1 ; RUN: llc < %s -mtriple=x86_64-apple-darwin -mattr=+mmx,+sse2 | FileCheck %s
2 ; There are no MMX operations here, so we use XMM or i64.
5 define void @ti8(double %a, double %b) nounwind {
7 %tmp1 = bitcast double %a to <8 x i8>
8 %tmp2 = bitcast double %b to <8 x i8>
9 %tmp3 = add <8 x i8> %tmp1, %tmp2
11 store <8 x i8> %tmp3, <8 x i8>* null
16 define void @ti16(double %a, double %b) nounwind {
18 %tmp1 = bitcast double %a to <4 x i16>
19 %tmp2 = bitcast double %b to <4 x i16>
20 %tmp3 = add <4 x i16> %tmp1, %tmp2
22 store <4 x i16> %tmp3, <4 x i16>* null
27 define void @ti32(double %a, double %b) nounwind {
29 %tmp1 = bitcast double %a to <2 x i32>
30 %tmp2 = bitcast double %b to <2 x i32>
31 %tmp3 = add <2 x i32> %tmp1, %tmp2
33 store <2 x i32> %tmp3, <2 x i32>* null
38 define void @ti64(double %a, double %b) nounwind {
40 %tmp1 = bitcast double %a to <1 x i64>
41 %tmp2 = bitcast double %b to <1 x i64>
42 %tmp3 = add <1 x i64> %tmp1, %tmp2
44 store <1 x i64> %tmp3, <1 x i64>* null
48 ; MMX intrinsics calls get us MMX instructions.
50 define void @ti8a(double %a, double %b) nounwind {
52 %tmp1 = bitcast double %a to x86_mmx
54 %tmp2 = bitcast double %b to x86_mmx
56 %tmp3 = tail call x86_mmx @llvm.x86.mmx.padd.b(x86_mmx %tmp1, x86_mmx %tmp2)
57 store x86_mmx %tmp3, x86_mmx* null
62 define void @ti16a(double %a, double %b) nounwind {
64 %tmp1 = bitcast double %a to x86_mmx
66 %tmp2 = bitcast double %b to x86_mmx
68 %tmp3 = tail call x86_mmx @llvm.x86.mmx.padd.w(x86_mmx %tmp1, x86_mmx %tmp2)
69 store x86_mmx %tmp3, x86_mmx* null
74 define void @ti32a(double %a, double %b) nounwind {
76 %tmp1 = bitcast double %a to x86_mmx
78 %tmp2 = bitcast double %b to x86_mmx
80 %tmp3 = tail call x86_mmx @llvm.x86.mmx.padd.d(x86_mmx %tmp1, x86_mmx %tmp2)
81 store x86_mmx %tmp3, x86_mmx* null
86 define void @ti64a(double %a, double %b) nounwind {
88 %tmp1 = bitcast double %a to x86_mmx
90 %tmp2 = bitcast double %b to x86_mmx
92 %tmp3 = tail call x86_mmx @llvm.x86.mmx.padd.q(x86_mmx %tmp1, x86_mmx %tmp2)
93 store x86_mmx %tmp3, x86_mmx* null
97 declare x86_mmx @llvm.x86.mmx.padd.b(x86_mmx, x86_mmx)
98 declare x86_mmx @llvm.x86.mmx.padd.w(x86_mmx, x86_mmx)
99 declare x86_mmx @llvm.x86.mmx.padd.d(x86_mmx, x86_mmx)
100 declare x86_mmx @llvm.x86.mmx.padd.q(x86_mmx, x86_mmx)