; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=i686-darwin -mattr=+mmx,+sse2 | FileCheck --check-prefix=X86 %s
; RUN: llc < %s -mtriple=x86_64-darwin -mattr=+mmx,+sse2 | FileCheck --check-prefix=X64 %s
; If there is no explicit MMX type usage, always promote to XMM.
define void @test0(ptr %x) nounwind {
; X86-LABEL: test0:
; X86: ## %bb.0: ## %entry
; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
; X86-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; X86-NEXT: shufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
; X86-NEXT: movlps %xmm0, (%eax)
; X86-NEXT: retl
;
; X64-LABEL: test0:
; X64: ## %bb.0: ## %entry
; X64-NEXT: movq {{.*#+}} xmm0 = mem[0],zero
; X64-NEXT: pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
; X64-NEXT: movq %xmm0, (%rdi)
; X64-NEXT: retq
entry:
%tmp2 = load <1 x i64>, ptr %x
%tmp6 = bitcast <1 x i64> %tmp2 to <2 x i32>
%tmp9 = shufflevector <2 x i32> %tmp6, <2 x i32> undef, <2 x i32> < i32 1, i32 1 >
%tmp10 = bitcast <2 x i32> %tmp9 to <1 x i64>
store <1 x i64> %tmp10, ptr %x
ret void
}
define void @test1() nounwind {
; X86-LABEL: test1:
; X86: ## %bb.0: ## %entry
; X86-NEXT: pushl %edi
; X86-NEXT: pxor %mm0, %mm0
; X86-NEXT: movq {{\.?LCPI[0-9]+_[0-9]+}}, %mm1 ## mm1 = 0x7070606040400000
; X86-NEXT: xorl %edi, %edi
; X86-NEXT: maskmovq %mm1, %mm0
; X86-NEXT: popl %edi
; X86-NEXT: retl
;
; X64-LABEL: test1:
; X64: ## %bb.0: ## %entry
; X64-NEXT: pxor %mm0, %mm0
; X64-NEXT: movq {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %mm1 ## mm1 = 0x7070606040400000
; X64-NEXT: xorl %edi, %edi
; X64-NEXT: maskmovq %mm1, %mm0
; X64-NEXT: retq
entry:
%tmp528 = bitcast <8 x i8> zeroinitializer to <2 x i32>
%tmp529 = and <2 x i32> %tmp528, bitcast (<4 x i16> < i16 -32640, i16 16448, i16 8224, i16 4112 > to <2 x i32>)
%tmp542 = bitcast <2 x i32> %tmp529 to <4 x i16>
%tmp543 = add <4 x i16> %tmp542, < i16 0, i16 16448, i16 24672, i16 28784 >
%tmp555 = bitcast <4 x i16> %tmp543 to <8 x i8>
%tmp556 = bitcast <8 x i8> %tmp555 to <1 x i64>
%tmp557 = bitcast <8 x i8> zeroinitializer to <1 x i64>
tail call void @llvm.x86.mmx.maskmovq( <1 x i64> %tmp557, <1 x i64> %tmp556, ptr null)
ret void
}
@tmp_V2i = common global <2 x i32> zeroinitializer
define void @test2() nounwind {
; X86-LABEL: test2:
; X86: ## %bb.0: ## %entry
; X86-NEXT: movl L_tmp_V2i$non_lazy_ptr, %eax
; X86-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; X86-NEXT: unpcklps {{.*#+}} xmm0 = xmm0[0,0,1,1]
; X86-NEXT: movlps %xmm0, (%eax)
; X86-NEXT: retl
;
; X64-LABEL: test2:
; X64: ## %bb.0: ## %entry
; X64-NEXT: movq _tmp_V2i@GOTPCREL(%rip), %rax
; X64-NEXT: movd {{.*#+}} xmm0 = mem[0],zero,zero,zero
; X64-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
; X64-NEXT: movq %xmm0, (%rax)
; X64-NEXT: retq
entry:
%0 = load <2 x i32>, ptr @tmp_V2i, align 8
%1 = shufflevector <2 x i32> %0, <2 x i32> undef, <2 x i32> zeroinitializer
store <2 x i32> %1, ptr @tmp_V2i, align 8
ret void
}
define <4 x float> @pr35869() nounwind {
; X86-LABEL: pr35869:
; X86: ## %bb.0:
; X86-NEXT: movl $64, %eax
; X86-NEXT: movd %eax, %mm0
; X86-NEXT: pxor %mm1, %mm1
; X86-NEXT: punpcklbw %mm1, %mm0 ## mm0 = mm0[0],mm1[0],mm0[1],mm1[1],mm0[2],mm1[2],mm0[3],mm1[3]
; X86-NEXT: pcmpgtw %mm0, %mm1
; X86-NEXT: movq %mm0, %mm2
; X86-NEXT: punpckhwd %mm1, %mm2 ## mm2 = mm2[2],mm1[2],mm2[3],mm1[3]
; X86-NEXT: xorps %xmm0, %xmm0
; X86-NEXT: cvtpi2ps %mm2, %xmm0
; X86-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0,0]
; X86-NEXT: punpcklwd %mm1, %mm0 ## mm0 = mm0[0],mm1[0],mm0[1],mm1[1]
; X86-NEXT: cvtpi2ps %mm0, %xmm0
; X86-NEXT: retl
;
; X64-LABEL: pr35869:
; X64: ## %bb.0:
; X64-NEXT: movl $64, %eax
; X64-NEXT: movd %eax, %mm0
; X64-NEXT: pxor %mm1, %mm1
; X64-NEXT: punpcklbw %mm1, %mm0 ## mm0 = mm0[0],mm1[0],mm0[1],mm1[1],mm0[2],mm1[2],mm0[3],mm1[3]
; X64-NEXT: pcmpgtw %mm0, %mm1
; X64-NEXT: movq %mm0, %mm2
; X64-NEXT: punpckhwd %mm1, %mm2 ## mm2 = mm2[2],mm1[2],mm2[3],mm1[3]
; X64-NEXT: xorps %xmm0, %xmm0
; X64-NEXT: cvtpi2ps %mm2, %xmm0
; X64-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0,0]
; X64-NEXT: punpcklwd %mm1, %mm0 ## mm0 = mm0[0],mm1[0],mm0[1],mm1[1]
; X64-NEXT: cvtpi2ps %mm0, %xmm0
; X64-NEXT: retq
%1 = tail call <1 x i64> @llvm.x86.mmx.punpcklbw(<1 x i64> bitcast (<8 x i8> <i8 64, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0> to <1 x i64>), <1 x i64> bitcast (<8 x i8> zeroinitializer to <1 x i64>))
%2 = tail call <1 x i64> @llvm.x86.mmx.pcmpgt.w(<1 x i64> bitcast (<4 x i16> zeroinitializer to <1 x i64>), <1 x i64> %1)
%3 = tail call <1 x i64> @llvm.x86.mmx.punpckhwd(<1 x i64> %1, <1 x i64> %2)
%4 = tail call <4 x float> @llvm.x86.sse.cvtpi2ps(<4 x float> zeroinitializer, <1 x i64> %3)
%5 = shufflevector <4 x float> %4, <4 x float> undef, <4 x i32> <i32 0, i32 1, i32 0, i32 1>
%6 = tail call <1 x i64> @llvm.x86.mmx.punpcklwd(<1 x i64> %1, <1 x i64> %2)
%7 = tail call <4 x float> @llvm.x86.sse.cvtpi2ps(<4 x float> %5, <1 x i64> %6)
ret <4 x float> %7
}
declare void @llvm.x86.mmx.maskmovq(<1 x i64>, <1 x i64>, ptr)
declare <1 x i64> @llvm.x86.mmx.pcmpgt.w(<1 x i64>, <1 x i64>)
declare <1 x i64> @llvm.x86.mmx.punpcklbw(<1 x i64>, <1 x i64>)
declare <1 x i64> @llvm.x86.mmx.punpcklwd(<1 x i64>, <1 x i64>)
declare <1 x i64> @llvm.x86.mmx.punpckhwd(<1 x i64>, <1 x i64>)
declare <4 x float> @llvm.x86.sse.cvtpi2ps(<4 x float>, <1 x i64>)
Codebase Browser
llvm