From 70ec7a66f88fafe5553f574f35d13e03ccb9b307 Mon Sep 17 00:00:00 2001 From: Carles Fernandez Date: Fri, 12 Feb 2016 21:24:07 +0100 Subject: [PATCH] fixes --- .../volk_gnsssdr_16ic_x2_dotprodxnpuppet_16ic.h | 8 ++++---- ...k_gnsssdr_16ic_x2_rotator_dot_prod_16ic_xn.h | 17 ++++++++--------- 2 files changed, 12 insertions(+), 13 deletions(-) diff --git a/src/algorithms/libs/volk_gnsssdr_module/volk_gnsssdr/kernels/volk_gnsssdr/volk_gnsssdr_16ic_x2_dotprodxnpuppet_16ic.h b/src/algorithms/libs/volk_gnsssdr_module/volk_gnsssdr/kernels/volk_gnsssdr/volk_gnsssdr_16ic_x2_dotprodxnpuppet_16ic.h index a435ca272..762fd4d8b 100644 --- a/src/algorithms/libs/volk_gnsssdr_module/volk_gnsssdr/kernels/volk_gnsssdr/volk_gnsssdr_16ic_x2_dotprodxnpuppet_16ic.h +++ b/src/algorithms/libs/volk_gnsssdr_module/volk_gnsssdr/kernels/volk_gnsssdr/volk_gnsssdr_16ic_x2_dotprodxnpuppet_16ic.h @@ -51,7 +51,7 @@ static inline void volk_gnsssdr_16ic_x2_dotprodxnpuppet_16ic_generic(lv_16sc_t* 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)); + //result = (lv_16sc_t*)calloc(num_points, sizeof(lv_16sc_t)); volk_gnsssdr_16ic_x2_dot_prod_16ic_xn_generic(result, local_code, (const lv_16sc_t**) in_a, num_a_vectors, num_points); for(unsigned int n = 0; n < num_a_vectors; n++) @@ -73,7 +73,7 @@ static inline void volk_gnsssdr_16ic_x2_dotprodxnpuppet_16ic_a_sse2(lv_16sc_t* r 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)); + //result = (lv_16sc_t*)calloc(num_points, sizeof(lv_16sc_t)); volk_gnsssdr_16ic_x2_dot_prod_16ic_xn_a_sse2(result, local_code, (const lv_16sc_t**) in_a, num_a_vectors, num_points); for(unsigned int n = 0; n < num_a_vectors; n++) @@ -96,7 +96,7 @@ static inline void volk_gnsssdr_16ic_x2_dotprodxnpuppet_16ic_u_sse2(lv_16sc_t* r 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)); + //result = (lv_16sc_t*)calloc(num_points, sizeof(lv_16sc_t)); volk_gnsssdr_16ic_x2_dot_prod_16ic_xn_u_sse2(result, local_code, (const lv_16sc_t**) in_a, num_a_vectors, num_points); for(unsigned int n = 0; n < num_a_vectors; n++) @@ -119,7 +119,7 @@ static inline void volk_gnsssdr_16ic_x2_dotprodxnpuppet_16ic_neon(lv_16sc_t* res 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)); + //result = (lv_16sc_t*)calloc(num_points, sizeof(lv_16sc_t)); volk_gnsssdr_16ic_x2_dot_prod_16ic_xn_neon(result, local_code, (const lv_16sc_t**) in_a, num_a_vectors, num_points); for(unsigned int n = 0; n < num_a_vectors; n++) diff --git a/src/algorithms/libs/volk_gnsssdr_module/volk_gnsssdr/kernels/volk_gnsssdr/volk_gnsssdr_16ic_x2_rotator_dot_prod_16ic_xn.h b/src/algorithms/libs/volk_gnsssdr_module/volk_gnsssdr/kernels/volk_gnsssdr/volk_gnsssdr_16ic_x2_rotator_dot_prod_16ic_xn.h index 902e98877..527b11b41 100644 --- a/src/algorithms/libs/volk_gnsssdr_module/volk_gnsssdr/kernels/volk_gnsssdr/volk_gnsssdr_16ic_x2_rotator_dot_prod_16ic_xn.h +++ b/src/algorithms/libs/volk_gnsssdr_module/volk_gnsssdr/kernels/volk_gnsssdr/volk_gnsssdr_16ic_x2_rotator_dot_prod_16ic_xn.h @@ -546,17 +546,16 @@ static inline void volk_gnsssdr_16ic_x2_rotator_dot_prod_16ic_xn_neon(lv_16sc_t* } - for (int n_vec = 0; n_vec < num_a_vectors; n_vec++) + for (unsigned int n = neon_iters * 4; n < num_points; n++) { - for(unsigned int n = neon_iters * 4; n < num_points; n++) + tmp16 = *in_common++; + tmp32 = lv_cmake((float)lv_creal(tmp16), (float)lv_cimag(tmp16)) * (*phase); + tmp16 = lv_cmake((int16_t)rintf(lv_creal(tmp32)), (int16_t)rintf(lv_cimag(tmp32))); + (*phase) *= phase_inc; + for (int n_vec = 0; n_vec < num_a_vectors; n_vec++) { - tmp16_ = *in_common++; - tmp32_ = lv_cmake((float)lv_creal(tmp16_), (float)lv_cimag(tmp16_)) * (*phase); - tmp16_ = lv_cmake((int16_t)rintf(lv_creal(tmp32_)), (int16_t)rintf(lv_cimag(tmp32_))); - (*phase) *= phase_inc; - tmp = tmp16_ * in_a[n_vec][n]; - _out[n_vec] = lv_cmake(sat_adds16i(lv_creal(_out[n_vec]), lv_creal(tmp)), - sat_adds16i(lv_cimag(_out[n_vec]), lv_cimag(tmp))); + tmp = tmp16 * in_a[n_vec][n]; + result[n_vec] = lv_cmake(sat_adds16i(lv_creal(result[n_vec]), lv_creal(tmp)), sat_adds16i(lv_cimag(result[n_vec]), lv_cimag(tmp))); } } }