Modified tracking in order to test the protokernel

Modified tracking in order to test the new 8 bits protokernels’
performance.

src/algorithms/libs/volk_gnsssdr/apps/volk_gnsssdr_profile.cc

@@ -106,25 +106,27 @@ int main(int argc, char *argv[]) {
     //GNSS-SDR PROTO-KERNELS
     //lv_32fc_t sfv = lv_cmake((float)1, (float)2);
     //example: VOLK_PROFILE(volk_gnsssdr_8ic_s8ic_multiply_8ic, 1e-4, sfv, 204602, 1000, &results, benchmark_mode, kernel_regex);
-     VOLK_PROFILE(volk_gnsssdr_32fc_convert_8ic, 1e-4, 0, 204602, 250, &results, benchmark_mode, kernel_regex);
+     VOLK_PROFILE(volk_gnsssdr_32fc_convert_8ic, 1e-4, 0, 16000, 250, &results, benchmark_mode, kernel_regex);
 
-     VOLK_PROFILE(volk_gnsssdr_8ic_x7_cw_vepl_corr_safe_32fc_x5, 1e-4, 0, 204602, 250, &results, benchmark_mode, kernel_regex);
+     VOLK_PROFILE(volk_gnsssdr_32fc_s32f_convert_8ic, 1e-4, 5, 16000, 250, &results, benchmark_mode, kernel_regex);
-     VOLK_PROFILE(volk_gnsssdr_8ic_x7_cw_vepl_corr_unsafe_32fc_x5, 1e-4, 0, 204602, 250, &results, benchmark_mode, kernel_regex);
-     VOLK_PROFILE(volk_gnsssdr_8ic_x7_cw_vepl_corr_TEST_32fc_x5, 1e-4, 0, 204602, 250, &results, benchmark_mode, kernel_regex);
-     VOLK_PROFILE(volk_gnsssdr_16ic_x5_cw_epl_corr_TEST_32fc_x3, 1e-4, 0, 204602, 250, &results, benchmark_mode, kernel_regex);
     
-     VOLK_PROFILE(volk_gnsssdr_8ic_x7_cw_vepl_corr_32fc_x5, 1e-4, 0, 204602, 250, &results, benchmark_mode, kernel_regex);
+     VOLK_PROFILE(volk_gnsssdr_8ic_x7_cw_vepl_corr_safe_32fc_x5, 1e-4, 0, 16000, 250, &results, benchmark_mode, kernel_regex);
-     VOLK_PROFILE(volk_gnsssdr_16ic_x7_cw_vepl_corr_32fc_x5, 1e-4, 0, 204602, 250, &results, benchmark_mode, kernel_regex);
+     VOLK_PROFILE(volk_gnsssdr_8ic_x7_cw_vepl_corr_unsafe_32fc_x5, 1e-4, 0, 16000, 250, &results, benchmark_mode, kernel_regex);
-     VOLK_PROFILE(volk_gnsssdr_32fc_x7_cw_vepl_corr_32fc_x5, 1e-4, 0, 204602, 250, &results, benchmark_mode, kernel_regex);
+     VOLK_PROFILE(volk_gnsssdr_8ic_x7_cw_vepl_corr_TEST_32fc_x5, 1e-4, 0, 16000, 250, &results, benchmark_mode, kernel_regex);
+     VOLK_PROFILE(volk_gnsssdr_16ic_x5_cw_epl_corr_TEST_32fc_x3, 1e-4, 0, 16000, 250, &results, benchmark_mode, kernel_regex);
     
-     VOLK_PROFILE(volk_gnsssdr_8ic_x5_cw_epl_corr_8ic_x3, 1e-4, 0, 204602, 250, &results, benchmark_mode, kernel_regex);
+     VOLK_PROFILE(volk_gnsssdr_8ic_x7_cw_vepl_corr_32fc_x5, 1e-4, 0, 16000, 250, &results, benchmark_mode, kernel_regex);
-     VOLK_PROFILE(volk_gnsssdr_8ic_x5_cw_epl_corr_32fc_x3, 1e-4, 0, 204602, 250, &results, benchmark_mode, kernel_regex);
+     VOLK_PROFILE(volk_gnsssdr_16ic_x7_cw_vepl_corr_32fc_x5, 1e-4, 0, 16000, 250, &results, benchmark_mode, kernel_regex);
-     VOLK_PROFILE(volk_gnsssdr_16ic_x5_cw_epl_corr_32fc_x3, 1e-4, 0, 204602, 250, &results, benchmark_mode, kernel_regex);
+     VOLK_PROFILE(volk_gnsssdr_32fc_x7_cw_vepl_corr_32fc_x5, 1e-4, 0, 16000, 250, &results, benchmark_mode, kernel_regex);
-     VOLK_PROFILE(volk_gnsssdr_32fc_x5_cw_epl_corr_32fc_x3, 1e-4, 0, 204602, 250, &results, benchmark_mode, kernel_regex);
     
-     VOLK_PROFILE(volk_gnsssdr_32fc_convert_16ic, 1e-4, 0, 204602, 250, &results, benchmark_mode, kernel_regex);
+     VOLK_PROFILE(volk_gnsssdr_8ic_x5_cw_epl_corr_8ic_x3, 1e-4, 0, 16000, 250, &results, benchmark_mode, kernel_regex);
+     VOLK_PROFILE(volk_gnsssdr_8ic_x5_cw_epl_corr_32fc_x3, 1e-4, 0, 16000, 250, &results, benchmark_mode, kernel_regex);
+     VOLK_PROFILE(volk_gnsssdr_16ic_x5_cw_epl_corr_32fc_x3, 1e-4, 0, 16000, 250, &results, benchmark_mode, kernel_regex);
+     VOLK_PROFILE(volk_gnsssdr_32fc_x5_cw_epl_corr_32fc_x3, 1e-4, 0, 16000, 250, &results, benchmark_mode, kernel_regex);
     
-     VOLK_PROFILE(volk_gnsssdr_32f_accumulator_s32f, 1e-4, 0, 204602, 10000, &results, benchmark_mode, kernel_regex);
+     VOLK_PROFILE(volk_gnsssdr_32fc_convert_16ic, 1e-4, 0, 16000, 250, &results, benchmark_mode, kernel_regex);
+    
+     /*VOLK_PROFILE(volk_gnsssdr_32f_accumulator_s32f, 1e-4, 0, 204602, 10000, &results, benchmark_mode, kernel_regex);
      VOLK_PROFILE(volk_gnsssdr_8i_accumulator_s8i, 1e-4, 0, 204602, 10000, &results, benchmark_mode, kernel_regex);
      VOLK_PROFILE(volk_gnsssdr_32f_index_max_16u, 3, 0, 204602, 5000, &results, benchmark_mode, kernel_regex);
      VOLK_PROFILE(volk_gnsssdr_8i_index_max_16u, 3, 0, 204602, 5000, &results, benchmark_mode, kernel_regex);
@@ -144,7 +146,7 @@ int main(int argc, char *argv[]) {
      VOLK_PROFILE(volk_gnsssdr_8u_x2_multiply_8u, 1e-4, 0, 204602, 1000, &results, benchmark_mode, kernel_regex);
      VOLK_PROFILE(volk_gnsssdr_64f_accumulator_64f, 1e-4, 0, 16000, 1000, &results, benchmark_mode, kernel_regex);
      VOLK_PROFILE(volk_gnsssdr_32f_s32f_convert_16i, 1e-4, 1, 204602, 250, &results, benchmark_mode, kernel_regex);
-     VOLK_PROFILE(volk_gnsssdr_16i_s32f_convert_32f, 1e-4, 1, 204602, 250, &results, benchmark_mode, kernel_regex);
+     VOLK_PROFILE(volk_gnsssdr_16i_s32f_convert_32f, 1e-4, 1, 204602, 250, &results, benchmark_mode, kernel_regex);*/
     if(store_results) {
         char path[1024];
         volk_gnsssdr_get_config_path(path);

src/algorithms/libs/volk_gnsssdr/kernels/volk_gnsssdr/volk_gnsssdr_32fc_s32f_convert_8ic.h

@@ -0,0 +1,231 @@
+/*!
+ * \file volk_gnsssdr_32fc_s32f_convert_8ic.h
+ * \brief Volk protokernel: converts float32 complex values to 8 integer complex values taking care of overflow
+ * \authors <ul>
+ *          <li> Andrés Cecilia, 2014. a.cecilia.luque(at)gmail.com
+ *          </ul>
+ *
+ * -------------------------------------------------------------------------
+ *
+ * Copyright (C) 2010-2014  (see AUTHORS file for a list of contributors)
+ *
+ * GNSS-SDR is a software defined Global Navigation
+ *          Satellite Systems receiver
+ *
+ * This file is part of GNSS-SDR.
+ *
+ * GNSS-SDR is free software: you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation, either version 3 of the License, or
+ * at your option) any later version.
+ *
+ * GNSS-SDR is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with GNSS-SDR. If not, see <http://www.gnu.org/licenses/>.
+ *
+ * -------------------------------------------------------------------------
+ */
+
+#ifndef INCLUDED_volk_gnsssdr_32fc_s32f_convert_8ic_u_H
+#define INCLUDED_volk_gnsssdr_32fc_s32f_convert_8ic_u_H
+
+#include <inttypes.h>
+#include <stdio.h>
+#include <math.h>
+
+#ifdef LV_HAVE_SSE2
+#include <emmintrin.h>
+/*!
+ \brief Converts a float vector of 64 bits (32 bits each part) into a 16 integer vector (8 bits each part)
+ \param inputVector The floating point input data buffer
+ \param outputVector The 16 bit output data buffer
+ \param num_points The number of data values to be converted
+ */
+static inline void volk_gnsssdr_32fc_s32f_convert_8ic_u_sse2(lv_8sc_t* outputVector, const lv_32fc_t* inputVector, const float scalar, unsigned int num_points){
+    const unsigned int sse_iters = num_points/8;
+    
+    float* inputVectorPtr = (float*)inputVector;
+    int8_t* outputVectorPtr = (int8_t*)outputVector;
+    __m128 invScalar = _mm_set_ps1(1.0/scalar);
+    
+    float min_val = -128;
+    float max_val = 127;
+    
+    __m128 inputVal1, inputVal2, inputVal3, inputVal4;
+    __m128i intInputVal1, intInputVal2, intInputVal3, intInputVal4;
+    __m128i int8InputVal;
+    __m128 ret1, ret2, ret3, ret4;
+    __m128 vmin_val = _mm_set_ps1(min_val);
+    __m128 vmax_val = _mm_set_ps1(max_val);
+    
+    for(unsigned int i = 0;i < sse_iters; i++){
+        inputVal1 = _mm_loadu_ps((float*)inputVectorPtr); inputVectorPtr += 4;
+        inputVal2 = _mm_loadu_ps((float*)inputVectorPtr); inputVectorPtr += 4;
+        inputVal3 = _mm_loadu_ps((float*)inputVectorPtr); inputVectorPtr += 4;
+        inputVal4 = _mm_loadu_ps((float*)inputVectorPtr); inputVectorPtr += 4;
+        
+        inputVal1 = _mm_mul_ps(inputVal1, invScalar);
+        inputVal2 = _mm_mul_ps(inputVal2, invScalar);
+        inputVal3 = _mm_mul_ps(inputVal3, invScalar);
+        inputVal4 = _mm_mul_ps(inputVal4, invScalar);
+        // Clip
+        ret1 = _mm_max_ps(_mm_min_ps(inputVal1, vmax_val), vmin_val);
+        ret2 = _mm_max_ps(_mm_min_ps(inputVal2, vmax_val), vmin_val);
+        ret3 = _mm_max_ps(_mm_min_ps(inputVal3, vmax_val), vmin_val);
+        ret4 = _mm_max_ps(_mm_min_ps(inputVal4, vmax_val), vmin_val);
+        
+        intInputVal1 = _mm_cvtps_epi32(ret1);
+        intInputVal2 = _mm_cvtps_epi32(ret2);
+        intInputVal3 = _mm_cvtps_epi32(ret3);
+        intInputVal4 = _mm_cvtps_epi32(ret4);
+        
+        intInputVal1 = _mm_packs_epi32(intInputVal1, intInputVal2);
+        intInputVal2 = _mm_packs_epi32(intInputVal3, intInputVal4);
+        int8InputVal = _mm_packs_epi16(intInputVal1, intInputVal2);
+        
+        _mm_storeu_si128((__m128i*)outputVectorPtr, int8InputVal);
+        outputVectorPtr += 16;
+    }
+    
+    float scaled = 0;
+    for(unsigned int i = 0; i < (num_points%4)*4; i++){
+        scaled = inputVectorPtr[i]/scalar;
+        if(scaled > max_val)
+            scaled = max_val;
+        else if(scaled < min_val)
+            scaled = min_val;
+        outputVectorPtr[i] = (int8_t)rintf(scaled);
+    }
+}
+#endif /* LV_HAVE_SSE2 */
+
+#ifdef LV_HAVE_GENERIC
+/*!
+ \brief Converts a float vector of 64 bits (32 bits each part) into a 16 integer vector (8 bits each part)
+ \param inputVector The floating point input data buffer
+ \param outputVector The 16 bit output data buffer
+ \param num_points The number of data values to be converted
+ */
+static inline void volk_gnsssdr_32fc_s32f_convert_8ic_generic(lv_8sc_t* outputVector, const lv_32fc_t* inputVector, const float scalar, unsigned int num_points){
+    float* inputVectorPtr = (float*)inputVector;
+    int8_t* outputVectorPtr = (int8_t*)outputVector;
+    float scaled = 0;
+    float min_val = -128;
+    float max_val = 127;
+    
+    for(unsigned int i = 0; i < num_points*2; i++){
+        scaled = (inputVectorPtr[i])/scalar;
+        if(scaled > max_val)
+            scaled = max_val;
+        else if(scaled < min_val)
+            scaled = min_val;
+        outputVectorPtr[i] = (int8_t)rintf(scaled);
+    }
+}
+#endif /* LV_HAVE_GENERIC */
+#endif /* INCLUDED_volk_gnsssdr_32fc_s32f_convert_8ic_u_H */
+
+
+#ifndef INCLUDED_volk_gnsssdr_32fc_s32f_convert_8ic_a_H
+#define INCLUDED_volk_gnsssdr_32fc_s32f_convert_8ic_a_H
+
+#include <volk/volk_common.h>
+#include <inttypes.h>
+#include <stdio.h>
+#include <math.h>
+
+#ifdef LV_HAVE_SSE2
+#include <emmintrin.h>
+/*!
+ \brief Converts a float vector of 64 bits (32 bits each part) into a 16 integer vector (8 bits each part)
+ \param inputVector The floating point input data buffer
+ \param outputVector The 16 bit output data buffer
+ \param num_points The number of data values to be converted
+ */
+static inline void volk_gnsssdr_32fc_s32f_convert_8ic_a_sse2(lv_8sc_t* outputVector, const lv_32fc_t* inputVector, const float scalar, unsigned int num_points){
+    const unsigned int sse_iters = num_points/8;
+    
+    float* inputVectorPtr = (float*)inputVector;
+    int8_t* outputVectorPtr = (int8_t*)outputVector;
+    __m128 invScalar = _mm_set_ps1(1.0/scalar);
+    
+    float min_val = -128;
+    float max_val = 127;
+    
+    __m128 inputVal1, inputVal2, inputVal3, inputVal4;
+    __m128i intInputVal1, intInputVal2, intInputVal3, intInputVal4;
+    __m128i int8InputVal;
+    __m128 ret1, ret2, ret3, ret4;
+    __m128 vmin_val = _mm_set_ps1(min_val);
+    __m128 vmax_val = _mm_set_ps1(max_val);
+    
+    for(unsigned int i = 0;i < sse_iters; i++){
+        inputVal1 = _mm_load_ps((float*)inputVectorPtr); inputVectorPtr += 4;
+        inputVal2 = _mm_load_ps((float*)inputVectorPtr); inputVectorPtr += 4;
+        inputVal3 = _mm_load_ps((float*)inputVectorPtr); inputVectorPtr += 4;
+        inputVal4 = _mm_load_ps((float*)inputVectorPtr); inputVectorPtr += 4;
+        
+        inputVal1 = _mm_mul_ps(inputVal1, invScalar);
+        inputVal2 = _mm_mul_ps(inputVal2, invScalar);
+        inputVal3 = _mm_mul_ps(inputVal3, invScalar);
+        inputVal4 = _mm_mul_ps(inputVal4, invScalar);
+        // Clip
+        ret1 = _mm_max_ps(_mm_min_ps(inputVal1, vmax_val), vmin_val);
+        ret2 = _mm_max_ps(_mm_min_ps(inputVal2, vmax_val), vmin_val);
+        ret3 = _mm_max_ps(_mm_min_ps(inputVal3, vmax_val), vmin_val);
+        ret4 = _mm_max_ps(_mm_min_ps(inputVal4, vmax_val), vmin_val);
+        
+        intInputVal1 = _mm_cvtps_epi32(ret1);
+        intInputVal2 = _mm_cvtps_epi32(ret2);
+        intInputVal3 = _mm_cvtps_epi32(ret3);
+        intInputVal4 = _mm_cvtps_epi32(ret4);
+        
+        intInputVal1 = _mm_packs_epi32(intInputVal1, intInputVal2);
+        intInputVal2 = _mm_packs_epi32(intInputVal3, intInputVal4);
+        int8InputVal = _mm_packs_epi16(intInputVal1, intInputVal2);
+        
+        _mm_store_si128((__m128i*)outputVectorPtr, int8InputVal);
+        outputVectorPtr += 16;
+    }
+    
+    float scaled = 0;
+    for(unsigned int i = 0; i < (num_points%4)*4; i++){
+        scaled = inputVectorPtr[i]/scalar;
+        if(scaled > max_val)
+            scaled = max_val;
+        else if(scaled < min_val)
+            scaled = min_val;
+        outputVectorPtr[i] = (int8_t)rintf(scaled);
+    }
+}
+#endif /* LV_HAVE_SSE2 */
+
+#ifdef LV_HAVE_GENERIC
+/*!
+ \brief Converts a float vector of 64 bits (32 bits each part) into a 16 integer vector (8 bits each part)
+ \param inputVector The floating point input data buffer
+ \param outputVector The 16 bit output data buffer
+ \param num_points The number of data values to be converted
+ */
+static inline void volk_gnsssdr_32fc_s32f_convert_8ic_a_generic(lv_8sc_t* outputVector, const lv_32fc_t* inputVector, const float scalar, unsigned int num_points){
+    float* inputVectorPtr = (float*)inputVector;
+    int8_t* outputVectorPtr = (int8_t*)outputVector;
+    float scaled = 0;
+    float min_val = -128;
+    float max_val = 127;
+    
+    for(unsigned int i = 0; i < num_points*2; i++){
+        scaled = inputVectorPtr[i]/scalar;
+        if(scaled > max_val)
+            scaled = max_val;
+        else if(scaled < min_val)
+            scaled = min_val;
+        outputVectorPtr[i] = (int8_t)rintf(scaled);
+    }
+}
+#endif /* LV_HAVE_GENERIC */
+#endif /* INCLUDED_volk_gnsssdr_32fc_s32f_convert_8ic_a_H */

src/algorithms/libs/volk_gnsssdr/lib/testqa.cc

@@ -50,6 +50,7 @@ VOLK_RUN_TESTS(volk_gnsssdr_8i_max_s8i, 3, 0, 20462, 1);
 VOLK_RUN_TESTS(volk_gnsssdr_64f_accumulator_64f, 3, 0, 20462, 1);
 
 VOLK_RUN_TESTS(volk_gnsssdr_32fc_convert_16ic, 3, 0, 20462, 1);
+VOLK_RUN_TESTS(volk_gnsssdr_32fc_s32f_convert_8ic, 3, 0, 20462, 1);
 VOLK_RUN_TESTS(volk_gnsssdr_32fc_convert_8ic, 3, 0, 20462, 1);
 VOLK_RUN_TESTS(volk_gnsssdr_16i_s32f_convert_32f, 3, 0, 20462, 1);
 

src/algorithms/tracking/gnuradio_blocks/galileo_e1_dll_pll_veml_tracking_cc.cc

@@ -328,7 +328,7 @@ galileo_e1_dll_pll_veml_tracking_cc::~galileo_e1_dll_pll_veml_tracking_cc()
     volk_free(d_late_code8);
     volk_free(d_very_late_code8);
     volk_free(d_carr_sign8);
-    volk_free(in16);
+    volk_free(in8);
 
     delete[] d_ca_code;
     delete[] d_Prompt_buffer;
@@ -376,7 +376,7 @@ int galileo_e1_dll_pll_veml_tracking_cc::general_work (int noutput_items,gr_vect
             update_local_code();
             update_local_carrier();
 
-            // perform carrier wipe-off and compute Very Early, Early, Prompt, Late and Very Late correlation
+            //perform carrier wipe-off and compute Very Early, Early, Prompt, Late and Very Late correlation
             d_correlator.Carrier_wipeoff_and_VEPL_volk(d_current_prn_length_samples,
                     in,
                     d_carr_sign,
@@ -402,18 +402,22 @@ int galileo_e1_dll_pll_veml_tracking_cc::general_work (int noutput_items,gr_vect
             volk_gnsssdr_32fc_convert_16ic(in16, in, d_current_prn_length_samples);
             volk_gnsssdr_32fc_convert_16ic(d_carr_sign16, d_carr_sign, d_current_prn_length_samples);
             
+            volk_gnsssdr_16ic_x7_cw_vepl_corr_32fc_x5(d_Very_Early, d_Early, d_Prompt, d_Late, d_Very_Late, in16, d_carr_sign16, d_very_early_code16, d_early_code16, d_prompt_code16, d_late_code16, d_very_late_code16, d_current_prn_length_samples);
+            
             volk_gnsssdr_32fc_convert_8ic(d_very_early_code8, d_very_early_code, d_current_prn_length_samples);
             volk_gnsssdr_32fc_convert_8ic(d_early_code8, d_early_code, d_current_prn_length_samples);
             volk_gnsssdr_32fc_convert_8ic(d_prompt_code8, d_prompt_code, d_current_prn_length_samples);
             volk_gnsssdr_32fc_convert_8ic(d_late_code8, d_late_code, d_current_prn_length_samples);
             volk_gnsssdr_32fc_convert_8ic(d_very_late_code8, d_very_late_code, d_current_prn_length_samples);
-            volk_gnsssdr_32fc_convert_8ic(in8, in, d_current_prn_length_samples);
             volk_gnsssdr_32fc_convert_8ic(d_carr_sign8, d_carr_sign, d_current_prn_length_samples);
-            
+            volk_gnsssdr_32fc_s32f_convert_8ic(in8, in, 4, d_current_prn_length_samples);
-            volk_gnsssdr_16ic_x7_cw_vepl_corr_32fc_x5(d_Very_Early, d_Early, d_Prompt, d_Late, d_Very_Late, in16, d_carr_sign16, d_very_early_code16, d_early_code16, d_prompt_code16, d_late_code16, d_very_late_code16, d_current_prn_length_samples);
             
             volk_gnsssdr_8ic_x7_cw_vepl_corr_32fc_x5(d_Very_Early, d_Early, d_Prompt, d_Late, d_Very_Late, in8, d_carr_sign8, d_very_early_code8, d_early_code8, d_prompt_code8, d_late_code8, d_very_late_code8, d_current_prn_length_samples);
 
+            volk_gnsssdr_8ic_x7_cw_vepl_corr_unsafe_32fc_x5(d_Very_Early, d_Early, d_Prompt, d_Late, d_Very_Late, in8, d_carr_sign8, d_very_early_code8, d_early_code8, d_prompt_code8, d_late_code8, d_very_late_code8, d_current_prn_length_samples);
+           
+            volk_gnsssdr_8ic_x7_cw_vepl_corr_safe_32fc_x5(d_Very_Early, d_Early, d_Prompt, d_Late, d_Very_Late, in8, d_carr_sign8, d_very_early_code8, d_early_code8, d_prompt_code8, d_late_code8, d_very_late_code8, d_current_prn_length_samples);
+            
             
             // ################## PLL ##########################################################
             // PLL discriminator