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,
const lv_16sc_t* inputVector,
unsigned int num_points)
56 const unsigned int avx_iters = num_points / 8;
57 unsigned int number = 0;
58 const int16_t* complexVectorPtr = (int16_t*)inputVector;
59 float* outputVectorPtr = (
float*)outputVector;
64 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++)
79 *outputVectorPtr++ = (float)*complexVectorPtr++;
85 #ifdef LV_HAVE_GENERIC
90 for(
i = 0;
i < num_points;
i++)
100 #include <emmintrin.h>
104 const unsigned int sse_iters = num_points / 2;
111 for(number = 0; number < sse_iters; number++)
114 _mm_store_ps((
float*)_out, a);
128 #include <immintrin.h>
132 const unsigned int sse_iters = num_points / 4;
137 unsigned int i, number;
139 for(number = 0; number < sse_iters; number++)
142 _mm256_store_ps((
float*)_out, a);
147 for (
i = 0;
i < (num_points % 4); ++
i)
158 #include <arm_neon.h>
162 const unsigned int sse_iters = num_points / 2;
170 unsigned int i, number;
172 for(number = 0; number < sse_iters; number++)
174 a16x4 = vld1_s16((
const int16_t*)_in);
176 a32x4 = vmovl_s16(a16x4);
177 f32x4 = vcvtq_f32_s32(a32x4);
178 vst1q_f32((float32_t*)_out, f32x4);
182 for (
i = 0;
i < (num_points % 2); ++
i)
192 #ifndef INCLUDED_volk_16ic_convert_32fc_u_H
193 #define INCLUDED_volk_16ic_convert_32fc_u_H
199 #include <immintrin.h>
201 static inline void volk_16ic_convert_32fc_u_avx2(
lv_32fc_t* outputVector,
const lv_16sc_t* inputVector,
unsigned int num_points)
203 const unsigned int avx_iters = num_points / 8;
204 unsigned int number = 0;
205 const int16_t* complexVectorPtr = (int16_t*)inputVector;
206 float* outputVectorPtr = (
float*)outputVector;
211 for(number = 0; number < avx_iters; number++)
213 cplxValue = _mm_loadu_si128((__m128i*)complexVectorPtr);
214 complexVectorPtr += 8;
216 outValInt = _mm256_cvtepi16_epi32(cplxValue);
217 outVal = _mm256_cvtepi32_ps(outValInt);
218 _mm256_storeu_ps((
float*)outputVectorPtr, outVal);
220 outputVectorPtr += 8;
223 number = avx_iters * 8;
224 for(; number < num_points*2; number++)
226 *outputVectorPtr++ = (float)*complexVectorPtr++;
233 #include <emmintrin.h>
237 const unsigned int sse_iters = num_points / 2;
244 for(number = 0; number < sse_iters; number++)
247 _mm_storeu_ps((
float*)_out, a);
262 #include <immintrin.h>
266 const unsigned int sse_iters = num_points / 4;
271 unsigned int i, number;
273 for(number = 0; number < sse_iters; number++)
276 _mm256_storeu_ps((
float*)_out, a);
281 for (
i = 0;
i < (num_points % 4); ++
i)