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