[X86] Use compare with 0 to fill an element with sign bits when sign extending to v2i64 pre-sse4.1

Previously we used an arithmetic shift right by 31, but that requires a copy to preserve the input. So we might as well materialize a zero and compare to it since the comparison will overwrite the register that contains the zeros. This should be one byte shorter.

llvm-svn: 347181

1 files changed, 34 insertions, 33 deletions

diff --git a/llvm/test/CodeGen/X86/vector-trunc-math.ll b/llvm/test/CodeGen/X86/vector-trunc-math.ll
index c8ccab02614..e3b3ee4bb22 100644
--- a/llvm/test/CodeGen/X86/vector-trunc-math.ll
+++ b/llvm/test/CodeGen/X86/vector-trunc-math.ll
@@ -5569,39 +5569,40 @@ define <4 x i32> @mul_add_const_v4i64_v4i32(<4 x i32> %a0, <4 x i32> %a1) nounwi
 define <4 x i32> @mul_add_self_v4i64_v4i32(<4 x i32> %a0, <4 x i32> %a1) nounwind {
 ; SSE-LABEL: mul_add_self_v4i64_v4i32:
 ; SSE:       # %bb.0:
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[2,3,0,1]
-; SSE-NEXT:    movdqa %xmm3, %xmm4
-; SSE-NEXT:    psrad $31, %xmm4
-; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
-; SSE-NEXT:    movdqa %xmm0, %xmm5
-; SSE-NEXT:    psrad $31, %xmm5
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm1[2,3,0,1]
-; SSE-NEXT:    movdqa %xmm6, %xmm7
-; SSE-NEXT:    psrad $31, %xmm7
-; SSE-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm7[0],xmm6[1],xmm7[1]
-; SSE-NEXT:    movdqa %xmm1, %xmm2
-; SSE-NEXT:    psrad $31, %xmm2
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,1,3]
-; SSE-NEXT:    pmuludq %xmm0, %xmm2
-; SSE-NEXT:    pmuludq %xmm1, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,1,1,3]
-; SSE-NEXT:    pmuludq %xmm1, %xmm5
-; SSE-NEXT:    paddq %xmm5, %xmm2
-; SSE-NEXT:    psllq $32, %xmm2
-; SSE-NEXT:    paddq %xmm0, %xmm2
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[0,1,1,3]
-; SSE-NEXT:    pmuludq %xmm3, %xmm0
-; SSE-NEXT:    pmuludq %xmm6, %xmm3
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[0,1,1,3]
-; SSE-NEXT:    pmuludq %xmm6, %xmm1
-; SSE-NEXT:    paddq %xmm1, %xmm0
-; SSE-NEXT:    psllq $32, %xmm0
-; SSE-NEXT:    paddq %xmm3, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,2],xmm0[0,2]
-; SSE-NEXT:    paddd %xmm2, %xmm2
-; SSE-NEXT:    movdqa %xmm2, %xmm0
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[2,3,0,1]
+; SSE-NEXT:    pxor %xmm8, %xmm8
+; SSE-NEXT:    pxor %xmm3, %xmm3
+; SSE-NEXT:    pcmpgtd %xmm2, %xmm3
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
+; SSE-NEXT:    pxor %xmm7, %xmm7
+; SSE-NEXT:    pcmpgtd %xmm0, %xmm7
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm7[0],xmm0[1],xmm7[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm1[2,3,0,1]
+; SSE-NEXT:    pxor %xmm6, %xmm6
+; SSE-NEXT:    pcmpgtd %xmm4, %xmm6
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm6[0],xmm4[1],xmm6[1]
+; SSE-NEXT:    pxor %xmm5, %xmm5
+; SSE-NEXT:    pcmpgtd %xmm1, %xmm5
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm5[0],xmm1[1],xmm5[1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm8[0],xmm7[1],xmm8[1]
+; SSE-NEXT:    pmuludq %xmm1, %xmm7
+; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm8[0],xmm5[1],xmm8[1]
+; SSE-NEXT:    pmuludq %xmm0, %xmm5
+; SSE-NEXT:    paddq %xmm7, %xmm5
+; SSE-NEXT:    psllq $32, %xmm5
+; SSE-NEXT:    pmuludq %xmm0, %xmm1
+; SSE-NEXT:    paddq %xmm5, %xmm1
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm8[0],xmm3[1],xmm8[1]
+; SSE-NEXT:    pmuludq %xmm4, %xmm3
+; SSE-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm8[0],xmm6[1],xmm8[1]
+; SSE-NEXT:    pmuludq %xmm2, %xmm6
+; SSE-NEXT:    paddq %xmm3, %xmm6
+; SSE-NEXT:    psllq $32, %xmm6
+; SSE-NEXT:    pmuludq %xmm2, %xmm4
+; SSE-NEXT:    paddq %xmm6, %xmm4
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,2],xmm4[0,2]
+; SSE-NEXT:    paddd %xmm1, %xmm1
+; SSE-NEXT:    movdqa %xmm1, %xmm0
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: mul_add_self_v4i64_v4i32: