; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse2 | FileCheck %s --check-prefix=CHECK --check-prefix=SSE2
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse4.1 | FileCheck %s --check-prefix=CHECK --check-prefix=SSE41
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse2 | FileCheck %s --check-prefixes=CHECK,SSE,SSE2
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse4.1 | FileCheck %s --check-prefixes=CHECK,SSE,SSE41
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx | FileCheck %s --check-prefixes=CHECK,AVX,AVX1
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefixes=CHECK,AVX,AVX2
define void @foo(<3 x float> %in, <4 x i8>* nocapture %out) nounwind {
; SSE2-LABEL: foo:
; SSE41-NEXT: pinsrb $3, %eax, %xmm0
; SSE41-NEXT: movd %xmm0, (%rdi)
; SSE41-NEXT: retq
+;
+; AVX-LABEL: foo:
+; AVX: # %bb.0:
+; AVX-NEXT: vcvttps2dq %xmm0, %xmm0
+; AVX-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,4,8],zero,xmm0[u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX-NEXT: movl $255, %eax
+; AVX-NEXT: vpinsrb $3, %eax, %xmm0, %xmm0
+; AVX-NEXT: vmovd %xmm0, (%rdi)
+; AVX-NEXT: retq
%t0 = fptoui <3 x float> %in to <3 x i8>
%t1 = shufflevector <3 x i8> %t0, <3 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 undef>
%t2 = insertelement <4 x i8> %t1, i8 -1, i32 3
; SSE41: # %bb.0: # %entry
; SSE41-NEXT: insertps {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[2],zero
; SSE41-NEXT: retq
+;
+; AVX-LABEL: test_negative_zero_1:
+; AVX: # %bb.0: # %entry
+; AVX-NEXT: vinsertps {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[2],zero
+; AVX-NEXT: retq
entry:
%0 = extractelement <4 x float> %A, i32 0
%1 = insertelement <4 x float> undef, float %0, i32 0
; SSE41: # %bb.0: # %entry
; SSE41-NEXT: blendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; SSE41-NEXT: retq
+;
+; AVX-LABEL: test_negative_zero_2:
+; AVX: # %bb.0: # %entry
+; AVX-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
+; AVX-NEXT: retq
entry:
%0 = extractelement <2 x double> %A, i32 0
%1 = insertelement <2 x double> undef, double %0, i32 0
; SSE41-NEXT: insertps {{.*#+}} xmm0 = xmm0[0,1],xmm2[0],xmm0[3]
; SSE41-NEXT: insertps {{.*#+}} xmm0 = xmm0[0,1,2],xmm3[0]
; SSE41-NEXT: retq
+;
+; AVX-LABEL: test_buildvector_v4f32_register:
+; AVX: # %bb.0:
+; AVX-NEXT: vinsertps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[2,3]
+; AVX-NEXT: vinsertps {{.*#+}} xmm0 = xmm0[0,1],xmm2[0],xmm0[3]
+; AVX-NEXT: vinsertps {{.*#+}} xmm0 = xmm0[0,1,2],xmm3[0]
+; AVX-NEXT: retq
%ins0 = insertelement <4 x float> undef, float %f0, i32 0
%ins1 = insertelement <4 x float> %ins0, float %f1, i32 1
%ins2 = insertelement <4 x float> %ins1, float %f2, i32 2
; SSE41-NEXT: insertps {{.*#+}} xmm0 = xmm0[0,1],mem[0],xmm0[3]
; SSE41-NEXT: insertps {{.*#+}} xmm0 = xmm0[0,1,2],mem[0]
; SSE41-NEXT: retq
+;
+; AVX-LABEL: test_buildvector_v4f32_load:
+; AVX: # %bb.0:
+; AVX-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; AVX-NEXT: vinsertps {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[2,3]
+; AVX-NEXT: vinsertps {{.*#+}} xmm0 = xmm0[0,1],mem[0],xmm0[3]
+; AVX-NEXT: vinsertps {{.*#+}} xmm0 = xmm0[0,1,2],mem[0]
+; AVX-NEXT: retq
%f0 = load float, float* %p0, align 4
%f1 = load float, float* %p1, align 4
%f2 = load float, float* %p2, align 4
; SSE41-NEXT: insertps {{.*#+}} xmm0 = xmm0[0,1],xmm2[0],xmm0[3]
; SSE41-NEXT: insertps {{.*#+}} xmm0 = xmm0[0,1,2],mem[0]
; SSE41-NEXT: retq
+;
+; AVX-LABEL: test_buildvector_v4f32_partial_load:
+; AVX: # %bb.0:
+; AVX-NEXT: vinsertps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[2,3]
+; AVX-NEXT: vinsertps {{.*#+}} xmm0 = xmm0[0,1],xmm2[0],xmm0[3]
+; AVX-NEXT: vinsertps {{.*#+}} xmm0 = xmm0[0,1,2],mem[0]
+; AVX-NEXT: retq
%f3 = load float, float* %p3, align 4
%ins0 = insertelement <4 x float> undef, float %f0, i32 0
%ins1 = insertelement <4 x float> %ins0, float %f1, i32 1
; SSE41-NEXT: pinsrd $2, %edx, %xmm0
; SSE41-NEXT: pinsrd $3, %ecx, %xmm0
; SSE41-NEXT: retq
+;
+; AVX-LABEL: test_buildvector_v4i32_register:
+; AVX: # %bb.0:
+; AVX-NEXT: vmovd %edi, %xmm0
+; AVX-NEXT: vpinsrd $1, %esi, %xmm0, %xmm0
+; AVX-NEXT: vpinsrd $2, %edx, %xmm0, %xmm0
+; AVX-NEXT: vpinsrd $3, %ecx, %xmm0, %xmm0
+; AVX-NEXT: retq
%ins0 = insertelement <4 x i32> undef, i32 %a0, i32 0
%ins1 = insertelement <4 x i32> %ins0, i32 %a1, i32 1
%ins2 = insertelement <4 x i32> %ins1, i32 %a2, i32 2
; SSE41-NEXT: movd %edi, %xmm0
; SSE41-NEXT: pinsrd $3, %esi, %xmm0
; SSE41-NEXT: retq
+;
+; AVX-LABEL: test_buildvector_v4i32_partial:
+; AVX: # %bb.0:
+; AVX-NEXT: vmovd %edi, %xmm0
+; AVX-NEXT: vpinsrd $3, %esi, %xmm0, %xmm0
+; AVX-NEXT: retq
%ins0 = insertelement <4 x i32> undef, i32 %a0, i32 0
%ins1 = insertelement <4 x i32> %ins0, i32 undef, i32 1
%ins2 = insertelement <4 x i32> %ins1, i32 undef, i32 2
}
define <4 x i32> @test_buildvector_v4i32_register_zero(i32 %a0, i32 %a2, i32 %a3) {
-; CHECK-LABEL: test_buildvector_v4i32_register_zero:
-; CHECK: # %bb.0:
-; CHECK-NEXT: movd %edx, %xmm0
-; CHECK-NEXT: movd %esi, %xmm1
-; CHECK-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; CHECK-NEXT: movd %edi, %xmm0
-; CHECK-NEXT: punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
-; CHECK-NEXT: retq
+; SSE-LABEL: test_buildvector_v4i32_register_zero:
+; SSE: # %bb.0:
+; SSE-NEXT: movd %edx, %xmm0
+; SSE-NEXT: movd %esi, %xmm1
+; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT: movd %edi, %xmm0
+; SSE-NEXT: punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; SSE-NEXT: retq
+;
+; AVX-LABEL: test_buildvector_v4i32_register_zero:
+; AVX: # %bb.0:
+; AVX-NEXT: vmovd %edx, %xmm0
+; AVX-NEXT: vmovd %esi, %xmm1
+; AVX-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; AVX-NEXT: vmovd %edi, %xmm1
+; AVX-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm1[0],xmm0[0]
+; AVX-NEXT: retq
%ins0 = insertelement <4 x i32> undef, i32 %a0, i32 0
%ins1 = insertelement <4 x i32> %ins0, i32 0, i32 1
%ins2 = insertelement <4 x i32> %ins1, i32 %a2, i32 2
}
define <4 x i32> @test_buildvector_v4i32_register_zero_2(i32 %a1, i32 %a2, i32 %a3) {
-; CHECK-LABEL: test_buildvector_v4i32_register_zero_2:
-; CHECK: # %bb.0:
-; CHECK-NEXT: movd %edx, %xmm0
-; CHECK-NEXT: movd %esi, %xmm1
-; CHECK-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; CHECK-NEXT: movd %edi, %xmm0
-; CHECK-NEXT: shufps {{.*#+}} xmm0 = xmm0[1,0],xmm1[0,1]
-; CHECK-NEXT: retq
+; SSE-LABEL: test_buildvector_v4i32_register_zero_2:
+; SSE: # %bb.0:
+; SSE-NEXT: movd %edx, %xmm0
+; SSE-NEXT: movd %esi, %xmm1
+; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT: movd %edi, %xmm0
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[1,0],xmm1[0,1]
+; SSE-NEXT: retq
+;
+; AVX-LABEL: test_buildvector_v4i32_register_zero_2:
+; AVX: # %bb.0:
+; AVX-NEXT: vmovd %edx, %xmm0
+; AVX-NEXT: vmovd %esi, %xmm1
+; AVX-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; AVX-NEXT: vmovd %edi, %xmm1
+; AVX-NEXT: vshufps {{.*#+}} xmm0 = xmm1[1,0],xmm0[0,1]
+; AVX-NEXT: retq
%ins0 = insertelement <4 x i32> undef, i32 0, i32 0
%ins1 = insertelement <4 x i32> %ins0, i32 %a1, i32 1
%ins2 = insertelement <4 x i32> %ins1, i32 %a2, i32 2
; SSE41-NEXT: pinsrw $6, {{[0-9]+}}(%rsp), %xmm0
; SSE41-NEXT: pinsrw $7, {{[0-9]+}}(%rsp), %xmm0
; SSE41-NEXT: retq
+;
+; AVX-LABEL: test_buildvector_v8i16_register:
+; AVX: # %bb.0:
+; AVX-NEXT: vmovd %edi, %xmm0
+; AVX-NEXT: vpinsrw $1, %esi, %xmm0, %xmm0
+; AVX-NEXT: vpinsrw $2, %edx, %xmm0, %xmm0
+; AVX-NEXT: vpinsrw $3, %ecx, %xmm0, %xmm0
+; AVX-NEXT: vpinsrw $4, %r8d, %xmm0, %xmm0
+; AVX-NEXT: vpinsrw $5, %r9d, %xmm0, %xmm0
+; AVX-NEXT: vpinsrw $6, {{[0-9]+}}(%rsp), %xmm0, %xmm0
+; AVX-NEXT: vpinsrw $7, {{[0-9]+}}(%rsp), %xmm0, %xmm0
+; AVX-NEXT: retq
%ins0 = insertelement <8 x i16> undef, i16 %a0, i32 0
%ins1 = insertelement <8 x i16> %ins0, i16 %a1, i32 1
%ins2 = insertelement <8 x i16> %ins1, i16 %a2, i32 2
}
define <8 x i16> @test_buildvector_v8i16_partial(i16 %a1, i16 %a3, i16 %a4, i16 %a5) {
-; CHECK-LABEL: test_buildvector_v8i16_partial:
-; CHECK: # %bb.0:
-; CHECK-NEXT: pxor %xmm0, %xmm0
-; CHECK-NEXT: pinsrw $1, %edi, %xmm0
-; CHECK-NEXT: pinsrw $3, %esi, %xmm0
-; CHECK-NEXT: pinsrw $4, %edx, %xmm0
-; CHECK-NEXT: pinsrw $5, %ecx, %xmm0
-; CHECK-NEXT: retq
+; SSE-LABEL: test_buildvector_v8i16_partial:
+; SSE: # %bb.0:
+; SSE-NEXT: pxor %xmm0, %xmm0
+; SSE-NEXT: pinsrw $1, %edi, %xmm0
+; SSE-NEXT: pinsrw $3, %esi, %xmm0
+; SSE-NEXT: pinsrw $4, %edx, %xmm0
+; SSE-NEXT: pinsrw $5, %ecx, %xmm0
+; SSE-NEXT: retq
+;
+; AVX-LABEL: test_buildvector_v8i16_partial:
+; AVX: # %bb.0:
+; AVX-NEXT: vpxor %xmm0, %xmm0, %xmm0
+; AVX-NEXT: vpinsrw $1, %edi, %xmm0, %xmm0
+; AVX-NEXT: vpinsrw $3, %esi, %xmm0, %xmm0
+; AVX-NEXT: vpinsrw $4, %edx, %xmm0, %xmm0
+; AVX-NEXT: vpinsrw $5, %ecx, %xmm0, %xmm0
+; AVX-NEXT: retq
%ins0 = insertelement <8 x i16> undef, i16 undef, i32 0
%ins1 = insertelement <8 x i16> %ins0, i16 %a1, i32 1
%ins2 = insertelement <8 x i16> %ins1, i16 undef, i32 2
}
define <8 x i16> @test_buildvector_v8i16_register_zero(i16 %a0, i16 %a3, i16 %a4, i16 %a5) {
-; CHECK-LABEL: test_buildvector_v8i16_register_zero:
-; CHECK: # %bb.0:
-; CHECK-NEXT: movzwl %di, %eax
-; CHECK-NEXT: movd %eax, %xmm0
-; CHECK-NEXT: pinsrw $3, %esi, %xmm0
-; CHECK-NEXT: pinsrw $4, %edx, %xmm0
-; CHECK-NEXT: pinsrw $5, %ecx, %xmm0
-; CHECK-NEXT: retq
+; SSE-LABEL: test_buildvector_v8i16_register_zero:
+; SSE: # %bb.0:
+; SSE-NEXT: movzwl %di, %eax
+; SSE-NEXT: movd %eax, %xmm0
+; SSE-NEXT: pinsrw $3, %esi, %xmm0
+; SSE-NEXT: pinsrw $4, %edx, %xmm0
+; SSE-NEXT: pinsrw $5, %ecx, %xmm0
+; SSE-NEXT: retq
+;
+; AVX-LABEL: test_buildvector_v8i16_register_zero:
+; AVX: # %bb.0:
+; AVX-NEXT: movzwl %di, %eax
+; AVX-NEXT: vmovd %eax, %xmm0
+; AVX-NEXT: vpinsrw $3, %esi, %xmm0, %xmm0
+; AVX-NEXT: vpinsrw $4, %edx, %xmm0, %xmm0
+; AVX-NEXT: vpinsrw $5, %ecx, %xmm0, %xmm0
+; AVX-NEXT: retq
%ins0 = insertelement <8 x i16> undef, i16 %a0, i32 0
%ins1 = insertelement <8 x i16> %ins0, i16 0, i32 1
%ins2 = insertelement <8 x i16> %ins1, i16 0, i32 2
}
define <8 x i16> @test_buildvector_v8i16_register_zero_2(i16 %a1, i16 %a3, i16 %a4, i16 %a5) {
-; CHECK-LABEL: test_buildvector_v8i16_register_zero_2:
-; CHECK: # %bb.0:
-; CHECK-NEXT: pxor %xmm0, %xmm0
-; CHECK-NEXT: pinsrw $1, %edi, %xmm0
-; CHECK-NEXT: pinsrw $3, %esi, %xmm0
-; CHECK-NEXT: pinsrw $4, %edx, %xmm0
-; CHECK-NEXT: pinsrw $5, %ecx, %xmm0
-; CHECK-NEXT: retq
+; SSE-LABEL: test_buildvector_v8i16_register_zero_2:
+; SSE: # %bb.0:
+; SSE-NEXT: pxor %xmm0, %xmm0
+; SSE-NEXT: pinsrw $1, %edi, %xmm0
+; SSE-NEXT: pinsrw $3, %esi, %xmm0
+; SSE-NEXT: pinsrw $4, %edx, %xmm0
+; SSE-NEXT: pinsrw $5, %ecx, %xmm0
+; SSE-NEXT: retq
+;
+; AVX-LABEL: test_buildvector_v8i16_register_zero_2:
+; AVX: # %bb.0:
+; AVX-NEXT: vpxor %xmm0, %xmm0, %xmm0
+; AVX-NEXT: vpinsrw $1, %edi, %xmm0, %xmm0
+; AVX-NEXT: vpinsrw $3, %esi, %xmm0, %xmm0
+; AVX-NEXT: vpinsrw $4, %edx, %xmm0, %xmm0
+; AVX-NEXT: vpinsrw $5, %ecx, %xmm0, %xmm0
+; AVX-NEXT: retq
%ins0 = insertelement <8 x i16> undef, i16 0, i32 0
%ins1 = insertelement <8 x i16> %ins0, i16 %a1, i32 1
%ins2 = insertelement <8 x i16> %ins1, i16 0, i32 2
; SSE41-NEXT: pinsrb $14, {{[0-9]+}}(%rsp), %xmm0
; SSE41-NEXT: pinsrb $15, {{[0-9]+}}(%rsp), %xmm0
; SSE41-NEXT: retq
+;
+; AVX-LABEL: test_buildvector_v16i8_register:
+; AVX: # %bb.0:
+; AVX-NEXT: vmovd %edi, %xmm0
+; AVX-NEXT: vpinsrb $1, %esi, %xmm0, %xmm0
+; AVX-NEXT: vpinsrb $2, %edx, %xmm0, %xmm0
+; AVX-NEXT: vpinsrb $3, %ecx, %xmm0, %xmm0
+; AVX-NEXT: vpinsrb $4, %r8d, %xmm0, %xmm0
+; AVX-NEXT: vpinsrb $5, %r9d, %xmm0, %xmm0
+; AVX-NEXT: vpinsrb $6, {{[0-9]+}}(%rsp), %xmm0, %xmm0
+; AVX-NEXT: vpinsrb $7, {{[0-9]+}}(%rsp), %xmm0, %xmm0
+; AVX-NEXT: vpinsrb $8, {{[0-9]+}}(%rsp), %xmm0, %xmm0
+; AVX-NEXT: vpinsrb $9, {{[0-9]+}}(%rsp), %xmm0, %xmm0
+; AVX-NEXT: vpinsrb $10, {{[0-9]+}}(%rsp), %xmm0, %xmm0
+; AVX-NEXT: vpinsrb $11, {{[0-9]+}}(%rsp), %xmm0, %xmm0
+; AVX-NEXT: vpinsrb $12, {{[0-9]+}}(%rsp), %xmm0, %xmm0
+; AVX-NEXT: vpinsrb $13, {{[0-9]+}}(%rsp), %xmm0, %xmm0
+; AVX-NEXT: vpinsrb $14, {{[0-9]+}}(%rsp), %xmm0, %xmm0
+; AVX-NEXT: vpinsrb $15, {{[0-9]+}}(%rsp), %xmm0, %xmm0
+; AVX-NEXT: retq
%ins0 = insertelement <16 x i8> undef, i8 %a0, i32 0
%ins1 = insertelement <16 x i8> %ins0, i8 %a1, i32 1
%ins2 = insertelement <16 x i8> %ins1, i8 %a2, i32 2
; SSE41-NEXT: pinsrb $12, %r8d, %xmm0
; SSE41-NEXT: pinsrb $15, %r9d, %xmm0
; SSE41-NEXT: retq
+;
+; AVX-LABEL: test_buildvector_v16i8_partial:
+; AVX: # %bb.0:
+; AVX-NEXT: vpxor %xmm0, %xmm0, %xmm0
+; AVX-NEXT: vpinsrb $2, %edi, %xmm0, %xmm0
+; AVX-NEXT: vpinsrb $6, %esi, %xmm0, %xmm0
+; AVX-NEXT: vpinsrb $8, %edx, %xmm0, %xmm0
+; AVX-NEXT: vpinsrb $11, %ecx, %xmm0, %xmm0
+; AVX-NEXT: vpinsrb $12, %r8d, %xmm0, %xmm0
+; AVX-NEXT: vpinsrb $15, %r9d, %xmm0, %xmm0
+; AVX-NEXT: retq
%ins0 = insertelement <16 x i8> undef, i8 undef, i32 0
%ins1 = insertelement <16 x i8> %ins0, i8 undef, i32 1
%ins2 = insertelement <16 x i8> %ins1, i8 %a2, i32 2
; SSE41-NEXT: pinsrb $12, %r9d, %xmm0
; SSE41-NEXT: pinsrb $15, {{[0-9]+}}(%rsp), %xmm0
; SSE41-NEXT: retq
+;
+; AVX-LABEL: test_buildvector_v16i8_register_zero:
+; AVX: # %bb.0:
+; AVX-NEXT: movzbl %dil, %eax
+; AVX-NEXT: vmovd %eax, %xmm0
+; AVX-NEXT: vpinsrb $4, %esi, %xmm0, %xmm0
+; AVX-NEXT: vpinsrb $6, %edx, %xmm0, %xmm0
+; AVX-NEXT: vpinsrb $8, %ecx, %xmm0, %xmm0
+; AVX-NEXT: vpinsrb $11, %r8d, %xmm0, %xmm0
+; AVX-NEXT: vpinsrb $12, %r9d, %xmm0, %xmm0
+; AVX-NEXT: vpinsrb $15, {{[0-9]+}}(%rsp), %xmm0, %xmm0
+; AVX-NEXT: retq
%ins0 = insertelement <16 x i8> undef, i8 %a0, i32 0
%ins1 = insertelement <16 x i8> %ins0, i8 0, i32 1
%ins2 = insertelement <16 x i8> %ins1, i8 0, i32 2
; SSE41-NEXT: pinsrb $12, %r9d, %xmm0
; SSE41-NEXT: pinsrb $15, {{[0-9]+}}(%rsp), %xmm0
; SSE41-NEXT: retq
+;
+; AVX-LABEL: test_buildvector_v16i8_register_zero_2:
+; AVX: # %bb.0:
+; AVX-NEXT: vpxor %xmm0, %xmm0, %xmm0
+; AVX-NEXT: vpinsrb $2, %edi, %xmm0, %xmm0
+; AVX-NEXT: vpinsrb $3, %esi, %xmm0, %xmm0
+; AVX-NEXT: vpinsrb $6, %edx, %xmm0, %xmm0
+; AVX-NEXT: vpinsrb $8, %ecx, %xmm0, %xmm0
+; AVX-NEXT: vpinsrb $11, %r8d, %xmm0, %xmm0
+; AVX-NEXT: vpinsrb $12, %r9d, %xmm0, %xmm0
+; AVX-NEXT: vpinsrb $15, {{[0-9]+}}(%rsp), %xmm0, %xmm0
+; AVX-NEXT: retq
%ins0 = insertelement <16 x i8> undef, i8 0, i32 0
%ins1 = insertelement <16 x i8> %ins0, i8 0, i32 1
%ins2 = insertelement <16 x i8> %ins1, i8 %a2, i32 2