45#ifndef INCLUDED_volk_16ic_convert_32fc_a_H
46#define INCLUDED_volk_16ic_convert_32fc_a_H
53static inline void volk_16ic_convert_32fc_a_avx2(
lv_32fc_t* outputVector,
55 unsigned int num_points)
57 const unsigned int avx_iters = num_points / 8;
58 unsigned int number = 0;
59 const int16_t* complexVectorPtr = (int16_t*)inputVector;
60 float* outputVectorPtr = (
float*)outputVector;
65 for (number = 0; number < avx_iters; number++) {
66 cplxValue = _mm_load_si128((__m128i*)complexVectorPtr);
67 complexVectorPtr += 8;
69 outValInt = _mm256_cvtepi16_epi32(cplxValue);
70 outVal = _mm256_cvtepi32_ps(outValInt);
71 _mm256_store_ps((
float*)outputVectorPtr, outVal);
76 number = avx_iters * 8;
77 for (; number < num_points * 2; number++) {
78 *outputVectorPtr++ = (float)*complexVectorPtr++;
88 unsigned int num_points)
91 for (
i = 0;
i < num_points;
i++) {
101#include <emmintrin.h>
105 unsigned int num_points)
107 const unsigned int sse_iters = num_points / 2;
114 for (number = 0; number < sse_iters; number++) {
121 _mm_store_ps((
float*)_out, a);
125 if (num_points & 1) {
134#include <immintrin.h>
138 unsigned int num_points)
140 const unsigned int sse_iters = num_points / 4;
145 unsigned int i, number;
147 for (number = 0; number < sse_iters; number++) {
158 _mm256_store_ps((
float*)_out, a);
163 for (
i = 0;
i < (num_points % 4); ++
i) {
177 unsigned int num_points)
179 const unsigned int sse_iters = num_points / 2;
187 unsigned int i, number;
189 for (number = 0; number < sse_iters; number++) {
190 a16x4 = vld1_s16((
const int16_t*)_in);
192 a32x4 = vmovl_s16(a16x4);
193 f32x4 = vcvtq_f32_s32(a32x4);
194 vst1q_f32((float32_t*)_out, f32x4);
198 for (
i = 0;
i < (num_points % 2); ++
i) {
207#ifndef INCLUDED_volk_16ic_convert_32fc_u_H
208#define INCLUDED_volk_16ic_convert_32fc_u_H
214#include <immintrin.h>
216static inline void volk_16ic_convert_32fc_u_avx2(
lv_32fc_t* outputVector,
218 unsigned int num_points)
220 const unsigned int avx_iters = num_points / 8;
221 unsigned int number = 0;
222 const int16_t* complexVectorPtr = (int16_t*)inputVector;
223 float* outputVectorPtr = (
float*)outputVector;
228 for (number = 0; number < avx_iters; number++) {
229 cplxValue = _mm_loadu_si128((__m128i*)complexVectorPtr);
230 complexVectorPtr += 8;
232 outValInt = _mm256_cvtepi16_epi32(cplxValue);
233 outVal = _mm256_cvtepi32_ps(outValInt);
234 _mm256_storeu_ps((
float*)outputVectorPtr, outVal);
236 outputVectorPtr += 8;
239 number = avx_iters * 8;
240 for (; number < num_points * 2; number++) {
241 *outputVectorPtr++ = (float)*complexVectorPtr++;
248#include <emmintrin.h>
252 unsigned int num_points)
254 const unsigned int sse_iters = num_points / 2;
261 for (number = 0; number < sse_iters; number++) {
268 _mm_storeu_ps((
float*)_out, a);
272 if (num_points & 1) {
282#include <immintrin.h>
286 unsigned int num_points)
288 const unsigned int sse_iters = num_points / 4;
293 unsigned int i, number;
295 for (number = 0; number < sse_iters; number++) {
306 _mm256_storeu_ps((
float*)_out, a);
311 for (
i = 0;
i < (num_points % 4); ++
i) {
320#include <riscv_vector.h>
322static inline void volk_16ic_convert_32fc_rvv(
lv_32fc_t* outputVector,
324 unsigned int num_points)
326 const int16_t* in = (
const int16_t*)inputVector;
327 float* out = (
float*)outputVector;
328 size_t n = num_points * 2;
329 for (
size_t vl; n > 0; n -= vl, in += vl, out += vl) {
330 vl = __riscv_vsetvl_e16m4(n);
331 vint16m4_t v = __riscv_vle16_v_i16m4(in, vl);
332 __riscv_vse32(out, __riscv_vfwcvt_f(v, vl), vl);
static void volk_16ic_convert_32fc_generic(lv_32fc_t *outputVector, const lv_16sc_t *inputVector, unsigned int num_points)
Definition volk_16ic_convert_32fc.h:86
static void volk_16ic_convert_32fc_u_avx(lv_32fc_t *outputVector, const lv_16sc_t *inputVector, unsigned int num_points)
Definition volk_16ic_convert_32fc.h:284
static void volk_16ic_convert_32fc_a_avx(lv_32fc_t *outputVector, const lv_16sc_t *inputVector, unsigned int num_points)
Definition volk_16ic_convert_32fc.h:136
static void volk_16ic_convert_32fc_u_sse2(lv_32fc_t *outputVector, const lv_16sc_t *inputVector, unsigned int num_points)
Definition volk_16ic_convert_32fc.h:250
static void volk_16ic_convert_32fc_neon(lv_32fc_t *outputVector, const lv_16sc_t *inputVector, unsigned int num_points)
Definition volk_16ic_convert_32fc.h:175
static void volk_16ic_convert_32fc_a_sse2(lv_32fc_t *outputVector, const lv_16sc_t *inputVector, unsigned int num_points)
Definition volk_16ic_convert_32fc.h:103
#define __VOLK_PREFETCH(addr)
Definition volk_common.h:68
#define lv_cimag(x)
Definition volk_complex.h:98
#define lv_cmake(r, i)
Definition volk_complex.h:77
#define lv_creal(x)
Definition volk_complex.h:96
float complex lv_32fc_t
Definition volk_complex.h:74
short complex lv_16sc_t
Definition volk_complex.h:71
for i
Definition volk_config_fixed.tmpl.h:13