summaryrefslogtreecommitdiffstats
path: root/llvm/test/CodeGen/X86/vector-rotate-128.ll
diff options
context:
space:
mode:
Diffstat (limited to 'llvm/test/CodeGen/X86/vector-rotate-128.ll')
-rw-r--r--llvm/test/CodeGen/X86/vector-rotate-128.ll266
1 files changed, 120 insertions, 146 deletions
diff --git a/llvm/test/CodeGen/X86/vector-rotate-128.ll b/llvm/test/CodeGen/X86/vector-rotate-128.ll
index 7ea470dfea3..166b75f3310 100644
--- a/llvm/test/CodeGen/X86/vector-rotate-128.ll
+++ b/llvm/test/CodeGen/X86/vector-rotate-128.ll
@@ -129,32 +129,28 @@ define <4 x i32> @var_rotate_v4i32(<4 x i32> %a, <4 x i32> %b) nounwind {
; SSE2-NEXT: cvttps2dq %xmm1, %xmm1
; SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm1[1,1,3,3]
; SSE2-NEXT: pmuludq %xmm0, %xmm1
+; SSE2-NEXT: pshufd {{.*#+}} xmm4 = xmm1[0,2,2,3]
+; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT: pmuludq %xmm3, %xmm1
; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
-; SSE2-NEXT: pshufd {{.*#+}} xmm4 = xmm0[1,1,3,3]
-; SSE2-NEXT: pmuludq %xmm3, %xmm4
-; SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm4[0,2,2,3]
-; SSE2-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
-; SSE2-NEXT: movdqa %xmm2, %xmm3
-; SSE2-NEXT: psrlq $32, %xmm3
-; SSE2-NEXT: movdqa %xmm0, %xmm4
-; SSE2-NEXT: psrld %xmm3, %xmm4
-; SSE2-NEXT: movdqa %xmm2, %xmm3
-; SSE2-NEXT: psrldq {{.*#+}} xmm3 = xmm3[12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; SSE2-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
+; SSE2-NEXT: pshuflw {{.*#+}} xmm1 = xmm2[2,3,3,3,4,5,6,7]
+; SSE2-NEXT: movdqa %xmm0, %xmm3
+; SSE2-NEXT: psrld %xmm1, %xmm3
+; SSE2-NEXT: pshuflw {{.*#+}} xmm5 = xmm2[0,1,1,1,4,5,6,7]
+; SSE2-NEXT: movdqa %xmm0, %xmm1
+; SSE2-NEXT: psrld %xmm5, %xmm1
+; SSE2-NEXT: punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm3[0]
+; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm2[2,3,0,1]
+; SSE2-NEXT: pshuflw {{.*#+}} xmm3 = xmm2[2,3,3,3,4,5,6,7]
; SSE2-NEXT: movdqa %xmm0, %xmm5
; SSE2-NEXT: psrld %xmm3, %xmm5
-; SSE2-NEXT: movsd {{.*#+}} xmm5 = xmm4[0],xmm5[1]
-; SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm5[1,3,2,3]
-; SSE2-NEXT: pxor %xmm4, %xmm4
-; SSE2-NEXT: movdqa %xmm2, %xmm5
-; SSE2-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
-; SSE2-NEXT: movdqa %xmm0, %xmm6
-; SSE2-NEXT: psrld %xmm5, %xmm6
-; SSE2-NEXT: punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; SSE2-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[0,1,1,1,4,5,6,7]
; SSE2-NEXT: psrld %xmm2, %xmm0
-; SSE2-NEXT: movsd {{.*#+}} xmm0 = xmm6[0],xmm0[1]
-; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
-; SSE2-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
-; SSE2-NEXT: por %xmm1, %xmm0
+; SSE2-NEXT: punpckhqdq {{.*#+}} xmm0 = xmm0[1],xmm5[1]
+; SSE2-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,3],xmm0[0,3]
+; SSE2-NEXT: orps %xmm4, %xmm1
+; SSE2-NEXT: movaps %xmm1, %xmm0
; SSE2-NEXT: retq
;
; SSE41-LABEL: var_rotate_v4i32:
@@ -165,23 +161,21 @@ define <4 x i32> @var_rotate_v4i32(<4 x i32> %a, <4 x i32> %b) nounwind {
; SSE41-NEXT: paddd {{.*}}(%rip), %xmm1
; SSE41-NEXT: cvttps2dq %xmm1, %xmm1
; SSE41-NEXT: pmulld %xmm0, %xmm1
-; SSE41-NEXT: movdqa %xmm2, %xmm3
-; SSE41-NEXT: psrldq {{.*#+}} xmm3 = xmm3[12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; SSE41-NEXT: pshuflw {{.*#+}} xmm3 = xmm2[2,3,3,3,4,5,6,7]
; SSE41-NEXT: movdqa %xmm0, %xmm4
; SSE41-NEXT: psrld %xmm3, %xmm4
-; SSE41-NEXT: movdqa %xmm2, %xmm3
-; SSE41-NEXT: psrlq $32, %xmm3
-; SSE41-NEXT: movdqa %xmm0, %xmm5
-; SSE41-NEXT: psrld %xmm3, %xmm5
-; SSE41-NEXT: pblendw {{.*#+}} xmm5 = xmm5[0,1,2,3],xmm4[4,5,6,7]
-; SSE41-NEXT: pxor %xmm3, %xmm3
-; SSE41-NEXT: pmovzxdq {{.*#+}} xmm4 = xmm2[0],zero,xmm2[1],zero
-; SSE41-NEXT: punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm3[2],xmm2[3],xmm3[3]
-; SSE41-NEXT: movdqa %xmm0, %xmm3
-; SSE41-NEXT: psrld %xmm2, %xmm3
-; SSE41-NEXT: psrld %xmm4, %xmm0
-; SSE41-NEXT: pblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm3[4,5,6,7]
-; SSE41-NEXT: pblendw {{.*#+}} xmm0 = xmm0[0,1],xmm5[2,3],xmm0[4,5],xmm5[6,7]
+; SSE41-NEXT: pshufd {{.*#+}} xmm3 = xmm2[2,3,0,1]
+; SSE41-NEXT: pshuflw {{.*#+}} xmm5 = xmm3[2,3,3,3,4,5,6,7]
+; SSE41-NEXT: movdqa %xmm0, %xmm6
+; SSE41-NEXT: psrld %xmm5, %xmm6
+; SSE41-NEXT: pblendw {{.*#+}} xmm6 = xmm4[0,1,2,3],xmm6[4,5,6,7]
+; SSE41-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[0,1,1,1,4,5,6,7]
+; SSE41-NEXT: movdqa %xmm0, %xmm4
+; SSE41-NEXT: psrld %xmm2, %xmm4
+; SSE41-NEXT: pshuflw {{.*#+}} xmm2 = xmm3[0,1,1,1,4,5,6,7]
+; SSE41-NEXT: psrld %xmm2, %xmm0
+; SSE41-NEXT: pblendw {{.*#+}} xmm0 = xmm4[0,1,2,3],xmm0[4,5,6,7]
+; SSE41-NEXT: pblendw {{.*#+}} xmm0 = xmm0[0,1],xmm6[2,3],xmm0[4,5],xmm6[6,7]
; SSE41-NEXT: por %xmm1, %xmm0
; SSE41-NEXT: retq
;
@@ -245,32 +239,28 @@ define <4 x i32> @var_rotate_v4i32(<4 x i32> %a, <4 x i32> %b) nounwind {
; X32-SSE-NEXT: cvttps2dq %xmm1, %xmm1
; X32-SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm1[1,1,3,3]
; X32-SSE-NEXT: pmuludq %xmm0, %xmm1
+; X32-SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm1[0,2,2,3]
+; X32-SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; X32-SSE-NEXT: pmuludq %xmm3, %xmm1
; X32-SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
-; X32-SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm0[1,1,3,3]
-; X32-SSE-NEXT: pmuludq %xmm3, %xmm4
-; X32-SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm4[0,2,2,3]
-; X32-SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
-; X32-SSE-NEXT: movdqa %xmm2, %xmm3
-; X32-SSE-NEXT: psrlq $32, %xmm3
-; X32-SSE-NEXT: movdqa %xmm0, %xmm4
-; X32-SSE-NEXT: psrld %xmm3, %xmm4
-; X32-SSE-NEXT: movdqa %xmm2, %xmm3
-; X32-SSE-NEXT: psrldq {{.*#+}} xmm3 = xmm3[12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X32-SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
+; X32-SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm2[2,3,3,3,4,5,6,7]
+; X32-SSE-NEXT: movdqa %xmm0, %xmm3
+; X32-SSE-NEXT: psrld %xmm1, %xmm3
+; X32-SSE-NEXT: pshuflw {{.*#+}} xmm5 = xmm2[0,1,1,1,4,5,6,7]
+; X32-SSE-NEXT: movdqa %xmm0, %xmm1
+; X32-SSE-NEXT: psrld %xmm5, %xmm1
+; X32-SSE-NEXT: punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm3[0]
+; X32-SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[2,3,0,1]
+; X32-SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm2[2,3,3,3,4,5,6,7]
; X32-SSE-NEXT: movdqa %xmm0, %xmm5
; X32-SSE-NEXT: psrld %xmm3, %xmm5
-; X32-SSE-NEXT: movsd {{.*#+}} xmm5 = xmm4[0],xmm5[1]
-; X32-SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm5[1,3,2,3]
-; X32-SSE-NEXT: pxor %xmm4, %xmm4
-; X32-SSE-NEXT: movdqa %xmm2, %xmm5
-; X32-SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
-; X32-SSE-NEXT: movdqa %xmm0, %xmm6
-; X32-SSE-NEXT: psrld %xmm5, %xmm6
-; X32-SSE-NEXT: punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; X32-SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[0,1,1,1,4,5,6,7]
; X32-SSE-NEXT: psrld %xmm2, %xmm0
-; X32-SSE-NEXT: movsd {{.*#+}} xmm0 = xmm6[0],xmm0[1]
-; X32-SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
-; X32-SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
-; X32-SSE-NEXT: por %xmm1, %xmm0
+; X32-SSE-NEXT: punpckhqdq {{.*#+}} xmm0 = xmm0[1],xmm5[1]
+; X32-SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,3],xmm0[0,3]
+; X32-SSE-NEXT: orps %xmm4, %xmm1
+; X32-SSE-NEXT: movaps %xmm1, %xmm0
; X32-SSE-NEXT: retl
%b32 = sub <4 x i32> <i32 32, i32 32, i32 32, i32 32>, %b
%shl = shl <4 x i32> %a, %b
@@ -910,35 +900,30 @@ define <2 x i64> @splatvar_rotate_v2i64(<2 x i64> %a, <2 x i64> %b) nounwind {
define <4 x i32> @splatvar_rotate_v4i32(<4 x i32> %a, <4 x i32> %b) nounwind {
; SSE2-LABEL: splatvar_rotate_v4i32:
; SSE2: # %bb.0:
-; SSE2-NEXT: pxor %xmm2, %xmm2
-; SSE2-NEXT: xorps %xmm3, %xmm3
-; SSE2-NEXT: movss {{.*#+}} xmm3 = xmm1[0],xmm3[1,2,3]
+; SSE2-NEXT: xorps %xmm2, %xmm2
+; SSE2-NEXT: movss {{.*#+}} xmm2 = xmm1[0],xmm2[1,2,3]
; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,0,0,0]
-; SSE2-NEXT: movdqa {{.*#+}} xmm4 = [32,32,32,32]
-; SSE2-NEXT: psubd %xmm1, %xmm4
+; SSE2-NEXT: movdqa {{.*#+}} xmm3 = [32,32,32,32]
+; SSE2-NEXT: psubd %xmm1, %xmm3
+; SSE2-NEXT: movdqa %xmm0, %xmm4
+; SSE2-NEXT: pslld %xmm2, %xmm4
+; SSE2-NEXT: pshuflw {{.*#+}} xmm1 = xmm3[2,3,3,3,4,5,6,7]
+; SSE2-NEXT: movdqa %xmm0, %xmm2
+; SSE2-NEXT: psrld %xmm1, %xmm2
+; SSE2-NEXT: pshuflw {{.*#+}} xmm5 = xmm3[0,1,1,1,4,5,6,7]
; SSE2-NEXT: movdqa %xmm0, %xmm1
-; SSE2-NEXT: pslld %xmm3, %xmm1
-; SSE2-NEXT: movdqa %xmm4, %xmm3
-; SSE2-NEXT: psrlq $32, %xmm3
+; SSE2-NEXT: psrld %xmm5, %xmm1
+; SSE2-NEXT: punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm3[2,3,0,1]
+; SSE2-NEXT: pshuflw {{.*#+}} xmm3 = xmm2[2,3,3,3,4,5,6,7]
; SSE2-NEXT: movdqa %xmm0, %xmm5
; SSE2-NEXT: psrld %xmm3, %xmm5
-; SSE2-NEXT: movdqa %xmm4, %xmm3
-; SSE2-NEXT: psrldq {{.*#+}} xmm3 = xmm3[12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; SSE2-NEXT: movdqa %xmm0, %xmm6
-; SSE2-NEXT: psrld %xmm3, %xmm6
-; SSE2-NEXT: movsd {{.*#+}} xmm6 = xmm5[0],xmm6[1]
-; SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm6[1,3,2,3]
-; SSE2-NEXT: movdqa %xmm4, %xmm5
-; SSE2-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm2[0],xmm5[1],xmm2[1]
-; SSE2-NEXT: movdqa %xmm0, %xmm6
-; SSE2-NEXT: psrld %xmm5, %xmm6
-; SSE2-NEXT: punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm2[2],xmm4[3],xmm2[3]
-; SSE2-NEXT: psrld %xmm4, %xmm0
-; SSE2-NEXT: movsd {{.*#+}} xmm0 = xmm6[0],xmm0[1]
-; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
-; SSE2-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
-; SSE2-NEXT: por %xmm0, %xmm1
-; SSE2-NEXT: movdqa %xmm1, %xmm0
+; SSE2-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[0,1,1,1,4,5,6,7]
+; SSE2-NEXT: psrld %xmm2, %xmm0
+; SSE2-NEXT: punpckhqdq {{.*#+}} xmm0 = xmm0[1],xmm5[1]
+; SSE2-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,3],xmm0[0,3]
+; SSE2-NEXT: orps %xmm4, %xmm1
+; SSE2-NEXT: movaps %xmm1, %xmm0
; SSE2-NEXT: retq
;
; SSE41-LABEL: splatvar_rotate_v4i32:
@@ -949,23 +934,21 @@ define <4 x i32> @splatvar_rotate_v4i32(<4 x i32> %a, <4 x i32> %b) nounwind {
; SSE41-NEXT: psubd %xmm1, %xmm3
; SSE41-NEXT: movdqa %xmm0, %xmm1
; SSE41-NEXT: pslld %xmm2, %xmm1
-; SSE41-NEXT: movdqa %xmm3, %xmm2
-; SSE41-NEXT: psrldq {{.*#+}} xmm2 = xmm2[12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; SSE41-NEXT: pshuflw {{.*#+}} xmm2 = xmm3[2,3,3,3,4,5,6,7]
; SSE41-NEXT: movdqa %xmm0, %xmm4
; SSE41-NEXT: psrld %xmm2, %xmm4
-; SSE41-NEXT: movdqa %xmm3, %xmm2
-; SSE41-NEXT: psrlq $32, %xmm2
-; SSE41-NEXT: movdqa %xmm0, %xmm5
-; SSE41-NEXT: psrld %xmm2, %xmm5
-; SSE41-NEXT: pblendw {{.*#+}} xmm5 = xmm5[0,1,2,3],xmm4[4,5,6,7]
-; SSE41-NEXT: pxor %xmm2, %xmm2
-; SSE41-NEXT: pmovzxdq {{.*#+}} xmm4 = xmm3[0],zero,xmm3[1],zero
-; SSE41-NEXT: punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm2[2],xmm3[3],xmm2[3]
-; SSE41-NEXT: movdqa %xmm0, %xmm2
-; SSE41-NEXT: psrld %xmm3, %xmm2
-; SSE41-NEXT: psrld %xmm4, %xmm0
-; SSE41-NEXT: pblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm2[4,5,6,7]
-; SSE41-NEXT: pblendw {{.*#+}} xmm0 = xmm0[0,1],xmm5[2,3],xmm0[4,5],xmm5[6,7]
+; SSE41-NEXT: pshufd {{.*#+}} xmm2 = xmm3[2,3,0,1]
+; SSE41-NEXT: pshuflw {{.*#+}} xmm5 = xmm2[2,3,3,3,4,5,6,7]
+; SSE41-NEXT: movdqa %xmm0, %xmm6
+; SSE41-NEXT: psrld %xmm5, %xmm6
+; SSE41-NEXT: pblendw {{.*#+}} xmm6 = xmm4[0,1,2,3],xmm6[4,5,6,7]
+; SSE41-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[0,1,1,1,4,5,6,7]
+; SSE41-NEXT: movdqa %xmm0, %xmm4
+; SSE41-NEXT: psrld %xmm3, %xmm4
+; SSE41-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[0,1,1,1,4,5,6,7]
+; SSE41-NEXT: psrld %xmm2, %xmm0
+; SSE41-NEXT: pblendw {{.*#+}} xmm0 = xmm4[0,1,2,3],xmm0[4,5,6,7]
+; SSE41-NEXT: pblendw {{.*#+}} xmm0 = xmm0[0,1],xmm6[2,3],xmm0[4,5],xmm6[6,7]
; SSE41-NEXT: por %xmm1, %xmm0
; SSE41-NEXT: retq
;
@@ -1031,35 +1014,30 @@ define <4 x i32> @splatvar_rotate_v4i32(<4 x i32> %a, <4 x i32> %b) nounwind {
;
; X32-SSE-LABEL: splatvar_rotate_v4i32:
; X32-SSE: # %bb.0:
-; X32-SSE-NEXT: pxor %xmm2, %xmm2
-; X32-SSE-NEXT: xorps %xmm3, %xmm3
-; X32-SSE-NEXT: movss {{.*#+}} xmm3 = xmm1[0],xmm3[1,2,3]
+; X32-SSE-NEXT: xorps %xmm2, %xmm2
+; X32-SSE-NEXT: movss {{.*#+}} xmm2 = xmm1[0],xmm2[1,2,3]
; X32-SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,0,0,0]
-; X32-SSE-NEXT: movdqa {{.*#+}} xmm4 = [32,32,32,32]
-; X32-SSE-NEXT: psubd %xmm1, %xmm4
+; X32-SSE-NEXT: movdqa {{.*#+}} xmm3 = [32,32,32,32]
+; X32-SSE-NEXT: psubd %xmm1, %xmm3
+; X32-SSE-NEXT: movdqa %xmm0, %xmm4
+; X32-SSE-NEXT: pslld %xmm2, %xmm4
+; X32-SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm3[2,3,3,3,4,5,6,7]
+; X32-SSE-NEXT: movdqa %xmm0, %xmm2
+; X32-SSE-NEXT: psrld %xmm1, %xmm2
+; X32-SSE-NEXT: pshuflw {{.*#+}} xmm5 = xmm3[0,1,1,1,4,5,6,7]
; X32-SSE-NEXT: movdqa %xmm0, %xmm1
-; X32-SSE-NEXT: pslld %xmm3, %xmm1
-; X32-SSE-NEXT: movdqa %xmm4, %xmm3
-; X32-SSE-NEXT: psrlq $32, %xmm3
+; X32-SSE-NEXT: psrld %xmm5, %xmm1
+; X32-SSE-NEXT: punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; X32-SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm3[2,3,0,1]
+; X32-SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm2[2,3,3,3,4,5,6,7]
; X32-SSE-NEXT: movdqa %xmm0, %xmm5
; X32-SSE-NEXT: psrld %xmm3, %xmm5
-; X32-SSE-NEXT: movdqa %xmm4, %xmm3
-; X32-SSE-NEXT: psrldq {{.*#+}} xmm3 = xmm3[12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; X32-SSE-NEXT: movdqa %xmm0, %xmm6
-; X32-SSE-NEXT: psrld %xmm3, %xmm6
-; X32-SSE-NEXT: movsd {{.*#+}} xmm6 = xmm5[0],xmm6[1]
-; X32-SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm6[1,3,2,3]
-; X32-SSE-NEXT: movdqa %xmm4, %xmm5
-; X32-SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm2[0],xmm5[1],xmm2[1]
-; X32-SSE-NEXT: movdqa %xmm0, %xmm6
-; X32-SSE-NEXT: psrld %xmm5, %xmm6
-; X32-SSE-NEXT: punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm2[2],xmm4[3],xmm2[3]
-; X32-SSE-NEXT: psrld %xmm4, %xmm0
-; X32-SSE-NEXT: movsd {{.*#+}} xmm0 = xmm6[0],xmm0[1]
-; X32-SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
-; X32-SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
-; X32-SSE-NEXT: por %xmm0, %xmm1
-; X32-SSE-NEXT: movdqa %xmm1, %xmm0
+; X32-SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[0,1,1,1,4,5,6,7]
+; X32-SSE-NEXT: psrld %xmm2, %xmm0
+; X32-SSE-NEXT: punpckhqdq {{.*#+}} xmm0 = xmm0[1],xmm5[1]
+; X32-SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,3],xmm0[0,3]
+; X32-SSE-NEXT: orps %xmm4, %xmm1
+; X32-SSE-NEXT: movaps %xmm1, %xmm0
; X32-SSE-NEXT: retl
%splat = shufflevector <4 x i32> %b, <4 x i32> undef, <4 x i32> zeroinitializer
%splat32 = sub <4 x i32> <i32 32, i32 32, i32 32, i32 32>, %splat
@@ -1673,18 +1651,16 @@ define <4 x i32> @constant_rotate_v4i32(<4 x i32> %a) nounwind {
; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm3[0,2,2,3]
; SSE2-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; SSE2-NEXT: movdqa %xmm0, %xmm1
-; SSE2-NEXT: psrld $27, %xmm1
+; SSE2-NEXT: psrld $25, %xmm1
; SSE2-NEXT: movdqa %xmm0, %xmm3
-; SSE2-NEXT: psrld $25, %xmm3
-; SSE2-NEXT: movsd {{.*#+}} xmm3 = xmm1[0],xmm3[1]
-; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm3[1,3,2,3]
-; SSE2-NEXT: movdqa %xmm0, %xmm3
-; SSE2-NEXT: psrld $28, %xmm3
-; SSE2-NEXT: psrld $26, %xmm0
-; SSE2-NEXT: movsd {{.*#+}} xmm0 = xmm3[0],xmm0[1]
-; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
-; SSE2-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE2-NEXT: por %xmm2, %xmm0
+; SSE2-NEXT: psrld $26, %xmm3
+; SSE2-NEXT: punpckhqdq {{.*#+}} xmm3 = xmm3[1],xmm1[1]
+; SSE2-NEXT: movdqa %xmm0, %xmm1
+; SSE2-NEXT: psrld $27, %xmm1
+; SSE2-NEXT: psrld $28, %xmm0
+; SSE2-NEXT: punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; SSE2-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,3],xmm3[0,3]
+; SSE2-NEXT: orps %xmm2, %xmm0
; SSE2-NEXT: retq
;
; SSE41-LABEL: constant_rotate_v4i32:
@@ -1755,18 +1731,16 @@ define <4 x i32> @constant_rotate_v4i32(<4 x i32> %a) nounwind {
; X32-SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm3[0,2,2,3]
; X32-SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; X32-SSE-NEXT: movdqa %xmm0, %xmm1
-; X32-SSE-NEXT: psrld $27, %xmm1
+; X32-SSE-NEXT: psrld $25, %xmm1
; X32-SSE-NEXT: movdqa %xmm0, %xmm3
-; X32-SSE-NEXT: psrld $25, %xmm3
-; X32-SSE-NEXT: movsd {{.*#+}} xmm3 = xmm1[0],xmm3[1]
-; X32-SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm3[1,3,2,3]
-; X32-SSE-NEXT: movdqa %xmm0, %xmm3
-; X32-SSE-NEXT: psrld $28, %xmm3
-; X32-SSE-NEXT: psrld $26, %xmm0
-; X32-SSE-NEXT: movsd {{.*#+}} xmm0 = xmm3[0],xmm0[1]
-; X32-SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
-; X32-SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; X32-SSE-NEXT: por %xmm2, %xmm0
+; X32-SSE-NEXT: psrld $26, %xmm3
+; X32-SSE-NEXT: punpckhqdq {{.*#+}} xmm3 = xmm3[1],xmm1[1]
+; X32-SSE-NEXT: movdqa %xmm0, %xmm1
+; X32-SSE-NEXT: psrld $27, %xmm1
+; X32-SSE-NEXT: psrld $28, %xmm0
+; X32-SSE-NEXT: punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; X32-SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,3],xmm3[0,3]
+; X32-SSE-NEXT: orps %xmm2, %xmm0
; X32-SSE-NEXT: retl
%shl = shl <4 x i32> %a, <i32 4, i32 5, i32 6, i32 7>
%lshr = lshr <4 x i32> %a, <i32 28, i32 27, i32 26, i32 25>
OpenPOWER on IntegriCloud