 65, i32 67, i32 69, i32 71, i32 73, i32 75, i32 77, i32 79, i32 81, i32 83, i32 85, i32 87, i32 89, i32 91, i32 93, i32 95, i32 97, i32 99, i32 101, i32 103, i32 105, i32 107, i32 109, i32 111, i32 113, i32 115, i32 117, i32 119, i32 121, i32 123, i32 125, i32 127>
store <64 x i64> %strided.vec0, ptr %out.vec0, align 64
store <64 x i64> %strided.vec1, ptr %out.vec1, align 64
ret void
}