xref: /llvm-project/llvm/test/CodeGen/X86/inline-sse.ll (revision 635f6d384596950e73b2485842c587a2954c655f)
1; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+sse | FileCheck %s --check-prefix=X86
2; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+sse2 | FileCheck %s --check-prefix=X86
3; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=-sse2 | FileCheck %s --check-prefix=X64
4; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse2 | FileCheck %s --check-prefix=X64
5
6; PR16133 - we must treat XMM registers as v4f32 as SSE1 targets don't permit other vector types.
7
8define void @nop() nounwind {
9; X86-LABEL: nop:
10; X86:       # %bb.0:
11; X86-NEXT:    pushl %ebp
12; X86-NEXT:    movl %esp, %ebp
13; X86-NEXT:    andl $-16, %esp
14; X86-NEXT:    subl $32, %esp
15; X86-NEXT:    #APP
16; X86-NEXT:    #NO_APP
17; X86-NEXT:    movaps %xmm0, (%esp)
18; X86-NEXT:    movl %ebp, %esp
19; X86-NEXT:    popl %ebp
20; X86-NEXT:    retl
21;
22; X64-LABEL: nop:
23; X64:       # %bb.0:
24; X64-NEXT:    #APP
25; X64-NEXT:    #NO_APP
26; X64-NEXT:    movaps %xmm0, -{{[0-9]+}}(%rsp)
27; X64-NEXT:    retq
28  %1 = alloca <4 x float>, align 16
29  %2 = call <4 x float> asm "", "=x,~{dirflag},~{fpsr},~{flags}"()
30  store <4 x float> %2, ptr %1, align 16
31  ret void
32}
33