58#ifndef INCLUDED_volk_32fc_x2_multiply_conjugate_32fc_u_H
59#define INCLUDED_volk_32fc_x2_multiply_conjugate_32fc_u_H
73 unsigned int num_points)
75 unsigned int number = 0;
76 const unsigned int quarterPoints = num_points / 4;
83 for (; number < quarterPoints; number++) {
89 _mm256_storeu_ps((
float*)c, z);
96 number = quarterPoints * 4;
98 for (; number < num_points; number++) {
106#include <pmmintrin.h>
112 unsigned int num_points)
114 unsigned int number = 0;
115 const unsigned int halfPoints = num_points / 2;
122 for (; number < halfPoints; number++) {
133 if ((num_points % 2) != 0) {
140#ifdef LV_HAVE_GENERIC
145 unsigned int num_points)
150 unsigned int number = 0;
152 for (number = 0; number < num_points; number++) {
153 *cPtr++ = (*aPtr++) *
lv_conj(*bPtr++);
160#ifndef INCLUDED_volk_32fc_x2_multiply_conjugate_32fc_a_H
161#define INCLUDED_volk_32fc_x2_multiply_conjugate_32fc_a_H
169#include <immintrin.h>
175 unsigned int num_points)
177 unsigned int number = 0;
178 const unsigned int quarterPoints = num_points / 4;
185 for (; number < quarterPoints; number++) {
186 x = _mm256_load_ps((
float*)a);
187 y = _mm256_load_ps((
float*)b);
189 _mm256_store_ps((
float*)c, z);
196 number = quarterPoints * 4;
198 for (; number < num_points; number++) {
206#include <pmmintrin.h>
212 unsigned int num_points)
214 unsigned int number = 0;
215 const unsigned int halfPoints = num_points / 2;
222 for (; number < halfPoints; number++) {
233 if ((num_points % 2) != 0) {
246 unsigned int num_points)
250 unsigned int quarter_points = num_points / 4;
251 float32x4x2_t a_val, b_val, c_val;
252 float32x4x2_t tmp_real, tmp_imag;
253 unsigned int number = 0;
255 for (number = 0; number < quarter_points; ++number) {
256 a_val = vld2q_f32((
float*)a_ptr);
257 b_val = vld2q_f32((
float*)b_ptr);
258 b_val.val[1] = vnegq_f32(b_val.val[1]);
264 tmp_real.val[0] = vmulq_f32(a_val.val[0], b_val.val[0]);
266 tmp_real.val[1] = vmulq_f32(a_val.val[1], b_val.val[1]);
270 tmp_imag.val[0] = vmulq_f32(a_val.val[0], b_val.val[1]);
272 tmp_imag.val[1] = vmulq_f32(a_val.val[1], b_val.val[0]);
275 c_val.val[0] = vsubq_f32(tmp_real.val[0], tmp_real.val[1]);
276 c_val.val[1] = vaddq_f32(tmp_imag.val[0], tmp_imag.val[1]);
277 vst2q_f32((
float*)cVector, c_val);
284 for (number = quarter_points * 4; number < num_points; number++) {
285 *cVector++ = (*a_ptr++) * conj(*b_ptr++);