74 #ifndef INCLUDED_volk_32f_x2_add_32f_u_H
75 #define INCLUDED_volk_32f_x2_add_32f_u_H
80 #ifdef LV_HAVE_AVX512F
81 #include <immintrin.h>
83 static inline void volk_32f_x2_add_32f_u_avx512f(
float* cVector,
86 unsigned int num_points)
88 unsigned int number = 0;
89 const unsigned int sixteenthPoints = num_points / 16;
91 float* cPtr = cVector;
92 const float* aPtr = aVector;
93 const float* bPtr = bVector;
95 __m512 aVal, bVal, cVal;
96 for (; number < sixteenthPoints; number++) {
98 aVal = _mm512_loadu_ps(aPtr);
99 bVal = _mm512_loadu_ps(bPtr);
101 cVal = _mm512_add_ps(aVal, bVal);
103 _mm512_storeu_ps(cPtr, cVal);
110 number = sixteenthPoints * 16;
112 for (; number < num_points; number++) {
113 *cPtr++ = (*aPtr++) + (*bPtr++);
121 #include <immintrin.h>
124 const float* aVector,
125 const float* bVector,
126 unsigned int num_points)
128 unsigned int number = 0;
129 const unsigned int eighthPoints = num_points / 8;
130 float* cPtr = cVector;
131 const float* aPtr = aVector;
132 const float* bPtr = bVector;
133 __m256 aVal, bVal, cVal;
134 for (; number < eighthPoints; number++) {
136 aVal = _mm256_loadu_ps(aPtr);
137 bVal = _mm256_loadu_ps(bPtr);
139 cVal = _mm256_add_ps(aVal, bVal);
141 _mm256_storeu_ps(cPtr, cVal);
148 number = eighthPoints * 8;
150 for (; number < num_points; number++) {
151 *cPtr++ = (*aPtr++) + (*bPtr++);
158 #include <xmmintrin.h>
161 const float* aVector,
162 const float* bVector,
163 unsigned int num_points)
165 unsigned int number = 0;
166 const unsigned int quarterPoints = num_points / 4;
168 float* cPtr = cVector;
169 const float* aPtr = aVector;
170 const float* bPtr = bVector;
172 __m128 aVal, bVal, cVal;
173 for (; number < quarterPoints; number++) {
175 aVal = _mm_loadu_ps(aPtr);
176 bVal = _mm_loadu_ps(bPtr);
178 cVal = _mm_add_ps(aVal, bVal);
180 _mm_storeu_ps(cPtr, cVal);
187 number = quarterPoints * 4;
188 for (; number < num_points; number++) {
189 *cPtr++ = (*aPtr++) + (*bPtr++);
195 #ifdef LV_HAVE_GENERIC
198 const float* aVector,
199 const float* bVector,
200 unsigned int num_points)
202 float* cPtr = cVector;
203 const float* aPtr = aVector;
204 const float* bPtr = bVector;
205 unsigned int number = 0;
207 for (number = 0; number < num_points; number++) {
208 *cPtr++ = (*aPtr++) + (*bPtr++);
215 #ifndef INCLUDED_volk_32f_x2_add_32f_a_H
216 #define INCLUDED_volk_32f_x2_add_32f_a_H
218 #include <inttypes.h>
221 #ifdef LV_HAVE_AVX512F
222 #include <immintrin.h>
224 static inline void volk_32f_x2_add_32f_a_avx512f(
float* cVector,
225 const float* aVector,
226 const float* bVector,
227 unsigned int num_points)
229 unsigned int number = 0;
230 const unsigned int sixteenthPoints = num_points / 16;
232 float* cPtr = cVector;
233 const float* aPtr = aVector;
234 const float* bPtr = bVector;
236 __m512 aVal, bVal, cVal;
237 for (; number < sixteenthPoints; number++) {
239 aVal = _mm512_load_ps(aPtr);
240 bVal = _mm512_load_ps(bPtr);
242 cVal = _mm512_add_ps(aVal, bVal);
244 _mm512_store_ps(cPtr, cVal);
251 number = sixteenthPoints * 16;
253 for (; number < num_points; number++) {
254 *cPtr++ = (*aPtr++) + (*bPtr++);
262 #include <immintrin.h>
265 const float* aVector,
266 const float* bVector,
267 unsigned int num_points)
269 unsigned int number = 0;
270 const unsigned int eighthPoints = num_points / 8;
272 float* cPtr = cVector;
273 const float* aPtr = aVector;
274 const float* bPtr = bVector;
276 __m256 aVal, bVal, cVal;
277 for (; number < eighthPoints; number++) {
279 aVal = _mm256_load_ps(aPtr);
280 bVal = _mm256_load_ps(bPtr);
282 cVal = _mm256_add_ps(aVal, bVal);
284 _mm256_store_ps(cPtr, cVal);
291 number = eighthPoints * 8;
292 for (; number < num_points; number++) {
293 *cPtr++ = (*aPtr++) + (*bPtr++);
299 #include <xmmintrin.h>
302 const float* aVector,
303 const float* bVector,
304 unsigned int num_points)
306 unsigned int number = 0;
307 const unsigned int quarterPoints = num_points / 4;
309 float* cPtr = cVector;
310 const float* aPtr = aVector;
311 const float* bPtr = bVector;
313 __m128 aVal, bVal, cVal;
314 for (; number < quarterPoints; number++) {
315 aVal = _mm_load_ps(aPtr);
316 bVal = _mm_load_ps(bPtr);
318 cVal = _mm_add_ps(aVal, bVal);
320 _mm_store_ps(cPtr, cVal);
327 number = quarterPoints * 4;
328 for (; number < num_points; number++) {
329 *cPtr++ = (*aPtr++) + (*bPtr++);
336 #include <arm_neon.h>
339 const float* aVector,
340 const float* bVector,
341 unsigned int num_points)
343 unsigned int number = 0;
344 const unsigned int quarterPoints = num_points / 4;
346 float* cPtr = cVector;
347 const float* aPtr = aVector;
348 const float* bPtr = bVector;
349 float32x4_t aVal, bVal, cVal;
350 for (number = 0; number < quarterPoints; number++) {
352 aVal = vld1q_f32(aPtr);
353 bVal = vld1q_f32(bPtr);
358 cVal = vaddq_f32(aVal, bVal);
360 vst1q_f32(cPtr, cVal);
367 number = quarterPoints * 4;
368 for (; number < num_points; number++) {
369 *cPtr++ = (*aPtr++) + (*bPtr++);
375 #ifdef LV_HAVE_NEONV7
376 extern void volk_32f_x2_add_32f_a_neonasm(
float* cVector,
377 const float* aVector,
378 const float* bVector,
379 unsigned int num_points);
382 #ifdef LV_HAVE_NEONV7
383 extern void volk_32f_x2_add_32f_a_neonpipeline(
float* cVector,
384 const float* aVector,
385 const float* bVector,
386 unsigned int num_points);
389 #ifdef LV_HAVE_GENERIC
392 const float* aVector,
393 const float* bVector,
394 unsigned int num_points)
396 float* cPtr = cVector;
397 const float* aPtr = aVector;
398 const float* bPtr = bVector;
399 unsigned int number = 0;
401 for (number = 0; number < num_points; number++) {
402 *cPtr++ = (*aPtr++) + (*bPtr++);
410 extern void volk_32f_x2_add_32f_a_orc_impl(
float* cVector,
411 const float* aVector,
412 const float* bVector,
413 unsigned int num_points);
415 static inline void volk_32f_x2_add_32f_u_orc(
float* cVector,
416 const float* aVector,
417 const float* bVector,
418 unsigned int num_points)
420 volk_32f_x2_add_32f_a_orc_impl(cVector, aVector, bVector, num_points);