diff options
Diffstat (limited to 'llvm/test/CodeGen/X86/widen_conv-4.ll')
| -rw-r--r-- | llvm/test/CodeGen/X86/widen_conv-4.ll | 32 |
1 files changed, 11 insertions, 21 deletions
diff --git a/llvm/test/CodeGen/X86/widen_conv-4.ll b/llvm/test/CodeGen/X86/widen_conv-4.ll index 3f8e203d407..5bdd7562d1e 100644 --- a/llvm/test/CodeGen/X86/widen_conv-4.ll +++ b/llvm/test/CodeGen/X86/widen_conv-4.ll @@ -77,7 +77,6 @@ define void @convert_v3i8_to_v3f32(<3 x float>* %dst.addr, <3 x i8>* %src.addr) ; X86-SSE2: # %bb.0: # %entry ; X86-SSE2-NEXT: pushl %ebp ; X86-SSE2-NEXT: movl %esp, %ebp -; X86-SSE2-NEXT: pushl %esi ; X86-SSE2-NEXT: andl $-16, %esp ; X86-SSE2-NEXT: subl $32, %esp ; X86-SSE2-NEXT: movl 8(%ebp), %eax @@ -88,15 +87,11 @@ define void @convert_v3i8_to_v3f32(<3 x float>* %dst.addr, <3 x i8>* %src.addr) ; X86-SSE2-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3] ; X86-SSE2-NEXT: movzbl 2(%ecx), %ecx ; X86-SSE2-NEXT: movdqa %xmm0, (%esp) -; X86-SSE2-NEXT: movl {{[0-9]+}}(%esp), %edx -; X86-SSE2-NEXT: shll $8, %edx -; X86-SSE2-NEXT: movzbl (%esp), %esi -; X86-SSE2-NEXT: orl %edx, %esi -; X86-SSE2-NEXT: movd %esi, %xmm0 -; X86-SSE2-NEXT: pinsrw $1, %ecx, %xmm0 -; X86-SSE2-NEXT: pxor %xmm1, %xmm1 -; X86-SSE2-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7] -; X86-SSE2-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3] +; X86-SSE2-NEXT: movzbl (%esp), %edx +; X86-SSE2-NEXT: movd %edx, %xmm0 +; X86-SSE2-NEXT: movzbl {{[0-9]+}}(%esp), %edx +; X86-SSE2-NEXT: pinsrw $2, %edx, %xmm0 +; X86-SSE2-NEXT: pinsrw $4, %ecx, %xmm0 ; X86-SSE2-NEXT: cvtdq2ps %xmm0, %xmm0 ; X86-SSE2-NEXT: movss %xmm0, (%eax) ; X86-SSE2-NEXT: movaps %xmm0, %xmm1 @@ -104,8 +99,7 @@ define void @convert_v3i8_to_v3f32(<3 x float>* %dst.addr, <3 x i8>* %src.addr) ; X86-SSE2-NEXT: movss %xmm1, 8(%eax) ; X86-SSE2-NEXT: shufps {{.*#+}} xmm0 = xmm0[1,1,2,3] ; X86-SSE2-NEXT: movss %xmm0, 4(%eax) -; X86-SSE2-NEXT: leal -4(%ebp), %esp -; X86-SSE2-NEXT: popl %esi +; X86-SSE2-NEXT: movl %ebp, %esp ; X86-SSE2-NEXT: popl %ebp ; X86-SSE2-NEXT: retl ; @@ -135,15 +129,11 @@ define void @convert_v3i8_to_v3f32(<3 x float>* %dst.addr, <3 x i8>* %src.addr) ; X64-SSE2-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3] ; X64-SSE2-NEXT: movzbl 2(%rsi), %eax ; X64-SSE2-NEXT: movdqa %xmm0, -{{[0-9]+}}(%rsp) -; X64-SSE2-NEXT: movl -{{[0-9]+}}(%rsp), %ecx -; X64-SSE2-NEXT: shll $8, %ecx -; X64-SSE2-NEXT: movzbl -{{[0-9]+}}(%rsp), %edx -; X64-SSE2-NEXT: orl %ecx, %edx -; X64-SSE2-NEXT: movd %edx, %xmm0 -; X64-SSE2-NEXT: pinsrw $1, %eax, %xmm0 -; X64-SSE2-NEXT: pxor %xmm1, %xmm1 -; X64-SSE2-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7] -; X64-SSE2-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3] +; X64-SSE2-NEXT: movzbl -{{[0-9]+}}(%rsp), %ecx +; X64-SSE2-NEXT: movd %ecx, %xmm0 +; X64-SSE2-NEXT: movzbl -{{[0-9]+}}(%rsp), %ecx +; X64-SSE2-NEXT: pinsrw $2, %ecx, %xmm0 +; X64-SSE2-NEXT: pinsrw $4, %eax, %xmm0 ; X64-SSE2-NEXT: cvtdq2ps %xmm0, %xmm0 ; X64-SSE2-NEXT: movlps %xmm0, (%rdi) ; X64-SSE2-NEXT: movhlps {{.*#+}} xmm0 = xmm0[1,1] |
