-
Notifications
You must be signed in to change notification settings - Fork 2
/
quantize.cpp
229 lines (196 loc) · 6.68 KB
/
quantize.cpp
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
#include "Python.h"
#include <cmath>
#include <cstdint>
#ifdef _OPENMP
#include <omp.h>
// OpenMP scheduling method
#ifndef OMP_SCHEDULER
#define OMP_SCHEDULER dynamic
#endif
#endif
#define NO_IMPORT_ARRAY
#define PY_ARRAY_UNIQUE_SYMBOL psr_ARRAY_API
#include "numpy/arrayobject.h"
#include "numpy/npy_math.h"
#include "psr.hpp"
template<uint8_t D>
PyObject *OptimizeDataLevels(PyObject *self, PyObject *args, PyObject *kwds) {
PyObject *spectra, *bzero=NULL, *bscale=NULL, *bdata=NULL, *output;
PyArrayObject *data=NULL, *zeroF=NULL, *scaleF=NULL, *dataF=NULL;
long ij, i, j, k, nStand, nSamps, nFFT;
int nChan = 64;
char const* kwlist[] = {"spectra", "nChan", "bzero", "bscale", "bdata", NULL};
if(!PyArg_ParseTupleAndKeywords(args, kwds, "Oi|OOO", const_cast<char **>(kwlist), &spectra, &nChan, &bzero, &bscale, &bdata)) {
PyErr_Format(PyExc_RuntimeError, "Invalid parameters");
goto fail;
}
// Bring the data into C and make it usable
data = (PyArrayObject *) PyArray_ContiguousFromObject(spectra, NPY_FLOAT32, 2, 2);
if( data == NULL ) {
PyErr_Format(PyExc_RuntimeError, "Cannot cast input spectra array to 2-D float32");
goto fail;
}
// Get the properties of the data
nStand = (long) PyArray_DIM(data, 0);
nSamps = (long) PyArray_DIM(data, 1);
// Find out how large the output array needs to be and initialize it
nFFT = nSamps / nChan;
npy_intp dims[2];
dims[0] = (npy_intp) nStand;
dims[1] = (npy_intp) nChan;
if( bzero != NULL && bzero != Py_None ) {
zeroF = (PyArrayObject*) PyArray_ContiguousFromObject(bzero, NPY_FLOAT32, 2, 2);
if(zeroF == NULL) {
PyErr_Format(PyExc_RuntimeError, "Cannot cast output bzero array to 2-D float32");
goto fail;
}
if(PyArray_DIM(zeroF, 0) != dims[0]) {
PyErr_Format(PyExc_RuntimeError, "bzero has an unexpected number of stands");
goto fail;
}
if(PyArray_DIM(zeroF, 1) != dims[1]) {
PyErr_Format(PyExc_RuntimeError, "bzero has an unexpected number of channels");
goto fail;
}
} else {
zeroF = (PyArrayObject*) PyArray_ZEROS(2, dims, NPY_FLOAT32, 0);
if(zeroF == NULL) {
PyErr_Format(PyExc_MemoryError, "Cannot create output array");
goto fail;
}
}
dims[0] = (npy_intp) nStand;
dims[1] = (npy_intp) nChan;
if( bscale != NULL && bscale != Py_None ) {
scaleF = (PyArrayObject*) PyArray_ContiguousFromObject(bzero, NPY_FLOAT32, 2, 2);
if(scaleF == NULL) {
PyErr_Format(PyExc_RuntimeError, "Cannot cast output bscale array to 2-D float32");
goto fail;
}
if(PyArray_DIM(scaleF, 0) != dims[0]) {
PyErr_Format(PyExc_RuntimeError, "bscale has an unexpected number of stands");
goto fail;
}
if(PyArray_DIM(scaleF, 1) != dims[1]) {
PyErr_Format(PyExc_RuntimeError, "bscale has an unexpected number of channels");
goto fail;
}
} else {
scaleF = (PyArrayObject*) PyArray_ZEROS(2, dims, NPY_FLOAT32, 0);
if(scaleF == NULL) {
PyErr_Format(PyExc_MemoryError, "Cannot create output array");
goto fail;
}
}
dims[0] = (npy_intp) nStand;
dims[1] = (npy_intp) nSamps;
if( bdata != NULL && bdata != Py_None ) {
dataF = (PyArrayObject*) PyArray_ContiguousFromObject(bdata, NPY_UINT8, 2, 2);
if(dataF == NULL) {
PyErr_Format(PyExc_RuntimeError, "Cannot cast output bdata array to 2-D uint8");
goto fail;
}
if(PyArray_DIM(dataF, 0) != dims[0]) {
PyErr_Format(PyExc_RuntimeError, "bdata has an unexpected number of stands");
goto fail;
}
if(PyArray_DIM(dataF, 1) != dims[1]) {
PyErr_Format(PyExc_RuntimeError, "bdata has an unexpected number of channels");
goto fail;
}
} else {
dataF = (PyArrayObject*) PyArray_ZEROS(2, dims, NPY_UINT8, 0);
if(dataF == NULL) {
PyErr_Format(PyExc_MemoryError, "Cannot create output array");
goto fail;
}
}
Py_BEGIN_ALLOW_THREADS
// Go!
long secStart;
float tempMin, tempMax, tempV;
float *a, *b, *c;
uint8_t *d;
a = (float *) PyArray_DATA(data);
b = (float *) PyArray_DATA(zeroF);
c = (float *) PyArray_DATA(scaleF);
d = (uint8_t *) PyArray_DATA(dataF);
#ifdef _OPENMP
#pragma omp parallel default(shared) private(secStart, i, j, k, tempV, tempMin, tempMax)
#endif
{
#ifdef _OPENMP
#pragma omp for schedule(OMP_SCHEDULER)
#endif
for(ij=0; ij<nStand*nChan; ij++) {
i = ij / nChan;
j = ij % nChan;
secStart = nSamps*i + j;
k = 0;
tempMin = *(a + secStart + nChan*k);
tempMax = *(a + secStart + nChan*k);
for(k=1; k<nFFT; k++) {
if( *(a + secStart + nChan*k) < tempMin ) {
tempMin = *(a + secStart + nChan*k);
} else if( *(a + secStart + nChan*k) > tempMax ) {
tempMax = *(a + secStart + nChan*k);
}
}
*(b + i*nChan + j) = tempMin;
*(c + i*nChan + j) = (tempMax - tempMin) / (float) ((1<<D) - 1);
for(k=0; k<nFFT; k++) {
tempV = *(a + secStart + nChan*k) - tempMin;
tempV /= *(c + i*nChan + j);
tempV = round(tempV);
*(d + secStart + nChan*k) = ((uint8_t) tempV) & ((1<<D) - 1);
}
}
}
Py_END_ALLOW_THREADS
output = Py_BuildValue("(OOO)", PyArray_Return(zeroF), PyArray_Return(scaleF), PyArray_Return(dataF));
Py_XDECREF(data);
Py_XDECREF(zeroF);
Py_XDECREF(scaleF);
Py_XDECREF(dataF);
return output;
fail:
Py_XDECREF(data);
Py_XDECREF(zeroF);
Py_XDECREF(scaleF);
Py_XDECREF(dataF);
return NULL;
}
PyObject *OptimizeDataLevels8Bit(PyObject *self, PyObject *args, PyObject *kwds) {
return OptimizeDataLevels<8>(self, args, kwds);
}
char OptimizeDataLevels8Bit_doc[] = PyDoc_STR(\
"Given the output of one of the 'Combine' functions, find the bzero and bscale\n\
values that yield the best representation of the data as unsigned characters\n\
and scale the data accordingly.\n\
\n\
Input arguments are:\n\
* signals: 2-D numpy.float64 (stands by samples) array of combined data\n\
* LFFT: number of channels per data stream\n\
\n\
Outputs:\n\
* bzero: 2-D numpy.float32 (stands by channels) array of bzero values\n\
* bscale: 2-D numpy.float32 (stand by channels) array of bscale values\n\
* spectra: 2-D numpy.unit8 (stands by samples) array of spectra\n\
");
PyObject *OptimizeDataLevels4Bit(PyObject *self, PyObject *args, PyObject *kwds) {
return OptimizeDataLevels<4>(self, args, kwds);
}
char OptimizeDataLevels4Bit_doc[] = PyDoc_STR(\
"Given the output of one of the 'Combine' functions, find the bzero and bscale\n\
values that yield the best representation of the data as 4-bit unsigned integers\n\
and scale the data accordingly.\n\
\n\
Input arguments are:\n\
* signals: 2-D numpy.float64 (stands by samples) array of combined data\n\
* LFFT: number of channels per data stream\n\
\n\
Outputs:\n\
* bzero: 2-D numpy.float32 (stands by channels) array of bzero values\n\
* bscale: 2-D numpy.float32 (stand by channels) array of bscale values\n\
* spectra: 2-D numpy.unit8 (stands by samples) array of spectra\n\
");