GNU Radio 3.7.2 C++ API
volk_16ic_magnitude_16i.h
Go to the documentation of this file.
1 #ifndef INCLUDED_volk_16ic_magnitude_16i_a_H
2 #define INCLUDED_volk_16ic_magnitude_16i_a_H
3 
4 #include <volk/volk_common.h>
5 #include <inttypes.h>
6 #include <stdio.h>
7 #include <math.h>
8 
9 #ifdef LV_HAVE_SSE3
10 #include <pmmintrin.h>
11 /*!
12  \brief Calculates the magnitude of the complexVector and stores the results in the magnitudeVector
13  \param complexVector The vector containing the complex input values
14  \param magnitudeVector The vector containing the real output values
15  \param num_points The number of complex values in complexVector to be calculated and stored into cVector
16 */
17 static inline void volk_16ic_magnitude_16i_a_sse3(int16_t* magnitudeVector, const lv_16sc_t* complexVector, unsigned int num_points){
18  unsigned int number = 0;
19  const unsigned int quarterPoints = num_points / 4;
20 
21  const int16_t* complexVectorPtr = (const int16_t*)complexVector;
22  int16_t* magnitudeVectorPtr = magnitudeVector;
23 
24  __m128 vScalar = _mm_set_ps1(32768.0);
25  __m128 invScalar = _mm_set_ps1(1.0/32768.0);
26 
27  __m128 cplxValue1, cplxValue2, result;
28 
29  __VOLK_ATTR_ALIGNED(16) float inputFloatBuffer[8];
30  __VOLK_ATTR_ALIGNED(16) float outputFloatBuffer[4];
31 
32  for(;number < quarterPoints; number++){
33 
34  inputFloatBuffer[0] = (float)(complexVectorPtr[0]);
35  inputFloatBuffer[1] = (float)(complexVectorPtr[1]);
36  inputFloatBuffer[2] = (float)(complexVectorPtr[2]);
37  inputFloatBuffer[3] = (float)(complexVectorPtr[3]);
38 
39  inputFloatBuffer[4] = (float)(complexVectorPtr[4]);
40  inputFloatBuffer[5] = (float)(complexVectorPtr[5]);
41  inputFloatBuffer[6] = (float)(complexVectorPtr[6]);
42  inputFloatBuffer[7] = (float)(complexVectorPtr[7]);
43 
44  cplxValue1 = _mm_load_ps(&inputFloatBuffer[0]);
45  cplxValue2 = _mm_load_ps(&inputFloatBuffer[4]);
46 
47  complexVectorPtr += 8;
48 
49  cplxValue1 = _mm_mul_ps(cplxValue1, invScalar);
50  cplxValue2 = _mm_mul_ps(cplxValue2, invScalar);
51 
52  cplxValue1 = _mm_mul_ps(cplxValue1, cplxValue1); // Square the values
53  cplxValue2 = _mm_mul_ps(cplxValue2, cplxValue2); // Square the Values
54 
55  result = _mm_hadd_ps(cplxValue1, cplxValue2); // Add the I2 and Q2 values
56 
57  result = _mm_sqrt_ps(result); // Square root the values
58 
59  result = _mm_mul_ps(result, vScalar); // Scale the results
60 
61  _mm_store_ps(outputFloatBuffer, result);
62  *magnitudeVectorPtr++ = (int16_t)(outputFloatBuffer[0]);
63  *magnitudeVectorPtr++ = (int16_t)(outputFloatBuffer[1]);
64  *magnitudeVectorPtr++ = (int16_t)(outputFloatBuffer[2]);
65  *magnitudeVectorPtr++ = (int16_t)(outputFloatBuffer[3]);
66  }
67 
68  number = quarterPoints * 4;
69  magnitudeVectorPtr = &magnitudeVector[number];
70  complexVectorPtr = (const int16_t*)&complexVector[number];
71  for(; number < num_points; number++){
72  const float val1Real = (float)(*complexVectorPtr++) / 32768.0;
73  const float val1Imag = (float)(*complexVectorPtr++) / 32768.0;
74  const float val1Result = sqrtf((val1Real * val1Real) + (val1Imag * val1Imag)) * 32768.0;
75  *magnitudeVectorPtr++ = (int16_t)(val1Result);
76  }
77 }
78 #endif /* LV_HAVE_SSE3 */
79 
80 #ifdef LV_HAVE_SSE
81 #include <xmmintrin.h>
82 /*!
83  \brief Calculates the magnitude of the complexVector and stores the results in the magnitudeVector
84  \param complexVector The vector containing the complex input values
85  \param magnitudeVector The vector containing the real output values
86  \param num_points The number of complex values in complexVector to be calculated and stored into cVector
87 */
88 static inline void volk_16ic_magnitude_16i_a_sse(int16_t* magnitudeVector, const lv_16sc_t* complexVector, unsigned int num_points){
89  unsigned int number = 0;
90  const unsigned int quarterPoints = num_points / 4;
91 
92  const int16_t* complexVectorPtr = (const int16_t*)complexVector;
93  int16_t* magnitudeVectorPtr = magnitudeVector;
94 
95  __m128 vScalar = _mm_set_ps1(32768.0);
96  __m128 invScalar = _mm_set_ps1(1.0/32768.0);
97 
98  __m128 cplxValue1, cplxValue2, iValue, qValue, result;
99 
100  __VOLK_ATTR_ALIGNED(16) float inputFloatBuffer[4];
101  __VOLK_ATTR_ALIGNED(16) float outputFloatBuffer[4];
102 
103  for(;number < quarterPoints; number++){
104 
105  inputFloatBuffer[0] = (float)(complexVectorPtr[0]);
106  inputFloatBuffer[1] = (float)(complexVectorPtr[1]);
107  inputFloatBuffer[2] = (float)(complexVectorPtr[2]);
108  inputFloatBuffer[3] = (float)(complexVectorPtr[3]);
109 
110  cplxValue1 = _mm_load_ps(inputFloatBuffer);
111  complexVectorPtr += 4;
112 
113  inputFloatBuffer[0] = (float)(complexVectorPtr[0]);
114  inputFloatBuffer[1] = (float)(complexVectorPtr[1]);
115  inputFloatBuffer[2] = (float)(complexVectorPtr[2]);
116  inputFloatBuffer[3] = (float)(complexVectorPtr[3]);
117 
118  cplxValue2 = _mm_load_ps(inputFloatBuffer);
119  complexVectorPtr += 4;
120 
121  cplxValue1 = _mm_mul_ps(cplxValue1, invScalar);
122  cplxValue2 = _mm_mul_ps(cplxValue2, invScalar);
123 
124  // Arrange in i1i2i3i4 format
125  iValue = _mm_shuffle_ps(cplxValue1, cplxValue2, _MM_SHUFFLE(2,0,2,0));
126  // Arrange in q1q2q3q4 format
127  qValue = _mm_shuffle_ps(cplxValue1, cplxValue2, _MM_SHUFFLE(3,1,3,1));
128 
129  iValue = _mm_mul_ps(iValue, iValue); // Square the I values
130  qValue = _mm_mul_ps(qValue, qValue); // Square the Q Values
131 
132  result = _mm_add_ps(iValue, qValue); // Add the I2 and Q2 values
133 
134  result = _mm_sqrt_ps(result); // Square root the values
135 
136  result = _mm_mul_ps(result, vScalar); // Scale the results
137 
138  _mm_store_ps(outputFloatBuffer, result);
139  *magnitudeVectorPtr++ = (int16_t)(outputFloatBuffer[0]);
140  *magnitudeVectorPtr++ = (int16_t)(outputFloatBuffer[1]);
141  *magnitudeVectorPtr++ = (int16_t)(outputFloatBuffer[2]);
142  *magnitudeVectorPtr++ = (int16_t)(outputFloatBuffer[3]);
143  }
144 
145  number = quarterPoints * 4;
146  magnitudeVectorPtr = &magnitudeVector[number];
147  complexVectorPtr = (const int16_t*)&complexVector[number];
148  for(; number < num_points; number++){
149  const float val1Real = (float)(*complexVectorPtr++) / 32768.0;
150  const float val1Imag = (float)(*complexVectorPtr++) / 32768.0;
151  const float val1Result = sqrtf((val1Real * val1Real) + (val1Imag * val1Imag)) * 32768.0;
152  *magnitudeVectorPtr++ = (int16_t)(val1Result);
153  }
154 }
155 #endif /* LV_HAVE_SSE */
156 
157 #ifdef LV_HAVE_GENERIC
158 /*!
159  \brief Calculates the magnitude of the complexVector and stores the results in the magnitudeVector
160  \param complexVector The vector containing the complex input values
161  \param magnitudeVector The vector containing the real output values
162  \param num_points The number of complex values in complexVector to be calculated and stored into cVector
163 */
164 static inline void volk_16ic_magnitude_16i_generic(int16_t* magnitudeVector, const lv_16sc_t* complexVector, unsigned int num_points){
165  const int16_t* complexVectorPtr = (const int16_t*)complexVector;
166  int16_t* magnitudeVectorPtr = magnitudeVector;
167  unsigned int number = 0;
168  const float scalar = 32768.0;
169  for(number = 0; number < num_points; number++){
170  float real = ((float)(*complexVectorPtr++)) / scalar;
171  float imag = ((float)(*complexVectorPtr++)) / scalar;
172  *magnitudeVectorPtr++ = (int16_t)(sqrtf((real*real) + (imag*imag)) * scalar);
173  }
174 }
175 #endif /* LV_HAVE_GENERIC */
176 
177 #ifdef LV_HAVE_ORC_DISABLED
178 /*!
179  \brief Calculates the magnitude of the complexVector and stores the results in the magnitudeVector
180  \param complexVector The vector containing the complex input values
181  \param magnitudeVector The vector containing the real output values
182  \param num_points The number of complex values in complexVector to be calculated and stored into cVector
183 */
184 extern void volk_16ic_magnitude_16i_a_orc_impl(int16_t* magnitudeVector, const lv_16sc_t* complexVector, float scalar, unsigned int num_points);
185 static inline void volk_16ic_magnitude_16i_u_orc(int16_t* magnitudeVector, const lv_16sc_t* complexVector, unsigned int num_points){
186  volk_16ic_magnitude_16i_a_orc_impl(magnitudeVector, complexVector, 32768.0, num_points);
187 }
188 #endif /* LV_HAVE_ORC */
189 
190 
191 #endif /* INCLUDED_volk_16ic_magnitude_16i_a_H */
lv_16sc_t
short complex lv_16sc_t
Definition: volk_complex.h:53
int16_t
signed short int16_t
Definition: stdint.h:76
__VOLK_ATTR_ALIGNED
#define __VOLK_ATTR_ALIGNED(x)
Definition: volk_common.h:27
volk_common.h