71 #ifndef INCLUDED_volk_32fc_x2_multiply_conjugate_32fc_u_H
72 #define INCLUDED_volk_32fc_x2_multiply_conjugate_32fc_u_H
80 #include <immintrin.h>
86 unsigned int num_points)
88 unsigned int number = 0;
89 const unsigned int quarterPoints = num_points / 4;
96 for (; number < quarterPoints; number++) {
102 _mm256_storeu_ps((
float*)c, z);
109 number = quarterPoints * 4;
111 for (; number < num_points; number++) {
119 #include <pmmintrin.h>
125 unsigned int num_points)
127 unsigned int number = 0;
128 const unsigned int halfPoints = num_points / 2;
135 for (; number < halfPoints; number++) {
136 x = _mm_loadu_ps((
float*)a);
137 y = _mm_loadu_ps((
float*)b);
139 _mm_storeu_ps((
float*)c, z);
146 if ((num_points % 2) != 0) {
153 #ifdef LV_HAVE_GENERIC
158 unsigned int num_points)
163 unsigned int number = 0;
165 for (number = 0; number < num_points; number++) {
166 *cPtr++ = (*aPtr++) *
lv_conj(*bPtr++);
173 #ifndef INCLUDED_volk_32fc_x2_multiply_conjugate_32fc_a_H
174 #define INCLUDED_volk_32fc_x2_multiply_conjugate_32fc_a_H
177 #include <inttypes.h>
182 #include <immintrin.h>
188 unsigned int num_points)
190 unsigned int number = 0;
191 const unsigned int quarterPoints = num_points / 4;
198 for (; number < quarterPoints; number++) {
199 x = _mm256_load_ps((
float*)a);
200 y = _mm256_load_ps((
float*)b);
202 _mm256_store_ps((
float*)c, z);
209 number = quarterPoints * 4;
211 for (; number < num_points; number++) {
219 #include <pmmintrin.h>
225 unsigned int num_points)
227 unsigned int number = 0;
228 const unsigned int halfPoints = num_points / 2;
235 for (; number < halfPoints; number++) {
236 x = _mm_load_ps((
float*)a);
237 y = _mm_load_ps((
float*)b);
239 _mm_store_ps((
float*)c, z);
246 if ((num_points % 2) != 0) {
254 #include <arm_neon.h>
259 unsigned int num_points)
263 unsigned int quarter_points = num_points / 4;
264 float32x4x2_t a_val, b_val, c_val;
265 float32x4x2_t tmp_real, tmp_imag;
266 unsigned int number = 0;
268 for (number = 0; number < quarter_points; ++number) {
269 a_val = vld2q_f32((
float*)a_ptr);
270 b_val = vld2q_f32((
float*)b_ptr);
271 b_val.val[1] = vnegq_f32(b_val.val[1]);
277 tmp_real.val[0] = vmulq_f32(a_val.val[0], b_val.val[0]);
279 tmp_real.val[1] = vmulq_f32(a_val.val[1], b_val.val[1]);
283 tmp_imag.val[0] = vmulq_f32(a_val.val[0], b_val.val[1]);
285 tmp_imag.val[1] = vmulq_f32(a_val.val[1], b_val.val[0]);
288 c_val.val[0] = vsubq_f32(tmp_real.val[0], tmp_real.val[1]);
289 c_val.val[1] = vaddq_f32(tmp_imag.val[0], tmp_imag.val[1]);
290 vst2q_f32((
float*)cVector, c_val);
297 for (number = quarter_points * 4; number < num_points; number++) {
298 *cVector++ = (*a_ptr++) * conj(*b_ptr++);
304 #ifdef LV_HAVE_GENERIC
310 unsigned int num_points)
315 unsigned int number = 0;
317 for (number = 0; number < num_points; number++) {
318 *cPtr++ = (*aPtr++) *
lv_conj(*bPtr++);