[X86][SSE] Combine UNPCKL with vector_shuffle into UNPCKH to save one instruction...

[oota-llvm.git] / test / CodeGen / X86 / vec_int_to_fp.ll

diff --git a/test/CodeGen/X86/vec_int_to_fp.ll b/test/CodeGen/X86/vec_int_to_fp.ll
index 32c040fc844c91db17c3d63820b2c62245d29172..fd98791815e71901103ffd4e59a9c535cdc5ab02 100644 (file)
--- a/test/CodeGen/X86/vec_int_to_fp.ll
+++ b/test/CodeGen/X86/vec_int_to_fp.ll
@@ -1035,8 +1035,7 @@ define <8 x float> @sitofp_8i16_to_8f32(<8 x i16> %a) {
 ; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
 ; SSE-NEXT:    psrad $16, %xmm1
 ; SSE-NEXT:    cvtdq2ps %xmm1, %xmm2
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3]
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4,4,5,5,6,6,7,7]
 ; SSE-NEXT:    psrad $16, %xmm0
 ; SSE-NEXT:    cvtdq2ps %xmm0, %xmm1
 ; SSE-NEXT:    movaps %xmm2, %xmm0
@@ -1820,15 +1819,14 @@ define void @aggregate_sitofp_8i16_to_8f32(%Arguments* nocapture readonly %a0) {
 ; SSE:       # BB#0:
 ; SSE-NEXT:    movq 24(%rdi), %rax
 ; SSE-NEXT:    movdqu 8(%rdi), %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0,0,1,1,2,2,3,3]
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
 ; SSE-NEXT:    psrad $16, %xmm1
 ; SSE-NEXT:    cvtdq2ps %xmm1, %xmm1
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3]
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4,4,5,5,6,6,7,7]
 ; SSE-NEXT:    psrad $16, %xmm0
 ; SSE-NEXT:    cvtdq2ps %xmm0, %xmm0
-; SSE-NEXT:    movaps %xmm0, (%rax)
-; SSE-NEXT:    movaps %xmm1, 16(%rax)
+; SSE-NEXT:    movaps %xmm0, 16(%rax)
+; SSE-NEXT:    movaps %xmm1, (%rax)
 ; SSE-NEXT:    retq
 ;
 ; AVX1-LABEL: aggregate_sitofp_8i16_to_8f32: