1 /* This file defines helper operations shared by all the tests. */
2
3 #ifndef _ARM_NEON_REF_H_
4 #define _ARM_NEON_REF_H_
5
6 #include <stdio.h>
7 #include <inttypes.h>
8
9 /* helper type, to help write floating point results in integer form. */
10 typedef uint16_t hfloat16_t;
11 typedef uint32_t hfloat32_t;
12 typedef uint64_t hfloat64_t;
13
14 typedef uint16_t hbfloat16_t;
15
16 extern void abort(void);
17 extern void *memset(void *, int, size_t);
18 extern void *memcpy(void *, const void *, size_t);
19 extern size_t strlen(const char *);
20
21 /* Helper macro to select FP16 tests. */
22 #if (defined (__ARM_FP16_FORMAT_IEEE) \
23 || defined (__ARM_FP16_FORMAT_ALTERNATIVE))
24 #define FP16_SUPPORTED (1)
25 #else
26 #undef FP16_SUPPORTED
27 #endif
28
29 /* Various string construction helpers. */
30
31 /*
32 The most useful at user-level are VECT_VAR and VECT_VAR_DECL, which
33 construct variable names or declarations, such as:
34 VECT_VAR(expected, int, 16, 4) -> expected_int16x4
35 VECT_VAR_DECL(expected, int, 16, 4) -> int16x4_t expected_int16x4
36 */
37 /* Some instructions don't exist on ARM.
38 Use this macro to guard against them. */
39 #ifdef __aarch64__
40 #define AARCH64_ONLY(X) X
41 #else
42 #define AARCH64_ONLY(X)
43 #endif
44
45 #define xSTR(X) #X
46 #define STR(X) xSTR(X)
47
48 #define xNAME1(V,T) V ## _ ## T
49 #define xNAME(V,T) xNAME1(V,T)
50
51 /* VAR(foo,int,16) -> foo_int16 */
52 #define VAR(V,T,W) xNAME(V,T##W)
53 /* VAR_DECL(foo,int,16) -> int16_t foo_int16 */
54 #define VAR_DECL(V, T, W) T##W##_t VAR(V,T,W)
55
56 /* VECT_NAME(int,16,4) -> int16x4 */
57 #define VECT_NAME(T, W, N) T##W##x##N
58 /* VECT_ARRAY_NAME(int,16,4,2) -> int16x4x2 */
59 #define VECT_ARRAY_NAME(T, W, N, L) T##W##x##N##x##L
60 /* VECT_TYPE(int,16,4) -> int16x4_t */
61 #define VECT_TYPE(T, W, N) xNAME(VECT_NAME(T,W,N),t)
62 /* VECT_ARRAY_TYPE(int,16,4,2) -> int16x4x2_t */
63 #define VECT_ARRAY_TYPE(T, W, N, L) xNAME(VECT_ARRAY_NAME(T,W,N,L),t)
64
65 /* VECT_VAR(foo,int,16,4) -> foo_int16x4 */
66 #define VECT_VAR(V,T,W,N) xNAME(V,VECT_NAME(T,W,N))
67 /* VECT_VAR_DECL(foo,int,16,4) -> int16_t foo_int16x4 */
68 #define VECT_VAR_DECL(V, T, W, N) T##W##_t VECT_VAR(V,T,W,N)
69
70 /* Array declarations. */
71 /* ARRAY(foo,int,16,4) -> int16_t foo_int16x4[4] */
72 #define ARRAY(V, T, W, N) VECT_VAR_DECL(V,T,W,N)[N]
73
74 /* Arrays of vectors. */
75 /* VECT_ARRAY_VAR(foo,int,16,4,2) -> foo_int16x4x2 */
76 #define VECT_ARRAY_VAR(V,T,W,N,L) xNAME(V,VECT_ARRAY_NAME(T,W,N,L))
77 /* VECT_ARRAY(foo,int,16,4,2) -> int16_t foo_int16x4x2[4*2] */
78 #define VECT_ARRAY(V, T, W, N, L) T##W##_t VECT_ARRAY_VAR(V,T,W,N,L)[N*L]
79
80 /* Check results vs expected values. Operates on one vector. */
81 #define CHECK(MSG,T,W,N,FMT,EXPECTED,COMMENT) \
82 { \
83 int i; \
84 for(i=0; i<N ; i++) \
85 { \
86 if (VECT_VAR(result, T, W, N)[i] != \
87 VECT_VAR(EXPECTED, T, W, N)[i]) { \
88 fprintf(stderr, \
89 "ERROR in %s (%s line %d in buffer '%s') at type %s " \
90 "index %d: got 0x%" FMT " != 0x%" FMT " %s\n", \
91 MSG, __FILE__, __LINE__, \
92 STR(EXPECTED), \
93 STR(VECT_NAME(T, W, N)), \
94 i, \
95 VECT_VAR(result, T, W, N)[i], \
96 VECT_VAR(EXPECTED, T, W, N)[i], \
97 strlen(COMMENT) > 0 ? COMMENT : ""); \
98 abort(); \
99 } \
100 } \
101 fprintf(stderr, "CHECKED %s %s\n", STR(VECT_TYPE(T, W, N)), MSG); \
102 }
103
104 /* Floating-point variant. */
105 #define CHECK_FP(MSG,T,W,N,FMT,EXPECTED,COMMENT) \
106 { \
107 int i; \
108 for(i=0; i<N ; i++) \
109 { \
110 union fp_operand { \
111 uint##W##_t i; \
112 T##W##_t f; \
113 } tmp_res, tmp_exp; \
114 tmp_res.f = VECT_VAR(result, T, W, N)[i]; \
115 tmp_exp.i = VECT_VAR(EXPECTED, h##T, W, N)[i]; \
116 if (tmp_res.i != tmp_exp.i) { \
117 fprintf(stderr, \
118 "ERROR in %s (%s line %d in buffer '%s') at type %s " \
119 "index %d: got 0x%" FMT " != 0x%" FMT " %s\n", \
120 MSG, __FILE__, __LINE__, \
121 STR(EXPECTED), \
122 STR(VECT_NAME(T, W, N)), \
123 i, \
124 tmp_res.i, \
125 tmp_exp.i, \
126 strlen(COMMENT) > 0 ? COMMENT : ""); \
127 abort(); \
128 } \
129 } \
130 fprintf(stderr, "CHECKED %s %s\n", STR(VECT_TYPE(T, W, N)), MSG); \
131 }
132
133 /* poly variant. */
134 #define CHECK_POLY(MSG,T,W,N,FMT,EXPECTED,COMMENT) \
135 { \
136 int i; \
137 for(i=0; i<N ; i++) \
138 { \
139 union poly_operand { \
140 uint##W##_t i; \
141 poly##W##_t p; \
142 } tmp_res, tmp_exp; \
143 tmp_res.p = VECT_VAR(result, T, W, N)[i]; \
144 tmp_exp.i = VECT_VAR(EXPECTED, T, W, N)[i]; \
145 if (tmp_res.i != tmp_exp.i) { \
146 fprintf(stderr, \
147 "ERROR in %s (%s line %d in buffer '%s') at type %s " \
148 "index %d: got 0x%" FMT " != 0x%" FMT " %s\n", \
149 MSG, __FILE__, __LINE__, \
150 STR(EXPECTED), \
151 STR(VECT_NAME(T, W, N)), \
152 i, \
153 tmp_res.i, \
154 tmp_exp.i, \
155 strlen(COMMENT) > 0 ? COMMENT : ""); \
156 abort(); \
157 } \
158 } \
159 fprintf(stderr, "CHECKED %s %s\n", STR(VECT_TYPE(T, W, N)), MSG); \
160 }
161
162 /* Clean buffer with a non-zero pattern to help diagnose buffer
163 overflows. */
164 #define CLEAN_PATTERN_8 0x33
165
166 #define CLEAN(VAR,T,W,N) \
167 memset(VECT_VAR(VAR, T, W, N), \
168 CLEAN_PATTERN_8, \
169 sizeof(VECT_VAR(VAR, T, W, N)));
170
171 /* Define output buffers, one of each size. */
172 static ARRAY(result, int, 8, 8);
173 static ARRAY(result, int, 16, 4);
174 static ARRAY(result, int, 32, 2);
175 static ARRAY(result, int, 64, 1);
176 static ARRAY(result, uint, 8, 8);
177 static ARRAY(result, uint, 16, 4);
178 static ARRAY(result, uint, 32, 2);
179 static ARRAY(result, uint, 64, 1);
180 static ARRAY(result, poly, 8, 8);
181 static ARRAY(result, poly, 16, 4);
182 #if defined (__ARM_FEATURE_CRYPTO)
183 static ARRAY(result, poly, 64, 1);
184 #endif
185 #if defined (__ARM_FP16_FORMAT_IEEE) || defined (__ARM_FP16_FORMAT_ALTERNATIVE)
186 static ARRAY(result, float, 16, 4);
187 #endif
188 static ARRAY(result, float, 32, 2);
189 #ifdef __aarch64__
190 static ARRAY(result, float, 64, 1);
191 #endif
192 static ARRAY(result, int, 8, 16);
193 static ARRAY(result, int, 16, 8);
194 static ARRAY(result, int, 32, 4);
195 static ARRAY(result, int, 64, 2);
196 static ARRAY(result, uint, 8, 16);
197 static ARRAY(result, uint, 16, 8);
198 static ARRAY(result, uint, 32, 4);
199 static ARRAY(result, uint, 64, 2);
200 static ARRAY(result, poly, 8, 16);
201 static ARRAY(result, poly, 16, 8);
202 #if defined (__ARM_FEATURE_CRYPTO)
203 static ARRAY(result, poly, 64, 2);
204 #endif
205 #if defined (__ARM_FP16_FORMAT_IEEE) || defined (__ARM_FP16_FORMAT_ALTERNATIVE)
206 static ARRAY(result, float, 16, 8);
207 #endif
208 static ARRAY(result, float, 32, 4);
209 #ifdef __aarch64__
210 static ARRAY(result, float, 64, 2);
211 #endif
212
213 /* Declare expected results, one of each size. They are defined and
214 initialized in each test file. */
215 extern ARRAY(expected, int, 8, 8);
216 extern ARRAY(expected, int, 16, 4);
217 extern ARRAY(expected, int, 32, 2);
218 extern ARRAY(expected, int, 64, 1);
219 extern ARRAY(expected, uint, 8, 8);
220 extern ARRAY(expected, uint, 16, 4);
221 extern ARRAY(expected, uint, 32, 2);
222 extern ARRAY(expected, uint, 64, 1);
223 extern ARRAY(expected, poly, 8, 8);
224 extern ARRAY(expected, poly, 16, 4);
225 extern ARRAY(expected, hfloat, 16, 4);
226 extern ARRAY(expected, hfloat, 32, 2);
227 extern ARRAY(expected, hfloat, 64, 1);
228 extern ARRAY(expected, int, 8, 16);
229 extern ARRAY(expected, int, 16, 8);
230 extern ARRAY(expected, int, 32, 4);
231 extern ARRAY(expected, int, 64, 2);
232 extern ARRAY(expected, uint, 8, 16);
233 extern ARRAY(expected, uint, 16, 8);
234 extern ARRAY(expected, uint, 32, 4);
235 extern ARRAY(expected, uint, 64, 2);
236 extern ARRAY(expected, poly, 8, 16);
237 extern ARRAY(expected, poly, 16, 8);
238 extern ARRAY(expected, hfloat, 16, 8);
239 extern ARRAY(expected, hfloat, 32, 4);
240 extern ARRAY(expected, hfloat, 64, 2);
241
242 #define CHECK_RESULTS_NAMED_NO_FP16(test_name,EXPECTED,comment) \
243 { \
244 CHECK(test_name, int, 8, 8, PRIx8, EXPECTED, comment); \
245 CHECK(test_name, int, 16, 4, PRIx16, EXPECTED, comment); \
246 CHECK(test_name, int, 32, 2, PRIx32, EXPECTED, comment); \
247 CHECK(test_name, int, 64, 1, PRIx64, EXPECTED, comment); \
248 CHECK(test_name, uint, 8, 8, PRIx8, EXPECTED, comment); \
249 CHECK(test_name, uint, 16, 4, PRIx16, EXPECTED, comment); \
250 CHECK(test_name, uint, 32, 2, PRIx32, EXPECTED, comment); \
251 CHECK(test_name, uint, 64, 1, PRIx64, EXPECTED, comment); \
252 CHECK_POLY(test_name, poly, 8, 8, PRIx8, EXPECTED, comment); \
253 CHECK_POLY(test_name, poly, 16, 4, PRIx16, EXPECTED, comment); \
254 CHECK_FP(test_name, float, 32, 2, PRIx32, EXPECTED, comment); \
255 \
256 CHECK(test_name, int, 8, 16, PRIx8, EXPECTED, comment); \
257 CHECK(test_name, int, 16, 8, PRIx16, EXPECTED, comment); \
258 CHECK(test_name, int, 32, 4, PRIx32, EXPECTED, comment); \
259 CHECK(test_name, int, 64, 2, PRIx64, EXPECTED, comment); \
260 CHECK(test_name, uint, 8, 16, PRIx8, EXPECTED, comment); \
261 CHECK(test_name, uint, 16, 8, PRIx16, EXPECTED, comment); \
262 CHECK(test_name, uint, 32, 4, PRIx32, EXPECTED, comment); \
263 CHECK(test_name, uint, 64, 2, PRIx64, EXPECTED, comment); \
264 CHECK_POLY(test_name, poly, 8, 16, PRIx8, EXPECTED, comment); \
265 CHECK_POLY(test_name, poly, 16, 8, PRIx16, EXPECTED, comment); \
266 CHECK_FP(test_name, float, 32, 4, PRIx32, EXPECTED, comment); \
267 } \
268
269 /* Check results against EXPECTED. Operates on all possible vector types. */
270 #if defined (__ARM_FP16_FORMAT_IEEE) || defined (__ARM_FP16_FORMAT_ALTERNATIVE)
271 #define CHECK_RESULTS_NAMED(test_name,EXPECTED,comment) \
272 { \
273 CHECK_RESULTS_NAMED_NO_FP16(test_name, EXPECTED, comment) \
274 CHECK_FP(test_name, float, 16, 4, PRIx16, EXPECTED, comment); \
275 CHECK_FP(test_name, float, 16, 8, PRIx16, EXPECTED, comment); \
276 }
277 #else
278 #define CHECK_RESULTS_NAMED(test_name,EXPECTED,comment) \
279 CHECK_RESULTS_NAMED_NO_FP16(test_name, EXPECTED, comment)
280 #endif
281
282 #define CHECK_RESULTS_NO_FP16(test_name,comment) \
283 CHECK_RESULTS_NAMED_NO_FP16(test_name, expected, comment)
284
285 #define CHECK_RESULTS(test_name,comment) \
286 CHECK_RESULTS_NAMED(test_name, expected, comment)
287
288
289 #if __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__
290
291 typedef union {
292 struct {
293 int _xxx:24;
294 unsigned int FZ:1;
295 unsigned int DN:1;
296 unsigned int AHP:1;
297 unsigned int QC:1;
298 int V:1;
299 int C:1;
300 int Z:1;
301 int N:1;
302 } b;
303 unsigned int word;
304 } _ARM_FPSCR;
305
306 #else /* __ORDER_BIG_ENDIAN__ */
307
308 typedef union {
309 struct {
310 int N:1;
311 int Z:1;
312 int C:1;
313 int V:1;
314 unsigned int QC:1;
315 unsigned int AHP:1;
316 unsigned int DN:1;
317 unsigned int FZ:1;
318 int _dnm:24;
319 } b;
320 unsigned int word;
321 } _ARM_FPSCR;
322
323 #endif /* __ORDER_BIG_ENDIAN__ */
324
325 #define Neon_Cumulative_Sat __read_neon_cumulative_sat()
326 /* We need a fake dependency to ensure correct ordering of asm
327 statements to preset the QC flag value, and Neon operators writing
328 to QC. */
329 #define Set_Neon_Cumulative_Sat(x, depend) \
330 __set_neon_cumulative_sat((x), (depend))
331
332 #if defined(__aarch64__)
333 static volatile int __read_neon_cumulative_sat (void) {
334 _ARM_FPSCR _afpscr_for_qc;
335 asm volatile ("mrs %0,fpsr" : "=r" (_afpscr_for_qc));
336 return _afpscr_for_qc.b.QC;
337 }
338 #define __set_neon_cumulative_sat(x, depend) { \
339 _ARM_FPSCR _afpscr_for_qc; \
340 asm volatile ("mrs %0,fpsr" : "=r" (_afpscr_for_qc)); \
341 _afpscr_for_qc.b.QC = x; \
342 asm volatile ("msr fpsr,%1" : "=X" (depend) : "r" (_afpscr_for_qc)); \
343 }
344 #else
345 static volatile int __read_neon_cumulative_sat (void) {
346 _ARM_FPSCR _afpscr_for_qc;
347 asm volatile ("vmrs %0,fpscr" : "=r" (_afpscr_for_qc));
348 return _afpscr_for_qc.b.QC;
349 }
350
351 #define __set_neon_cumulative_sat(x, depend) { \
352 _ARM_FPSCR _afpscr_for_qc; \
353 asm volatile ("vmrs %0,fpscr" : "=r" (_afpscr_for_qc)); \
354 _afpscr_for_qc.b.QC = x; \
355 asm volatile ("vmsr fpscr,%1" : "=X" (depend) : "r" (_afpscr_for_qc)); \
356 }
357 #endif
358
359 /* Clean output buffers before execution. */
360 static void clean_results (void)
361 {
362 CLEAN(result, int, 8, 8);
363 CLEAN(result, int, 16, 4);
364 CLEAN(result, int, 32, 2);
365 CLEAN(result, int, 64, 1);
366 CLEAN(result, uint, 8, 8);
367 CLEAN(result, uint, 16, 4);
368 CLEAN(result, uint, 32, 2);
369 CLEAN(result, uint, 64, 1);
370 CLEAN(result, poly, 8, 8);
371 CLEAN(result, poly, 16, 4);
372 #if defined (__ARM_FEATURE_CRYPTO)
373 CLEAN(result, poly, 64, 1);
374 #endif
375 #if defined (__ARM_FP16_FORMAT_IEEE) || defined (__ARM_FP16_FORMAT_ALTERNATIVE)
376 CLEAN(result, float, 16, 4);
377 #endif
378 CLEAN(result, float, 32, 2);
379
380 CLEAN(result, int, 8, 16);
381 CLEAN(result, int, 16, 8);
382 CLEAN(result, int, 32, 4);
383 CLEAN(result, int, 64, 2);
384 CLEAN(result, uint, 8, 16);
385 CLEAN(result, uint, 16, 8);
386 CLEAN(result, uint, 32, 4);
387 CLEAN(result, uint, 64, 2);
388 CLEAN(result, poly, 8, 16);
389 CLEAN(result, poly, 16, 8);
390 #if defined (__ARM_FEATURE_CRYPTO)
391 CLEAN(result, poly, 64, 2);
392 #endif
393 #if defined (__ARM_FP16_FORMAT_IEEE) || defined (__ARM_FP16_FORMAT_ALTERNATIVE)
394 CLEAN(result, float, 16, 8);
395 #endif
396 CLEAN(result, float, 32, 4);
397
398 AARCH64_ONLY(CLEAN(result, float, 64, 2));
399
400 #if defined(__aarch64__)
401 /* On AArch64, make sure to return DefaultNaN to have the same
402 results as on AArch32. */
403 _ARM_FPSCR _afpscr;
404 asm volatile ("mrs %0,fpcr" : "=r" (_afpscr));
405 _afpscr.b.DN = 1;
406
407 /* On AArch64, make sure to flush to zero by default, as on
408 AArch32. */
409 _afpscr.b.FZ = 1;
410
411 asm volatile ("msr fpcr,%0" : : "r" (_afpscr));
412 #endif
413 }
414
415
416 /* Helpers to declare variables of various types. */
417 #define DECL_VARIABLE(VAR, T1, W, N) \
418 VECT_TYPE(T1, W, N) VECT_VAR(VAR, T1, W, N)
419
420 #if defined (__ARM_FEATURE_CRYPTO)
421 #define DECL_VARIABLE_CRYPTO(VAR, T1, W, N) \
422 DECL_VARIABLE(VAR, T1, W, N)
423 #else
424 #define DECL_VARIABLE_CRYPTO(VAR, T1, W, N)
425 #endif
426
427 /* Declare only 64 bits signed variants. */
428 #define DECL_VARIABLE_64BITS_SIGNED_VARIANTS(VAR) \
429 DECL_VARIABLE(VAR, int, 8, 8); \
430 DECL_VARIABLE(VAR, int, 16, 4); \
431 DECL_VARIABLE(VAR, int, 32, 2); \
432 DECL_VARIABLE(VAR, int, 64, 1)
433
434 /* Declare only 64 bits unsigned variants. */
435 #define DECL_VARIABLE_64BITS_UNSIGNED_VARIANTS(VAR) \
436 DECL_VARIABLE(VAR, uint, 8, 8); \
437 DECL_VARIABLE(VAR, uint, 16, 4); \
438 DECL_VARIABLE(VAR, uint, 32, 2); \
439 DECL_VARIABLE(VAR, uint, 64, 1)
440
441 /* Declare only 128 bits signed variants. */
442 #define DECL_VARIABLE_128BITS_SIGNED_VARIANTS(VAR) \
443 DECL_VARIABLE(VAR, int, 8, 16); \
444 DECL_VARIABLE(VAR, int, 16, 8); \
445 DECL_VARIABLE(VAR, int, 32, 4); \
446 DECL_VARIABLE(VAR, int, 64, 2)
447
448 /* Declare only 128 bits unsigned variants. */
449 #define DECL_VARIABLE_128BITS_UNSIGNED_VARIANTS(VAR) \
450 DECL_VARIABLE(VAR, uint, 8, 16); \
451 DECL_VARIABLE(VAR, uint, 16, 8); \
452 DECL_VARIABLE(VAR, uint, 32, 4); \
453 DECL_VARIABLE(VAR, uint, 64, 2)
454
455 /* Declare all 64 bits variants. */
456 #if defined (__ARM_FP16_FORMAT_IEEE) || defined (__ARM_FP16_FORMAT_ALTERNATIVE)
457 #define DECL_VARIABLE_64BITS_VARIANTS(VAR) \
458 DECL_VARIABLE_64BITS_SIGNED_VARIANTS(VAR); \
459 DECL_VARIABLE_64BITS_UNSIGNED_VARIANTS(VAR); \
460 DECL_VARIABLE(VAR, poly, 8, 8); \
461 DECL_VARIABLE(VAR, poly, 16, 4); \
462 DECL_VARIABLE_CRYPTO(VAR, poly, 64, 1); \
463 DECL_VARIABLE(VAR, float, 16, 4); \
464 DECL_VARIABLE(VAR, float, 32, 2)
465 #else
466 #define DECL_VARIABLE_64BITS_VARIANTS(VAR) \
467 DECL_VARIABLE_64BITS_SIGNED_VARIANTS(VAR); \
468 DECL_VARIABLE_64BITS_UNSIGNED_VARIANTS(VAR); \
469 DECL_VARIABLE(VAR, poly, 8, 8); \
470 DECL_VARIABLE(VAR, poly, 16, 4); \
471 DECL_VARIABLE_CRYPTO(VAR, poly, 64, 1); \
472 DECL_VARIABLE(VAR, float, 32, 2)
473 #endif
474
475 /* Declare all 128 bits variants. */
476 #if defined (__ARM_FP16_FORMAT_IEEE) || defined (__ARM_FP16_FORMAT_ALTERNATIVE)
477 #define DECL_VARIABLE_128BITS_VARIANTS(VAR) \
478 DECL_VARIABLE_128BITS_SIGNED_VARIANTS(VAR); \
479 DECL_VARIABLE_128BITS_UNSIGNED_VARIANTS(VAR); \
480 DECL_VARIABLE(VAR, poly, 8, 16); \
481 DECL_VARIABLE(VAR, poly, 16, 8); \
482 DECL_VARIABLE_CRYPTO(VAR, poly, 64, 2); \
483 DECL_VARIABLE(VAR, float, 16, 8); \
484 DECL_VARIABLE(VAR, float, 32, 4); \
485 AARCH64_ONLY(DECL_VARIABLE(VAR, float, 64, 2))
486 #else
487 #define DECL_VARIABLE_128BITS_VARIANTS(VAR) \
488 DECL_VARIABLE_128BITS_SIGNED_VARIANTS(VAR); \
489 DECL_VARIABLE_128BITS_UNSIGNED_VARIANTS(VAR); \
490 DECL_VARIABLE(VAR, poly, 8, 16); \
491 DECL_VARIABLE(VAR, poly, 16, 8); \
492 DECL_VARIABLE_CRYPTO(VAR, poly, 64, 2); \
493 DECL_VARIABLE(VAR, float, 32, 4); \
494 AARCH64_ONLY(DECL_VARIABLE(VAR, float, 64, 2))
495 #endif
496 /* Declare all variants. */
497 #define DECL_VARIABLE_ALL_VARIANTS(VAR) \
498 DECL_VARIABLE_64BITS_VARIANTS(VAR); \
499 DECL_VARIABLE_128BITS_VARIANTS(VAR)
500
501 /* Declare all signed variants. */
502 #define DECL_VARIABLE_SIGNED_VARIANTS(VAR) \
503 DECL_VARIABLE_64BITS_SIGNED_VARIANTS(VAR); \
504 DECL_VARIABLE_128BITS_SIGNED_VARIANTS(VAR)
505
506 /* Declare all unsigned variants. */
507 #define DECL_VARIABLE_UNSIGNED_VARIANTS(VAR) \
508 DECL_VARIABLE_64BITS_UNSIGNED_VARIANTS(VAR); \
509 DECL_VARIABLE_128BITS_UNSIGNED_VARIANTS(VAR)
510
511 /* Helpers to initialize vectors. */
512 #define VDUP(VAR, Q, T1, T2, W, N, V) \
513 VECT_VAR(VAR, T1, W, N) = vdup##Q##_n_##T2##W(V)
514
515 #define VSET_LANE(VAR, Q, T1, T2, W, N, L, V) \
516 VECT_VAR(VAR, T1, W, N) = vset##Q##_lane_##T2##W(V, \
517 VECT_VAR(VAR, T1, W, N), \
518 L)
519
520 /* We need to load initial values first, so rely on VLD1. */
521 #define VLOAD(VAR, BUF, Q, T1, T2, W, N) \
522 VECT_VAR(VAR, T1, W, N) = vld1##Q##_##T2##W(VECT_VAR(BUF, T1, W, N))
523
524 /* Helpers to call macros with 1 constant and 5 variable
525 arguments. */
526 #if defined (__ARM_FEATURE_CRYPTO)
527 #define MACRO_CRYPTO(MACRO, VAR1, VAR2, T1, T2, T3, W, N) \
528 MACRO(VAR1, VAR2, T1, T2, T3, W, N)
529 #else
530 #define MACRO_CRYPTO(MACRO, VAR1, VAR2, T1, T2, T3, W, N)
531 #endif
532
533 #define TEST_MACRO_64BITS_SIGNED_VARIANTS_1_5(MACRO, VAR) \
534 MACRO(VAR, , int, s, 8, 8); \
535 MACRO(VAR, , int, s, 16, 4); \
536 MACRO(VAR, , int, s, 32, 2); \
537 MACRO(VAR, , int, s, 64, 1)
538
539 #define TEST_MACRO_64BITS_UNSIGNED_VARIANTS_1_5(MACRO, VAR) \
540 MACRO(VAR, , uint, u, 8, 8); \
541 MACRO(VAR, , uint, u, 16, 4); \
542 MACRO(VAR, , uint, u, 32, 2); \
543 MACRO(VAR, , uint, u, 64, 1)
544
545 #define TEST_MACRO_128BITS_SIGNED_VARIANTS_1_5(MACRO, VAR) \
546 MACRO(VAR, q, int, s, 8, 16); \
547 MACRO(VAR, q, int, s, 16, 8); \
548 MACRO(VAR, q, int, s, 32, 4); \
549 MACRO(VAR, q, int, s, 64, 2)
550
551 #define TEST_MACRO_128BITS_UNSIGNED_VARIANTS_1_5(MACRO,VAR) \
552 MACRO(VAR, q, uint, u, 8, 16); \
553 MACRO(VAR, q, uint, u, 16, 8); \
554 MACRO(VAR, q, uint, u, 32, 4); \
555 MACRO(VAR, q, uint, u, 64, 2)
556
557 #define TEST_MACRO_64BITS_VARIANTS_1_5(MACRO, VAR) \
558 TEST_MACRO_64BITS_SIGNED_VARIANTS_1_5(MACRO, VAR); \
559 TEST_MACRO_64BITS_UNSIGNED_VARIANTS_1_5(MACRO, VAR)
560
561 #define TEST_MACRO_128BITS_VARIANTS_1_5(MACRO, VAR) \
562 TEST_MACRO_128BITS_SIGNED_VARIANTS_1_5(MACRO, VAR); \
563 TEST_MACRO_128BITS_UNSIGNED_VARIANTS_1_5(MACRO, VAR)
564
565 #define TEST_MACRO_ALL_VARIANTS_1_5(MACRO, VAR) \
566 TEST_MACRO_64BITS_VARIANTS_1_5(MACRO, VAR); \
567 TEST_MACRO_128BITS_VARIANTS_1_5(MACRO, VAR)
568
569 #define TEST_MACRO_SIGNED_VARIANTS_1_5(MACRO, VAR) \
570 TEST_MACRO_64BITS_SIGNED_VARIANTS_1_5(MACRO, VAR); \
571 TEST_MACRO_128BITS_SIGNED_VARIANTS_1_5(MACRO, VAR)
572
573 /* Helpers to call macros with 2 constant and 5 variable
574 arguments. */
575 #define TEST_MACRO_64BITS_SIGNED_VARIANTS_2_5(MACRO, VAR1, VAR2) \
576 MACRO(VAR1, VAR2, , int, s, 8, 8); \
577 MACRO(VAR1, VAR2, , int, s, 16, 4); \
578 MACRO(VAR1, VAR2, , int, s, 32, 2); \
579 MACRO(VAR1, VAR2 , , int, s, 64, 1)
580
581 #define TEST_MACRO_64BITS_UNSIGNED_VARIANTS_2_5(MACRO, VAR1, VAR2) \
582 MACRO(VAR1, VAR2, , uint, u, 8, 8); \
583 MACRO(VAR1, VAR2, , uint, u, 16, 4); \
584 MACRO(VAR1, VAR2, , uint, u, 32, 2); \
585 MACRO(VAR1, VAR2, , uint, u, 64, 1)
586
587 #define TEST_MACRO_128BITS_SIGNED_VARIANTS_2_5(MACRO, VAR1, VAR2) \
588 MACRO(VAR1, VAR2, q, int, s, 8, 16); \
589 MACRO(VAR1, VAR2, q, int, s, 16, 8); \
590 MACRO(VAR1, VAR2, q, int, s, 32, 4); \
591 MACRO(VAR1, VAR2, q, int, s, 64, 2)
592
593 #define TEST_MACRO_128BITS_UNSIGNED_VARIANTS_2_5(MACRO, VAR1, VAR2) \
594 MACRO(VAR1, VAR2, q, uint, u, 8, 16); \
595 MACRO(VAR1, VAR2, q, uint, u, 16, 8); \
596 MACRO(VAR1, VAR2, q, uint, u, 32, 4); \
597 MACRO(VAR1, VAR2, q, uint, u, 64, 2)
598
599 #define TEST_MACRO_64BITS_VARIANTS_2_5(MACRO, VAR1, VAR2) \
600 TEST_MACRO_64BITS_SIGNED_VARIANTS_2_5(MACRO, VAR1, VAR2); \
601 TEST_MACRO_64BITS_UNSIGNED_VARIANTS_2_5(MACRO, VAR1, VAR2); \
602 MACRO(VAR1, VAR2, , poly, p, 8, 8); \
603 MACRO(VAR1, VAR2, , poly, p, 16, 4); \
604 MACRO_CRYPTO(MACRO, VAR1, VAR2, , poly, p, 64, 1)
605
606 #define TEST_MACRO_128BITS_VARIANTS_2_5(MACRO, VAR1, VAR2) \
607 TEST_MACRO_128BITS_SIGNED_VARIANTS_2_5(MACRO, VAR1, VAR2); \
608 TEST_MACRO_128BITS_UNSIGNED_VARIANTS_2_5(MACRO, VAR1, VAR2); \
609 MACRO(VAR1, VAR2, q, poly, p, 8, 16); \
610 MACRO(VAR1, VAR2, q, poly, p, 16, 8); \
611 MACRO_CRYPTO(MACRO, VAR1, VAR2, q, poly, p, 64, 2)
612
613 #define TEST_MACRO_ALL_VARIANTS_2_5(MACRO, VAR1, VAR2) \
614 TEST_MACRO_64BITS_VARIANTS_2_5(MACRO, VAR1, VAR2); \
615 TEST_MACRO_128BITS_VARIANTS_2_5(MACRO, VAR1, VAR2)
616
617 #define TEST_MACRO_SIGNED_VARIANTS_2_5(MACRO, VAR1, VAR2) \
618 TEST_MACRO_64BITS_SIGNED_VARIANTS_2_5(MACRO, VAR1, VAR2); \
619 TEST_MACRO_128BITS_SIGNED_VARIANTS_2_5(MACRO, VAR1, VAR2)
620
621 #endif /* _ARM_NEON_REF_H_ */