51#ifndef INCLUDED_volk_32fc_accumulator_s32fc_a_H
52#define INCLUDED_volk_32fc_accumulator_s32fc_a_H
60 unsigned int num_points)
63 unsigned int number = 0;
66 for (; number < num_points; number++) {
67 returnValue += (*aPtr++);
69 *result = returnValue;
78 unsigned int num_points)
81 unsigned int number = 0;
82 const unsigned int quarterPoints = num_points / 4;
87 __m256 accumulator = _mm256_setzero_ps();
88 __m256 aVal = _mm256_setzero_ps();
90 for (; number < quarterPoints; number++) {
91 aVal = _mm256_loadu_ps((
float*)aPtr);
92 accumulator = _mm256_add_ps(accumulator, aVal);
96 _mm256_store_ps(tempBuffer, accumulator);
98 returnValue =
lv_cmake(tempBuffer[0], tempBuffer[1]);
99 returnValue +=
lv_cmake(tempBuffer[2], tempBuffer[3]);
100 returnValue +=
lv_cmake(tempBuffer[4], tempBuffer[5]);
101 returnValue +=
lv_cmake(tempBuffer[6], tempBuffer[7]);
103 number = quarterPoints * 4;
104 for (; number < num_points; number++) {
105 returnValue += (*aPtr++);
107 *result = returnValue;
112#include <xmmintrin.h>
116 unsigned int num_points)
119 unsigned int number = 0;
120 const unsigned int halfPoints = num_points / 2;
125 __m128 accumulator = _mm_setzero_ps();
126 __m128 aVal = _mm_setzero_ps();
128 for (; number < halfPoints; number++) {
129 aVal = _mm_loadu_ps((
float*)aPtr);
130 accumulator = _mm_add_ps(accumulator, aVal);
134 _mm_store_ps(tempBuffer, accumulator);
136 returnValue =
lv_cmake(tempBuffer[0], tempBuffer[1]);
137 returnValue +=
lv_cmake(tempBuffer[2], tempBuffer[3]);
139 number = halfPoints * 2;
140 for (; number < num_points; number++) {
141 returnValue += (*aPtr++);
143 *result = returnValue;
148#include <immintrin.h>
152 unsigned int num_points)
155 unsigned int number = 0;
156 const unsigned int quarterPoints = num_points / 4;
161 __m256 accumulator = _mm256_setzero_ps();
162 __m256 aVal = _mm256_setzero_ps();
164 for (; number < quarterPoints; number++) {
165 aVal = _mm256_load_ps((
float*)aPtr);
166 accumulator = _mm256_add_ps(accumulator, aVal);
170 _mm256_store_ps(tempBuffer, accumulator);
172 returnValue =
lv_cmake(tempBuffer[0], tempBuffer[1]);
173 returnValue +=
lv_cmake(tempBuffer[2], tempBuffer[3]);
174 returnValue +=
lv_cmake(tempBuffer[4], tempBuffer[5]);
175 returnValue +=
lv_cmake(tempBuffer[6], tempBuffer[7]);
177 number = quarterPoints * 4;
178 for (; number < num_points; number++) {
179 returnValue += (*aPtr++);
181 *result = returnValue;
186#include <xmmintrin.h>
190 unsigned int num_points)
193 unsigned int number = 0;
194 const unsigned int halfPoints = num_points / 2;
199 __m128 accumulator = _mm_setzero_ps();
200 __m128 aVal = _mm_setzero_ps();
202 for (; number < halfPoints; number++) {
203 aVal = _mm_load_ps((
float*)aPtr);
204 accumulator = _mm_add_ps(accumulator, aVal);
208 _mm_store_ps(tempBuffer, accumulator);
210 returnValue =
lv_cmake(tempBuffer[0], tempBuffer[1]);
211 returnValue +=
lv_cmake(tempBuffer[2], tempBuffer[3]);
213 number = halfPoints * 2;
214 for (; number < num_points; number++) {
215 returnValue += (*aPtr++);
217 *result = returnValue;
225 unsigned int num_points)
228 unsigned int number = 0;
230 unsigned int eighthPoints = num_points / 8;
232 float32x4_t out_vec0 = { 0.f, 0.f, 0.f, 0.f };
233 float32x4_t out_vec1 = { 0.f, 0.f, 0.f, 0.f };
234 float32x4_t out_vec2 = { 0.f, 0.f, 0.f, 0.f };
235 float32x4_t out_vec3 = { 0.f, 0.f, 0.f, 0.f };
238 for (; number < eighthPoints; number++) {
239 in_vec = vld1q_f32((
float*)aPtr);
240 out_vec0 = vaddq_f32(in_vec, out_vec0);
243 in_vec = vld1q_f32((
float*)aPtr);
244 out_vec1 = vaddq_f32(in_vec, out_vec1);
247 in_vec = vld1q_f32((
float*)aPtr);
248 out_vec2 = vaddq_f32(in_vec, out_vec2);
251 in_vec = vld1q_f32((
float*)aPtr);
252 out_vec3 = vaddq_f32(in_vec, out_vec3);
255 vst1q_f32(tempBuffer, out_vec0);
256 returnValue =
lv_cmake(tempBuffer[0], tempBuffer[1]);
257 returnValue +=
lv_cmake(tempBuffer[2], tempBuffer[3]);
259 vst1q_f32(tempBuffer, out_vec1);
260 returnValue +=
lv_cmake(tempBuffer[0], tempBuffer[1]);
261 returnValue +=
lv_cmake(tempBuffer[2], tempBuffer[3]);
263 vst1q_f32(tempBuffer, out_vec2);
264 returnValue +=
lv_cmake(tempBuffer[0], tempBuffer[1]);
265 returnValue +=
lv_cmake(tempBuffer[2], tempBuffer[3]);
267 vst1q_f32(tempBuffer, out_vec3);
268 returnValue +=
lv_cmake(tempBuffer[0], tempBuffer[1]);
269 returnValue +=
lv_cmake(tempBuffer[2], tempBuffer[3]);
271 number = eighthPoints * 8;
272 for (; number < num_points; number++) {
273 returnValue += (*aPtr++);
275 *result = returnValue;
280#include <riscv_vector.h>
283static inline void volk_32fc_accumulator_s32fc_rvv(
lv_32fc_t* result,
285 unsigned int num_points)
287 size_t vlmax = __riscv_vsetvlmax_e32m8();
288 vfloat32m8_t vsum = __riscv_vfmv_v_f_f32m8(0, vlmax);
289 const float* in = (
const float*)inputBuffer;
290 size_t n = num_points * 2;
291 for (
size_t vl; n > 0; n -= vl, in += vl) {
292 vl = __riscv_vsetvl_e32m8(n < vlmax ? n : vlmax);
293 vfloat32m8_t v = __riscv_vle32_v_f32m8(in, vl);
294 vsum = __riscv_vfadd_tu(vsum, vsum, v, vl);
296 vuint64m8_t vsumu = __riscv_vreinterpret_u64m8(__riscv_vreinterpret_u32m8(vsum));
297 vfloat32m4_t vsum1 = __riscv_vreinterpret_f32m4(__riscv_vnsrl(vsumu, 0, vlmax));
298 vfloat32m4_t vsum2 = __riscv_vreinterpret_f32m4(__riscv_vnsrl(vsumu, 32, vlmax));
299 vlmax = __riscv_vsetvlmax_e32m1();
302 vfloat32m1_t z = __riscv_vfmv_s_f_f32m1(0, vlmax);
303 *result =
lv_cmake(__riscv_vfmv_f(__riscv_vfredusum(vr, z, vlmax)),
304 __riscv_vfmv_f(__riscv_vfredusum(vi, z, vlmax)));
static void volk_32fc_accumulator_s32fc_generic(lv_32fc_t *result, const lv_32fc_t *inputBuffer, unsigned int num_points)
Definition volk_32fc_accumulator_s32fc.h:58
static void volk_32fc_accumulator_s32fc_a_sse(lv_32fc_t *result, const lv_32fc_t *inputBuffer, unsigned int num_points)
Definition volk_32fc_accumulator_s32fc.h:188
static void volk_32fc_accumulator_s32fc_u_sse(lv_32fc_t *result, const lv_32fc_t *inputBuffer, unsigned int num_points)
Definition volk_32fc_accumulator_s32fc.h:114
static void volk_32fc_accumulator_s32fc_a_avx(lv_32fc_t *result, const lv_32fc_t *inputBuffer, unsigned int num_points)
Definition volk_32fc_accumulator_s32fc.h:150
static void volk_32fc_accumulator_s32fc_neon(lv_32fc_t *result, const lv_32fc_t *inputBuffer, unsigned int num_points)
Definition volk_32fc_accumulator_s32fc.h:223
static void volk_32fc_accumulator_s32fc_u_avx(lv_32fc_t *result, const lv_32fc_t *inputBuffer, unsigned int num_points)
Definition volk_32fc_accumulator_s32fc.h:76
#define __VOLK_ATTR_ALIGNED(x)
Definition volk_common.h:62
#define lv_cmake(r, i)
Definition volk_complex.h:77
float complex lv_32fc_t
Definition volk_complex.h:74
#define RISCV_SHRINK4(op, T, S, v)
Definition volk_rvv_intrinsics.h:24