volk_32f_x2_subtract_32f.h

Go to the documentation of this file.

/* -*- c++ -*- */
/*
 * Copyright 2012, 2014 Free Software Foundation, Inc.
 *
 * This file is part of VOLK
 *
 * SPDX-License-Identifier: LGPL-3.0-or-later
 */
 
#ifndef INCLUDED_volk_32f_x2_subtract_32f_a_H
#define INCLUDED_volk_32f_x2_subtract_32f_a_H
 
#include <inttypes.h>
#include <stdio.h>
 
#ifdef LV_HAVE_AVX512F
#include <immintrin.h>
 
static inline void volk_32f_x2_subtract_32f_a_avx512f(float* cVector,
                                                      const float* aVector,
                                                      const float* bVector,
                                                      unsigned int num_points)
{
    unsigned int number = 0;
    const unsigned int sixteenthPoints = num_points / 16;
 
    float* cPtr = cVector;
    const float* aPtr = aVector;
    const float* bPtr = bVector;
 
    __m512 aVal, bVal, cVal;
    for (; number < sixteenthPoints; number++) {
 
        aVal = _mm512_load_ps(aPtr);
        bVal = _mm512_load_ps(bPtr);
 
        cVal = _mm512_sub_ps(aVal, bVal);
 
        _mm512_store_ps(cPtr, cVal); // Store the results back into the C container
 
        aPtr += 16;
        bPtr += 16;
        cPtr += 16;
    }
 
    number = sixteenthPoints * 16;
    for (; number < num_points; number++) {
        *cPtr++ = (*aPtr++) - (*bPtr++);
    }
}
#endif /* LV_HAVE_AVX512F */
 
#ifdef LV_HAVE_AVX
#include <immintrin.h>
 
static inline void volk_32f_x2_subtract_32f_a_avx(float* cVector,
                                                  const float* aVector,
                                                  const float* bVector,
                                                  unsigned int num_points)
{
    unsigned int number = 0;
    const unsigned int eighthPoints = num_points / 8;
 
    float* cPtr = cVector;
    const float* aPtr = aVector;
    const float* bPtr = bVector;
 
    __m256 aVal, bVal, cVal;
    for (; number < eighthPoints; number++) {
 
        aVal = _mm256_load_ps(aPtr);
        bVal = _mm256_load_ps(bPtr);
 
        cVal = _mm256_sub_ps(aVal, bVal);
 
        _mm256_store_ps(cPtr, cVal); // Store the results back into the C container
 
        aPtr += 8;
        bPtr += 8;
        cPtr += 8;
    }
 
    number = eighthPoints * 8;
    for (; number < num_points; number++) {
        *cPtr++ = (*aPtr++) - (*bPtr++);
    }
}
#endif /* LV_HAVE_AVX */
 
#ifdef LV_HAVE_SSE
#include <xmmintrin.h>
 
static inline void volk_32f_x2_subtract_32f_a_sse(float* cVector,
                                                  const float* aVector,
                                                  const float* bVector,
                                                  unsigned int num_points)
{
    unsigned int number = 0;
    const unsigned int quarterPoints = num_points / 4;
 
    float* cPtr = cVector;
    const float* aPtr = aVector;
    const float* bPtr = bVector;
 
    __m128 aVal, bVal, cVal;
    for (; number < quarterPoints; number++) {
 
        aVal = _mm_load_ps(aPtr);
        bVal = _mm_load_ps(bPtr);
 
        cVal = _mm_sub_ps(aVal, bVal);
 
        _mm_store_ps(cPtr, cVal); // Store the results back into the C container
 
        aPtr += 4;
        bPtr += 4;
        cPtr += 4;
    }
 
    number = quarterPoints * 4;
    for (; number < num_points; number++) {
        *cPtr++ = (*aPtr++) - (*bPtr++);
    }
}
#endif /* LV_HAVE_SSE */
 
 
#ifdef LV_HAVE_GENERIC
 
static inline void volk_32f_x2_subtract_32f_generic(float* cVector,
                                                    const float* aVector,
                                                    const float* bVector,
                                                    unsigned int num_points)
{
    float* cPtr = cVector;
    const float* aPtr = aVector;
    const float* bPtr = bVector;
    unsigned int number = 0;
 
    for (number = 0; number < num_points; number++) {
        *cPtr++ = (*aPtr++) - (*bPtr++);
    }
}
#endif /* LV_HAVE_GENERIC */
 
 
#ifdef LV_HAVE_NEON
#include <arm_neon.h>
 
static inline void volk_32f_x2_subtract_32f_neon(float* cVector,
                                                 const float* aVector,
                                                 const float* bVector,
                                                 unsigned int num_points)
{
    float* cPtr = cVector;
    const float* aPtr = aVector;
    const float* bPtr = bVector;
    unsigned int number = 0;
    unsigned int quarter_points = num_points / 4;
 
    float32x4_t a_vec, b_vec, c_vec;
 
    for (number = 0; number < quarter_points; number++) {
        a_vec = vld1q_f32(aPtr);
        b_vec = vld1q_f32(bPtr);
        c_vec = vsubq_f32(a_vec, b_vec);
        vst1q_f32(cPtr, c_vec);
        aPtr += 4;
        bPtr += 4;
        cPtr += 4;
    }
 
    for (number = quarter_points * 4; number < num_points; number++) {
        *cPtr++ = (*aPtr++) - (*bPtr++);
    }
}
#endif /* LV_HAVE_NEON */
 
 
#ifdef LV_HAVE_ORC
extern void volk_32f_x2_subtract_32f_a_orc_impl(float* cVector,
                                                const float* aVector,
                                                const float* bVector,
                                                unsigned int num_points);
 
static inline void volk_32f_x2_subtract_32f_u_orc(float* cVector,
                                                  const float* aVector,
                                                  const float* bVector,
                                                  unsigned int num_points)
{
    volk_32f_x2_subtract_32f_a_orc_impl(cVector, aVector, bVector, num_points);
}
#endif /* LV_HAVE_ORC */
 
 
#endif /* INCLUDED_volk_32f_x2_subtract_32f_a_H */
 
 
#ifndef INCLUDED_volk_32f_x2_subtract_32f_u_H
#define INCLUDED_volk_32f_x2_subtract_32f_u_H
 
#include <inttypes.h>
#include <stdio.h>
 
#ifdef LV_HAVE_AVX512F
#include <immintrin.h>
 
static inline void volk_32f_x2_subtract_32f_u_avx512f(float* cVector,
                                                      const float* aVector,
                                                      const float* bVector,
                                                      unsigned int num_points)
{
    unsigned int number = 0;
    const unsigned int sixteenthPoints = num_points / 16;
 
    float* cPtr = cVector;
    const float* aPtr = aVector;
    const float* bPtr = bVector;
 
    __m512 aVal, bVal, cVal;
    for (; number < sixteenthPoints; number++) {
 
        aVal = _mm512_loadu_ps(aPtr);
        bVal = _mm512_loadu_ps(bPtr);
 
        cVal = _mm512_sub_ps(aVal, bVal);
 
        _mm512_storeu_ps(cPtr, cVal); // Store the results back into the C container
 
        aPtr += 16;
        bPtr += 16;
        cPtr += 16;
    }
 
    number = sixteenthPoints * 16;
    for (; number < num_points; number++) {
        *cPtr++ = (*aPtr++) - (*bPtr++);
    }
}
#endif /* LV_HAVE_AVX512F */
 
 
#ifdef LV_HAVE_AVX
#include <immintrin.h>
 
static inline void volk_32f_x2_subtract_32f_u_avx(float* cVector,
                                                  const float* aVector,
                                                  const float* bVector,
                                                  unsigned int num_points)
{
    unsigned int number = 0;
    const unsigned int eighthPoints = num_points / 8;
 
    float* cPtr = cVector;
    const float* aPtr = aVector;
    const float* bPtr = bVector;
 
    __m256 aVal, bVal, cVal;
    for (; number < eighthPoints; number++) {
 
        aVal = _mm256_loadu_ps(aPtr);
        bVal = _mm256_loadu_ps(bPtr);
 
        cVal = _mm256_sub_ps(aVal, bVal);
 
        _mm256_storeu_ps(cPtr, cVal); // Store the results back into the C container
 
        aPtr += 8;
        bPtr += 8;
        cPtr += 8;
    }
 
    number = eighthPoints * 8;
    for (; number < num_points; number++) {
        *cPtr++ = (*aPtr++) - (*bPtr++);
    }
}
#endif /* LV_HAVE_AVX */
 
#endif /* INCLUDED_volk_32f_x2_subtract_32f_u_H */