mirrors/gnss-sdr
1
0
Fork 0
mirror of https://github.com/gnss-sdr/gnss-sdr synced 2024-07-02 01:43:20 +00:00
Code Issues Releases Wiki Activity

fixes

Browse Source
This commit is contained in:
Carles Fernandez 2016-02-13 12:53:39 +01:00
parent 5ed27443fe
commit 83d7b72743
2 changed files with 17 additions and 7 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

14
src/algorithms/libs/volk_gnsssdr_module/volk_gnsssdr/kernels/volk_gnsssdr/volk_gnsssdr_16ic_x2_rotator_dot_prod_16ic_xn.h
View File

@ -371,7 +371,7 @@ static inline void volk_gnsssdr_16ic_x2_rotator_dot_prod_16ic_xn_u_sse3(lv_16sc_
_mm_storeu_si128((__m128i*)dotProductVector, result); // Store the results back into the dot product vector
dotProduct = lv_cmake(0,0);
for (int i = 0; i<4; ++i)
for (int i = 0; i < 4; ++i)
{
dotProduct = lv_cmake(sat_adds16i(lv_creal(dotProduct), lv_creal(dotProductVector[i])),
sat_adds16i(lv_cimag(dotProduct), lv_cimag(dotProductVector[i])));
@ -404,6 +404,16 @@ static inline void volk_gnsssdr_16ic_x2_rotator_dot_prod_16ic_xn_u_sse3(lv_16sc_
#ifdef LV_HAVE_NEON
#include <arm_neon.h>
/*!
\brief Rotates and multiplies the reference complex vector with multiple versions of another complex vector, accumulates the results and stores them in the output vector
\param[out] result Array of num_a_vectors components with the multiple versions of in_a multiplied and accumulated The vector where the accumulated result will be stored
\param[in] in_common Pointer to one of the vectors to be rotated, multiplied and accumulated (reference vector)
\param[in] phase_inc Phase increment = lv_cmake(cos(phase_step_rad), -sin(phase_step_rad))
\param[in,out] phase Initial / final phase
\param[in] in_a Pointer to an array of pointers to multiple versions of the other vector to be multiplied and accumulated
\param[in] num_a_vectors Number of vectors to be multiplied by the reference vector and accumulated
\param[in] num_points The Number of complex values to be multiplied together, accumulated and stored into result
*/
static inline void volk_gnsssdr_16ic_x2_rotator_dot_prod_16ic_xn_neon(lv_16sc_t* out, const lv_16sc_t* in_common, const lv_32fc_t phase_inc, lv_32fc_t* phase, const lv_16sc_t** in_a, int num_a_vectors, unsigned int num_points)
{
const unsigned int neon_iters = num_points / 4;
@ -428,7 +438,7 @@ static inline void volk_gnsssdr_16ic_x2_rotator_dot_prod_16ic_xn_neon(lv_16sc_t*
lv_32fc_t phase2 = (lv_32fc_t)(*phase) * phase_inc;
lv_32fc_t phase3 = phase2 * phase_inc;
lv_32fc_t phase4 = phase3 * phase_inc;;
lv_32fc_t phase4 = phase3 * phase_inc;
__VOLK_ATTR_ALIGNED(16) float32_t __phase_real[4] = { lv_creal((*phase)), lv_creal(phase2), lv_creal(phase3), lv_creal(phase4) };
__VOLK_ATTR_ALIGNED(16) float32_t __phase_imag[4] = { lv_cimag((*phase)), lv_cimag(phase2), lv_cimag(phase3), lv_cimag(phase4) };

10
src/algorithms/libs/volk_gnsssdr_module/volk_gnsssdr/kernels/volk_gnsssdr/volk_gnsssdr_16ic_x2_rotator_dotprodxnpuppet_16ic.h
View File

@ -57,7 +57,7 @@ static inline void volk_gnsssdr_16ic_x2_rotator_dotprodxnpuppet_16ic_generic(lv_
for(unsigned int n = 0; n < num_a_vectors; n++)
{
in_a[n] = (lv_16sc_t*)volk_gnsssdr_malloc(sizeof(lv_16sc_t) * num_points, volk_gnsssdr_get_alignment());
memcpy(in_a[n], in, sizeof(lv_16sc_t) * num_points);
memcpy((lv_16sc_t*)in_a[n], (lv_16sc_t*)in, sizeof(lv_16sc_t) * num_points);
}
//result = (lv_16sc_t*)calloc(num_points, sizeof(lv_16sc_t));
volk_gnsssdr_16ic_x2_rotator_dot_prod_16ic_xn_generic(&result[0], local_code, phase_inc[0], phase,(const lv_16sc_t**) in_a, num_a_vectors, num_points);
@ -117,8 +117,8 @@ static inline void volk_gnsssdr_16ic_x2_rotator_dotprodxnpuppet_16ic_u_sse3(lv_1
lv_16sc_t** in_a = (lv_16sc_t**)volk_gnsssdr_malloc(sizeof(lv_16sc_t*) * num_a_vectors, volk_gnsssdr_get_alignment());
for(unsigned int n = 0; n < num_a_vectors; n++)
{
in_a[n] = (lv_16sc_t*)volk_gnsssdr_malloc(sizeof(lv_16sc_t)*num_points, volk_gnsssdr_get_alignment());
memcpy(in_a[n], in, sizeof(lv_16sc_t)*num_points);
in_a[n] = (lv_16sc_t*)volk_gnsssdr_malloc(sizeof(lv_16sc_t) * num_points, volk_gnsssdr_get_alignment());
memcpy(in_a[n], in, sizeof(lv_16sc_t) * num_points);
}
//result = (lv_16sc_t*)calloc(num_points, sizeof(lv_16sc_t));
volk_gnsssdr_16ic_x2_rotator_dot_prod_16ic_xn_u_sse3(&result[0], local_code, phase_inc[0], phase, (const lv_16sc_t**) in_a, num_a_vectors, num_points);
@ -148,8 +148,8 @@ static inline void volk_gnsssdr_16ic_x2_rotator_dotprodxnpuppet_16ic_neon(lv_16s
lv_16sc_t** in_a = (lv_16sc_t**)volk_gnsssdr_malloc(sizeof(lv_16sc_t*) * num_a_vectors, volk_gnsssdr_get_alignment());
for(unsigned int n = 0; n < num_a_vectors; n++)
{
in_a[n] = (lv_16sc_t*)volk_gnsssdr_malloc(sizeof(lv_16sc_t)*num_points, volk_gnsssdr_get_alignment());
memcpy(in_a[n], in, sizeof(lv_16sc_t)*num_points);
in_a[n] = (lv_16sc_t*)volk_gnsssdr_malloc(sizeof(lv_16sc_t) * num_points, volk_gnsssdr_get_alignment());
memcpy((lv_16sc_t*)in_a[n], (lv_16sc_t*)in, sizeof(lv_16sc_t) * num_points);
}
//result = (lv_16sc_t*)calloc(num_points, sizeof(lv_16sc_t));
volk_gnsssdr_16ic_x2_rotator_dot_prod_16ic_xn_neon(result, local_code, phase_inc[0], phase, (const lv_16sc_t**) in_a, num_a_vectors, num_points);