mirror of
https://github.com/gnss-sdr/gnss-sdr
synced 2024-06-24 22:13:15 +00:00
Add kernel volk_gnsssdr_32f_index_max_32u
This is a copy of VOLK's volk_32f_index_max32u. It allows acquisition with rata rates higher than 16 Msps when working with 4ms code periods, while not requiring a very recent version of VOLK
This commit is contained in:
Carles Fernandez
parent
8d8249247f
commit
7e0e99dc4b
|
|
@ -0,0 +1,221 @@
|
|||
/*!
|
||||
* \file volk_gnsssdr_32f_index_max_32u.h
|
||||
* \brief VOLK_GNSSSDR kernel: Finds and returns the index which contains the maximum value in the given vector.
|
||||
*
|
||||
*
|
||||
* -------------------------------------------------------------------------
|
||||
*
|
||||
* 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_32f_index_max_32u.h
|
||||
*
|
||||
* \b Overview
|
||||
*
|
||||
* Finds and returns the index which contains the maximum value in the given vector.
|
||||
*
|
||||
* <b>Dispatcher Prototype</b>
|
||||
* \code
|
||||
* volk_gnsssdr_32f_index_max_32u(uint32_t* target, const float* src0, uint32_t num_points)
|
||||
* \endcode
|
||||
*
|
||||
* \b Inputs
|
||||
* \li src0: The input vector of floats.
|
||||
* \li num_points: The number of data points.
|
||||
*
|
||||
* \b Outputs
|
||||
* \li target: The index of the maximum value in the input buffer.
|
||||
*
|
||||
*/
|
||||
|
||||
|
||||
#ifndef INCLUDED_volk_gnsssdr_32f_index_max_32u_H
|
||||
#define INCLUDED_volk_gnsssdr_32f_index_max_32u_H
|
||||
|
||||
#include <volk_gnsssdr/volk_gnsssdr_common.h>
|
||||
#include <inttypes.h>
|
||||
#include <stdio.h>
|
||||
|
||||
#ifdef LV_HAVE_SSE4_1
|
||||
#include<smmintrin.h>
|
||||
|
||||
static inline void volk_gnsssdr_32f_index_max_32u_a_sse4_1(uint32_t* target, const float* src0, uint32_t num_points)
|
||||
{
|
||||
if(num_points > 0)
|
||||
{
|
||||
uint32_t number = 0;
|
||||
const uint32_t quarterPoints = num_points / 4;
|
||||
|
||||
float* inputPtr = (float*)src0;
|
||||
|
||||
__m128 indexIncrementValues = _mm_set1_ps(4);
|
||||
__m128 currentIndexes = _mm_set_ps(-1,-2,-3,-4);
|
||||
|
||||
float max = src0[0];
|
||||
float index = 0;
|
||||
__m128 maxValues = _mm_set1_ps(max);
|
||||
__m128 maxValuesIndex = _mm_setzero_ps();
|
||||
__m128 compareResults;
|
||||
__m128 currentValues;
|
||||
|
||||
__VOLK_ATTR_ALIGNED(16) float maxValuesBuffer[4];
|
||||
__VOLK_ATTR_ALIGNED(16) float maxIndexesBuffer[4];
|
||||
|
||||
for(;number < quarterPoints; number++)
|
||||
{
|
||||
|
||||
currentValues = _mm_load_ps(inputPtr); inputPtr += 4;
|
||||
currentIndexes = _mm_add_ps(currentIndexes, indexIncrementValues);
|
||||
|
||||
compareResults = _mm_cmpgt_ps(maxValues, currentValues);
|
||||
|
||||
maxValuesIndex = _mm_blendv_ps(currentIndexes, maxValuesIndex, compareResults);
|
||||
maxValues = _mm_blendv_ps(currentValues, maxValues, compareResults);
|
||||
}
|
||||
|
||||
// Calculate the largest value from the remaining 4 points
|
||||
_mm_store_ps(maxValuesBuffer, maxValues);
|
||||
_mm_store_ps(maxIndexesBuffer, maxValuesIndex);
|
||||
|
||||
for(number = 0; number < 4; number++)
|
||||
{
|
||||
if(maxValuesBuffer[number] > max)
|
||||
{
|
||||
index = maxIndexesBuffer[number];
|
||||
max = maxValuesBuffer[number];
|
||||
}
|
||||
}
|
||||
|
||||
number = quarterPoints * 4;
|
||||
for(;number < num_points; number++)
|
||||
{
|
||||
if(src0[number] > max)
|
||||
{
|
||||
index = number;
|
||||
max = src0[number];
|
||||
}
|
||||
}
|
||||
target[0] = (uint32_t)index;
|
||||
}
|
||||
}
|
||||
|
||||
#endif /*LV_HAVE_SSE4_1*/
|
||||
|
||||
|
||||
#ifdef LV_HAVE_SSE
|
||||
|
||||
#include<xmmintrin.h>
|
||||
|
||||
static inline void volk_gnsssdr_32f_index_max_32u_a_sse(uint32_t* target, const float* src0, uint32_t num_points)
|
||||
{
|
||||
if(num_points > 0)
|
||||
{
|
||||
uint32_t number = 0;
|
||||
const uint32_t quarterPoints = num_points / 4;
|
||||
|
||||
float* inputPtr = (float*)src0;
|
||||
|
||||
__m128 indexIncrementValues = _mm_set1_ps(4);
|
||||
__m128 currentIndexes = _mm_set_ps(-1,-2,-3,-4);
|
||||
|
||||
float max = src0[0];
|
||||
float index = 0;
|
||||
__m128 maxValues = _mm_set1_ps(max);
|
||||
__m128 maxValuesIndex = _mm_setzero_ps();
|
||||
__m128 compareResults;
|
||||
__m128 currentValues;
|
||||
|
||||
__VOLK_ATTR_ALIGNED(16) float maxValuesBuffer[4];
|
||||
__VOLK_ATTR_ALIGNED(16) float maxIndexesBuffer[4];
|
||||
|
||||
for(;number < quarterPoints; number++)
|
||||
{
|
||||
currentValues = _mm_load_ps(inputPtr); inputPtr += 4;
|
||||
currentIndexes = _mm_add_ps(currentIndexes, indexIncrementValues);
|
||||
|
||||
compareResults = _mm_cmpgt_ps(maxValues, currentValues);
|
||||
|
||||
maxValuesIndex = _mm_or_ps(_mm_and_ps(compareResults, maxValuesIndex) , _mm_andnot_ps(compareResults, currentIndexes));
|
||||
|
||||
maxValues = _mm_or_ps(_mm_and_ps(compareResults, maxValues) , _mm_andnot_ps(compareResults, currentValues));
|
||||
}
|
||||
|
||||
// Calculate the largest value from the remaining 4 points
|
||||
_mm_store_ps(maxValuesBuffer, maxValues);
|
||||
_mm_store_ps(maxIndexesBuffer, maxValuesIndex);
|
||||
|
||||
for(number = 0; number < 4; number++)
|
||||
{
|
||||
if(maxValuesBuffer[number] > max)
|
||||
{
|
||||
index = maxIndexesBuffer[number];
|
||||
max = maxValuesBuffer[number];
|
||||
}
|
||||
}
|
||||
|
||||
number = quarterPoints * 4;
|
||||
for(;number < num_points; number++)
|
||||
{
|
||||
if(src0[number] > max)
|
||||
{
|
||||
index = number;
|
||||
max = src0[number];
|
||||
}
|
||||
}
|
||||
target[0] = (uint32_t)index;
|
||||
}
|
||||
}
|
||||
|
||||
#endif /*LV_HAVE_SSE*/
|
||||
|
||||
|
||||
#ifdef LV_HAVE_GENERIC
|
||||
|
||||
static inline void volk_gnsssdr_32f_index_max_32u_generic(uint32_t* target, const float* src0, uint32_t num_points)
|
||||
{
|
||||
if(num_points > 0)
|
||||
{
|
||||
float max = src0[0];
|
||||
uint32_t index = 0;
|
||||
|
||||
uint32_t i = 1;
|
||||
|
||||
for(; i < num_points; ++i)
|
||||
{
|
||||
if(src0[i] > max)
|
||||
{
|
||||
index = i;
|
||||
max = src0[i];
|
||||
}
|
||||
}
|
||||
target[0] = index;
|
||||
}
|
||||
}
|
||||
|
||||
#endif /*LV_HAVE_GENERIC*/
|
||||
|
||||
|
||||
#endif /*INCLUDED_volk_gnsssdr_32f_index_max_32u_H*/
|
||||
|
|
@ -78,6 +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_32f_index_max_32u, test_params))
|
||||
(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))
|
||||
|
|
|
|||
Loading…
Reference in New Issue
Block a user