56 #ifndef INCLUDED_volk_8ic_s32f_deinterleave_32f_x2_a_H
57 #define INCLUDED_volk_8ic_s32f_deinterleave_32f_x2_a_H
64 #include <smmintrin.h>
67 volk_8ic_s32f_deinterleave_32f_x2_a_sse4_1(
float* iBuffer,
float* qBuffer,
const lv_8sc_t* complexVector,
68 const float scalar,
unsigned int num_points)
70 float* iBufferPtr = iBuffer;
71 float* qBufferPtr = qBuffer;
73 unsigned int number = 0;
74 const unsigned int eighthPoints = num_points / 8;
75 __m128 iFloatValue, qFloatValue;
77 const float iScalar= 1.0 / scalar;
78 __m128 invScalar = _mm_set_ps1(iScalar);
79 __m128i complexVal, iIntVal, qIntVal, iComplexVal, qComplexVal;
82 __m128i iMoveMask = _mm_set_epi8(0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 14, 12, 10, 8, 6, 4, 2, 0);
83 __m128i qMoveMask = _mm_set_epi8(0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 15, 13, 11, 9, 7, 5, 3, 1);
85 for(;number < eighthPoints; number++){
86 complexVal = _mm_load_si128((__m128i*)complexVectorPtr); complexVectorPtr += 16;
87 iComplexVal = _mm_shuffle_epi8(complexVal, iMoveMask);
88 qComplexVal = _mm_shuffle_epi8(complexVal, qMoveMask);
90 iIntVal = _mm_cvtepi8_epi32(iComplexVal);
91 iFloatValue = _mm_cvtepi32_ps(iIntVal);
92 iFloatValue = _mm_mul_ps(iFloatValue, invScalar);
93 _mm_store_ps(iBufferPtr, iFloatValue);
96 iComplexVal = _mm_srli_si128(iComplexVal, 4);
98 iIntVal = _mm_cvtepi8_epi32(iComplexVal);
99 iFloatValue = _mm_cvtepi32_ps(iIntVal);
100 iFloatValue = _mm_mul_ps(iFloatValue, invScalar);
101 _mm_store_ps(iBufferPtr, iFloatValue);
104 qIntVal = _mm_cvtepi8_epi32(qComplexVal);
105 qFloatValue = _mm_cvtepi32_ps(qIntVal);
106 qFloatValue = _mm_mul_ps(qFloatValue, invScalar);
107 _mm_store_ps(qBufferPtr, qFloatValue);
110 qComplexVal = _mm_srli_si128(qComplexVal, 4);
112 qIntVal = _mm_cvtepi8_epi32(qComplexVal);
113 qFloatValue = _mm_cvtepi32_ps(qIntVal);
114 qFloatValue = _mm_mul_ps(qFloatValue, invScalar);
115 _mm_store_ps(qBufferPtr, qFloatValue);
120 number = eighthPoints * 8;
121 for(; number < num_points; number++){
122 *iBufferPtr++ = (float)(*complexVectorPtr++) * iScalar;
123 *qBufferPtr++ = (float)(*complexVectorPtr++) * iScalar;
131 #include <xmmintrin.h>
134 volk_8ic_s32f_deinterleave_32f_x2_a_sse(
float* iBuffer,
float* qBuffer,
136 const float scalar,
unsigned int num_points)
138 float* iBufferPtr = iBuffer;
139 float* qBufferPtr = qBuffer;
141 unsigned int number = 0;
142 const unsigned int quarterPoints = num_points / 4;
143 __m128 cplxValue1, cplxValue2, iValue, qValue;
145 __m128 invScalar = _mm_set_ps1(1.0/scalar);
150 for(;number < quarterPoints; number++){
151 floatBuffer[0] = (float)(complexVectorPtr[0]);
152 floatBuffer[1] = (float)(complexVectorPtr[1]);
153 floatBuffer[2] = (float)(complexVectorPtr[2]);
154 floatBuffer[3] = (float)(complexVectorPtr[3]);
156 floatBuffer[4] = (float)(complexVectorPtr[4]);
157 floatBuffer[5] = (float)(complexVectorPtr[5]);
158 floatBuffer[6] = (float)(complexVectorPtr[6]);
159 floatBuffer[7] = (float)(complexVectorPtr[7]);
161 cplxValue1 = _mm_load_ps(&floatBuffer[0]);
162 cplxValue2 = _mm_load_ps(&floatBuffer[4]);
164 complexVectorPtr += 8;
166 cplxValue1 = _mm_mul_ps(cplxValue1, invScalar);
167 cplxValue2 = _mm_mul_ps(cplxValue2, invScalar);
170 iValue = _mm_shuffle_ps(cplxValue1, cplxValue2, _MM_SHUFFLE(2,0,2,0));
171 qValue = _mm_shuffle_ps(cplxValue1, cplxValue2, _MM_SHUFFLE(3,1,3,1));
173 _mm_store_ps(iBufferPtr, iValue);
174 _mm_store_ps(qBufferPtr, qValue);
180 number = quarterPoints * 4;
181 complexVectorPtr = (
int8_t*)&complexVector[number];
182 for(; number < num_points; number++){
183 *iBufferPtr++ = (float)(*complexVectorPtr++) / scalar;
184 *qBufferPtr++ = (float)(*complexVectorPtr++) / scalar;
190 #ifdef LV_HAVE_GENERIC
193 volk_8ic_s32f_deinterleave_32f_x2_generic(
float* iBuffer,
float* qBuffer,
195 const float scalar,
unsigned int num_points)
197 const int8_t* complexVectorPtr = (
const int8_t*)complexVector;
198 float* iBufferPtr = iBuffer;
199 float* qBufferPtr = qBuffer;
201 const float invScalar = 1.0 / scalar;
202 for(number = 0; number < num_points; number++){
203 *iBufferPtr++ = (float)(*complexVectorPtr++)*invScalar;
204 *qBufferPtr++ = (float)(*complexVectorPtr++)*invScalar;
signed char int8_t
Definition: stdint.h:75
#define __VOLK_ATTR_ALIGNED(x)
Definition: volk_common.h:27
char complex lv_8sc_t
Provide typedefs and operators for all complex types in C and C++.
Definition: volk_complex.h:52