46 #ifndef INCLUDED_volk_16ic_convert_32fc_a_H 47 #define INCLUDED_volk_16ic_convert_32fc_a_H 52 #include <immintrin.h> 54 static inline void volk_16ic_convert_32fc_a_avx2(
lv_32fc_t* outputVector,
56 unsigned int num_points)
58 const unsigned int avx_iters = num_points / 8;
59 unsigned int number = 0;
60 const int16_t* complexVectorPtr = (int16_t*)inputVector;
61 float* outputVectorPtr = (
float*)outputVector;
66 for (number = 0; number < avx_iters; number++) {
67 cplxValue = _mm_load_si128((__m128i*)complexVectorPtr);
68 complexVectorPtr += 8;
70 outValInt = _mm256_cvtepi16_epi32(cplxValue);
71 outVal = _mm256_cvtepi32_ps(outValInt);
72 _mm256_store_ps((
float*)outputVectorPtr, outVal);
77 number = avx_iters * 8;
78 for (; number < num_points * 2; number++) {
79 *outputVectorPtr++ = (float)*complexVectorPtr++;
85 #ifdef LV_HAVE_GENERIC 89 unsigned int num_points)
92 for (i = 0; i < num_points; i++) {
102 #include <emmintrin.h> 106 unsigned int num_points)
108 const unsigned int sse_iters = num_points / 2;
115 for (number = 0; number < sse_iters; number++) {
122 _mm_store_ps((
float*)_out, a);
126 if (num_points & 1) {
135 #include <immintrin.h> 139 unsigned int num_points)
141 const unsigned int sse_iters = num_points / 4;
146 unsigned int i, number;
148 for (number = 0; number < sse_iters; number++) {
159 _mm256_store_ps((
float*)_out, a);
164 for (i = 0; i < (num_points % 4); ++
i) {
174 #include <arm_neon.h> 178 unsigned int num_points)
180 const unsigned int sse_iters = num_points / 2;
188 unsigned int i, number;
190 for (number = 0; number < sse_iters; number++) {
191 a16x4 = vld1_s16((
const int16_t*)_in);
193 a32x4 = vmovl_s16(a16x4);
194 f32x4 = vcvtq_f32_s32(a32x4);
195 vst1q_f32((float32_t*)_out, f32x4);
199 for (i = 0; i < (num_points % 2); ++
i) {
208 #ifndef INCLUDED_volk_16ic_convert_32fc_u_H 209 #define INCLUDED_volk_16ic_convert_32fc_u_H 215 #include <immintrin.h> 217 static inline void volk_16ic_convert_32fc_u_avx2(
lv_32fc_t* outputVector,
219 unsigned int num_points)
221 const unsigned int avx_iters = num_points / 8;
222 unsigned int number = 0;
223 const int16_t* complexVectorPtr = (int16_t*)inputVector;
224 float* outputVectorPtr = (
float*)outputVector;
229 for (number = 0; number < avx_iters; number++) {
230 cplxValue = _mm_loadu_si128((__m128i*)complexVectorPtr);
231 complexVectorPtr += 8;
233 outValInt = _mm256_cvtepi16_epi32(cplxValue);
234 outVal = _mm256_cvtepi32_ps(outValInt);
235 _mm256_storeu_ps((
float*)outputVectorPtr, outVal);
237 outputVectorPtr += 8;
240 number = avx_iters * 8;
241 for (; number < num_points * 2; number++) {
242 *outputVectorPtr++ = (float)*complexVectorPtr++;
249 #include <emmintrin.h> 253 unsigned int num_points)
255 const unsigned int sse_iters = num_points / 2;
262 for (number = 0; number < sse_iters; number++) {
269 _mm_storeu_ps((
float*)_out, a);
273 if (num_points & 1) {
283 #include <immintrin.h> 287 unsigned int num_points)
289 const unsigned int sse_iters = num_points / 4;
294 unsigned int i, number;
296 for (number = 0; number < sse_iters; number++) {
307 _mm256_storeu_ps((
float*)_out, a);
312 for (i = 0; i < (num_points % 4); ++
i) {
short complex lv_16sc_t
Definition: volk_complex.h:67
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:137
#define lv_cmake(r, i)
Definition: volk_complex.h:73
#define __VOLK_PREFETCH(addr)
Definition: volk_common.h:62
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:104
for i
Definition: volk_config_fixed.tmpl.h:25
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:176
float complex lv_32fc_t
Definition: volk_complex.h:70
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:285
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:87
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:251
#define lv_creal(x)
Definition: volk_complex.h:92
#define lv_cimag(x)
Definition: volk_complex.h:94