volk_32fc_s32f_magnitude_16i.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_32fc_s32f_magnitude_16i_a_H
#define INCLUDED_volk_32fc_s32f_magnitude_16i_a_H
 
#include <inttypes.h>
#include <math.h>
#include <stdio.h>
#include <volk/volk_common.h>
 
#ifdef LV_HAVE_GENERIC
 
static inline void volk_32fc_s32f_magnitude_16i_generic(int16_t* magnitudeVector,
                                                        const lv_32fc_t* complexVector,
                                                        const float scalar,
                                                        unsigned int num_points)
{
    const float* complexVectorPtr = (float*)complexVector;
    int16_t* magnitudeVectorPtr = magnitudeVector;
    unsigned int number = 0;
    for (number = 0; number < num_points; number++) {
        float real = *complexVectorPtr++;
        float imag = *complexVectorPtr++;
        *magnitudeVectorPtr++ =
            (int16_t)rintf(scalar * sqrtf((real * real) + (imag * imag)));
    }
}
#endif /* LV_HAVE_GENERIC */
 
#ifdef LV_HAVE_AVX2
#include <immintrin.h>
 
static inline void volk_32fc_s32f_magnitude_16i_a_avx2(int16_t* magnitudeVector,
                                                       const lv_32fc_t* complexVector,
                                                       const float scalar,
                                                       unsigned int num_points)
{
    unsigned int number = 0;
    const unsigned int eighthPoints = num_points / 8;
 
    const float* complexVectorPtr = (const float*)complexVector;
    int16_t* magnitudeVectorPtr = magnitudeVector;
 
    __m256 vScalar = _mm256_set1_ps(scalar);
    __m256i idx = _mm256_set_epi32(0, 0, 0, 0, 5, 1, 4, 0);
    __m256 cplxValue1, cplxValue2, result;
    __m256i resultInt;
    __m128i resultShort;
 
    for (; number < eighthPoints; number++) {
        cplxValue1 = _mm256_load_ps(complexVectorPtr);
        complexVectorPtr += 8;
 
        cplxValue2 = _mm256_load_ps(complexVectorPtr);
        complexVectorPtr += 8;
 
        cplxValue1 = _mm256_mul_ps(cplxValue1, cplxValue1); // Square the values
        cplxValue2 = _mm256_mul_ps(cplxValue2, cplxValue2); // Square the Values
 
        result = _mm256_hadd_ps(cplxValue1, cplxValue2); // Add the I2 and Q2 values
 
        result = _mm256_sqrt_ps(result);
 
        result = _mm256_mul_ps(result, vScalar);
 
        resultInt = _mm256_cvtps_epi32(result);
        resultInt = _mm256_packs_epi32(resultInt, resultInt);
        resultInt = _mm256_permutevar8x32_epi32(
            resultInt, idx); // permute to compensate for shuffling in hadd and packs
        resultShort = _mm256_extracti128_si256(resultInt, 0);
        _mm_store_si128((__m128i*)magnitudeVectorPtr, resultShort);
        magnitudeVectorPtr += 8;
    }
 
    number = eighthPoints * 8;
    volk_32fc_s32f_magnitude_16i_generic(
        magnitudeVector + number, complexVector + number, scalar, num_points - number);
}
#endif /* LV_HAVE_AVX2 */
 
#ifdef LV_HAVE_SSE3
#include <pmmintrin.h>
 
static inline void volk_32fc_s32f_magnitude_16i_a_sse3(int16_t* magnitudeVector,
                                                       const lv_32fc_t* complexVector,
                                                       const float scalar,
                                                       unsigned int num_points)
{
    unsigned int number = 0;
    const unsigned int quarterPoints = num_points / 4;
 
    const float* complexVectorPtr = (const float*)complexVector;
    int16_t* magnitudeVectorPtr = magnitudeVector;
 
    __m128 vScalar = _mm_set_ps1(scalar);
 
    __m128 cplxValue1, cplxValue2, result;
 
    __VOLK_ATTR_ALIGNED(16) float floatBuffer[4];
 
    for (; number < quarterPoints; number++) {
        cplxValue1 = _mm_load_ps(complexVectorPtr);
        complexVectorPtr += 4;
 
        cplxValue2 = _mm_load_ps(complexVectorPtr);
        complexVectorPtr += 4;
 
        cplxValue1 = _mm_mul_ps(cplxValue1, cplxValue1); // Square the values
        cplxValue2 = _mm_mul_ps(cplxValue2, cplxValue2); // Square the Values
 
        result = _mm_hadd_ps(cplxValue1, cplxValue2); // Add the I2 and Q2 values
 
        result = _mm_sqrt_ps(result);
 
        result = _mm_mul_ps(result, vScalar);
 
        _mm_store_ps(floatBuffer, result);
        *magnitudeVectorPtr++ = (int16_t)rintf(floatBuffer[0]);
        *magnitudeVectorPtr++ = (int16_t)rintf(floatBuffer[1]);
        *magnitudeVectorPtr++ = (int16_t)rintf(floatBuffer[2]);
        *magnitudeVectorPtr++ = (int16_t)rintf(floatBuffer[3]);
    }
 
    number = quarterPoints * 4;
    volk_32fc_s32f_magnitude_16i_generic(
        magnitudeVector + number, complexVector + number, scalar, num_points - number);
}
#endif /* LV_HAVE_SSE3 */
 
 
#ifdef LV_HAVE_SSE
#include <xmmintrin.h>
 
static inline void volk_32fc_s32f_magnitude_16i_a_sse(int16_t* magnitudeVector,
                                                      const lv_32fc_t* complexVector,
                                                      const float scalar,
                                                      unsigned int num_points)
{
    unsigned int number = 0;
    const unsigned int quarterPoints = num_points / 4;
 
    const float* complexVectorPtr = (const float*)complexVector;
    int16_t* magnitudeVectorPtr = magnitudeVector;
 
    __m128 vScalar = _mm_set_ps1(scalar);
 
    __m128 cplxValue1, cplxValue2, result;
    __m128 iValue, qValue;
 
    __VOLK_ATTR_ALIGNED(16) float floatBuffer[4];
 
    for (; number < quarterPoints; number++) {
        cplxValue1 = _mm_load_ps(complexVectorPtr);
        complexVectorPtr += 4;
 
        cplxValue2 = _mm_load_ps(complexVectorPtr);
        complexVectorPtr += 4;
 
        // Arrange in i1i2i3i4 format
        iValue = _mm_shuffle_ps(cplxValue1, cplxValue2, _MM_SHUFFLE(2, 0, 2, 0));
        // Arrange in q1q2q3q4 format
        qValue = _mm_shuffle_ps(cplxValue1, cplxValue2, _MM_SHUFFLE(3, 1, 3, 1));
 
        __m128 iValue2 = _mm_mul_ps(iValue, iValue); // Square the I values
        __m128 qValue2 = _mm_mul_ps(qValue, qValue); // Square the Q Values
 
        result = _mm_add_ps(iValue2, qValue2); // Add the I2 and Q2 values
 
        result = _mm_sqrt_ps(result);
 
        result = _mm_mul_ps(result, vScalar);
 
        _mm_store_ps(floatBuffer, result);
        *magnitudeVectorPtr++ = (int16_t)rintf(floatBuffer[0]);
        *magnitudeVectorPtr++ = (int16_t)rintf(floatBuffer[1]);
        *magnitudeVectorPtr++ = (int16_t)rintf(floatBuffer[2]);
        *magnitudeVectorPtr++ = (int16_t)rintf(floatBuffer[3]);
    }
 
    number = quarterPoints * 4;
    volk_32fc_s32f_magnitude_16i_generic(
        magnitudeVector + number, complexVector + number, scalar, num_points - number);
}
#endif /* LV_HAVE_SSE */
 
 
#endif /* INCLUDED_volk_32fc_s32f_magnitude_16i_a_H */
 
#ifndef INCLUDED_volk_32fc_s32f_magnitude_16i_u_H
#define INCLUDED_volk_32fc_s32f_magnitude_16i_u_H
 
#include <inttypes.h>
#include <math.h>
#include <stdio.h>
#include <volk/volk_common.h>
 
#ifdef LV_HAVE_AVX2
#include <immintrin.h>
 
static inline void volk_32fc_s32f_magnitude_16i_u_avx2(int16_t* magnitudeVector,
                                                       const lv_32fc_t* complexVector,
                                                       const float scalar,
                                                       unsigned int num_points)
{
    unsigned int number = 0;
    const unsigned int eighthPoints = num_points / 8;
 
    const float* complexVectorPtr = (const float*)complexVector;
    int16_t* magnitudeVectorPtr = magnitudeVector;
 
    __m256 vScalar = _mm256_set1_ps(scalar);
    __m256i idx = _mm256_set_epi32(0, 0, 0, 0, 5, 1, 4, 0);
    __m256 cplxValue1, cplxValue2, result;
    __m256i resultInt;
    __m128i resultShort;
 
    for (; number < eighthPoints; number++) {
        cplxValue1 = _mm256_loadu_ps(complexVectorPtr);
        complexVectorPtr += 8;
 
        cplxValue2 = _mm256_loadu_ps(complexVectorPtr);
        complexVectorPtr += 8;
 
        cplxValue1 = _mm256_mul_ps(cplxValue1, cplxValue1); // Square the values
        cplxValue2 = _mm256_mul_ps(cplxValue2, cplxValue2); // Square the Values
 
        result = _mm256_hadd_ps(cplxValue1, cplxValue2); // Add the I2 and Q2 values
 
        result = _mm256_sqrt_ps(result);
 
        result = _mm256_mul_ps(result, vScalar);
 
        resultInt = _mm256_cvtps_epi32(result);
        resultInt = _mm256_packs_epi32(resultInt, resultInt);
        resultInt = _mm256_permutevar8x32_epi32(
            resultInt, idx); // permute to compensate for shuffling in hadd and packs
        resultShort = _mm256_extracti128_si256(resultInt, 0);
        _mm_storeu_si128((__m128i*)magnitudeVectorPtr, resultShort);
        magnitudeVectorPtr += 8;
    }
 
    number = eighthPoints * 8;
    volk_32fc_s32f_magnitude_16i_generic(
        magnitudeVector + number, complexVector + number, scalar, num_points - number);
}
#endif /* LV_HAVE_AVX2 */
 
#ifdef LV_HAVE_NEON
#include <arm_neon.h>
 
static inline void volk_32fc_s32f_magnitude_16i_neon(int16_t* magnitudeVector,
                                                     const lv_32fc_t* complexVector,
                                                     const float scalar,
                                                     unsigned int num_points)
{
    unsigned int number = 0;
    const unsigned int quarter_points = num_points / 4;
 
    const float* complexVectorPtr = (const float*)complexVector;
    int16_t* magnitudeVectorPtr = magnitudeVector;
    float32x4_t vScalar = vdupq_n_f32(scalar);
 
    float32x4_t half = vdupq_n_f32(0.5f);
 
    for (; number < quarter_points; number++) {
        float32x4x2_t input = vld2q_f32(complexVectorPtr);
        complexVectorPtr += 8;
 
        float32x4_t realSquared = vmulq_f32(input.val[0], input.val[0]);
        float32x4_t imagSquared = vmulq_f32(input.val[1], input.val[1]);
        float32x4_t sumSquared = vaddq_f32(realSquared, imagSquared);
 
        /* Use reciprocal square root estimate with Newton-Raphson refinement */
        float32x4_t rsqrt = vrsqrteq_f32(sumSquared);
        rsqrt = vmulq_f32(rsqrt, vrsqrtsq_f32(vmulq_f32(sumSquared, rsqrt), rsqrt));
        rsqrt = vmulq_f32(rsqrt, vrsqrtsq_f32(vmulq_f32(sumSquared, rsqrt), rsqrt));
        float32x4_t magnitude = vmulq_f32(sumSquared, rsqrt);
 
        /* Handle zero case */
        uint32x4_t zero_mask = vceqq_f32(sumSquared, vdupq_n_f32(0.0f));
        magnitude = vbslq_f32(zero_mask, sumSquared, magnitude);
 
        // Magnitude is always non-negative, so just add 0.5 for rounding
        float32x4_t scaled = vaddq_f32(vmulq_f32(magnitude, vScalar), half);
        int32x4_t intVal = vcvtq_s32_f32(scaled);
        int16x4_t shortVal = vqmovn_s32(intVal);
 
        vst1_s16(magnitudeVectorPtr, shortVal);
        magnitudeVectorPtr += 4;
    }
 
    number = quarter_points * 4;
    for (; number < num_points; number++) {
        float real = *complexVectorPtr++;
        float imag = *complexVectorPtr++;
        *magnitudeVectorPtr++ =
            (int16_t)rintf(scalar * sqrtf((real * real) + (imag * imag)));
    }
}
#endif /* LV_HAVE_NEON */
 
#ifdef LV_HAVE_NEONV8
#include <arm_neon.h>
 
static inline void volk_32fc_s32f_magnitude_16i_neonv8(int16_t* magnitudeVector,
                                                       const lv_32fc_t* complexVector,
                                                       const float scalar,
                                                       unsigned int num_points)
{
    unsigned int number = 0;
    const unsigned int eighth_points = num_points / 8;
 
    const float* complexVectorPtr = (const float*)complexVector;
    int16_t* magnitudeVectorPtr = magnitudeVector;
    float32x4_t vScalar = vdupq_n_f32(scalar);
 
    for (; number < eighth_points; number++) {
        float32x4x2_t input0 = vld2q_f32(complexVectorPtr);
        float32x4x2_t input1 = vld2q_f32(complexVectorPtr + 8);
        complexVectorPtr += 16;
        __VOLK_PREFETCH(complexVectorPtr + 16);
 
        float32x4_t sumSquared0 = vfmaq_f32(
            vmulq_f32(input0.val[1], input0.val[1]), input0.val[0], input0.val[0]);
        float32x4_t sumSquared1 = vfmaq_f32(
            vmulq_f32(input1.val[1], input1.val[1]), input1.val[0], input1.val[0]);
 
        float32x4_t magnitude0 = vsqrtq_f32(sumSquared0);
        float32x4_t magnitude1 = vsqrtq_f32(sumSquared1);
 
        float32x4_t scaled0 = vmulq_f32(magnitude0, vScalar);
        float32x4_t scaled1 = vmulq_f32(magnitude1, vScalar);
 
        int32x4_t intVal0 = vcvtnq_s32_f32(scaled0);
        int32x4_t intVal1 = vcvtnq_s32_f32(scaled1);
 
        int16x4_t shortVal0 = vqmovn_s32(intVal0);
        int16x4_t shortVal1 = vqmovn_s32(intVal1);
 
        vst1_s16(magnitudeVectorPtr, shortVal0);
        vst1_s16(magnitudeVectorPtr + 4, shortVal1);
        magnitudeVectorPtr += 8;
    }
 
    number = eighth_points * 8;
    for (; number < num_points; number++) {
        float real = *complexVectorPtr++;
        float imag = *complexVectorPtr++;
        *magnitudeVectorPtr++ =
            (int16_t)rintf(scalar * sqrtf((real * real) + (imag * imag)));
    }
}
#endif /* LV_HAVE_NEONV8 */
 
#ifdef LV_HAVE_RVV
#include <riscv_vector.h>
 
static inline void volk_32fc_s32f_magnitude_16i_rvv(int16_t* magnitudeVector,
                                                    const lv_32fc_t* complexVector,
                                                    const float scalar,
                                                    unsigned int num_points)
{
    size_t n = num_points;
    for (size_t vl; n > 0; n -= vl, complexVector += vl, magnitudeVector += vl) {
        vl = __riscv_vsetvl_e32m4(n);
        vuint64m8_t vc = __riscv_vle64_v_u64m8((const uint64_t*)complexVector, vl);
        vfloat32m4_t vr = __riscv_vreinterpret_f32m4(__riscv_vnsrl(vc, 0, vl));
        vfloat32m4_t vi = __riscv_vreinterpret_f32m4(__riscv_vnsrl(vc, 32, vl));
        vfloat32m4_t v = __riscv_vfmacc(__riscv_vfmul(vi, vi, vl), vr, vr, vl);
        v = __riscv_vfmul(__riscv_vfsqrt(v, vl), scalar, vl);
        __riscv_vse16(magnitudeVector, __riscv_vfncvt_x(v, vl), vl);
    }
}
#endif /*LV_HAVE_RVV*/
 
#ifdef LV_HAVE_RVVSEG
#include <riscv_vector.h>
 
static inline void volk_32fc_s32f_magnitude_16i_rvvseg(int16_t* magnitudeVector,
                                                       const lv_32fc_t* complexVector,
                                                       const float scalar,
                                                       unsigned int num_points)
{
    size_t n = num_points;
    for (size_t vl; n > 0; n -= vl, complexVector += vl, magnitudeVector += vl) {
        vl = __riscv_vsetvl_e32m4(n);
        vfloat32m4x2_t vc = __riscv_vlseg2e32_v_f32m4x2((const float*)complexVector, vl);
        vfloat32m4_t vr = __riscv_vget_f32m4(vc, 0);
        vfloat32m4_t vi = __riscv_vget_f32m4(vc, 1);
        vfloat32m4_t v = __riscv_vfmacc(__riscv_vfmul(vi, vi, vl), vr, vr, vl);
        v = __riscv_vfmul(__riscv_vfsqrt(v, vl), scalar, vl);
        __riscv_vse16(magnitudeVector, __riscv_vfncvt_x(v, vl), vl);
    }
}
#endif /*LV_HAVE_RVVSEG*/
 
#endif /* INCLUDED_volk_32fc_s32f_magnitude_16i_u_H */