1 ; RUN: llc < %s -march=arm -mattr=+neon | FileCheck %s
3 define void @vst2lanei8(i8* %A, <8 x i8>* %B) nounwind {
6 %tmp1 = load <8 x i8>* %B
7 call void @llvm.arm.neon.vst2lane.v8i8(i8* %A, <8 x i8> %tmp1, <8 x i8> %tmp1, i32 1)
11 define void @vst2lanei16(i16* %A, <4 x i16>* %B) nounwind {
14 %tmp1 = load <4 x i16>* %B
15 call void @llvm.arm.neon.vst2lane.v4i16(i16* %A, <4 x i16> %tmp1, <4 x i16> %tmp1, i32 1)
19 define void @vst2lanei32(i32* %A, <2 x i32>* %B) nounwind {
22 %tmp1 = load <2 x i32>* %B
23 call void @llvm.arm.neon.vst2lane.v2i32(i32* %A, <2 x i32> %tmp1, <2 x i32> %tmp1, i32 1)
27 define void @vst2lanef(float* %A, <2 x float>* %B) nounwind {
30 %tmp1 = load <2 x float>* %B
31 call void @llvm.arm.neon.vst2lane.v2f32(float* %A, <2 x float> %tmp1, <2 x float> %tmp1, i32 1)
35 define void @vst2laneQi16(i16* %A, <8 x i16>* %B) nounwind {
38 %tmp1 = load <8 x i16>* %B
39 call void @llvm.arm.neon.vst2lane.v8i16(i16* %A, <8 x i16> %tmp1, <8 x i16> %tmp1, i32 1)
43 define void @vst2laneQi32(i32* %A, <4 x i32>* %B) nounwind {
46 %tmp1 = load <4 x i32>* %B
47 call void @llvm.arm.neon.vst2lane.v4i32(i32* %A, <4 x i32> %tmp1, <4 x i32> %tmp1, i32 2)
51 define void @vst2laneQf(float* %A, <4 x float>* %B) nounwind {
54 %tmp1 = load <4 x float>* %B
55 call void @llvm.arm.neon.vst2lane.v4f32(float* %A, <4 x float> %tmp1, <4 x float> %tmp1, i32 3)
59 declare void @llvm.arm.neon.vst2lane.v8i8(i8*, <8 x i8>, <8 x i8>, i32) nounwind
60 declare void @llvm.arm.neon.vst2lane.v4i16(i8*, <4 x i16>, <4 x i16>, i32) nounwind
61 declare void @llvm.arm.neon.vst2lane.v2i32(i8*, <2 x i32>, <2 x i32>, i32) nounwind
62 declare void @llvm.arm.neon.vst2lane.v2f32(i8*, <2 x float>, <2 x float>, i32) nounwind
64 declare void @llvm.arm.neon.vst2lane.v8i16(i8*, <8 x i16>, <8 x i16>, i32) nounwind
65 declare void @llvm.arm.neon.vst2lane.v4i32(i8*, <4 x i32>, <4 x i32>, i32) nounwind
66 declare void @llvm.arm.neon.vst2lane.v4f32(i8*, <4 x float>, <4 x float>, i32) nounwind
68 define void @vst3lanei8(i8* %A, <8 x i8>* %B) nounwind {
71 %tmp1 = load <8 x i8>* %B
72 call void @llvm.arm.neon.vst3lane.v8i8(i8* %A, <8 x i8> %tmp1, <8 x i8> %tmp1, <8 x i8> %tmp1, i32 1)
76 define void @vst3lanei16(i16* %A, <4 x i16>* %B) nounwind {
79 %tmp1 = load <4 x i16>* %B
80 call void @llvm.arm.neon.vst3lane.v4i16(i16* %A, <4 x i16> %tmp1, <4 x i16> %tmp1, <4 x i16> %tmp1, i32 1)
84 define void @vst3lanei32(i32* %A, <2 x i32>* %B) nounwind {
87 %tmp1 = load <2 x i32>* %B
88 call void @llvm.arm.neon.vst3lane.v2i32(i32* %A, <2 x i32> %tmp1, <2 x i32> %tmp1, <2 x i32> %tmp1, i32 1)
92 define void @vst3lanef(float* %A, <2 x float>* %B) nounwind {
95 %tmp1 = load <2 x float>* %B
96 call void @llvm.arm.neon.vst3lane.v2f32(float* %A, <2 x float> %tmp1, <2 x float> %tmp1, <2 x float> %tmp1, i32 1)
100 define void @vst3laneQi16(i16* %A, <8 x i16>* %B) nounwind {
101 ;CHECK: vst3laneQi16:
103 %tmp1 = load <8 x i16>* %B
104 call void @llvm.arm.neon.vst3lane.v8i16(i16* %A, <8 x i16> %tmp1, <8 x i16> %tmp1, <8 x i16> %tmp1, i32 6)
108 define void @vst3laneQi32(i32* %A, <4 x i32>* %B) nounwind {
109 ;CHECK: vst3laneQi32:
111 %tmp1 = load <4 x i32>* %B
112 call void @llvm.arm.neon.vst3lane.v4i32(i32* %A, <4 x i32> %tmp1, <4 x i32> %tmp1, <4 x i32> %tmp1, i32 0)
116 define void @vst3laneQf(float* %A, <4 x float>* %B) nounwind {
119 %tmp1 = load <4 x float>* %B
120 call void @llvm.arm.neon.vst3lane.v4f32(float* %A, <4 x float> %tmp1, <4 x float> %tmp1, <4 x float> %tmp1, i32 1)
124 declare void @llvm.arm.neon.vst3lane.v8i8(i8*, <8 x i8>, <8 x i8>, <8 x i8>, i32) nounwind
125 declare void @llvm.arm.neon.vst3lane.v4i16(i8*, <4 x i16>, <4 x i16>, <4 x i16>, i32) nounwind
126 declare void @llvm.arm.neon.vst3lane.v2i32(i8*, <2 x i32>, <2 x i32>, <2 x i32>, i32) nounwind
127 declare void @llvm.arm.neon.vst3lane.v2f32(i8*, <2 x float>, <2 x float>, <2 x float>, i32) nounwind
129 declare void @llvm.arm.neon.vst3lane.v8i16(i8*, <8 x i16>, <8 x i16>, <8 x i16>, i32) nounwind
130 declare void @llvm.arm.neon.vst3lane.v4i32(i8*, <4 x i32>, <4 x i32>, <4 x i32>, i32) nounwind
131 declare void @llvm.arm.neon.vst3lane.v4f32(i8*, <4 x float>, <4 x float>, <4 x float>, i32) nounwind
134 define void @vst4lanei8(i8* %A, <8 x i8>* %B) nounwind {
137 %tmp1 = load <8 x i8>* %B
138 call void @llvm.arm.neon.vst4lane.v8i8(i8* %A, <8 x i8> %tmp1, <8 x i8> %tmp1, <8 x i8> %tmp1, <8 x i8> %tmp1, i32 1)
142 define void @vst4lanei16(i16* %A, <4 x i16>* %B) nounwind {
145 %tmp1 = load <4 x i16>* %B
146 call void @llvm.arm.neon.vst4lane.v4i16(i16* %A, <4 x i16> %tmp1, <4 x i16> %tmp1, <4 x i16> %tmp1, <4 x i16> %tmp1, i32 1)
150 define void @vst4lanei32(i32* %A, <2 x i32>* %B) nounwind {
153 %tmp1 = load <2 x i32>* %B
154 call void @llvm.arm.neon.vst4lane.v2i32(i32* %A, <2 x i32> %tmp1, <2 x i32> %tmp1, <2 x i32> %tmp1, <2 x i32> %tmp1, i32 1)
158 define void @vst4lanef(float* %A, <2 x float>* %B) nounwind {
161 %tmp1 = load <2 x float>* %B
162 call void @llvm.arm.neon.vst4lane.v2f32(float* %A, <2 x float> %tmp1, <2 x float> %tmp1, <2 x float> %tmp1, <2 x float> %tmp1, i32 1)
166 define void @vst4laneQi16(i16* %A, <8 x i16>* %B) nounwind {
167 ;CHECK: vst4laneQi16:
169 %tmp1 = load <8 x i16>* %B
170 call void @llvm.arm.neon.vst4lane.v8i16(i16* %A, <8 x i16> %tmp1, <8 x i16> %tmp1, <8 x i16> %tmp1, <8 x i16> %tmp1, i32 7)
174 define void @vst4laneQi32(i32* %A, <4 x i32>* %B) nounwind {
175 ;CHECK: vst4laneQi32:
177 %tmp1 = load <4 x i32>* %B
178 call void @llvm.arm.neon.vst4lane.v4i32(i32* %A, <4 x i32> %tmp1, <4 x i32> %tmp1, <4 x i32> %tmp1, <4 x i32> %tmp1, i32 2)
182 define void @vst4laneQf(float* %A, <4 x float>* %B) nounwind {
185 %tmp1 = load <4 x float>* %B
186 call void @llvm.arm.neon.vst4lane.v4f32(float* %A, <4 x float> %tmp1, <4 x float> %tmp1, <4 x float> %tmp1, <4 x float> %tmp1, i32 1)
190 declare void @llvm.arm.neon.vst4lane.v8i8(i8*, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, i32) nounwind
191 declare void @llvm.arm.neon.vst4lane.v4i16(i8*, <4 x i16>, <4 x i16>, <4 x i16>, <4 x i16>, i32) nounwind
192 declare void @llvm.arm.neon.vst4lane.v2i32(i8*, <2 x i32>, <2 x i32>, <2 x i32>, <2 x i32>, i32) nounwind
193 declare void @llvm.arm.neon.vst4lane.v2f32(i8*, <2 x float>, <2 x float>, <2 x float>, <2 x float>, i32) nounwind
195 declare void @llvm.arm.neon.vst4lane.v8i16(i8*, <8 x i16>, <8 x i16>, <8 x i16>, <8 x i16>, i32) nounwind
196 declare void @llvm.arm.neon.vst4lane.v4i32(i8*, <4 x i32>, <4 x i32>, <4 x i32>, <4 x i32>, i32) nounwind
197 declare void @llvm.arm.neon.vst4lane.v4f32(i8*, <4 x float>, <4 x float>, <4 x float>, <4 x float>, i32) nounwind