1 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2 ; RUN: llc < %s -mtriple=arm-eabi -mattr=+neon | FileCheck %s --check-prefixes=CHECK,ALLOC,NOALLOC
3 ; RUN: llc < %s -mtriple=arm-eabi -mattr=+neon -regalloc=basic | FileCheck %s --check-prefixes=CHECK,BASIC
5 define <8 x i8> @vcgts8(<8 x i8>* %A, <8 x i8>* %B) nounwind {
8 ; ALLOC-NEXT: vldr d16, [r1]
9 ; ALLOC-NEXT: vldr d17, [r0]
10 ; ALLOC-NEXT: vcgt.s8 d16, d17, d16
11 ; ALLOC-NEXT: vmov r0, r1, d16
12 ; ALLOC-NEXT: mov pc, lr
14 ; BASIC-LABEL: vcgts8:
16 ; BASIC-NEXT: vldr d17, [r1]
17 ; BASIC-NEXT: vldr d16, [r0]
18 ; BASIC-NEXT: vcgt.s8 d16, d16, d17
19 ; BASIC-NEXT: vmov r0, r1, d16
20 ; BASIC-NEXT: mov pc, lr
21 %tmp1 = load <8 x i8>, <8 x i8>* %A
22 %tmp2 = load <8 x i8>, <8 x i8>* %B
23 %tmp3 = icmp sgt <8 x i8> %tmp1, %tmp2
24 %tmp4 = sext <8 x i1> %tmp3 to <8 x i8>
28 define <4 x i16> @vcgts16(<4 x i16>* %A, <4 x i16>* %B) nounwind {
29 ; ALLOC-LABEL: vcgts16:
31 ; ALLOC-NEXT: vldr d16, [r1]
32 ; ALLOC-NEXT: vldr d17, [r0]
33 ; ALLOC-NEXT: vcgt.s16 d16, d17, d16
34 ; ALLOC-NEXT: vmov r0, r1, d16
35 ; ALLOC-NEXT: mov pc, lr
37 ; BASIC-LABEL: vcgts16:
39 ; BASIC-NEXT: vldr d17, [r1]
40 ; BASIC-NEXT: vldr d16, [r0]
41 ; BASIC-NEXT: vcgt.s16 d16, d16, d17
42 ; BASIC-NEXT: vmov r0, r1, d16
43 ; BASIC-NEXT: mov pc, lr
44 %tmp1 = load <4 x i16>, <4 x i16>* %A
45 %tmp2 = load <4 x i16>, <4 x i16>* %B
46 %tmp3 = icmp sgt <4 x i16> %tmp1, %tmp2
47 %tmp4 = sext <4 x i1> %tmp3 to <4 x i16>
51 define <2 x i32> @vcgts32(<2 x i32>* %A, <2 x i32>* %B) nounwind {
52 ; ALLOC-LABEL: vcgts32:
54 ; ALLOC-NEXT: vldr d16, [r1]
55 ; ALLOC-NEXT: vldr d17, [r0]
56 ; ALLOC-NEXT: vcgt.s32 d16, d17, d16
57 ; ALLOC-NEXT: vmov r0, r1, d16
58 ; ALLOC-NEXT: mov pc, lr
60 ; BASIC-LABEL: vcgts32:
62 ; BASIC-NEXT: vldr d17, [r1]
63 ; BASIC-NEXT: vldr d16, [r0]
64 ; BASIC-NEXT: vcgt.s32 d16, d16, d17
65 ; BASIC-NEXT: vmov r0, r1, d16
66 ; BASIC-NEXT: mov pc, lr
67 %tmp1 = load <2 x i32>, <2 x i32>* %A
68 %tmp2 = load <2 x i32>, <2 x i32>* %B
69 %tmp3 = icmp sgt <2 x i32> %tmp1, %tmp2
70 %tmp4 = sext <2 x i1> %tmp3 to <2 x i32>
74 define <8 x i8> @vcgtu8(<8 x i8>* %A, <8 x i8>* %B) nounwind {
75 ; ALLOC-LABEL: vcgtu8:
77 ; ALLOC-NEXT: vldr d16, [r1]
78 ; ALLOC-NEXT: vldr d17, [r0]
79 ; ALLOC-NEXT: vcgt.u8 d16, d17, d16
80 ; ALLOC-NEXT: vmov r0, r1, d16
81 ; ALLOC-NEXT: mov pc, lr
83 ; BASIC-LABEL: vcgtu8:
85 ; BASIC-NEXT: vldr d17, [r1]
86 ; BASIC-NEXT: vldr d16, [r0]
87 ; BASIC-NEXT: vcgt.u8 d16, d16, d17
88 ; BASIC-NEXT: vmov r0, r1, d16
89 ; BASIC-NEXT: mov pc, lr
90 %tmp1 = load <8 x i8>, <8 x i8>* %A
91 %tmp2 = load <8 x i8>, <8 x i8>* %B
92 %tmp3 = icmp ugt <8 x i8> %tmp1, %tmp2
93 %tmp4 = sext <8 x i1> %tmp3 to <8 x i8>
97 define <4 x i16> @vcgtu16(<4 x i16>* %A, <4 x i16>* %B) nounwind {
98 ; ALLOC-LABEL: vcgtu16:
100 ; ALLOC-NEXT: vldr d16, [r1]
101 ; ALLOC-NEXT: vldr d17, [r0]
102 ; ALLOC-NEXT: vcgt.u16 d16, d17, d16
103 ; ALLOC-NEXT: vmov r0, r1, d16
104 ; ALLOC-NEXT: mov pc, lr
106 ; BASIC-LABEL: vcgtu16:
108 ; BASIC-NEXT: vldr d17, [r1]
109 ; BASIC-NEXT: vldr d16, [r0]
110 ; BASIC-NEXT: vcgt.u16 d16, d16, d17
111 ; BASIC-NEXT: vmov r0, r1, d16
112 ; BASIC-NEXT: mov pc, lr
113 %tmp1 = load <4 x i16>, <4 x i16>* %A
114 %tmp2 = load <4 x i16>, <4 x i16>* %B
115 %tmp3 = icmp ugt <4 x i16> %tmp1, %tmp2
116 %tmp4 = sext <4 x i1> %tmp3 to <4 x i16>
120 define <2 x i32> @vcgtu32(<2 x i32>* %A, <2 x i32>* %B) nounwind {
121 ; ALLOC-LABEL: vcgtu32:
123 ; ALLOC-NEXT: vldr d16, [r1]
124 ; ALLOC-NEXT: vldr d17, [r0]
125 ; ALLOC-NEXT: vcgt.u32 d16, d17, d16
126 ; ALLOC-NEXT: vmov r0, r1, d16
127 ; ALLOC-NEXT: mov pc, lr
129 ; BASIC-LABEL: vcgtu32:
131 ; BASIC-NEXT: vldr d17, [r1]
132 ; BASIC-NEXT: vldr d16, [r0]
133 ; BASIC-NEXT: vcgt.u32 d16, d16, d17
134 ; BASIC-NEXT: vmov r0, r1, d16
135 ; BASIC-NEXT: mov pc, lr
136 %tmp1 = load <2 x i32>, <2 x i32>* %A
137 %tmp2 = load <2 x i32>, <2 x i32>* %B
138 %tmp3 = icmp ugt <2 x i32> %tmp1, %tmp2
139 %tmp4 = sext <2 x i1> %tmp3 to <2 x i32>
143 define <2 x i32> @vcgtf32(<2 x float>* %A, <2 x float>* %B) nounwind {
144 ; ALLOC-LABEL: vcgtf32:
146 ; ALLOC-NEXT: vldr d16, [r1]
147 ; ALLOC-NEXT: vldr d17, [r0]
148 ; ALLOC-NEXT: vcgt.f32 d16, d17, d16
149 ; ALLOC-NEXT: vmov r0, r1, d16
150 ; ALLOC-NEXT: mov pc, lr
152 ; BASIC-LABEL: vcgtf32:
154 ; BASIC-NEXT: vldr d17, [r1]
155 ; BASIC-NEXT: vldr d16, [r0]
156 ; BASIC-NEXT: vcgt.f32 d16, d16, d17
157 ; BASIC-NEXT: vmov r0, r1, d16
158 ; BASIC-NEXT: mov pc, lr
159 %tmp1 = load <2 x float>, <2 x float>* %A
160 %tmp2 = load <2 x float>, <2 x float>* %B
161 %tmp3 = fcmp ogt <2 x float> %tmp1, %tmp2
162 %tmp4 = sext <2 x i1> %tmp3 to <2 x i32>
166 define <16 x i8> @vcgtQs8(<16 x i8>* %A, <16 x i8>* %B) nounwind {
167 ; ALLOC-LABEL: vcgtQs8:
169 ; ALLOC-NEXT: vld1.64 {d16, d17}, [r1]
170 ; ALLOC-NEXT: vld1.64 {d18, d19}, [r0]
171 ; ALLOC-NEXT: vcgt.s8 q8, q9, q8
172 ; ALLOC-NEXT: vmov r0, r1, d16
173 ; ALLOC-NEXT: vmov r2, r3, d17
174 ; ALLOC-NEXT: mov pc, lr
176 ; BASIC-LABEL: vcgtQs8:
178 ; BASIC-NEXT: vld1.64 {d18, d19}, [r1]
179 ; BASIC-NEXT: vld1.64 {d16, d17}, [r0]
180 ; BASIC-NEXT: vcgt.s8 q8, q8, q9
181 ; BASIC-NEXT: vmov r0, r1, d16
182 ; BASIC-NEXT: vmov r2, r3, d17
183 ; BASIC-NEXT: mov pc, lr
184 %tmp1 = load <16 x i8>, <16 x i8>* %A
185 %tmp2 = load <16 x i8>, <16 x i8>* %B
186 %tmp3 = icmp sgt <16 x i8> %tmp1, %tmp2
187 %tmp4 = sext <16 x i1> %tmp3 to <16 x i8>
191 define <8 x i16> @vcgtQs16(<8 x i16>* %A, <8 x i16>* %B) nounwind {
192 ; ALLOC-LABEL: vcgtQs16:
194 ; ALLOC-NEXT: vld1.64 {d16, d17}, [r1]
195 ; ALLOC-NEXT: vld1.64 {d18, d19}, [r0]
196 ; ALLOC-NEXT: vcgt.s16 q8, q9, q8
197 ; ALLOC-NEXT: vmov r0, r1, d16
198 ; ALLOC-NEXT: vmov r2, r3, d17
199 ; ALLOC-NEXT: mov pc, lr
201 ; BASIC-LABEL: vcgtQs16:
203 ; BASIC-NEXT: vld1.64 {d18, d19}, [r1]
204 ; BASIC-NEXT: vld1.64 {d16, d17}, [r0]
205 ; BASIC-NEXT: vcgt.s16 q8, q8, q9
206 ; BASIC-NEXT: vmov r0, r1, d16
207 ; BASIC-NEXT: vmov r2, r3, d17
208 ; BASIC-NEXT: mov pc, lr
209 %tmp1 = load <8 x i16>, <8 x i16>* %A
210 %tmp2 = load <8 x i16>, <8 x i16>* %B
211 %tmp3 = icmp sgt <8 x i16> %tmp1, %tmp2
212 %tmp4 = sext <8 x i1> %tmp3 to <8 x i16>
216 define <4 x i32> @vcgtQs32(<4 x i32>* %A, <4 x i32>* %B) nounwind {
217 ; ALLOC-LABEL: vcgtQs32:
219 ; ALLOC-NEXT: vld1.64 {d16, d17}, [r1]
220 ; ALLOC-NEXT: vld1.64 {d18, d19}, [r0]
221 ; ALLOC-NEXT: vcgt.s32 q8, q9, q8
222 ; ALLOC-NEXT: vmov r0, r1, d16
223 ; ALLOC-NEXT: vmov r2, r3, d17
224 ; ALLOC-NEXT: mov pc, lr
226 ; BASIC-LABEL: vcgtQs32:
228 ; BASIC-NEXT: vld1.64 {d18, d19}, [r1]
229 ; BASIC-NEXT: vld1.64 {d16, d17}, [r0]
230 ; BASIC-NEXT: vcgt.s32 q8, q8, q9
231 ; BASIC-NEXT: vmov r0, r1, d16
232 ; BASIC-NEXT: vmov r2, r3, d17
233 ; BASIC-NEXT: mov pc, lr
234 %tmp1 = load <4 x i32>, <4 x i32>* %A
235 %tmp2 = load <4 x i32>, <4 x i32>* %B
236 %tmp3 = icmp sgt <4 x i32> %tmp1, %tmp2
237 %tmp4 = sext <4 x i1> %tmp3 to <4 x i32>
241 define <16 x i8> @vcgtQu8(<16 x i8>* %A, <16 x i8>* %B) nounwind {
242 ; ALLOC-LABEL: vcgtQu8:
244 ; ALLOC-NEXT: vld1.64 {d16, d17}, [r1]
245 ; ALLOC-NEXT: vld1.64 {d18, d19}, [r0]
246 ; ALLOC-NEXT: vcgt.u8 q8, q9, q8
247 ; ALLOC-NEXT: vmov r0, r1, d16
248 ; ALLOC-NEXT: vmov r2, r3, d17
249 ; ALLOC-NEXT: mov pc, lr
251 ; BASIC-LABEL: vcgtQu8:
253 ; BASIC-NEXT: vld1.64 {d18, d19}, [r1]
254 ; BASIC-NEXT: vld1.64 {d16, d17}, [r0]
255 ; BASIC-NEXT: vcgt.u8 q8, q8, q9
256 ; BASIC-NEXT: vmov r0, r1, d16
257 ; BASIC-NEXT: vmov r2, r3, d17
258 ; BASIC-NEXT: mov pc, lr
259 %tmp1 = load <16 x i8>, <16 x i8>* %A
260 %tmp2 = load <16 x i8>, <16 x i8>* %B
261 %tmp3 = icmp ugt <16 x i8> %tmp1, %tmp2
262 %tmp4 = sext <16 x i1> %tmp3 to <16 x i8>
266 define <8 x i16> @vcgtQu16(<8 x i16>* %A, <8 x i16>* %B) nounwind {
267 ; ALLOC-LABEL: vcgtQu16:
269 ; ALLOC-NEXT: vld1.64 {d16, d17}, [r1]
270 ; ALLOC-NEXT: vld1.64 {d18, d19}, [r0]
271 ; ALLOC-NEXT: vcgt.u16 q8, q9, q8
272 ; ALLOC-NEXT: vmov r0, r1, d16
273 ; ALLOC-NEXT: vmov r2, r3, d17
274 ; ALLOC-NEXT: mov pc, lr
276 ; BASIC-LABEL: vcgtQu16:
278 ; BASIC-NEXT: vld1.64 {d18, d19}, [r1]
279 ; BASIC-NEXT: vld1.64 {d16, d17}, [r0]
280 ; BASIC-NEXT: vcgt.u16 q8, q8, q9
281 ; BASIC-NEXT: vmov r0, r1, d16
282 ; BASIC-NEXT: vmov r2, r3, d17
283 ; BASIC-NEXT: mov pc, lr
284 %tmp1 = load <8 x i16>, <8 x i16>* %A
285 %tmp2 = load <8 x i16>, <8 x i16>* %B
286 %tmp3 = icmp ugt <8 x i16> %tmp1, %tmp2
287 %tmp4 = sext <8 x i1> %tmp3 to <8 x i16>
291 define <4 x i32> @vcgtQu32(<4 x i32>* %A, <4 x i32>* %B) nounwind {
292 ; ALLOC-LABEL: vcgtQu32:
294 ; ALLOC-NEXT: vld1.64 {d16, d17}, [r1]
295 ; ALLOC-NEXT: vld1.64 {d18, d19}, [r0]
296 ; ALLOC-NEXT: vcgt.u32 q8, q9, q8
297 ; ALLOC-NEXT: vmov r0, r1, d16
298 ; ALLOC-NEXT: vmov r2, r3, d17
299 ; ALLOC-NEXT: mov pc, lr
301 ; BASIC-LABEL: vcgtQu32:
303 ; BASIC-NEXT: vld1.64 {d18, d19}, [r1]
304 ; BASIC-NEXT: vld1.64 {d16, d17}, [r0]
305 ; BASIC-NEXT: vcgt.u32 q8, q8, q9
306 ; BASIC-NEXT: vmov r0, r1, d16
307 ; BASIC-NEXT: vmov r2, r3, d17
308 ; BASIC-NEXT: mov pc, lr
309 %tmp1 = load <4 x i32>, <4 x i32>* %A
310 %tmp2 = load <4 x i32>, <4 x i32>* %B
311 %tmp3 = icmp ugt <4 x i32> %tmp1, %tmp2
312 %tmp4 = sext <4 x i1> %tmp3 to <4 x i32>
316 define <4 x i32> @vcgtQf32(<4 x float>* %A, <4 x float>* %B) nounwind {
317 ; ALLOC-LABEL: vcgtQf32:
319 ; ALLOC-NEXT: vld1.64 {d16, d17}, [r1]
320 ; ALLOC-NEXT: vld1.64 {d18, d19}, [r0]
321 ; ALLOC-NEXT: vcgt.f32 q8, q9, q8
322 ; ALLOC-NEXT: vmov r0, r1, d16
323 ; ALLOC-NEXT: vmov r2, r3, d17
324 ; ALLOC-NEXT: mov pc, lr
326 ; BASIC-LABEL: vcgtQf32:
328 ; BASIC-NEXT: vld1.64 {d18, d19}, [r1]
329 ; BASIC-NEXT: vld1.64 {d16, d17}, [r0]
330 ; BASIC-NEXT: vcgt.f32 q8, q8, q9
331 ; BASIC-NEXT: vmov r0, r1, d16
332 ; BASIC-NEXT: vmov r2, r3, d17
333 ; BASIC-NEXT: mov pc, lr
334 %tmp1 = load <4 x float>, <4 x float>* %A
335 %tmp2 = load <4 x float>, <4 x float>* %B
336 %tmp3 = fcmp ogt <4 x float> %tmp1, %tmp2
337 %tmp4 = sext <4 x i1> %tmp3 to <4 x i32>
341 define <2 x i32> @vacgtf32(<2 x float>* %A, <2 x float>* %B) nounwind {
342 ; ALLOC-LABEL: vacgtf32:
344 ; ALLOC-NEXT: vldr d16, [r1]
345 ; ALLOC-NEXT: vldr d17, [r0]
346 ; ALLOC-NEXT: vacgt.f32 d16, d17, d16
347 ; ALLOC-NEXT: vmov r0, r1, d16
348 ; ALLOC-NEXT: mov pc, lr
350 ; BASIC-LABEL: vacgtf32:
352 ; BASIC-NEXT: vldr d17, [r1]
353 ; BASIC-NEXT: vldr d16, [r0]
354 ; BASIC-NEXT: vacgt.f32 d16, d16, d17
355 ; BASIC-NEXT: vmov r0, r1, d16
356 ; BASIC-NEXT: mov pc, lr
357 %tmp1 = load <2 x float>, <2 x float>* %A
358 %tmp2 = load <2 x float>, <2 x float>* %B
359 %tmp3 = call <2 x i32> @llvm.arm.neon.vacgt.v2i32.v2f32(<2 x float> %tmp1, <2 x float> %tmp2)
363 define <4 x i32> @vacgtQf32(<4 x float>* %A, <4 x float>* %B) nounwind {
364 ; ALLOC-LABEL: vacgtQf32:
366 ; ALLOC-NEXT: vld1.64 {d16, d17}, [r1]
367 ; ALLOC-NEXT: vld1.64 {d18, d19}, [r0]
368 ; ALLOC-NEXT: vacgt.f32 q8, q9, q8
369 ; ALLOC-NEXT: vmov r0, r1, d16
370 ; ALLOC-NEXT: vmov r2, r3, d17
371 ; ALLOC-NEXT: mov pc, lr
373 ; BASIC-LABEL: vacgtQf32:
375 ; BASIC-NEXT: vld1.64 {d18, d19}, [r1]
376 ; BASIC-NEXT: vld1.64 {d16, d17}, [r0]
377 ; BASIC-NEXT: vacgt.f32 q8, q8, q9
378 ; BASIC-NEXT: vmov r0, r1, d16
379 ; BASIC-NEXT: vmov r2, r3, d17
380 ; BASIC-NEXT: mov pc, lr
381 %tmp1 = load <4 x float>, <4 x float>* %A
382 %tmp2 = load <4 x float>, <4 x float>* %B
383 %tmp3 = call <4 x i32> @llvm.arm.neon.vacgt.v4i32.v4f32(<4 x float> %tmp1, <4 x float> %tmp2)
388 define <4 x i32> @vcgt_zext(<4 x float>* %A, <4 x float>* %B) nounwind {
389 ; ALLOC-LABEL: vcgt_zext:
391 ; ALLOC-NEXT: vld1.64 {d16, d17}, [r1]
392 ; ALLOC-NEXT: vld1.64 {d18, d19}, [r0]
393 ; ALLOC-NEXT: vcgt.f32 q8, q9, q8
394 ; ALLOC-NEXT: vmov.i32 q9, #0x1
395 ; ALLOC-NEXT: vand q8, q8, q9
396 ; ALLOC-NEXT: vmov r0, r1, d16
397 ; ALLOC-NEXT: vmov r2, r3, d17
398 ; ALLOC-NEXT: mov pc, lr
400 ; BASIC-LABEL: vcgt_zext:
402 ; BASIC-NEXT: vld1.64 {d18, d19}, [r1]
403 ; BASIC-NEXT: vld1.64 {d16, d17}, [r0]
404 ; BASIC-NEXT: vcgt.f32 q9, q8, q9
405 ; BASIC-NEXT: vmov.i32 q8, #0x1
406 ; BASIC-NEXT: vand q8, q9, q8
407 ; BASIC-NEXT: vmov r0, r1, d16
408 ; BASIC-NEXT: vmov r2, r3, d17
409 ; BASIC-NEXT: mov pc, lr
410 %tmp1 = load <4 x float>, <4 x float>* %A
411 %tmp2 = load <4 x float>, <4 x float>* %B
412 %tmp3 = fcmp ogt <4 x float> %tmp1, %tmp2
413 %tmp4 = zext <4 x i1> %tmp3 to <4 x i32>
417 declare <2 x i32> @llvm.arm.neon.vacgt.v2i32.v2f32(<2 x float>, <2 x float>) nounwind readnone
418 declare <4 x i32> @llvm.arm.neon.vacgt.v4i32.v4f32(<4 x float>, <4 x float>) nounwind readnone
420 define <8 x i8> @vcgti8Z(<8 x i8>* %A) nounwind {
421 ; CHECK-LABEL: vcgti8Z:
423 ; CHECK-NEXT: vldr d16, [r0]
424 ; CHECK-NEXT: vcgt.s8 d16, d16, #0
425 ; CHECK-NEXT: vmov r0, r1, d16
426 ; CHECK-NEXT: mov pc, lr
427 %tmp1 = load <8 x i8>, <8 x i8>* %A
428 %tmp3 = icmp sgt <8 x i8> %tmp1, <i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0>
429 %tmp4 = sext <8 x i1> %tmp3 to <8 x i8>
433 define <8 x i8> @vclti8Z(<8 x i8>* %A) nounwind {
434 ; CHECK-LABEL: vclti8Z:
436 ; CHECK-NEXT: vldr d16, [r0]
437 ; CHECK-NEXT: vclt.s8 d16, d16, #0
438 ; CHECK-NEXT: vmov r0, r1, d16
439 ; CHECK-NEXT: mov pc, lr
440 %tmp1 = load <8 x i8>, <8 x i8>* %A
441 %tmp3 = icmp slt <8 x i8> %tmp1, <i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0>
442 %tmp4 = sext <8 x i1> %tmp3 to <8 x i8>