1 ; RUN: llc -mtriple=arm-eabi -mattr=+neon %s -o - | FileCheck %s
3 define <8 x i8> @vshls8(ptr %A, ptr %B) nounwind {
6 %tmp1 = load <8 x i8>, ptr %A
7 %tmp2 = load <8 x i8>, ptr %B
8 %tmp3 = call <8 x i8> @llvm.arm.neon.vshifts.v8i8(<8 x i8> %tmp1, <8 x i8> %tmp2)
12 define <4 x i16> @vshls16(ptr %A, ptr %B) nounwind {
13 ;CHECK-LABEL: vshls16:
15 %tmp1 = load <4 x i16>, ptr %A
16 %tmp2 = load <4 x i16>, ptr %B
17 %tmp3 = call <4 x i16> @llvm.arm.neon.vshifts.v4i16(<4 x i16> %tmp1, <4 x i16> %tmp2)
21 define <2 x i32> @vshls32(ptr %A, ptr %B) nounwind {
22 ;CHECK-LABEL: vshls32:
24 %tmp1 = load <2 x i32>, ptr %A
25 %tmp2 = load <2 x i32>, ptr %B
26 %tmp3 = call <2 x i32> @llvm.arm.neon.vshifts.v2i32(<2 x i32> %tmp1, <2 x i32> %tmp2)
30 define <1 x i64> @vshls64(ptr %A, ptr %B) nounwind {
31 ;CHECK-LABEL: vshls64:
33 %tmp1 = load <1 x i64>, ptr %A
34 %tmp2 = load <1 x i64>, ptr %B
35 %tmp3 = call <1 x i64> @llvm.arm.neon.vshifts.v1i64(<1 x i64> %tmp1, <1 x i64> %tmp2)
39 define <8 x i8> @vshlu8(ptr %A, ptr %B) nounwind {
42 %tmp1 = load <8 x i8>, ptr %A
43 %tmp2 = load <8 x i8>, ptr %B
44 %tmp3 = call <8 x i8> @llvm.arm.neon.vshiftu.v8i8(<8 x i8> %tmp1, <8 x i8> %tmp2)
48 define <4 x i16> @vshlu16(ptr %A, ptr %B) nounwind {
49 ;CHECK-LABEL: vshlu16:
51 %tmp1 = load <4 x i16>, ptr %A
52 %tmp2 = load <4 x i16>, ptr %B
53 %tmp3 = call <4 x i16> @llvm.arm.neon.vshiftu.v4i16(<4 x i16> %tmp1, <4 x i16> %tmp2)
57 define <2 x i32> @vshlu32(ptr %A, ptr %B) nounwind {
58 ;CHECK-LABEL: vshlu32:
60 %tmp1 = load <2 x i32>, ptr %A
61 %tmp2 = load <2 x i32>, ptr %B
62 %tmp3 = call <2 x i32> @llvm.arm.neon.vshiftu.v2i32(<2 x i32> %tmp1, <2 x i32> %tmp2)
66 define <1 x i64> @vshlu64(ptr %A, ptr %B) nounwind {
67 ;CHECK-LABEL: vshlu64:
69 %tmp1 = load <1 x i64>, ptr %A
70 %tmp2 = load <1 x i64>, ptr %B
71 %tmp3 = call <1 x i64> @llvm.arm.neon.vshiftu.v1i64(<1 x i64> %tmp1, <1 x i64> %tmp2)
75 define <16 x i8> @vshlQs8(ptr %A, ptr %B) nounwind {
76 ;CHECK-LABEL: vshlQs8:
78 %tmp1 = load <16 x i8>, ptr %A
79 %tmp2 = load <16 x i8>, ptr %B
80 %tmp3 = call <16 x i8> @llvm.arm.neon.vshifts.v16i8(<16 x i8> %tmp1, <16 x i8> %tmp2)
84 define <8 x i16> @vshlQs16(ptr %A, ptr %B) nounwind {
85 ;CHECK-LABEL: vshlQs16:
87 %tmp1 = load <8 x i16>, ptr %A
88 %tmp2 = load <8 x i16>, ptr %B
89 %tmp3 = call <8 x i16> @llvm.arm.neon.vshifts.v8i16(<8 x i16> %tmp1, <8 x i16> %tmp2)
93 define <4 x i32> @vshlQs32(ptr %A, ptr %B) nounwind {
94 ;CHECK-LABEL: vshlQs32:
96 %tmp1 = load <4 x i32>, ptr %A
97 %tmp2 = load <4 x i32>, ptr %B
98 %tmp3 = call <4 x i32> @llvm.arm.neon.vshifts.v4i32(<4 x i32> %tmp1, <4 x i32> %tmp2)
102 define <2 x i64> @vshlQs64(ptr %A, ptr %B) nounwind {
103 ;CHECK-LABEL: vshlQs64:
105 %tmp1 = load <2 x i64>, ptr %A
106 %tmp2 = load <2 x i64>, ptr %B
107 %tmp3 = call <2 x i64> @llvm.arm.neon.vshifts.v2i64(<2 x i64> %tmp1, <2 x i64> %tmp2)
111 define <16 x i8> @vshlQu8(ptr %A, ptr %B) nounwind {
112 ;CHECK-LABEL: vshlQu8:
114 %tmp1 = load <16 x i8>, ptr %A
115 %tmp2 = load <16 x i8>, ptr %B
116 %tmp3 = call <16 x i8> @llvm.arm.neon.vshiftu.v16i8(<16 x i8> %tmp1, <16 x i8> %tmp2)
120 define <8 x i16> @vshlQu16(ptr %A, ptr %B) nounwind {
121 ;CHECK-LABEL: vshlQu16:
123 %tmp1 = load <8 x i16>, ptr %A
124 %tmp2 = load <8 x i16>, ptr %B
125 %tmp3 = call <8 x i16> @llvm.arm.neon.vshiftu.v8i16(<8 x i16> %tmp1, <8 x i16> %tmp2)
129 define <4 x i32> @vshlQu32(ptr %A, ptr %B) nounwind {
130 ;CHECK-LABEL: vshlQu32:
132 %tmp1 = load <4 x i32>, ptr %A
133 %tmp2 = load <4 x i32>, ptr %B
134 %tmp3 = call <4 x i32> @llvm.arm.neon.vshiftu.v4i32(<4 x i32> %tmp1, <4 x i32> %tmp2)
138 define <2 x i64> @vshlQu64(ptr %A, ptr %B) nounwind {
139 ;CHECK-LABEL: vshlQu64:
141 %tmp1 = load <2 x i64>, ptr %A
142 %tmp2 = load <2 x i64>, ptr %B
143 %tmp3 = call <2 x i64> @llvm.arm.neon.vshiftu.v2i64(<2 x i64> %tmp1, <2 x i64> %tmp2)
147 ; For left shifts by immediates, the signedness is irrelevant.
148 ; Test a mix of both signed and unsigned intrinsics.
150 define <8 x i8> @vshli8(ptr %A) nounwind {
151 ;CHECK-LABEL: vshli8:
153 %tmp1 = load <8 x i8>, ptr %A
154 %tmp2 = call <8 x i8> @llvm.arm.neon.vshifts.v8i8(<8 x i8> %tmp1, <8 x i8> < i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7 >)
158 define <4 x i16> @vshli16(ptr %A) nounwind {
159 ;CHECK-LABEL: vshli16:
161 %tmp1 = load <4 x i16>, ptr %A
162 %tmp2 = call <4 x i16> @llvm.arm.neon.vshiftu.v4i16(<4 x i16> %tmp1, <4 x i16> < i16 15, i16 15, i16 15, i16 15 >)
166 define <2 x i32> @vshli32(ptr %A) nounwind {
167 ;CHECK-LABEL: vshli32:
169 %tmp1 = load <2 x i32>, ptr %A
170 %tmp2 = call <2 x i32> @llvm.arm.neon.vshifts.v2i32(<2 x i32> %tmp1, <2 x i32> < i32 31, i32 31 >)
174 define <1 x i64> @vshli64(ptr %A) nounwind {
175 ;CHECK-LABEL: vshli64:
177 %tmp1 = load <1 x i64>, ptr %A
178 %tmp2 = call <1 x i64> @llvm.arm.neon.vshiftu.v1i64(<1 x i64> %tmp1, <1 x i64> < i64 63 >)
182 define <16 x i8> @vshlQi8(ptr %A) nounwind {
183 ;CHECK-LABEL: vshlQi8:
185 %tmp1 = load <16 x i8>, ptr %A
186 %tmp2 = call <16 x i8> @llvm.arm.neon.vshifts.v16i8(<16 x i8> %tmp1, <16 x i8> < i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7 >)
190 define <8 x i16> @vshlQi16(ptr %A) nounwind {
191 ;CHECK-LABEL: vshlQi16:
193 %tmp1 = load <8 x i16>, ptr %A
194 %tmp2 = call <8 x i16> @llvm.arm.neon.vshiftu.v8i16(<8 x i16> %tmp1, <8 x i16> < i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15 >)
198 define <4 x i32> @vshlQi32(ptr %A) nounwind {
199 ;CHECK-LABEL: vshlQi32:
201 %tmp1 = load <4 x i32>, ptr %A
202 %tmp2 = call <4 x i32> @llvm.arm.neon.vshifts.v4i32(<4 x i32> %tmp1, <4 x i32> < i32 31, i32 31, i32 31, i32 31 >)
206 define <2 x i64> @vshlQi64(ptr %A) nounwind {
207 ;CHECK-LABEL: vshlQi64:
209 %tmp1 = load <2 x i64>, ptr %A
210 %tmp2 = call <2 x i64> @llvm.arm.neon.vshiftu.v2i64(<2 x i64> %tmp1, <2 x i64> < i64 63, i64 63 >)
214 ; Right shift by immediate:
216 define <8 x i8> @vshrs8(ptr %A) nounwind {
217 ;CHECK-LABEL: vshrs8:
219 %tmp1 = load <8 x i8>, ptr %A
220 %tmp2 = call <8 x i8> @llvm.arm.neon.vshifts.v8i8(<8 x i8> %tmp1, <8 x i8> < i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8 >)
224 define <4 x i16> @vshrs16(ptr %A) nounwind {
225 ;CHECK-LABEL: vshrs16:
227 %tmp1 = load <4 x i16>, ptr %A
228 %tmp2 = call <4 x i16> @llvm.arm.neon.vshifts.v4i16(<4 x i16> %tmp1, <4 x i16> < i16 -16, i16 -16, i16 -16, i16 -16 >)
232 define <2 x i32> @vshrs32(ptr %A) nounwind {
233 ;CHECK-LABEL: vshrs32:
235 %tmp1 = load <2 x i32>, ptr %A
236 %tmp2 = call <2 x i32> @llvm.arm.neon.vshifts.v2i32(<2 x i32> %tmp1, <2 x i32> < i32 -32, i32 -32 >)
240 define <1 x i64> @vshrs64(ptr %A) nounwind {
241 ;CHECK-LABEL: vshrs64:
243 %tmp1 = load <1 x i64>, ptr %A
244 %tmp2 = call <1 x i64> @llvm.arm.neon.vshifts.v1i64(<1 x i64> %tmp1, <1 x i64> < i64 -64 >)
248 define <8 x i8> @vshru8(ptr %A) nounwind {
249 ;CHECK-LABEL: vshru8:
251 %tmp1 = load <8 x i8>, ptr %A
252 %tmp2 = call <8 x i8> @llvm.arm.neon.vshiftu.v8i8(<8 x i8> %tmp1, <8 x i8> < i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8 >)
256 define <4 x i16> @vshru16(ptr %A) nounwind {
257 ;CHECK-LABEL: vshru16:
259 %tmp1 = load <4 x i16>, ptr %A
260 %tmp2 = call <4 x i16> @llvm.arm.neon.vshiftu.v4i16(<4 x i16> %tmp1, <4 x i16> < i16 -16, i16 -16, i16 -16, i16 -16 >)
264 define <2 x i32> @vshru32(ptr %A) nounwind {
265 ;CHECK-LABEL: vshru32:
267 %tmp1 = load <2 x i32>, ptr %A
268 %tmp2 = call <2 x i32> @llvm.arm.neon.vshiftu.v2i32(<2 x i32> %tmp1, <2 x i32> < i32 -32, i32 -32 >)
272 define <1 x i64> @vshru64(ptr %A) nounwind {
273 ;CHECK-LABEL: vshru64:
275 %tmp1 = load <1 x i64>, ptr %A
276 %tmp2 = call <1 x i64> @llvm.arm.neon.vshiftu.v1i64(<1 x i64> %tmp1, <1 x i64> < i64 -64 >)
280 define <16 x i8> @vshrQs8(ptr %A) nounwind {
281 ;CHECK-LABEL: vshrQs8:
283 %tmp1 = load <16 x i8>, ptr %A
284 %tmp2 = call <16 x i8> @llvm.arm.neon.vshifts.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 >)
288 define <8 x i16> @vshrQs16(ptr %A) nounwind {
289 ;CHECK-LABEL: vshrQs16:
291 %tmp1 = load <8 x i16>, ptr %A
292 %tmp2 = call <8 x i16> @llvm.arm.neon.vshifts.v8i16(<8 x i16> %tmp1, <8 x i16> < i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16 >)
296 define <4 x i32> @vshrQs32(ptr %A) nounwind {
297 ;CHECK-LABEL: vshrQs32:
299 %tmp1 = load <4 x i32>, ptr %A
300 %tmp2 = call <4 x i32> @llvm.arm.neon.vshifts.v4i32(<4 x i32> %tmp1, <4 x i32> < i32 -32, i32 -32, i32 -32, i32 -32 >)
304 define <2 x i64> @vshrQs64(ptr %A) nounwind {
305 ;CHECK-LABEL: vshrQs64:
307 %tmp1 = load <2 x i64>, ptr %A
308 %tmp2 = call <2 x i64> @llvm.arm.neon.vshifts.v2i64(<2 x i64> %tmp1, <2 x i64> < i64 -64, i64 -64 >)
312 define <16 x i8> @vshrQu8(ptr %A) nounwind {
313 ;CHECK-LABEL: vshrQu8:
315 %tmp1 = load <16 x i8>, ptr %A
316 %tmp2 = call <16 x i8> @llvm.arm.neon.vshiftu.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 >)
320 define <8 x i16> @vshrQu16(ptr %A) nounwind {
321 ;CHECK-LABEL: vshrQu16:
323 %tmp1 = load <8 x i16>, ptr %A
324 %tmp2 = call <8 x i16> @llvm.arm.neon.vshiftu.v8i16(<8 x i16> %tmp1, <8 x i16> < i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16 >)
328 define <4 x i32> @vshrQu32(ptr %A) nounwind {
329 ;CHECK-LABEL: vshrQu32:
331 %tmp1 = load <4 x i32>, ptr %A
332 %tmp2 = call <4 x i32> @llvm.arm.neon.vshiftu.v4i32(<4 x i32> %tmp1, <4 x i32> < i32 -32, i32 -32, i32 -32, i32 -32 >)
336 define <2 x i64> @vshrQu64(ptr %A) nounwind {
337 ;CHECK-LABEL: vshrQu64:
339 %tmp1 = load <2 x i64>, ptr %A
340 %tmp2 = call <2 x i64> @llvm.arm.neon.vshiftu.v2i64(<2 x i64> %tmp1, <2 x i64> < i64 -64, i64 -64 >)
344 declare <8 x i8> @llvm.arm.neon.vshifts.v8i8(<8 x i8>, <8 x i8>) nounwind readnone
345 declare <4 x i16> @llvm.arm.neon.vshifts.v4i16(<4 x i16>, <4 x i16>) nounwind readnone
346 declare <2 x i32> @llvm.arm.neon.vshifts.v2i32(<2 x i32>, <2 x i32>) nounwind readnone
347 declare <1 x i64> @llvm.arm.neon.vshifts.v1i64(<1 x i64>, <1 x i64>) nounwind readnone
349 declare <8 x i8> @llvm.arm.neon.vshiftu.v8i8(<8 x i8>, <8 x i8>) nounwind readnone
350 declare <4 x i16> @llvm.arm.neon.vshiftu.v4i16(<4 x i16>, <4 x i16>) nounwind readnone
351 declare <2 x i32> @llvm.arm.neon.vshiftu.v2i32(<2 x i32>, <2 x i32>) nounwind readnone
352 declare <1 x i64> @llvm.arm.neon.vshiftu.v1i64(<1 x i64>, <1 x i64>) nounwind readnone
354 declare <16 x i8> @llvm.arm.neon.vshifts.v16i8(<16 x i8>, <16 x i8>) nounwind readnone
355 declare <8 x i16> @llvm.arm.neon.vshifts.v8i16(<8 x i16>, <8 x i16>) nounwind readnone
356 declare <4 x i32> @llvm.arm.neon.vshifts.v4i32(<4 x i32>, <4 x i32>) nounwind readnone
357 declare <2 x i64> @llvm.arm.neon.vshifts.v2i64(<2 x i64>, <2 x i64>) nounwind readnone
359 declare <16 x i8> @llvm.arm.neon.vshiftu.v16i8(<16 x i8>, <16 x i8>) nounwind readnone
360 declare <8 x i16> @llvm.arm.neon.vshiftu.v8i16(<8 x i16>, <8 x i16>) nounwind readnone
361 declare <4 x i32> @llvm.arm.neon.vshiftu.v4i32(<4 x i32>, <4 x i32>) nounwind readnone
362 declare <2 x i64> @llvm.arm.neon.vshiftu.v2i64(<2 x i64>, <2 x i64>) nounwind readnone
364 define <8 x i8> @vrshls8(ptr %A, ptr %B) nounwind {
365 ;CHECK-LABEL: vrshls8:
367 %tmp1 = load <8 x i8>, ptr %A
368 %tmp2 = load <8 x i8>, ptr %B
369 %tmp3 = call <8 x i8> @llvm.arm.neon.vrshifts.v8i8(<8 x i8> %tmp1, <8 x i8> %tmp2)
373 define <4 x i16> @vrshls16(ptr %A, ptr %B) nounwind {
374 ;CHECK-LABEL: vrshls16:
376 %tmp1 = load <4 x i16>, ptr %A
377 %tmp2 = load <4 x i16>, ptr %B
378 %tmp3 = call <4 x i16> @llvm.arm.neon.vrshifts.v4i16(<4 x i16> %tmp1, <4 x i16> %tmp2)
382 define <2 x i32> @vrshls32(ptr %A, ptr %B) nounwind {
383 ;CHECK-LABEL: vrshls32:
385 %tmp1 = load <2 x i32>, ptr %A
386 %tmp2 = load <2 x i32>, ptr %B
387 %tmp3 = call <2 x i32> @llvm.arm.neon.vrshifts.v2i32(<2 x i32> %tmp1, <2 x i32> %tmp2)
391 define <1 x i64> @vrshls64(ptr %A, ptr %B) nounwind {
392 ;CHECK-LABEL: vrshls64:
394 %tmp1 = load <1 x i64>, ptr %A
395 %tmp2 = load <1 x i64>, ptr %B
396 %tmp3 = call <1 x i64> @llvm.arm.neon.vrshifts.v1i64(<1 x i64> %tmp1, <1 x i64> %tmp2)
400 define <8 x i8> @vrshlu8(ptr %A, ptr %B) nounwind {
401 ;CHECK-LABEL: vrshlu8:
403 %tmp1 = load <8 x i8>, ptr %A
404 %tmp2 = load <8 x i8>, ptr %B
405 %tmp3 = call <8 x i8> @llvm.arm.neon.vrshiftu.v8i8(<8 x i8> %tmp1, <8 x i8> %tmp2)
409 define <4 x i16> @vrshlu16(ptr %A, ptr %B) nounwind {
410 ;CHECK-LABEL: vrshlu16:
412 %tmp1 = load <4 x i16>, ptr %A
413 %tmp2 = load <4 x i16>, ptr %B
414 %tmp3 = call <4 x i16> @llvm.arm.neon.vrshiftu.v4i16(<4 x i16> %tmp1, <4 x i16> %tmp2)
418 define <2 x i32> @vrshlu32(ptr %A, ptr %B) nounwind {
419 ;CHECK-LABEL: vrshlu32:
421 %tmp1 = load <2 x i32>, ptr %A
422 %tmp2 = load <2 x i32>, ptr %B
423 %tmp3 = call <2 x i32> @llvm.arm.neon.vrshiftu.v2i32(<2 x i32> %tmp1, <2 x i32> %tmp2)
427 define <1 x i64> @vrshlu64(ptr %A, ptr %B) nounwind {
428 ;CHECK-LABEL: vrshlu64:
430 %tmp1 = load <1 x i64>, ptr %A
431 %tmp2 = load <1 x i64>, ptr %B
432 %tmp3 = call <1 x i64> @llvm.arm.neon.vrshiftu.v1i64(<1 x i64> %tmp1, <1 x i64> %tmp2)
436 define <16 x i8> @vrshlQs8(ptr %A, ptr %B) nounwind {
437 ;CHECK-LABEL: vrshlQs8:
439 %tmp1 = load <16 x i8>, ptr %A
440 %tmp2 = load <16 x i8>, ptr %B
441 %tmp3 = call <16 x i8> @llvm.arm.neon.vrshifts.v16i8(<16 x i8> %tmp1, <16 x i8> %tmp2)
445 define <8 x i16> @vrshlQs16(ptr %A, ptr %B) nounwind {
446 ;CHECK-LABEL: vrshlQs16:
448 %tmp1 = load <8 x i16>, ptr %A
449 %tmp2 = load <8 x i16>, ptr %B
450 %tmp3 = call <8 x i16> @llvm.arm.neon.vrshifts.v8i16(<8 x i16> %tmp1, <8 x i16> %tmp2)
454 define <4 x i32> @vrshlQs32(ptr %A, ptr %B) nounwind {
455 ;CHECK-LABEL: vrshlQs32:
457 %tmp1 = load <4 x i32>, ptr %A
458 %tmp2 = load <4 x i32>, ptr %B
459 %tmp3 = call <4 x i32> @llvm.arm.neon.vrshifts.v4i32(<4 x i32> %tmp1, <4 x i32> %tmp2)
463 define <2 x i64> @vrshlQs64(ptr %A, ptr %B) nounwind {
464 ;CHECK-LABEL: vrshlQs64:
466 %tmp1 = load <2 x i64>, ptr %A
467 %tmp2 = load <2 x i64>, ptr %B
468 %tmp3 = call <2 x i64> @llvm.arm.neon.vrshifts.v2i64(<2 x i64> %tmp1, <2 x i64> %tmp2)
472 define <16 x i8> @vrshlQu8(ptr %A, ptr %B) nounwind {
473 ;CHECK-LABEL: vrshlQu8:
475 %tmp1 = load <16 x i8>, ptr %A
476 %tmp2 = load <16 x i8>, ptr %B
477 %tmp3 = call <16 x i8> @llvm.arm.neon.vrshiftu.v16i8(<16 x i8> %tmp1, <16 x i8> %tmp2)
481 define <8 x i16> @vrshlQu16(ptr %A, ptr %B) nounwind {
482 ;CHECK-LABEL: vrshlQu16:
484 %tmp1 = load <8 x i16>, ptr %A
485 %tmp2 = load <8 x i16>, ptr %B
486 %tmp3 = call <8 x i16> @llvm.arm.neon.vrshiftu.v8i16(<8 x i16> %tmp1, <8 x i16> %tmp2)
490 define <4 x i32> @vrshlQu32(ptr %A, ptr %B) nounwind {
491 ;CHECK-LABEL: vrshlQu32:
493 %tmp1 = load <4 x i32>, ptr %A
494 %tmp2 = load <4 x i32>, ptr %B
495 %tmp3 = call <4 x i32> @llvm.arm.neon.vrshiftu.v4i32(<4 x i32> %tmp1, <4 x i32> %tmp2)
499 define <2 x i64> @vrshlQu64(ptr %A, ptr %B) nounwind {
500 ;CHECK-LABEL: vrshlQu64:
502 %tmp1 = load <2 x i64>, ptr %A
503 %tmp2 = load <2 x i64>, ptr %B
504 %tmp3 = call <2 x i64> @llvm.arm.neon.vrshiftu.v2i64(<2 x i64> %tmp1, <2 x i64> %tmp2)
508 define <8 x i8> @vrshrs8(ptr %A) nounwind {
509 ;CHECK-LABEL: vrshrs8:
511 %tmp1 = load <8 x i8>, ptr %A
512 %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 >)
516 define <4 x i16> @vrshrs16(ptr %A) nounwind {
517 ;CHECK-LABEL: vrshrs16:
519 %tmp1 = load <4 x i16>, ptr %A
520 %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 >)
524 define <2 x i32> @vrshrs32(ptr %A) nounwind {
525 ;CHECK-LABEL: vrshrs32:
527 %tmp1 = load <2 x i32>, ptr %A
528 %tmp2 = call <2 x i32> @llvm.arm.neon.vrshifts.v2i32(<2 x i32> %tmp1, <2 x i32> < i32 -32, i32 -32 >)
532 define <1 x i64> @vrshrs64(ptr %A) nounwind {
533 ;CHECK-LABEL: vrshrs64:
535 %tmp1 = load <1 x i64>, ptr %A
536 %tmp2 = call <1 x i64> @llvm.arm.neon.vrshifts.v1i64(<1 x i64> %tmp1, <1 x i64> < i64 -64 >)
540 define <8 x i8> @vrshru8(ptr %A) nounwind {
541 ;CHECK-LABEL: vrshru8:
543 %tmp1 = load <8 x i8>, ptr %A
544 %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 >)
548 define <4 x i16> @vrshru16(ptr %A) nounwind {
549 ;CHECK-LABEL: vrshru16:
551 %tmp1 = load <4 x i16>, ptr %A
552 %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 >)
556 define <2 x i32> @vrshru32(ptr %A) nounwind {
557 ;CHECK-LABEL: vrshru32:
559 %tmp1 = load <2 x i32>, ptr %A
560 %tmp2 = call <2 x i32> @llvm.arm.neon.vrshiftu.v2i32(<2 x i32> %tmp1, <2 x i32> < i32 -32, i32 -32 >)
564 define <1 x i64> @vrshru64(ptr %A) nounwind {
565 ;CHECK-LABEL: vrshru64:
567 %tmp1 = load <1 x i64>, ptr %A
568 %tmp2 = call <1 x i64> @llvm.arm.neon.vrshiftu.v1i64(<1 x i64> %tmp1, <1 x i64> < i64 -64 >)
572 define <16 x i8> @vrshrQs8(ptr %A) nounwind {
573 ;CHECK-LABEL: vrshrQs8:
575 %tmp1 = load <16 x i8>, ptr %A
576 %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 >)
580 define <8 x i16> @vrshrQs16(ptr %A) nounwind {
581 ;CHECK-LABEL: vrshrQs16:
583 %tmp1 = load <8 x i16>, ptr %A
584 %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 >)
588 define <4 x i32> @vrshrQs32(ptr %A) nounwind {
589 ;CHECK-LABEL: vrshrQs32:
591 %tmp1 = load <4 x i32>, ptr %A
592 %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 >)
596 define <2 x i64> @vrshrQs64(ptr %A) nounwind {
597 ;CHECK-LABEL: vrshrQs64:
599 %tmp1 = load <2 x i64>, ptr %A
600 %tmp2 = call <2 x i64> @llvm.arm.neon.vrshifts.v2i64(<2 x i64> %tmp1, <2 x i64> < i64 -64, i64 -64 >)
604 define <16 x i8> @vrshrQu8(ptr %A) nounwind {
605 ;CHECK-LABEL: vrshrQu8:
607 %tmp1 = load <16 x i8>, ptr %A
608 %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 >)
612 define <8 x i16> @vrshrQu16(ptr %A) nounwind {
613 ;CHECK-LABEL: vrshrQu16:
615 %tmp1 = load <8 x i16>, ptr %A
616 %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 >)
620 define <4 x i32> @vrshrQu32(ptr %A) nounwind {
621 ;CHECK-LABEL: vrshrQu32:
623 %tmp1 = load <4 x i32>, ptr %A
624 %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 >)
628 define <2 x i64> @vrshrQu64(ptr %A) nounwind {
629 ;CHECK-LABEL: vrshrQu64:
631 %tmp1 = load <2 x i64>, ptr %A
632 %tmp2 = call <2 x i64> @llvm.arm.neon.vrshiftu.v2i64(<2 x i64> %tmp1, <2 x i64> < i64 -64, i64 -64 >)
636 declare <8 x i8> @llvm.arm.neon.vrshifts.v8i8(<8 x i8>, <8 x i8>) nounwind readnone
637 declare <4 x i16> @llvm.arm.neon.vrshifts.v4i16(<4 x i16>, <4 x i16>) nounwind readnone
638 declare <2 x i32> @llvm.arm.neon.vrshifts.v2i32(<2 x i32>, <2 x i32>) nounwind readnone
639 declare <1 x i64> @llvm.arm.neon.vrshifts.v1i64(<1 x i64>, <1 x i64>) nounwind readnone
641 declare <8 x i8> @llvm.arm.neon.vrshiftu.v8i8(<8 x i8>, <8 x i8>) nounwind readnone
642 declare <4 x i16> @llvm.arm.neon.vrshiftu.v4i16(<4 x i16>, <4 x i16>) nounwind readnone
643 declare <2 x i32> @llvm.arm.neon.vrshiftu.v2i32(<2 x i32>, <2 x i32>) nounwind readnone
644 declare <1 x i64> @llvm.arm.neon.vrshiftu.v1i64(<1 x i64>, <1 x i64>) nounwind readnone
646 declare <16 x i8> @llvm.arm.neon.vrshifts.v16i8(<16 x i8>, <16 x i8>) nounwind readnone
647 declare <8 x i16> @llvm.arm.neon.vrshifts.v8i16(<8 x i16>, <8 x i16>) nounwind readnone
648 declare <4 x i32> @llvm.arm.neon.vrshifts.v4i32(<4 x i32>, <4 x i32>) nounwind readnone
649 declare <2 x i64> @llvm.arm.neon.vrshifts.v2i64(<2 x i64>, <2 x i64>) nounwind readnone
651 declare <16 x i8> @llvm.arm.neon.vrshiftu.v16i8(<16 x i8>, <16 x i8>) nounwind readnone
652 declare <8 x i16> @llvm.arm.neon.vrshiftu.v8i16(<8 x i16>, <8 x i16>) nounwind readnone
653 declare <4 x i32> @llvm.arm.neon.vrshiftu.v4i32(<4 x i32>, <4 x i32>) nounwind readnone
654 declare <2 x i64> @llvm.arm.neon.vrshiftu.v2i64(<2 x i64>, <2 x i64>) nounwind readnone