Adding and integrating sincos kernel

src/algorithms/acquisition/gnuradio_blocks/galileo_e5a_noncoherent_iq_acquisition_caf_cc.cc

@@ -40,7 +40,7 @@
 #include <gnuradio/io_signature.h>
 #include <glog/logging.h>
 #include <volk/volk.h>
-#include "gnss_signal_processing.h"
+#include <volk_gnsssdr/volk_gnsssdr.h>
 #include "control_message_factory.h"
 
 using google::LogMessage;
@@ -277,6 +277,7 @@ void galileo_e5a_noncoherentIQ_acquisition_caf_cc::init()
     d_gnss_synchro->Acq_samplestamp_samples = 0;
     d_mag = 0.0;
     d_input_power = 0.0;
+    const double GALILEO_TWO_PI = 6.283185307179600;
 
     // Count the number of bins
     d_num_doppler_bins = 0;
@@ -293,7 +294,9 @@ void galileo_e5a_noncoherentIQ_acquisition_caf_cc::init()
         {
             d_grid_doppler_wipeoffs[doppler_index] = static_cast<gr_complex*>(volk_malloc(d_fft_size * sizeof(gr_complex), volk_get_alignment()));
             int doppler = -static_cast<int>(d_doppler_max) + d_doppler_step * doppler_index;
-            complex_exp_gen_conj(d_grid_doppler_wipeoffs[doppler_index], d_freq + doppler, d_fs_in, d_fft_size);
+            float phase_step_rad = GALILEO_TWO_PI * (d_freq + doppler) / static_cast<float>(d_fs_in);
+
+            volk_gnsssdr_s32f_sincos_32fc(d_grid_doppler_wipeoffs[doppler_index], - phase_step_rad, d_fft_size);
         }
 
     /* CAF Filtering to resolve doppler ambiguity. Phase and quadrature must be processed

src/algorithms/acquisition/gnuradio_blocks/galileo_pcps_8ms_acquisition_cc.cc

@@ -34,7 +34,7 @@
 #include <glog/logging.h>
 #include <gnuradio/io_signature.h>
 #include <volk/volk.h>
-#include "gnss_signal_processing.h"
+#include <volk_gnsssdr/volk_gnsssdr.h>
 #include "control_message_factory.h"
 
 using google::LogMessage;
@@ -157,7 +157,7 @@ void galileo_pcps_8ms_acquisition_cc::init()
     d_gnss_synchro->Acq_samplestamp_samples = 0;
     d_mag = 0.0;
     d_input_power = 0.0;
-
+    const double GALILEO_TWO_PI = 6.283185307179600;
     // Count the number of bins
     d_num_doppler_bins = 0;
     for (int doppler = static_cast<int>(-d_doppler_max);
@@ -173,7 +173,8 @@ void galileo_pcps_8ms_acquisition_cc::init()
         {
             d_grid_doppler_wipeoffs[doppler_index] = static_cast<gr_complex*>(volk_malloc(d_fft_size * sizeof(gr_complex), volk_get_alignment()));
             int doppler = -static_cast<int>(d_doppler_max) + d_doppler_step * doppler_index;
-            complex_exp_gen_conj(d_grid_doppler_wipeoffs[doppler_index], d_freq + doppler, d_fs_in, d_fft_size);
+            float phase_step_rad = static_cast<float>(GALILEO_TWO_PI) * (d_freq + doppler) / static_cast<float>(d_fs_in);
+            volk_gnsssdr_s32f_sincos_32fc(d_grid_doppler_wipeoffs[doppler_index], - phase_step_rad, d_fft_size);
         }
 }
 

src/algorithms/acquisition/gnuradio_blocks/pcps_acquisition_cc.cc

@@ -38,9 +38,9 @@
 #include <gnuradio/io_signature.h>
 #include <glog/logging.h>
 #include <volk/volk.h>
+#include <volk_gnsssdr/volk_gnsssdr.h>
 #include "gnss_signal_processing.h"
 #include "control_message_factory.h"
-#include <gnuradio/fxpt.h>  // fixed point sine and cosine
 #include "GPS_L1_CA.h" //GPS_TWO_PI
 
 
@@ -174,18 +174,8 @@ void pcps_acquisition_cc::set_local_code(std::complex<float> * code)
 
 void pcps_acquisition_cc::update_local_carrier(gr_complex* carrier_vector, int correlator_length_samples, float freq)
 {
-    float sin_f, cos_f;
+    float phase_step_rad = GPS_TWO_PI * freq / static_cast<float>(d_fs_in);
-    float phase_step_rad= GPS_TWO_PI * freq/ static_cast<float>(d_fs_in);
+    volk_gnsssdr_s32f_sincos_32fc(carrier_vector, - phase_step_rad, correlator_length_samples);
-
-    int phase_step_rad_i = gr::fxpt::float_to_fixed(phase_step_rad);
-    int phase_rad_i = 0;
-
-    for(int i = 0; i < correlator_length_samples; i++)
-        {
-            gr::fxpt::sincos(phase_rad_i, &sin_f, &cos_f);
-            carrier_vector[i] = gr_complex(cos_f, -sin_f);
-            phase_rad_i += phase_step_rad_i;
-        }
 }
 
 void pcps_acquisition_cc::init()

src/algorithms/acquisition/gnuradio_blocks/pcps_acquisition_fine_doppler_cc.cc

@@ -36,7 +36,7 @@
 #include <glog/logging.h>
 #include <gnuradio/io_signature.h>
 #include <volk/volk.h>
-#include "nco_lib.h"
+#include <volk_gnsssdr/volk_gnsssdr.h>
 #include "concurrent_map.h"
 #include "gnss_signal_processing.h"
 #include "gps_sdr_signal_processing.h"
@@ -184,14 +184,16 @@ void pcps_acquisition_fine_doppler_cc::forecast (int noutput_items,
 void pcps_acquisition_fine_doppler_cc::reset_grid()
 {
     d_well_count = 0;
-    for (int i=0; i<d_num_doppler_points; i++)
+    for (int i = 0; i < d_num_doppler_points; i++)
         {
-            for (unsigned int j=0; j < d_fft_size; j++)
+            for (unsigned int j = 0; j < d_fft_size; j++)
                 {
                     d_grid_data[i][j] = 0.0;
                 }
         }
 }
+
+
 void pcps_acquisition_fine_doppler_cc::update_carrier_wipeoff()
 {
     // create the carrier Doppler wipeoff signals
@@ -200,13 +202,12 @@ void pcps_acquisition_fine_doppler_cc::update_carrier_wipeoff()
     d_grid_doppler_wipeoffs = new gr_complex*[d_num_doppler_points];
     for (int doppler_index = 0; doppler_index < d_num_doppler_points; doppler_index++)
         {
-
             doppler_hz = d_config_doppler_min + d_doppler_step*doppler_index;
             // doppler search steps
             // compute the carrier doppler wipe-off signal and store it
             phase_step_rad = static_cast<float>(GPS_TWO_PI) * ( d_freq + doppler_hz ) / static_cast<float>(d_fs_in);
             d_grid_doppler_wipeoffs[doppler_index] = new gr_complex[d_fft_size];
-            fxp_nco(d_grid_doppler_wipeoffs[doppler_index], d_fft_size,0, phase_step_rad);
+            volk_gnsssdr_s32f_sincos_32fc(d_grid_doppler_wipeoffs[doppler_index], - phase_step_rad, d_fft_size);
         }
 }
 

src/algorithms/acquisition/gnuradio_blocks/pcps_acquisition_sc.cc

@@ -41,7 +41,6 @@
 #include "gnss_signal_processing.h"
 #include "control_message_factory.h"
 #include <volk_gnsssdr/volk_gnsssdr.h>
-#include <gnuradio/fxpt.h>  // fixed point sine and cosine
 #include "GPS_L1_CA.h" //GPS_TWO_PI
 
 using google::LogMessage;
@@ -177,18 +176,8 @@ void pcps_acquisition_sc::set_local_code(std::complex<float> * code)
 
 void pcps_acquisition_sc::update_local_carrier(gr_complex* carrier_vector, int correlator_length_samples, float freq)
 {
-    float sin_f, cos_f;
+    float phase_step_rad = GPS_TWO_PI * freq / static_cast<float>(d_fs_in);
-    float phase_step_rad= GPS_TWO_PI * freq/ static_cast<float>(d_fs_in);
+    volk_gnsssdr_s32f_sincos_32fc(carrier_vector, - phase_step_rad, correlator_length_samples);
-
-    int phase_step_rad_i = gr::fxpt::float_to_fixed(phase_step_rad);
-    int phase_rad_i = 0;
-
-    for(int i = 0; i < correlator_length_samples; i++)
-        {
-            gr::fxpt::sincos(phase_rad_i, &sin_f, &cos_f);
-            carrier_vector[i] = gr_complex(cos_f, -sin_f);
-            phase_rad_i += phase_step_rad_i;
-        }
 }
 
 void pcps_acquisition_sc::init()

src/algorithms/acquisition/gnuradio_blocks/pcps_assisted_acquisition_cc.cc

@@ -35,7 +35,7 @@
 #include <glog/logging.h>
 #include <gnuradio/io_signature.h>
 #include <volk/volk.h>
-#include "nco_lib.h"
+#include <volk_gnsssdr/volk_gnsssdr.h>
 #include "concurrent_map.h"
 #include "gnss_signal_processing.h"
 #include "control_message_factory.h"
@@ -252,7 +252,7 @@ void pcps_assisted_acquisition_cc::redefine_grid()
             // compute the carrier doppler wipe-off signal and store it
             phase_step_rad = static_cast<float>(GPS_TWO_PI) * doppler_hz / static_cast<float>(d_fs_in);
             d_grid_doppler_wipeoffs[doppler_index] = new gr_complex[d_fft_size];
-            fxp_nco(d_grid_doppler_wipeoffs[doppler_index], d_fft_size, 0, phase_step_rad);
+            volk_gnsssdr_s32f_sincos_32fc(d_grid_doppler_wipeoffs[doppler_index], - phase_step_rad, d_fft_size);
         }
 }
 

src/algorithms/acquisition/gnuradio_blocks/pcps_cccwsr_acquisition_cc.cc

@@ -39,8 +39,9 @@
 #include <glog/logging.h>
 #include <gnuradio/io_signature.h>
 #include <volk/volk.h>
-#include "gnss_signal_processing.h"
+#include <volk_gnsssdr/volk_gnsssdr.h>
 #include "control_message_factory.h"
+#include "GPS_L1_CA.h" //GPS_TWO_PI
 
 
 using google::LogMessage;
@@ -188,8 +189,9 @@ void pcps_cccwsr_acquisition_cc::init()
             d_grid_doppler_wipeoffs[doppler_index] = static_cast<gr_complex*>(volk_malloc(d_fft_size * sizeof(gr_complex), volk_get_alignment()));
 
             int doppler = -static_cast<int>(d_doppler_max) + d_doppler_step * doppler_index;
-            complex_exp_gen_conj(d_grid_doppler_wipeoffs[doppler_index],
+            float phase_step_rad = GPS_TWO_PI * (d_freq + doppler) / static_cast<float>(d_fs_in);
-                                 d_freq + doppler, d_fs_in, d_fft_size);
+
+            volk_gnsssdr_s32f_sincos_32fc(d_grid_doppler_wipeoffs[doppler_index], - phase_step_rad, d_fft_size);
         }
 }
 

src/algorithms/acquisition/gnuradio_blocks/pcps_multithread_acquisition_cc.cc

@@ -39,8 +39,9 @@
 #include <glog/logging.h>
 #include <gnuradio/io_signature.h>
 #include <volk/volk.h>
-#include "gnss_signal_processing.h"
+#include <volk_gnsssdr/volk_gnsssdr.h>
 #include "control_message_factory.h"
+#include "GPS_L1_CA.h" //GPS_TWO_PI
 
 using google::LogMessage;
 
@@ -174,8 +175,8 @@ void pcps_multithread_acquisition_cc::init()
             d_grid_doppler_wipeoffs[doppler_index] = static_cast<gr_complex*>(volk_malloc(d_fft_size * sizeof(gr_complex), volk_get_alignment()));
 
             int doppler = -(int)d_doppler_max + d_doppler_step * doppler_index;
-            complex_exp_gen_conj(d_grid_doppler_wipeoffs[doppler_index],
+            float phase_step_rad = static_cast<float>(GPS_TWO_PI) * (d_freq + doppler) / static_cast<float>(d_fs_in);
-                                 d_freq + doppler, d_fs_in, d_fft_size);
+            volk_gnsssdr_s32f_sincos_32fc(d_grid_doppler_wipeoffs[doppler_index], - phase_step_rad, d_fft_size);
         }
 }
 

src/algorithms/acquisition/gnuradio_blocks/pcps_opencl_acquisition_cc.cc

@@ -56,11 +56,11 @@
 #include <glog/logging.h>
 #include <gnuradio/io_signature.h>
 #include <volk/volk.h>
-#include "gnss_signal_processing.h"
+#include <volk_gnsssdr/volk_gnsssdr.h>
 #include "control_message_factory.h"
 #include "fft_base_kernels.h"
 #include "fft_internal.h"
-
+#include "GPS_L1_CA.h" //GPS_TWO_PI
 
 
 using google::LogMessage;
@@ -315,9 +315,9 @@ void pcps_opencl_acquisition_cc::init()
         {
             d_grid_doppler_wipeoffs[doppler_index] = static_cast<gr_complex*>(volk_malloc(d_fft_size * sizeof(gr_complex), volk_get_alignment()));
 
-            int doppler= -static_cast<int>(d_doppler_max) + d_doppler_step * doppler_index;
+            int doppler = -static_cast<int>(d_doppler_max) + d_doppler_step * doppler_index;
-            complex_exp_gen_conj(d_grid_doppler_wipeoffs[doppler_index],
+            float phase_step_rad = static_cast<float>(GPS_TWO_PI) * (d_freq + doppler) / static_cast<float>(d_fs_in);
-                                 d_freq + doppler, d_fs_in, d_fft_size);
+            volk_gnsssdr_s32f_sincos_32fc(d_grid_doppler_wipeoffs[doppler_index], - phase_step_rad, d_fft_size);
 
             if (d_opencl == 0)
                 {

src/algorithms/acquisition/gnuradio_blocks/pcps_quicksync_acquisition_cc.cc

@@ -34,9 +34,9 @@
 #include <gnuradio/io_signature.h>
 #include <glog/logging.h>
 #include <volk/volk.h>
+#include <volk_gnsssdr/volk_gnsssdr.h>
 #include "control_message_factory.h"
-#include "gnss_signal_processing.h"
+#include "GPS_L1_CA.h"
-
 
 
 using google::LogMessage;
@@ -220,9 +220,9 @@ void pcps_quicksync_acquisition_cc::init()
         {
             d_grid_doppler_wipeoffs[doppler_index] = static_cast<gr_complex*>(volk_malloc(d_samples_per_code * d_folding_factor * sizeof(gr_complex), volk_get_alignment()));
             int doppler = -static_cast<int>(d_doppler_max) + d_doppler_step * doppler_index;
-            complex_exp_gen_conj(d_grid_doppler_wipeoffs[doppler_index],
+            float phase_step_rad = GPS_TWO_PI * (d_freq + doppler) / static_cast<float>(d_fs_in);
-                    d_freq + doppler, d_fs_in,
+
-                    d_samples_per_code * d_folding_factor);
+            volk_gnsssdr_s32f_sincos_32fc(d_grid_doppler_wipeoffs[doppler_index], - phase_step_rad, d_samples_per_code * d_folding_factor);
         }
     // DLOG(INFO) << "end init";
 }

src/algorithms/acquisition/gnuradio_blocks/pcps_tong_acquisition_cc.cc

@@ -53,8 +53,9 @@
 #include <glog/logging.h>
 #include <gnuradio/io_signature.h>
 #include <volk/volk.h>
+#include <volk_gnsssdr/volk_gnsssdr.h>
 #include "control_message_factory.h"
-#include "gnss_signal_processing.h"
+#include "GPS_L1_CA.h" //GPS_TWO_PI
 
 using google::LogMessage;
 
@@ -185,9 +186,9 @@ void pcps_tong_acquisition_cc::init()
             d_grid_doppler_wipeoffs[doppler_index] = static_cast<gr_complex*>(volk_malloc(d_fft_size * sizeof(gr_complex), volk_get_alignment()));
 
             int doppler = -static_cast<int>(d_doppler_max) + d_doppler_step * doppler_index;
+            float phase_step_rad = GPS_TWO_PI * (d_freq + doppler) / static_cast<float>(d_fs_in);
 
-            complex_exp_gen_conj(d_grid_doppler_wipeoffs[doppler_index],
+            volk_gnsssdr_s32f_sincos_32fc(d_grid_doppler_wipeoffs[doppler_index], - phase_step_rad, d_fft_size);
-                                 d_freq + doppler, d_fs_in, d_fft_size);
 
             d_grid_data[doppler_index] = static_cast<float*>(volk_malloc(d_fft_size * sizeof(float), volk_get_alignment()));
 

src/algorithms/libs/volk_gnsssdr_module/volk_gnsssdr/kernels/volk_gnsssdr/volk_gnsssdr_s32f_sincos_32fc.h

@@ -0,0 +1,543 @@
+/*!
+ * \file volk_gnsssdr_s32f_sincos_32fc.h
+ * \brief VOLK_GNSSSDR kernel: Computes the sine and cosine of a vector of floats.
+ * \authors <ul>
+ *          <li> Carles Fernandez-Prades, 2016. cfernandez(at)cttc.es
+ *          </ul>
+ *
+ * VOLK_GNSSSDR kernel that computes the sine and cosine of a vector of floats.
+ *
+ * -------------------------------------------------------------------------
+ *
+ * Copyright (C) 2010-2015  (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/>.
+ *
+ * -------------------------------------------------------------------------
+ */
+
+/*!
+ * \page volk_gnsssdr_s32f_sincos_32fc
+ *
+ * \b Overview
+ *
+ * VOLK_GNSSSDR kernel that computes the sine and cosine with a fixed
+ * phase increment \p phase_inc per sample, providing the output in a complex vector (cosine, sine)
+ *
+ * <b>Dispatcher Prototype</b>
+ * \code
+ * void volk_gnsssdr_s32f_sincos_32fc(lv_32fc_t* out, const float phase_inc, unsigned int num_points)
+ * \endcode
+ *
+ * \b Inputs
+ * \li phase_inc:      Phase increment per sample, in radians.
+ * \li num_points:     Number of components in \p in to be computed.
+ *
+ * \b Outputs
+ * \li out:            Vector of the form lv_32fc_t out[n] = lv_cmake(cos(in[n]), sin(in[n]))
+ *
+ */
+
+
+#ifndef INCLUDED_volk_gnsssdr_s32f_sincos_32fc_H
+#define INCLUDED_volk_gnsssdr_s32f_sincos_32fc_H
+
+#include <math.h>
+#include <volk_gnsssdr/volk_gnsssdr_common.h>
+#include <volk_gnsssdr/volk_gnsssdr_complex.h>
+
+
+#ifdef LV_HAVE_SSE2
+#include <emmintrin.h>
+/* Adapted from http://gruntthepeon.free.fr/ssemath/sse_mathfun.h, original code from Julien Pommier  */
+/* Based on algorithms from the cephes library http://www.netlib.org/cephes/   */
+static inline void volk_gnsssdr_s32f_sincos_32fc_a_sse2(lv_32fc_t* out, const float phase_inc, unsigned int num_points)
+{
+    lv_32fc_t* bPtr = out;
+
+    const unsigned int sse_iters = num_points / 4;
+    unsigned int number = 0;
+    float _phase;
+
+    __m128 sine, cosine, aux, x, four_phases_reg;
+    __m128 xmm1, xmm2, xmm3 = _mm_setzero_ps(), sign_bit_sin, y;
+    __m128i emm0, emm2, emm4;
+
+    /* declare some SSE constants */
+    static const int _ps_inv_sign_mask[4] __attribute__((aligned(16))) = { ~0x80000000, ~0x80000000, ~0x80000000, ~0x80000000 };
+    static const int _ps_sign_mask[4] __attribute__((aligned(16))) = { (int)0x80000000, (int)0x80000000, (int)0x80000000, (int)0x80000000 };
+
+    static const float _ps_cephes_FOPI[4] __attribute__((aligned(16))) = { 1.27323954473516, 1.27323954473516, 1.27323954473516, 1.27323954473516 };
+    static const int _pi32_1[4]  __attribute__((aligned(16))) = { 1, 1, 1, 1 };
+    static const int _pi32_inv1[4]  __attribute__((aligned(16))) = { ~1, ~1, ~1, ~1 };
+    static const int _pi32_2[4]  __attribute__((aligned(16))) = { 2, 2, 2, 2};
+    static const int _pi32_4[4]  __attribute__((aligned(16))) = { 4, 4, 4, 4};
+
+    static const float _ps_minus_cephes_DP1[4] __attribute__((aligned(16))) = { -0.78515625, -0.78515625, -0.78515625, -0.78515625 };
+    static const float _ps_minus_cephes_DP2[4] __attribute__((aligned(16))) = { -2.4187564849853515625e-4, -2.4187564849853515625e-4, -2.4187564849853515625e-4, -2.4187564849853515625e-4 };
+    static const float _ps_minus_cephes_DP3[4] __attribute__((aligned(16))) = { -3.77489497744594108e-8, -3.77489497744594108e-8, -3.77489497744594108e-8, -3.77489497744594108e-8 };
+    static const float _ps_coscof_p0[4] __attribute__((aligned(16))) = { 2.443315711809948E-005, 2.443315711809948E-005, 2.443315711809948E-005, 2.443315711809948E-005 };
+    static const float _ps_coscof_p1[4] __attribute__((aligned(16))) = { -1.388731625493765E-003, -1.388731625493765E-003, -1.388731625493765E-003, -1.388731625493765E-003 };
+    static const float _ps_coscof_p2[4] __attribute__((aligned(16))) = { 4.166664568298827E-002, 4.166664568298827E-002, 4.166664568298827E-002, 4.166664568298827E-002 };
+    static const float _ps_sincof_p0[4] __attribute__((aligned(16))) = { -1.9515295891E-4, -1.9515295891E-4, -1.9515295891E-4, -1.9515295891E-4 };
+    static const float _ps_sincof_p1[4] __attribute__((aligned(16))) = { 8.3321608736E-3, 8.3321608736E-3, 8.3321608736E-3, 8.3321608736E-3 };
+    static const float _ps_sincof_p2[4] __attribute__((aligned(16))) = { -1.6666654611E-1, -1.6666654611E-1, -1.6666654611E-1, -1.6666654611E-1 };
+    static const float _ps_0p5[4] __attribute__((aligned(16))) = { 0.5f, 0.5f, 0.5f, 0.5f };
+    static const float _ps_1[4] __attribute__((aligned(16))) = { 1.0f, 1.0f, 1.0f, 1.0f };
+
+    float four_phases[4] __attribute__((aligned(16))) = { 0.0f, phase_inc, 2 * phase_inc, 3 * phase_inc };
+    float four_phases_inc[4] __attribute__((aligned(16))) = { 4 * phase_inc, 4 * phase_inc, 4 * phase_inc, 4 * phase_inc };
+    four_phases_reg = _mm_load_ps(four_phases);
+    const __m128 four_phases_inc_reg = _mm_load_ps(four_phases_inc);
+
+    for(;number < sse_iters; number++)
+        {
+            x = four_phases_reg;
+
+            sign_bit_sin = x;
+            /* take the absolute value */
+            x = _mm_and_ps(x, *(__m128*)_ps_inv_sign_mask);
+            /* extract the sign bit (upper one) */
+            sign_bit_sin = _mm_and_ps(sign_bit_sin, *(__m128*)_ps_sign_mask);
+
+            /* scale by 4/Pi */
+            y = _mm_mul_ps(x, *(__m128*)_ps_cephes_FOPI);
+
+            /* store the integer part of y in emm2 */
+            emm2 = _mm_cvttps_epi32(y);
+
+            /* j=(j+1) & (~1) (see the cephes sources) */
+            emm2 = _mm_add_epi32(emm2, *(__m128i *)_pi32_1);
+            emm2 = _mm_and_si128(emm2, *(__m128i *)_pi32_inv1);
+            y = _mm_cvtepi32_ps(emm2);
+
+            emm4 = emm2;
+
+            /* get the swap sign flag for the sine */
+            emm0 = _mm_and_si128(emm2, *(__m128i *)_pi32_4);
+            emm0 = _mm_slli_epi32(emm0, 29);
+            __m128 swap_sign_bit_sin = _mm_castsi128_ps(emm0);
+
+            /* get the polynom selection mask for the sine*/
+            emm2 = _mm_and_si128(emm2, *(__m128i *)_pi32_2);
+            emm2 = _mm_cmpeq_epi32(emm2, _mm_setzero_si128());
+            __m128 poly_mask = _mm_castsi128_ps(emm2);
+
+            /* The magic pass: "Extended precision modular arithmetic”
+               x = ((x - y * DP1) - y * DP2) - y * DP3; */
+            xmm1 = *(__m128*)_ps_minus_cephes_DP1;
+            xmm2 = *(__m128*)_ps_minus_cephes_DP2;
+            xmm3 = *(__m128*)_ps_minus_cephes_DP3;
+            xmm1 = _mm_mul_ps(y, xmm1);
+            xmm2 = _mm_mul_ps(y, xmm2);
+            xmm3 = _mm_mul_ps(y, xmm3);
+            x = _mm_add_ps(x, xmm1);
+            x = _mm_add_ps(x, xmm2);
+            x = _mm_add_ps(x, xmm3);
+
+            emm4 = _mm_sub_epi32(emm4, *(__m128i *)_pi32_2);
+            emm4 = _mm_andnot_si128(emm4, *(__m128i *)_pi32_4);
+            emm4 = _mm_slli_epi32(emm4, 29);
+            __m128 sign_bit_cos = _mm_castsi128_ps(emm4);
+
+            sign_bit_sin = _mm_xor_ps(sign_bit_sin, swap_sign_bit_sin);
+
+            /* Evaluate the first polynom  (0 <= x <= Pi/4) */
+            __m128 z = _mm_mul_ps(x,x);
+            y = *(__m128*)_ps_coscof_p0;
+
+            y = _mm_mul_ps(y, z);
+            y = _mm_add_ps(y, *(__m128*)_ps_coscof_p1);
+            y = _mm_mul_ps(y, z);
+            y = _mm_add_ps(y, *(__m128*)_ps_coscof_p2);
+            y = _mm_mul_ps(y, z);
+            y = _mm_mul_ps(y, z);
+            __m128 tmp = _mm_mul_ps(z, *(__m128*)_ps_0p5);
+            y = _mm_sub_ps(y, tmp);
+            y = _mm_add_ps(y, *(__m128*)_ps_1);
+
+            /* Evaluate the second polynom  (Pi/4 <= x <= 0) */
+            __m128 y2 = *(__m128*)_ps_sincof_p0;
+            y2 = _mm_mul_ps(y2, z);
+            y2 = _mm_add_ps(y2, *(__m128*)_ps_sincof_p1);
+            y2 = _mm_mul_ps(y2, z);
+            y2 = _mm_add_ps(y2, *(__m128*)_ps_sincof_p2);
+            y2 = _mm_mul_ps(y2, z);
+            y2 = _mm_mul_ps(y2, x);
+            y2 = _mm_add_ps(y2, x);
+
+            /* select the correct result from the two polynoms */
+            xmm3 = poly_mask;
+            __m128 ysin2 = _mm_and_ps(xmm3, y2);
+            __m128 ysin1 = _mm_andnot_ps(xmm3, y);
+            y2 = _mm_sub_ps(y2,ysin2);
+            y = _mm_sub_ps(y, ysin1);
+
+            xmm1 = _mm_add_ps(ysin1,ysin2);
+            xmm2 = _mm_add_ps(y,y2);
+
+            /* update the sign */
+            sine = _mm_xor_ps(xmm1, sign_bit_sin);
+            cosine = _mm_xor_ps(xmm2, sign_bit_cos);
+
+            /* write the output */
+            aux = _mm_unpacklo_ps(cosine, sine);
+            _mm_store_ps((float*)bPtr, aux);
+            bPtr += 2;
+            aux = _mm_unpackhi_ps(cosine, sine);
+            _mm_store_ps((float*)bPtr, aux);
+            bPtr += 2;
+
+            four_phases_reg = _mm_add_ps(four_phases_reg, four_phases_inc_reg);
+        }
+
+    _phase = phase_inc * (sse_iters * 4);
+    for(number = sse_iters * 4; number < num_points; number++)
+        {
+            *bPtr++ = lv_cmake((float)cos(_phase), (float)sin(_phase) );
+            _phase += phase_inc;
+        }
+}
+
+#endif /* LV_HAVE_SSE2  */
+
+
+#ifdef LV_HAVE_SSE2
+#include <emmintrin.h>
+/* Adapted from http://gruntthepeon.free.fr/ssemath/sse_mathfun.h, original code from Julien Pommier  */
+/* Based on algorithms from the cephes library http://www.netlib.org/cephes/   */
+static inline void volk_gnsssdr_s32f_sincos_32fc_u_sse2(lv_32fc_t* out, const float phase_inc, unsigned int num_points)
+{
+    lv_32fc_t* bPtr = out;
+
+    const unsigned int sse_iters = num_points / 4;
+    unsigned int number = 0;
+    float _phase;
+
+    __m128 sine, cosine, aux, x, four_phases_reg;
+    __m128 xmm1, xmm2, xmm3 = _mm_setzero_ps(), sign_bit_sin, y;
+    __m128i emm0, emm2, emm4;
+
+    /* declare some SSE constants */
+    static const int _ps_inv_sign_mask[4] __attribute__((aligned(16))) = { ~0x80000000, ~0x80000000, ~0x80000000, ~0x80000000 };
+    static const int _ps_sign_mask[4] __attribute__((aligned(16))) = { (int)0x80000000, (int)0x80000000, (int)0x80000000, (int)0x80000000 };
+
+    static const float _ps_cephes_FOPI[4] __attribute__((aligned(16))) = { 1.27323954473516, 1.27323954473516, 1.27323954473516, 1.27323954473516 };
+    static const int _pi32_1[4]  __attribute__((aligned(16))) = { 1, 1, 1, 1 };
+    static const int _pi32_inv1[4]  __attribute__((aligned(16))) = { ~1, ~1, ~1, ~1 };
+    static const int _pi32_2[4]  __attribute__((aligned(16))) = { 2, 2, 2, 2};
+    static const int _pi32_4[4]  __attribute__((aligned(16))) = { 4, 4, 4, 4};
+
+    static const float _ps_minus_cephes_DP1[4] __attribute__((aligned(16))) = { -0.78515625, -0.78515625, -0.78515625, -0.78515625 };
+    static const float _ps_minus_cephes_DP2[4] __attribute__((aligned(16))) = { -2.4187564849853515625e-4, -2.4187564849853515625e-4, -2.4187564849853515625e-4, -2.4187564849853515625e-4 };
+    static const float _ps_minus_cephes_DP3[4] __attribute__((aligned(16))) = { -3.77489497744594108e-8, -3.77489497744594108e-8, -3.77489497744594108e-8, -3.77489497744594108e-8 };
+    static const float _ps_coscof_p0[4] __attribute__((aligned(16))) = { 2.443315711809948E-005, 2.443315711809948E-005, 2.443315711809948E-005, 2.443315711809948E-005 };
+    static const float _ps_coscof_p1[4] __attribute__((aligned(16))) = { -1.388731625493765E-003, -1.388731625493765E-003, -1.388731625493765E-003, -1.388731625493765E-003 };
+    static const float _ps_coscof_p2[4] __attribute__((aligned(16))) = { 4.166664568298827E-002, 4.166664568298827E-002, 4.166664568298827E-002, 4.166664568298827E-002 };
+    static const float _ps_sincof_p0[4] __attribute__((aligned(16))) = { -1.9515295891E-4, -1.9515295891E-4, -1.9515295891E-4, -1.9515295891E-4 };
+    static const float _ps_sincof_p1[4] __attribute__((aligned(16))) = { 8.3321608736E-3, 8.3321608736E-3, 8.3321608736E-3, 8.3321608736E-3 };
+    static const float _ps_sincof_p2[4] __attribute__((aligned(16))) = { -1.6666654611E-1, -1.6666654611E-1, -1.6666654611E-1, -1.6666654611E-1 };
+    static const float _ps_0p5[4] __attribute__((aligned(16))) = { 0.5f, 0.5f, 0.5f, 0.5f };
+    static const float _ps_1[4] __attribute__((aligned(16))) = { 1.0f, 1.0f, 1.0f, 1.0f };
+
+    float four_phases[4] __attribute__((aligned(16))) = { 0.0f, phase_inc, 2 * phase_inc, 3 * phase_inc };
+    float four_phases_inc[4] __attribute__((aligned(16))) = { 4 * phase_inc, 4 * phase_inc, 4 * phase_inc, 4 * phase_inc };
+    four_phases_reg = _mm_load_ps(four_phases);
+    const __m128 four_phases_inc_reg = _mm_load_ps(four_phases_inc);
+
+    for(;number < sse_iters; number++)
+        {
+            x = four_phases_reg;
+
+            sign_bit_sin = x;
+            /* take the absolute value */
+            x = _mm_and_ps(x, *(__m128*)_ps_inv_sign_mask);
+            /* extract the sign bit (upper one) */
+            sign_bit_sin = _mm_and_ps(sign_bit_sin, *(__m128*)_ps_sign_mask);
+
+            /* scale by 4/Pi */
+            y = _mm_mul_ps(x, *(__m128*)_ps_cephes_FOPI);
+
+            /* store the integer part of y in emm2 */
+            emm2 = _mm_cvttps_epi32(y);
+
+            /* j=(j+1) & (~1) (see the cephes sources) */
+            emm2 = _mm_add_epi32(emm2, *(__m128i *)_pi32_1);
+            emm2 = _mm_and_si128(emm2, *(__m128i *)_pi32_inv1);
+            y = _mm_cvtepi32_ps(emm2);
+
+            emm4 = emm2;
+
+            /* get the swap sign flag for the sine */
+            emm0 = _mm_and_si128(emm2, *(__m128i *)_pi32_4);
+            emm0 = _mm_slli_epi32(emm0, 29);
+            __m128 swap_sign_bit_sin = _mm_castsi128_ps(emm0);
+
+            /* get the polynom selection mask for the sine*/
+            emm2 = _mm_and_si128(emm2, *(__m128i *)_pi32_2);
+            emm2 = _mm_cmpeq_epi32(emm2, _mm_setzero_si128());
+            __m128 poly_mask = _mm_castsi128_ps(emm2);
+
+            /* The magic pass: "Extended precision modular arithmetic”
+               x = ((x - y * DP1) - y * DP2) - y * DP3; */
+            xmm1 = *(__m128*)_ps_minus_cephes_DP1;
+            xmm2 = *(__m128*)_ps_minus_cephes_DP2;
+            xmm3 = *(__m128*)_ps_minus_cephes_DP3;
+            xmm1 = _mm_mul_ps(y, xmm1);
+            xmm2 = _mm_mul_ps(y, xmm2);
+            xmm3 = _mm_mul_ps(y, xmm3);
+            x = _mm_add_ps(x, xmm1);
+            x = _mm_add_ps(x, xmm2);
+            x = _mm_add_ps(x, xmm3);
+
+            emm4 = _mm_sub_epi32(emm4, *(__m128i *)_pi32_2);
+            emm4 = _mm_andnot_si128(emm4, *(__m128i *)_pi32_4);
+            emm4 = _mm_slli_epi32(emm4, 29);
+            __m128 sign_bit_cos = _mm_castsi128_ps(emm4);
+
+            sign_bit_sin = _mm_xor_ps(sign_bit_sin, swap_sign_bit_sin);
+
+            /* Evaluate the first polynom  (0 <= x <= Pi/4) */
+            __m128 z = _mm_mul_ps(x,x);
+            y = *(__m128*)_ps_coscof_p0;
+
+            y = _mm_mul_ps(y, z);
+            y = _mm_add_ps(y, *(__m128*)_ps_coscof_p1);
+            y = _mm_mul_ps(y, z);
+            y = _mm_add_ps(y, *(__m128*)_ps_coscof_p2);
+            y = _mm_mul_ps(y, z);
+            y = _mm_mul_ps(y, z);
+            __m128 tmp = _mm_mul_ps(z, *(__m128*)_ps_0p5);
+            y = _mm_sub_ps(y, tmp);
+            y = _mm_add_ps(y, *(__m128*)_ps_1);
+
+            /* Evaluate the second polynom  (Pi/4 <= x <= 0) */
+            __m128 y2 = *(__m128*)_ps_sincof_p0;
+            y2 = _mm_mul_ps(y2, z);
+            y2 = _mm_add_ps(y2, *(__m128*)_ps_sincof_p1);
+            y2 = _mm_mul_ps(y2, z);
+            y2 = _mm_add_ps(y2, *(__m128*)_ps_sincof_p2);
+            y2 = _mm_mul_ps(y2, z);
+            y2 = _mm_mul_ps(y2, x);
+            y2 = _mm_add_ps(y2, x);
+
+            /* select the correct result from the two polynoms */
+            xmm3 = poly_mask;
+            __m128 ysin2 = _mm_and_ps(xmm3, y2);
+            __m128 ysin1 = _mm_andnot_ps(xmm3, y);
+            y2 = _mm_sub_ps(y2,ysin2);
+            y = _mm_sub_ps(y, ysin1);
+
+            xmm1 = _mm_add_ps(ysin1,ysin2);
+            xmm2 = _mm_add_ps(y,y2);
+
+            /* update the sign */
+            sine = _mm_xor_ps(xmm1, sign_bit_sin);
+            cosine = _mm_xor_ps(xmm2, sign_bit_cos);
+
+            /* write the output */
+            aux = _mm_unpacklo_ps(cosine, sine);
+            _mm_storeu_ps((float*)bPtr, aux);
+            bPtr += 2;
+            aux = _mm_unpackhi_ps(cosine, sine);
+            _mm_storeu_ps((float*)bPtr, aux);
+            bPtr += 2;
+
+            four_phases_reg = _mm_add_ps(four_phases_reg, four_phases_inc_reg);
+        }
+
+    _phase = phase_inc * (sse_iters * 4);
+    for(number = sse_iters * 4; number < num_points; number++)
+        {
+            *bPtr++ = lv_cmake((float)cos(_phase), (float)sin(_phase) );
+            _phase += phase_inc;
+        }
+}
+
+#endif /* LV_HAVE_SSE2  */
+
+#ifdef LV_HAVE_GENERIC
+
+static inline void volk_gnsssdr_s32f_sincos_32fc_generic(lv_32fc_t* out, const float phase_inc, unsigned int num_points)
+{
+    float _phase = 0.0;
+    for(unsigned int i = 0; i < num_points; i++)
+        {
+            *out++ = lv_cmake((float)cos(_phase), (float)sin(_phase) );
+            _phase += phase_inc;
+        }
+}
+
+#endif /* LV_HAVE_GENERIC  */
+
+
+#ifdef LV_HAVE_GENERIC
+#include <volk_gnsssdr/volk_gnsssdr_sine_table.h>
+#include <stdint.h>
+static inline void volk_gnsssdr_s32f_sincos_32fc_generic_fxpt(lv_32fc_t* out, const float phase_inc, unsigned int num_points)
+{
+    float _in, s, c;
+    int32_t x, sin_index, cos_index, d;
+    const float PI = 3.14159265358979323846;
+    const float TWO_TO_THE_31_DIV_PI = 2147483648.0 / PI;
+    const float TWO_PI = PI * 2;
+    const int32_t bitlength = 32;
+    const int32_t Nbits = 10;
+    const int32_t diffbits = bitlength - Nbits;
+    uint32_t ux;
+    float _phase = 0.0;
+    for(unsigned int i = 0; i < num_points; i++)
+        {
+            _in = _phase;
+            d = (int32_t)floor(_in / TWO_PI + 0.5);
+            _in -= d * TWO_PI;
+            x = (int32_t) ((float)_in * TWO_TO_THE_31_DIV_PI);
+
+            ux = x;
+            sin_index = ux >> diffbits;
+            s = sine_table_10bits[sin_index][0] * (ux >> 1) + sine_table_10bits[sin_index][1];
+
+            ux = x + 0x40000000;
+            cos_index = ux >> diffbits;
+            c = sine_table_10bits[cos_index][0] * (ux >> 1) + sine_table_10bits[cos_index][1];
+
+            *out++ = lv_cmake((float)c, (float)s );
+            _phase += phase_inc;
+        }
+}
+
+#endif /* LV_HAVE_GENERIC  */
+
+
+#ifdef LV_HAVE_NEON
+#include <arm_neon.h>
+/* Adapted from http://gruntthepeon.free.fr/ssemath/neon_mathfun.h, original code from Julien Pommier  */
+/* Based on algorithms from the cephes library http://www.netlib.org/cephes/   */
+static inline void volk_gnsssdr_s32f_sincos_32fc_neon(lv_32fc_t* out, const float phase_inc, unsigned int num_points)
+{
+    lv_32fc_t* bPtr = out;
+    const unsigned int neon_iters = num_points / 4;
+
+    __VOLK_ATTR_ALIGNED(16) float32_t four_phases[4] = { 0.0f , phase_inc, 2 * phase_inc,  3 * phase_inc };
+    float four_inc = 4 * phase_inc;
+    __VOLK_ATTR_ALIGNED(16) float32_t four_phases_inc[4] = { four_inc, four_inc, four_inc, four_inc };
+
+    float32x4_t four_phases_reg = vld1q_f32(four_phases);
+    float32x4_t four_phases_inc_reg = vld1q_f32(four_phases_inc);
+
+    const float32_t c_minus_cephes_DP1 = -0.78515625;
+    const float32_t c_minus_cephes_DP2 = -2.4187564849853515625e-4;
+    const float32_t c_minus_cephes_DP3 = -3.77489497744594108e-8;
+    const float32_t c_sincof_p0 = -1.9515295891E-4;
+    const float32_t c_sincof_p1 = 8.3321608736E-3;
+    const float32_t c_sincof_p2 = -1.6666654611E-1;
+    const float32_t c_coscof_p0 = 2.443315711809948E-005;
+    const float32_t c_coscof_p1 = -1.388731625493765E-003;
+    const float32_t c_coscof_p2 = 4.166664568298827E-002;
+    const float32_t c_cephes_FOPI = 1.27323954473516;
+
+    unsigned int number = 0;
+    float _phase;
+
+    float32x4_t x, xmm1, xmm2, xmm3, y, y1, y2, ys, yc, z;
+    float32x4x2_t result;
+
+    uint32x4_t emm2, poly_mask, sign_mask_sin, sign_mask_cos;
+
+    for(;number < neon_iters; number++)
+        {
+            x = four_phases_reg;
+            __builtin_prefetch(aPtr + 8);
+
+            sign_mask_sin = vcltq_f32(x, vdupq_n_f32(0));
+            x = vabsq_f32(x);
+
+            /* scale by 4/Pi */
+            y = vmulq_f32(x, vdupq_n_f32(c_cephes_FOPI));
+
+            /* store the integer part of y in mm0 */
+            emm2 = vcvtq_u32_f32(y);
+            /* j=(j+1) & (~1) (see the cephes sources) */
+            emm2 = vaddq_u32(emm2, vdupq_n_u32(1));
+            emm2 = vandq_u32(emm2, vdupq_n_u32(~1));
+            y = vcvtq_f32_u32(emm2);
+
+            /* get the polynom selection mask
+                    there is one polynom for 0 <= x <= Pi/4
+                    and another one for Pi/4<x<=Pi/2
+
+                    Both branches will be computed.
+             */
+            poly_mask = vtstq_u32(emm2, vdupq_n_u32(2));
+
+            /* The magic pass: "Extended precision modular arithmetic"
+                    x = ((x - y * DP1) - y * DP2) - y * DP3; */
+            xmm1 = vmulq_n_f32(y, c_minus_cephes_DP1);
+            xmm2 = vmulq_n_f32(y, c_minus_cephes_DP2);
+            xmm3 = vmulq_n_f32(y, c_minus_cephes_DP3);
+            x = vaddq_f32(x, xmm1);
+            x = vaddq_f32(x, xmm2);
+            x = vaddq_f32(x, xmm3);
+
+            sign_mask_sin = veorq_u32(sign_mask_sin, vtstq_u32(emm2, vdupq_n_u32(4)));
+            sign_mask_cos = vtstq_u32(vsubq_u32(emm2, vdupq_n_u32(2)), vdupq_n_u32(4));
+
+            /* Evaluate the first polynom  (0 <= x <= Pi/4) in y1,
+                    and the second polynom      (Pi/4 <= x <= 0) in y2 */
+            z = vmulq_f32(x,x);
+
+            y1 = vmulq_n_f32(z, c_coscof_p0);
+            y2 = vmulq_n_f32(z, c_sincof_p0);
+            y1 = vaddq_f32(y1, vdupq_n_f32(c_coscof_p1));
+            y2 = vaddq_f32(y2, vdupq_n_f32(c_sincof_p1));
+            y1 = vmulq_f32(y1, z);
+            y2 = vmulq_f32(y2, z);
+            y1 = vaddq_f32(y1, vdupq_n_f32(c_coscof_p2));
+            y2 = vaddq_f32(y2, vdupq_n_f32(c_sincof_p2));
+            y1 = vmulq_f32(y1, z);
+            y2 = vmulq_f32(y2, z);
+            y1 = vmulq_f32(y1, z);
+            y2 = vmulq_f32(y2, x);
+            y1 = vsubq_f32(y1, vmulq_f32(z, vdupq_n_f32(0.5f)));
+            y2 = vaddq_f32(y2, x);
+            y1 = vaddq_f32(y1, vdupq_n_f32(1));
+
+            /* select the correct result from the two polynoms */
+            ys = vbslq_f32(poly_mask, y1, y2);
+            yc = vbslq_f32(poly_mask, y2, y1);
+            result.val[1] = vbslq_f32(sign_mask_sin, vnegq_f32(ys), ys);
+            result.val[0] = vbslq_f32(sign_mask_cos, yc, vnegq_f32(yc));
+
+            vst2q_f32((float32_t*)bPtr, result);
+            bPtr += 4;
+
+            four_phases_reg = vaddq_f32(four_phases_reg, four_phases_inc_reg);
+        }
+
+    _phase = phase_inc * (neon_iters * 4);
+    for(number = neon_iters * 4; number < num_points; number++)
+        {
+            *bPtr++ = lv_cmake((float)cos(_phase), (float)sin(_phase) );
+            _phase += phase_inc;
+        }
+}
+
+#endif /* LV_HAVE_NEON  */
+
+#endif /* INCLUDED_volk_gnsssdr_s32f_sincos_32fc_H */

src/algorithms/libs/volk_gnsssdr_module/volk_gnsssdr/lib/kernel_tests.h

@@ -61,6 +61,8 @@ std::vector<volk_gnsssdr_test_case_t> init_test_list(volk_gnsssdr_test_params_t
     // ... or more tolerance *****  ADDED BY GNSS-SDR
     volk_gnsssdr_test_params_t test_params_int16 = volk_gnsssdr_test_params_t(16, test_params.scalar(),
                 test_params.vlen(), test_params.iter(), test_params.benchmark_mode(), test_params.kernel_regex());
+    volk_gnsssdr_test_params_t test_params_inacc2 = volk_gnsssdr_test_params_t(2e-1, test_params.scalar(),
+                test_params.vlen(), test_params.iter(), test_params.benchmark_mode(), test_params.kernel_regex());
 
     std::vector<volk_gnsssdr_test_case_t> test_cases = boost::assign::list_of
 
@@ -76,6 +78,7 @@ std::vector<volk_gnsssdr_test_case_t> init_test_list(volk_gnsssdr_test_params_t
         (VOLK_INIT_TEST(volk_gnsssdr_8u_x2_multiply_8u, test_params_more_iters))
         (VOLK_INIT_TEST(volk_gnsssdr_64f_accumulator_64f, test_params))
         (VOLK_INIT_TEST(volk_gnsssdr_32f_sincos_32fc, test_params_inacc))
+        (VOLK_INIT_TEST(volk_gnsssdr_s32f_sincos_32fc, test_params_inacc2))
         (VOLK_INIT_TEST(volk_gnsssdr_32fc_convert_8ic, test_params))
         (VOLK_INIT_TEST(volk_gnsssdr_32fc_convert_16ic, test_params_more_iters))
         (VOLK_INIT_TEST(volk_gnsssdr_16ic_x2_dot_prod_16ic, test_params))