[llvm] r239146 - [X86][AVX2] Added tests for v32i8 vector shifts
Simon Pilgrim
llvm-dev at redking.me.uk
Fri Jun 5 05:35:36 PDT 2015
Author: rksimon
Date: Fri Jun 5 07:35:36 2015
New Revision: 239146
URL: http://llvm.org/viewvc/llvm-project?rev=239146&view=rev
Log:
[X86][AVX2] Added tests for v32i8 vector shifts
Currently still scalarized, but D9474 should remedy that.
Modified:
llvm/trunk/test/CodeGen/X86/avx2-vector-shifts.ll
Modified: llvm/trunk/test/CodeGen/X86/avx2-vector-shifts.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/avx2-vector-shifts.ll?rev=239146&r1=239145&r2=239146&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/X86/avx2-vector-shifts.ll (original)
+++ llvm/trunk/test/CodeGen/X86/avx2-vector-shifts.ll Fri Jun 5 07:35:36 2015
@@ -300,6 +300,56 @@ define <16 x i16> @shl_16i16(<16 x i16>
ret <16 x i16> %shl
}
+define <32 x i8> @shl_32i8(<32 x i8> %r, <32 x i8> %a) nounwind {
+; CHECK-LABEL: shl_32i8
+; CHECK: vextracti128 $1, %ymm0, %xmm3
+; CHECK-NEXT: vpsllw $4, %xmm3, %xmm2
+; CHECK-NEXT: vmovdqa {{.*#+}} xmm8 = [240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240]
+; CHECK-NEXT: vpand %xmm8, %xmm2, %xmm5
+; CHECK-NEXT: vextracti128 $1, %ymm1, %xmm2
+; CHECK-NEXT: vpsllw $5, %xmm2, %xmm2
+; CHECK-NEXT: vmovdqa {{.*#+}} xmm9 = [224,224,224,224,224,224,224,224,224,224,224,224,224,224,224,224]
+; CHECK-NEXT: vpand %xmm9, %xmm2, %xmm7
+; CHECK-NEXT: vmovdqa {{.*#+}} xmm2 = [128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128]
+; CHECK-NEXT: vpand %xmm7, %xmm2, %xmm4
+; CHECK-NEXT: vpcmpeqb %xmm2, %xmm4, %xmm4
+; CHECK-NEXT: vpblendvb %xmm4, %xmm5, %xmm3, %xmm3
+; CHECK-NEXT: vpsllw $2, %xmm3, %xmm4
+; CHECK-NEXT: vmovdqa {{.*#+}} xmm5 = [252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252]
+; CHECK-NEXT: vpand %xmm5, %xmm4, %xmm4
+; CHECK-NEXT: vpaddb %xmm7, %xmm7, %xmm7
+; CHECK-NEXT: vpand %xmm7, %xmm2, %xmm6
+; CHECK-NEXT: vpcmpeqb %xmm2, %xmm6, %xmm6
+; CHECK-NEXT: vpblendvb %xmm6, %xmm4, %xmm3, %xmm3
+; CHECK-NEXT: vpaddb %xmm3, %xmm3, %xmm4
+; CHECK-NEXT: vpaddb %xmm7, %xmm7, %xmm6
+; CHECK-NEXT: vpand %xmm6, %xmm2, %xmm6
+; CHECK-NEXT: vpcmpeqb %xmm2, %xmm6, %xmm6
+; CHECK-NEXT: vpblendvb %xmm6, %xmm4, %xmm3, %xmm3
+; CHECK-NEXT: vpsllw $4, %xmm0, %xmm4
+; CHECK-NEXT: vpand %xmm8, %xmm4, %xmm4
+; CHECK-NEXT: vpsllw $5, %xmm1, %xmm1
+; CHECK-NEXT: vpand %xmm9, %xmm1, %xmm1
+; CHECK-NEXT: vpand %xmm1, %xmm2, %xmm6
+; CHECK-NEXT: vpcmpeqb %xmm2, %xmm6, %xmm6
+; CHECK-NEXT: vpblendvb %xmm6, %xmm4, %xmm0, %xmm0
+; CHECK-NEXT: vpsllw $2, %xmm0, %xmm4
+; CHECK-NEXT: vpand %xmm5, %xmm4, %xmm4
+; CHECK-NEXT: vpaddb %xmm1, %xmm1, %xmm1
+; CHECK-NEXT: vpand %xmm1, %xmm2, %xmm5
+; CHECK-NEXT: vpcmpeqb %xmm2, %xmm5, %xmm5
+; CHECK-NEXT: vpblendvb %xmm5, %xmm4, %xmm0, %xmm0
+; CHECK-NEXT: vpaddb %xmm0, %xmm0, %xmm4
+; CHECK-NEXT: vpaddb %xmm1, %xmm1, %xmm1
+; CHECK-NEXT: vpand %xmm1, %xmm2, %xmm1
+; CHECK-NEXT: vpcmpeqb %xmm2, %xmm1, %xmm1
+; CHECK-NEXT: vpblendvb %xmm1, %xmm4, %xmm0, %xmm0
+; CHECK-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm0
+; CHECK-NEXT: retq
+ %shl = shl <32 x i8> %r, %a
+ ret <32 x i8> %shl
+}
+
define <8 x i16> @ashr_8i16(<8 x i16> %r, <8 x i16> %a) nounwind {
; CHECK-LABEL: ashr_8i16
; CHECK: vpmovzxwd {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero
@@ -329,6 +379,176 @@ define <16 x i16> @ashr_16i16(<16 x i16>
ret <16 x i16> %ashr
}
+define <32 x i8> @ashr_32i8(<32 x i8> %r, <32 x i8> %a) nounwind {
+; CHECK-LABEL: ashr_32i8
+; CHECK: vextracti128 $1, %ymm1, %xmm2
+; CHECK-NEXT: vpextrb $1, %xmm2, %ecx
+; CHECK-NEXT: vextracti128 $1, %ymm0, %xmm3
+; CHECK-NEXT: vpextrb $1, %xmm3, %eax
+; CHECK-NEXT: sarb %cl, %al
+; CHECK-NEXT: vpextrb $0, %xmm2, %ecx
+; CHECK-NEXT: vpextrb $0, %xmm3, %edx
+; CHECK-NEXT: sarb %cl, %dl
+; CHECK-NEXT: movzbl %al, %eax
+; CHECK-NEXT: movzbl %dl, %edx
+; CHECK-NEXT: vpextrb $2, %xmm2, %ecx
+; CHECK-NEXT: vpextrb $2, %xmm3, %esi
+; CHECK-NEXT: sarb %cl, %sil
+; CHECK-NEXT: vmovd %edx, %xmm4
+; CHECK-NEXT: vpinsrb $1, %eax, %xmm4, %xmm4
+; CHECK-NEXT: movzbl %sil, %eax
+; CHECK-NEXT: vpextrb $3, %xmm2, %ecx
+; CHECK-NEXT: vpextrb $3, %xmm3, %edx
+; CHECK-NEXT: sarb %cl, %dl
+; CHECK-NEXT: vpinsrb $2, %eax, %xmm4, %xmm4
+; CHECK-NEXT: movzbl %dl, %eax
+; CHECK-NEXT: vpinsrb $3, %eax, %xmm4, %xmm4
+; CHECK-NEXT: vpextrb $4, %xmm2, %ecx
+; CHECK-NEXT: vpextrb $4, %xmm3, %eax
+; CHECK-NEXT: sarb %cl, %al
+; CHECK-NEXT: movzbl %al, %eax
+; CHECK-NEXT: vpinsrb $4, %eax, %xmm4, %xmm4
+; CHECK-NEXT: vpextrb $5, %xmm2, %ecx
+; CHECK-NEXT: vpextrb $5, %xmm3, %eax
+; CHECK-NEXT: sarb %cl, %al
+; CHECK-NEXT: vpextrb $6, %xmm2, %ecx
+; CHECK-NEXT: vpextrb $6, %xmm3, %edx
+; CHECK-NEXT: sarb %cl, %dl
+; CHECK-NEXT: movzbl %al, %eax
+; CHECK-NEXT: vpinsrb $5, %eax, %xmm4, %xmm4
+; CHECK-NEXT: movzbl %dl, %eax
+; CHECK-NEXT: vpextrb $7, %xmm2, %ecx
+; CHECK-NEXT: vpextrb $7, %xmm3, %edx
+; CHECK-NEXT: sarb %cl, %dl
+; CHECK-NEXT: vpinsrb $6, %eax, %xmm4, %xmm4
+; CHECK-NEXT: movzbl %dl, %eax
+; CHECK-NEXT: vpinsrb $7, %eax, %xmm4, %xmm4
+; CHECK-NEXT: vpextrb $8, %xmm2, %ecx
+; CHECK-NEXT: vpextrb $8, %xmm3, %eax
+; CHECK-NEXT: sarb %cl, %al
+; CHECK-NEXT: movzbl %al, %eax
+; CHECK-NEXT: vpinsrb $8, %eax, %xmm4, %xmm4
+; CHECK-NEXT: vpextrb $9, %xmm2, %ecx
+; CHECK-NEXT: vpextrb $9, %xmm3, %eax
+; CHECK-NEXT: sarb %cl, %al
+; CHECK-NEXT: vpextrb $10, %xmm2, %ecx
+; CHECK-NEXT: vpextrb $10, %xmm3, %edx
+; CHECK-NEXT: sarb %cl, %dl
+; CHECK-NEXT: movzbl %al, %eax
+; CHECK-NEXT: vpinsrb $9, %eax, %xmm4, %xmm4
+; CHECK-NEXT: movzbl %dl, %eax
+; CHECK-NEXT: vpextrb $11, %xmm2, %ecx
+; CHECK-NEXT: vpextrb $11, %xmm3, %edx
+; CHECK-NEXT: sarb %cl, %dl
+; CHECK-NEXT: vpinsrb $10, %eax, %xmm4, %xmm4
+; CHECK-NEXT: movzbl %dl, %eax
+; CHECK-NEXT: vpinsrb $11, %eax, %xmm4, %xmm4
+; CHECK-NEXT: vpextrb $12, %xmm2, %ecx
+; CHECK-NEXT: vpextrb $12, %xmm3, %eax
+; CHECK-NEXT: sarb %cl, %al
+; CHECK-NEXT: movzbl %al, %eax
+; CHECK-NEXT: vpinsrb $12, %eax, %xmm4, %xmm4
+; CHECK-NEXT: vpextrb $13, %xmm2, %ecx
+; CHECK-NEXT: vpextrb $13, %xmm3, %eax
+; CHECK-NEXT: sarb %cl, %al
+; CHECK-NEXT: vpextrb $14, %xmm2, %ecx
+; CHECK-NEXT: vpextrb $14, %xmm3, %edx
+; CHECK-NEXT: sarb %cl, %dl
+; CHECK-NEXT: movzbl %al, %eax
+; CHECK-NEXT: vpinsrb $13, %eax, %xmm4, %xmm4
+; CHECK-NEXT: vpextrb $15, %xmm2, %ecx
+; CHECK-NEXT: vpextrb $15, %xmm3, %eax
+; CHECK-NEXT: sarb %cl, %al
+; CHECK-NEXT: vpextrb $1, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $1, %xmm0, %esi
+; CHECK-NEXT: sarb %cl, %sil
+; CHECK-NEXT: movzbl %dl, %ecx
+; CHECK-NEXT: vpinsrb $14, %ecx, %xmm4, %xmm2
+; CHECK-NEXT: vpextrb $0, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $0, %xmm0, %edx
+; CHECK-NEXT: sarb %cl, %dl
+; CHECK-NEXT: vpextrb $2, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $2, %xmm0, %edi
+; CHECK-NEXT: sarb %cl, %dil
+; CHECK-NEXT: movzbl %al, %eax
+; CHECK-NEXT: vpinsrb $15, %eax, %xmm2, %xmm2
+; CHECK-NEXT: movzbl %sil, %eax
+; CHECK-NEXT: movzbl %dl, %ecx
+; CHECK-NEXT: vmovd %ecx, %xmm3
+; CHECK-NEXT: vpinsrb $1, %eax, %xmm3, %xmm3
+; CHECK-NEXT: movzbl %dil, %eax
+; CHECK-NEXT: vpextrb $3, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $3, %xmm0, %edx
+; CHECK-NEXT: sarb %cl, %dl
+; CHECK-NEXT: vpinsrb $2, %eax, %xmm3, %xmm3
+; CHECK-NEXT: movzbl %dl, %eax
+; CHECK-NEXT: vpinsrb $3, %eax, %xmm3, %xmm3
+; CHECK-NEXT: vpextrb $4, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $4, %xmm0, %eax
+; CHECK-NEXT: sarb %cl, %al
+; CHECK-NEXT: movzbl %al, %eax
+; CHECK-NEXT: vpinsrb $4, %eax, %xmm3, %xmm3
+; CHECK-NEXT: vpextrb $5, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $5, %xmm0, %eax
+; CHECK-NEXT: sarb %cl, %al
+; CHECK-NEXT: vpextrb $6, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $6, %xmm0, %edx
+; CHECK-NEXT: sarb %cl, %dl
+; CHECK-NEXT: movzbl %al, %eax
+; CHECK-NEXT: vpinsrb $5, %eax, %xmm3, %xmm3
+; CHECK-NEXT: movzbl %dl, %eax
+; CHECK-NEXT: vpextrb $7, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $7, %xmm0, %edx
+; CHECK-NEXT: sarb %cl, %dl
+; CHECK-NEXT: vpinsrb $6, %eax, %xmm3, %xmm3
+; CHECK-NEXT: movzbl %dl, %eax
+; CHECK-NEXT: vpinsrb $7, %eax, %xmm3, %xmm3
+; CHECK-NEXT: vpextrb $8, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $8, %xmm0, %eax
+; CHECK-NEXT: sarb %cl, %al
+; CHECK-NEXT: movzbl %al, %eax
+; CHECK-NEXT: vpinsrb $8, %eax, %xmm3, %xmm3
+; CHECK-NEXT: vpextrb $9, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $9, %xmm0, %eax
+; CHECK-NEXT: sarb %cl, %al
+; CHECK-NEXT: vpextrb $10, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $10, %xmm0, %edx
+; CHECK-NEXT: sarb %cl, %dl
+; CHECK-NEXT: movzbl %al, %eax
+; CHECK-NEXT: vpinsrb $9, %eax, %xmm3, %xmm3
+; CHECK-NEXT: movzbl %dl, %eax
+; CHECK-NEXT: vpextrb $11, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $11, %xmm0, %edx
+; CHECK-NEXT: sarb %cl, %dl
+; CHECK-NEXT: vpinsrb $10, %eax, %xmm3, %xmm3
+; CHECK-NEXT: movzbl %dl, %eax
+; CHECK-NEXT: vpinsrb $11, %eax, %xmm3, %xmm3
+; CHECK-NEXT: vpextrb $12, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $12, %xmm0, %eax
+; CHECK-NEXT: sarb %cl, %al
+; CHECK-NEXT: movzbl %al, %eax
+; CHECK-NEXT: vpinsrb $12, %eax, %xmm3, %xmm3
+; CHECK-NEXT: vpextrb $13, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $13, %xmm0, %eax
+; CHECK-NEXT: sarb %cl, %al
+; CHECK-NEXT: vpextrb $14, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $14, %xmm0, %edx
+; CHECK-NEXT: sarb %cl, %dl
+; CHECK-NEXT: movzbl %al, %eax
+; CHECK-NEXT: vpinsrb $13, %eax, %xmm3, %xmm3
+; CHECK-NEXT: movzbl %dl, %eax
+; CHECK-NEXT: vpextrb $15, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $15, %xmm0, %edx
+; CHECK-NEXT: sarb %cl, %dl
+; CHECK-NEXT: vpinsrb $14, %eax, %xmm3, %xmm0
+; CHECK-NEXT: movzbl %dl, %eax
+; CHECK-NEXT: vpinsrb $15, %eax, %xmm0, %xmm0
+; CHECK-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm0
+; CHECK-NEXT: retq
+ %ashr = ashr <32 x i8> %r, %a
+ ret <32 x i8> %ashr
+}
+
define <8 x i16> @lshr_8i16(<8 x i16> %r, <8 x i16> %a) nounwind {
; CHECK-LABEL: lshr_8i16
; CHECK: vpmovzxwd {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero
@@ -357,3 +577,173 @@ define <16 x i16> @lshr_16i16(<16 x i16>
%lshr = lshr <16 x i16> %r, %a
ret <16 x i16> %lshr
}
+
+define <32 x i8> @lshr_32i8(<32 x i8> %r, <32 x i8> %a) nounwind {
+; CHECK-LABEL: lshr_32i8
+; CHECK: vextracti128 $1, %ymm1, %xmm2
+; CHECK-NEXT: vpextrb $1, %xmm2, %ecx
+; CHECK-NEXT: vextracti128 $1, %ymm0, %xmm3
+; CHECK-NEXT: vpextrb $1, %xmm3, %eax
+; CHECK-NEXT: shrb %cl, %al
+; CHECK-NEXT: vpextrb $0, %xmm2, %ecx
+; CHECK-NEXT: vpextrb $0, %xmm3, %edx
+; CHECK-NEXT: shrb %cl, %dl
+; CHECK-NEXT: movzbl %al, %eax
+; CHECK-NEXT: movzbl %dl, %edx
+; CHECK-NEXT: vpextrb $2, %xmm2, %ecx
+; CHECK-NEXT: vpextrb $2, %xmm3, %esi
+; CHECK-NEXT: shrb %cl, %sil
+; CHECK-NEXT: vmovd %edx, %xmm4
+; CHECK-NEXT: vpinsrb $1, %eax, %xmm4, %xmm4
+; CHECK-NEXT: movzbl %sil, %eax
+; CHECK-NEXT: vpextrb $3, %xmm2, %ecx
+; CHECK-NEXT: vpextrb $3, %xmm3, %edx
+; CHECK-NEXT: shrb %cl, %dl
+; CHECK-NEXT: vpinsrb $2, %eax, %xmm4, %xmm4
+; CHECK-NEXT: movzbl %dl, %eax
+; CHECK-NEXT: vpinsrb $3, %eax, %xmm4, %xmm4
+; CHECK-NEXT: vpextrb $4, %xmm2, %ecx
+; CHECK-NEXT: vpextrb $4, %xmm3, %eax
+; CHECK-NEXT: shrb %cl, %al
+; CHECK-NEXT: movzbl %al, %eax
+; CHECK-NEXT: vpinsrb $4, %eax, %xmm4, %xmm4
+; CHECK-NEXT: vpextrb $5, %xmm2, %ecx
+; CHECK-NEXT: vpextrb $5, %xmm3, %eax
+; CHECK-NEXT: shrb %cl, %al
+; CHECK-NEXT: vpextrb $6, %xmm2, %ecx
+; CHECK-NEXT: vpextrb $6, %xmm3, %edx
+; CHECK-NEXT: shrb %cl, %dl
+; CHECK-NEXT: movzbl %al, %eax
+; CHECK-NEXT: vpinsrb $5, %eax, %xmm4, %xmm4
+; CHECK-NEXT: movzbl %dl, %eax
+; CHECK-NEXT: vpextrb $7, %xmm2, %ecx
+; CHECK-NEXT: vpextrb $7, %xmm3, %edx
+; CHECK-NEXT: shrb %cl, %dl
+; CHECK-NEXT: vpinsrb $6, %eax, %xmm4, %xmm4
+; CHECK-NEXT: movzbl %dl, %eax
+; CHECK-NEXT: vpinsrb $7, %eax, %xmm4, %xmm4
+; CHECK-NEXT: vpextrb $8, %xmm2, %ecx
+; CHECK-NEXT: vpextrb $8, %xmm3, %eax
+; CHECK-NEXT: shrb %cl, %al
+; CHECK-NEXT: movzbl %al, %eax
+; CHECK-NEXT: vpinsrb $8, %eax, %xmm4, %xmm4
+; CHECK-NEXT: vpextrb $9, %xmm2, %ecx
+; CHECK-NEXT: vpextrb $9, %xmm3, %eax
+; CHECK-NEXT: shrb %cl, %al
+; CHECK-NEXT: vpextrb $10, %xmm2, %ecx
+; CHECK-NEXT: vpextrb $10, %xmm3, %edx
+; CHECK-NEXT: shrb %cl, %dl
+; CHECK-NEXT: movzbl %al, %eax
+; CHECK-NEXT: vpinsrb $9, %eax, %xmm4, %xmm4
+; CHECK-NEXT: movzbl %dl, %eax
+; CHECK-NEXT: vpextrb $11, %xmm2, %ecx
+; CHECK-NEXT: vpextrb $11, %xmm3, %edx
+; CHECK-NEXT: shrb %cl, %dl
+; CHECK-NEXT: vpinsrb $10, %eax, %xmm4, %xmm4
+; CHECK-NEXT: movzbl %dl, %eax
+; CHECK-NEXT: vpinsrb $11, %eax, %xmm4, %xmm4
+; CHECK-NEXT: vpextrb $12, %xmm2, %ecx
+; CHECK-NEXT: vpextrb $12, %xmm3, %eax
+; CHECK-NEXT: shrb %cl, %al
+; CHECK-NEXT: movzbl %al, %eax
+; CHECK-NEXT: vpinsrb $12, %eax, %xmm4, %xmm4
+; CHECK-NEXT: vpextrb $13, %xmm2, %ecx
+; CHECK-NEXT: vpextrb $13, %xmm3, %eax
+; CHECK-NEXT: shrb %cl, %al
+; CHECK-NEXT: vpextrb $14, %xmm2, %ecx
+; CHECK-NEXT: vpextrb $14, %xmm3, %edx
+; CHECK-NEXT: shrb %cl, %dl
+; CHECK-NEXT: movzbl %al, %eax
+; CHECK-NEXT: vpinsrb $13, %eax, %xmm4, %xmm4
+; CHECK-NEXT: vpextrb $15, %xmm2, %ecx
+; CHECK-NEXT: vpextrb $15, %xmm3, %eax
+; CHECK-NEXT: shrb %cl, %al
+; CHECK-NEXT: vpextrb $1, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $1, %xmm0, %esi
+; CHECK-NEXT: shrb %cl, %sil
+; CHECK-NEXT: movzbl %dl, %ecx
+; CHECK-NEXT: vpinsrb $14, %ecx, %xmm4, %xmm2
+; CHECK-NEXT: vpextrb $0, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $0, %xmm0, %edx
+; CHECK-NEXT: shrb %cl, %dl
+; CHECK-NEXT: vpextrb $2, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $2, %xmm0, %edi
+; CHECK-NEXT: shrb %cl, %dil
+; CHECK-NEXT: movzbl %al, %eax
+; CHECK-NEXT: vpinsrb $15, %eax, %xmm2, %xmm2
+; CHECK-NEXT: movzbl %sil, %eax
+; CHECK-NEXT: movzbl %dl, %ecx
+; CHECK-NEXT: vmovd %ecx, %xmm3
+; CHECK-NEXT: vpinsrb $1, %eax, %xmm3, %xmm3
+; CHECK-NEXT: movzbl %dil, %eax
+; CHECK-NEXT: vpextrb $3, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $3, %xmm0, %edx
+; CHECK-NEXT: shrb %cl, %dl
+; CHECK-NEXT: vpinsrb $2, %eax, %xmm3, %xmm3
+; CHECK-NEXT: movzbl %dl, %eax
+; CHECK-NEXT: vpinsrb $3, %eax, %xmm3, %xmm3
+; CHECK-NEXT: vpextrb $4, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $4, %xmm0, %eax
+; CHECK-NEXT: shrb %cl, %al
+; CHECK-NEXT: movzbl %al, %eax
+; CHECK-NEXT: vpinsrb $4, %eax, %xmm3, %xmm3
+; CHECK-NEXT: vpextrb $5, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $5, %xmm0, %eax
+; CHECK-NEXT: shrb %cl, %al
+; CHECK-NEXT: vpextrb $6, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $6, %xmm0, %edx
+; CHECK-NEXT: shrb %cl, %dl
+; CHECK-NEXT: movzbl %al, %eax
+; CHECK-NEXT: vpinsrb $5, %eax, %xmm3, %xmm3
+; CHECK-NEXT: movzbl %dl, %eax
+; CHECK-NEXT: vpextrb $7, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $7, %xmm0, %edx
+; CHECK-NEXT: shrb %cl, %dl
+; CHECK-NEXT: vpinsrb $6, %eax, %xmm3, %xmm3
+; CHECK-NEXT: movzbl %dl, %eax
+; CHECK-NEXT: vpinsrb $7, %eax, %xmm3, %xmm3
+; CHECK-NEXT: vpextrb $8, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $8, %xmm0, %eax
+; CHECK-NEXT: shrb %cl, %al
+; CHECK-NEXT: movzbl %al, %eax
+; CHECK-NEXT: vpinsrb $8, %eax, %xmm3, %xmm3
+; CHECK-NEXT: vpextrb $9, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $9, %xmm0, %eax
+; CHECK-NEXT: shrb %cl, %al
+; CHECK-NEXT: vpextrb $10, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $10, %xmm0, %edx
+; CHECK-NEXT: shrb %cl, %dl
+; CHECK-NEXT: movzbl %al, %eax
+; CHECK-NEXT: vpinsrb $9, %eax, %xmm3, %xmm3
+; CHECK-NEXT: movzbl %dl, %eax
+; CHECK-NEXT: vpextrb $11, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $11, %xmm0, %edx
+; CHECK-NEXT: shrb %cl, %dl
+; CHECK-NEXT: vpinsrb $10, %eax, %xmm3, %xmm3
+; CHECK-NEXT: movzbl %dl, %eax
+; CHECK-NEXT: vpinsrb $11, %eax, %xmm3, %xmm3
+; CHECK-NEXT: vpextrb $12, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $12, %xmm0, %eax
+; CHECK-NEXT: shrb %cl, %al
+; CHECK-NEXT: movzbl %al, %eax
+; CHECK-NEXT: vpinsrb $12, %eax, %xmm3, %xmm3
+; CHECK-NEXT: vpextrb $13, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $13, %xmm0, %eax
+; CHECK-NEXT: shrb %cl, %al
+; CHECK-NEXT: vpextrb $14, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $14, %xmm0, %edx
+; CHECK-NEXT: shrb %cl, %dl
+; CHECK-NEXT: movzbl %al, %eax
+; CHECK-NEXT: vpinsrb $13, %eax, %xmm3, %xmm3
+; CHECK-NEXT: movzbl %dl, %eax
+; CHECK-NEXT: vpextrb $15, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $15, %xmm0, %edx
+; CHECK-NEXT: shrb %cl, %dl
+; CHECK-NEXT: vpinsrb $14, %eax, %xmm3, %xmm0
+; CHECK-NEXT: movzbl %dl, %eax
+; CHECK-NEXT: vpinsrb $15, %eax, %xmm0, %xmm0
+; CHECK-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm0
+; CHECK-NEXT: retq
+ %lshr = lshr <32 x i8> %r, %a
+ ret <32 x i8> %lshr
+}
More information about the llvm-commits
mailing list