volk_32fc_index_min_16u.h

Go to the documentation of this file.

/* -*- c++ -*- */
/*
 * Copyright 2021 Free Software Foundation, Inc.
 *
 * This file is part of VOLK
 *
 * SPDX-License-Identifier: LGPL-3.0-or-later
 */
 
#ifndef INCLUDED_volk_32fc_index_min_16u_a_H
#define INCLUDED_volk_32fc_index_min_16u_a_H
 
#include <inttypes.h>
#include <limits.h>
#include <stdio.h>
#include <volk/volk_common.h>
#include <volk/volk_complex.h>
 
#ifdef LV_HAVE_AVX2
#include <immintrin.h>
#include <volk/volk_avx2_intrinsics.h>
 
static inline void volk_32fc_index_min_16u_a_avx2_variant_0(uint16_t* target,
                                                            const lv_32fc_t* source,
                                                            uint32_t num_points)
{
    num_points = (num_points > USHRT_MAX) ? USHRT_MAX : num_points;
 
    const __m256i indices_increment = _mm256_set1_epi32(8);
    /*
     * At the start of each loop iteration current_indices holds the indices of
     * the complex numbers loaded from memory. Explanation for odd order is given
     * in implementation of vector_32fc_index_min_variant0().
     */
    __m256i current_indices = _mm256_set_epi32(7, 6, 3, 2, 5, 4, 1, 0);
 
    __m256 min_values = _mm256_set1_ps(FLT_MAX);
    __m256i min_indices = _mm256_setzero_si256();
 
    for (unsigned i = 0; i < num_points / 8u; ++i) {
        __m256 in0 = _mm256_load_ps((float*)source);
        __m256 in1 = _mm256_load_ps((float*)(source + 4));
        vector_32fc_index_min_variant0(
            in0, in1, &min_values, &min_indices, &current_indices, indices_increment);
        source += 8;
    }
 
    // determine minimum value and index in the result of the vectorized loop
    __VOLK_ATTR_ALIGNED(32) float min_values_buffer[8];
    __VOLK_ATTR_ALIGNED(32) uint32_t min_indices_buffer[8];
    _mm256_store_ps(min_values_buffer, min_values);
    _mm256_store_si256((__m256i*)min_indices_buffer, min_indices);
 
    float min = FLT_MAX;
    uint32_t index = 0;
    for (unsigned i = 0; i < 8; i++) {
        if (min_values_buffer[i] < min) {
            min = min_values_buffer[i];
            index = min_indices_buffer[i];
        }
    }
 
    // handle tail not processed by the vectorized loop
    for (unsigned i = num_points & (~7u); i < num_points; ++i) {
        const float abs_squared =
            lv_creal(*source) * lv_creal(*source) + lv_cimag(*source) * lv_cimag(*source);
        if (abs_squared < min) {
            min = abs_squared;
            index = i;
        }
        ++source;
    }
 
    *target = index;
}
 
#endif /*LV_HAVE_AVX2*/
 
#ifdef LV_HAVE_AVX2
#include <immintrin.h>
#include <volk/volk_avx2_intrinsics.h>
 
static inline void volk_32fc_index_min_16u_a_avx2_variant_1(uint16_t* target,
                                                            const lv_32fc_t* source,
                                                            uint32_t num_points)
{
    num_points = (num_points > USHRT_MAX) ? USHRT_MAX : num_points;
 
    const __m256i indices_increment = _mm256_set1_epi32(8);
    /*
     * At the start of each loop iteration current_indices holds the indices of
     * the complex numbers loaded from memory. Explanation for odd order is given
     * in implementation of vector_32fc_index_min_variant0().
     */
    __m256i current_indices = _mm256_set_epi32(7, 6, 3, 2, 5, 4, 1, 0);
 
    __m256 min_values = _mm256_set1_ps(FLT_MAX);
    __m256i min_indices = _mm256_setzero_si256();
 
    for (unsigned i = 0; i < num_points / 8u; ++i) {
        __m256 in0 = _mm256_load_ps((float*)source);
        __m256 in1 = _mm256_load_ps((float*)(source + 4));
        vector_32fc_index_min_variant1(
            in0, in1, &min_values, &min_indices, &current_indices, indices_increment);
        source += 8;
    }
 
    // determine minimum value and index in the result of the vectorized loop
    __VOLK_ATTR_ALIGNED(32) float min_values_buffer[8];
    __VOLK_ATTR_ALIGNED(32) uint32_t min_indices_buffer[8];
    _mm256_store_ps(min_values_buffer, min_values);
    _mm256_store_si256((__m256i*)min_indices_buffer, min_indices);
 
    float min = FLT_MAX;
    uint32_t index = 0;
    for (unsigned i = 0; i < 8; i++) {
        if (min_values_buffer[i] < min) {
            min = min_values_buffer[i];
            index = min_indices_buffer[i];
        }
    }
 
    // handle tail not processed by the vectorized loop
    for (unsigned i = num_points & (~7u); i < num_points; ++i) {
        const float abs_squared =
            lv_creal(*source) * lv_creal(*source) + lv_cimag(*source) * lv_cimag(*source);
        if (abs_squared < min) {
            min = abs_squared;
            index = i;
        }
        ++source;
    }
 
    *target = index;
}
 
#endif /*LV_HAVE_AVX2*/
 
#ifdef LV_HAVE_SSE3
#include <pmmintrin.h>
#include <xmmintrin.h>
 
static inline void volk_32fc_index_min_16u_a_sse3(uint16_t* target,
                                                  const lv_32fc_t* source,
                                                  uint32_t num_points)
{
    num_points = (num_points > USHRT_MAX) ? USHRT_MAX : num_points;
 
    union bit128 holderf;
    union bit128 holderi;
    float sq_dist = 0.0;
 
    union bit128 xmm5, xmm4;
    __m128 xmm1, xmm2, xmm3;
    __m128i xmm8, xmm11, xmm12, xmm9, xmm10;
 
    xmm5.int_vec = _mm_setzero_si128();
    xmm4.int_vec = _mm_setzero_si128();
    holderf.int_vec = _mm_setzero_si128();
    holderi.int_vec = _mm_setzero_si128();
 
    xmm8 = _mm_setr_epi32(0, 1, 2, 3);
    xmm9 = _mm_setzero_si128();
    xmm10 = _mm_setr_epi32(4, 4, 4, 4);
    xmm3 = _mm_set_ps1(FLT_MAX);
 
    int bound = num_points >> 2;
 
    for (int i = 0; i < bound; ++i) {
        xmm1 = _mm_load_ps((float*)source);
        xmm2 = _mm_load_ps((float*)&source[2]);
 
        source += 4;
 
        xmm1 = _mm_mul_ps(xmm1, xmm1);
        xmm2 = _mm_mul_ps(xmm2, xmm2);
 
        xmm1 = _mm_hadd_ps(xmm1, xmm2);
 
        xmm3 = _mm_min_ps(xmm1, xmm3);
 
        xmm4.float_vec = _mm_cmpgt_ps(xmm1, xmm3);
        xmm5.float_vec = _mm_cmpeq_ps(xmm1, xmm3);
 
        xmm11 = _mm_and_si128(xmm8, xmm5.int_vec);
        xmm12 = _mm_and_si128(xmm9, xmm4.int_vec);
 
        xmm9 = _mm_add_epi32(xmm11, xmm12);
 
        xmm8 = _mm_add_epi32(xmm8, xmm10);
    }
 
    if (num_points >> 1 & 1) {
        xmm2 = _mm_load_ps((float*)source);
 
        xmm1 = _mm_movelh_ps(bit128_p(&xmm8)->float_vec, bit128_p(&xmm8)->float_vec);
        xmm8 = bit128_p(&xmm1)->int_vec;
 
        xmm2 = _mm_mul_ps(xmm2, xmm2);
 
        source += 2;
 
        xmm1 = _mm_hadd_ps(xmm2, xmm2);
 
        xmm3 = _mm_min_ps(xmm1, xmm3);
 
        xmm10 = _mm_setr_epi32(2, 2, 2, 2);
 
        xmm4.float_vec = _mm_cmpgt_ps(xmm1, xmm3);
        xmm5.float_vec = _mm_cmpeq_ps(xmm1, xmm3);
 
        xmm11 = _mm_and_si128(xmm8, xmm5.int_vec);
        xmm12 = _mm_and_si128(xmm9, xmm4.int_vec);
 
        xmm9 = _mm_add_epi32(xmm11, xmm12);
 
        xmm8 = _mm_add_epi32(xmm8, xmm10);
    }
 
    if (num_points & 1) {
        sq_dist = lv_creal(source[0]) * lv_creal(source[0]) +
                  lv_cimag(source[0]) * lv_cimag(source[0]);
 
        xmm2 = _mm_load1_ps(&sq_dist);
 
        xmm1 = xmm3;
 
        xmm3 = _mm_min_ss(xmm3, xmm2);
 
        xmm4.float_vec = _mm_cmpgt_ps(xmm1, xmm3);
        xmm5.float_vec = _mm_cmpeq_ps(xmm1, xmm3);
 
        xmm8 = _mm_shuffle_epi32(xmm8, 0x00);
 
        xmm11 = _mm_and_si128(xmm8, xmm4.int_vec);
        xmm12 = _mm_and_si128(xmm9, xmm5.int_vec);
 
        xmm9 = _mm_add_epi32(xmm11, xmm12);
    }
 
    _mm_store_ps((float*)&(holderf.f), xmm3);
    _mm_store_si128(&(holderi.int_vec), xmm9);
 
    target[0] = holderi.i[0];
    sq_dist = holderf.f[0];
    target[0] = (holderf.f[1] < sq_dist) ? holderi.i[1] : target[0];
    sq_dist = (holderf.f[1] < sq_dist) ? holderf.f[1] : sq_dist;
    target[0] = (holderf.f[2] < sq_dist) ? holderi.i[2] : target[0];
    sq_dist = (holderf.f[2] < sq_dist) ? holderf.f[2] : sq_dist;
    target[0] = (holderf.f[3] < sq_dist) ? holderi.i[3] : target[0];
    sq_dist = (holderf.f[3] < sq_dist) ? holderf.f[3] : sq_dist;
}
 
#endif /*LV_HAVE_SSE3*/
 
 
#ifdef LV_HAVE_NEON
#include <arm_neon.h>
#include <float.h>
#include <limits.h>
#include <volk/volk_neon_intrinsics.h>
 
static inline void volk_32fc_index_min_16u_neon(uint16_t* target,
                                                const lv_32fc_t* source,
                                                uint32_t num_points)
{
    num_points = (num_points > USHRT_MAX) ? USHRT_MAX : num_points;
 
    if (num_points == 0)
        return;
 
    const uint32_t quarter_points = num_points / 4;
    const lv_32fc_t* inputPtr = source;
 
    // Use integer indices directly
    uint32x4_t vec_indices = { 0, 1, 2, 3 };
    const uint32x4_t vec_incr = vdupq_n_u32(4);
 
    float32x4_t vec_min = vdupq_n_f32(FLT_MAX);
    uint32x4_t vec_min_idx = vdupq_n_u32(0);
 
    for (uint32_t i = 0; i < quarter_points; i++) {
        // Load and deinterleave complex values
        float32x4x2_t cplx = vld2q_f32((const float*)inputPtr);
        inputPtr += 4;
 
        // Magnitude squared: re*re + im*im
        float32x4_t mag2 =
            vmlaq_f32(vmulq_f32(cplx.val[0], cplx.val[0]), cplx.val[1], cplx.val[1]);
 
        // Compare BEFORE min update to know which lanes change
        uint32x4_t lt_mask = vcltq_f32(mag2, vec_min);
        vec_min_idx = vbslq_u32(lt_mask, vec_indices, vec_min_idx);
 
        // vminq_f32 is single-cycle, no dependency on comparison result
        vec_min = vminq_f32(mag2, vec_min);
 
        vec_indices = vaddq_u32(vec_indices, vec_incr);
    }
 
    // Scalar reduction
    __VOLK_ATTR_ALIGNED(16) float min_buf[4];
    __VOLK_ATTR_ALIGNED(16) uint32_t idx_buf[4];
    vst1q_f32(min_buf, vec_min);
    vst1q_u32(idx_buf, vec_min_idx);
 
    float min_val = min_buf[0];
    uint32_t result_idx = idx_buf[0];
    for (int i = 1; i < 4; i++) {
        if (min_buf[i] < min_val) {
            min_val = min_buf[i];
            result_idx = idx_buf[i];
        } else if (min_buf[i] == min_val && idx_buf[i] < result_idx) {
            result_idx = idx_buf[i];
        }
    }
 
    // Handle tail
    for (uint32_t i = quarter_points * 4; i < num_points; i++) {
        float re = lv_creal(source[i]);
        float im = lv_cimag(source[i]);
        float mag2 = re * re + im * im;
        if (mag2 < min_val) {
            min_val = mag2;
            result_idx = i;
        }
    }
 
    *target = (uint16_t)result_idx;
}
 
#endif /*LV_HAVE_NEON*/
 
 
#ifdef LV_HAVE_NEONV8
#include <arm_neon.h>
#include <float.h>
#include <limits.h>
 
static inline void volk_32fc_index_min_16u_neonv8(uint16_t* target,
                                                  const lv_32fc_t* source,
                                                  uint32_t num_points)
{
    num_points = (num_points > USHRT_MAX) ? USHRT_MAX : num_points;
 
    if (num_points == 0)
        return;
 
    const uint32_t quarter_points = num_points / 4;
    const lv_32fc_t* inputPtr = source;
 
    // Use integer indices directly (no float conversion overhead)
    uint32x4_t vec_indices = { 0, 1, 2, 3 };
    const uint32x4_t vec_incr = vdupq_n_u32(4);
 
    float32x4_t vec_min = vdupq_n_f32(FLT_MAX);
    uint32x4_t vec_min_idx = vdupq_n_u32(0);
 
    for (uint32_t i = 0; i < quarter_points; i++) {
        // Load and deinterleave complex values
        float32x4x2_t cplx = vld2q_f32((const float*)inputPtr);
        inputPtr += 4;
 
        // Magnitude squared using FMA: re*re + im*im
        float32x4_t mag2 =
            vfmaq_f32(vmulq_f32(cplx.val[0], cplx.val[0]), cplx.val[1], cplx.val[1]);
 
        // Compare BEFORE min update to know which lanes change
        uint32x4_t lt_mask = vcltq_f32(mag2, vec_min);
        vec_min_idx = vbslq_u32(lt_mask, vec_indices, vec_min_idx);
 
        // vminq_f32 is single-cycle, no dependency on comparison result
        vec_min = vminq_f32(mag2, vec_min);
 
        vec_indices = vaddq_u32(vec_indices, vec_incr);
    }
 
    // ARMv8 horizontal reduction
    float min_val = vminvq_f32(vec_min);
    uint32x4_t min_mask = vceqq_f32(vec_min, vdupq_n_f32(min_val));
    uint32x4_t idx_masked = vbslq_u32(min_mask, vec_min_idx, vdupq_n_u32(UINT32_MAX));
    uint32_t result_idx = vminvq_u32(idx_masked);
 
    // Handle tail
    for (uint32_t i = quarter_points * 4; i < num_points; i++) {
        float re = lv_creal(source[i]);
        float im = lv_cimag(source[i]);
        float mag2 = re * re + im * im;
        if (mag2 < min_val) {
            min_val = mag2;
            result_idx = i;
        }
    }
 
    *target = (uint16_t)result_idx;
}
 
#endif /*LV_HAVE_NEONV8*/
 
 
#ifdef LV_HAVE_GENERIC
static inline void volk_32fc_index_min_16u_generic(uint16_t* target,
                                                   const lv_32fc_t* source,
                                                   uint32_t num_points)
{
    num_points = (num_points > USHRT_MAX) ? USHRT_MAX : num_points;
 
    float sq_dist = 0.0;
    float min = FLT_MAX;
    uint16_t index = 0;
 
    for (uint32_t i = 0; i < num_points; ++i) {
        sq_dist = lv_creal(source[i]) * lv_creal(source[i]) +
                  lv_cimag(source[i]) * lv_cimag(source[i]);
 
        if (sq_dist < min) {
            index = i;
            min = sq_dist;
        }
    }
    target[0] = index;
}
 
#endif /*LV_HAVE_GENERIC*/
 
#ifdef LV_HAVE_AVX512F
#include <float.h>
#include <immintrin.h>
#include <limits.h>
 
static inline void volk_32fc_index_min_16u_a_avx512f(uint16_t* target,
                                                     const lv_32fc_t* src0,
                                                     uint32_t num_points)
{
    num_points = (num_points > USHRT_MAX) ? USHRT_MAX : num_points;
 
    const lv_32fc_t* src0Ptr = src0;
    const uint32_t sixteenthPoints = num_points / 16;
 
    // Index ordering after shuffle: [0,1,8,9, 2,3,10,11, 4,5,12,13, 6,7,14,15]
    __m512 currentIndexes =
        _mm512_setr_ps(0, 1, 8, 9, 2, 3, 10, 11, 4, 5, 12, 13, 6, 7, 14, 15);
    const __m512 indexIncrement = _mm512_set1_ps(16);
 
    __m512 minValues = _mm512_set1_ps(FLT_MAX);
    __m512 minIndices = _mm512_setzero_ps();
 
    for (uint32_t number = 0; number < sixteenthPoints; number++) {
        // Load 16 complex values (32 floats)
        __m512 in0 = _mm512_load_ps((const float*)src0Ptr);
        __m512 in1 = _mm512_load_ps((const float*)(src0Ptr + 8));
        src0Ptr += 16;
 
        // Square all values
        in0 = _mm512_mul_ps(in0, in0);
        in1 = _mm512_mul_ps(in1, in1);
 
        // Add adjacent pairs (re² + im²) using within-lane shuffle
        // 0xB1 = _MM_SHUFFLE(2,3,0,1) swaps adjacent elements
        __m512 sw0 = _mm512_shuffle_ps(in0, in0, 0xB1);
        __m512 sw1 = _mm512_shuffle_ps(in1, in1, 0xB1);
        __m512 sum0 = _mm512_add_ps(in0, sw0);
        __m512 sum1 = _mm512_add_ps(in1, sw1);
 
        // Compact: pick elements 0,2 from sum0 and sum1 per 128-bit lane
        // 0x88 = _MM_SHUFFLE(2,0,2,0)
        __m512 mag_sq = _mm512_shuffle_ps(sum0, sum1, 0x88);
 
        // Compare and update minimums
        __mmask16 cmpMask = _mm512_cmp_ps_mask(mag_sq, minValues, _CMP_LT_OS);
        minIndices = _mm512_mask_blend_ps(cmpMask, minIndices, currentIndexes);
        minValues = _mm512_min_ps(mag_sq, minValues);
 
        currentIndexes = _mm512_add_ps(currentIndexes, indexIncrement);
    }
 
    // Reduce 16 values to find minimum
    __VOLK_ATTR_ALIGNED(64) float minValuesBuffer[16];
    __VOLK_ATTR_ALIGNED(64) float minIndexesBuffer[16];
    _mm512_store_ps(minValuesBuffer, minValues);
    _mm512_store_ps(minIndexesBuffer, minIndices);
 
    float min = FLT_MAX;
    uint32_t index = 0;
    for (uint32_t i = 0; i < 16; i++) {
        if (minValuesBuffer[i] < min) {
            min = minValuesBuffer[i];
            index = (uint32_t)minIndexesBuffer[i];
        } else if (minValuesBuffer[i] == min) {
            if ((uint32_t)minIndexesBuffer[i] < index)
                index = (uint32_t)minIndexesBuffer[i];
        }
    }
 
    // Handle tail
    for (uint32_t number = sixteenthPoints * 16; number < num_points; number++) {
        const float re = lv_creal(*src0Ptr);
        const float im = lv_cimag(*src0Ptr);
        const float sq_dist = re * re + im * im;
        if (sq_dist < min) {
            min = sq_dist;
            index = number;
        }
        src0Ptr++;
    }
    *target = (uint16_t)index;
}
 
#endif /*LV_HAVE_AVX512F*/
 
#endif /*INCLUDED_volk_32fc_index_min_16u_a_H*/
 
#ifndef INCLUDED_volk_32fc_index_min_16u_u_H
#define INCLUDED_volk_32fc_index_min_16u_u_H
 
#include <inttypes.h>
#include <limits.h>
#include <stdio.h>
#include <volk/volk_common.h>
#include <volk/volk_complex.h>
 
#ifdef LV_HAVE_AVX2
#include <immintrin.h>
#include <volk/volk_avx2_intrinsics.h>
 
static inline void volk_32fc_index_min_16u_u_avx2_variant_0(uint16_t* target,
                                                            const lv_32fc_t* source,
                                                            uint32_t num_points)
{
    num_points = (num_points > USHRT_MAX) ? USHRT_MAX : num_points;
 
    const __m256i indices_increment = _mm256_set1_epi32(8);
    /*
     * At the start of each loop iteration current_indices holds the indices of
     * the complex numbers loaded from memory. Explanation for odd order is given
     * in implementation of vector_32fc_index_min_variant0().
     */
    __m256i current_indices = _mm256_set_epi32(7, 6, 3, 2, 5, 4, 1, 0);
 
    __m256 min_values = _mm256_set1_ps(FLT_MAX);
    __m256i min_indices = _mm256_setzero_si256();
 
    for (unsigned i = 0; i < num_points / 8u; ++i) {
        __m256 in0 = _mm256_loadu_ps((float*)source);
        __m256 in1 = _mm256_loadu_ps((float*)(source + 4));
        vector_32fc_index_min_variant0(
            in0, in1, &min_values, &min_indices, &current_indices, indices_increment);
        source += 8;
    }
 
    // determine minimum value and index in the result of the vectorized loop
    __VOLK_ATTR_ALIGNED(32) float min_values_buffer[8];
    __VOLK_ATTR_ALIGNED(32) uint32_t min_indices_buffer[8];
    _mm256_store_ps(min_values_buffer, min_values);
    _mm256_store_si256((__m256i*)min_indices_buffer, min_indices);
 
    float min = FLT_MAX;
    uint32_t index = 0;
    for (unsigned i = 0; i < 8; i++) {
        if (min_values_buffer[i] < min) {
            min = min_values_buffer[i];
            index = min_indices_buffer[i];
        }
    }
 
    // handle tail not processed by the vectorized loop
    for (unsigned i = num_points & (~7u); i < num_points; ++i) {
        const float abs_squared =
            lv_creal(*source) * lv_creal(*source) + lv_cimag(*source) * lv_cimag(*source);
        if (abs_squared < min) {
            min = abs_squared;
            index = i;
        }
        ++source;
    }
 
    *target = index;
}
 
#endif /*LV_HAVE_AVX2*/
 
#ifdef LV_HAVE_AVX2
#include <immintrin.h>
#include <volk/volk_avx2_intrinsics.h>
 
static inline void volk_32fc_index_min_16u_u_avx2_variant_1(uint16_t* target,
                                                            const lv_32fc_t* source,
                                                            uint32_t num_points)
{
    num_points = (num_points > USHRT_MAX) ? USHRT_MAX : num_points;
 
    const __m256i indices_increment = _mm256_set1_epi32(8);
    /*
     * At the start of each loop iteration current_indices holds the indices of
     * the complex numbers loaded from memory. Explanation for odd order is given
     * in implementation of vector_32fc_index_min_variant0().
     */
    __m256i current_indices = _mm256_set_epi32(7, 6, 3, 2, 5, 4, 1, 0);
 
    __m256 min_values = _mm256_set1_ps(FLT_MAX);
    __m256i min_indices = _mm256_setzero_si256();
 
    for (unsigned i = 0; i < num_points / 8u; ++i) {
        __m256 in0 = _mm256_loadu_ps((float*)source);
        __m256 in1 = _mm256_loadu_ps((float*)(source + 4));
        vector_32fc_index_min_variant1(
            in0, in1, &min_values, &min_indices, &current_indices, indices_increment);
        source += 8;
    }
 
    // determine minimum value and index in the result of the vectorized loop
    __VOLK_ATTR_ALIGNED(32) float min_values_buffer[8];
    __VOLK_ATTR_ALIGNED(32) uint32_t min_indices_buffer[8];
    _mm256_store_ps(min_values_buffer, min_values);
    _mm256_store_si256((__m256i*)min_indices_buffer, min_indices);
 
    float min = FLT_MAX;
    uint32_t index = 0;
    for (unsigned i = 0; i < 8; i++) {
        if (min_values_buffer[i] < min) {
            min = min_values_buffer[i];
            index = min_indices_buffer[i];
        }
    }
 
    // handle tail not processed by the vectorized loop
    for (unsigned i = num_points & (~7u); i < num_points; ++i) {
        const float abs_squared =
            lv_creal(*source) * lv_creal(*source) + lv_cimag(*source) * lv_cimag(*source);
        if (abs_squared < min) {
            min = abs_squared;
            index = i;
        }
        ++source;
    }
 
    *target = index;
}
 
#endif /*LV_HAVE_AVX2*/
 
#ifdef LV_HAVE_AVX512F
#include <float.h>
#include <immintrin.h>
#include <limits.h>
 
static inline void volk_32fc_index_min_16u_u_avx512f(uint16_t* target,
                                                     const lv_32fc_t* src0,
                                                     uint32_t num_points)
{
    num_points = (num_points > USHRT_MAX) ? USHRT_MAX : num_points;
 
    const lv_32fc_t* src0Ptr = src0;
    const uint32_t sixteenthPoints = num_points / 16;
 
    // Index ordering after shuffle: [0,1,8,9, 2,3,10,11, 4,5,12,13, 6,7,14,15]
    __m512 currentIndexes =
        _mm512_setr_ps(0, 1, 8, 9, 2, 3, 10, 11, 4, 5, 12, 13, 6, 7, 14, 15);
    const __m512 indexIncrement = _mm512_set1_ps(16);
 
    __m512 minValues = _mm512_set1_ps(FLT_MAX);
    __m512 minIndices = _mm512_setzero_ps();
 
    for (uint32_t number = 0; number < sixteenthPoints; number++) {
        // Load 16 complex values (32 floats)
        __m512 in0 = _mm512_loadu_ps((const float*)src0Ptr);
        __m512 in1 = _mm512_loadu_ps((const float*)(src0Ptr + 8));
        src0Ptr += 16;
 
        // Square all values
        in0 = _mm512_mul_ps(in0, in0);
        in1 = _mm512_mul_ps(in1, in1);
 
        // Add adjacent pairs (re² + im²) using within-lane shuffle
        // 0xB1 = _MM_SHUFFLE(2,3,0,1) swaps adjacent elements
        __m512 sw0 = _mm512_shuffle_ps(in0, in0, 0xB1);
        __m512 sw1 = _mm512_shuffle_ps(in1, in1, 0xB1);
        __m512 sum0 = _mm512_add_ps(in0, sw0);
        __m512 sum1 = _mm512_add_ps(in1, sw1);
 
        // Compact: pick elements 0,2 from sum0 and sum1 per 128-bit lane
        // 0x88 = _MM_SHUFFLE(2,0,2,0)
        __m512 mag_sq = _mm512_shuffle_ps(sum0, sum1, 0x88);
 
        // Compare and update minimums
        __mmask16 cmpMask = _mm512_cmp_ps_mask(mag_sq, minValues, _CMP_LT_OS);
        minIndices = _mm512_mask_blend_ps(cmpMask, minIndices, currentIndexes);
        minValues = _mm512_min_ps(mag_sq, minValues);
 
        currentIndexes = _mm512_add_ps(currentIndexes, indexIncrement);
    }
 
    // Reduce 16 values to find minimum
    __VOLK_ATTR_ALIGNED(64) float minValuesBuffer[16];
    __VOLK_ATTR_ALIGNED(64) float minIndexesBuffer[16];
    _mm512_store_ps(minValuesBuffer, minValues);
    _mm512_store_ps(minIndexesBuffer, minIndices);
 
    float min = FLT_MAX;
    uint32_t index = 0;
    for (uint32_t i = 0; i < 16; i++) {
        if (minValuesBuffer[i] < min) {
            min = minValuesBuffer[i];
            index = (uint32_t)minIndexesBuffer[i];
        } else if (minValuesBuffer[i] == min) {
            if ((uint32_t)minIndexesBuffer[i] < index)
                index = (uint32_t)minIndexesBuffer[i];
        }
    }
 
    // Handle tail
    for (uint32_t number = sixteenthPoints * 16; number < num_points; number++) {
        const float re = lv_creal(*src0Ptr);
        const float im = lv_cimag(*src0Ptr);
        const float sq_dist = re * re + im * im;
        if (sq_dist < min) {
            min = sq_dist;
            index = number;
        }
        src0Ptr++;
    }
    *target = (uint16_t)index;
}
 
#endif /*LV_HAVE_AVX512F*/
 
#ifdef LV_HAVE_RVV
#include <float.h>
#include <riscv_vector.h>
 
static inline void volk_32fc_index_min_16u_rvv(uint16_t* target,
                                               const lv_32fc_t* source,
                                               uint32_t num_points)
{
    vfloat32m4_t vmin = __riscv_vfmv_v_f_f32m4(FLT_MAX, __riscv_vsetvlmax_e32m4());
    vuint16m2_t vmini = __riscv_vmv_v_x_u16m2(0, __riscv_vsetvlmax_e16m2());
    vuint16m2_t vidx = __riscv_vid_v_u16m2(__riscv_vsetvlmax_e16m2());
    size_t n = (num_points > USHRT_MAX) ? USHRT_MAX : num_points;
    for (size_t vl; n > 0; n -= vl, source += vl) {
        vl = __riscv_vsetvl_e32m4(n);
        vuint64m8_t vc = __riscv_vle64_v_u64m8((const uint64_t*)source, 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(vr, vr, vl), vi, vi, vl);
        vbool8_t m = __riscv_vmfgt(vmin, v, vl);
        vmin = __riscv_vfmin_tu(vmin, vmin, v, vl);
        vmini = __riscv_vmerge_tu(vmini, vmini, vidx, m, vl);
        vidx = __riscv_vadd(vidx, vl, __riscv_vsetvlmax_e16m4());
    }
    size_t vl = __riscv_vsetvlmax_e32m4();
    float min = __riscv_vfmv_f(__riscv_vfredmin(RISCV_SHRINK4(vfmin, f, 32, vmin),
                                                __riscv_vfmv_v_f_f32m1(FLT_MAX, 1),
                                                __riscv_vsetvlmax_e32m1()));
    // Find minimum index among lanes with min value
    __attribute__((aligned(32))) float values[128];
    __attribute__((aligned(32))) uint16_t indices[128];
    __riscv_vse32(values, vmin, vl);
    __riscv_vse16(indices, vmini, vl);
    uint16_t min_idx = UINT16_MAX;
    for (size_t i = 0; i < vl; i++) {
        if (values[i] == min && indices[i] < min_idx) {
            min_idx = indices[i];
        }
    }
    *target = min_idx;
}
#endif /*LV_HAVE_RVV*/
 
#ifdef LV_HAVE_RVVSEG
#include <float.h>
#include <riscv_vector.h>
 
static inline void volk_32fc_index_min_16u_rvvseg(uint16_t* target,
                                                  const lv_32fc_t* source,
                                                  uint32_t num_points)
{
    vfloat32m4_t vmin = __riscv_vfmv_v_f_f32m4(FLT_MAX, __riscv_vsetvlmax_e32m4());
    vuint16m2_t vmini = __riscv_vmv_v_x_u16m2(0, __riscv_vsetvlmax_e16m2());
    vuint16m2_t vidx = __riscv_vid_v_u16m2(__riscv_vsetvlmax_e16m2());
    size_t n = (num_points > USHRT_MAX) ? USHRT_MAX : num_points;
    for (size_t vl; n > 0; n -= vl, source += vl) {
        vl = __riscv_vsetvl_e32m4(n);
        vfloat32m4x2_t vc = __riscv_vlseg2e32_v_f32m4x2((const float*)source, vl);
        vfloat32m4_t vr = __riscv_vget_f32m4(vc, 0), vi = __riscv_vget_f32m4(vc, 1);
        vfloat32m4_t v = __riscv_vfmacc(__riscv_vfmul(vr, vr, vl), vi, vi, vl);
        vbool8_t m = __riscv_vmfgt(vmin, v, vl);
        vmin = __riscv_vfmin_tu(vmin, vmin, v, vl);
        vmini = __riscv_vmerge_tu(vmini, vmini, vidx, m, vl);
        vidx = __riscv_vadd(vidx, vl, __riscv_vsetvlmax_e16m4());
    }
    size_t vl = __riscv_vsetvlmax_e32m4();
    float min = __riscv_vfmv_f(__riscv_vfredmin(RISCV_SHRINK4(vfmin, f, 32, vmin),
                                                __riscv_vfmv_v_f_f32m1(FLT_MAX, 1),
                                                __riscv_vsetvlmax_e32m1()));
    // Find minimum index among lanes with min value
    __attribute__((aligned(32))) float values[128];
    __attribute__((aligned(32))) uint16_t indices[128];
    __riscv_vse32(values, vmin, vl);
    __riscv_vse16(indices, vmini, vl);
    uint16_t min_idx = UINT16_MAX;
    for (size_t i = 0; i < vl; i++) {
        if (values[i] == min && indices[i] < min_idx) {
            min_idx = indices[i];
        }
    }
    *target = min_idx;
}
#endif /*LV_HAVE_RVVSEG*/
 
#endif /*INCLUDED_volk_32fc_index_min_16u_u_H*/