| Line | Branch | Exec | Source |
|---|---|---|---|
| 1 | /* -*- c++ -*- */ | ||
| 2 | /* | ||
| 3 | * Copyright 2012, 2014 Free Software Foundation, Inc. | ||
| 4 | * | ||
| 5 | * This file is part of VOLK | ||
| 6 | * | ||
| 7 | * SPDX-License-Identifier: LGPL-3.0-or-later | ||
| 8 | */ | ||
| 9 | |||
| 10 | /*! | ||
| 11 | * \page volk_32fc_s32fc_multiply_32fc | ||
| 12 | * | ||
| 13 | * \b Overview | ||
| 14 | * | ||
| 15 | * Multiplies the input complex vector by a complex scalar and returns | ||
| 16 | * the results. | ||
| 17 | * | ||
| 18 | * <b>Dispatcher Prototype</b> | ||
| 19 | * \code | ||
| 20 | * void volk_32fc_s32fc_multiply_32fc(lv_32fc_t* cVector, const lv_32fc_t* aVector, const | ||
| 21 | * lv_32fc_t scalar, unsigned int num_points); \endcode | ||
| 22 | * | ||
| 23 | * \b Inputs | ||
| 24 | * \li aVector: The input vector to be multiplied. | ||
| 25 | * \li scalar The complex scalar to multiply against aVector. | ||
| 26 | * \li num_points: The number of complex values in aVector. | ||
| 27 | * | ||
| 28 | * \b Outputs | ||
| 29 | * \li cVector: The vector where the results will be stored. | ||
| 30 | * | ||
| 31 | * \b Example | ||
| 32 | * Generate points around the unit circle and shift the phase pi/3 rad. | ||
| 33 | * \code | ||
| 34 | * int N = 10; | ||
| 35 | * unsigned int alignment = volk_get_alignment(); | ||
| 36 | * lv_32fc_t* in = (lv_32fc_t*)volk_malloc(sizeof(lv_32fc_t)*N, alignment); | ||
| 37 | * lv_32fc_t* out = (lv_32fc_t*)volk_malloc(sizeof(lv_32fc_t)*N, alignment); | ||
| 38 | * lv_32fc_t scalar = lv_cmake((float)std::cos(M_PI/3.f), (float)std::sin(M_PI/3.f)); | ||
| 39 | * | ||
| 40 | * float delta = 2.f*M_PI / (float)N; | ||
| 41 | * for(unsigned int ii = 0; ii < N/2; ++ii){ | ||
| 42 | * // Generate points around the unit circle | ||
| 43 | * float real = std::cos(delta * (float)ii); | ||
| 44 | * float imag = std::sin(delta * (float)ii); | ||
| 45 | * in[ii] = lv_cmake(real, imag); | ||
| 46 | * in[ii+N/2] = lv_cmake(-real, -imag); | ||
| 47 | * } | ||
| 48 | * | ||
| 49 | * volk_32fc_s32fc_multiply_32fc(out, in, scalar, N); | ||
| 50 | * | ||
| 51 | * printf(" mag phase | mag phase\n"); | ||
| 52 | * for(unsigned int ii = 0; ii < N; ++ii){ | ||
| 53 | * printf("%+1.2f %+1.2f | %+1.2f %+1.2f\n", | ||
| 54 | * std::abs(in[ii]), std::arg(in[ii]), | ||
| 55 | * std::abs(out[ii]), std::arg(out[ii])); | ||
| 56 | * } | ||
| 57 | * | ||
| 58 | * volk_free(in); | ||
| 59 | * volk_free(out); | ||
| 60 | * \endcode | ||
| 61 | */ | ||
| 62 | |||
| 63 | #ifndef INCLUDED_volk_32fc_s32fc_multiply_32fc_u_H | ||
| 64 | #define INCLUDED_volk_32fc_s32fc_multiply_32fc_u_H | ||
| 65 | |||
| 66 | #include <float.h> | ||
| 67 | #include <inttypes.h> | ||
| 68 | #include <stdio.h> | ||
| 69 | #include <volk/volk_complex.h> | ||
| 70 | |||
| 71 | #if LV_HAVE_AVX && LV_HAVE_FMA | ||
| 72 | #include <immintrin.h> | ||
| 73 | |||
| 74 | 2 | static inline void volk_32fc_s32fc_multiply_32fc_u_avx_fma(lv_32fc_t* cVector, | |
| 75 | const lv_32fc_t* aVector, | ||
| 76 | const lv_32fc_t scalar, | ||
| 77 | unsigned int num_points) | ||
| 78 | { | ||
| 79 | 2 | unsigned int number = 0; | |
| 80 | 2 | unsigned int i = 0; | |
| 81 | 2 | const unsigned int quarterPoints = num_points / 4; | |
| 82 | 2 | unsigned int isodd = num_points & 3; | |
| 83 | __m256 x, yl, yh, z, tmp1, tmp2; | ||
| 84 | 2 | lv_32fc_t* c = cVector; | |
| 85 | 2 | const lv_32fc_t* a = aVector; | |
| 86 | |||
| 87 | // Set up constant scalar vector | ||
| 88 | 2 | yl = _mm256_set1_ps(lv_creal(scalar)); | |
| 89 | 2 | yh = _mm256_set1_ps(lv_cimag(scalar)); | |
| 90 | |||
| 91 |
2/2✓ Branch 0 taken 65534 times.
✓ Branch 1 taken 2 times.
|
65536 | for (; number < quarterPoints; number++) { |
| 92 | 65534 | x = _mm256_loadu_ps((float*)a); // Load the ar + ai, br + bi as ar,ai,br,bi | |
| 93 | |||
| 94 | 65534 | tmp1 = x; | |
| 95 | |||
| 96 | 65534 | x = _mm256_shuffle_ps(x, x, 0xB1); // Re-arrange x to be ai,ar,bi,br | |
| 97 | |||
| 98 | 65534 | tmp2 = _mm256_mul_ps(x, yh); // tmp2 = ai*ci,ar*ci,bi*di,br*di | |
| 99 | |||
| 100 | 65534 | z = _mm256_fmaddsub_ps( | |
| 101 | tmp1, yl, tmp2); // ar*cr-ai*ci, ai*cr+ar*ci, br*dr-bi*di, bi*dr+br*di | ||
| 102 | |||
| 103 | _mm256_storeu_ps((float*)c, z); // Store the results back into the C container | ||
| 104 | |||
| 105 | 65534 | a += 4; | |
| 106 | 65534 | c += 4; | |
| 107 | } | ||
| 108 | |||
| 109 |
2/2✓ Branch 0 taken 6 times.
✓ Branch 1 taken 2 times.
|
8 | for (i = num_points - isodd; i < num_points; i++) { |
| 110 | 6 | *c++ = (*a++) * scalar; | |
| 111 | } | ||
| 112 | 2 | } | |
| 113 | #endif /* LV_HAVE_AVX && LV_HAVE_FMA */ | ||
| 114 | |||
| 115 | #ifdef LV_HAVE_AVX | ||
| 116 | #include <immintrin.h> | ||
| 117 | |||
| 118 | 2 | static inline void volk_32fc_s32fc_multiply_32fc_u_avx(lv_32fc_t* cVector, | |
| 119 | const lv_32fc_t* aVector, | ||
| 120 | const lv_32fc_t scalar, | ||
| 121 | unsigned int num_points) | ||
| 122 | { | ||
| 123 | 2 | unsigned int number = 0; | |
| 124 | 2 | unsigned int i = 0; | |
| 125 | 2 | const unsigned int quarterPoints = num_points / 4; | |
| 126 | 2 | unsigned int isodd = num_points & 3; | |
| 127 | __m256 x, yl, yh, z, tmp1, tmp2; | ||
| 128 | 2 | lv_32fc_t* c = cVector; | |
| 129 | 2 | const lv_32fc_t* a = aVector; | |
| 130 | |||
| 131 | // Set up constant scalar vector | ||
| 132 | 2 | yl = _mm256_set1_ps(lv_creal(scalar)); | |
| 133 | 2 | yh = _mm256_set1_ps(lv_cimag(scalar)); | |
| 134 | |||
| 135 |
2/2✓ Branch 0 taken 65534 times.
✓ Branch 1 taken 2 times.
|
65536 | for (; number < quarterPoints; number++) { |
| 136 | 65534 | x = _mm256_loadu_ps((float*)a); // Load the ar + ai, br + bi as ar,ai,br,bi | |
| 137 | |||
| 138 | 65534 | tmp1 = _mm256_mul_ps(x, yl); // tmp1 = ar*cr,ai*cr,br*dr,bi*dr | |
| 139 | |||
| 140 | 65534 | x = _mm256_shuffle_ps(x, x, 0xB1); // Re-arrange x to be ai,ar,bi,br | |
| 141 | |||
| 142 | 65534 | tmp2 = _mm256_mul_ps(x, yh); // tmp2 = ai*ci,ar*ci,bi*di,br*di | |
| 143 | |||
| 144 | 65534 | z = _mm256_addsub_ps(tmp1, | |
| 145 | tmp2); // ar*cr-ai*ci, ai*cr+ar*ci, br*dr-bi*di, bi*dr+br*di | ||
| 146 | |||
| 147 | _mm256_storeu_ps((float*)c, z); // Store the results back into the C container | ||
| 148 | |||
| 149 | 65534 | a += 4; | |
| 150 | 65534 | c += 4; | |
| 151 | } | ||
| 152 | |||
| 153 |
2/2✓ Branch 0 taken 6 times.
✓ Branch 1 taken 2 times.
|
8 | for (i = num_points - isodd; i < num_points; i++) { |
| 154 | 6 | *c++ = (*a++) * scalar; | |
| 155 | } | ||
| 156 | 2 | } | |
| 157 | #endif /* LV_HAVE_AVX */ | ||
| 158 | |||
| 159 | #ifdef LV_HAVE_SSE3 | ||
| 160 | #include <pmmintrin.h> | ||
| 161 | |||
| 162 | 2 | static inline void volk_32fc_s32fc_multiply_32fc_u_sse3(lv_32fc_t* cVector, | |
| 163 | const lv_32fc_t* aVector, | ||
| 164 | const lv_32fc_t scalar, | ||
| 165 | unsigned int num_points) | ||
| 166 | { | ||
| 167 | 2 | unsigned int number = 0; | |
| 168 | 2 | const unsigned int halfPoints = num_points / 2; | |
| 169 | |||
| 170 | __m128 x, yl, yh, z, tmp1, tmp2; | ||
| 171 | 2 | lv_32fc_t* c = cVector; | |
| 172 | 2 | const lv_32fc_t* a = aVector; | |
| 173 | |||
| 174 | // Set up constant scalar vector | ||
| 175 | 2 | yl = _mm_set_ps1(lv_creal(scalar)); | |
| 176 | 2 | yh = _mm_set_ps1(lv_cimag(scalar)); | |
| 177 | |||
| 178 |
2/2✓ Branch 0 taken 131070 times.
✓ Branch 1 taken 2 times.
|
131072 | for (; number < halfPoints; number++) { |
| 179 | |||
| 180 | 131070 | x = _mm_loadu_ps((float*)a); // Load the ar + ai, br + bi as ar,ai,br,bi | |
| 181 | |||
| 182 | 131070 | tmp1 = _mm_mul_ps(x, yl); // tmp1 = ar*cr,ai*cr,br*dr,bi*dr | |
| 183 | |||
| 184 | 131070 | x = _mm_shuffle_ps(x, x, 0xB1); // Re-arrange x to be ai,ar,bi,br | |
| 185 | |||
| 186 | 131070 | tmp2 = _mm_mul_ps(x, yh); // tmp2 = ai*ci,ar*ci,bi*di,br*di | |
| 187 | |||
| 188 | 131070 | z = _mm_addsub_ps(tmp1, | |
| 189 | tmp2); // ar*cr-ai*ci, ai*cr+ar*ci, br*dr-bi*di, bi*dr+br*di | ||
| 190 | |||
| 191 | _mm_storeu_ps((float*)c, z); // Store the results back into the C container | ||
| 192 | |||
| 193 | 131070 | a += 2; | |
| 194 | 131070 | c += 2; | |
| 195 | } | ||
| 196 | |||
| 197 |
1/2✓ Branch 0 taken 2 times.
✗ Branch 1 not taken.
|
2 | if ((num_points % 2) != 0) { |
| 198 | 2 | *c = (*a) * scalar; | |
| 199 | } | ||
| 200 | 2 | } | |
| 201 | #endif /* LV_HAVE_SSE */ | ||
| 202 | |||
| 203 | #ifdef LV_HAVE_GENERIC | ||
| 204 | |||
| 205 | 2 | static inline void volk_32fc_s32fc_multiply_32fc_generic(lv_32fc_t* cVector, | |
| 206 | const lv_32fc_t* aVector, | ||
| 207 | const lv_32fc_t scalar, | ||
| 208 | unsigned int num_points) | ||
| 209 | { | ||
| 210 | 2 | lv_32fc_t* cPtr = cVector; | |
| 211 | 2 | const lv_32fc_t* aPtr = aVector; | |
| 212 | 2 | unsigned int number = num_points; | |
| 213 | |||
| 214 | // unwrap loop | ||
| 215 |
2/2✓ Branch 0 taken 32766 times.
✓ Branch 1 taken 2 times.
|
32768 | while (number >= 8) { |
| 216 | 32766 | *cPtr++ = (*aPtr++) * scalar; | |
| 217 | 32766 | *cPtr++ = (*aPtr++) * scalar; | |
| 218 | 32766 | *cPtr++ = (*aPtr++) * scalar; | |
| 219 | 32766 | *cPtr++ = (*aPtr++) * scalar; | |
| 220 | 32766 | *cPtr++ = (*aPtr++) * scalar; | |
| 221 | 32766 | *cPtr++ = (*aPtr++) * scalar; | |
| 222 | 32766 | *cPtr++ = (*aPtr++) * scalar; | |
| 223 | 32766 | *cPtr++ = (*aPtr++) * scalar; | |
| 224 | 32766 | number -= 8; | |
| 225 | } | ||
| 226 | |||
| 227 | // clean up any remaining | ||
| 228 |
2/2✓ Branch 0 taken 14 times.
✓ Branch 1 taken 2 times.
|
16 | while (number-- > 0) |
| 229 | 14 | *cPtr++ = *aPtr++ * scalar; | |
| 230 | 2 | } | |
| 231 | #endif /* LV_HAVE_GENERIC */ | ||
| 232 | |||
| 233 | |||
| 234 | #endif /* INCLUDED_volk_32fc_x2_multiply_32fc_u_H */ | ||
| 235 | #ifndef INCLUDED_volk_32fc_s32fc_multiply_32fc_a_H | ||
| 236 | #define INCLUDED_volk_32fc_s32fc_multiply_32fc_a_H | ||
| 237 | |||
| 238 | #include <float.h> | ||
| 239 | #include <inttypes.h> | ||
| 240 | #include <stdio.h> | ||
| 241 | #include <volk/volk_complex.h> | ||
| 242 | |||
| 243 | #if LV_HAVE_AVX && LV_HAVE_FMA | ||
| 244 | #include <immintrin.h> | ||
| 245 | |||
| 246 | 2 | static inline void volk_32fc_s32fc_multiply_32fc_a_avx_fma(lv_32fc_t* cVector, | |
| 247 | const lv_32fc_t* aVector, | ||
| 248 | const lv_32fc_t scalar, | ||
| 249 | unsigned int num_points) | ||
| 250 | { | ||
| 251 | 2 | unsigned int number = 0; | |
| 252 | 2 | unsigned int i = 0; | |
| 253 | 2 | const unsigned int quarterPoints = num_points / 4; | |
| 254 | 2 | unsigned int isodd = num_points & 3; | |
| 255 | __m256 x, yl, yh, z, tmp1, tmp2; | ||
| 256 | 2 | lv_32fc_t* c = cVector; | |
| 257 | 2 | const lv_32fc_t* a = aVector; | |
| 258 | |||
| 259 | // Set up constant scalar vector | ||
| 260 | 2 | yl = _mm256_set1_ps(lv_creal(scalar)); | |
| 261 | 2 | yh = _mm256_set1_ps(lv_cimag(scalar)); | |
| 262 | |||
| 263 |
2/2✓ Branch 0 taken 65534 times.
✓ Branch 1 taken 2 times.
|
65536 | for (; number < quarterPoints; number++) { |
| 264 | 65534 | x = _mm256_load_ps((float*)a); // Load the ar + ai, br + bi as ar,ai,br,bi | |
| 265 | |||
| 266 | 65534 | tmp1 = x; | |
| 267 | |||
| 268 | 65534 | x = _mm256_shuffle_ps(x, x, 0xB1); // Re-arrange x to be ai,ar,bi,br | |
| 269 | |||
| 270 | 65534 | tmp2 = _mm256_mul_ps(x, yh); // tmp2 = ai*ci,ar*ci,bi*di,br*di | |
| 271 | |||
| 272 | 65534 | z = _mm256_fmaddsub_ps( | |
| 273 | tmp1, yl, tmp2); // ar*cr-ai*ci, ai*cr+ar*ci, br*dr-bi*di, bi*dr+br*di | ||
| 274 | |||
| 275 | _mm256_store_ps((float*)c, z); // Store the results back into the C container | ||
| 276 | |||
| 277 | 65534 | a += 4; | |
| 278 | 65534 | c += 4; | |
| 279 | } | ||
| 280 | |||
| 281 |
2/2✓ Branch 0 taken 6 times.
✓ Branch 1 taken 2 times.
|
8 | for (i = num_points - isodd; i < num_points; i++) { |
| 282 | 6 | *c++ = (*a++) * scalar; | |
| 283 | } | ||
| 284 | 2 | } | |
| 285 | #endif /* LV_HAVE_AVX && LV_HAVE_FMA */ | ||
| 286 | |||
| 287 | |||
| 288 | #ifdef LV_HAVE_AVX | ||
| 289 | #include <immintrin.h> | ||
| 290 | |||
| 291 | 2 | static inline void volk_32fc_s32fc_multiply_32fc_a_avx(lv_32fc_t* cVector, | |
| 292 | const lv_32fc_t* aVector, | ||
| 293 | const lv_32fc_t scalar, | ||
| 294 | unsigned int num_points) | ||
| 295 | { | ||
| 296 | 2 | unsigned int number = 0; | |
| 297 | 2 | unsigned int i = 0; | |
| 298 | 2 | const unsigned int quarterPoints = num_points / 4; | |
| 299 | 2 | unsigned int isodd = num_points & 3; | |
| 300 | __m256 x, yl, yh, z, tmp1, tmp2; | ||
| 301 | 2 | lv_32fc_t* c = cVector; | |
| 302 | 2 | const lv_32fc_t* a = aVector; | |
| 303 | |||
| 304 | // Set up constant scalar vector | ||
| 305 | 2 | yl = _mm256_set1_ps(lv_creal(scalar)); | |
| 306 | 2 | yh = _mm256_set1_ps(lv_cimag(scalar)); | |
| 307 | |||
| 308 |
2/2✓ Branch 0 taken 65534 times.
✓ Branch 1 taken 2 times.
|
65536 | for (; number < quarterPoints; number++) { |
| 309 | 65534 | x = _mm256_load_ps((float*)a); // Load the ar + ai, br + bi as ar,ai,br,bi | |
| 310 | |||
| 311 | 65534 | tmp1 = _mm256_mul_ps(x, yl); // tmp1 = ar*cr,ai*cr,br*dr,bi*dr | |
| 312 | |||
| 313 | 65534 | x = _mm256_shuffle_ps(x, x, 0xB1); // Re-arrange x to be ai,ar,bi,br | |
| 314 | |||
| 315 | 65534 | tmp2 = _mm256_mul_ps(x, yh); // tmp2 = ai*ci,ar*ci,bi*di,br*di | |
| 316 | |||
| 317 | 65534 | z = _mm256_addsub_ps(tmp1, | |
| 318 | tmp2); // ar*cr-ai*ci, ai*cr+ar*ci, br*dr-bi*di, bi*dr+br*di | ||
| 319 | |||
| 320 | _mm256_store_ps((float*)c, z); // Store the results back into the C container | ||
| 321 | |||
| 322 | 65534 | a += 4; | |
| 323 | 65534 | c += 4; | |
| 324 | } | ||
| 325 | |||
| 326 |
2/2✓ Branch 0 taken 6 times.
✓ Branch 1 taken 2 times.
|
8 | for (i = num_points - isodd; i < num_points; i++) { |
| 327 | 6 | *c++ = (*a++) * scalar; | |
| 328 | } | ||
| 329 | 2 | } | |
| 330 | #endif /* LV_HAVE_AVX */ | ||
| 331 | |||
| 332 | #ifdef LV_HAVE_SSE3 | ||
| 333 | #include <pmmintrin.h> | ||
| 334 | |||
| 335 | 2 | static inline void volk_32fc_s32fc_multiply_32fc_a_sse3(lv_32fc_t* cVector, | |
| 336 | const lv_32fc_t* aVector, | ||
| 337 | const lv_32fc_t scalar, | ||
| 338 | unsigned int num_points) | ||
| 339 | { | ||
| 340 | 2 | unsigned int number = 0; | |
| 341 | 2 | const unsigned int halfPoints = num_points / 2; | |
| 342 | |||
| 343 | __m128 x, yl, yh, z, tmp1, tmp2; | ||
| 344 | 2 | lv_32fc_t* c = cVector; | |
| 345 | 2 | const lv_32fc_t* a = aVector; | |
| 346 | |||
| 347 | // Set up constant scalar vector | ||
| 348 | 2 | yl = _mm_set_ps1(lv_creal(scalar)); | |
| 349 | 2 | yh = _mm_set_ps1(lv_cimag(scalar)); | |
| 350 | |||
| 351 |
2/2✓ Branch 0 taken 131070 times.
✓ Branch 1 taken 2 times.
|
131072 | for (; number < halfPoints; number++) { |
| 352 | |||
| 353 | 131070 | x = _mm_load_ps((float*)a); // Load the ar + ai, br + bi as ar,ai,br,bi | |
| 354 | |||
| 355 | 131070 | tmp1 = _mm_mul_ps(x, yl); // tmp1 = ar*cr,ai*cr,br*dr,bi*dr | |
| 356 | |||
| 357 | 131070 | x = _mm_shuffle_ps(x, x, 0xB1); // Re-arrange x to be ai,ar,bi,br | |
| 358 | |||
| 359 | 131070 | tmp2 = _mm_mul_ps(x, yh); // tmp2 = ai*ci,ar*ci,bi*di,br*di | |
| 360 | |||
| 361 | 131070 | z = _mm_addsub_ps(tmp1, | |
| 362 | tmp2); // ar*cr-ai*ci, ai*cr+ar*ci, br*dr-bi*di, bi*dr+br*di | ||
| 363 | |||
| 364 | _mm_store_ps((float*)c, z); // Store the results back into the C container | ||
| 365 | |||
| 366 | 131070 | a += 2; | |
| 367 | 131070 | c += 2; | |
| 368 | } | ||
| 369 | |||
| 370 |
1/2✓ Branch 0 taken 2 times.
✗ Branch 1 not taken.
|
2 | if ((num_points % 2) != 0) { |
| 371 | 2 | *c = (*a) * scalar; | |
| 372 | } | ||
| 373 | 2 | } | |
| 374 | #endif /* LV_HAVE_SSE */ | ||
| 375 | |||
| 376 | #ifdef LV_HAVE_NEON | ||
| 377 | #include <arm_neon.h> | ||
| 378 | |||
| 379 | static inline void volk_32fc_s32fc_multiply_32fc_neon(lv_32fc_t* cVector, | ||
| 380 | const lv_32fc_t* aVector, | ||
| 381 | const lv_32fc_t scalar, | ||
| 382 | unsigned int num_points) | ||
| 383 | { | ||
| 384 | lv_32fc_t* cPtr = cVector; | ||
| 385 | const lv_32fc_t* aPtr = aVector; | ||
| 386 | unsigned int number = num_points; | ||
| 387 | unsigned int quarter_points = num_points / 4; | ||
| 388 | |||
| 389 | float32x4x2_t a_val, scalar_val; | ||
| 390 | float32x4x2_t tmp_imag; | ||
| 391 | |||
| 392 | scalar_val.val[0] = vld1q_dup_f32((const float*)&scalar); | ||
| 393 | scalar_val.val[1] = vld1q_dup_f32(((const float*)&scalar) + 1); | ||
| 394 | for (number = 0; number < quarter_points; ++number) { | ||
| 395 | a_val = vld2q_f32((float*)aPtr); | ||
| 396 | tmp_imag.val[1] = vmulq_f32(a_val.val[1], scalar_val.val[0]); | ||
| 397 | tmp_imag.val[0] = vmulq_f32(a_val.val[0], scalar_val.val[0]); | ||
| 398 | |||
| 399 | tmp_imag.val[1] = vmlaq_f32(tmp_imag.val[1], a_val.val[0], scalar_val.val[1]); | ||
| 400 | tmp_imag.val[0] = vmlsq_f32(tmp_imag.val[0], a_val.val[1], scalar_val.val[1]); | ||
| 401 | |||
| 402 | vst2q_f32((float*)cPtr, tmp_imag); | ||
| 403 | aPtr += 4; | ||
| 404 | cPtr += 4; | ||
| 405 | } | ||
| 406 | |||
| 407 | for (number = quarter_points * 4; number < num_points; number++) { | ||
| 408 | *cPtr++ = *aPtr++ * scalar; | ||
| 409 | } | ||
| 410 | } | ||
| 411 | #endif /* LV_HAVE_NEON */ | ||
| 412 | |||
| 413 | #ifdef LV_HAVE_GENERIC | ||
| 414 | |||
| 415 | 2 | static inline void volk_32fc_s32fc_multiply_32fc_a_generic(lv_32fc_t* cVector, | |
| 416 | const lv_32fc_t* aVector, | ||
| 417 | const lv_32fc_t scalar, | ||
| 418 | unsigned int num_points) | ||
| 419 | { | ||
| 420 | 2 | lv_32fc_t* cPtr = cVector; | |
| 421 | 2 | const lv_32fc_t* aPtr = aVector; | |
| 422 | 2 | unsigned int number = num_points; | |
| 423 | |||
| 424 | // unwrap loop | ||
| 425 |
2/2✓ Branch 0 taken 32766 times.
✓ Branch 1 taken 2 times.
|
32768 | while (number >= 8) { |
| 426 | 32766 | *cPtr++ = (*aPtr++) * scalar; | |
| 427 | 32766 | *cPtr++ = (*aPtr++) * scalar; | |
| 428 | 32766 | *cPtr++ = (*aPtr++) * scalar; | |
| 429 | 32766 | *cPtr++ = (*aPtr++) * scalar; | |
| 430 | 32766 | *cPtr++ = (*aPtr++) * scalar; | |
| 431 | 32766 | *cPtr++ = (*aPtr++) * scalar; | |
| 432 | 32766 | *cPtr++ = (*aPtr++) * scalar; | |
| 433 | 32766 | *cPtr++ = (*aPtr++) * scalar; | |
| 434 | 32766 | number -= 8; | |
| 435 | } | ||
| 436 | |||
| 437 | // clean up any remaining | ||
| 438 |
2/2✓ Branch 0 taken 14 times.
✓ Branch 1 taken 2 times.
|
16 | while (number-- > 0) |
| 439 | 14 | *cPtr++ = *aPtr++ * scalar; | |
| 440 | 2 | } | |
| 441 | #endif /* LV_HAVE_GENERIC */ | ||
| 442 | |||
| 443 | #endif /* INCLUDED_volk_32fc_x2_multiply_32fc_a_H */ | ||
| 444 |