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>
85 const lv_32fc_t* bVector,
unsigned int num_points)
87 unsigned int number = 0;
88 const unsigned int quarterPoints = num_points / 4;
95 for(; number < quarterPoints; number++){
96 x = _mm256_loadu_ps((
float*) a);
97 y = _mm256_loadu_ps((
float*) b);
99 _mm256_storeu_ps((
float*) c, z);
106 number = quarterPoints * 4;
108 for(; number < num_points; number++){
116 #include <pmmintrin.h>
121 const lv_32fc_t* bVector,
unsigned int num_points)
123 unsigned int number = 0;
124 const unsigned int halfPoints = num_points / 2;
131 for(; number < halfPoints; number++){
132 x = _mm_loadu_ps((
float*) a);
133 y = _mm_loadu_ps((
float*) b);
135 _mm_storeu_ps((
float*) c, z);
142 if((num_points % 2) != 0){
149 #ifdef LV_HAVE_GENERIC
153 const lv_32fc_t* bVector,
unsigned int num_points)
158 unsigned int number = 0;
160 for(number = 0; number < num_points; number++){
161 *cPtr++ = (*aPtr++) *
lv_conj(*bPtr++);
169 #ifndef INCLUDED_volk_32fc_x2_multiply_conjugate_32fc_a_H
170 #define INCLUDED_volk_32fc_x2_multiply_conjugate_32fc_a_H
172 #include <inttypes.h>
178 #include <immintrin.h>
183 const lv_32fc_t* bVector,
unsigned int num_points)
185 unsigned int number = 0;
186 const unsigned int quarterPoints = num_points / 4;
193 for(; number < quarterPoints; number++){
194 x = _mm256_load_ps((
float*) a);
195 y = _mm256_load_ps((
float*) b);
197 _mm256_store_ps((
float*) c, z);
204 number = quarterPoints * 4;
206 for(; number < num_points; number++){
214 #include <pmmintrin.h>
219 const lv_32fc_t* bVector,
unsigned int num_points)
221 unsigned int number = 0;
222 const unsigned int halfPoints = num_points / 2;
229 for(; number < halfPoints; number++){
230 x = _mm_load_ps((
float*) a);
231 y = _mm_load_ps((
float*) b);
233 _mm_store_ps((
float*) c, z);
240 if((num_points % 2) != 0){
248 #include <arm_neon.h>
252 const lv_32fc_t* bVector,
unsigned int num_points)
256 unsigned int quarter_points = num_points / 4;
257 float32x4x2_t a_val, b_val, c_val;
258 float32x4x2_t tmp_real, tmp_imag;
259 unsigned int number = 0;
261 for(number = 0; number < quarter_points; ++number) {
262 a_val = vld2q_f32((
float*)a_ptr);
263 b_val = vld2q_f32((
float*)b_ptr);
264 b_val.val[1] = vnegq_f32(b_val.val[1]);
270 tmp_real.val[0] = vmulq_f32(a_val.val[0], b_val.val[0]);
272 tmp_real.val[1] = vmulq_f32(a_val.val[1], b_val.val[1]);
276 tmp_imag.val[0] = vmulq_f32(a_val.val[0], b_val.val[1]);
278 tmp_imag.val[1] = vmulq_f32(a_val.val[1], b_val.val[0]);
281 c_val.val[0] = vsubq_f32(tmp_real.val[0], tmp_real.val[1]);
282 c_val.val[1] = vaddq_f32(tmp_imag.val[0], tmp_imag.val[1]);
283 vst2q_f32((
float*)cVector, c_val);
290 for(number = quarter_points*4; number < num_points; number++){
291 *cVector++ = (*a_ptr++) * conj(*b_ptr++);
297 #ifdef LV_HAVE_GENERIC
301 const lv_32fc_t* bVector,
unsigned int num_points)
306 unsigned int number = 0;
308 for(number = 0; number < num_points; number++){
309 *cPtr++ = (*aPtr++) *
lv_conj(*bPtr++);