| 1 |
#ifndef HAVE_VEC_SAT_ARITH |
| 2 |
#define HAVE_VEC_SAT_ARITH |
| 3 |
|
| 4 |
#include <stdint-gcc.h> |
| 5 |
#include <stdbool.h> |
| 6 |
|
| 7 |
#define VALIDATE_RESULT(out, expect, N) \ |
| 8 |
do \ |
| 9 |
{ \ |
| 10 |
for (unsigned i = 0; i < N; i++) \ |
| 11 |
if (out[i] != expect[i]) __builtin_abort (); \ |
| 12 |
} \ |
| 13 |
while (false) |
| 14 |
|
| 15 |
/******************************************************************************/ |
| 16 |
/* Saturation Add (unsigned and signed) */ |
| 17 |
/******************************************************************************/ |
| 18 |
#define DEF_VEC_SAT_U_ADD_FMT_1(T) \ |
| 19 |
void __attribute__((noinline)) \ |
| 20 |
vec_sat_u_add_##T##_fmt_1 (T *out, T *op_1, T *op_2, unsigned limit) \ |
| 21 |
{ \ |
| 22 |
unsigned i; \ |
| 23 |
for (i = 0; i < limit; i++) \ |
| 24 |
{ \ |
| 25 |
T x = op_1[i]; \ |
| 26 |
T y = op_2[i]; \ |
| 27 |
out[i] = (x + y) | (-(T)((T)(x + y) < x)); \ |
| 28 |
} \ |
| 29 |
} |
| 30 |
|
| 31 |
#define DEF_VEC_SAT_U_ADD_FMT_2(T) \ |
| 32 |
void __attribute__((noinline)) \ |
| 33 |
vec_sat_u_add_##T##_fmt_2 (T *out, T *op_1, T *op_2, unsigned limit) \ |
| 34 |
{ \ |
| 35 |
unsigned i; \ |
| 36 |
for (i = 0; i < limit; i++) \ |
| 37 |
{ \ |
| 38 |
T x = op_1[i]; \ |
| 39 |
T y = op_2[i]; \ |
| 40 |
out[i] = (T)(x + y) >= x ? (x + y) : -1; \ |
| 41 |
} \ |
| 42 |
} |
| 43 |
|
| 44 |
#define DEF_VEC_SAT_U_ADD_FMT_3(T) \ |
| 45 |
void __attribute__((noinline)) \ |
| 46 |
vec_sat_u_add_##T##_fmt_3 (T *out, T *op_1, T *op_2, unsigned limit) \ |
| 47 |
{ \ |
| 48 |
unsigned i; \ |
| 49 |
for (i = 0; i < limit; i++) \ |
| 50 |
{ \ |
| 51 |
T x = op_1[i]; \ |
| 52 |
T y = op_2[i]; \ |
| 53 |
T ret; \ |
| 54 |
T overflow = __builtin_add_overflow (x, y, &ret); \ |
| 55 |
out[i] = (T)(-overflow) | ret; \ |
| 56 |
} \ |
| 57 |
} |
| 58 |
|
| 59 |
#define DEF_VEC_SAT_U_ADD_FMT_4(T) \ |
| 60 |
void __attribute__((noinline)) \ |
| 61 |
vec_sat_u_add_##T##_fmt_4 (T *out, T *op_1, T *op_2, unsigned limit) \ |
| 62 |
{ \ |
| 63 |
unsigned i; \ |
| 64 |
for (i = 0; i < limit; i++) \ |
| 65 |
{ \ |
| 66 |
T x = op_1[i]; \ |
| 67 |
T y = op_2[i]; \ |
| 68 |
T ret; \ |
| 69 |
out[i] = __builtin_add_overflow (x, y, &ret) ? -1 : ret; \ |
| 70 |
} \ |
| 71 |
} |
| 72 |
|
| 73 |
#define DEF_VEC_SAT_U_ADD_FMT_5(T) \ |
| 74 |
void __attribute__((noinline)) \ |
| 75 |
vec_sat_u_add_##T##_fmt_5 (T *out, T *op_1, T *op_2, unsigned limit) \ |
| 76 |
{ \ |
| 77 |
unsigned i; \ |
| 78 |
for (i = 0; i < limit; i++) \ |
| 79 |
{ \ |
| 80 |
T x = op_1[i]; \ |
| 81 |
T y = op_2[i]; \ |
| 82 |
T ret; \ |
| 83 |
out[i] = __builtin_add_overflow (x, y, &ret) == 0 ? ret : -1; \ |
| 84 |
} \ |
| 85 |
} |
| 86 |
|
| 87 |
#define DEF_VEC_SAT_U_ADD_FMT_6(T) \ |
| 88 |
void __attribute__((noinline)) \ |
| 89 |
vec_sat_u_add_##T##_fmt_6 (T *out, T *op_1, T *op_2, unsigned limit) \ |
| 90 |
{ \ |
| 91 |
unsigned i; \ |
| 92 |
for (i = 0; i < limit; i++) \ |
| 93 |
{ \ |
| 94 |
T x = op_1[i]; \ |
| 95 |
T y = op_2[i]; \ |
| 96 |
out[i] = x <= (T)(x + y) ? (x + y) : -1; \ |
| 97 |
} \ |
| 98 |
} |
| 99 |
|
| 100 |
#define DEF_VEC_SAT_U_ADD_FMT_7(T) \ |
| 101 |
void __attribute__((noinline)) \ |
| 102 |
vec_sat_u_add_##T##_fmt_7 (T *out, T *op_1, T *op_2, unsigned limit) \ |
| 103 |
{ \ |
| 104 |
unsigned i; \ |
| 105 |
for (i = 0; i < limit; i++) \ |
| 106 |
{ \ |
| 107 |
T x = op_1[i]; \ |
| 108 |
T y = op_2[i]; \ |
| 109 |
out[i] = (T)(x + y) < x ? -1 : (x + y); \ |
| 110 |
} \ |
| 111 |
} |
| 112 |
|
| 113 |
#define DEF_VEC_SAT_U_ADD_FMT_8(T) \ |
| 114 |
void __attribute__((noinline)) \ |
| 115 |
vec_sat_u_add_##T##_fmt_8 (T *out, T *op_1, T *op_2, unsigned limit) \ |
| 116 |
{ \ |
| 117 |
unsigned i; \ |
| 118 |
for (i = 0; i < limit; i++) \ |
| 119 |
{ \ |
| 120 |
T x = op_1[i]; \ |
| 121 |
T y = op_2[i]; \ |
| 122 |
out[i] = x > (T)(x + y) ? -1 : (x + y); \ |
| 123 |
} \ |
| 124 |
} |
| 125 |
|
| 126 |
#define RUN_VEC_SAT_U_ADD_FMT_1(T, out, op_1, op_2, N) \ |
| 127 |
vec_sat_u_add_##T##_fmt_1(out, op_1, op_2, N) |
| 128 |
|
| 129 |
#define RUN_VEC_SAT_U_ADD_FMT_2(T, out, op_1, op_2, N) \ |
| 130 |
vec_sat_u_add_##T##_fmt_2(out, op_1, op_2, N) |
| 131 |
|
| 132 |
#define RUN_VEC_SAT_U_ADD_FMT_3(T, out, op_1, op_2, N) \ |
| 133 |
vec_sat_u_add_##T##_fmt_3(out, op_1, op_2, N) |
| 134 |
|
| 135 |
#define RUN_VEC_SAT_U_ADD_FMT_4(T, out, op_1, op_2, N) \ |
| 136 |
vec_sat_u_add_##T##_fmt_4(out, op_1, op_2, N) |
| 137 |
|
| 138 |
#define RUN_VEC_SAT_U_ADD_FMT_5(T, out, op_1, op_2, N) \ |
| 139 |
vec_sat_u_add_##T##_fmt_5(out, op_1, op_2, N) |
| 140 |
|
| 141 |
#define RUN_VEC_SAT_U_ADD_FMT_6(T, out, op_1, op_2, N) \ |
| 142 |
vec_sat_u_add_##T##_fmt_6(out, op_1, op_2, N) |
| 143 |
|
| 144 |
#define RUN_VEC_SAT_U_ADD_FMT_7(T, out, op_1, op_2, N) \ |
| 145 |
vec_sat_u_add_##T##_fmt_7(out, op_1, op_2, N) |
| 146 |
|
| 147 |
#define RUN_VEC_SAT_U_ADD_FMT_8(T, out, op_1, op_2, N) \ |
| 148 |
vec_sat_u_add_##T##_fmt_8(out, op_1, op_2, N) |
| 149 |
|
| 150 |
#define DEF_VEC_SAT_U_ADD_IMM_FMT_1(T, IMM) \ |
| 151 |
T __attribute__((noinline)) \ |
| 152 |
vec_sat_u_add_imm##IMM##_##T##_fmt_1 (T *out, T *in, unsigned limit) \ |
| 153 |
{ \ |
| 154 |
unsigned i; \ |
| 155 |
for (i = 0; i < limit; i++) \ |
| 156 |
out[i] = (T)(in[i] + IMM) >= in[i] ? (in[i] + IMM) : -1; \ |
| 157 |
} |
| 158 |
#define DEF_VEC_SAT_U_ADD_IMM_FMT_1_WRAP(T, IMM) \ |
| 159 |
DEF_VEC_SAT_U_ADD_IMM_FMT_1(T, IMM) |
| 160 |
|
| 161 |
#define DEF_VEC_SAT_U_ADD_IMM_FMT_2(T, IMM) \ |
| 162 |
T __attribute__((noinline)) \ |
| 163 |
vec_sat_u_add_imm##IMM##_##T##_fmt_2 (T *out, T *in, unsigned limit) \ |
| 164 |
{ \ |
| 165 |
unsigned i; \ |
| 166 |
for (i = 0; i < limit; i++) \ |
| 167 |
out[i] = (T)(in[i] + IMM) < in[i] ? -1 : (in[i] + IMM); \ |
| 168 |
} |
| 169 |
#define DEF_VEC_SAT_U_ADD_IMM_FMT_2_WRAP(T, IMM) \ |
| 170 |
DEF_VEC_SAT_U_ADD_IMM_FMT_2(T, IMM) |
| 171 |
|
| 172 |
#define DEF_VEC_SAT_U_ADD_IMM_FMT_3(T, IMM) \ |
| 173 |
T __attribute__((noinline)) \ |
| 174 |
vec_sat_u_add_imm##IMM##_##T##_fmt_3 (T *out, T *in, unsigned limit) \ |
| 175 |
{ \ |
| 176 |
unsigned i; \ |
| 177 |
T ret; \ |
| 178 |
for (i = 0; i < limit; i++) \ |
| 179 |
{ \ |
| 180 |
out[i] = __builtin_add_overflow (in[i], IMM, &ret) ? -1 : ret; \ |
| 181 |
} \ |
| 182 |
} |
| 183 |
#define DEF_VEC_SAT_U_ADD_IMM_FMT_3_WRAP(T, IMM) \ |
| 184 |
DEF_VEC_SAT_U_ADD_IMM_FMT_3(T, IMM) |
| 185 |
|
| 186 |
#define DEF_VEC_SAT_U_ADD_IMM_FMT_4(T, IMM) \ |
| 187 |
T __attribute__((noinline)) \ |
| 188 |
vec_sat_u_add_imm##IMM##_##T##_fmt_4 (T *out, T *in, unsigned limit) \ |
| 189 |
{ \ |
| 190 |
unsigned i; \ |
| 191 |
T ret; \ |
| 192 |
for (i = 0; i < limit; i++) \ |
| 193 |
{ \ |
| 194 |
out[i] = __builtin_add_overflow (in[i], IMM, &ret) == 0 ? ret : -1; \ |
| 195 |
} \ |
| 196 |
} |
| 197 |
#define DEF_VEC_SAT_U_ADD_IMM_FMT_4_WRAP(T, IMM) \ |
| 198 |
DEF_VEC_SAT_U_ADD_IMM_FMT_4(T, IMM) |
| 199 |
|
| 200 |
#define RUN_VEC_SAT_U_ADD_IMM_FMT_1(T, out, op_1, expect, IMM, N) \ |
| 201 |
vec_sat_u_add_imm##IMM##_##T##_fmt_1(out, op_1, N); \ |
| 202 |
VALIDATE_RESULT (out, expect, N) |
| 203 |
#define RUN_VEC_SAT_U_ADD_IMM_FMT_1_WRAP(T, out, op_1, expect, IMM, N) \ |
| 204 |
RUN_VEC_SAT_U_ADD_IMM_FMT_1(T, out, op_1, expect, IMM, N) |
| 205 |
|
| 206 |
#define RUN_VEC_SAT_U_ADD_IMM_FMT_2(T, out, op_1, expect, IMM, N) \ |
| 207 |
vec_sat_u_add_imm##IMM##_##T##_fmt_2(out, op_1, N); \ |
| 208 |
VALIDATE_RESULT (out, expect, N) |
| 209 |
#define RUN_VEC_SAT_U_ADD_IMM_FMT_2_WRAP(T, out, op_1, expect, IMM, N) \ |
| 210 |
RUN_VEC_SAT_U_ADD_IMM_FMT_2(T, out, op_1, expect, IMM, N) |
| 211 |
|
| 212 |
#define RUN_VEC_SAT_U_ADD_IMM_FMT_3(T, out, op_1, expect, IMM, N) \ |
| 213 |
vec_sat_u_add_imm##IMM##_##T##_fmt_3(out, op_1, N); \ |
| 214 |
VALIDATE_RESULT (out, expect, N) |
| 215 |
#define RUN_VEC_SAT_U_ADD_IMM_FMT_3_WRAP(T, out, op_1, expect, IMM, N) \ |
| 216 |
RUN_VEC_SAT_U_ADD_IMM_FMT_3(T, out, op_1, expect, IMM, N) |
| 217 |
|
| 218 |
#define RUN_VEC_SAT_U_ADD_IMM_FMT_4(T, out, op_1, expect, IMM, N) \ |
| 219 |
vec_sat_u_add_imm##IMM##_##T##_fmt_4(out, op_1, N); \ |
| 220 |
VALIDATE_RESULT (out, expect, N) |
| 221 |
#define RUN_VEC_SAT_U_ADD_IMM_FMT_4_WRAP(T, out, op_1, expect, IMM, N) \ |
| 222 |
RUN_VEC_SAT_U_ADD_IMM_FMT_4(T, out, op_1, expect, IMM, N) |
| 223 |
|
| 224 |
#define DEF_VEC_SAT_S_ADD_FMT_1(T, UT, MIN, MAX) \ |
| 225 |
void __attribute__((noinline)) \ |
| 226 |
vec_sat_s_add_##T##_fmt_1 (T *out, T *op_1, T *op_2, unsigned limit) \ |
| 227 |
{ \ |
| 228 |
unsigned i; \ |
| 229 |
for (i = 0; i < limit; i++) \ |
| 230 |
{ \ |
| 231 |
T x = op_1[i]; \ |
| 232 |
T y = op_2[i]; \ |
| 233 |
T sum = (UT)x + (UT)y; \ |
| 234 |
out[i] = (x ^ y) < 0 \ |
| 235 |
? sum \ |
| 236 |
: (sum ^ x) >= 0 \ |
| 237 |
? sum \ |
| 238 |
: x < 0 ? MIN : MAX; \ |
| 239 |
} \ |
| 240 |
} |
| 241 |
#define DEF_VEC_SAT_S_ADD_FMT_1_WRAP(T, UT, MIN, MAX) \ |
| 242 |
DEF_VEC_SAT_S_ADD_FMT_1(T, UT, MIN, MAX) |
| 243 |
|
| 244 |
#define DEF_VEC_SAT_S_ADD_FMT_2(T, UT, MIN, MAX) \ |
| 245 |
void __attribute__((noinline)) \ |
| 246 |
vec_sat_s_add_##T##_fmt_2 (T *out, T *op_1, T *op_2, unsigned limit) \ |
| 247 |
{ \ |
| 248 |
unsigned i; \ |
| 249 |
for (i = 0; i < limit; i++) \ |
| 250 |
{ \ |
| 251 |
T x = op_1[i]; \ |
| 252 |
T y = op_2[i]; \ |
| 253 |
T sum = (UT)x + (UT)y; \ |
| 254 |
if ((x ^ y) < 0 || (sum ^ x) >= 0) \ |
| 255 |
out[i] = sum; \ |
| 256 |
else \ |
| 257 |
out[i] = x < 0 ? MIN : MAX; \ |
| 258 |
} \ |
| 259 |
} |
| 260 |
#define DEF_VEC_SAT_S_ADD_FMT_2_WRAP(T, UT, MIN, MAX) \ |
| 261 |
DEF_VEC_SAT_S_ADD_FMT_2(T, UT, MIN, MAX) |
| 262 |
|
| 263 |
#define DEF_VEC_SAT_S_ADD_FMT_3(T, UT, MIN, MAX) \ |
| 264 |
void __attribute__((noinline)) \ |
| 265 |
vec_sat_s_add_##T##_fmt_3 (T *out, T *op_1, T *op_2, unsigned limit) \ |
| 266 |
{ \ |
| 267 |
unsigned i; \ |
| 268 |
for (i = 0; i < limit; i++) \ |
| 269 |
{ \ |
| 270 |
T x = op_1[i]; \ |
| 271 |
T y = op_2[i]; \ |
| 272 |
T sum; \ |
| 273 |
bool overflow = __builtin_add_overflow (x, y, &sum); \ |
| 274 |
out[i] = overflow ? x < 0 ? MIN : MAX : sum; \ |
| 275 |
} \ |
| 276 |
} |
| 277 |
#define DEF_VEC_SAT_S_ADD_FMT_3_WRAP(T, UT, MIN, MAX) \ |
| 278 |
DEF_VEC_SAT_S_ADD_FMT_3(T, UT, MIN, MAX) |
| 279 |
|
| 280 |
#define DEF_VEC_SAT_S_ADD_FMT_4(T, UT, MIN, MAX) \ |
| 281 |
void __attribute__((noinline)) \ |
| 282 |
vec_sat_s_add_##T##_fmt_4 (T *out, T *op_1, T *op_2, unsigned limit) \ |
| 283 |
{ \ |
| 284 |
unsigned i; \ |
| 285 |
for (i = 0; i < limit; i++) \ |
| 286 |
{ \ |
| 287 |
T x = op_1[i]; \ |
| 288 |
T y = op_2[i]; \ |
| 289 |
T sum; \ |
| 290 |
bool overflow = __builtin_add_overflow (x, y, &sum); \ |
| 291 |
out[i] = !overflow ? sum : x < 0 ? MIN : MAX; \ |
| 292 |
} \ |
| 293 |
} |
| 294 |
#define DEF_VEC_SAT_S_ADD_FMT_4_WRAP(T, UT, MIN, MAX) \ |
| 295 |
DEF_VEC_SAT_S_ADD_FMT_4(T, UT, MIN, MAX) |
| 296 |
|
| 297 |
#define RUN_VEC_SAT_S_ADD_FMT_1(T, out, op_1, op_2, N) \ |
| 298 |
vec_sat_s_add_##T##_fmt_1(out, op_1, op_2, N) |
| 299 |
#define RUN_VEC_SAT_S_ADD_FMT_1_WRAP(T, out, op_1, op_2, N) \ |
| 300 |
RUN_VEC_SAT_S_ADD_FMT_1(T, out, op_1, op_2, N) |
| 301 |
|
| 302 |
#define RUN_VEC_SAT_S_ADD_FMT_2(T, out, op_1, op_2, N) \ |
| 303 |
vec_sat_s_add_##T##_fmt_2(out, op_1, op_2, N) |
| 304 |
#define RUN_VEC_SAT_S_ADD_FMT_2_WRAP(T, out, op_1, op_2, N) \ |
| 305 |
RUN_VEC_SAT_S_ADD_FMT_2(T, out, op_1, op_2, N) |
| 306 |
|
| 307 |
#define RUN_VEC_SAT_S_ADD_FMT_3(T, out, op_1, op_2, N) \ |
| 308 |
vec_sat_s_add_##T##_fmt_3(out, op_1, op_2, N) |
| 309 |
#define RUN_VEC_SAT_S_ADD_FMT_3_WRAP(T, out, op_1, op_2, N) \ |
| 310 |
RUN_VEC_SAT_S_ADD_FMT_3(T, out, op_1, op_2, N) |
| 311 |
|
| 312 |
#define RUN_VEC_SAT_S_ADD_FMT_4(T, out, op_1, op_2, N) \ |
| 313 |
vec_sat_s_add_##T##_fmt_4(out, op_1, op_2, N) |
| 314 |
#define RUN_VEC_SAT_S_ADD_FMT_4_WRAP(T, out, op_1, op_2, N) \ |
| 315 |
RUN_VEC_SAT_S_ADD_FMT_4(T, out, op_1, op_2, N) |
| 316 |
|
| 317 |
/******************************************************************************/ |
| 318 |
/* Saturation Sub (Unsigned and Signed) */ |
| 319 |
/******************************************************************************/ |
| 320 |
#define DEF_VEC_SAT_U_SUB_FMT_1(T) \ |
| 321 |
void __attribute__((noinline)) \ |
| 322 |
vec_sat_u_sub_##T##_fmt_1 (T *out, T *op_1, T *op_2, unsigned limit) \ |
| 323 |
{ \ |
| 324 |
unsigned i; \ |
| 325 |
for (i = 0; i < limit; i++) \ |
| 326 |
{ \ |
| 327 |
T x = op_1[i]; \ |
| 328 |
T y = op_2[i]; \ |
| 329 |
out[i] = (x - y) & (-(T)(x >= y)); \ |
| 330 |
} \ |
| 331 |
} |
| 332 |
|
| 333 |
#define DEF_VEC_SAT_U_SUB_FMT_2(T) \ |
| 334 |
void __attribute__((noinline)) \ |
| 335 |
vec_sat_u_sub_##T##_fmt_2 (T *out, T *op_1, T *op_2, unsigned limit) \ |
| 336 |
{ \ |
| 337 |
unsigned i; \ |
| 338 |
for (i = 0; i < limit; i++) \ |
| 339 |
{ \ |
| 340 |
T x = op_1[i]; \ |
| 341 |
T y = op_2[i]; \ |
| 342 |
out[i] = (x - y) & (-(T)(x > y)); \ |
| 343 |
} \ |
| 344 |
} |
| 345 |
|
| 346 |
#define DEF_VEC_SAT_U_SUB_FMT_3(T) \ |
| 347 |
void __attribute__((noinline)) \ |
| 348 |
vec_sat_u_sub_##T##_fmt_3 (T *out, T *op_1, T *op_2, unsigned limit) \ |
| 349 |
{ \ |
| 350 |
unsigned i; \ |
| 351 |
for (i = 0; i < limit; i++) \ |
| 352 |
{ \ |
| 353 |
T x = op_1[i]; \ |
| 354 |
T y = op_2[i]; \ |
| 355 |
out[i] = x > y ? x - y : 0; \ |
| 356 |
} \ |
| 357 |
} |
| 358 |
|
| 359 |
#define DEF_VEC_SAT_U_SUB_FMT_4(T) \ |
| 360 |
void __attribute__((noinline)) \ |
| 361 |
vec_sat_u_sub_##T##_fmt_4 (T *out, T *op_1, T *op_2, unsigned limit) \ |
| 362 |
{ \ |
| 363 |
unsigned i; \ |
| 364 |
for (i = 0; i < limit; i++) \ |
| 365 |
{ \ |
| 366 |
T x = op_1[i]; \ |
| 367 |
T y = op_2[i]; \ |
| 368 |
out[i] = x >= y ? x - y : 0; \ |
| 369 |
} \ |
| 370 |
} |
| 371 |
|
| 372 |
#define DEF_VEC_SAT_U_SUB_FMT_5(T) \ |
| 373 |
void __attribute__((noinline)) \ |
| 374 |
vec_sat_u_sub_##T##_fmt_5 (T *out, T *op_1, T *op_2, unsigned limit) \ |
| 375 |
{ \ |
| 376 |
unsigned i; \ |
| 377 |
for (i = 0; i < limit; i++) \ |
| 378 |
{ \ |
| 379 |
T x = op_1[i]; \ |
| 380 |
T y = op_2[i]; \ |
| 381 |
out[i] = x < y ? 0 : x - y; \ |
| 382 |
} \ |
| 383 |
} |
| 384 |
|
| 385 |
#define DEF_VEC_SAT_U_SUB_FMT_6(T) \ |
| 386 |
void __attribute__((noinline)) \ |
| 387 |
vec_sat_u_sub_##T##_fmt_6 (T *out, T *op_1, T *op_2, unsigned limit) \ |
| 388 |
{ \ |
| 389 |
unsigned i; \ |
| 390 |
for (i = 0; i < limit; i++) \ |
| 391 |
{ \ |
| 392 |
T x = op_1[i]; \ |
| 393 |
T y = op_2[i]; \ |
| 394 |
out[i] = x <= y ? 0 : x - y; \ |
| 395 |
} \ |
| 396 |
} |
| 397 |
|
| 398 |
#define DEF_VEC_SAT_U_SUB_FMT_7(T) \ |
| 399 |
void __attribute__((noinline)) \ |
| 400 |
vec_sat_u_sub_##T##_fmt_7 (T *out, T *op_1, T *op_2, unsigned limit) \ |
| 401 |
{ \ |
| 402 |
unsigned i; \ |
| 403 |
for (i = 0; i < limit; i++) \ |
| 404 |
{ \ |
| 405 |
T x = op_1[i]; \ |
| 406 |
T y = op_2[i]; \ |
| 407 |
T ret; \ |
| 408 |
T overflow = __builtin_sub_overflow (x, y, &ret); \ |
| 409 |
out[i] = ret & (T)(overflow - 1); \ |
| 410 |
} \ |
| 411 |
} |
| 412 |
|
| 413 |
#define DEF_VEC_SAT_U_SUB_FMT_8(T) \ |
| 414 |
void __attribute__((noinline)) \ |
| 415 |
vec_sat_u_sub_##T##_fmt_8 (T *out, T *op_1, T *op_2, unsigned limit) \ |
| 416 |
{ \ |
| 417 |
unsigned i; \ |
| 418 |
for (i = 0; i < limit; i++) \ |
| 419 |
{ \ |
| 420 |
T x = op_1[i]; \ |
| 421 |
T y = op_2[i]; \ |
| 422 |
T ret; \ |
| 423 |
T overflow = __builtin_sub_overflow (x, y, &ret); \ |
| 424 |
out[i] = ret & (T)-(!overflow); \ |
| 425 |
} \ |
| 426 |
} |
| 427 |
|
| 428 |
#define DEF_VEC_SAT_U_SUB_FMT_9(T) \ |
| 429 |
void __attribute__((noinline)) \ |
| 430 |
vec_sat_u_sub_##T##_fmt_9 (T *out, T *op_1, T *op_2, unsigned limit) \ |
| 431 |
{ \ |
| 432 |
unsigned i; \ |
| 433 |
for (i = 0; i < limit; i++) \ |
| 434 |
{ \ |
| 435 |
T x = op_1[i]; \ |
| 436 |
T y = op_2[i]; \ |
| 437 |
T ret; \ |
| 438 |
bool overflow = __builtin_sub_overflow (x, y, &ret); \ |
| 439 |
out[i] = overflow ? 0 : ret; \ |
| 440 |
} \ |
| 441 |
} |
| 442 |
|
| 443 |
#define DEF_VEC_SAT_U_SUB_FMT_10(T) \ |
| 444 |
void __attribute__((noinline)) \ |
| 445 |
vec_sat_u_sub_##T##_fmt_10 (T *out, T *op_1, T *op_2, unsigned limit) \ |
| 446 |
{ \ |
| 447 |
unsigned i; \ |
| 448 |
for (i = 0; i < limit; i++) \ |
| 449 |
{ \ |
| 450 |
T x = op_1[i]; \ |
| 451 |
T y = op_2[i]; \ |
| 452 |
T ret; \ |
| 453 |
bool overflow = __builtin_sub_overflow (x, y, &ret); \ |
| 454 |
out[i] = !overflow ? ret : 0; \ |
| 455 |
} \ |
| 456 |
} |
| 457 |
|
| 458 |
#define DEF_VEC_SAT_U_SUB_ZIP(T1, T2) \ |
| 459 |
void __attribute__((noinline)) \ |
| 460 |
vec_sat_u_sub_##T1##_##T2##_fmt_zip (T1 *x, T2 b, unsigned limit) \ |
| 461 |
{ \ |
| 462 |
T2 a; \ |
| 463 |
T1 *p = x; \ |
| 464 |
do { \ |
| 465 |
a = *--p; \ |
| 466 |
*p = (T1)(a >= b ? a - b : 0); \ |
| 467 |
} while (--limit); \ |
| 468 |
} |
| 469 |
#define DEF_VEC_SAT_U_SUB_ZIP_WRAP(T1, T2) DEF_VEC_SAT_U_SUB_ZIP(T1, T2) |
| 470 |
|
| 471 |
#define DEF_VEC_SAT_U_SUB_IMM_FMT_1(T, IMM) \ |
| 472 |
void __attribute__((noinline)) \ |
| 473 |
vec_sat_u_sub_imm##IMM##_##T##_fmt_1 (T *out, T *in, unsigned limit) \ |
| 474 |
{ \ |
| 475 |
unsigned i; \ |
| 476 |
for (i = 0; i < limit; i++) \ |
| 477 |
out[i] = (T)IMM >= in[i] ? (T)IMM - in[i] : 0; \ |
| 478 |
} |
| 479 |
|
| 480 |
#define DEF_VEC_SAT_U_SUB_IMM_FMT_1_WRAP(T, IMM) \ |
| 481 |
DEF_VEC_SAT_U_SUB_IMM_FMT_1(T, IMM) |
| 482 |
|
| 483 |
#define DEF_VEC_SAT_U_SUB_IMM_FMT_2(T, IMM) \ |
| 484 |
void __attribute__((noinline)) \ |
| 485 |
vec_sat_u_sub_imm##IMM##_##T##_fmt_2 (T *out, T *in, unsigned limit) \ |
| 486 |
{ \ |
| 487 |
unsigned i; \ |
| 488 |
for (i = 0; i < limit; i++) \ |
| 489 |
out[i] = in[i] >= (T)IMM ? in[i] - (T)IMM : 0; \ |
| 490 |
} |
| 491 |
|
| 492 |
#define DEF_VEC_SAT_U_SUB_IMM_FMT_2_WRAP(T, IMM) \ |
| 493 |
DEF_VEC_SAT_U_SUB_IMM_FMT_2(T, IMM) |
| 494 |
|
| 495 |
#define DEF_VEC_SAT_U_SUB_IMM_FMT_3(T, IMM) \ |
| 496 |
void __attribute__((noinline)) \ |
| 497 |
vec_sat_u_sub_imm##IMM##_##T##_fmt_3 (T *out, T *in, unsigned limit) \ |
| 498 |
{ \ |
| 499 |
unsigned i; \ |
| 500 |
for (i = 0; i < limit; i++) \ |
| 501 |
out[i] = (T)IMM > in[i] ? (T)IMM - in[i] : 0; \ |
| 502 |
} |
| 503 |
|
| 504 |
#define DEF_VEC_SAT_U_SUB_IMM_FMT_3_WRAP(T, IMM) \ |
| 505 |
DEF_VEC_SAT_U_SUB_IMM_FMT_3(T, IMM) |
| 506 |
|
| 507 |
#define DEF_VEC_SAT_U_SUB_IMM_FMT_4(T, IMM) \ |
| 508 |
void __attribute__((noinline)) \ |
| 509 |
vec_sat_u_sub_imm##IMM##_##T##_fmt_4 (T *out, T *in, unsigned limit) \ |
| 510 |
{ \ |
| 511 |
unsigned i; \ |
| 512 |
for (i = 0; i < limit; i++) \ |
| 513 |
out[i] = in[i] > (T)IMM ? in[i] - (T)IMM : 0; \ |
| 514 |
} |
| 515 |
|
| 516 |
#define DEF_VEC_SAT_U_SUB_IMM_FMT_4_WRAP(T, IMM) \ |
| 517 |
DEF_VEC_SAT_U_SUB_IMM_FMT_4(T, IMM) |
| 518 |
|
| 519 |
#define RUN_VEC_SAT_U_SUB_IMM_FMT_1(T, out, op_1, expect, IMM, N) \ |
| 520 |
vec_sat_u_sub_imm##IMM##_##T##_fmt_1(out, op_1, N); \ |
| 521 |
VALIDATE_RESULT (out, expect, N) |
| 522 |
#define RUN_VEC_SAT_U_SUB_IMM_FMT_1_WRAP(T, out, op_1, expect, IMM, N) \ |
| 523 |
RUN_VEC_SAT_U_SUB_IMM_FMT_1(T, out, op_1, expect, IMM, N) |
| 524 |
|
| 525 |
#define RUN_VEC_SAT_U_SUB_IMM_FMT_2(T, out, op_1, expect, IMM, N) \ |
| 526 |
vec_sat_u_sub_imm##IMM##_##T##_fmt_2(out, op_1, N); \ |
| 527 |
VALIDATE_RESULT (out, expect, N) |
| 528 |
#define RUN_VEC_SAT_U_SUB_IMM_FMT_2_WRAP(T, out, op_1, expect, IMM, N) \ |
| 529 |
RUN_VEC_SAT_U_SUB_IMM_FMT_2(T, out, op_1, expect, IMM, N) |
| 530 |
|
| 531 |
#define RUN_VEC_SAT_U_SUB_IMM_FMT_3(T, out, op_1, expect, IMM, N) \ |
| 532 |
vec_sat_u_sub_imm##IMM##_##T##_fmt_3(out, op_1, N); \ |
| 533 |
VALIDATE_RESULT (out, expect, N) |
| 534 |
#define RUN_VEC_SAT_U_SUB_IMM_FMT_3_WRAP(T, out, op_1, expect, IMM, N) \ |
| 535 |
RUN_VEC_SAT_U_SUB_IMM_FMT_3(T, out, op_1, expect, IMM, N) |
| 536 |
|
| 537 |
#define RUN_VEC_SAT_U_SUB_IMM_FMT_4(T, out, op_1, expect, IMM, N) \ |
| 538 |
vec_sat_u_sub_imm##IMM##_##T##_fmt_4(out, op_1, N); \ |
| 539 |
VALIDATE_RESULT (out, expect, N) |
| 540 |
#define RUN_VEC_SAT_U_SUB_IMM_FMT_4_WRAP(T, out, op_1, expect, IMM, N) \ |
| 541 |
RUN_VEC_SAT_U_SUB_IMM_FMT_4(T, out, op_1, expect, IMM, N) |
| 542 |
|
| 543 |
#define DEF_VEC_SAT_S_SUB_FMT_1(T, UT, MIN, MAX) \ |
| 544 |
void __attribute__((noinline)) \ |
| 545 |
vec_sat_s_sub_##T##_fmt_1 (T *out, T *op_1, T *op_2, unsigned limit) \ |
| 546 |
{ \ |
| 547 |
unsigned i; \ |
| 548 |
for (i = 0; i < limit; i++) \ |
| 549 |
{ \ |
| 550 |
T x = op_1[i]; \ |
| 551 |
T y = op_2[i]; \ |
| 552 |
T minus = (UT)x - (UT)y; \ |
| 553 |
out[i] = (x ^ y) >= 0 \ |
| 554 |
? minus \ |
| 555 |
: (minus ^ x) >= 0 \ |
| 556 |
? minus \ |
| 557 |
: x < 0 ? MIN : MAX; \ |
| 558 |
} \ |
| 559 |
} |
| 560 |
#define DEF_VEC_SAT_S_SUB_FMT_1_WRAP(T, UT, MIN, MAX) \ |
| 561 |
DEF_VEC_SAT_S_SUB_FMT_1(T, UT, MIN, MAX) |
| 562 |
|
| 563 |
#define DEF_VEC_SAT_S_SUB_FMT_2(T, UT, MIN, MAX) \ |
| 564 |
void __attribute__((noinline)) \ |
| 565 |
vec_sat_s_sub_##T##_fmt_2 (T *out, T *op_1, T *op_2, unsigned limit) \ |
| 566 |
{ \ |
| 567 |
unsigned i; \ |
| 568 |
for (i = 0; i < limit; i++) \ |
| 569 |
{ \ |
| 570 |
T x = op_1[i]; \ |
| 571 |
T y = op_2[i]; \ |
| 572 |
T minus = (UT)x - (UT)y; \ |
| 573 |
out[i] = (x ^ y) >= 0 || (minus ^ x) >= 0 \ |
| 574 |
? minus : x < 0 ? MIN : MAX; \ |
| 575 |
} \ |
| 576 |
} |
| 577 |
#define DEF_VEC_SAT_S_SUB_FMT_2_WRAP(T, UT, MIN, MAX) \ |
| 578 |
DEF_VEC_SAT_S_SUB_FMT_2(T, UT, MIN, MAX) |
| 579 |
|
| 580 |
#define DEF_VEC_SAT_S_SUB_FMT_3(T, UT, MIN, MAX) \ |
| 581 |
void __attribute__((noinline)) \ |
| 582 |
vec_sat_s_sub_##T##_fmt_3 (T *out, T *op_1, T *op_2, unsigned limit) \ |
| 583 |
{ \ |
| 584 |
unsigned i; \ |
| 585 |
for (i = 0; i < limit; i++) \ |
| 586 |
{ \ |
| 587 |
T x = op_1[i]; \ |
| 588 |
T y = op_2[i]; \ |
| 589 |
T minus; \ |
| 590 |
bool overflow = __builtin_sub_overflow (x, y, &minus); \ |
| 591 |
out[i] = overflow ? x < 0 ? MIN : MAX : minus; \ |
| 592 |
} \ |
| 593 |
} |
| 594 |
#define DEF_VEC_SAT_S_SUB_FMT_3_WRAP(T, UT, MIN, MAX) \ |
| 595 |
DEF_VEC_SAT_S_SUB_FMT_3(T, UT, MIN, MAX) |
| 596 |
|
| 597 |
#define DEF_VEC_SAT_S_SUB_FMT_4(T, UT, MIN, MAX) \ |
| 598 |
void __attribute__((noinline)) \ |
| 599 |
vec_sat_s_sub_##T##_fmt_4 (T *out, T *op_1, T *op_2, unsigned limit) \ |
| 600 |
{ \ |
| 601 |
unsigned i; \ |
| 602 |
for (i = 0; i < limit; i++) \ |
| 603 |
{ \ |
| 604 |
T x = op_1[i]; \ |
| 605 |
T y = op_2[i]; \ |
| 606 |
T minus; \ |
| 607 |
bool overflow = __builtin_sub_overflow (x, y, &minus); \ |
| 608 |
out[i] = !overflow ? minus : x < 0 ? MIN : MAX; \ |
| 609 |
} \ |
| 610 |
} |
| 611 |
#define DEF_VEC_SAT_S_SUB_FMT_4_WRAP(T, UT, MIN, MAX) \ |
| 612 |
DEF_VEC_SAT_S_SUB_FMT_4(T, UT, MIN, MAX) |
| 613 |
|
| 614 |
#define RUN_VEC_SAT_U_SUB_FMT_1(T, out, op_1, op_2, N) \ |
| 615 |
vec_sat_u_sub_##T##_fmt_1(out, op_1, op_2, N) |
| 616 |
|
| 617 |
#define RUN_VEC_SAT_U_SUB_FMT_2(T, out, op_1, op_2, N) \ |
| 618 |
vec_sat_u_sub_##T##_fmt_2(out, op_1, op_2, N) |
| 619 |
|
| 620 |
#define RUN_VEC_SAT_U_SUB_FMT_3(T, out, op_1, op_2, N) \ |
| 621 |
vec_sat_u_sub_##T##_fmt_3(out, op_1, op_2, N) |
| 622 |
|
| 623 |
#define RUN_VEC_SAT_U_SUB_FMT_4(T, out, op_1, op_2, N) \ |
| 624 |
vec_sat_u_sub_##T##_fmt_4(out, op_1, op_2, N) |
| 625 |
|
| 626 |
#define RUN_VEC_SAT_U_SUB_FMT_5(T, out, op_1, op_2, N) \ |
| 627 |
vec_sat_u_sub_##T##_fmt_5(out, op_1, op_2, N) |
| 628 |
|
| 629 |
#define RUN_VEC_SAT_U_SUB_FMT_6(T, out, op_1, op_2, N) \ |
| 630 |
vec_sat_u_sub_##T##_fmt_6(out, op_1, op_2, N) |
| 631 |
|
| 632 |
#define RUN_VEC_SAT_U_SUB_FMT_7(T, out, op_1, op_2, N) \ |
| 633 |
vec_sat_u_sub_##T##_fmt_7(out, op_1, op_2, N) |
| 634 |
|
| 635 |
#define RUN_VEC_SAT_U_SUB_FMT_8(T, out, op_1, op_2, N) \ |
| 636 |
vec_sat_u_sub_##T##_fmt_8(out, op_1, op_2, N) |
| 637 |
|
| 638 |
#define RUN_VEC_SAT_U_SUB_FMT_9(T, out, op_1, op_2, N) \ |
| 639 |
vec_sat_u_sub_##T##_fmt_9(out, op_1, op_2, N) |
| 640 |
|
| 641 |
#define RUN_VEC_SAT_U_SUB_FMT_10(T, out, op_1, op_2, N) \ |
| 642 |
vec_sat_u_sub_##T##_fmt_10(out, op_1, op_2, N) |
| 643 |
|
| 644 |
#define RUN_VEC_SAT_U_SUB_FMT_ZIP(T1, T2, x, b, N) \ |
| 645 |
vec_sat_u_sub_##T1##_##T2##_fmt_zip(x, b, N) |
| 646 |
#define RUN_VEC_SAT_U_SUB_FMT_ZIP_WRAP(T1, T2, x, b, N) \ |
| 647 |
RUN_VEC_SAT_U_SUB_FMT_ZIP(T1, T2, x, b, N) \ |
| 648 |
|
| 649 |
#define RUN_VEC_SAT_S_SUB_FMT_1(T, out, op_1, op_2, N) \ |
| 650 |
vec_sat_s_sub_##T##_fmt_1(out, op_1, op_2, N) |
| 651 |
#define RUN_VEC_SAT_S_SUB_FMT_1_WRAP(T, out, op_1, op_2, N) \ |
| 652 |
RUN_VEC_SAT_S_SUB_FMT_1(T, out, op_1, op_2, N) |
| 653 |
|
| 654 |
#define RUN_VEC_SAT_S_SUB_FMT_2(T, out, op_1, op_2, N) \ |
| 655 |
vec_sat_s_sub_##T##_fmt_2(out, op_1, op_2, N) |
| 656 |
#define RUN_VEC_SAT_S_SUB_FMT_2_WRAP(T, out, op_1, op_2, N) \ |
| 657 |
RUN_VEC_SAT_S_SUB_FMT_2(T, out, op_1, op_2, N) |
| 658 |
|
| 659 |
#define RUN_VEC_SAT_S_SUB_FMT_3(T, out, op_1, op_2, N) \ |
| 660 |
vec_sat_s_sub_##T##_fmt_3(out, op_1, op_2, N) |
| 661 |
#define RUN_VEC_SAT_S_SUB_FMT_3_WRAP(T, out, op_1, op_2, N) \ |
| 662 |
RUN_VEC_SAT_S_SUB_FMT_3(T, out, op_1, op_2, N) |
| 663 |
|
| 664 |
#define RUN_VEC_SAT_S_SUB_FMT_4(T, out, op_1, op_2, N) \ |
| 665 |
vec_sat_s_sub_##T##_fmt_4(out, op_1, op_2, N) |
| 666 |
#define RUN_VEC_SAT_S_SUB_FMT_4_WRAP(T, out, op_1, op_2, N) \ |
| 667 |
RUN_VEC_SAT_S_SUB_FMT_4(T, out, op_1, op_2, N) |
| 668 |
|
| 669 |
/******************************************************************************/ |
| 670 |
/* Saturation Sub Truncated (Unsigned and Signed) */ |
| 671 |
/******************************************************************************/ |
| 672 |
#define DEF_VEC_SAT_U_SUB_TRUNC_FMT_1(OUT_T, IN_T) \ |
| 673 |
void __attribute__((noinline)) \ |
| 674 |
vec_sat_u_sub_trunc_##OUT_T##_fmt_1 (OUT_T *out, IN_T *op_1, IN_T y, \ |
| 675 |
unsigned limit) \ |
| 676 |
{ \ |
| 677 |
unsigned i; \ |
| 678 |
for (i = 0; i < limit; i++) \ |
| 679 |
{ \ |
| 680 |
IN_T x = op_1[i]; \ |
| 681 |
out[i] = (OUT_T)(x >= y ? x - y : 0); \ |
| 682 |
} \ |
| 683 |
} |
| 684 |
|
| 685 |
#define RUN_VEC_SAT_U_SUB_TRUNC_FMT_1(OUT_T, IN_T, out, op_1, y, N) \ |
| 686 |
vec_sat_u_sub_trunc_##OUT_T##_fmt_1(out, op_1, y, N) |
| 687 |
|
| 688 |
/******************************************************************************/ |
| 689 |
/* Saturation Truncation (Unsigned and Signed) */ |
| 690 |
/******************************************************************************/ |
| 691 |
#define DEF_VEC_SAT_U_TRUNC_FMT_1(NT, WT) \ |
| 692 |
void __attribute__((noinline)) \ |
| 693 |
vec_sat_u_trunc_##NT##_##WT##_fmt_1 (NT *out, WT *in, unsigned limit) \ |
| 694 |
{ \ |
| 695 |
unsigned i; \ |
| 696 |
for (i = 0; i < limit; i++) \ |
| 697 |
{ \ |
| 698 |
WT x = in[i]; \ |
| 699 |
bool overflow = x > (WT)(NT)(-1); \ |
| 700 |
out[i] = ((NT)x) | (NT)-overflow; \ |
| 701 |
} \ |
| 702 |
} |
| 703 |
#define DEF_VEC_SAT_U_TRUNC_FMT_1_WRAP(NT, WT) DEF_VEC_SAT_U_TRUNC_FMT_1(NT, WT) |
| 704 |
|
| 705 |
#define DEF_VEC_SAT_U_TRUNC_FMT_2(NT, WT) \ |
| 706 |
void __attribute__((noinline)) \ |
| 707 |
vec_sat_u_trunc_##NT##_##WT##_fmt_2 (NT *out, WT *in, unsigned limit) \ |
| 708 |
{ \ |
| 709 |
unsigned i; \ |
| 710 |
for (i = 0; i < limit; i++) \ |
| 711 |
{ \ |
| 712 |
WT max = (WT)(NT)-1; \ |
| 713 |
out[i] = in[i] > max ? (NT)max : (NT)in[i]; \ |
| 714 |
} \ |
| 715 |
} |
| 716 |
#define DEF_VEC_SAT_U_TRUNC_FMT_2_WRAP(NT, WT) DEF_VEC_SAT_U_TRUNC_FMT_2(NT, WT) |
| 717 |
|
| 718 |
#define DEF_VEC_SAT_U_TRUNC_FMT_3(NT, WT) \ |
| 719 |
void __attribute__((noinline)) \ |
| 720 |
vec_sat_u_trunc_##NT##_##WT##_fmt_3 (NT *out, WT *in, unsigned limit) \ |
| 721 |
{ \ |
| 722 |
unsigned i; \ |
| 723 |
for (i = 0; i < limit; i++) \ |
| 724 |
{ \ |
| 725 |
WT max = (WT)(NT)-1; \ |
| 726 |
out[i] = in[i] <= max ? (NT)in[i] : (NT)max; \ |
| 727 |
} \ |
| 728 |
} |
| 729 |
#define DEF_VEC_SAT_U_TRUNC_FMT_3_WRAP(NT, WT) DEF_VEC_SAT_U_TRUNC_FMT_3(NT, WT) |
| 730 |
|
| 731 |
#define DEF_VEC_SAT_U_TRUNC_FMT_4(NT, WT) \ |
| 732 |
void __attribute__((noinline)) \ |
| 733 |
vec_sat_u_trunc_##NT##_##WT##_fmt_4 (NT *out, WT *in, unsigned limit) \ |
| 734 |
{ \ |
| 735 |
unsigned i; \ |
| 736 |
for (i = 0; i < limit; i++) \ |
| 737 |
{ \ |
| 738 |
bool not_overflow = in[i] <= (WT)(NT)(-1); \ |
| 739 |
out[i] = ((NT)in[i]) | (NT)((NT)not_overflow - 1); \ |
| 740 |
} \ |
| 741 |
} |
| 742 |
#define DEF_VEC_SAT_U_TRUNC_FMT_4_WRAP(NT, WT) DEF_VEC_SAT_U_TRUNC_FMT_4(NT, WT) |
| 743 |
|
| 744 |
#define DEF_VEC_SAT_S_TRUNC_FMT_1(NT, WT, NT_MIN, NT_MAX) \ |
| 745 |
void __attribute__((noinline)) \ |
| 746 |
vec_sat_s_trunc_##NT##_##WT##_fmt_1 (NT *out, WT *in, unsigned limit) \ |
| 747 |
{ \ |
| 748 |
unsigned i; \ |
| 749 |
for (i = 0; i < limit; i++) \ |
| 750 |
{ \ |
| 751 |
WT x = in[i]; \ |
| 752 |
NT trunc = (NT)x; \ |
| 753 |
out[i] = (WT)NT_MIN <= x && x <= (WT)NT_MAX \ |
| 754 |
? trunc \ |
| 755 |
: x < 0 ? NT_MIN : NT_MAX; \ |
| 756 |
} \ |
| 757 |
} |
| 758 |
#define DEF_VEC_SAT_S_TRUNC_FMT_1_WRAP(NT, WT, NT_MIN, NT_MAX) \ |
| 759 |
DEF_VEC_SAT_S_TRUNC_FMT_1(NT, WT, NT_MIN, NT_MAX) |
| 760 |
|
| 761 |
#define DEF_VEC_SAT_S_TRUNC_FMT_2(NT, WT, NT_MIN, NT_MAX) \ |
| 762 |
void __attribute__((noinline)) \ |
| 763 |
vec_sat_s_trunc_##NT##_##WT##_fmt_2 (NT *out, WT *in, unsigned limit) \ |
| 764 |
{ \ |
| 765 |
unsigned i; \ |
| 766 |
for (i = 0; i < limit; i++) \ |
| 767 |
{ \ |
| 768 |
WT x = in[i]; \ |
| 769 |
NT trunc = (NT)x; \ |
| 770 |
out[i] = (WT)NT_MIN < x && x < (WT)NT_MAX \ |
| 771 |
? trunc \ |
| 772 |
: x < 0 ? NT_MIN : NT_MAX; \ |
| 773 |
} \ |
| 774 |
} |
| 775 |
#define DEF_VEC_SAT_S_TRUNC_FMT_2_WRAP(NT, WT, NT_MIN, NT_MAX) \ |
| 776 |
DEF_VEC_SAT_S_TRUNC_FMT_2(NT, WT, NT_MIN, NT_MAX) |
| 777 |
|
| 778 |
#define DEF_VEC_SAT_S_TRUNC_FMT_3(NT, WT, NT_MIN, NT_MAX) \ |
| 779 |
void __attribute__((noinline)) \ |
| 780 |
vec_sat_s_trunc_##NT##_##WT##_fmt_3 (NT *out, WT *in, unsigned limit) \ |
| 781 |
{ \ |
| 782 |
unsigned i; \ |
| 783 |
for (i = 0; i < limit; i++) \ |
| 784 |
{ \ |
| 785 |
WT x = in[i]; \ |
| 786 |
NT trunc = (NT)x; \ |
| 787 |
out[i] = (WT)NT_MIN < x && x <= (WT)NT_MAX \ |
| 788 |
? trunc \ |
| 789 |
: x < 0 ? NT_MIN : NT_MAX; \ |
| 790 |
} \ |
| 791 |
} |
| 792 |
#define DEF_VEC_SAT_S_TRUNC_FMT_3_WRAP(NT, WT, NT_MIN, NT_MAX) \ |
| 793 |
DEF_VEC_SAT_S_TRUNC_FMT_3(NT, WT, NT_MIN, NT_MAX) |
| 794 |
|
| 795 |
#define DEF_VEC_SAT_S_TRUNC_FMT_4(NT, WT, NT_MIN, NT_MAX) \ |
| 796 |
void __attribute__((noinline)) \ |
| 797 |
vec_sat_s_trunc_##NT##_##WT##_fmt_4 (NT *out, WT *in, unsigned limit) \ |
| 798 |
{ \ |
| 799 |
unsigned i; \ |
| 800 |
for (i = 0; i < limit; i++) \ |
| 801 |
{ \ |
| 802 |
WT x = in[i]; \ |
| 803 |
NT trunc = (NT)x; \ |
| 804 |
out[i] = (WT)NT_MIN <= x && x < (WT)NT_MAX \ |
| 805 |
? trunc \ |
| 806 |
: x < 0 ? NT_MIN : NT_MAX; \ |
| 807 |
} \ |
| 808 |
} |
| 809 |
#define DEF_VEC_SAT_S_TRUNC_FMT_4_WRAP(NT, WT, NT_MIN, NT_MAX) \ |
| 810 |
DEF_VEC_SAT_S_TRUNC_FMT_4(NT, WT, NT_MIN, NT_MAX) |
| 811 |
|
| 812 |
#define DEF_VEC_SAT_S_TRUNC_FMT_5(NT, WT, NT_MIN, NT_MAX) \ |
| 813 |
void __attribute__((noinline)) \ |
| 814 |
vec_sat_s_trunc_##NT##_##WT##_fmt_5 (NT *out, WT *in, unsigned limit) \ |
| 815 |
{ \ |
| 816 |
unsigned i; \ |
| 817 |
for (i = 0; i < limit; i++) \ |
| 818 |
{ \ |
| 819 |
WT x = in[i]; \ |
| 820 |
NT trunc = (NT)x; \ |
| 821 |
out[i] = (WT)NT_MIN > x || x > (WT)NT_MAX \ |
| 822 |
? x < 0 ? NT_MIN : NT_MAX \ |
| 823 |
: trunc; \ |
| 824 |
} \ |
| 825 |
} |
| 826 |
#define DEF_VEC_SAT_S_TRUNC_FMT_5_WRAP(NT, WT, NT_MIN, NT_MAX) \ |
| 827 |
DEF_VEC_SAT_S_TRUNC_FMT_5(NT, WT, NT_MIN, NT_MAX) |
| 828 |
|
| 829 |
#define DEF_VEC_SAT_S_TRUNC_FMT_6(NT, WT, NT_MIN, NT_MAX) \ |
| 830 |
void __attribute__((noinline)) \ |
| 831 |
vec_sat_s_trunc_##NT##_##WT##_fmt_6 (NT *out, WT *in, unsigned limit) \ |
| 832 |
{ \ |
| 833 |
unsigned i; \ |
| 834 |
for (i = 0; i < limit; i++) \ |
| 835 |
{ \ |
| 836 |
WT x = in[i]; \ |
| 837 |
NT trunc = (NT)x; \ |
| 838 |
out[i] = (WT)NT_MIN >= x || x > (WT)NT_MAX \ |
| 839 |
? x < 0 ? NT_MIN : NT_MAX \ |
| 840 |
: trunc; \ |
| 841 |
} \ |
| 842 |
} |
| 843 |
#define DEF_VEC_SAT_S_TRUNC_FMT_6_WRAP(NT, WT, NT_MIN, NT_MAX) \ |
| 844 |
DEF_VEC_SAT_S_TRUNC_FMT_6(NT, WT, NT_MIN, NT_MAX) |
| 845 |
|
| 846 |
#define DEF_VEC_SAT_S_TRUNC_FMT_7(NT, WT, NT_MIN, NT_MAX) \ |
| 847 |
void __attribute__((noinline)) \ |
| 848 |
vec_sat_s_trunc_##NT##_##WT##_fmt_7 (NT *out, WT *in, unsigned limit) \ |
| 849 |
{ \ |
| 850 |
unsigned i; \ |
| 851 |
for (i = 0; i < limit; i++) \ |
| 852 |
{ \ |
| 853 |
WT x = in[i]; \ |
| 854 |
NT trunc = (NT)x; \ |
| 855 |
out[i] = (WT)NT_MIN > x || x >= (WT)NT_MAX \ |
| 856 |
? x < 0 ? NT_MIN : NT_MAX \ |
| 857 |
: trunc; \ |
| 858 |
} \ |
| 859 |
} |
| 860 |
#define DEF_VEC_SAT_S_TRUNC_FMT_7_WRAP(NT, WT, NT_MIN, NT_MAX) \ |
| 861 |
DEF_VEC_SAT_S_TRUNC_FMT_7(NT, WT, NT_MIN, NT_MAX) |
| 862 |
|
| 863 |
#define DEF_VEC_SAT_S_TRUNC_FMT_8(NT, WT, NT_MIN, NT_MAX) \ |
| 864 |
void __attribute__((noinline)) \ |
| 865 |
vec_sat_s_trunc_##NT##_##WT##_fmt_8 (NT *out, WT *in, unsigned limit) \ |
| 866 |
{ \ |
| 867 |
unsigned i; \ |
| 868 |
for (i = 0; i < limit; i++) \ |
| 869 |
{ \ |
| 870 |
WT x = in[i]; \ |
| 871 |
NT trunc = (NT)x; \ |
| 872 |
out[i] = (WT)NT_MIN >= x || x >= (WT)NT_MAX \ |
| 873 |
? x < 0 ? NT_MIN : NT_MAX \ |
| 874 |
: trunc; \ |
| 875 |
} \ |
| 876 |
} |
| 877 |
#define DEF_VEC_SAT_S_TRUNC_FMT_8_WRAP(NT, WT, NT_MIN, NT_MAX) \ |
| 878 |
DEF_VEC_SAT_S_TRUNC_FMT_8(NT, WT, NT_MIN, NT_MAX) |
| 879 |
|
| 880 |
#define RUN_VEC_SAT_U_TRUNC_FMT_1(NT, WT, out, in, N) \ |
| 881 |
vec_sat_u_trunc_##NT##_##WT##_fmt_1 (out, in, N) |
| 882 |
#define RUN_VEC_SAT_U_TRUNC_FMT_1_WRAP(NT, WT, out, in, N) \ |
| 883 |
RUN_VEC_SAT_U_TRUNC_FMT_1(NT, WT, out, in, N) |
| 884 |
|
| 885 |
#define RUN_VEC_SAT_U_TRUNC_FMT_2(NT, WT, out, in, N) \ |
| 886 |
vec_sat_u_trunc_##NT##_##WT##_fmt_2 (out, in, N) |
| 887 |
#define RUN_VEC_SAT_U_TRUNC_FMT_2_WRAP(NT, WT, out, in, N) \ |
| 888 |
RUN_VEC_SAT_U_TRUNC_FMT_2(NT, WT, out, in, N) |
| 889 |
|
| 890 |
#define RUN_VEC_SAT_U_TRUNC_FMT_3(NT, WT, out, in, N) \ |
| 891 |
vec_sat_u_trunc_##NT##_##WT##_fmt_3 (out, in, N) |
| 892 |
#define RUN_VEC_SAT_U_TRUNC_FMT_3_WRAP(NT, WT, out, in, N) \ |
| 893 |
RUN_VEC_SAT_U_TRUNC_FMT_3(NT, WT, out, in, N) |
| 894 |
|
| 895 |
#define RUN_VEC_SAT_U_TRUNC_FMT_4(NT, WT, out, in, N) \ |
| 896 |
vec_sat_u_trunc_##NT##_##WT##_fmt_4 (out, in, N) |
| 897 |
#define RUN_VEC_SAT_U_TRUNC_FMT_4_WRAP(NT, WT, out, in, N) \ |
| 898 |
RUN_VEC_SAT_U_TRUNC_FMT_4(NT, WT, out, in, N) |
| 899 |
|
| 900 |
#define RUN_VEC_SAT_S_TRUNC_FMT_1(NT, WT, out, in, N) \ |
| 901 |
vec_sat_s_trunc_##NT##_##WT##_fmt_1 (out, in, N) |
| 902 |
#define RUN_VEC_SAT_S_TRUNC_FMT_1_WRAP(NT, WT, out, in, N) \ |
| 903 |
RUN_VEC_SAT_S_TRUNC_FMT_1(NT, WT, out, in, N) |
| 904 |
|
| 905 |
#define RUN_VEC_SAT_S_TRUNC_FMT_2(NT, WT, out, in, N) \ |
| 906 |
vec_sat_s_trunc_##NT##_##WT##_fmt_2 (out, in, N) |
| 907 |
#define RUN_VEC_SAT_S_TRUNC_FMT_2_WRAP(NT, WT, out, in, N) \ |
| 908 |
RUN_VEC_SAT_S_TRUNC_FMT_2(NT, WT, out, in, N) |
| 909 |
|
| 910 |
#define RUN_VEC_SAT_S_TRUNC_FMT_3(NT, WT, out, in, N) \ |
| 911 |
vec_sat_s_trunc_##NT##_##WT##_fmt_3 (out, in, N) |
| 912 |
#define RUN_VEC_SAT_S_TRUNC_FMT_3_WRAP(NT, WT, out, in, N) \ |
| 913 |
RUN_VEC_SAT_S_TRUNC_FMT_3(NT, WT, out, in, N) |
| 914 |
|
| 915 |
#define RUN_VEC_SAT_S_TRUNC_FMT_4(NT, WT, out, in, N) \ |
| 916 |
vec_sat_s_trunc_##NT##_##WT##_fmt_4 (out, in, N) |
| 917 |
#define RUN_VEC_SAT_S_TRUNC_FMT_4_WRAP(NT, WT, out, in, N) \ |
| 918 |
RUN_VEC_SAT_S_TRUNC_FMT_4(NT, WT, out, in, N) |
| 919 |
|
| 920 |
#define RUN_VEC_SAT_S_TRUNC_FMT_5(NT, WT, out, in, N) \ |
| 921 |
vec_sat_s_trunc_##NT##_##WT##_fmt_5 (out, in, N) |
| 922 |
#define RUN_VEC_SAT_S_TRUNC_FMT_5_WRAP(NT, WT, out, in, N) \ |
| 923 |
RUN_VEC_SAT_S_TRUNC_FMT_5(NT, WT, out, in, N) |
| 924 |
|
| 925 |
#define RUN_VEC_SAT_S_TRUNC_FMT_6(NT, WT, out, in, N) \ |
| 926 |
vec_sat_s_trunc_##NT##_##WT##_fmt_6 (out, in, N) |
| 927 |
#define RUN_VEC_SAT_S_TRUNC_FMT_6_WRAP(NT, WT, out, in, N) \ |
| 928 |
RUN_VEC_SAT_S_TRUNC_FMT_6(NT, WT, out, in, N) |
| 929 |
|
| 930 |
#define RUN_VEC_SAT_S_TRUNC_FMT_7(NT, WT, out, in, N) \ |
| 931 |
vec_sat_s_trunc_##NT##_##WT##_fmt_7 (out, in, N) |
| 932 |
#define RUN_VEC_SAT_S_TRUNC_FMT_7_WRAP(NT, WT, out, in, N) \ |
| 933 |
RUN_VEC_SAT_S_TRUNC_FMT_7(NT, WT, out, in, N) |
| 934 |
|
| 935 |
#define RUN_VEC_SAT_S_TRUNC_FMT_8(NT, WT, out, in, N) \ |
| 936 |
vec_sat_s_trunc_##NT##_##WT##_fmt_8 (out, in, N) |
| 937 |
#define RUN_VEC_SAT_S_TRUNC_FMT_8_WRAP(NT, WT, out, in, N) \ |
| 938 |
RUN_VEC_SAT_S_TRUNC_FMT_8(NT, WT, out, in, N) |
| 939 |
|
| 940 |
#endif |