1 /* { dg-require-effective-target arm_v8_1m_mve_fp_ok } */
2 /* { dg-add-options arm_v8_1m_mve_fp } */
3 /* { dg-additional-options "-O2" } */
4 /* { dg-final { check-function-bodies "**" "" } } */
5
6 #include "arm_mve.h"
7 /*
8 **off_load8_0:
9 ** ...
10 ** vldrb.8 q[0-7], \[r0, #16\]
11 ** ...
12 */
13 int8x16_t off_load8_0 (int8_t * a)
14 {
15 return vld1q_s8 (a + 16);
16 }
17
18 /*
19 **off_load8_1:
20 ** ...
21 ** vldrb.u16 q[0-7], \[r0, #1\]
22 ** ...
23 */
24 uint16x8_t off_load8_1 (uint8_t * a)
25 {
26 return vldrbq_u16 (a + 1);
27 }
28
29 /*
30 **off_load8_2:
31 ** ...
32 ** vldrb.s32 q[0-7], \[r0, #127\]
33 ** ...
34 */
35 int32x4_t off_load8_2 (int8_t * a)
36 {
37 return vldrbq_s32 (a + 127);
38 }
39
40 /*
41 **off_load8_3:
42 ** ...
43 ** vldrb.8 q[0-7], \[r0, #-127\]
44 ** ...
45 */
46 uint8x16_t off_load8_3 (uint8_t * a)
47 {
48 return vldrbq_u8 (a - 127);
49 }
50
51 /*
52 **not_off_load8_0:
53 ** ...
54 ** vldrb.8 q[0-7], \[r[0-7]+\]
55 ** ...
56 */
57 int8x16_t not_off_load8_0 (int8_t * a)
58 {
59 return vld1q_s8 (a + 128);
60 }
61
62 /*
63 **off_loadfp16_0:
64 ** ...
65 ** vldrh.16 q[0-7], \[r0, #-244\]
66 ** ...
67 */
68 float16x8_t off_loadfp16_0 (float16_t *a)
69 {
70 return vld1q_f16 (a - 122);
71 }
72
73 /*
74 **off_load16_0:
75 ** ...
76 ** vldrh.16 q[0-7], \[r0, #-2\]
77 ** ...
78 */
79 uint16x8_t off_load16_0 (uint16_t * a)
80 {
81 return vld1q_u16 (a - 1);
82 }
83
84 /*
85 **off_load16_1:
86 ** ...
87 ** vldrh.u32 q[0-7], \[r0, #254\]
88 ** ...
89 */
90 uint32x4_t off_load16_1 (uint16_t * a)
91 {
92 return vldrhq_u32 (a + 127);
93 }
94
95 /*
96 **not_off_load16_0:
97 ** ...
98 ** vldrh.16 q[0-7], \[r[0-7]+\]
99 ** ...
100 */
101 int16x8_t not_off_load16_0 (int8_t * a)
102 {
103 return vld1q_s16 ((int16_t *)(a + 1));
104 }
105
106 /*
107 **not_off_load16_1:
108 ** ...
109 ** vldrh.u32 q[0-7], \[r[0-7]+\]
110 ** ...
111 */
112 uint32x4_t not_off_load16_1 (uint16_t * a)
113 {
114 return vldrhq_u32 ((a - 128));
115 }
116
117 /*
118 **off_loadfp32_0:
119 ** ...
120 ** vldrw.32 q[0-7], \[r0, #24\]
121 ** ...
122 */
123 float32x4_t off_loadfp32_0 (float32_t *a)
124 {
125 return vld1q_f32 (a + 6);
126 }
127
128 /*
129 **off_load32_0:
130 ** ...
131 ** vldrw.32 q[0-7], \[r0, #4\]
132 ** ...
133 */
134 uint32x4_t off_load32_0 (uint32_t * a)
135 {
136 return vld1q_u32 (a + 1);
137 }
138
139 /*
140 **off_load32_1:
141 ** ...
142 ** vldrw.32 q[0-7], \[r0, #-508\]
143 ** ...
144 */
145 int32x4_t off_load32_1 (int32_t * a)
146 {
147 return vldrwq_s32 (a - 127);
148 }
149 /*
150 **pre_load8_0:
151 ** ...
152 ** vldrb.8 q[0-7], \[r0, #16\]!
153 ** ...
154 */
155 int8_t* pre_load8_0 (int8_t * a, int8x16_t *v)
156 {
157 a += 16;
158 *v = vld1q_s8 (a);
159 return a;
160 }
161
162 /*
163 **pre_load8_1:
164 ** ...
165 ** vldrb.u16 q[0-7], \[r0, #4\]!
166 ** ...
167 */
168 uint8_t* pre_load8_1 (uint8_t * a, uint16x8_t *v)
169 {
170 a += 4;
171 *v = vldrbq_u16 (a);
172 return a;
173 }
174
175 /*
176 **pre_loadfp16_0:
177 ** ...
178 ** vldrh.16 q[0-7], \[r0, #128\]!
179 ** ...
180 */
181 float16_t* pre_loadfp16_0 (float16_t *a, float16x8_t *v)
182 {
183 a += 64;
184 *v = vld1q_f16 (a);
185 return a;
186 }
187
188 /*
189 **pre_load16_0:
190 ** ...
191 ** vldrh.16 q[0-7], \[r0, #-254\]!
192 ** ...
193 */
194 int16_t* pre_load16_0 (int16_t * a, int16x8_t *v)
195 {
196 a -= 127;
197 *v = vldrhq_s16 (a);
198 return a;
199 }
200
201 /*
202 **pre_load16_1:
203 ** ...
204 ** vldrh.s32 q[0-7], \[r0, #52\]!
205 ** ...
206 */
207 int16_t* pre_load16_1 (int16_t * a, int32x4_t *v)
208 {
209 a += 26;
210 *v = vldrhq_s32 (a);
211 return a;
212 }
213
214 /*
215 **pre_loadfp32_0:
216 ** ...
217 ** vldrw.32 q[0-7], \[r0, #-72\]!
218 ** ...
219 */
220 float32_t* pre_loadfp32_0 (float32_t *a, float32x4_t *v)
221 {
222 a -= 18;
223 *v = vld1q_f32 (a);
224 return a;
225 }
226
227
228 /*
229 **pre_load32_0:
230 ** ...
231 ** vldrw.32 q[0-7], \[r0, #-4\]!
232 ** ...
233 */
234 uint32_t* pre_load32_0 (uint32_t * a, uint32x4_t *v)
235 {
236 a -= 1;
237 *v = vld1q_u32 (a);
238 return a;
239 }
240
241
242 /*
243 **post_load8_0:
244 ** ...
245 ** vldrb.8 q[0-7], \[r0\], #26
246 ** ...
247 */
248 uint8_t* post_load8_0 (uint8_t * a, uint8x16_t *v)
249 {
250 *v = vld1q_u8 (a);
251 a += 26;
252 return a;
253 }
254
255 /*
256 **post_load8_1:
257 ** ...
258 ** vldrb.s16 q[0-7], \[r0\], #-1
259 ** ...
260 */
261 int8_t* post_load8_1 (int8_t * a, int16x8_t *v)
262 {
263 *v = vldrbq_s16 (a);
264 a--;
265 return a;
266 }
267
268 /*
269 **post_load8_2:
270 ** ...
271 ** vldrb.8 q[0-7], \[r0\], #26
272 ** ...
273 */
274 uint8_t* post_load8_2 (uint8_t * a, uint8x16_t *v)
275 {
276 *v = vld1q_u8 (a);
277 a += 26;
278 return a;
279 }
280
281 /*
282 **post_load8_3:
283 ** ...
284 ** vldrb.s16 q[0-7], \[r0\], #-1
285 ** ...
286 */
287 int8_t* post_load8_3 (int8_t * a, int16x8_t *v)
288 {
289 *v = vldrbq_s16 (a);
290 a--;
291 return a;
292 }
293
294 /*
295 **post_loadfp16_0:
296 ** ...
297 ** vldrh.16 q[0-7], \[r0\], #-24
298 ** ...
299 */
300 float16_t* post_loadfp16_0 (float16_t *a, float16x8_t *v)
301 {
302 *v = vld1q_f16 (a);
303 a -= 12;
304 return a;
305 }
306
307 /*
308 **post_load16_0:
309 ** ...
310 ** vldrh.16 q[0-7], \[r0\], #-126
311 ** ...
312 */
313 uint16_t* post_load16_0 (uint16_t * a, uint16x8_t *v)
314 {
315 *v = vldrhq_u16 (a);
316 a -= 63;
317 return a;
318 }
319
320 /*
321 **post_load16_1:
322 ** ...
323 ** vldrh.u32 q[0-7], \[r0\], #16
324 ** ...
325 */
326 uint16_t* post_load16_1 (uint16_t * a, uint32x4_t *v)
327 {
328 *v = vldrhq_u32 (a);
329 a += 8;
330 return a;
331 }
332
333 /*
334 **post_loadfp32_0:
335 ** ...
336 ** vldrw.32 q[0-7], \[r0\], #4
337 ** ...
338 */
339 float32_t* post_loadfp32_0 (float32_t *a, float32x4_t *v)
340 {
341 *v = vld1q_f32 (a);
342 a++;
343 return a;
344 }
345
346 /*
347 **post_load32_0:
348 ** ...
349 ** vldrw.32 q[0-7], \[r0\], #-16
350 ** ...
351 */
352 int32_t* post_load32_0 (int32_t * a, int32x4_t *v)
353 {
354 *v = vld1q_s32 (a);
355 a -= 4;
356 return a;
357 }