mirrors/gnss-sdr
1
0
Fork 0
mirror of https://github.com/gnss-sdr/gnss-sdr synced 2024-12-15 04:30:33 +00:00
Code Issues Releases Wiki Activity

Still some bugs to fix in 16sc dot product. All fixed now.

Browse Source
This commit is contained in:
Javier Arribas 2016-01-18 10:43:09 +01:00
parent fd2af02aec
commit 4ba75a3fbe
1 changed files with 49 additions and 53 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

12
src/algorithms/libs/volk_gnsssdr_module/volk_gnsssdr/kernels/volk_gnsssdr/volk_gnsssdr_16ic_x2_dot_prod_16ic.h
View File

@ -40,7 +40,6 @@
#include <volk_gnsssdr/volk_gnsssdr_common.h> #include <volk_gnsssdr/volk_gnsssdr_common.h>
#include <volk_gnsssdr/volk_gnsssdr_complex.h> #include <volk_gnsssdr/volk_gnsssdr_complex.h>
#include <volk_gnsssdr/saturation_arithmetic.h> #include <volk_gnsssdr/saturation_arithmetic.h>
//#include <stdio.h>
#ifdef LV_HAVE_GENERIC #ifdef LV_HAVE_GENERIC
@ -56,11 +55,9 @@ static inline void volk_gnsssdr_16ic_x2_dot_prod_16ic_generic(lv_16sc_t* result,
result[0] = lv_cmake((int16_t)0, (int16_t)0); result[0] = lv_cmake((int16_t)0, (int16_t)0);
for (unsigned int n = 0; n < num_points; n++) for (unsigned int n = 0; n < num_points; n++)
{ {
//r*a.r - i*a.i, i*a.r + r*a.i
lv_16sc_t tmp = in_a[n] * in_b[n]; lv_16sc_t tmp = in_a[n] * in_b[n];
result[0] += lv_cmake(sat_adds16i(lv_creal(tmp), lv_creal(tmp)), sat_adds16i(lv_cimag(tmp), lv_cimag(tmp) )); result[0] = lv_cmake(sat_adds16i(lv_creal(result[0]), lv_creal(tmp)), sat_adds16i(lv_cimag(result[0]), lv_cimag(tmp) ));
} }
//printf("generic result = %i,%i", lv_creal(result[0]),lv_cimag(result[0]));
} }
#endif /*LV_HAVE_GENERIC*/ #endif /*LV_HAVE_GENERIC*/
@ -70,7 +67,6 @@ static inline void volk_gnsssdr_16ic_x2_dot_prod_16ic_generic(lv_16sc_t* result,
#include <emmintrin.h> #include <emmintrin.h>
static inline void volk_gnsssdr_16ic_x2_dot_prod_16ic_a_sse2(lv_16sc_t* out, const lv_16sc_t* in_a, const lv_16sc_t* in_b, unsigned int num_points) static inline void volk_gnsssdr_16ic_x2_dot_prod_16ic_a_sse2(lv_16sc_t* out, const lv_16sc_t* in_a, const lv_16sc_t* in_b, unsigned int num_points)
{ {
lv_16sc_t dotProduct = lv_cmake((int16_t)0, (int16_t)0); lv_16sc_t dotProduct = lv_cmake((int16_t)0, (int16_t)0);
const unsigned int sse_iters = num_points / 4; const unsigned int sse_iters = num_points / 4;
@ -127,14 +123,14 @@ static inline void volk_gnsssdr_16ic_x2_dot_prod_16ic_a_sse2(lv_16sc_t* out, con
for (int i = 0; i < 4; ++i) for (int i = 0; i < 4; ++i)
{ {
dotProduct += lv_cmake(sat_adds16i(lv_creal(dotProductVector[i]), lv_creal(dotProductVector[i])), sat_adds16i(lv_cimag(dotProductVector[i]), lv_cimag(dotProductVector[i]))); dotProduct = lv_cmake(sat_adds16i(lv_creal(dotProduct), lv_creal(dotProductVector[i])), sat_adds16i(lv_cimag(dotProduct), lv_cimag(dotProductVector[i])));
} }
} }
for (unsigned int i = sse_iters * 4; i < num_points; ++i) for (unsigned int i = 0; i < (num_points % 4); ++i)
{ {
lv_16sc_t tmp = (*_in_a++) * (*_in_b++); lv_16sc_t tmp = (*_in_a++) * (*_in_b++);
dotProduct += lv_cmake( sat_adds16i(lv_creal(tmp), lv_creal(tmp)), sat_adds16i(lv_cimag(tmp), lv_cimag(tmp))); dotProduct = lv_cmake( sat_adds16i(lv_creal(dotProduct), lv_creal(tmp)), sat_adds16i(lv_cimag(dotProduct), lv_cimag(tmp)));
} }
*_out = dotProduct; *_out = dotProduct;