1 ; RUN: llc < %s -march=arm -mattr=+neon | FileCheck %s
3 define <8 x i8> @vrshls8(<8 x i8>* %A, <8 x i8>* %B) nounwind {
6 %tmp1 = load <8 x i8>* %A
7 %tmp2 = load <8 x i8>* %B
8 %tmp3 = call <8 x i8> @llvm.arm.neon.vrshifts.v8i8(<8 x i8> %tmp1, <8 x i8> %tmp2)
12 define <4 x i16> @vrshls16(<4 x i16>* %A, <4 x i16>* %B) nounwind {
15 %tmp1 = load <4 x i16>* %A
16 %tmp2 = load <4 x i16>* %B
17 %tmp3 = call <4 x i16> @llvm.arm.neon.vrshifts.v4i16(<4 x i16> %tmp1, <4 x i16> %tmp2)
21 define <2 x i32> @vrshls32(<2 x i32>* %A, <2 x i32>* %B) nounwind {
24 %tmp1 = load <2 x i32>* %A
25 %tmp2 = load <2 x i32>* %B
26 %tmp3 = call <2 x i32> @llvm.arm.neon.vrshifts.v2i32(<2 x i32> %tmp1, <2 x i32> %tmp2)
30 define <1 x i64> @vrshls64(<1 x i64>* %A, <1 x i64>* %B) nounwind {
33 %tmp1 = load <1 x i64>* %A
34 %tmp2 = load <1 x i64>* %B
35 %tmp3 = call <1 x i64> @llvm.arm.neon.vrshifts.v1i64(<1 x i64> %tmp1, <1 x i64> %tmp2)
39 define <8 x i8> @vrshlu8(<8 x i8>* %A, <8 x i8>* %B) nounwind {
42 %tmp1 = load <8 x i8>* %A
43 %tmp2 = load <8 x i8>* %B
44 %tmp3 = call <8 x i8> @llvm.arm.neon.vrshiftu.v8i8(<8 x i8> %tmp1, <8 x i8> %tmp2)
48 define <4 x i16> @vrshlu16(<4 x i16>* %A, <4 x i16>* %B) nounwind {
51 %tmp1 = load <4 x i16>* %A
52 %tmp2 = load <4 x i16>* %B
53 %tmp3 = call <4 x i16> @llvm.arm.neon.vrshiftu.v4i16(<4 x i16> %tmp1, <4 x i16> %tmp2)
57 define <2 x i32> @vrshlu32(<2 x i32>* %A, <2 x i32>* %B) nounwind {
60 %tmp1 = load <2 x i32>* %A
61 %tmp2 = load <2 x i32>* %B
62 %tmp3 = call <2 x i32> @llvm.arm.neon.vrshiftu.v2i32(<2 x i32> %tmp1, <2 x i32> %tmp2)
66 define <1 x i64> @vrshlu64(<1 x i64>* %A, <1 x i64>* %B) nounwind {
69 %tmp1 = load <1 x i64>* %A
70 %tmp2 = load <1 x i64>* %B
71 %tmp3 = call <1 x i64> @llvm.arm.neon.vrshiftu.v1i64(<1 x i64> %tmp1, <1 x i64> %tmp2)
75 define <16 x i8> @vrshlQs8(<16 x i8>* %A, <16 x i8>* %B) nounwind {
78 %tmp1 = load <16 x i8>* %A
79 %tmp2 = load <16 x i8>* %B
80 %tmp3 = call <16 x i8> @llvm.arm.neon.vrshifts.v16i8(<16 x i8> %tmp1, <16 x i8> %tmp2)
84 define <8 x i16> @vrshlQs16(<8 x i16>* %A, <8 x i16>* %B) nounwind {
87 %tmp1 = load <8 x i16>* %A
88 %tmp2 = load <8 x i16>* %B
89 %tmp3 = call <8 x i16> @llvm.arm.neon.vrshifts.v8i16(<8 x i16> %tmp1, <8 x i16> %tmp2)
93 define <4 x i32> @vrshlQs32(<4 x i32>* %A, <4 x i32>* %B) nounwind {
96 %tmp1 = load <4 x i32>* %A
97 %tmp2 = load <4 x i32>* %B
98 %tmp3 = call <4 x i32> @llvm.arm.neon.vrshifts.v4i32(<4 x i32> %tmp1, <4 x i32> %tmp2)
102 define <2 x i64> @vrshlQs64(<2 x i64>* %A, <2 x i64>* %B) nounwind {
105 %tmp1 = load <2 x i64>* %A
106 %tmp2 = load <2 x i64>* %B
107 %tmp3 = call <2 x i64> @llvm.arm.neon.vrshifts.v2i64(<2 x i64> %tmp1, <2 x i64> %tmp2)
111 define <16 x i8> @vrshlQu8(<16 x i8>* %A, <16 x i8>* %B) nounwind {
114 %tmp1 = load <16 x i8>* %A
115 %tmp2 = load <16 x i8>* %B
116 %tmp3 = call <16 x i8> @llvm.arm.neon.vrshiftu.v16i8(<16 x i8> %tmp1, <16 x i8> %tmp2)
120 define <8 x i16> @vrshlQu16(<8 x i16>* %A, <8 x i16>* %B) nounwind {
123 %tmp1 = load <8 x i16>* %A
124 %tmp2 = load <8 x i16>* %B
125 %tmp3 = call <8 x i16> @llvm.arm.neon.vrshiftu.v8i16(<8 x i16> %tmp1, <8 x i16> %tmp2)
129 define <4 x i32> @vrshlQu32(<4 x i32>* %A, <4 x i32>* %B) nounwind {
132 %tmp1 = load <4 x i32>* %A
133 %tmp2 = load <4 x i32>* %B
134 %tmp3 = call <4 x i32> @llvm.arm.neon.vrshiftu.v4i32(<4 x i32> %tmp1, <4 x i32> %tmp2)
138 define <2 x i64> @vrshlQu64(<2 x i64>* %A, <2 x i64>* %B) nounwind {
141 %tmp1 = load <2 x i64>* %A
142 %tmp2 = load <2 x i64>* %B
143 %tmp3 = call <2 x i64> @llvm.arm.neon.vrshiftu.v2i64(<2 x i64> %tmp1, <2 x i64> %tmp2)
147 define <8 x i8> @vrshrs8(<8 x i8>* %A) nounwind {
150 %tmp1 = load <8 x i8>* %A
151 %tmp2 = call <8 x i8> @llvm.arm.neon.vrshifts.v8i8(<8 x i8> %tmp1, <8 x i8> < i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8 >)
155 define <4 x i16> @vrshrs16(<4 x i16>* %A) nounwind {
158 %tmp1 = load <4 x i16>* %A
159 %tmp2 = call <4 x i16> @llvm.arm.neon.vrshifts.v4i16(<4 x i16> %tmp1, <4 x i16> < i16 -16, i16 -16, i16 -16, i16 -16 >)
163 define <2 x i32> @vrshrs32(<2 x i32>* %A) nounwind {
166 %tmp1 = load <2 x i32>* %A
167 %tmp2 = call <2 x i32> @llvm.arm.neon.vrshifts.v2i32(<2 x i32> %tmp1, <2 x i32> < i32 -32, i32 -32 >)
171 define <1 x i64> @vrshrs64(<1 x i64>* %A) nounwind {
174 %tmp1 = load <1 x i64>* %A
175 %tmp2 = call <1 x i64> @llvm.arm.neon.vrshifts.v1i64(<1 x i64> %tmp1, <1 x i64> < i64 -64 >)
179 define <8 x i8> @vrshru8(<8 x i8>* %A) nounwind {
182 %tmp1 = load <8 x i8>* %A
183 %tmp2 = call <8 x i8> @llvm.arm.neon.vrshiftu.v8i8(<8 x i8> %tmp1, <8 x i8> < i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8 >)
187 define <4 x i16> @vrshru16(<4 x i16>* %A) nounwind {
190 %tmp1 = load <4 x i16>* %A
191 %tmp2 = call <4 x i16> @llvm.arm.neon.vrshiftu.v4i16(<4 x i16> %tmp1, <4 x i16> < i16 -16, i16 -16, i16 -16, i16 -16 >)
195 define <2 x i32> @vrshru32(<2 x i32>* %A) nounwind {
198 %tmp1 = load <2 x i32>* %A
199 %tmp2 = call <2 x i32> @llvm.arm.neon.vrshiftu.v2i32(<2 x i32> %tmp1, <2 x i32> < i32 -32, i32 -32 >)
203 define <1 x i64> @vrshru64(<1 x i64>* %A) nounwind {
206 %tmp1 = load <1 x i64>* %A
207 %tmp2 = call <1 x i64> @llvm.arm.neon.vrshiftu.v1i64(<1 x i64> %tmp1, <1 x i64> < i64 -64 >)
211 define <16 x i8> @vrshrQs8(<16 x i8>* %A) nounwind {
214 %tmp1 = load <16 x i8>* %A
215 %tmp2 = call <16 x i8> @llvm.arm.neon.vrshifts.v16i8(<16 x i8> %tmp1, <16 x i8> < i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8 >)
219 define <8 x i16> @vrshrQs16(<8 x i16>* %A) nounwind {
222 %tmp1 = load <8 x i16>* %A
223 %tmp2 = call <8 x i16> @llvm.arm.neon.vrshifts.v8i16(<8 x i16> %tmp1, <8 x i16> < i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16 >)
227 define <4 x i32> @vrshrQs32(<4 x i32>* %A) nounwind {
230 %tmp1 = load <4 x i32>* %A
231 %tmp2 = call <4 x i32> @llvm.arm.neon.vrshifts.v4i32(<4 x i32> %tmp1, <4 x i32> < i32 -32, i32 -32, i32 -32, i32 -32 >)
235 define <2 x i64> @vrshrQs64(<2 x i64>* %A) nounwind {
238 %tmp1 = load <2 x i64>* %A
239 %tmp2 = call <2 x i64> @llvm.arm.neon.vrshifts.v2i64(<2 x i64> %tmp1, <2 x i64> < i64 -64, i64 -64 >)
243 define <16 x i8> @vrshrQu8(<16 x i8>* %A) nounwind {
246 %tmp1 = load <16 x i8>* %A
247 %tmp2 = call <16 x i8> @llvm.arm.neon.vrshiftu.v16i8(<16 x i8> %tmp1, <16 x i8> < i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8 >)
251 define <8 x i16> @vrshrQu16(<8 x i16>* %A) nounwind {
254 %tmp1 = load <8 x i16>* %A
255 %tmp2 = call <8 x i16> @llvm.arm.neon.vrshiftu.v8i16(<8 x i16> %tmp1, <8 x i16> < i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16 >)
259 define <4 x i32> @vrshrQu32(<4 x i32>* %A) nounwind {
262 %tmp1 = load <4 x i32>* %A
263 %tmp2 = call <4 x i32> @llvm.arm.neon.vrshiftu.v4i32(<4 x i32> %tmp1, <4 x i32> < i32 -32, i32 -32, i32 -32, i32 -32 >)
267 define <2 x i64> @vrshrQu64(<2 x i64>* %A) nounwind {
270 %tmp1 = load <2 x i64>* %A
271 %tmp2 = call <2 x i64> @llvm.arm.neon.vrshiftu.v2i64(<2 x i64> %tmp1, <2 x i64> < i64 -64, i64 -64 >)
275 declare <8 x i8> @llvm.arm.neon.vrshifts.v8i8(<8 x i8>, <8 x i8>) nounwind readnone
276 declare <4 x i16> @llvm.arm.neon.vrshifts.v4i16(<4 x i16>, <4 x i16>) nounwind readnone
277 declare <2 x i32> @llvm.arm.neon.vrshifts.v2i32(<2 x i32>, <2 x i32>) nounwind readnone
278 declare <1 x i64> @llvm.arm.neon.vrshifts.v1i64(<1 x i64>, <1 x i64>) nounwind readnone
280 declare <8 x i8> @llvm.arm.neon.vrshiftu.v8i8(<8 x i8>, <8 x i8>) nounwind readnone
281 declare <4 x i16> @llvm.arm.neon.vrshiftu.v4i16(<4 x i16>, <4 x i16>) nounwind readnone
282 declare <2 x i32> @llvm.arm.neon.vrshiftu.v2i32(<2 x i32>, <2 x i32>) nounwind readnone
283 declare <1 x i64> @llvm.arm.neon.vrshiftu.v1i64(<1 x i64>, <1 x i64>) nounwind readnone
285 declare <16 x i8> @llvm.arm.neon.vrshifts.v16i8(<16 x i8>, <16 x i8>) nounwind readnone
286 declare <8 x i16> @llvm.arm.neon.vrshifts.v8i16(<8 x i16>, <8 x i16>) nounwind readnone
287 declare <4 x i32> @llvm.arm.neon.vrshifts.v4i32(<4 x i32>, <4 x i32>) nounwind readnone
288 declare <2 x i64> @llvm.arm.neon.vrshifts.v2i64(<2 x i64>, <2 x i64>) nounwind readnone
290 declare <16 x i8> @llvm.arm.neon.vrshiftu.v16i8(<16 x i8>, <16 x i8>) nounwind readnone
291 declare <8 x i16> @llvm.arm.neon.vrshiftu.v8i16(<8 x i16>, <8 x i16>) nounwind readnone
292 declare <4 x i32> @llvm.arm.neon.vrshiftu.v4i32(<4 x i32>, <4 x i32>) nounwind readnone
293 declare <2 x i64> @llvm.arm.neon.vrshiftu.v2i64(<2 x i64>, <2 x i64>) nounwind readnone