55 #ifndef INCLUDED_volk_32fc_s32f_x2_power_spectral_density_32f_a_H 56 #define INCLUDED_volk_32fc_s32f_x2_power_spectral_density_32f_a_H 63 #include <immintrin.h> 65 #ifdef LV_HAVE_LIB_SIMDMATH 72 const float normalizationFactor,
74 unsigned int num_points)
76 const float* inputPtr = (
const float*)complexFFTInput;
77 float* destPtr = logPowerOutput;
79 const float iRBW = 1.0 / rbw;
80 const float iNormalizationFactor = 1.0 / normalizationFactor;
82 #ifdef LV_HAVE_LIB_SIMDMATH 83 __m256 magScalar = _mm256_set1_ps(10.0);
84 magScalar = _mm256_div_ps(magScalar, logf4(magScalar));
86 __m256 invRBW = _mm256_set1_ps(iRBW);
88 __m256 invNormalizationFactor = _mm256_set1_ps(iNormalizationFactor);
91 __m256 input1, input2;
92 const uint64_t eighthPoints = num_points / 8;
93 for (; number < eighthPoints; number++) {
95 input1 = _mm256_load_ps(inputPtr);
97 input2 = _mm256_load_ps(inputPtr);
101 input1 = _mm256_mul_ps(input1, invNormalizationFactor);
102 input2 = _mm256_mul_ps(input2, invNormalizationFactor);
106 input1 = _mm256_mul_ps(input1, input1);
108 input2 = _mm256_mul_ps(input2, input2);
112 inputVal1 = _mm256_permute2f128_ps(input1, input2, 0x20);
113 inputVal2 = _mm256_permute2f128_ps(input1, input2, 0x31);
115 power = _mm256_hadd_ps(inputVal1, inputVal2);
118 power = _mm256_mul_ps(power, invRBW);
121 power = logf4(power);
124 power = _mm256_mul_ps(power, magScalar);
127 _mm256_store_ps(destPtr, power);
132 number = eighthPoints * 8;
135 for (; number < num_points; number++) {
142 const float real = *inputPtr++ * iNormalizationFactor;
143 const float imag = *inputPtr++ * iNormalizationFactor;
153 #include <pmmintrin.h> 156 #ifdef LV_HAVE_LIB_SIMDMATH 157 #include <simdmath.h> 163 const float normalizationFactor,
165 unsigned int num_points)
167 const float* inputPtr = (
const float*)complexFFTInput;
168 float* destPtr = logPowerOutput;
170 const float iRBW = 1.0 / rbw;
171 const float iNormalizationFactor = 1.0 / normalizationFactor;
173 #ifdef LV_HAVE_LIB_SIMDMATH 174 __m128 magScalar = _mm_set_ps1(10.0);
175 magScalar = _mm_div_ps(magScalar, logf4(magScalar));
177 __m128 invRBW = _mm_set_ps1(iRBW);
179 __m128 invNormalizationFactor = _mm_set_ps1(iNormalizationFactor);
182 __m128 input1, input2;
183 const uint64_t quarterPoints = num_points / 4;
184 for (; number < quarterPoints; number++) {
186 input1 = _mm_load_ps(inputPtr);
188 input2 = _mm_load_ps(inputPtr);
192 input1 = _mm_mul_ps(input1, invNormalizationFactor);
193 input2 = _mm_mul_ps(input2, invNormalizationFactor);
197 input1 = _mm_mul_ps(input1, input1);
199 input2 = _mm_mul_ps(input2, input2);
203 power = _mm_hadd_ps(input1, input2);
206 power = _mm_mul_ps(power, invRBW);
209 power = logf4(power);
212 power = _mm_mul_ps(power, magScalar);
215 _mm_store_ps(destPtr, power);
220 number = quarterPoints * 4;
223 for (; number < num_points; number++) {
230 const float real = *inputPtr++ * iNormalizationFactor;
231 const float imag = *inputPtr++ * iNormalizationFactor;
241 #ifdef LV_HAVE_GENERIC 246 const float normalizationFactor,
248 unsigned int num_points)
251 volk_32fc_s32f_power_spectrum_32f(
252 logPowerOutput, complexFFTInput, normalizationFactor * sqrt(rbw), num_points);
254 volk_32fc_s32f_power_spectrum_32f(
255 logPowerOutput, complexFFTInput, normalizationFactor, num_points);
static float log2f_non_ieee(float f)
Definition: volk_common.h:150
static void volk_32fc_s32f_x2_power_spectral_density_32f_generic(float *logPowerOutput, const lv_32fc_t *complexFFTInput, const float normalizationFactor, const float rbw, unsigned int num_points)
Definition: volk_32fc_s32f_x2_power_spectral_density_32f.h:244
#define volk_log2to10factor
Definition: volk_common.h:160
static void volk_32fc_s32f_x2_power_spectral_density_32f_a_sse3(float *logPowerOutput, const lv_32fc_t *complexFFTInput, const float normalizationFactor, const float rbw, unsigned int num_points)
Definition: volk_32fc_s32f_x2_power_spectral_density_32f.h:161
float complex lv_32fc_t
Definition: volk_complex.h:70
static void volk_32fc_s32f_x2_power_spectral_density_32f_a_avx(float *logPowerOutput, const lv_32fc_t *complexFFTInput, const float normalizationFactor, const float rbw, unsigned int num_points)
Definition: volk_32fc_s32f_x2_power_spectral_density_32f.h:70