mirrors/gnss-sdr
1
0
Fork 0
mirror of https://github.com/gnss-sdr/gnss-sdr synced 2025-01-30 19:04:51 +00:00
Code Issues Releases Wiki Activity

Add AVX2

Browse Source
This commit is contained in:
Carles Fernandez 2017-11-16 19:51:15 +01:00
parent 6d216fa5a1
commit eecfa6c582
1 changed files with 59 additions and 58 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

117
src/algorithms/libs/volk_gnsssdr_module/volk_gnsssdr/kernels/volk_gnsssdr/volk_gnsssdr_16ic_conjugate_16ic.h
View File

@ -60,35 +60,35 @@
#include <volk_gnsssdr/volk_gnsssdr_complex.h> #include <volk_gnsssdr/volk_gnsssdr_complex.h>
//#ifdef LV_HAVE_AVX2 #ifdef LV_HAVE_AVX2
//#include <immintrin.h> #include <immintrin.h>
//
//static inline void volk_gnsssdr_16ic_conjugate_16ic_u_avx2(lv_16sc_t* cVector, const lv_16sc_t* aVector, unsigned int num_points) static inline void volk_gnsssdr_16ic_conjugate_16ic_u_avx2(lv_16sc_t* cVector, const lv_16sc_t* aVector, unsigned int num_points)
//{ {
// const unsigned int avx2_iters = num_points / 16; const unsigned int avx2_iters = num_points / 8;
// unsigned int i; unsigned int i;
// lv_16sc_t* c = cVector; lv_16sc_t* c = cVector;
// const lv_16sc_t* a = aVector; const lv_16sc_t* a = aVector;
//
// __m256i tmp; __m256i tmp;
// __m256i conjugator = _mm256_setr_epi8(1, -1, 1, -1, 1, -1, 1, -1, 1, -1, 1, -1, 1, -1, 1, -1, 1, -1, 1, -1, 1, -1, 1, -1, 1, -1, 1, -1, 1, -1, 1, -1); __m256i conjugator = _mm256_setr_epi16(1, -1, 1, -1, 1, -1, 1, -1, 1, -1, 1, -1, 1, -1, 1, -1);
//
// for (i = 0; i < avx2_iters; ++i) for (i = 0; i < avx2_iters; ++i)
// { {
// tmp = _mm256_loadu_si256((__m256i*)a); tmp = _mm256_loadu_si256((__m256i*)a);
// tmp = _mm256_sign_epi8(tmp, conjugator); tmp = _mm256_sign_epi16(tmp, conjugator);
// _mm256_storeu_si256((__m256i*)c, tmp); _mm256_storeu_si256((__m256i*)c, tmp);
//
// a += 16; a += 8;
// c += 16; c += 8;
// } }
//
// for (i = avx2_iters * 16; i < num_points; ++i) for (i = avx2_iters * 8; i < num_points; ++i)
// { {
// *c++ = lv_conj(*a++); *c++ = lv_conj(*a++);
// } }
//} }
//#endif /* LV_HAVE_AVX2 */ #endif /* LV_HAVE_AVX2 */
//#ifdef LV_HAVE_AVX //#ifdef LV_HAVE_AVX
@ -249,35 +249,36 @@ static inline void volk_gnsssdr_16ic_conjugate_16ic_generic(lv_16sc_t* cVector,
//#endif /* LV_HAVE_AVX */ //#endif /* LV_HAVE_AVX */
// //
// //
//#ifdef LV_HAVE_AVX2 #ifdef LV_HAVE_AVX2
//#include <immintrin.h> #include <immintrin.h>
//
//static inline void volk_gnsssdr_16ic_conjugate_16ic_a_avx2(lv_16sc_t* cVector, const lv_16sc_t* aVector, unsigned int num_points) static inline void volk_gnsssdr_16ic_conjugate_16ic_a_avx2(lv_16sc_t* cVector, const lv_16sc_t* aVector, unsigned int num_points)
//{ {
// const unsigned int avx2_iters = num_points / 16; const unsigned int avx2_iters = num_points / 8;
// unsigned int i; unsigned int i;
// lv_16sc_t* c = cVector; lv_16sc_t* c = cVector;
// const lv_16sc_t* a = aVector; const lv_16sc_t* a = aVector;
//
// __m256i tmp; __m256i tmp;
// __m256i conjugator = _mm256_setr_epi8(1, -1, 1, -1, 1, -1, 1, -1, 1, -1, 1, -1, 1, -1, 1, -1, 1, -1, 1, -1, 1, -1, 1, -1, 1, -1, 1, -1, 1, -1, 1, -1); __m256i conjugator = _mm256_setr_epi16(1, -1, 1, -1, 1, -1, 1, -1, 1, -1, 1, -1, 1, -1, 1, -1);
//
// for (i = 0; i < avx2_iters; ++i) for (i = 0; i < avx2_iters; ++i)
// { {
// tmp = _mm256_load_si256((__m256i*)a); tmp = _mm256_load_si256((__m256i*)a);
// tmp = _mm256_sign_epi8(tmp, conjugator); tmp = _mm256_sign_epi16(tmp, conjugator);
// _mm256_store_si256((__m256i*)c, tmp); _mm256_store_si256((__m256i*)c, tmp);
//
// a += 16; a += 8;
// c += 16; c += 8;
// } }
//
// for (i = avx2_iters * 16; i < num_points; ++i) for (i = avx2_iters * 8; i < num_points; ++i)
// { {
// *c++ = lv_conj(*a++); *c++ = lv_conj(*a++);
// } }
//} }
//#endif /* LV_HAVE_AVX2 */
#endif /* LV_HAVE_AVX2 */
// //
// //
#ifdef LV_HAVE_SSSE3 #ifdef LV_HAVE_SSSE3