; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=x86-64-v2 | FileCheck %s --check-prefixes=SSE
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=sandybridge | FileCheck %s --check-prefixes=AVX,AVX1
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=x86-64-v3 | FileCheck %s --check-prefixes=AVX,AVX2
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=x86-64-v4 | FileCheck %s --check-prefixes=AVX,AVX512

define <4 x double> @concat_round_v4f64_v2f64(<2 x double> %a0, <2 x double> %a1) {
; SSE-LABEL: concat_round_v4f64_v2f64:
; SSE:       # %bb.0:
; SSE-NEXT:    movapd {{.*#+}} xmm2 = [-0.0E+0,-0.0E+0]
; SSE-NEXT:    movapd %xmm0, %xmm3
; SSE-NEXT:    andpd %xmm2, %xmm3
; SSE-NEXT:    movapd {{.*#+}} xmm4 = [4.9999999999999994E-1,4.9999999999999994E-1]
; SSE-NEXT:    orpd %xmm4, %xmm3
; SSE-NEXT:    addpd %xmm0, %xmm3
; SSE-NEXT:    roundpd $11, %xmm3, %xmm0
; SSE-NEXT:    andpd %xmm1, %xmm2
; SSE-NEXT:    orpd %xmm4, %xmm2
; SSE-NEXT:    addpd %xmm1, %xmm2
; SSE-NEXT:    roundpd $11, %xmm2, %xmm1
; SSE-NEXT:    retq
;
; AVX1-LABEL: concat_round_v4f64_v2f64:
; AVX1:       # %bb.0:
; AVX1-NEXT:    # kill: def $xmm0 killed $xmm0 def $ymm0
; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
; AVX1-NEXT:    vandpd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm1
; AVX1-NEXT:    vorpd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm1
; AVX1-NEXT:    vaddpd %ymm1, %ymm0, %ymm0
; AVX1-NEXT:    vroundpd $11, %ymm0, %ymm0
; AVX1-NEXT:    retq
;
; AVX2-LABEL: concat_round_v4f64_v2f64:
; AVX2:       # %bb.0:
; AVX2-NEXT:    # kill: def $xmm0 killed $xmm0 def $ymm0
; AVX2-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
; AVX2-NEXT:    vbroadcastsd {{.*#+}} ymm1 = [-0.0E+0,-0.0E+0,-0.0E+0,-0.0E+0]
; AVX2-NEXT:    vandpd %ymm1, %ymm0, %ymm1
; AVX2-NEXT:    vbroadcastsd {{.*#+}} ymm2 = [4.9999999999999994E-1,4.9999999999999994E-1,4.9999999999999994E-1,4.9999999999999994E-1]
; AVX2-NEXT:    vorpd %ymm2, %ymm1, %ymm1
; AVX2-NEXT:    vaddpd %ymm1, %ymm0, %ymm0
; AVX2-NEXT:    vroundpd $11, %ymm0, %ymm0
; AVX2-NEXT:    retq
;
; AVX512-LABEL: concat_round_v4f64_v2f64:
; AVX512:       # %bb.0:
; AVX512-NEXT:    # kill: def $xmm0 killed $xmm0 def $ymm0
; AVX512-NEXT:    vpbroadcastq {{.*#+}} ymm2 = [4.9999999999999994E-1,4.9999999999999994E-1,4.9999999999999994E-1,4.9999999999999994E-1]
; AVX512-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
; AVX512-NEXT:    vpternlogq {{.*#+}} ymm2 = ymm2 | (ymm0 & m64bcst)
; AVX512-NEXT:    vaddpd %ymm2, %ymm0, %ymm0
; AVX512-NEXT:    vroundpd $11, %ymm0, %ymm0
; AVX512-NEXT:    retq
  %v0 = call <2 x double> @llvm.round.v2f64(<2 x double> %a0)
  %v1 = call <2 x double> @llvm.round.v2f64(<2 x double> %a1)
  %res  = shufflevector <2 x double> %v0, <2 x double> %v1, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  ret <4 x double> %res
}

define <8 x float> @concat_round_v8f32_v4f32(<4 x float> %a0, <4 x float> %a1) {
; SSE-LABEL: concat_round_v8f32_v4f32:
; SSE:       # %bb.0:
; SSE-NEXT:    movaps {{.*#+}} xmm2 = [-0.0E+0,-0.0E+0,-0.0E+0,-0.0E+0]
; SSE-NEXT:    movaps %xmm0, %xmm3
; SSE-NEXT:    andps %xmm2, %xmm3
; SSE-NEXT:    movaps {{.*#+}} xmm4 = [4.9999997E-1,4.9999997E-1,4.9999997E-1,4.9999997E-1]
; SSE-NEXT:    orps %xmm4, %xmm3
; SSE-NEXT:    addps %xmm0, %xmm3
; SSE-NEXT:    roundps $11, %xmm3, %xmm0
; SSE-NEXT:    andps %xmm1, %xmm2
; SSE-NEXT:    orps %xmm4, %xmm2
; SSE-NEXT:    addps %xmm1, %xmm2
; SSE-NEXT:    roundps $11, %xmm2, %xmm1
; SSE-NEXT:    retq
;
; AVX1-LABEL: concat_round_v8f32_v4f32:
; AVX1:       # %bb.0:
; AVX1-NEXT:    # kill: def $xmm0 killed $xmm0 def $ymm0
; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
; AVX1-NEXT:    vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm1
; AVX1-NEXT:    vorps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm1
; AVX1-NEXT:    vaddps %ymm1, %ymm0, %ymm0
; AVX1-NEXT:    vroundps $11, %ymm0, %ymm0
; AVX1-NEXT:    retq
;
; AVX2-LABEL: concat_round_v8f32_v4f32:
; AVX2:       # %bb.0:
; AVX2-NEXT:    # kill: def $xmm0 killed $xmm0 def $ymm0
; AVX2-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
; AVX2-NEXT:    vbroadcastss {{.*#+}} ymm1 = [-0.0E+0,-0.0E+0,-0.0E+0,-0.0E+0,-0.0E+0,-0.0E+0,-0.0E+0,-0.0E+0]
; AVX2-NEXT:    vandps %ymm1, %ymm0, %ymm1
; AVX2-NEXT:    vbroadcastss {{.*#+}} ymm2 = [4.9999997E-1,4.9999997E-1,4.9999997E-1,4.9999997E-1,4.9999997E-1,4.9999997E-1,4.9999997E-1,4.9999997E-1]
; AVX2-NEXT:    vorps %ymm2, %ymm1, %ymm1
; AVX2-NEXT:    vaddps %ymm1, %ymm0, %ymm0
; AVX2-NEXT:    vroundps $11, %ymm0, %ymm0
; AVX2-NEXT:    retq
;
; AVX512-LABEL: concat_round_v8f32_v4f32:
; AVX512:       # %bb.0:
; AVX512-NEXT:    # kill: def $xmm0 killed $xmm0 def $ymm0
; AVX512-NEXT:    vpbroadcastd {{.*#+}} ymm2 = [4.9999997E-1,4.9999997E-1,4.9999997E-1,4.9999997E-1,4.9999997E-1,4.9999997E-1,4.9999997E-1,4.9999997E-1]
; AVX512-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
; AVX512-NEXT:    vpternlogd {{.*#+}} ymm2 = ymm2 | (ymm0 & m32bcst)
; AVX512-NEXT:    vaddps %ymm2, %ymm0, %ymm0
; AVX512-NEXT:    vroundps $11, %ymm0, %ymm0
; AVX512-NEXT:    retq
  %v0 = call <4 x float> @llvm.round.v4f32(<4 x float> %a0)
  %v1 = call <4 x float> @llvm.round.v4f32(<4 x float> %a1)
  %res  = shufflevector <4 x float> %v0, <4 x float> %v1, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
  ret <8 x float> %res
}

define <8 x double> @concat_round_v8f64_v2f64(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2, <2 x double> %a3) {
; SSE-LABEL: concat_round_v8f64_v2f64:
; SSE:       # %bb.0:
; SSE-NEXT:    movapd {{.*#+}} xmm4 = [-0.0E+0,-0.0E+0]
; SSE-NEXT:    movapd %xmm0, %xmm5
; SSE-NEXT:    andpd %xmm4, %xmm5
; SSE-NEXT:    movapd {{.*#+}} xmm6 = [4.9999999999999994E-1,4.9999999999999994E-1]
; SSE-NEXT:    orpd %xmm6, %xmm5
; SSE-NEXT:    addpd %xmm0, %xmm5
; SSE-NEXT:    roundpd $11, %xmm5, %xmm0
; SSE-NEXT:    movapd %xmm1, %xmm5
; SSE-NEXT:    andpd %xmm4, %xmm5
; SSE-NEXT:    orpd %xmm6, %xmm5
; SSE-NEXT:    addpd %xmm1, %xmm5
; SSE-NEXT:    roundpd $11, %xmm5, %xmm1
; SSE-NEXT:    movapd %xmm2, %xmm5
; SSE-NEXT:    andpd %xmm4, %xmm5
; SSE-NEXT:    orpd %xmm6, %xmm5
; SSE-NEXT:    addpd %xmm2, %xmm5
; SSE-NEXT:    roundpd $11, %xmm5, %xmm2
; SSE-NEXT:    andpd %xmm3, %xmm4
; SSE-NEXT:    orpd %xmm6, %xmm4
; SSE-NEXT:    addpd %xmm3, %xmm4
; SSE-NEXT:    roundpd $11, %xmm4, %xmm3
; SSE-NEXT:    retq
;
; AVX1-LABEL: concat_round_v8f64_v2f64:
; AVX1:       # %bb.0:
; AVX1-NEXT:    # kill: def $xmm2 killed $xmm2 def $ymm2
; AVX1-NEXT:    # kill: def $xmm0 killed $xmm0 def $ymm0
; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
; AVX1-NEXT:    vmovapd {{.*#+}} ymm1 = [-0.0E+0,-0.0E+0,-0.0E+0,-0.0E+0]
; AVX1-NEXT:    vandpd %ymm1, %ymm0, %ymm4
; AVX1-NEXT:    vmovapd {{.*#+}} ymm5 = [4.9999999999999994E-1,4.9999999999999994E-1,4.9999999999999994E-1,4.9999999999999994E-1]
; AVX1-NEXT:    vorpd %ymm5, %ymm4, %ymm4
; AVX1-NEXT:    vaddpd %ymm4, %ymm0, %ymm0
; AVX1-NEXT:    vroundpd $11, %ymm0, %ymm0
; AVX1-NEXT:    vinsertf128 $1, %xmm3, %ymm2, %ymm2
; AVX1-NEXT:    vandpd %ymm1, %ymm2, %ymm1
; AVX1-NEXT:    vorpd %ymm5, %ymm1, %ymm1
; AVX1-NEXT:    vaddpd %ymm1, %ymm2, %ymm1
; AVX1-NEXT:    vroundpd $11, %ymm1, %ymm1
; AVX1-NEXT:    retq
;
; AVX2-LABEL: concat_round_v8f64_v2f64:
; AVX2:       # %bb.0:
; AVX2-NEXT:    # kill: def $xmm2 killed $xmm2 def $ymm2
; AVX2-NEXT:    # kill: def $xmm0 killed $xmm0 def $ymm0
; AVX2-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
; AVX2-NEXT:    vbroadcastsd {{.*#+}} ymm1 = [-0.0E+0,-0.0E+0,-0.0E+0,-0.0E+0]
; AVX2-NEXT:    vandpd %ymm1, %ymm0, %ymm4
; AVX2-NEXT:    vbroadcastsd {{.*#+}} ymm5 = [4.9999999999999994E-1,4.9999999999999994E-1,4.9999999999999994E-1,4.9999999999999994E-1]
; AVX2-NEXT:    vorpd %ymm5, %ymm4, %ymm4
; AVX2-NEXT:    vaddpd %ymm4, %ymm0, %ymm0
; AVX2-NEXT:    vroundpd $11, %ymm0, %ymm0
; AVX2-NEXT:    vinsertf128 $1, %xmm3, %ymm2, %ymm2
; AVX2-NEXT:    vandpd %ymm1, %ymm2, %ymm1
; AVX2-NEXT:    vorpd %ymm5, %ymm1, %ymm1
; AVX2-NEXT:    vaddpd %ymm1, %ymm2, %ymm1
; AVX2-NEXT:    vroundpd $11, %ymm1, %ymm1
; AVX2-NEXT:    retq
;
; AVX512-LABEL: concat_round_v8f64_v2f64:
; AVX512:       # %bb.0:
; AVX512-NEXT:    # kill: def $xmm2 killed $xmm2 def $ymm2
; AVX512-NEXT:    # kill: def $xmm0 killed $xmm0 def $ymm0
; AVX512-NEXT:    vinserti128 $1, %xmm3, %ymm2, %ymm2
; AVX512-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
; AVX512-NEXT:    vpbroadcastq {{.*#+}} zmm1 = [4.9999999999999994E-1,4.9999999999999994E-1,4.9999999999999994E-1,4.9999999999999994E-1,4.9999999999999994E-1,4.9999999999999994E-1,4.9999999999999994E-1,4.9999999999999994E-1]
; AVX512-NEXT:    vinserti64x4 $1, %ymm2, %zmm0, %zmm0
; AVX512-NEXT:    vpternlogq {{.*#+}} zmm1 = zmm1 | (zmm0 & m64bcst)
; AVX512-NEXT:    vaddpd %zmm1, %zmm0, %zmm0
; AVX512-NEXT:    vrndscalepd $11, %zmm0, %zmm0
; AVX512-NEXT:    retq
  %v0 = call <2 x double> @llvm.round.v2f64(<2 x double> %a0)
  %v1 = call <2 x double> @llvm.round.v2f64(<2 x double> %a1)
  %v2 = call <2 x double> @llvm.round.v2f64(<2 x double> %a2)
  %v3 = call <2 x double> @llvm.round.v2f64(<2 x double> %a3)
  %r01 = shufflevector <2 x double> %v0, <2 x double> %v1, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %r23 = shufflevector <2 x double> %v2, <2 x double> %v3, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %res  = shufflevector <4 x double> %r01, <4 x double> %r23, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
  ret <8 x double> %res
}

define <16 x float> @concat_round_v16f32_v4f32(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2, <4 x float> %a3) {
; SSE-LABEL: concat_round_v16f32_v4f32:
; SSE:       # %bb.0:
; SSE-NEXT:    movaps {{.*#+}} xmm4 = [-0.0E+0,-0.0E+0,-0.0E+0,-0.0E+0]
; SSE-NEXT:    movaps %xmm0, %xmm5
; SSE-NEXT:    andps %xmm4, %xmm5
; SSE-NEXT:    movaps {{.*#+}} xmm6 = [4.9999997E-1,4.9999997E-1,4.9999997E-1,4.9999997E-1]
; SSE-NEXT:    orps %xmm6, %xmm5
; SSE-NEXT:    addps %xmm0, %xmm5
; SSE-NEXT:    roundps $11, %xmm5, %xmm0
; SSE-NEXT:    movaps %xmm1, %xmm5
; SSE-NEXT:    andps %xmm4, %xmm5
; SSE-NEXT:    orps %xmm6, %xmm5
; SSE-NEXT:    addps %xmm1, %xmm5
; SSE-NEXT:    roundps $11, %xmm5, %xmm1
; SSE-NEXT:    movaps %xmm2, %xmm5
; SSE-NEXT:    andps %xmm4, %xmm5
; SSE-NEXT:    orps %xmm6, %xmm5
; SSE-NEXT:    addps %xmm2, %xmm5
; SSE-NEXT:    roundps $11, %xmm5, %xmm2
; SSE-NEXT:    andps %xmm3, %xmm4
; SSE-NEXT:    orps %xmm6, %xmm4
; SSE-NEXT:    addps %xmm3, %xmm4
; SSE-NEXT:    roundps $11, %xmm4, %xmm3
; SSE-NEXT:    retq
;
; AVX1-LABEL: concat_round_v16f32_v4f32:
; AVX1:       # %bb.0:
; AVX1-NEXT:    # kill: def $xmm2 killed $xmm2 def $ymm2
; AVX1-NEXT:    # kill: def $xmm0 killed $xmm0 def $ymm0
; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
; AVX1-NEXT:    vmovaps {{.*#+}} ymm1 = [-0.0E+0,-0.0E+0,-0.0E+0,-0.0E+0,-0.0E+0,-0.0E+0,-0.0E+0,-0.0E+0]
; AVX1-NEXT:    vandps %ymm1, %ymm0, %ymm4
; AVX1-NEXT:    vmovaps {{.*#+}} ymm5 = [4.9999997E-1,4.9999997E-1,4.9999997E-1,4.9999997E-1,4.9999997E-1,4.9999997E-1,4.9999997E-1,4.9999997E-1]
; AVX1-NEXT:    vorps %ymm5, %ymm4, %ymm4
; AVX1-NEXT:    vaddps %ymm4, %ymm0, %ymm0
; AVX1-NEXT:    vroundps $11, %ymm0, %ymm0
; AVX1-NEXT:    vinsertf128 $1, %xmm3, %ymm2, %ymm2
; AVX1-NEXT:    vandps %ymm1, %ymm2, %ymm1
; AVX1-NEXT:    vorps %ymm5, %ymm1, %ymm1
; AVX1-NEXT:    vaddps %ymm1, %ymm2, %ymm1
; AVX1-NEXT:    vroundps $11, %ymm1, %ymm1
; AVX1-NEXT:    retq
;
; AVX2-LABEL: concat_round_v16f32_v4f32:
; AVX2:       # %bb.0:
; AVX2-NEXT:    # kill: def $xmm2 killed $xmm2 def $ymm2
; AVX2-NEXT:    # kill: def $xmm0 killed $xmm0 def $ymm0
; AVX2-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
; AVX2-NEXT:    vbroadcastss {{.*#+}} ymm1 = [-0.0E+0,-0.0E+0,-0.0E+0,-0.0E+0,-0.0E+0,-0.0E+0,-0.0E+0,-0.0E+0]
; AVX2-NEXT:    vandps %ymm1, %ymm0, %ymm4
; AVX2-NEXT:    vbroadcastss {{.*#+}} ymm5 = [4.9999997E-1,4.9999997E-1,4.9999997E-1,4.9999997E-1,4.9999997E-1,4.9999997E-1,4.9999997E-1,4.9999997E-1]
; AVX2-NEXT:    vorps %ymm5, %ymm4, %ymm4
; AVX2-NEXT:    vaddps %ymm4, %ymm0, %ymm0
; AVX2-NEXT:    vroundps $11, %ymm0, %ymm0
; AVX2-NEXT:    vinsertf128 $1, %xmm3, %ymm2, %ymm2
; AVX2-NEXT:    vandps %ymm1, %ymm2, %ymm1
; AVX2-NEXT:    vorps %ymm5, %ymm1, %ymm1
; AVX2-NEXT:    vaddps %ymm1, %ymm2, %ymm1
; AVX2-NEXT:    vroundps $11, %ymm1, %ymm1
; AVX2-NEXT:    retq
;
; AVX512-LABEL: concat_round_v16f32_v4f32:
; AVX512:       # %bb.0:
; AVX512-NEXT:    # kill: def $xmm2 killed $xmm2 def $ymm2
; AVX512-NEXT:    # kill: def $xmm0 killed $xmm0 def $ymm0
; AVX512-NEXT:    vinserti128 $1, %xmm3, %ymm2, %ymm2
; AVX512-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
; AVX512-NEXT:    vpbroadcastd {{.*#+}} zmm1 = [4.9999997E-1,4.9999997E-1,4.9999997E-1,4.9999997E-1,4.9999997E-1,4.9999997E-1,4.9999997E-1,4.9999997E-1,4.9999997E-1,4.9999997E-1,4.9999997E-1,4.9999997E-1,4.9999997E-1,4.9999997E-1,4.9999997E-1,4.9999997E-1]
; AVX512-NEXT:    vinserti64x4 $1, %ymm2, %zmm0, %zmm0
; AVX512-NEXT:    vpternlogd {{.*#+}} zmm1 = zmm1 | (zmm0 & m32bcst)
; AVX512-NEXT:    vaddps %zmm1, %zmm0, %zmm0
; AVX512-NEXT:    vrndscaleps $11, %zmm0, %zmm0
; AVX512-NEXT:    retq
  %v0 = call <4 x float> @llvm.round.v4f32(<4 x float> %a0)
  %v1 = call <4 x float> @llvm.round.v4f32(<4 x float> %a1)
  %v2 = call <4 x float> @llvm.round.v4f32(<4 x float> %a2)
  %v3 = call <4 x float> @llvm.round.v4f32(<4 x float> %a3)
  %r01 = shufflevector <4 x float> %v0, <4 x float> %v1, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
  %r23 = shufflevector <4 x float> %v2, <4 x float> %v3, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
  %res  = shufflevector <8 x float> %r01, <8 x float> %r23, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
  ret <16 x float> %res
}

define <8 x double> @concat_round_v8f64_v4f64(<4 x double> %a0, <4 x double> %a1) {
; SSE-LABEL: concat_round_v8f64_v4f64:
; SSE:       # %bb.0:
; SSE-NEXT:    movapd {{.*#+}} xmm4 = [-0.0E+0,-0.0E+0]
; SSE-NEXT:    movapd %xmm0, %xmm5
; SSE-NEXT:    andpd %xmm4, %xmm5
; SSE-NEXT:    movapd {{.*#+}} xmm6 = [4.9999999999999994E-1,4.9999999999999994E-1]
; SSE-NEXT:    orpd %xmm6, %xmm5
; SSE-NEXT:    addpd %xmm0, %xmm5
; SSE-NEXT:    roundpd $11, %xmm5, %xmm0
; SSE-NEXT:    movapd %xmm1, %xmm5
; SSE-NEXT:    andpd %xmm4, %xmm5
; SSE-NEXT:    orpd %xmm6, %xmm5
; SSE-NEXT:    addpd %xmm1, %xmm5
; SSE-NEXT:    roundpd $11, %xmm5, %xmm1
; SSE-NEXT:    movapd %xmm2, %xmm5
; SSE-NEXT:    andpd %xmm4, %xmm5
; SSE-NEXT:    orpd %xmm6, %xmm5
; SSE-NEXT:    addpd %xmm2, %xmm5
; SSE-NEXT:    roundpd $11, %xmm5, %xmm2
; SSE-NEXT:    andpd %xmm3, %xmm4
; SSE-NEXT:    orpd %xmm6, %xmm4
; SSE-NEXT:    addpd %xmm3, %xmm4
; SSE-NEXT:    roundpd $11, %xmm4, %xmm3
; SSE-NEXT:    retq
;
; AVX1-LABEL: concat_round_v8f64_v4f64:
; AVX1:       # %bb.0:
; AVX1-NEXT:    vmovapd {{.*#+}} ymm2 = [-0.0E+0,-0.0E+0,-0.0E+0,-0.0E+0]
; AVX1-NEXT:    vandpd %ymm2, %ymm0, %ymm3
; AVX1-NEXT:    vmovapd {{.*#+}} ymm4 = [4.9999999999999994E-1,4.9999999999999994E-1,4.9999999999999994E-1,4.9999999999999994E-1]
; AVX1-NEXT:    vorpd %ymm4, %ymm3, %ymm3
; AVX1-NEXT:    vaddpd %ymm3, %ymm0, %ymm0
; AVX1-NEXT:    vroundpd $11, %ymm0, %ymm0
; AVX1-NEXT:    vandpd %ymm2, %ymm1, %ymm2
; AVX1-NEXT:    vorpd %ymm4, %ymm2, %ymm2
; AVX1-NEXT:    vaddpd %ymm2, %ymm1, %ymm1
; AVX1-NEXT:    vroundpd $11, %ymm1, %ymm1
; AVX1-NEXT:    retq
;
; AVX2-LABEL: concat_round_v8f64_v4f64:
; AVX2:       # %bb.0:
; AVX2-NEXT:    vbroadcastsd {{.*#+}} ymm2 = [-0.0E+0,-0.0E+0,-0.0E+0,-0.0E+0]
; AVX2-NEXT:    vandpd %ymm2, %ymm0, %ymm3
; AVX2-NEXT:    vbroadcastsd {{.*#+}} ymm4 = [4.9999999999999994E-1,4.9999999999999994E-1,4.9999999999999994E-1,4.9999999999999994E-1]
; AVX2-NEXT:    vorpd %ymm4, %ymm3, %ymm3
; AVX2-NEXT:    vaddpd %ymm3, %ymm0, %ymm0
; AVX2-NEXT:    vroundpd $11, %ymm0, %ymm0
; AVX2-NEXT:    vandpd %ymm2, %ymm1, %ymm2
; AVX2-NEXT:    vorpd %ymm4, %ymm2, %ymm2
; AVX2-NEXT:    vaddpd %ymm2, %ymm1, %ymm1
; AVX2-NEXT:    vroundpd $11, %ymm1, %ymm1
; AVX2-NEXT:    retq
;
; AVX512-LABEL: concat_round_v8f64_v4f64:
; AVX512:       # %bb.0:
; AVX512-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
; AVX512-NEXT:    vpbroadcastq {{.*#+}} zmm2 = [4.9999999999999994E-1,4.9999999999999994E-1,4.9999999999999994E-1,4.9999999999999994E-1,4.9999999999999994E-1,4.9999999999999994E-1,4.9999999999999994E-1,4.9999999999999994E-1]
; AVX512-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
; AVX512-NEXT:    vpternlogq {{.*#+}} zmm2 = zmm2 | (zmm0 & m64bcst)
; AVX512-NEXT:    vaddpd %zmm2, %zmm0, %zmm0
; AVX512-NEXT:    vrndscalepd $11, %zmm0, %zmm0
; AVX512-NEXT:    retq
  %v0 = call <4 x double> @llvm.round.v4f64(<4 x double> %a0)
  %v1 = call <4 x double> @llvm.round.v4f64(<4 x double> %a1)
  %res  = shufflevector <4 x double> %v0, <4 x double> %v1, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
  ret <8 x double> %res
}

define <16 x float> @concat_round_v16f32_v8f32(<8 x float> %a0, <8 x float> %a1) {
; SSE-LABEL: concat_round_v16f32_v8f32:
; SSE:       # %bb.0:
; SSE-NEXT:    movaps {{.*#+}} xmm4 = [-0.0E+0,-0.0E+0,-0.0E+0,-0.0E+0]
; SSE-NEXT:    movaps %xmm0, %xmm5
; SSE-NEXT:    andps %xmm4, %xmm5
; SSE-NEXT:    movaps {{.*#+}} xmm6 = [4.9999997E-1,4.9999997E-1,4.9999997E-1,4.9999997E-1]
; SSE-NEXT:    orps %xmm6, %xmm5
; SSE-NEXT:    addps %xmm0, %xmm5
; SSE-NEXT:    roundps $11, %xmm5, %xmm0
; SSE-NEXT:    movaps %xmm1, %xmm5
; SSE-NEXT:    andps %xmm4, %xmm5
; SSE-NEXT:    orps %xmm6, %xmm5
; SSE-NEXT:    addps %xmm1, %xmm5
; SSE-NEXT:    roundps $11, %xmm5, %xmm1
; SSE-NEXT:    movaps %xmm2, %xmm5
; SSE-NEXT:    andps %xmm4, %xmm5
; SSE-NEXT:    orps %xmm6, %xmm5
; SSE-NEXT:    addps %xmm2, %xmm5
; SSE-NEXT:    roundps $11, %xmm5, %xmm2
; SSE-NEXT:    andps %xmm3, %xmm4
; SSE-NEXT:    orps %xmm6, %xmm4
; SSE-NEXT:    addps %xmm3, %xmm4
; SSE-NEXT:    roundps $11, %xmm4, %xmm3
; SSE-NEXT:    retq
;
; AVX1-LABEL: concat_round_v16f32_v8f32:
; AVX1:       # %bb.0:
; AVX1-NEXT:    vmovaps {{.*#+}} ymm2 = [-0.0E+0,-0.0E+0,-0.0E+0,-0.0E+0,-0.0E+0,-0.0E+0,-0.0E+0,-0.0E+0]
; AVX1-NEXT:    vandps %ymm2, %ymm0, %ymm3
; AVX1-NEXT:    vmovaps {{.*#+}} ymm4 = [4.9999997E-1,4.9999997E-1,4.9999997E-1,4.9999997E-1,4.9999997E-1,4.9999997E-1,4.9999997E-1,4.9999997E-1]
; AVX1-NEXT:    vorps %ymm4, %ymm3, %ymm3
; AVX1-NEXT:    vaddps %ymm3, %ymm0, %ymm0
; AVX1-NEXT:    vroundps $11, %ymm0, %ymm0
; AVX1-NEXT:    vandps %ymm2, %ymm1, %ymm2
; AVX1-NEXT:    vorps %ymm4, %ymm2, %ymm2
; AVX1-NEXT:    vaddps %ymm2, %ymm1, %ymm1
; AVX1-NEXT:    vroundps $11, %ymm1, %ymm1
; AVX1-NEXT:    retq
;
; AVX2-LABEL: concat_round_v16f32_v8f32:
; AVX2:       # %bb.0:
; AVX2-NEXT:    vbroadcastss {{.*#+}} ymm2 = [-0.0E+0,-0.0E+0,-0.0E+0,-0.0E+0,-0.0E+0,-0.0E+0,-0.0E+0,-0.0E+0]
; AVX2-NEXT:    vandps %ymm2, %ymm0, %ymm3
; AVX2-NEXT:    vbroadcastss {{.*#+}} ymm4 = [4.9999997E-1,4.9999997E-1,4.9999997E-1,4.9999997E-1,4.9999997E-1,4.9999997E-1,4.9999997E-1,4.9999997E-1]
; AVX2-NEXT:    vorps %ymm4, %ymm3, %ymm3
; AVX2-NEXT:    vaddps %ymm3, %ymm0, %ymm0
; AVX2-NEXT:    vroundps $11, %ymm0, %ymm0
; AVX2-NEXT:    vandps %ymm2, %ymm1, %ymm2
; AVX2-NEXT:    vorps %ymm4, %ymm2, %ymm2
; AVX2-NEXT:    vaddps %ymm2, %ymm1, %ymm1
; AVX2-NEXT:    vroundps $11, %ymm1, %ymm1
; AVX2-NEXT:    retq
;
; AVX512-LABEL: concat_round_v16f32_v8f32:
; AVX512:       # %bb.0:
; AVX512-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
; AVX512-NEXT:    vpbroadcastd {{.*#+}} zmm2 = [4.9999997E-1,4.9999997E-1,4.9999997E-1,4.9999997E-1,4.9999997E-1,4.9999997E-1,4.9999997E-1,4.9999997E-1,4.9999997E-1,4.9999997E-1,4.9999997E-1,4.9999997E-1,4.9999997E-1,4.9999997E-1,4.9999997E-1,4.9999997E-1]
; AVX512-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
; AVX512-NEXT:    vpternlogd {{.*#+}} zmm2 = zmm2 | (zmm0 & m32bcst)
; AVX512-NEXT:    vaddps %zmm2, %zmm0, %zmm0
; AVX512-NEXT:    vrndscaleps $11, %zmm0, %zmm0
; AVX512-NEXT:    retq
  %v0 = call <8 x float> @llvm.round.v8f32(<8 x float> %a0)
  %v1 = call <8 x float> @llvm.round.v8f32(<8 x float> %a1)
  %res  = shufflevector <8 x float> %v0, <8 x float> %v1, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
  ret <16 x float> %res
}
;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
; AVX: {{.*}}