1 // NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py
2 // RUN: %clang_cc1 -triple thumbv8.1m.main-none-none-eabi -target-feature +mve -mfloat-abi hard -O0 -disable-O0-optnone -emit-llvm -o - %s | opt -S -passes=mem2reg | FileCheck %s
3 // RUN: %clang_cc1 -triple thumbv8.1m.main-none-none-eabi -target-feature +mve -mfloat-abi hard -O0 -disable-O0-optnone -DPOLYMORPHIC -emit-llvm -o - %s | opt -S -passes=mem2reg | FileCheck %s
4
5 // REQUIRES: aarch64-registered-target || arm-registered-target
6
7 #include <arm_mve.h>
8
9 // CHECK-LABEL: @test_vaddvq_s8(
10 // CHECK-NEXT: entry:
11 // CHECK-NEXT: [[TMP0:%.*]] = call i32 @llvm.arm.mve.addv.v16i8(<16 x i8> [[A:%.*]], i32 0)
12 // CHECK-NEXT: ret i32 [[TMP0]]
13 //
test_vaddvq_s8(int8x16_t a)14 int32_t test_vaddvq_s8(int8x16_t a) {
15 #ifdef POLYMORPHIC
16 return vaddvq(a);
17 #else /* POLYMORPHIC */
18 return vaddvq_s8(a);
19 #endif /* POLYMORPHIC */
20 }
21
22 // CHECK-LABEL: @test_vaddvq_s16(
23 // CHECK-NEXT: entry:
24 // CHECK-NEXT: [[TMP0:%.*]] = call i32 @llvm.arm.mve.addv.v8i16(<8 x i16> [[A:%.*]], i32 0)
25 // CHECK-NEXT: ret i32 [[TMP0]]
26 //
test_vaddvq_s16(int16x8_t a)27 int32_t test_vaddvq_s16(int16x8_t a) {
28 #ifdef POLYMORPHIC
29 return vaddvq(a);
30 #else /* POLYMORPHIC */
31 return vaddvq_s16(a);
32 #endif /* POLYMORPHIC */
33 }
34
35 // CHECK-LABEL: @test_vaddvq_s32(
36 // CHECK-NEXT: entry:
37 // CHECK-NEXT: [[TMP0:%.*]] = call i32 @llvm.arm.mve.addv.v4i32(<4 x i32> [[A:%.*]], i32 0)
38 // CHECK-NEXT: ret i32 [[TMP0]]
39 //
test_vaddvq_s32(int32x4_t a)40 int32_t test_vaddvq_s32(int32x4_t a) {
41 #ifdef POLYMORPHIC
42 return vaddvq(a);
43 #else /* POLYMORPHIC */
44 return vaddvq_s32(a);
45 #endif /* POLYMORPHIC */
46 }
47
48 // CHECK-LABEL: @test_vaddvq_u8(
49 // CHECK-NEXT: entry:
50 // CHECK-NEXT: [[TMP0:%.*]] = call i32 @llvm.arm.mve.addv.v16i8(<16 x i8> [[A:%.*]], i32 1)
51 // CHECK-NEXT: ret i32 [[TMP0]]
52 //
test_vaddvq_u8(uint8x16_t a)53 uint32_t test_vaddvq_u8(uint8x16_t a) {
54 #ifdef POLYMORPHIC
55 return vaddvq(a);
56 #else /* POLYMORPHIC */
57 return vaddvq_u8(a);
58 #endif /* POLYMORPHIC */
59 }
60
61 // CHECK-LABEL: @test_vaddvq_u16(
62 // CHECK-NEXT: entry:
63 // CHECK-NEXT: [[TMP0:%.*]] = call i32 @llvm.arm.mve.addv.v8i16(<8 x i16> [[A:%.*]], i32 1)
64 // CHECK-NEXT: ret i32 [[TMP0]]
65 //
test_vaddvq_u16(uint16x8_t a)66 uint32_t test_vaddvq_u16(uint16x8_t a) {
67 #ifdef POLYMORPHIC
68 return vaddvq(a);
69 #else /* POLYMORPHIC */
70 return vaddvq_u16(a);
71 #endif /* POLYMORPHIC */
72 }
73
74 // CHECK-LABEL: @test_vaddvq_u32(
75 // CHECK-NEXT: entry:
76 // CHECK-NEXT: [[TMP0:%.*]] = call i32 @llvm.arm.mve.addv.v4i32(<4 x i32> [[A:%.*]], i32 1)
77 // CHECK-NEXT: ret i32 [[TMP0]]
78 //
test_vaddvq_u32(uint32x4_t a)79 uint32_t test_vaddvq_u32(uint32x4_t a) {
80 #ifdef POLYMORPHIC
81 return vaddvq(a);
82 #else /* POLYMORPHIC */
83 return vaddvq_u32(a);
84 #endif /* POLYMORPHIC */
85 }
86
87 // CHECK-LABEL: @test_vaddvaq_s8(
88 // CHECK-NEXT: entry:
89 // CHECK-NEXT: [[TMP0:%.*]] = call i32 @llvm.arm.mve.addv.v16i8(<16 x i8> [[B:%.*]], i32 0)
90 // CHECK-NEXT: [[TMP1:%.*]] = add i32 [[TMP0]], [[A:%.*]]
91 // CHECK-NEXT: ret i32 [[TMP1]]
92 //
test_vaddvaq_s8(int32_t a,int8x16_t b)93 int32_t test_vaddvaq_s8(int32_t a, int8x16_t b) {
94 #ifdef POLYMORPHIC
95 return vaddvaq(a, b);
96 #else /* POLYMORPHIC */
97 return vaddvaq_s8(a, b);
98 #endif /* POLYMORPHIC */
99 }
100
101 // CHECK-LABEL: @test_vaddvaq_s16(
102 // CHECK-NEXT: entry:
103 // CHECK-NEXT: [[TMP0:%.*]] = call i32 @llvm.arm.mve.addv.v8i16(<8 x i16> [[B:%.*]], i32 0)
104 // CHECK-NEXT: [[TMP1:%.*]] = add i32 [[TMP0]], [[A:%.*]]
105 // CHECK-NEXT: ret i32 [[TMP1]]
106 //
test_vaddvaq_s16(int32_t a,int16x8_t b)107 int32_t test_vaddvaq_s16(int32_t a, int16x8_t b) {
108 #ifdef POLYMORPHIC
109 return vaddvaq(a, b);
110 #else /* POLYMORPHIC */
111 return vaddvaq_s16(a, b);
112 #endif /* POLYMORPHIC */
113 }
114
115 // CHECK-LABEL: @test_vaddvaq_s32(
116 // CHECK-NEXT: entry:
117 // CHECK-NEXT: [[TMP0:%.*]] = call i32 @llvm.arm.mve.addv.v4i32(<4 x i32> [[B:%.*]], i32 0)
118 // CHECK-NEXT: [[TMP1:%.*]] = add i32 [[TMP0]], [[A:%.*]]
119 // CHECK-NEXT: ret i32 [[TMP1]]
120 //
test_vaddvaq_s32(int32_t a,int32x4_t b)121 int32_t test_vaddvaq_s32(int32_t a, int32x4_t b) {
122 #ifdef POLYMORPHIC
123 return vaddvaq(a, b);
124 #else /* POLYMORPHIC */
125 return vaddvaq_s32(a, b);
126 #endif /* POLYMORPHIC */
127 }
128
129 // CHECK-LABEL: @test_vaddvaq_u8(
130 // CHECK-NEXT: entry:
131 // CHECK-NEXT: [[TMP0:%.*]] = call i32 @llvm.arm.mve.addv.v16i8(<16 x i8> [[B:%.*]], i32 1)
132 // CHECK-NEXT: [[TMP1:%.*]] = add i32 [[TMP0]], [[A:%.*]]
133 // CHECK-NEXT: ret i32 [[TMP1]]
134 //
test_vaddvaq_u8(uint32_t a,uint8x16_t b)135 uint32_t test_vaddvaq_u8(uint32_t a, uint8x16_t b) {
136 #ifdef POLYMORPHIC
137 return vaddvaq(a, b);
138 #else /* POLYMORPHIC */
139 return vaddvaq_u8(a, b);
140 #endif /* POLYMORPHIC */
141 }
142
143 // CHECK-LABEL: @test_vaddvaq_u16(
144 // CHECK-NEXT: entry:
145 // CHECK-NEXT: [[TMP0:%.*]] = call i32 @llvm.arm.mve.addv.v8i16(<8 x i16> [[B:%.*]], i32 1)
146 // CHECK-NEXT: [[TMP1:%.*]] = add i32 [[TMP0]], [[A:%.*]]
147 // CHECK-NEXT: ret i32 [[TMP1]]
148 //
test_vaddvaq_u16(uint32_t a,uint16x8_t b)149 uint32_t test_vaddvaq_u16(uint32_t a, uint16x8_t b) {
150 #ifdef POLYMORPHIC
151 return vaddvaq(a, b);
152 #else /* POLYMORPHIC */
153 return vaddvaq_u16(a, b);
154 #endif /* POLYMORPHIC */
155 }
156
157 // CHECK-LABEL: @test_vaddvaq_u32(
158 // CHECK-NEXT: entry:
159 // CHECK-NEXT: [[TMP0:%.*]] = call i32 @llvm.arm.mve.addv.v4i32(<4 x i32> [[B:%.*]], i32 1)
160 // CHECK-NEXT: [[TMP1:%.*]] = add i32 [[TMP0]], [[A:%.*]]
161 // CHECK-NEXT: ret i32 [[TMP1]]
162 //
test_vaddvaq_u32(uint32_t a,uint32x4_t b)163 uint32_t test_vaddvaq_u32(uint32_t a, uint32x4_t b) {
164 #ifdef POLYMORPHIC
165 return vaddvaq(a, b);
166 #else /* POLYMORPHIC */
167 return vaddvaq_u32(a, b);
168 #endif /* POLYMORPHIC */
169 }
170
171 // CHECK-LABEL: @test_vaddvq_p_s8(
172 // CHECK-NEXT: entry:
173 // CHECK-NEXT: [[TMP0:%.*]] = zext i16 [[P:%.*]] to i32
174 // CHECK-NEXT: [[TMP1:%.*]] = call <16 x i1> @llvm.arm.mve.pred.i2v.v16i1(i32 [[TMP0]])
175 // CHECK-NEXT: [[TMP2:%.*]] = call i32 @llvm.arm.mve.addv.predicated.v16i8.v16i1(<16 x i8> [[A:%.*]], i32 0, <16 x i1> [[TMP1]])
176 // CHECK-NEXT: ret i32 [[TMP2]]
177 //
test_vaddvq_p_s8(int8x16_t a,mve_pred16_t p)178 int32_t test_vaddvq_p_s8(int8x16_t a, mve_pred16_t p) {
179 #ifdef POLYMORPHIC
180 return vaddvq_p(a, p);
181 #else /* POLYMORPHIC */
182 return vaddvq_p_s8(a, p);
183 #endif /* POLYMORPHIC */
184 }
185
186 // CHECK-LABEL: @test_vaddvq_p_s16(
187 // CHECK-NEXT: entry:
188 // CHECK-NEXT: [[TMP0:%.*]] = zext i16 [[P:%.*]] to i32
189 // CHECK-NEXT: [[TMP1:%.*]] = call <8 x i1> @llvm.arm.mve.pred.i2v.v8i1(i32 [[TMP0]])
190 // CHECK-NEXT: [[TMP2:%.*]] = call i32 @llvm.arm.mve.addv.predicated.v8i16.v8i1(<8 x i16> [[A:%.*]], i32 0, <8 x i1> [[TMP1]])
191 // CHECK-NEXT: ret i32 [[TMP2]]
192 //
test_vaddvq_p_s16(int16x8_t a,mve_pred16_t p)193 int32_t test_vaddvq_p_s16(int16x8_t a, mve_pred16_t p) {
194 #ifdef POLYMORPHIC
195 return vaddvq_p(a, p);
196 #else /* POLYMORPHIC */
197 return vaddvq_p_s16(a, p);
198 #endif /* POLYMORPHIC */
199 }
200
201 // CHECK-LABEL: @test_vaddvq_p_s32(
202 // CHECK-NEXT: entry:
203 // CHECK-NEXT: [[TMP0:%.*]] = zext i16 [[P:%.*]] to i32
204 // CHECK-NEXT: [[TMP1:%.*]] = call <4 x i1> @llvm.arm.mve.pred.i2v.v4i1(i32 [[TMP0]])
205 // CHECK-NEXT: [[TMP2:%.*]] = call i32 @llvm.arm.mve.addv.predicated.v4i32.v4i1(<4 x i32> [[A:%.*]], i32 0, <4 x i1> [[TMP1]])
206 // CHECK-NEXT: ret i32 [[TMP2]]
207 //
test_vaddvq_p_s32(int32x4_t a,mve_pred16_t p)208 int32_t test_vaddvq_p_s32(int32x4_t a, mve_pred16_t p) {
209 #ifdef POLYMORPHIC
210 return vaddvq_p(a, p);
211 #else /* POLYMORPHIC */
212 return vaddvq_p_s32(a, p);
213 #endif /* POLYMORPHIC */
214 }
215
216 // CHECK-LABEL: @test_vaddvq_p_u8(
217 // CHECK-NEXT: entry:
218 // CHECK-NEXT: [[TMP0:%.*]] = zext i16 [[P:%.*]] to i32
219 // CHECK-NEXT: [[TMP1:%.*]] = call <16 x i1> @llvm.arm.mve.pred.i2v.v16i1(i32 [[TMP0]])
220 // CHECK-NEXT: [[TMP2:%.*]] = call i32 @llvm.arm.mve.addv.predicated.v16i8.v16i1(<16 x i8> [[A:%.*]], i32 1, <16 x i1> [[TMP1]])
221 // CHECK-NEXT: ret i32 [[TMP2]]
222 //
test_vaddvq_p_u8(uint8x16_t a,mve_pred16_t p)223 uint32_t test_vaddvq_p_u8(uint8x16_t a, mve_pred16_t p) {
224 #ifdef POLYMORPHIC
225 return vaddvq_p(a, p);
226 #else /* POLYMORPHIC */
227 return vaddvq_p_u8(a, p);
228 #endif /* POLYMORPHIC */
229 }
230
231 // CHECK-LABEL: @test_vaddvq_p_u16(
232 // CHECK-NEXT: entry:
233 // CHECK-NEXT: [[TMP0:%.*]] = zext i16 [[P:%.*]] to i32
234 // CHECK-NEXT: [[TMP1:%.*]] = call <8 x i1> @llvm.arm.mve.pred.i2v.v8i1(i32 [[TMP0]])
235 // CHECK-NEXT: [[TMP2:%.*]] = call i32 @llvm.arm.mve.addv.predicated.v8i16.v8i1(<8 x i16> [[A:%.*]], i32 1, <8 x i1> [[TMP1]])
236 // CHECK-NEXT: ret i32 [[TMP2]]
237 //
test_vaddvq_p_u16(uint16x8_t a,mve_pred16_t p)238 uint32_t test_vaddvq_p_u16(uint16x8_t a, mve_pred16_t p) {
239 #ifdef POLYMORPHIC
240 return vaddvq_p(a, p);
241 #else /* POLYMORPHIC */
242 return vaddvq_p_u16(a, p);
243 #endif /* POLYMORPHIC */
244 }
245
246 // CHECK-LABEL: @test_vaddvq_p_u32(
247 // CHECK-NEXT: entry:
248 // CHECK-NEXT: [[TMP0:%.*]] = zext i16 [[P:%.*]] to i32
249 // CHECK-NEXT: [[TMP1:%.*]] = call <4 x i1> @llvm.arm.mve.pred.i2v.v4i1(i32 [[TMP0]])
250 // CHECK-NEXT: [[TMP2:%.*]] = call i32 @llvm.arm.mve.addv.predicated.v4i32.v4i1(<4 x i32> [[A:%.*]], i32 1, <4 x i1> [[TMP1]])
251 // CHECK-NEXT: ret i32 [[TMP2]]
252 //
test_vaddvq_p_u32(uint32x4_t a,mve_pred16_t p)253 uint32_t test_vaddvq_p_u32(uint32x4_t a, mve_pred16_t p) {
254 #ifdef POLYMORPHIC
255 return vaddvq_p(a, p);
256 #else /* POLYMORPHIC */
257 return vaddvq_p_u32(a, p);
258 #endif /* POLYMORPHIC */
259 }
260
261 // CHECK-LABEL: @test_vaddvaq_p_s8(
262 // CHECK-NEXT: entry:
263 // CHECK-NEXT: [[TMP0:%.*]] = zext i16 [[P:%.*]] to i32
264 // CHECK-NEXT: [[TMP1:%.*]] = call <16 x i1> @llvm.arm.mve.pred.i2v.v16i1(i32 [[TMP0]])
265 // CHECK-NEXT: [[TMP2:%.*]] = call i32 @llvm.arm.mve.addv.predicated.v16i8.v16i1(<16 x i8> [[B:%.*]], i32 0, <16 x i1> [[TMP1]])
266 // CHECK-NEXT: [[TMP3:%.*]] = add i32 [[TMP2]], [[A:%.*]]
267 // CHECK-NEXT: ret i32 [[TMP3]]
268 //
test_vaddvaq_p_s8(int32_t a,int8x16_t b,mve_pred16_t p)269 int32_t test_vaddvaq_p_s8(int32_t a, int8x16_t b, mve_pred16_t p) {
270 #ifdef POLYMORPHIC
271 return vaddvaq_p(a, b, p);
272 #else /* POLYMORPHIC */
273 return vaddvaq_p_s8(a, b, p);
274 #endif /* POLYMORPHIC */
275 }
276
277 // CHECK-LABEL: @test_vaddvaq_p_s16(
278 // CHECK-NEXT: entry:
279 // CHECK-NEXT: [[TMP0:%.*]] = zext i16 [[P:%.*]] to i32
280 // CHECK-NEXT: [[TMP1:%.*]] = call <8 x i1> @llvm.arm.mve.pred.i2v.v8i1(i32 [[TMP0]])
281 // CHECK-NEXT: [[TMP2:%.*]] = call i32 @llvm.arm.mve.addv.predicated.v8i16.v8i1(<8 x i16> [[B:%.*]], i32 0, <8 x i1> [[TMP1]])
282 // CHECK-NEXT: [[TMP3:%.*]] = add i32 [[TMP2]], [[A:%.*]]
283 // CHECK-NEXT: ret i32 [[TMP3]]
284 //
test_vaddvaq_p_s16(int32_t a,int16x8_t b,mve_pred16_t p)285 int32_t test_vaddvaq_p_s16(int32_t a, int16x8_t b, mve_pred16_t p) {
286 #ifdef POLYMORPHIC
287 return vaddvaq_p(a, b, p);
288 #else /* POLYMORPHIC */
289 return vaddvaq_p_s16(a, b, p);
290 #endif /* POLYMORPHIC */
291 }
292
293 // CHECK-LABEL: @test_vaddvaq_p_s32(
294 // CHECK-NEXT: entry:
295 // CHECK-NEXT: [[TMP0:%.*]] = zext i16 [[P:%.*]] to i32
296 // CHECK-NEXT: [[TMP1:%.*]] = call <4 x i1> @llvm.arm.mve.pred.i2v.v4i1(i32 [[TMP0]])
297 // CHECK-NEXT: [[TMP2:%.*]] = call i32 @llvm.arm.mve.addv.predicated.v4i32.v4i1(<4 x i32> [[B:%.*]], i32 0, <4 x i1> [[TMP1]])
298 // CHECK-NEXT: [[TMP3:%.*]] = add i32 [[TMP2]], [[A:%.*]]
299 // CHECK-NEXT: ret i32 [[TMP3]]
300 //
test_vaddvaq_p_s32(int32_t a,int32x4_t b,mve_pred16_t p)301 int32_t test_vaddvaq_p_s32(int32_t a, int32x4_t b, mve_pred16_t p) {
302 #ifdef POLYMORPHIC
303 return vaddvaq_p(a, b, p);
304 #else /* POLYMORPHIC */
305 return vaddvaq_p_s32(a, b, p);
306 #endif /* POLYMORPHIC */
307 }
308
309 // CHECK-LABEL: @test_vaddvaq_p_u8(
310 // CHECK-NEXT: entry:
311 // CHECK-NEXT: [[TMP0:%.*]] = zext i16 [[P:%.*]] to i32
312 // CHECK-NEXT: [[TMP1:%.*]] = call <16 x i1> @llvm.arm.mve.pred.i2v.v16i1(i32 [[TMP0]])
313 // CHECK-NEXT: [[TMP2:%.*]] = call i32 @llvm.arm.mve.addv.predicated.v16i8.v16i1(<16 x i8> [[B:%.*]], i32 1, <16 x i1> [[TMP1]])
314 // CHECK-NEXT: [[TMP3:%.*]] = add i32 [[TMP2]], [[A:%.*]]
315 // CHECK-NEXT: ret i32 [[TMP3]]
316 //
test_vaddvaq_p_u8(uint32_t a,uint8x16_t b,mve_pred16_t p)317 uint32_t test_vaddvaq_p_u8(uint32_t a, uint8x16_t b, mve_pred16_t p) {
318 #ifdef POLYMORPHIC
319 return vaddvaq_p(a, b, p);
320 #else /* POLYMORPHIC */
321 return vaddvaq_p_u8(a, b, p);
322 #endif /* POLYMORPHIC */
323 }
324
325 // CHECK-LABEL: @test_vaddvaq_p_u16(
326 // CHECK-NEXT: entry:
327 // CHECK-NEXT: [[TMP0:%.*]] = zext i16 [[P:%.*]] to i32
328 // CHECK-NEXT: [[TMP1:%.*]] = call <8 x i1> @llvm.arm.mve.pred.i2v.v8i1(i32 [[TMP0]])
329 // CHECK-NEXT: [[TMP2:%.*]] = call i32 @llvm.arm.mve.addv.predicated.v8i16.v8i1(<8 x i16> [[B:%.*]], i32 1, <8 x i1> [[TMP1]])
330 // CHECK-NEXT: [[TMP3:%.*]] = add i32 [[TMP2]], [[A:%.*]]
331 // CHECK-NEXT: ret i32 [[TMP3]]
332 //
test_vaddvaq_p_u16(uint32_t a,uint16x8_t b,mve_pred16_t p)333 uint32_t test_vaddvaq_p_u16(uint32_t a, uint16x8_t b, mve_pred16_t p) {
334 #ifdef POLYMORPHIC
335 return vaddvaq_p(a, b, p);
336 #else /* POLYMORPHIC */
337 return vaddvaq_p_u16(a, b, p);
338 #endif /* POLYMORPHIC */
339 }
340
341 // CHECK-LABEL: @test_vaddvaq_p_u32(
342 // CHECK-NEXT: entry:
343 // CHECK-NEXT: [[TMP0:%.*]] = zext i16 [[P:%.*]] to i32
344 // CHECK-NEXT: [[TMP1:%.*]] = call <4 x i1> @llvm.arm.mve.pred.i2v.v4i1(i32 [[TMP0]])
345 // CHECK-NEXT: [[TMP2:%.*]] = call i32 @llvm.arm.mve.addv.predicated.v4i32.v4i1(<4 x i32> [[B:%.*]], i32 1, <4 x i1> [[TMP1]])
346 // CHECK-NEXT: [[TMP3:%.*]] = add i32 [[TMP2]], [[A:%.*]]
347 // CHECK-NEXT: ret i32 [[TMP3]]
348 //
test_vaddvaq_p_u32(uint32_t a,uint32x4_t b,mve_pred16_t p)349 uint32_t test_vaddvaq_p_u32(uint32_t a, uint32x4_t b, mve_pred16_t p) {
350 #ifdef POLYMORPHIC
351 return vaddvaq_p(a, b, p);
352 #else /* POLYMORPHIC */
353 return vaddvaq_p_u32(a, b, p);
354 #endif /* POLYMORPHIC */
355 }
356
357 // CHECK-LABEL: @test_vaddlvq_s32(
358 // CHECK-NEXT: entry:
359 // CHECK-NEXT: [[TMP0:%.*]] = call i64 @llvm.arm.mve.addlv.v4i32(<4 x i32> [[A:%.*]], i32 0)
360 // CHECK-NEXT: ret i64 [[TMP0]]
361 //
test_vaddlvq_s32(int32x4_t a)362 int64_t test_vaddlvq_s32(int32x4_t a) {
363 #ifdef POLYMORPHIC
364 return vaddlvq(a);
365 #else /* POLYMORPHIC */
366 return vaddlvq_s32(a);
367 #endif /* POLYMORPHIC */
368 }
369
370 // CHECK-LABEL: @test_vaddlvq_u32(
371 // CHECK-NEXT: entry:
372 // CHECK-NEXT: [[TMP0:%.*]] = call i64 @llvm.arm.mve.addlv.v4i32(<4 x i32> [[A:%.*]], i32 1)
373 // CHECK-NEXT: ret i64 [[TMP0]]
374 //
test_vaddlvq_u32(uint32x4_t a)375 uint64_t test_vaddlvq_u32(uint32x4_t a) {
376 #ifdef POLYMORPHIC
377 return vaddlvq(a);
378 #else /* POLYMORPHIC */
379 return vaddlvq_u32(a);
380 #endif /* POLYMORPHIC */
381 }
382
383 // CHECK-LABEL: @test_vaddlvaq_s32(
384 // CHECK-NEXT: entry:
385 // CHECK-NEXT: [[TMP0:%.*]] = call i64 @llvm.arm.mve.addlv.v4i32(<4 x i32> [[B:%.*]], i32 0)
386 // CHECK-NEXT: [[TMP1:%.*]] = add i64 [[TMP0]], [[A:%.*]]
387 // CHECK-NEXT: ret i64 [[TMP1]]
388 //
test_vaddlvaq_s32(int64_t a,int32x4_t b)389 int64_t test_vaddlvaq_s32(int64_t a, int32x4_t b) {
390 #ifdef POLYMORPHIC
391 return vaddlvaq(a, b);
392 #else /* POLYMORPHIC */
393 return vaddlvaq_s32(a, b);
394 #endif /* POLYMORPHIC */
395 }
396
397 // CHECK-LABEL: @test_vaddlvaq_u32(
398 // CHECK-NEXT: entry:
399 // CHECK-NEXT: [[TMP0:%.*]] = call i64 @llvm.arm.mve.addlv.v4i32(<4 x i32> [[B:%.*]], i32 1)
400 // CHECK-NEXT: [[TMP1:%.*]] = add i64 [[TMP0]], [[A:%.*]]
401 // CHECK-NEXT: ret i64 [[TMP1]]
402 //
test_vaddlvaq_u32(uint64_t a,uint32x4_t b)403 uint64_t test_vaddlvaq_u32(uint64_t a, uint32x4_t b) {
404 #ifdef POLYMORPHIC
405 return vaddlvaq(a, b);
406 #else /* POLYMORPHIC */
407 return vaddlvaq_u32(a, b);
408 #endif /* POLYMORPHIC */
409 }
410
411 // CHECK-LABEL: @test_vaddlvq_p_s32(
412 // CHECK-NEXT: entry:
413 // CHECK-NEXT: [[TMP0:%.*]] = zext i16 [[P:%.*]] to i32
414 // CHECK-NEXT: [[TMP1:%.*]] = call <4 x i1> @llvm.arm.mve.pred.i2v.v4i1(i32 [[TMP0]])
415 // CHECK-NEXT: [[TMP2:%.*]] = call i64 @llvm.arm.mve.addlv.predicated.v4i32.v4i1(<4 x i32> [[A:%.*]], i32 0, <4 x i1> [[TMP1]])
416 // CHECK-NEXT: ret i64 [[TMP2]]
417 //
test_vaddlvq_p_s32(int32x4_t a,mve_pred16_t p)418 int64_t test_vaddlvq_p_s32(int32x4_t a, mve_pred16_t p) {
419 #ifdef POLYMORPHIC
420 return vaddlvq_p(a, p);
421 #else /* POLYMORPHIC */
422 return vaddlvq_p_s32(a, p);
423 #endif /* POLYMORPHIC */
424 }
425
426 // CHECK-LABEL: @test_vaddlvq_p_u32(
427 // CHECK-NEXT: entry:
428 // CHECK-NEXT: [[TMP0:%.*]] = zext i16 [[P:%.*]] to i32
429 // CHECK-NEXT: [[TMP1:%.*]] = call <4 x i1> @llvm.arm.mve.pred.i2v.v4i1(i32 [[TMP0]])
430 // CHECK-NEXT: [[TMP2:%.*]] = call i64 @llvm.arm.mve.addlv.predicated.v4i32.v4i1(<4 x i32> [[A:%.*]], i32 1, <4 x i1> [[TMP1]])
431 // CHECK-NEXT: ret i64 [[TMP2]]
432 //
test_vaddlvq_p_u32(uint32x4_t a,mve_pred16_t p)433 uint64_t test_vaddlvq_p_u32(uint32x4_t a, mve_pred16_t p) {
434 #ifdef POLYMORPHIC
435 return vaddlvq_p(a, p);
436 #else /* POLYMORPHIC */
437 return vaddlvq_p_u32(a, p);
438 #endif /* POLYMORPHIC */
439 }
440
441 // CHECK-LABEL: @test_vaddlvaq_p_s32(
442 // CHECK-NEXT: entry:
443 // CHECK-NEXT: [[TMP0:%.*]] = zext i16 [[P:%.*]] to i32
444 // CHECK-NEXT: [[TMP1:%.*]] = call <4 x i1> @llvm.arm.mve.pred.i2v.v4i1(i32 [[TMP0]])
445 // CHECK-NEXT: [[TMP2:%.*]] = call i64 @llvm.arm.mve.addlv.predicated.v4i32.v4i1(<4 x i32> [[B:%.*]], i32 0, <4 x i1> [[TMP1]])
446 // CHECK-NEXT: [[TMP3:%.*]] = add i64 [[TMP2]], [[A:%.*]]
447 // CHECK-NEXT: ret i64 [[TMP3]]
448 //
test_vaddlvaq_p_s32(int64_t a,int32x4_t b,mve_pred16_t p)449 int64_t test_vaddlvaq_p_s32(int64_t a, int32x4_t b, mve_pred16_t p) {
450 #ifdef POLYMORPHIC
451 return vaddlvaq_p(a, b, p);
452 #else /* POLYMORPHIC */
453 return vaddlvaq_p_s32(a, b, p);
454 #endif /* POLYMORPHIC */
455 }
456
457 // CHECK-LABEL: @test_vaddlvaq_p_u32(
458 // CHECK-NEXT: entry:
459 // CHECK-NEXT: [[TMP0:%.*]] = zext i16 [[P:%.*]] to i32
460 // CHECK-NEXT: [[TMP1:%.*]] = call <4 x i1> @llvm.arm.mve.pred.i2v.v4i1(i32 [[TMP0]])
461 // CHECK-NEXT: [[TMP2:%.*]] = call i64 @llvm.arm.mve.addlv.predicated.v4i32.v4i1(<4 x i32> [[B:%.*]], i32 1, <4 x i1> [[TMP1]])
462 // CHECK-NEXT: [[TMP3:%.*]] = add i64 [[TMP2]], [[A:%.*]]
463 // CHECK-NEXT: ret i64 [[TMP3]]
464 //
test_vaddlvaq_p_u32(uint64_t a,uint32x4_t b,mve_pred16_t p)465 uint64_t test_vaddlvaq_p_u32(uint64_t a, uint32x4_t b, mve_pred16_t p) {
466 #ifdef POLYMORPHIC
467 return vaddlvaq_p(a, b, p);
468 #else /* POLYMORPHIC */
469 return vaddlvaq_p_u32(a, b, p);
470 #endif /* POLYMORPHIC */
471 }
472
473