/src/gdal/frmts/vrt/pixelfunctions.cpp
Line | Count | Source (jump to first uncovered line) |
1 | | /****************************************************************************** |
2 | | * |
3 | | * Project: GDAL |
4 | | * Purpose: Implementation of a set of GDALDerivedPixelFunc(s) to be used |
5 | | * with source raster band of virtual GDAL datasets. |
6 | | * Author: Antonio Valentino <antonio.valentino@tiscali.it> |
7 | | * |
8 | | ****************************************************************************** |
9 | | * Copyright (c) 2008-2014,2022 Antonio Valentino <antonio.valentino@tiscali.it> |
10 | | * |
11 | | * SPDX-License-Identifier: MIT |
12 | | *****************************************************************************/ |
13 | | |
14 | | #include <array> |
15 | | #include <cmath> |
16 | | #include "gdal.h" |
17 | | #include "vrtdataset.h" |
18 | | #include "vrtexpression.h" |
19 | | #include "vrtreclassifier.h" |
20 | | #include "cpl_float.h" |
21 | | |
22 | | #if defined(__x86_64) || defined(_M_X64) || defined(USE_NEON_OPTIMIZATIONS) |
23 | | #define USE_SSE2 |
24 | | #include "gdalsse_priv.h" |
25 | | |
26 | | #if !defined(USE_NEON_OPTIMIZATIONS) |
27 | | #define LIBDIVIDE_SSE2 |
28 | | #ifdef __GNUC__ |
29 | | #pragma GCC diagnostic push |
30 | | #pragma GCC diagnostic ignored "-Wold-style-cast" |
31 | | #pragma GCC diagnostic ignored "-Weffc++" |
32 | | #endif |
33 | | #include "../../third_party/libdivide/libdivide.h" |
34 | | #ifdef __GNUC__ |
35 | | #pragma GCC diagnostic pop |
36 | | #endif |
37 | | #endif |
38 | | |
39 | | #endif |
40 | | |
41 | | #include "gdal_priv_templates.hpp" |
42 | | |
43 | | #include <algorithm> |
44 | | #include <cassert> |
45 | | #include <limits> |
46 | | |
47 | | namespace gdal |
48 | | { |
49 | 0 | MathExpression::~MathExpression() = default; |
50 | | } |
51 | | |
52 | | template <typename T> |
53 | | inline double GetSrcVal(const void *pSource, GDALDataType eSrcType, T ii) |
54 | 0 | { |
55 | 0 | switch (eSrcType) |
56 | 0 | { |
57 | 0 | case GDT_Unknown: |
58 | 0 | return 0; |
59 | 0 | case GDT_Byte: |
60 | 0 | return static_cast<const GByte *>(pSource)[ii]; |
61 | 0 | case GDT_Int8: |
62 | 0 | return static_cast<const GInt8 *>(pSource)[ii]; |
63 | 0 | case GDT_UInt16: |
64 | 0 | return static_cast<const GUInt16 *>(pSource)[ii]; |
65 | 0 | case GDT_Int16: |
66 | 0 | return static_cast<const GInt16 *>(pSource)[ii]; |
67 | 0 | case GDT_UInt32: |
68 | 0 | return static_cast<const GUInt32 *>(pSource)[ii]; |
69 | 0 | case GDT_Int32: |
70 | 0 | return static_cast<const GInt32 *>(pSource)[ii]; |
71 | | // Precision loss currently for int64/uint64 |
72 | 0 | case GDT_UInt64: |
73 | 0 | return static_cast<double>( |
74 | 0 | static_cast<const uint64_t *>(pSource)[ii]); |
75 | 0 | case GDT_Int64: |
76 | 0 | return static_cast<double>( |
77 | 0 | static_cast<const int64_t *>(pSource)[ii]); |
78 | 0 | case GDT_Float16: |
79 | 0 | return static_cast<const GFloat16 *>(pSource)[ii]; |
80 | 0 | case GDT_Float32: |
81 | 0 | return static_cast<const float *>(pSource)[ii]; |
82 | 0 | case GDT_Float64: |
83 | 0 | return static_cast<const double *>(pSource)[ii]; |
84 | 0 | case GDT_CInt16: |
85 | 0 | return static_cast<const GInt16 *>(pSource)[2 * ii]; |
86 | 0 | case GDT_CInt32: |
87 | 0 | return static_cast<const GInt32 *>(pSource)[2 * ii]; |
88 | 0 | case GDT_CFloat16: |
89 | 0 | return static_cast<const GFloat16 *>(pSource)[2 * ii]; |
90 | 0 | case GDT_CFloat32: |
91 | 0 | return static_cast<const float *>(pSource)[2 * ii]; |
92 | 0 | case GDT_CFloat64: |
93 | 0 | return static_cast<const double *>(pSource)[2 * ii]; |
94 | 0 | case GDT_TypeCount: |
95 | 0 | break; |
96 | 0 | } |
97 | 0 | return 0; |
98 | 0 | } |
99 | | |
100 | | static bool IsNoData(double dfVal, double dfNoData) |
101 | 0 | { |
102 | 0 | return dfVal == dfNoData || (std::isnan(dfVal) && std::isnan(dfNoData)); |
103 | 0 | } |
104 | | |
105 | | static CPLErr FetchDoubleArg(CSLConstList papszArgs, const char *pszName, |
106 | | double *pdfX, double *pdfDefault = nullptr) |
107 | 0 | { |
108 | 0 | const char *pszVal = CSLFetchNameValue(papszArgs, pszName); |
109 | |
|
110 | 0 | if (pszVal == nullptr) |
111 | 0 | { |
112 | 0 | if (pdfDefault == nullptr) |
113 | 0 | { |
114 | 0 | CPLError(CE_Failure, CPLE_AppDefined, |
115 | 0 | "Missing pixel function argument: %s", pszName); |
116 | 0 | return CE_Failure; |
117 | 0 | } |
118 | 0 | else |
119 | 0 | { |
120 | 0 | *pdfX = *pdfDefault; |
121 | 0 | return CE_None; |
122 | 0 | } |
123 | 0 | } |
124 | | |
125 | 0 | char *pszEnd = nullptr; |
126 | 0 | *pdfX = std::strtod(pszVal, &pszEnd); |
127 | 0 | if (pszEnd == pszVal) |
128 | 0 | { |
129 | 0 | CPLError(CE_Failure, CPLE_AppDefined, |
130 | 0 | "Failed to parse pixel function argument: %s", pszName); |
131 | 0 | return CE_Failure; |
132 | 0 | } |
133 | | |
134 | 0 | return CE_None; |
135 | 0 | } |
136 | | |
137 | | static CPLErr RealPixelFunc(void **papoSources, int nSources, void *pData, |
138 | | int nXSize, int nYSize, GDALDataType eSrcType, |
139 | | GDALDataType eBufType, int nPixelSpace, |
140 | | int nLineSpace) |
141 | 0 | { |
142 | | /* ---- Init ---- */ |
143 | 0 | if (nSources != 1) |
144 | 0 | return CE_Failure; |
145 | | |
146 | 0 | const int nPixelSpaceSrc = GDALGetDataTypeSizeBytes(eSrcType); |
147 | 0 | const size_t nLineSpaceSrc = static_cast<size_t>(nPixelSpaceSrc) * nXSize; |
148 | | |
149 | | /* ---- Set pixels ---- */ |
150 | 0 | for (int iLine = 0; iLine < nYSize; ++iLine) |
151 | 0 | { |
152 | 0 | GDALCopyWords(static_cast<GByte *>(papoSources[0]) + |
153 | 0 | nLineSpaceSrc * iLine, |
154 | 0 | eSrcType, nPixelSpaceSrc, |
155 | 0 | static_cast<GByte *>(pData) + |
156 | 0 | static_cast<GSpacing>(nLineSpace) * iLine, |
157 | 0 | eBufType, nPixelSpace, nXSize); |
158 | 0 | } |
159 | | |
160 | | /* ---- Return success ---- */ |
161 | 0 | return CE_None; |
162 | 0 | } // RealPixelFunc |
163 | | |
164 | | static CPLErr ImagPixelFunc(void **papoSources, int nSources, void *pData, |
165 | | int nXSize, int nYSize, GDALDataType eSrcType, |
166 | | GDALDataType eBufType, int nPixelSpace, |
167 | | int nLineSpace) |
168 | 0 | { |
169 | | /* ---- Init ---- */ |
170 | 0 | if (nSources != 1) |
171 | 0 | return CE_Failure; |
172 | | |
173 | 0 | if (GDALDataTypeIsComplex(eSrcType)) |
174 | 0 | { |
175 | 0 | const GDALDataType eSrcBaseType = GDALGetNonComplexDataType(eSrcType); |
176 | 0 | const int nPixelSpaceSrc = GDALGetDataTypeSizeBytes(eSrcType); |
177 | 0 | const size_t nLineSpaceSrc = |
178 | 0 | static_cast<size_t>(nPixelSpaceSrc) * nXSize; |
179 | |
|
180 | 0 | const void *const pImag = static_cast<GByte *>(papoSources[0]) + |
181 | 0 | GDALGetDataTypeSizeBytes(eSrcType) / 2; |
182 | | |
183 | | /* ---- Set pixels ---- */ |
184 | 0 | for (int iLine = 0; iLine < nYSize; ++iLine) |
185 | 0 | { |
186 | 0 | GDALCopyWords(static_cast<const GByte *>(pImag) + |
187 | 0 | nLineSpaceSrc * iLine, |
188 | 0 | eSrcBaseType, nPixelSpaceSrc, |
189 | 0 | static_cast<GByte *>(pData) + |
190 | 0 | static_cast<GSpacing>(nLineSpace) * iLine, |
191 | 0 | eBufType, nPixelSpace, nXSize); |
192 | 0 | } |
193 | 0 | } |
194 | 0 | else |
195 | 0 | { |
196 | 0 | const double dfImag = 0; |
197 | | |
198 | | /* ---- Set pixels ---- */ |
199 | 0 | for (int iLine = 0; iLine < nYSize; ++iLine) |
200 | 0 | { |
201 | | // Always copy from the same location. |
202 | 0 | GDALCopyWords(&dfImag, eSrcType, 0, |
203 | 0 | static_cast<GByte *>(pData) + |
204 | 0 | static_cast<GSpacing>(nLineSpace) * iLine, |
205 | 0 | eBufType, nPixelSpace, nXSize); |
206 | 0 | } |
207 | 0 | } |
208 | | |
209 | | /* ---- Return success ---- */ |
210 | 0 | return CE_None; |
211 | 0 | } // ImagPixelFunc |
212 | | |
213 | | static CPLErr ComplexPixelFunc(void **papoSources, int nSources, void *pData, |
214 | | int nXSize, int nYSize, GDALDataType eSrcType, |
215 | | GDALDataType eBufType, int nPixelSpace, |
216 | | int nLineSpace) |
217 | 0 | { |
218 | | /* ---- Init ---- */ |
219 | 0 | if (nSources != 2) |
220 | 0 | return CE_Failure; |
221 | | |
222 | 0 | const void *const pReal = papoSources[0]; |
223 | 0 | const void *const pImag = papoSources[1]; |
224 | | |
225 | | /* ---- Set pixels ---- */ |
226 | 0 | size_t ii = 0; |
227 | 0 | for (int iLine = 0; iLine < nYSize; ++iLine) |
228 | 0 | { |
229 | 0 | for (int iCol = 0; iCol < nXSize; ++iCol, ++ii) |
230 | 0 | { |
231 | | // Source raster pixels may be obtained with GetSrcVal macro. |
232 | 0 | const double adfPixVal[2] = { |
233 | 0 | GetSrcVal(pReal, eSrcType, ii), // re |
234 | 0 | GetSrcVal(pImag, eSrcType, ii) // im |
235 | 0 | }; |
236 | |
|
237 | 0 | GDALCopyWords(adfPixVal, GDT_CFloat64, 0, |
238 | 0 | static_cast<GByte *>(pData) + |
239 | 0 | static_cast<GSpacing>(nLineSpace) * iLine + |
240 | 0 | iCol * nPixelSpace, |
241 | 0 | eBufType, nPixelSpace, 1); |
242 | 0 | } |
243 | 0 | } |
244 | | |
245 | | /* ---- Return success ---- */ |
246 | 0 | return CE_None; |
247 | 0 | } // ComplexPixelFunc |
248 | | |
249 | | typedef enum |
250 | | { |
251 | | GAT_amplitude, |
252 | | GAT_intensity, |
253 | | GAT_dB |
254 | | } PolarAmplitudeType; |
255 | | |
256 | | static const char pszPolarPixelFuncMetadata[] = |
257 | | "<PixelFunctionArgumentsList>" |
258 | | " <Argument name='amplitude_type' description='Amplitude Type' " |
259 | | "type='string-select' default='AMPLITUDE'>" |
260 | | " <Value>INTENSITY</Value>" |
261 | | " <Value>dB</Value>" |
262 | | " <Value>AMPLITUDE</Value>" |
263 | | " </Argument>" |
264 | | "</PixelFunctionArgumentsList>"; |
265 | | |
266 | | static CPLErr PolarPixelFunc(void **papoSources, int nSources, void *pData, |
267 | | int nXSize, int nYSize, GDALDataType eSrcType, |
268 | | GDALDataType eBufType, int nPixelSpace, |
269 | | int nLineSpace, CSLConstList papszArgs) |
270 | 0 | { |
271 | | /* ---- Init ---- */ |
272 | 0 | if (nSources != 2) |
273 | 0 | return CE_Failure; |
274 | | |
275 | 0 | const char pszName[] = "amplitude_type"; |
276 | 0 | const char *pszVal = CSLFetchNameValue(papszArgs, pszName); |
277 | 0 | PolarAmplitudeType amplitudeType = GAT_amplitude; |
278 | 0 | if (pszVal != nullptr) |
279 | 0 | { |
280 | 0 | if (strcmp(pszVal, "INTENSITY") == 0) |
281 | 0 | amplitudeType = GAT_intensity; |
282 | 0 | else if (strcmp(pszVal, "dB") == 0) |
283 | 0 | amplitudeType = GAT_dB; |
284 | 0 | else if (strcmp(pszVal, "AMPLITUDE") != 0) |
285 | 0 | { |
286 | 0 | CPLError(CE_Failure, CPLE_AppDefined, |
287 | 0 | "Invalid value for pixel function argument '%s': %s", |
288 | 0 | pszName, pszVal); |
289 | 0 | return CE_Failure; |
290 | 0 | } |
291 | 0 | } |
292 | | |
293 | 0 | const void *const pAmp = papoSources[0]; |
294 | 0 | const void *const pPhase = papoSources[1]; |
295 | | |
296 | | /* ---- Set pixels ---- */ |
297 | 0 | size_t ii = 0; |
298 | 0 | for (int iLine = 0; iLine < nYSize; ++iLine) |
299 | 0 | { |
300 | 0 | for (int iCol = 0; iCol < nXSize; ++iCol, ++ii) |
301 | 0 | { |
302 | | // Source raster pixels may be obtained with GetSrcVal macro. |
303 | 0 | double dfAmp = GetSrcVal(pAmp, eSrcType, ii); |
304 | 0 | switch (amplitudeType) |
305 | 0 | { |
306 | 0 | case GAT_intensity: |
307 | | // clip to zero |
308 | 0 | dfAmp = dfAmp <= 0 ? 0 : std::sqrt(dfAmp); |
309 | 0 | break; |
310 | 0 | case GAT_dB: |
311 | 0 | dfAmp = dfAmp <= 0 |
312 | 0 | ? -std::numeric_limits<double>::infinity() |
313 | 0 | : pow(10, dfAmp / 20.); |
314 | 0 | break; |
315 | 0 | case GAT_amplitude: |
316 | 0 | break; |
317 | 0 | } |
318 | 0 | const double dfPhase = GetSrcVal(pPhase, eSrcType, ii); |
319 | 0 | const double adfPixVal[2] = { |
320 | 0 | dfAmp * std::cos(dfPhase), // re |
321 | 0 | dfAmp * std::sin(dfPhase) // im |
322 | 0 | }; |
323 | |
|
324 | 0 | GDALCopyWords(adfPixVal, GDT_CFloat64, 0, |
325 | 0 | static_cast<GByte *>(pData) + |
326 | 0 | static_cast<GSpacing>(nLineSpace) * iLine + |
327 | 0 | iCol * nPixelSpace, |
328 | 0 | eBufType, nPixelSpace, 1); |
329 | 0 | } |
330 | 0 | } |
331 | | |
332 | | /* ---- Return success ---- */ |
333 | 0 | return CE_None; |
334 | 0 | } // PolarPixelFunc |
335 | | |
336 | | static CPLErr ModulePixelFunc(void **papoSources, int nSources, void *pData, |
337 | | int nXSize, int nYSize, GDALDataType eSrcType, |
338 | | GDALDataType eBufType, int nPixelSpace, |
339 | | int nLineSpace) |
340 | 0 | { |
341 | | /* ---- Init ---- */ |
342 | 0 | if (nSources != 1) |
343 | 0 | return CE_Failure; |
344 | | |
345 | 0 | if (GDALDataTypeIsComplex(eSrcType)) |
346 | 0 | { |
347 | 0 | const void *pReal = papoSources[0]; |
348 | 0 | const void *pImag = static_cast<GByte *>(papoSources[0]) + |
349 | 0 | GDALGetDataTypeSizeBytes(eSrcType) / 2; |
350 | | |
351 | | /* ---- Set pixels ---- */ |
352 | 0 | size_t ii = 0; |
353 | 0 | for (int iLine = 0; iLine < nYSize; ++iLine) |
354 | 0 | { |
355 | 0 | for (int iCol = 0; iCol < nXSize; ++iCol, ++ii) |
356 | 0 | { |
357 | | // Source raster pixels may be obtained with GetSrcVal macro. |
358 | 0 | const double dfReal = GetSrcVal(pReal, eSrcType, ii); |
359 | 0 | const double dfImag = GetSrcVal(pImag, eSrcType, ii); |
360 | |
|
361 | 0 | const double dfPixVal = sqrt(dfReal * dfReal + dfImag * dfImag); |
362 | |
|
363 | 0 | GDALCopyWords(&dfPixVal, GDT_Float64, 0, |
364 | 0 | static_cast<GByte *>(pData) + |
365 | 0 | static_cast<GSpacing>(nLineSpace) * iLine + |
366 | 0 | iCol * nPixelSpace, |
367 | 0 | eBufType, nPixelSpace, 1); |
368 | 0 | } |
369 | 0 | } |
370 | 0 | } |
371 | 0 | else |
372 | 0 | { |
373 | | /* ---- Set pixels ---- */ |
374 | 0 | size_t ii = 0; |
375 | 0 | for (int iLine = 0; iLine < nYSize; ++iLine) |
376 | 0 | { |
377 | 0 | for (int iCol = 0; iCol < nXSize; ++iCol, ++ii) |
378 | 0 | { |
379 | | // Source raster pixels may be obtained with GetSrcVal macro. |
380 | 0 | const double dfPixVal = |
381 | 0 | fabs(GetSrcVal(papoSources[0], eSrcType, ii)); |
382 | |
|
383 | 0 | GDALCopyWords(&dfPixVal, GDT_Float64, 0, |
384 | 0 | static_cast<GByte *>(pData) + |
385 | 0 | static_cast<GSpacing>(nLineSpace) * iLine + |
386 | 0 | iCol * nPixelSpace, |
387 | 0 | eBufType, nPixelSpace, 1); |
388 | 0 | } |
389 | 0 | } |
390 | 0 | } |
391 | | |
392 | | /* ---- Return success ---- */ |
393 | 0 | return CE_None; |
394 | 0 | } // ModulePixelFunc |
395 | | |
396 | | static CPLErr PhasePixelFunc(void **papoSources, int nSources, void *pData, |
397 | | int nXSize, int nYSize, GDALDataType eSrcType, |
398 | | GDALDataType eBufType, int nPixelSpace, |
399 | | int nLineSpace) |
400 | 0 | { |
401 | | /* ---- Init ---- */ |
402 | 0 | if (nSources != 1) |
403 | 0 | return CE_Failure; |
404 | | |
405 | 0 | if (GDALDataTypeIsComplex(eSrcType)) |
406 | 0 | { |
407 | 0 | const void *const pReal = papoSources[0]; |
408 | 0 | const void *const pImag = static_cast<GByte *>(papoSources[0]) + |
409 | 0 | GDALGetDataTypeSizeBytes(eSrcType) / 2; |
410 | | |
411 | | /* ---- Set pixels ---- */ |
412 | 0 | size_t ii = 0; |
413 | 0 | for (int iLine = 0; iLine < nYSize; ++iLine) |
414 | 0 | { |
415 | 0 | for (int iCol = 0; iCol < nXSize; ++iCol, ++ii) |
416 | 0 | { |
417 | | // Source raster pixels may be obtained with GetSrcVal macro. |
418 | 0 | const double dfReal = GetSrcVal(pReal, eSrcType, ii); |
419 | 0 | const double dfImag = GetSrcVal(pImag, eSrcType, ii); |
420 | |
|
421 | 0 | const double dfPixVal = atan2(dfImag, dfReal); |
422 | |
|
423 | 0 | GDALCopyWords(&dfPixVal, GDT_Float64, 0, |
424 | 0 | static_cast<GByte *>(pData) + |
425 | 0 | static_cast<GSpacing>(nLineSpace) * iLine + |
426 | 0 | iCol * nPixelSpace, |
427 | 0 | eBufType, nPixelSpace, 1); |
428 | 0 | } |
429 | 0 | } |
430 | 0 | } |
431 | 0 | else if (GDALDataTypeIsInteger(eSrcType) && !GDALDataTypeIsSigned(eSrcType)) |
432 | 0 | { |
433 | 0 | constexpr double dfZero = 0; |
434 | 0 | for (int iLine = 0; iLine < nYSize; ++iLine) |
435 | 0 | { |
436 | 0 | GDALCopyWords(&dfZero, GDT_Float64, 0, |
437 | 0 | static_cast<GByte *>(pData) + |
438 | 0 | static_cast<GSpacing>(nLineSpace) * iLine, |
439 | 0 | eBufType, nPixelSpace, nXSize); |
440 | 0 | } |
441 | 0 | } |
442 | 0 | else |
443 | 0 | { |
444 | | /* ---- Set pixels ---- */ |
445 | 0 | size_t ii = 0; |
446 | 0 | for (int iLine = 0; iLine < nYSize; ++iLine) |
447 | 0 | { |
448 | 0 | for (int iCol = 0; iCol < nXSize; ++iCol, ++ii) |
449 | 0 | { |
450 | 0 | const void *const pReal = papoSources[0]; |
451 | | |
452 | | // Source raster pixels may be obtained with GetSrcVal macro. |
453 | 0 | const double dfReal = GetSrcVal(pReal, eSrcType, ii); |
454 | 0 | const double dfPixVal = (dfReal < 0) ? M_PI : 0.0; |
455 | |
|
456 | 0 | GDALCopyWords(&dfPixVal, GDT_Float64, 0, |
457 | 0 | static_cast<GByte *>(pData) + |
458 | 0 | static_cast<GSpacing>(nLineSpace) * iLine + |
459 | 0 | iCol * nPixelSpace, |
460 | 0 | eBufType, nPixelSpace, 1); |
461 | 0 | } |
462 | 0 | } |
463 | 0 | } |
464 | | |
465 | | /* ---- Return success ---- */ |
466 | 0 | return CE_None; |
467 | 0 | } // PhasePixelFunc |
468 | | |
469 | | static CPLErr ConjPixelFunc(void **papoSources, int nSources, void *pData, |
470 | | int nXSize, int nYSize, GDALDataType eSrcType, |
471 | | GDALDataType eBufType, int nPixelSpace, |
472 | | int nLineSpace) |
473 | 0 | { |
474 | | /* ---- Init ---- */ |
475 | 0 | if (nSources != 1) |
476 | 0 | return CE_Failure; |
477 | | |
478 | 0 | if (GDALDataTypeIsComplex(eSrcType) && GDALDataTypeIsComplex(eBufType)) |
479 | 0 | { |
480 | 0 | const int nOffset = GDALGetDataTypeSizeBytes(eSrcType) / 2; |
481 | 0 | const void *const pReal = papoSources[0]; |
482 | 0 | const void *const pImag = |
483 | 0 | static_cast<GByte *>(papoSources[0]) + nOffset; |
484 | | |
485 | | /* ---- Set pixels ---- */ |
486 | 0 | size_t ii = 0; |
487 | 0 | for (int iLine = 0; iLine < nYSize; ++iLine) |
488 | 0 | { |
489 | 0 | for (int iCol = 0; iCol < nXSize; ++iCol, ++ii) |
490 | 0 | { |
491 | | // Source raster pixels may be obtained with GetSrcVal macro. |
492 | 0 | const double adfPixVal[2] = { |
493 | 0 | +GetSrcVal(pReal, eSrcType, ii), // re |
494 | 0 | -GetSrcVal(pImag, eSrcType, ii) // im |
495 | 0 | }; |
496 | |
|
497 | 0 | GDALCopyWords(adfPixVal, GDT_CFloat64, 0, |
498 | 0 | static_cast<GByte *>(pData) + |
499 | 0 | static_cast<GSpacing>(nLineSpace) * iLine + |
500 | 0 | iCol * nPixelSpace, |
501 | 0 | eBufType, nPixelSpace, 1); |
502 | 0 | } |
503 | 0 | } |
504 | 0 | } |
505 | 0 | else |
506 | 0 | { |
507 | | // No complex data type. |
508 | 0 | return RealPixelFunc(papoSources, nSources, pData, nXSize, nYSize, |
509 | 0 | eSrcType, eBufType, nPixelSpace, nLineSpace); |
510 | 0 | } |
511 | | |
512 | | /* ---- Return success ---- */ |
513 | 0 | return CE_None; |
514 | 0 | } // ConjPixelFunc |
515 | | |
516 | | #ifdef USE_SSE2 |
517 | | |
518 | | /************************************************************************/ |
519 | | /* OptimizedSumToFloat_SSE2() */ |
520 | | /************************************************************************/ |
521 | | |
522 | | template <typename Tsrc> |
523 | | static void OptimizedSumToFloat_SSE2(double dfK, void *pOutBuffer, |
524 | | int nLineSpace, int nXSize, int nYSize, |
525 | | int nSources, |
526 | | const void *const *papoSources) |
527 | 0 | { |
528 | 0 | const XMMReg4Float cst = XMMReg4Float::Set1(static_cast<float>(dfK)); |
529 | |
|
530 | 0 | for (int iLine = 0; iLine < nYSize; ++iLine) |
531 | 0 | { |
532 | 0 | float *CPL_RESTRICT const pDest = reinterpret_cast<float *>( |
533 | 0 | static_cast<GByte *>(pOutBuffer) + |
534 | 0 | static_cast<GSpacing>(nLineSpace) * iLine); |
535 | 0 | const size_t iOffsetLine = static_cast<size_t>(iLine) * nXSize; |
536 | |
|
537 | 0 | constexpr int VALUES_PER_REG = 4; |
538 | 0 | constexpr int UNROLLING = 4 * VALUES_PER_REG; |
539 | 0 | int iCol = 0; |
540 | 0 | for (; iCol < nXSize - (UNROLLING - 1); iCol += UNROLLING) |
541 | 0 | { |
542 | 0 | XMMReg4Float d0(cst); |
543 | 0 | XMMReg4Float d1(cst); |
544 | 0 | XMMReg4Float d2(cst); |
545 | 0 | XMMReg4Float d3(cst); |
546 | 0 | for (int iSrc = 0; iSrc < nSources; ++iSrc) |
547 | 0 | { |
548 | 0 | XMMReg4Float t0, t1, t2, t3; |
549 | 0 | XMMReg4Float::Load16Val( |
550 | 0 | static_cast<const Tsrc * CPL_RESTRICT>(papoSources[iSrc]) + |
551 | 0 | iOffsetLine + iCol, |
552 | 0 | t0, t1, t2, t3); |
553 | 0 | d0 += t0; |
554 | 0 | d1 += t1; |
555 | 0 | d2 += t2; |
556 | 0 | d3 += t3; |
557 | 0 | } |
558 | 0 | d0.Store4Val(pDest + iCol + VALUES_PER_REG * 0); |
559 | 0 | d1.Store4Val(pDest + iCol + VALUES_PER_REG * 1); |
560 | 0 | d2.Store4Val(pDest + iCol + VALUES_PER_REG * 2); |
561 | 0 | d3.Store4Val(pDest + iCol + VALUES_PER_REG * 3); |
562 | 0 | } |
563 | |
|
564 | 0 | for (; iCol < nXSize; iCol++) |
565 | 0 | { |
566 | 0 | float d = static_cast<float>(dfK); |
567 | 0 | for (int iSrc = 0; iSrc < nSources; ++iSrc) |
568 | 0 | { |
569 | 0 | d += static_cast<const Tsrc * CPL_RESTRICT>( |
570 | 0 | papoSources[iSrc])[iOffsetLine + iCol]; |
571 | 0 | } |
572 | 0 | pDest[iCol] = d; |
573 | 0 | } |
574 | 0 | } |
575 | 0 | } Unexecuted instantiation: pixelfunctions.cpp:void OptimizedSumToFloat_SSE2<unsigned char>(double, void*, int, int, int, int, void const* const*) Unexecuted instantiation: pixelfunctions.cpp:void OptimizedSumToFloat_SSE2<unsigned short>(double, void*, int, int, int, int, void const* const*) Unexecuted instantiation: pixelfunctions.cpp:void OptimizedSumToFloat_SSE2<short>(double, void*, int, int, int, int, void const* const*) Unexecuted instantiation: pixelfunctions.cpp:void OptimizedSumToFloat_SSE2<int>(double, void*, int, int, int, int, void const* const*) Unexecuted instantiation: pixelfunctions.cpp:void OptimizedSumToFloat_SSE2<float>(double, void*, int, int, int, int, void const* const*) |
576 | | |
577 | | /************************************************************************/ |
578 | | /* OptimizedSumToDouble_SSE2() */ |
579 | | /************************************************************************/ |
580 | | |
581 | | template <typename Tsrc> |
582 | | static void OptimizedSumToDouble_SSE2(double dfK, void *pOutBuffer, |
583 | | int nLineSpace, int nXSize, int nYSize, |
584 | | int nSources, |
585 | | const void *const *papoSources) |
586 | 0 | { |
587 | 0 | const XMMReg4Double cst = XMMReg4Double::Set1(dfK); |
588 | |
|
589 | 0 | for (int iLine = 0; iLine < nYSize; ++iLine) |
590 | 0 | { |
591 | 0 | double *CPL_RESTRICT const pDest = reinterpret_cast<double *>( |
592 | 0 | static_cast<GByte *>(pOutBuffer) + |
593 | 0 | static_cast<GSpacing>(nLineSpace) * iLine); |
594 | 0 | const size_t iOffsetLine = static_cast<size_t>(iLine) * nXSize; |
595 | |
|
596 | 0 | constexpr int VALUES_PER_REG = 4; |
597 | 0 | constexpr int UNROLLING = 2 * VALUES_PER_REG; |
598 | 0 | int iCol = 0; |
599 | 0 | for (; iCol < nXSize - (UNROLLING - 1); iCol += UNROLLING) |
600 | 0 | { |
601 | 0 | XMMReg4Double d0(cst); |
602 | 0 | XMMReg4Double d1(cst); |
603 | 0 | for (int iSrc = 0; iSrc < nSources; ++iSrc) |
604 | 0 | { |
605 | 0 | XMMReg4Double t0, t1; |
606 | 0 | XMMReg4Double::Load8Val( |
607 | 0 | static_cast<const Tsrc * CPL_RESTRICT>(papoSources[iSrc]) + |
608 | 0 | iOffsetLine + iCol, |
609 | 0 | t0, t1); |
610 | 0 | d0 += t0; |
611 | 0 | d1 += t1; |
612 | 0 | } |
613 | 0 | d0.Store4Val(pDest + iCol + VALUES_PER_REG * 0); |
614 | 0 | d1.Store4Val(pDest + iCol + VALUES_PER_REG * 1); |
615 | 0 | } |
616 | |
|
617 | 0 | for (; iCol < nXSize; iCol++) |
618 | 0 | { |
619 | 0 | double d = dfK; |
620 | 0 | for (int iSrc = 0; iSrc < nSources; ++iSrc) |
621 | 0 | { |
622 | 0 | d += static_cast<const Tsrc * CPL_RESTRICT>( |
623 | 0 | papoSources[iSrc])[iOffsetLine + iCol]; |
624 | 0 | } |
625 | 0 | pDest[iCol] = d; |
626 | 0 | } |
627 | 0 | } |
628 | 0 | } Unexecuted instantiation: pixelfunctions.cpp:void OptimizedSumToDouble_SSE2<unsigned char>(double, void*, int, int, int, int, void const* const*) Unexecuted instantiation: pixelfunctions.cpp:void OptimizedSumToDouble_SSE2<unsigned short>(double, void*, int, int, int, int, void const* const*) Unexecuted instantiation: pixelfunctions.cpp:void OptimizedSumToDouble_SSE2<short>(double, void*, int, int, int, int, void const* const*) Unexecuted instantiation: pixelfunctions.cpp:void OptimizedSumToDouble_SSE2<int>(double, void*, int, int, int, int, void const* const*) Unexecuted instantiation: pixelfunctions.cpp:void OptimizedSumToDouble_SSE2<float>(double, void*, int, int, int, int, void const* const*) Unexecuted instantiation: pixelfunctions.cpp:void OptimizedSumToDouble_SSE2<double>(double, void*, int, int, int, int, void const* const*) |
629 | | |
630 | | /************************************************************************/ |
631 | | /* OptimizedSumSameType_SSE2() */ |
632 | | /************************************************************************/ |
633 | | |
634 | | template <typename T, typename Tsigned, typename Tacc, class SSEWrapper> |
635 | | static void OptimizedSumSameType_SSE2(double dfK, void *pOutBuffer, |
636 | | int nLineSpace, int nXSize, int nYSize, |
637 | | int nSources, |
638 | | const void *const *papoSources) |
639 | 0 | { |
640 | 0 | static_assert(std::numeric_limits<T>::is_integer); |
641 | 0 | static_assert(!std::numeric_limits<T>::is_signed); |
642 | 0 | static_assert(std::numeric_limits<Tsigned>::is_integer); |
643 | 0 | static_assert(std::numeric_limits<Tsigned>::is_signed); |
644 | 0 | static_assert(sizeof(T) == sizeof(Tsigned)); |
645 | 0 | const T nK = static_cast<T>(dfK); |
646 | 0 | Tsigned nKSigned; |
647 | 0 | memcpy(&nKSigned, &nK, sizeof(T)); |
648 | 0 | const __m128i valInit = SSEWrapper::Set1(nKSigned); |
649 | 0 | constexpr int VALUES_PER_REG = |
650 | 0 | static_cast<int>(sizeof(valInit) / sizeof(T)); |
651 | 0 | for (int iLine = 0; iLine < nYSize; ++iLine) |
652 | 0 | { |
653 | 0 | T *CPL_RESTRICT const pDest = |
654 | 0 | reinterpret_cast<T *>(static_cast<GByte *>(pOutBuffer) + |
655 | 0 | static_cast<GSpacing>(nLineSpace) * iLine); |
656 | 0 | const size_t iOffsetLine = static_cast<size_t>(iLine) * nXSize; |
657 | 0 | int iCol = 0; |
658 | 0 | for (; iCol < nXSize - (4 * VALUES_PER_REG - 1); |
659 | 0 | iCol += 4 * VALUES_PER_REG) |
660 | 0 | { |
661 | 0 | __m128i reg0 = valInit; |
662 | 0 | __m128i reg1 = valInit; |
663 | 0 | __m128i reg2 = valInit; |
664 | 0 | __m128i reg3 = valInit; |
665 | 0 | for (int iSrc = 0; iSrc < nSources; ++iSrc) |
666 | 0 | { |
667 | 0 | reg0 = SSEWrapper::AddSaturate( |
668 | 0 | reg0, |
669 | 0 | _mm_loadu_si128(reinterpret_cast<const __m128i *>( |
670 | 0 | static_cast<const T * CPL_RESTRICT>(papoSources[iSrc]) + |
671 | 0 | iOffsetLine + iCol))); |
672 | 0 | reg1 = SSEWrapper::AddSaturate( |
673 | 0 | reg1, |
674 | 0 | _mm_loadu_si128(reinterpret_cast<const __m128i *>( |
675 | 0 | static_cast<const T * CPL_RESTRICT>(papoSources[iSrc]) + |
676 | 0 | iOffsetLine + iCol + VALUES_PER_REG))); |
677 | 0 | reg2 = SSEWrapper::AddSaturate( |
678 | 0 | reg2, |
679 | 0 | _mm_loadu_si128(reinterpret_cast<const __m128i *>( |
680 | 0 | static_cast<const T * CPL_RESTRICT>(papoSources[iSrc]) + |
681 | 0 | iOffsetLine + iCol + 2 * VALUES_PER_REG))); |
682 | 0 | reg3 = SSEWrapper::AddSaturate( |
683 | 0 | reg3, |
684 | 0 | _mm_loadu_si128(reinterpret_cast<const __m128i *>( |
685 | 0 | static_cast<const T * CPL_RESTRICT>(papoSources[iSrc]) + |
686 | 0 | iOffsetLine + iCol + 3 * VALUES_PER_REG))); |
687 | 0 | } |
688 | 0 | _mm_storeu_si128(reinterpret_cast<__m128i *>(pDest + iCol), reg0); |
689 | 0 | _mm_storeu_si128( |
690 | 0 | reinterpret_cast<__m128i *>(pDest + iCol + VALUES_PER_REG), |
691 | 0 | reg1); |
692 | 0 | _mm_storeu_si128( |
693 | 0 | reinterpret_cast<__m128i *>(pDest + iCol + 2 * VALUES_PER_REG), |
694 | 0 | reg2); |
695 | 0 | _mm_storeu_si128( |
696 | 0 | reinterpret_cast<__m128i *>(pDest + iCol + 3 * VALUES_PER_REG), |
697 | 0 | reg3); |
698 | 0 | } |
699 | 0 | for (; iCol < nXSize; ++iCol) |
700 | 0 | { |
701 | 0 | Tacc nAcc = nK; |
702 | 0 | for (int iSrc = 0; iSrc < nSources; ++iSrc) |
703 | 0 | { |
704 | 0 | nAcc = std::min<Tacc>( |
705 | 0 | nAcc + static_cast<const T * CPL_RESTRICT>( |
706 | 0 | papoSources[iSrc])[iOffsetLine + iCol], |
707 | 0 | std::numeric_limits<T>::max()); |
708 | 0 | } |
709 | 0 | pDest[iCol] = static_cast<T>(nAcc); |
710 | 0 | } |
711 | 0 | } |
712 | 0 | } Unexecuted instantiation: pixelfunctions.cpp:void OptimizedSumSameType_SSE2<unsigned char, signed char, unsigned int, SumPixelFunc(void**, int, void*, int, int, GDALDataType, GDALDataType, int, int, char const* const*)::SSEWrapper>(double, void*, int, int, int, int, void const* const*) Unexecuted instantiation: pixelfunctions.cpp:void OptimizedSumSameType_SSE2<unsigned short, short, unsigned int, SumPixelFunc(void**, int, void*, int, int, GDALDataType, GDALDataType, int, int, char const* const*)::SSEWrapper>(double, void*, int, int, int, int, void const* const*) |
713 | | #endif // USE_SSE2 |
714 | | |
715 | | /************************************************************************/ |
716 | | /* OptimizedSumPackedOutput() */ |
717 | | /************************************************************************/ |
718 | | |
719 | | template <typename Tsrc, typename Tdest> |
720 | | static void OptimizedSumPackedOutput(double dfK, void *pOutBuffer, |
721 | | int nLineSpace, int nXSize, int nYSize, |
722 | | int nSources, |
723 | | const void *const *papoSources) |
724 | 0 | { |
725 | 0 | #ifdef USE_SSE2 |
726 | | if constexpr (std::is_same_v<Tdest, float> && !std::is_same_v<Tsrc, double>) |
727 | 0 | { |
728 | 0 | OptimizedSumToFloat_SSE2<Tsrc>(dfK, pOutBuffer, nLineSpace, nXSize, |
729 | 0 | nYSize, nSources, papoSources); |
730 | | } |
731 | | else if constexpr (std::is_same_v<Tdest, double>) |
732 | 0 | { |
733 | 0 | OptimizedSumToDouble_SSE2<Tsrc>(dfK, pOutBuffer, nLineSpace, nXSize, |
734 | 0 | nYSize, nSources, papoSources); |
735 | | } |
736 | | else |
737 | | #endif // USE_SSE2 |
738 | 0 | { |
739 | 0 | const Tdest nCst = static_cast<Tdest>(dfK); |
740 | 0 | for (int iLine = 0; iLine < nYSize; ++iLine) |
741 | 0 | { |
742 | 0 | Tdest *CPL_RESTRICT const pDest = reinterpret_cast<Tdest *>( |
743 | 0 | static_cast<GByte *>(pOutBuffer) + |
744 | 0 | static_cast<GSpacing>(nLineSpace) * iLine); |
745 | 0 | const size_t iOffsetLine = static_cast<size_t>(iLine) * nXSize; |
746 | |
|
747 | 0 | #define LOAD_SRCVAL(iSrc_, j_) \ |
748 | 0 | static_cast<Tdest>(static_cast<const Tsrc * CPL_RESTRICT>( \ |
749 | 0 | papoSources[(iSrc_)])[iOffsetLine + iCol + (j_)]) |
750 | |
|
751 | 0 | constexpr int UNROLLING = 4; |
752 | 0 | int iCol = 0; |
753 | 0 | for (; iCol < nXSize - (UNROLLING - 1); iCol += UNROLLING) |
754 | 0 | { |
755 | 0 | Tdest d[4] = {nCst, nCst, nCst, nCst}; |
756 | 0 | for (int iSrc = 0; iSrc < nSources; ++iSrc) |
757 | 0 | { |
758 | 0 | d[0] += LOAD_SRCVAL(iSrc, 0); |
759 | 0 | d[1] += LOAD_SRCVAL(iSrc, 1); |
760 | 0 | d[2] += LOAD_SRCVAL(iSrc, 2); |
761 | 0 | d[3] += LOAD_SRCVAL(iSrc, 3); |
762 | 0 | } |
763 | 0 | pDest[iCol + 0] = d[0]; |
764 | 0 | pDest[iCol + 1] = d[1]; |
765 | 0 | pDest[iCol + 2] = d[2]; |
766 | 0 | pDest[iCol + 3] = d[3]; |
767 | 0 | } |
768 | 0 | for (; iCol < nXSize; iCol++) |
769 | 0 | { |
770 | 0 | Tdest d0 = nCst; |
771 | 0 | for (int iSrc = 0; iSrc < nSources; ++iSrc) |
772 | 0 | { |
773 | 0 | d0 += LOAD_SRCVAL(iSrc, 0); |
774 | 0 | } |
775 | 0 | pDest[iCol] = d0; |
776 | 0 | } |
777 | 0 | #undef LOAD_SRCVAL |
778 | 0 | } |
779 | 0 | } |
780 | 0 | } Unexecuted instantiation: pixelfunctions.cpp:void OptimizedSumPackedOutput<unsigned char, float>(double, void*, int, int, int, int, void const* const*) Unexecuted instantiation: pixelfunctions.cpp:void OptimizedSumPackedOutput<unsigned short, float>(double, void*, int, int, int, int, void const* const*) Unexecuted instantiation: pixelfunctions.cpp:void OptimizedSumPackedOutput<short, float>(double, void*, int, int, int, int, void const* const*) Unexecuted instantiation: pixelfunctions.cpp:void OptimizedSumPackedOutput<int, float>(double, void*, int, int, int, int, void const* const*) Unexecuted instantiation: pixelfunctions.cpp:void OptimizedSumPackedOutput<float, float>(double, void*, int, int, int, int, void const* const*) Unexecuted instantiation: pixelfunctions.cpp:void OptimizedSumPackedOutput<double, float>(double, void*, int, int, int, int, void const* const*) Unexecuted instantiation: pixelfunctions.cpp:void OptimizedSumPackedOutput<unsigned char, double>(double, void*, int, int, int, int, void const* const*) Unexecuted instantiation: pixelfunctions.cpp:void OptimizedSumPackedOutput<unsigned short, double>(double, void*, int, int, int, int, void const* const*) Unexecuted instantiation: pixelfunctions.cpp:void OptimizedSumPackedOutput<short, double>(double, void*, int, int, int, int, void const* const*) Unexecuted instantiation: pixelfunctions.cpp:void OptimizedSumPackedOutput<int, double>(double, void*, int, int, int, int, void const* const*) Unexecuted instantiation: pixelfunctions.cpp:void OptimizedSumPackedOutput<float, double>(double, void*, int, int, int, int, void const* const*) Unexecuted instantiation: pixelfunctions.cpp:void OptimizedSumPackedOutput<double, double>(double, void*, int, int, int, int, void const* const*) Unexecuted instantiation: pixelfunctions.cpp:void OptimizedSumPackedOutput<unsigned char, int>(double, void*, int, int, int, int, void const* const*) |
781 | | |
782 | | /************************************************************************/ |
783 | | /* OptimizedSumPackedOutput() */ |
784 | | /************************************************************************/ |
785 | | |
786 | | template <typename Tdest> |
787 | | static bool OptimizedSumPackedOutput(GDALDataType eSrcType, double dfK, |
788 | | void *pOutBuffer, int nLineSpace, |
789 | | int nXSize, int nYSize, int nSources, |
790 | | const void *const *papoSources) |
791 | 0 | { |
792 | 0 | switch (eSrcType) |
793 | 0 | { |
794 | 0 | case GDT_Byte: |
795 | 0 | OptimizedSumPackedOutput<uint8_t, Tdest>(dfK, pOutBuffer, |
796 | 0 | nLineSpace, nXSize, nYSize, |
797 | 0 | nSources, papoSources); |
798 | 0 | return true; |
799 | | |
800 | 0 | case GDT_UInt16: |
801 | 0 | OptimizedSumPackedOutput<uint16_t, Tdest>( |
802 | 0 | dfK, pOutBuffer, nLineSpace, nXSize, nYSize, nSources, |
803 | 0 | papoSources); |
804 | 0 | return true; |
805 | | |
806 | 0 | case GDT_Int16: |
807 | 0 | OptimizedSumPackedOutput<int16_t, Tdest>(dfK, pOutBuffer, |
808 | 0 | nLineSpace, nXSize, nYSize, |
809 | 0 | nSources, papoSources); |
810 | 0 | return true; |
811 | | |
812 | 0 | case GDT_Int32: |
813 | 0 | OptimizedSumPackedOutput<int32_t, Tdest>(dfK, pOutBuffer, |
814 | 0 | nLineSpace, nXSize, nYSize, |
815 | 0 | nSources, papoSources); |
816 | 0 | return true; |
817 | | |
818 | 0 | case GDT_Float32: |
819 | 0 | OptimizedSumPackedOutput<float, Tdest>(dfK, pOutBuffer, nLineSpace, |
820 | 0 | nXSize, nYSize, nSources, |
821 | 0 | papoSources); |
822 | 0 | return true; |
823 | | |
824 | 0 | case GDT_Float64: |
825 | 0 | OptimizedSumPackedOutput<double, Tdest>(dfK, pOutBuffer, nLineSpace, |
826 | 0 | nXSize, nYSize, nSources, |
827 | 0 | papoSources); |
828 | 0 | return true; |
829 | | |
830 | 0 | default: |
831 | 0 | break; |
832 | 0 | } |
833 | 0 | return false; |
834 | 0 | } Unexecuted instantiation: pixelfunctions.cpp:bool OptimizedSumPackedOutput<float>(GDALDataType, double, void*, int, int, int, int, void const* const*) Unexecuted instantiation: pixelfunctions.cpp:bool OptimizedSumPackedOutput<double>(GDALDataType, double, void*, int, int, int, int, void const* const*) |
835 | | |
836 | | /************************************************************************/ |
837 | | /* OptimizedSumThroughLargerType() */ |
838 | | /************************************************************************/ |
839 | | |
840 | | namespace |
841 | | { |
842 | | template <typename Tsrc, typename Tdest, typename Enable = void> |
843 | | struct TintermediateS |
844 | | { |
845 | | using type = double; |
846 | | }; |
847 | | |
848 | | template <typename Tsrc, typename Tdest> |
849 | | struct TintermediateS< |
850 | | Tsrc, Tdest, |
851 | | std::enable_if_t< |
852 | | (std::is_same_v<Tsrc, uint8_t> || std::is_same_v<Tsrc, int16_t> || |
853 | | std::is_same_v<Tsrc, uint16_t>)&&(std::is_same_v<Tdest, uint8_t> || |
854 | | std::is_same_v<Tdest, int16_t> || |
855 | | std::is_same_v<Tdest, uint16_t>), |
856 | | bool>> |
857 | | { |
858 | | using type = int32_t; |
859 | | }; |
860 | | |
861 | | } // namespace |
862 | | |
863 | | template <typename Tsrc, typename Tdest> |
864 | | static bool OptimizedSumThroughLargerType(double dfK, void *pOutBuffer, |
865 | | int nPixelSpace, int nLineSpace, |
866 | | int nXSize, int nYSize, int nSources, |
867 | | const void *const *papoSources) |
868 | 0 | { |
869 | 0 | using Tintermediate = typename TintermediateS<Tsrc, Tdest>::type; |
870 | 0 | const Tintermediate k = static_cast<Tintermediate>(dfK); |
871 | |
|
872 | 0 | size_t ii = 0; |
873 | 0 | for (int iLine = 0; iLine < nYSize; ++iLine) |
874 | 0 | { |
875 | 0 | GByte *CPL_RESTRICT pDest = static_cast<GByte *>(pOutBuffer) + |
876 | 0 | static_cast<GSpacing>(nLineSpace) * iLine; |
877 | |
|
878 | 0 | constexpr int UNROLLING = 4; |
879 | 0 | int iCol = 0; |
880 | 0 | for (; iCol < nXSize - (UNROLLING - 1); |
881 | 0 | iCol += UNROLLING, ii += UNROLLING) |
882 | 0 | { |
883 | 0 | Tintermediate aSum[4] = {k, k, k, k}; |
884 | |
|
885 | 0 | for (int iSrc = 0; iSrc < nSources; ++iSrc) |
886 | 0 | { |
887 | 0 | aSum[0] += static_cast<const Tsrc *>(papoSources[iSrc])[ii + 0]; |
888 | 0 | aSum[1] += static_cast<const Tsrc *>(papoSources[iSrc])[ii + 1]; |
889 | 0 | aSum[2] += static_cast<const Tsrc *>(papoSources[iSrc])[ii + 2]; |
890 | 0 | aSum[3] += static_cast<const Tsrc *>(papoSources[iSrc])[ii + 3]; |
891 | 0 | } |
892 | |
|
893 | 0 | GDALCopyWord(aSum[0], *reinterpret_cast<Tdest *>(pDest)); |
894 | 0 | pDest += nPixelSpace; |
895 | 0 | GDALCopyWord(aSum[1], *reinterpret_cast<Tdest *>(pDest)); |
896 | 0 | pDest += nPixelSpace; |
897 | 0 | GDALCopyWord(aSum[2], *reinterpret_cast<Tdest *>(pDest)); |
898 | 0 | pDest += nPixelSpace; |
899 | 0 | GDALCopyWord(aSum[3], *reinterpret_cast<Tdest *>(pDest)); |
900 | 0 | pDest += nPixelSpace; |
901 | 0 | } |
902 | 0 | for (; iCol < nXSize; ++iCol, ++ii, pDest += nPixelSpace) |
903 | 0 | { |
904 | 0 | Tintermediate sum = k; |
905 | 0 | for (int iSrc = 0; iSrc < nSources; ++iSrc) |
906 | 0 | { |
907 | 0 | sum += static_cast<const Tsrc *>(papoSources[iSrc])[ii]; |
908 | 0 | } |
909 | |
|
910 | 0 | auto pDst = reinterpret_cast<Tdest *>(pDest); |
911 | 0 | GDALCopyWord(sum, *pDst); |
912 | 0 | } |
913 | 0 | } |
914 | 0 | return true; |
915 | 0 | } Unexecuted instantiation: pixelfunctions.cpp:bool OptimizedSumThroughLargerType<unsigned char, unsigned char>(double, void*, int, int, int, int, int, void const* const*) Unexecuted instantiation: pixelfunctions.cpp:bool OptimizedSumThroughLargerType<unsigned char, unsigned short>(double, void*, int, int, int, int, int, void const* const*) Unexecuted instantiation: pixelfunctions.cpp:bool OptimizedSumThroughLargerType<unsigned char, short>(double, void*, int, int, int, int, int, void const* const*) Unexecuted instantiation: pixelfunctions.cpp:bool OptimizedSumThroughLargerType<unsigned char, int>(double, void*, int, int, int, int, int, void const* const*) Unexecuted instantiation: pixelfunctions.cpp:bool OptimizedSumThroughLargerType<unsigned short, unsigned char>(double, void*, int, int, int, int, int, void const* const*) Unexecuted instantiation: pixelfunctions.cpp:bool OptimizedSumThroughLargerType<unsigned short, unsigned short>(double, void*, int, int, int, int, int, void const* const*) Unexecuted instantiation: pixelfunctions.cpp:bool OptimizedSumThroughLargerType<unsigned short, short>(double, void*, int, int, int, int, int, void const* const*) Unexecuted instantiation: pixelfunctions.cpp:bool OptimizedSumThroughLargerType<unsigned short, int>(double, void*, int, int, int, int, int, void const* const*) Unexecuted instantiation: pixelfunctions.cpp:bool OptimizedSumThroughLargerType<short, unsigned char>(double, void*, int, int, int, int, int, void const* const*) Unexecuted instantiation: pixelfunctions.cpp:bool OptimizedSumThroughLargerType<short, unsigned short>(double, void*, int, int, int, int, int, void const* const*) Unexecuted instantiation: pixelfunctions.cpp:bool OptimizedSumThroughLargerType<short, short>(double, void*, int, int, int, int, int, void const* const*) Unexecuted instantiation: pixelfunctions.cpp:bool OptimizedSumThroughLargerType<short, int>(double, void*, int, int, int, int, int, void const* const*) Unexecuted instantiation: pixelfunctions.cpp:bool OptimizedSumThroughLargerType<int, unsigned char>(double, void*, int, int, int, int, int, void const* const*) Unexecuted instantiation: pixelfunctions.cpp:bool OptimizedSumThroughLargerType<int, unsigned short>(double, void*, int, int, int, int, int, void const* const*) Unexecuted instantiation: pixelfunctions.cpp:bool OptimizedSumThroughLargerType<int, short>(double, void*, int, int, int, int, int, void const* const*) Unexecuted instantiation: pixelfunctions.cpp:bool OptimizedSumThroughLargerType<int, int>(double, void*, int, int, int, int, int, void const* const*) Unexecuted instantiation: pixelfunctions.cpp:bool OptimizedSumThroughLargerType<float, unsigned char>(double, void*, int, int, int, int, int, void const* const*) Unexecuted instantiation: pixelfunctions.cpp:bool OptimizedSumThroughLargerType<float, unsigned short>(double, void*, int, int, int, int, int, void const* const*) Unexecuted instantiation: pixelfunctions.cpp:bool OptimizedSumThroughLargerType<float, short>(double, void*, int, int, int, int, int, void const* const*) Unexecuted instantiation: pixelfunctions.cpp:bool OptimizedSumThroughLargerType<float, int>(double, void*, int, int, int, int, int, void const* const*) Unexecuted instantiation: pixelfunctions.cpp:bool OptimizedSumThroughLargerType<double, unsigned char>(double, void*, int, int, int, int, int, void const* const*) Unexecuted instantiation: pixelfunctions.cpp:bool OptimizedSumThroughLargerType<double, unsigned short>(double, void*, int, int, int, int, int, void const* const*) Unexecuted instantiation: pixelfunctions.cpp:bool OptimizedSumThroughLargerType<double, short>(double, void*, int, int, int, int, int, void const* const*) Unexecuted instantiation: pixelfunctions.cpp:bool OptimizedSumThroughLargerType<double, int>(double, void*, int, int, int, int, int, void const* const*) |
916 | | |
917 | | /************************************************************************/ |
918 | | /* OptimizedSumThroughLargerType() */ |
919 | | /************************************************************************/ |
920 | | |
921 | | template <typename Tsrc> |
922 | | static bool OptimizedSumThroughLargerType(GDALDataType eBufType, double dfK, |
923 | | void *pOutBuffer, int nPixelSpace, |
924 | | int nLineSpace, int nXSize, |
925 | | int nYSize, int nSources, |
926 | | const void *const *papoSources) |
927 | 0 | { |
928 | 0 | switch (eBufType) |
929 | 0 | { |
930 | 0 | case GDT_Byte: |
931 | 0 | return OptimizedSumThroughLargerType<Tsrc, uint8_t>( |
932 | 0 | dfK, pOutBuffer, nPixelSpace, nLineSpace, nXSize, nYSize, |
933 | 0 | nSources, papoSources); |
934 | | |
935 | 0 | case GDT_UInt16: |
936 | 0 | return OptimizedSumThroughLargerType<Tsrc, uint16_t>( |
937 | 0 | dfK, pOutBuffer, nPixelSpace, nLineSpace, nXSize, nYSize, |
938 | 0 | nSources, papoSources); |
939 | | |
940 | 0 | case GDT_Int16: |
941 | 0 | return OptimizedSumThroughLargerType<Tsrc, int16_t>( |
942 | 0 | dfK, pOutBuffer, nPixelSpace, nLineSpace, nXSize, nYSize, |
943 | 0 | nSources, papoSources); |
944 | | |
945 | 0 | case GDT_Int32: |
946 | 0 | return OptimizedSumThroughLargerType<Tsrc, int32_t>( |
947 | 0 | dfK, pOutBuffer, nPixelSpace, nLineSpace, nXSize, nYSize, |
948 | 0 | nSources, papoSources); |
949 | | |
950 | | // Float32 and Float64 already covered by OptimizedSum() for packed case |
951 | 0 | default: |
952 | 0 | break; |
953 | 0 | } |
954 | 0 | return false; |
955 | 0 | } Unexecuted instantiation: pixelfunctions.cpp:bool OptimizedSumThroughLargerType<unsigned char>(GDALDataType, double, void*, int, int, int, int, int, void const* const*) Unexecuted instantiation: pixelfunctions.cpp:bool OptimizedSumThroughLargerType<unsigned short>(GDALDataType, double, void*, int, int, int, int, int, void const* const*) Unexecuted instantiation: pixelfunctions.cpp:bool OptimizedSumThroughLargerType<short>(GDALDataType, double, void*, int, int, int, int, int, void const* const*) Unexecuted instantiation: pixelfunctions.cpp:bool OptimizedSumThroughLargerType<int>(GDALDataType, double, void*, int, int, int, int, int, void const* const*) Unexecuted instantiation: pixelfunctions.cpp:bool OptimizedSumThroughLargerType<float>(GDALDataType, double, void*, int, int, int, int, int, void const* const*) Unexecuted instantiation: pixelfunctions.cpp:bool OptimizedSumThroughLargerType<double>(GDALDataType, double, void*, int, int, int, int, int, void const* const*) |
956 | | |
957 | | /************************************************************************/ |
958 | | /* OptimizedSumThroughLargerType() */ |
959 | | /************************************************************************/ |
960 | | |
961 | | static bool OptimizedSumThroughLargerType(GDALDataType eSrcType, |
962 | | GDALDataType eBufType, double dfK, |
963 | | void *pOutBuffer, int nPixelSpace, |
964 | | int nLineSpace, int nXSize, |
965 | | int nYSize, int nSources, |
966 | | const void *const *papoSources) |
967 | 0 | { |
968 | 0 | switch (eSrcType) |
969 | 0 | { |
970 | 0 | case GDT_Byte: |
971 | 0 | return OptimizedSumThroughLargerType<uint8_t>( |
972 | 0 | eBufType, dfK, pOutBuffer, nPixelSpace, nLineSpace, nXSize, |
973 | 0 | nYSize, nSources, papoSources); |
974 | | |
975 | 0 | case GDT_UInt16: |
976 | 0 | return OptimizedSumThroughLargerType<uint16_t>( |
977 | 0 | eBufType, dfK, pOutBuffer, nPixelSpace, nLineSpace, nXSize, |
978 | 0 | nYSize, nSources, papoSources); |
979 | | |
980 | 0 | case GDT_Int16: |
981 | 0 | return OptimizedSumThroughLargerType<int16_t>( |
982 | 0 | eBufType, dfK, pOutBuffer, nPixelSpace, nLineSpace, nXSize, |
983 | 0 | nYSize, nSources, papoSources); |
984 | | |
985 | 0 | case GDT_Int32: |
986 | 0 | return OptimizedSumThroughLargerType<int32_t>( |
987 | 0 | eBufType, dfK, pOutBuffer, nPixelSpace, nLineSpace, nXSize, |
988 | 0 | nYSize, nSources, papoSources); |
989 | | |
990 | 0 | case GDT_Float32: |
991 | 0 | return OptimizedSumThroughLargerType<float>( |
992 | 0 | eBufType, dfK, pOutBuffer, nPixelSpace, nLineSpace, nXSize, |
993 | 0 | nYSize, nSources, papoSources); |
994 | | |
995 | 0 | case GDT_Float64: |
996 | 0 | return OptimizedSumThroughLargerType<double>( |
997 | 0 | eBufType, dfK, pOutBuffer, nPixelSpace, nLineSpace, nXSize, |
998 | 0 | nYSize, nSources, papoSources); |
999 | | |
1000 | 0 | default: |
1001 | 0 | break; |
1002 | 0 | } |
1003 | | |
1004 | 0 | return false; |
1005 | 0 | } |
1006 | | |
1007 | | /************************************************************************/ |
1008 | | /* SumPixelFunc() */ |
1009 | | /************************************************************************/ |
1010 | | |
1011 | | static const char pszSumPixelFuncMetadata[] = |
1012 | | "<PixelFunctionArgumentsList>" |
1013 | | " <Argument name='k' description='Optional constant term' type='double' " |
1014 | | "default='0.0' />" |
1015 | | " <Argument name='propagateNoData' description='Whether the output value " |
1016 | | "should be NoData as as soon as one source is NoData' type='boolean' " |
1017 | | "default='false' />" |
1018 | | " <Argument type='builtin' value='NoData' optional='true' />" |
1019 | | "</PixelFunctionArgumentsList>"; |
1020 | | |
1021 | | static CPLErr SumPixelFunc(void **papoSources, int nSources, void *pData, |
1022 | | int nXSize, int nYSize, GDALDataType eSrcType, |
1023 | | GDALDataType eBufType, int nPixelSpace, |
1024 | | int nLineSpace, CSLConstList papszArgs) |
1025 | 0 | { |
1026 | | /* ---- Init ---- */ |
1027 | 0 | if (nSources < 1) |
1028 | 0 | { |
1029 | 0 | CPLError(CE_Failure, CPLE_AppDefined, |
1030 | 0 | "sum requires at least one source"); |
1031 | 0 | return CE_Failure; |
1032 | 0 | } |
1033 | | |
1034 | 0 | double dfK = 0.0; |
1035 | 0 | if (FetchDoubleArg(papszArgs, "k", &dfK, &dfK) != CE_None) |
1036 | 0 | return CE_Failure; |
1037 | | |
1038 | 0 | double dfNoData{0}; |
1039 | 0 | bool bHasNoData = CSLFindName(papszArgs, "NoData") != -1; |
1040 | 0 | if (bHasNoData && FetchDoubleArg(papszArgs, "NoData", &dfNoData) != CE_None) |
1041 | 0 | return CE_Failure; |
1042 | | |
1043 | 0 | const bool bPropagateNoData = CPLTestBool( |
1044 | 0 | CSLFetchNameValueDef(papszArgs, "propagateNoData", "false")); |
1045 | |
|
1046 | 0 | if (dfNoData == 0 && !bPropagateNoData) |
1047 | 0 | bHasNoData = false; |
1048 | | |
1049 | | /* ---- Set pixels ---- */ |
1050 | 0 | if (GDALDataTypeIsComplex(eSrcType)) |
1051 | 0 | { |
1052 | 0 | const int nOffset = GDALGetDataTypeSizeBytes(eSrcType) / 2; |
1053 | | |
1054 | | /* ---- Set pixels ---- */ |
1055 | 0 | size_t ii = 0; |
1056 | 0 | for (int iLine = 0; iLine < nYSize; ++iLine) |
1057 | 0 | { |
1058 | 0 | for (int iCol = 0; iCol < nXSize; ++iCol, ++ii) |
1059 | 0 | { |
1060 | 0 | double adfSum[2] = {dfK, 0.0}; |
1061 | |
|
1062 | 0 | for (int iSrc = 0; iSrc < nSources; ++iSrc) |
1063 | 0 | { |
1064 | 0 | const void *const pReal = papoSources[iSrc]; |
1065 | 0 | const void *const pImag = |
1066 | 0 | static_cast<const GByte *>(pReal) + nOffset; |
1067 | | |
1068 | | // Source raster pixels may be obtained with GetSrcVal |
1069 | | // macro. |
1070 | 0 | adfSum[0] += GetSrcVal(pReal, eSrcType, ii); |
1071 | 0 | adfSum[1] += GetSrcVal(pImag, eSrcType, ii); |
1072 | 0 | } |
1073 | |
|
1074 | 0 | GDALCopyWords(adfSum, GDT_CFloat64, 0, |
1075 | 0 | static_cast<GByte *>(pData) + |
1076 | 0 | static_cast<GSpacing>(nLineSpace) * iLine + |
1077 | 0 | iCol * nPixelSpace, |
1078 | 0 | eBufType, nPixelSpace, 1); |
1079 | 0 | } |
1080 | 0 | } |
1081 | 0 | } |
1082 | 0 | else |
1083 | 0 | { |
1084 | | /* ---- Set pixels ---- */ |
1085 | 0 | bool bGeneralCase = true; |
1086 | 0 | if (dfNoData == 0 && !bPropagateNoData) |
1087 | 0 | { |
1088 | 0 | #ifdef USE_SSE2 |
1089 | 0 | if (eBufType == GDT_Byte && nPixelSpace == sizeof(uint8_t) && |
1090 | 0 | eSrcType == GDT_Byte && |
1091 | 0 | dfK >= std::numeric_limits<uint8_t>::min() && |
1092 | 0 | dfK <= std::numeric_limits<uint8_t>::max() && |
1093 | 0 | static_cast<int>(dfK) == dfK) |
1094 | 0 | { |
1095 | 0 | bGeneralCase = false; |
1096 | |
|
1097 | 0 | struct SSEWrapper |
1098 | 0 | { |
1099 | 0 | inline static __m128i Set1(int8_t x) |
1100 | 0 | { |
1101 | 0 | return _mm_set1_epi8(x); |
1102 | 0 | } |
1103 | |
|
1104 | 0 | inline static __m128i AddSaturate(__m128i x, __m128i y) |
1105 | 0 | { |
1106 | 0 | return _mm_adds_epu8(x, y); |
1107 | 0 | } |
1108 | 0 | }; |
1109 | |
|
1110 | 0 | OptimizedSumSameType_SSE2<uint8_t, int8_t, uint32_t, |
1111 | 0 | SSEWrapper>(dfK, pData, nLineSpace, |
1112 | 0 | nXSize, nYSize, nSources, |
1113 | 0 | papoSources); |
1114 | 0 | } |
1115 | 0 | else if (eBufType == GDT_UInt16 && |
1116 | 0 | nPixelSpace == sizeof(uint16_t) && |
1117 | 0 | eSrcType == GDT_UInt16 && |
1118 | 0 | dfK >= std::numeric_limits<uint16_t>::min() && |
1119 | 0 | dfK <= std::numeric_limits<uint16_t>::max() && |
1120 | 0 | static_cast<int>(dfK) == dfK) |
1121 | 0 | { |
1122 | 0 | bGeneralCase = false; |
1123 | |
|
1124 | 0 | struct SSEWrapper |
1125 | 0 | { |
1126 | 0 | inline static __m128i Set1(int16_t x) |
1127 | 0 | { |
1128 | 0 | return _mm_set1_epi16(x); |
1129 | 0 | } |
1130 | |
|
1131 | 0 | inline static __m128i AddSaturate(__m128i x, __m128i y) |
1132 | 0 | { |
1133 | 0 | return _mm_adds_epu16(x, y); |
1134 | 0 | } |
1135 | 0 | }; |
1136 | |
|
1137 | 0 | OptimizedSumSameType_SSE2<uint16_t, int16_t, uint32_t, |
1138 | 0 | SSEWrapper>(dfK, pData, nLineSpace, |
1139 | 0 | nXSize, nYSize, nSources, |
1140 | 0 | papoSources); |
1141 | 0 | } |
1142 | 0 | else |
1143 | 0 | #endif |
1144 | 0 | if (eBufType == GDT_Float32 && nPixelSpace == sizeof(float)) |
1145 | 0 | { |
1146 | 0 | bGeneralCase = !OptimizedSumPackedOutput<float>( |
1147 | 0 | eSrcType, dfK, pData, nLineSpace, nXSize, nYSize, nSources, |
1148 | 0 | papoSources); |
1149 | 0 | } |
1150 | 0 | else if (eBufType == GDT_Float64 && nPixelSpace == sizeof(double)) |
1151 | 0 | { |
1152 | 0 | bGeneralCase = !OptimizedSumPackedOutput<double>( |
1153 | 0 | eSrcType, dfK, pData, nLineSpace, nXSize, nYSize, nSources, |
1154 | 0 | papoSources); |
1155 | 0 | } |
1156 | 0 | else if ( |
1157 | 0 | dfK >= 0 && dfK <= INT_MAX && eBufType == GDT_Int32 && |
1158 | 0 | nPixelSpace == sizeof(int32_t) && eSrcType == GDT_Byte && |
1159 | | // Limitation to avoid overflow of int32 if all source values are at the max of their data type |
1160 | 0 | nSources <= |
1161 | 0 | (INT_MAX - dfK) / std::numeric_limits<uint8_t>::max()) |
1162 | 0 | { |
1163 | 0 | bGeneralCase = false; |
1164 | 0 | OptimizedSumPackedOutput<uint8_t, int32_t>( |
1165 | 0 | dfK, pData, nLineSpace, nXSize, nYSize, nSources, |
1166 | 0 | papoSources); |
1167 | 0 | } |
1168 | |
|
1169 | 0 | if (bGeneralCase && dfK >= 0 && dfK <= INT_MAX && |
1170 | 0 | nSources <= |
1171 | 0 | (INT_MAX - dfK) / std::numeric_limits<uint16_t>::max()) |
1172 | 0 | { |
1173 | 0 | bGeneralCase = !OptimizedSumThroughLargerType( |
1174 | 0 | eSrcType, eBufType, dfK, pData, nPixelSpace, nLineSpace, |
1175 | 0 | nXSize, nYSize, nSources, papoSources); |
1176 | 0 | } |
1177 | 0 | } |
1178 | |
|
1179 | 0 | if (bGeneralCase) |
1180 | 0 | { |
1181 | 0 | size_t ii = 0; |
1182 | 0 | for (int iLine = 0; iLine < nYSize; ++iLine) |
1183 | 0 | { |
1184 | 0 | for (int iCol = 0; iCol < nXSize; ++iCol, ++ii) |
1185 | 0 | { |
1186 | 0 | double dfSum = dfK; |
1187 | 0 | for (int iSrc = 0; iSrc < nSources; ++iSrc) |
1188 | 0 | { |
1189 | 0 | const double dfVal = |
1190 | 0 | GetSrcVal(papoSources[iSrc], eSrcType, ii); |
1191 | |
|
1192 | 0 | if (bHasNoData && IsNoData(dfVal, dfNoData)) |
1193 | 0 | { |
1194 | 0 | if (bPropagateNoData) |
1195 | 0 | { |
1196 | 0 | dfSum = dfNoData; |
1197 | 0 | break; |
1198 | 0 | } |
1199 | 0 | } |
1200 | 0 | else |
1201 | 0 | { |
1202 | 0 | dfSum += dfVal; |
1203 | 0 | } |
1204 | 0 | } |
1205 | |
|
1206 | 0 | GDALCopyWords(&dfSum, GDT_Float64, 0, |
1207 | 0 | static_cast<GByte *>(pData) + |
1208 | 0 | static_cast<GSpacing>(nLineSpace) * |
1209 | 0 | iLine + |
1210 | 0 | iCol * nPixelSpace, |
1211 | 0 | eBufType, nPixelSpace, 1); |
1212 | 0 | } |
1213 | 0 | } |
1214 | 0 | } |
1215 | 0 | } |
1216 | | |
1217 | | /* ---- Return success ---- */ |
1218 | 0 | return CE_None; |
1219 | 0 | } /* SumPixelFunc */ |
1220 | | |
1221 | | static const char pszDiffPixelFuncMetadata[] = |
1222 | | "<PixelFunctionArgumentsList>" |
1223 | | " <Argument type='builtin' value='NoData' optional='true' />" |
1224 | | "</PixelFunctionArgumentsList>"; |
1225 | | |
1226 | | static CPLErr DiffPixelFunc(void **papoSources, int nSources, void *pData, |
1227 | | int nXSize, int nYSize, GDALDataType eSrcType, |
1228 | | GDALDataType eBufType, int nPixelSpace, |
1229 | | int nLineSpace, CSLConstList papszArgs) |
1230 | 0 | { |
1231 | | /* ---- Init ---- */ |
1232 | 0 | if (nSources != 2) |
1233 | 0 | return CE_Failure; |
1234 | | |
1235 | 0 | double dfNoData{0}; |
1236 | 0 | const bool bHasNoData = CSLFindName(papszArgs, "NoData") != -1; |
1237 | 0 | if (bHasNoData && FetchDoubleArg(papszArgs, "NoData", &dfNoData) != CE_None) |
1238 | 0 | return CE_Failure; |
1239 | | |
1240 | 0 | if (GDALDataTypeIsComplex(eSrcType)) |
1241 | 0 | { |
1242 | 0 | const int nOffset = GDALGetDataTypeSizeBytes(eSrcType) / 2; |
1243 | 0 | const void *const pReal0 = papoSources[0]; |
1244 | 0 | const void *const pImag0 = |
1245 | 0 | static_cast<GByte *>(papoSources[0]) + nOffset; |
1246 | 0 | const void *const pReal1 = papoSources[1]; |
1247 | 0 | const void *const pImag1 = |
1248 | 0 | static_cast<GByte *>(papoSources[1]) + nOffset; |
1249 | | |
1250 | | /* ---- Set pixels ---- */ |
1251 | 0 | size_t ii = 0; |
1252 | 0 | for (int iLine = 0; iLine < nYSize; ++iLine) |
1253 | 0 | { |
1254 | 0 | for (int iCol = 0; iCol < nXSize; ++iCol, ++ii) |
1255 | 0 | { |
1256 | | // Source raster pixels may be obtained with GetSrcVal macro. |
1257 | 0 | double adfPixVal[2] = {GetSrcVal(pReal0, eSrcType, ii) - |
1258 | 0 | GetSrcVal(pReal1, eSrcType, ii), |
1259 | 0 | GetSrcVal(pImag0, eSrcType, ii) - |
1260 | 0 | GetSrcVal(pImag1, eSrcType, ii)}; |
1261 | |
|
1262 | 0 | GDALCopyWords(adfPixVal, GDT_CFloat64, 0, |
1263 | 0 | static_cast<GByte *>(pData) + |
1264 | 0 | static_cast<GSpacing>(nLineSpace) * iLine + |
1265 | 0 | iCol * nPixelSpace, |
1266 | 0 | eBufType, nPixelSpace, 1); |
1267 | 0 | } |
1268 | 0 | } |
1269 | 0 | } |
1270 | 0 | else |
1271 | 0 | { |
1272 | | /* ---- Set pixels ---- */ |
1273 | 0 | size_t ii = 0; |
1274 | 0 | for (int iLine = 0; iLine < nYSize; ++iLine) |
1275 | 0 | { |
1276 | 0 | for (int iCol = 0; iCol < nXSize; ++iCol, ++ii) |
1277 | 0 | { |
1278 | 0 | const double dfA = GetSrcVal(papoSources[0], eSrcType, ii); |
1279 | 0 | const double dfB = GetSrcVal(papoSources[1], eSrcType, ii); |
1280 | |
|
1281 | 0 | const double dfPixVal = |
1282 | 0 | bHasNoData && |
1283 | 0 | (IsNoData(dfA, dfNoData) || IsNoData(dfB, dfNoData)) |
1284 | 0 | ? dfNoData |
1285 | 0 | : dfA - dfB; |
1286 | |
|
1287 | 0 | GDALCopyWords(&dfPixVal, GDT_Float64, 0, |
1288 | 0 | static_cast<GByte *>(pData) + |
1289 | 0 | static_cast<GSpacing>(nLineSpace) * iLine + |
1290 | 0 | iCol * nPixelSpace, |
1291 | 0 | eBufType, nPixelSpace, 1); |
1292 | 0 | } |
1293 | 0 | } |
1294 | 0 | } |
1295 | | |
1296 | | /* ---- Return success ---- */ |
1297 | 0 | return CE_None; |
1298 | 0 | } // DiffPixelFunc |
1299 | | |
1300 | | static const char pszMulPixelFuncMetadata[] = |
1301 | | "<PixelFunctionArgumentsList>" |
1302 | | " <Argument name='k' description='Optional constant factor' " |
1303 | | "type='double' default='1.0' />" |
1304 | | " <Argument name='propagateNoData' description='Whether the output value " |
1305 | | "should be NoData as as soon as one source is NoData' type='boolean' " |
1306 | | "default='false' />" |
1307 | | " <Argument type='builtin' value='NoData' optional='true' />" |
1308 | | "</PixelFunctionArgumentsList>"; |
1309 | | |
1310 | | static CPLErr MulPixelFunc(void **papoSources, int nSources, void *pData, |
1311 | | int nXSize, int nYSize, GDALDataType eSrcType, |
1312 | | GDALDataType eBufType, int nPixelSpace, |
1313 | | int nLineSpace, CSLConstList papszArgs) |
1314 | 0 | { |
1315 | | /* ---- Init ---- */ |
1316 | 0 | if (nSources < 2 && CSLFetchNameValue(papszArgs, "k") == nullptr) |
1317 | 0 | { |
1318 | 0 | CPLError(CE_Failure, CPLE_AppDefined, |
1319 | 0 | "mul requires at least two sources or a specified constant k"); |
1320 | 0 | return CE_Failure; |
1321 | 0 | } |
1322 | | |
1323 | 0 | double dfK = 1.0; |
1324 | 0 | if (FetchDoubleArg(papszArgs, "k", &dfK, &dfK) != CE_None) |
1325 | 0 | return CE_Failure; |
1326 | | |
1327 | 0 | double dfNoData{0}; |
1328 | 0 | const bool bHasNoData = CSLFindName(papszArgs, "NoData") != -1; |
1329 | 0 | if (bHasNoData && FetchDoubleArg(papszArgs, "NoData", &dfNoData) != CE_None) |
1330 | 0 | return CE_Failure; |
1331 | | |
1332 | 0 | const bool bPropagateNoData = CPLTestBool( |
1333 | 0 | CSLFetchNameValueDef(papszArgs, "propagateNoData", "false")); |
1334 | | |
1335 | | /* ---- Set pixels ---- */ |
1336 | 0 | if (GDALDataTypeIsComplex(eSrcType)) |
1337 | 0 | { |
1338 | 0 | const int nOffset = GDALGetDataTypeSizeBytes(eSrcType) / 2; |
1339 | | |
1340 | | /* ---- Set pixels ---- */ |
1341 | 0 | size_t ii = 0; |
1342 | 0 | for (int iLine = 0; iLine < nYSize; ++iLine) |
1343 | 0 | { |
1344 | 0 | for (int iCol = 0; iCol < nXSize; ++iCol, ++ii) |
1345 | 0 | { |
1346 | 0 | double adfPixVal[2] = {dfK, 0.0}; |
1347 | |
|
1348 | 0 | for (int iSrc = 0; iSrc < nSources; ++iSrc) |
1349 | 0 | { |
1350 | 0 | const void *const pReal = papoSources[iSrc]; |
1351 | 0 | const void *const pImag = |
1352 | 0 | static_cast<const GByte *>(pReal) + nOffset; |
1353 | |
|
1354 | 0 | const double dfOldR = adfPixVal[0]; |
1355 | 0 | const double dfOldI = adfPixVal[1]; |
1356 | | |
1357 | | // Source raster pixels may be obtained with GetSrcVal |
1358 | | // macro. |
1359 | 0 | const double dfNewR = GetSrcVal(pReal, eSrcType, ii); |
1360 | 0 | const double dfNewI = GetSrcVal(pImag, eSrcType, ii); |
1361 | |
|
1362 | 0 | adfPixVal[0] = dfOldR * dfNewR - dfOldI * dfNewI; |
1363 | 0 | adfPixVal[1] = dfOldR * dfNewI + dfOldI * dfNewR; |
1364 | 0 | } |
1365 | |
|
1366 | 0 | GDALCopyWords(adfPixVal, GDT_CFloat64, 0, |
1367 | 0 | static_cast<GByte *>(pData) + |
1368 | 0 | static_cast<GSpacing>(nLineSpace) * iLine + |
1369 | 0 | iCol * nPixelSpace, |
1370 | 0 | eBufType, nPixelSpace, 1); |
1371 | 0 | } |
1372 | 0 | } |
1373 | 0 | } |
1374 | 0 | else |
1375 | 0 | { |
1376 | | /* ---- Set pixels ---- */ |
1377 | 0 | size_t ii = 0; |
1378 | 0 | for (int iLine = 0; iLine < nYSize; ++iLine) |
1379 | 0 | { |
1380 | 0 | for (int iCol = 0; iCol < nXSize; ++iCol, ++ii) |
1381 | 0 | { |
1382 | 0 | double dfPixVal = dfK; // Not complex. |
1383 | |
|
1384 | 0 | for (int iSrc = 0; iSrc < nSources; ++iSrc) |
1385 | 0 | { |
1386 | 0 | const double dfVal = |
1387 | 0 | GetSrcVal(papoSources[iSrc], eSrcType, ii); |
1388 | |
|
1389 | 0 | if (bHasNoData && IsNoData(dfVal, dfNoData)) |
1390 | 0 | { |
1391 | 0 | if (bPropagateNoData) |
1392 | 0 | { |
1393 | 0 | dfPixVal = dfNoData; |
1394 | 0 | break; |
1395 | 0 | } |
1396 | 0 | } |
1397 | 0 | else |
1398 | 0 | { |
1399 | 0 | dfPixVal *= dfVal; |
1400 | 0 | } |
1401 | 0 | } |
1402 | |
|
1403 | 0 | GDALCopyWords(&dfPixVal, GDT_Float64, 0, |
1404 | 0 | static_cast<GByte *>(pData) + |
1405 | 0 | static_cast<GSpacing>(nLineSpace) * iLine + |
1406 | 0 | iCol * nPixelSpace, |
1407 | 0 | eBufType, nPixelSpace, 1); |
1408 | 0 | } |
1409 | 0 | } |
1410 | 0 | } |
1411 | | |
1412 | | /* ---- Return success ---- */ |
1413 | 0 | return CE_None; |
1414 | 0 | } // MulPixelFunc |
1415 | | |
1416 | | static const char pszDivPixelFuncMetadata[] = |
1417 | | "<PixelFunctionArgumentsList>" |
1418 | | " " |
1419 | | "<Argument type='builtin' value='NoData' optional='true' />" |
1420 | | "</PixelFunctionArgumentsList>"; |
1421 | | |
1422 | | static CPLErr DivPixelFunc(void **papoSources, int nSources, void *pData, |
1423 | | int nXSize, int nYSize, GDALDataType eSrcType, |
1424 | | GDALDataType eBufType, int nPixelSpace, |
1425 | | int nLineSpace, CSLConstList papszArgs) |
1426 | 0 | { |
1427 | | /* ---- Init ---- */ |
1428 | 0 | if (nSources != 2) |
1429 | 0 | return CE_Failure; |
1430 | | |
1431 | 0 | double dfNoData{0}; |
1432 | 0 | const bool bHasNoData = CSLFindName(papszArgs, "NoData") != -1; |
1433 | 0 | if (bHasNoData && FetchDoubleArg(papszArgs, "NoData", &dfNoData) != CE_None) |
1434 | 0 | return CE_Failure; |
1435 | | |
1436 | | /* ---- Set pixels ---- */ |
1437 | 0 | if (GDALDataTypeIsComplex(eSrcType)) |
1438 | 0 | { |
1439 | 0 | const int nOffset = GDALGetDataTypeSizeBytes(eSrcType) / 2; |
1440 | 0 | const void *const pReal0 = papoSources[0]; |
1441 | 0 | const void *const pImag0 = |
1442 | 0 | static_cast<GByte *>(papoSources[0]) + nOffset; |
1443 | 0 | const void *const pReal1 = papoSources[1]; |
1444 | 0 | const void *const pImag1 = |
1445 | 0 | static_cast<GByte *>(papoSources[1]) + nOffset; |
1446 | | |
1447 | | /* ---- Set pixels ---- */ |
1448 | 0 | size_t ii = 0; |
1449 | 0 | for (int iLine = 0; iLine < nYSize; ++iLine) |
1450 | 0 | { |
1451 | 0 | for (int iCol = 0; iCol < nXSize; ++iCol, ++ii) |
1452 | 0 | { |
1453 | | // Source raster pixels may be obtained with GetSrcVal macro. |
1454 | 0 | const double dfReal0 = GetSrcVal(pReal0, eSrcType, ii); |
1455 | 0 | const double dfReal1 = GetSrcVal(pReal1, eSrcType, ii); |
1456 | 0 | const double dfImag0 = GetSrcVal(pImag0, eSrcType, ii); |
1457 | 0 | const double dfImag1 = GetSrcVal(pImag1, eSrcType, ii); |
1458 | 0 | const double dfAux = dfReal1 * dfReal1 + dfImag1 * dfImag1; |
1459 | |
|
1460 | 0 | const double adfPixVal[2] = { |
1461 | 0 | dfAux == 0 |
1462 | 0 | ? std::numeric_limits<double>::infinity() |
1463 | 0 | : dfReal0 * dfReal1 / dfAux + dfImag0 * dfImag1 / dfAux, |
1464 | 0 | dfAux == 0 ? std::numeric_limits<double>::infinity() |
1465 | 0 | : dfReal1 / dfAux * dfImag0 - |
1466 | 0 | dfReal0 * dfImag1 / dfAux}; |
1467 | |
|
1468 | 0 | GDALCopyWords(adfPixVal, GDT_CFloat64, 0, |
1469 | 0 | static_cast<GByte *>(pData) + |
1470 | 0 | static_cast<GSpacing>(nLineSpace) * iLine + |
1471 | 0 | iCol * nPixelSpace, |
1472 | 0 | eBufType, nPixelSpace, 1); |
1473 | 0 | } |
1474 | 0 | } |
1475 | 0 | } |
1476 | 0 | else |
1477 | 0 | { |
1478 | | /* ---- Set pixels ---- */ |
1479 | 0 | size_t ii = 0; |
1480 | 0 | for (int iLine = 0; iLine < nYSize; ++iLine) |
1481 | 0 | { |
1482 | 0 | for (int iCol = 0; iCol < nXSize; ++iCol, ++ii) |
1483 | 0 | { |
1484 | 0 | const double dfNum = GetSrcVal(papoSources[0], eSrcType, ii); |
1485 | 0 | const double dfDenom = GetSrcVal(papoSources[1], eSrcType, ii); |
1486 | |
|
1487 | 0 | double dfPixVal = dfNoData; |
1488 | 0 | if (!bHasNoData || (!IsNoData(dfNum, dfNoData) && |
1489 | 0 | !IsNoData(dfDenom, dfNoData))) |
1490 | 0 | { |
1491 | | // coverity[divide_by_zero] |
1492 | 0 | dfPixVal = |
1493 | 0 | dfDenom == 0 |
1494 | 0 | ? std::numeric_limits<double>::infinity() |
1495 | 0 | : dfNum / |
1496 | | #ifdef __COVERITY__ |
1497 | | (dfDenom + std::numeric_limits<double>::min()) |
1498 | | #else |
1499 | 0 | dfDenom |
1500 | 0 | #endif |
1501 | 0 | ; |
1502 | 0 | } |
1503 | |
|
1504 | 0 | GDALCopyWords(&dfPixVal, GDT_Float64, 0, |
1505 | 0 | static_cast<GByte *>(pData) + |
1506 | 0 | static_cast<GSpacing>(nLineSpace) * iLine + |
1507 | 0 | iCol * nPixelSpace, |
1508 | 0 | eBufType, nPixelSpace, 1); |
1509 | 0 | } |
1510 | 0 | } |
1511 | 0 | } |
1512 | | |
1513 | | /* ---- Return success ---- */ |
1514 | 0 | return CE_None; |
1515 | 0 | } // DivPixelFunc |
1516 | | |
1517 | | static CPLErr CMulPixelFunc(void **papoSources, int nSources, void *pData, |
1518 | | int nXSize, int nYSize, GDALDataType eSrcType, |
1519 | | GDALDataType eBufType, int nPixelSpace, |
1520 | | int nLineSpace) |
1521 | 0 | { |
1522 | | /* ---- Init ---- */ |
1523 | 0 | if (nSources != 2) |
1524 | 0 | return CE_Failure; |
1525 | | |
1526 | | /* ---- Set pixels ---- */ |
1527 | 0 | if (GDALDataTypeIsComplex(eSrcType)) |
1528 | 0 | { |
1529 | 0 | const int nOffset = GDALGetDataTypeSizeBytes(eSrcType) / 2; |
1530 | 0 | const void *const pReal0 = papoSources[0]; |
1531 | 0 | const void *const pImag0 = |
1532 | 0 | static_cast<GByte *>(papoSources[0]) + nOffset; |
1533 | 0 | const void *const pReal1 = papoSources[1]; |
1534 | 0 | const void *const pImag1 = |
1535 | 0 | static_cast<GByte *>(papoSources[1]) + nOffset; |
1536 | |
|
1537 | 0 | size_t ii = 0; |
1538 | 0 | for (int iLine = 0; iLine < nYSize; ++iLine) |
1539 | 0 | { |
1540 | 0 | for (int iCol = 0; iCol < nXSize; ++iCol, ++ii) |
1541 | 0 | { |
1542 | | // Source raster pixels may be obtained with GetSrcVal macro. |
1543 | 0 | const double dfReal0 = GetSrcVal(pReal0, eSrcType, ii); |
1544 | 0 | const double dfReal1 = GetSrcVal(pReal1, eSrcType, ii); |
1545 | 0 | const double dfImag0 = GetSrcVal(pImag0, eSrcType, ii); |
1546 | 0 | const double dfImag1 = GetSrcVal(pImag1, eSrcType, ii); |
1547 | 0 | const double adfPixVal[2] = { |
1548 | 0 | dfReal0 * dfReal1 + dfImag0 * dfImag1, |
1549 | 0 | dfReal1 * dfImag0 - dfReal0 * dfImag1}; |
1550 | |
|
1551 | 0 | GDALCopyWords(adfPixVal, GDT_CFloat64, 0, |
1552 | 0 | static_cast<GByte *>(pData) + |
1553 | 0 | static_cast<GSpacing>(nLineSpace) * iLine + |
1554 | 0 | iCol * nPixelSpace, |
1555 | 0 | eBufType, nPixelSpace, 1); |
1556 | 0 | } |
1557 | 0 | } |
1558 | 0 | } |
1559 | 0 | else |
1560 | 0 | { |
1561 | 0 | size_t ii = 0; |
1562 | 0 | for (int iLine = 0; iLine < nYSize; ++iLine) |
1563 | 0 | { |
1564 | 0 | for (int iCol = 0; iCol < nXSize; ++iCol, ++ii) |
1565 | 0 | { |
1566 | | // Source raster pixels may be obtained with GetSrcVal macro. |
1567 | | // Not complex. |
1568 | 0 | const double adfPixVal[2] = { |
1569 | 0 | GetSrcVal(papoSources[0], eSrcType, ii) * |
1570 | 0 | GetSrcVal(papoSources[1], eSrcType, ii), |
1571 | 0 | 0.0}; |
1572 | |
|
1573 | 0 | GDALCopyWords(adfPixVal, GDT_CFloat64, 0, |
1574 | 0 | static_cast<GByte *>(pData) + |
1575 | 0 | static_cast<GSpacing>(nLineSpace) * iLine + |
1576 | 0 | iCol * nPixelSpace, |
1577 | 0 | eBufType, nPixelSpace, 1); |
1578 | 0 | } |
1579 | 0 | } |
1580 | 0 | } |
1581 | | |
1582 | | /* ---- Return success ---- */ |
1583 | 0 | return CE_None; |
1584 | 0 | } // CMulPixelFunc |
1585 | | |
1586 | | static const char pszInvPixelFuncMetadata[] = |
1587 | | "<PixelFunctionArgumentsList>" |
1588 | | " <Argument name='k' description='Optional constant factor' " |
1589 | | "type='double' default='1.0' />" |
1590 | | " " |
1591 | | "<Argument type='builtin' value='NoData' optional='true' />" |
1592 | | "</PixelFunctionArgumentsList>"; |
1593 | | |
1594 | | static CPLErr InvPixelFunc(void **papoSources, int nSources, void *pData, |
1595 | | int nXSize, int nYSize, GDALDataType eSrcType, |
1596 | | GDALDataType eBufType, int nPixelSpace, |
1597 | | int nLineSpace, CSLConstList papszArgs) |
1598 | 0 | { |
1599 | | /* ---- Init ---- */ |
1600 | 0 | if (nSources != 1) |
1601 | 0 | return CE_Failure; |
1602 | | |
1603 | 0 | double dfK = 1.0; |
1604 | 0 | if (FetchDoubleArg(papszArgs, "k", &dfK, &dfK) != CE_None) |
1605 | 0 | return CE_Failure; |
1606 | | |
1607 | 0 | double dfNoData{0}; |
1608 | 0 | const bool bHasNoData = CSLFindName(papszArgs, "NoData") != -1; |
1609 | 0 | if (bHasNoData && FetchDoubleArg(papszArgs, "NoData", &dfNoData) != CE_None) |
1610 | 0 | return CE_Failure; |
1611 | | |
1612 | | /* ---- Set pixels ---- */ |
1613 | 0 | if (GDALDataTypeIsComplex(eSrcType)) |
1614 | 0 | { |
1615 | 0 | const int nOffset = GDALGetDataTypeSizeBytes(eSrcType) / 2; |
1616 | 0 | const void *const pReal = papoSources[0]; |
1617 | 0 | const void *const pImag = |
1618 | 0 | static_cast<GByte *>(papoSources[0]) + nOffset; |
1619 | |
|
1620 | 0 | size_t ii = 0; |
1621 | 0 | for (int iLine = 0; iLine < nYSize; ++iLine) |
1622 | 0 | { |
1623 | 0 | for (int iCol = 0; iCol < nXSize; ++iCol, ++ii) |
1624 | 0 | { |
1625 | | // Source raster pixels may be obtained with GetSrcVal macro. |
1626 | 0 | const double dfReal = GetSrcVal(pReal, eSrcType, ii); |
1627 | 0 | const double dfImag = GetSrcVal(pImag, eSrcType, ii); |
1628 | 0 | const double dfAux = dfReal * dfReal + dfImag * dfImag; |
1629 | 0 | const double adfPixVal[2] = { |
1630 | 0 | dfAux == 0 ? std::numeric_limits<double>::infinity() |
1631 | 0 | : dfK * dfReal / dfAux, |
1632 | 0 | dfAux == 0 ? std::numeric_limits<double>::infinity() |
1633 | 0 | : -dfK * dfImag / dfAux}; |
1634 | |
|
1635 | 0 | GDALCopyWords(adfPixVal, GDT_CFloat64, 0, |
1636 | 0 | static_cast<GByte *>(pData) + |
1637 | 0 | static_cast<GSpacing>(nLineSpace) * iLine + |
1638 | 0 | iCol * nPixelSpace, |
1639 | 0 | eBufType, nPixelSpace, 1); |
1640 | 0 | } |
1641 | 0 | } |
1642 | 0 | } |
1643 | 0 | else |
1644 | 0 | { |
1645 | | /* ---- Set pixels ---- */ |
1646 | 0 | size_t ii = 0; |
1647 | 0 | for (int iLine = 0; iLine < nYSize; ++iLine) |
1648 | 0 | { |
1649 | 0 | for (int iCol = 0; iCol < nXSize; ++iCol, ++ii) |
1650 | 0 | { |
1651 | | // Source raster pixels may be obtained with GetSrcVal macro. |
1652 | | // Not complex. |
1653 | 0 | const double dfVal = GetSrcVal(papoSources[0], eSrcType, ii); |
1654 | 0 | double dfPixVal = dfNoData; |
1655 | |
|
1656 | 0 | if (!bHasNoData || !IsNoData(dfVal, dfNoData)) |
1657 | 0 | { |
1658 | 0 | dfPixVal = |
1659 | 0 | dfVal == 0 |
1660 | 0 | ? std::numeric_limits<double>::infinity() |
1661 | 0 | : dfK / |
1662 | | #ifdef __COVERITY__ |
1663 | | (dfVal + std::numeric_limits<double>::min()) |
1664 | | #else |
1665 | 0 | dfVal |
1666 | 0 | #endif |
1667 | 0 | ; |
1668 | 0 | } |
1669 | |
|
1670 | 0 | GDALCopyWords(&dfPixVal, GDT_Float64, 0, |
1671 | 0 | static_cast<GByte *>(pData) + |
1672 | 0 | static_cast<GSpacing>(nLineSpace) * iLine + |
1673 | 0 | iCol * nPixelSpace, |
1674 | 0 | eBufType, nPixelSpace, 1); |
1675 | 0 | } |
1676 | 0 | } |
1677 | 0 | } |
1678 | | |
1679 | | /* ---- Return success ---- */ |
1680 | 0 | return CE_None; |
1681 | 0 | } // InvPixelFunc |
1682 | | |
1683 | | static CPLErr IntensityPixelFunc(void **papoSources, int nSources, void *pData, |
1684 | | int nXSize, int nYSize, GDALDataType eSrcType, |
1685 | | GDALDataType eBufType, int nPixelSpace, |
1686 | | int nLineSpace) |
1687 | 0 | { |
1688 | | /* ---- Init ---- */ |
1689 | 0 | if (nSources != 1) |
1690 | 0 | return CE_Failure; |
1691 | | |
1692 | 0 | if (GDALDataTypeIsComplex(eSrcType)) |
1693 | 0 | { |
1694 | 0 | const int nOffset = GDALGetDataTypeSizeBytes(eSrcType) / 2; |
1695 | 0 | const void *const pReal = papoSources[0]; |
1696 | 0 | const void *const pImag = |
1697 | 0 | static_cast<GByte *>(papoSources[0]) + nOffset; |
1698 | | |
1699 | | /* ---- Set pixels ---- */ |
1700 | 0 | size_t ii = 0; |
1701 | 0 | for (int iLine = 0; iLine < nYSize; ++iLine) |
1702 | 0 | { |
1703 | 0 | for (int iCol = 0; iCol < nXSize; ++iCol, ++ii) |
1704 | 0 | { |
1705 | | // Source raster pixels may be obtained with GetSrcVal macro. |
1706 | 0 | const double dfReal = GetSrcVal(pReal, eSrcType, ii); |
1707 | 0 | const double dfImag = GetSrcVal(pImag, eSrcType, ii); |
1708 | |
|
1709 | 0 | const double dfPixVal = dfReal * dfReal + dfImag * dfImag; |
1710 | |
|
1711 | 0 | GDALCopyWords(&dfPixVal, GDT_Float64, 0, |
1712 | 0 | static_cast<GByte *>(pData) + |
1713 | 0 | static_cast<GSpacing>(nLineSpace) * iLine + |
1714 | 0 | iCol * nPixelSpace, |
1715 | 0 | eBufType, nPixelSpace, 1); |
1716 | 0 | } |
1717 | 0 | } |
1718 | 0 | } |
1719 | 0 | else |
1720 | 0 | { |
1721 | | /* ---- Set pixels ---- */ |
1722 | 0 | size_t ii = 0; |
1723 | 0 | for (int iLine = 0; iLine < nYSize; ++iLine) |
1724 | 0 | { |
1725 | 0 | for (int iCol = 0; iCol < nXSize; ++iCol, ++ii) |
1726 | 0 | { |
1727 | | // Source raster pixels may be obtained with GetSrcVal macro. |
1728 | 0 | double dfPixVal = GetSrcVal(papoSources[0], eSrcType, ii); |
1729 | 0 | dfPixVal *= dfPixVal; |
1730 | |
|
1731 | 0 | GDALCopyWords(&dfPixVal, GDT_Float64, 0, |
1732 | 0 | static_cast<GByte *>(pData) + |
1733 | 0 | static_cast<GSpacing>(nLineSpace) * iLine + |
1734 | 0 | iCol * nPixelSpace, |
1735 | 0 | eBufType, nPixelSpace, 1); |
1736 | 0 | } |
1737 | 0 | } |
1738 | 0 | } |
1739 | | |
1740 | | /* ---- Return success ---- */ |
1741 | 0 | return CE_None; |
1742 | 0 | } // IntensityPixelFunc |
1743 | | |
1744 | | static const char pszSqrtPixelFuncMetadata[] = |
1745 | | "<PixelFunctionArgumentsList>" |
1746 | | " <Argument type='builtin' value='NoData' optional='true'/>" |
1747 | | "</PixelFunctionArgumentsList>"; |
1748 | | |
1749 | | static CPLErr SqrtPixelFunc(void **papoSources, int nSources, void *pData, |
1750 | | int nXSize, int nYSize, GDALDataType eSrcType, |
1751 | | GDALDataType eBufType, int nPixelSpace, |
1752 | | int nLineSpace, CSLConstList papszArgs) |
1753 | 0 | { |
1754 | | /* ---- Init ---- */ |
1755 | 0 | if (nSources != 1) |
1756 | 0 | return CE_Failure; |
1757 | 0 | if (GDALDataTypeIsComplex(eSrcType)) |
1758 | 0 | return CE_Failure; |
1759 | | |
1760 | 0 | double dfNoData{0}; |
1761 | 0 | const bool bHasNoData = CSLFindName(papszArgs, "NoData") != -1; |
1762 | 0 | if (bHasNoData && FetchDoubleArg(papszArgs, "NoData", &dfNoData) != CE_None) |
1763 | 0 | return CE_Failure; |
1764 | | |
1765 | | /* ---- Set pixels ---- */ |
1766 | 0 | size_t ii = 0; |
1767 | 0 | for (int iLine = 0; iLine < nYSize; ++iLine) |
1768 | 0 | { |
1769 | 0 | for (int iCol = 0; iCol < nXSize; ++iCol, ++ii) |
1770 | 0 | { |
1771 | | // Source raster pixels may be obtained with GetSrcVal macro. |
1772 | 0 | double dfPixVal = GetSrcVal(papoSources[0], eSrcType, ii); |
1773 | |
|
1774 | 0 | if (bHasNoData && IsNoData(dfPixVal, dfNoData)) |
1775 | 0 | { |
1776 | 0 | dfPixVal = dfNoData; |
1777 | 0 | } |
1778 | 0 | else |
1779 | 0 | { |
1780 | 0 | dfPixVal = std::sqrt(dfPixVal); |
1781 | 0 | } |
1782 | |
|
1783 | 0 | GDALCopyWords(&dfPixVal, GDT_Float64, 0, |
1784 | 0 | static_cast<GByte *>(pData) + |
1785 | 0 | static_cast<GSpacing>(nLineSpace) * iLine + |
1786 | 0 | iCol * nPixelSpace, |
1787 | 0 | eBufType, nPixelSpace, 1); |
1788 | 0 | } |
1789 | 0 | } |
1790 | | |
1791 | | /* ---- Return success ---- */ |
1792 | 0 | return CE_None; |
1793 | 0 | } // SqrtPixelFunc |
1794 | | |
1795 | | static CPLErr Log10PixelFuncHelper(void **papoSources, int nSources, |
1796 | | void *pData, int nXSize, int nYSize, |
1797 | | GDALDataType eSrcType, GDALDataType eBufType, |
1798 | | int nPixelSpace, int nLineSpace, |
1799 | | CSLConstList papszArgs, double fact) |
1800 | 0 | { |
1801 | | /* ---- Init ---- */ |
1802 | 0 | if (nSources != 1) |
1803 | 0 | return CE_Failure; |
1804 | | |
1805 | 0 | double dfNoData{0}; |
1806 | 0 | const bool bHasNoData = CSLFindName(papszArgs, "NoData") != -1; |
1807 | 0 | if (bHasNoData && FetchDoubleArg(papszArgs, "NoData", &dfNoData) != CE_None) |
1808 | 0 | return CE_Failure; |
1809 | | |
1810 | 0 | if (GDALDataTypeIsComplex(eSrcType)) |
1811 | 0 | { |
1812 | | // Complex input datatype. |
1813 | 0 | const int nOffset = GDALGetDataTypeSizeBytes(eSrcType) / 2; |
1814 | 0 | const void *const pReal = papoSources[0]; |
1815 | 0 | const void *const pImag = |
1816 | 0 | static_cast<GByte *>(papoSources[0]) + nOffset; |
1817 | | |
1818 | | /* We should compute fact * log10( sqrt( dfReal * dfReal + dfImag * |
1819 | | * dfImag ) ) */ |
1820 | | /* Given that log10(sqrt(x)) = 0.5 * log10(x) */ |
1821 | | /* we can remove the sqrt() by multiplying fact by 0.5 */ |
1822 | 0 | fact *= 0.5; |
1823 | | |
1824 | | /* ---- Set pixels ---- */ |
1825 | 0 | size_t ii = 0; |
1826 | 0 | for (int iLine = 0; iLine < nYSize; ++iLine) |
1827 | 0 | { |
1828 | 0 | for (int iCol = 0; iCol < nXSize; ++iCol, ++ii) |
1829 | 0 | { |
1830 | | // Source raster pixels may be obtained with GetSrcVal macro. |
1831 | 0 | const double dfReal = GetSrcVal(pReal, eSrcType, ii); |
1832 | 0 | const double dfImag = GetSrcVal(pImag, eSrcType, ii); |
1833 | |
|
1834 | 0 | const double dfPixVal = |
1835 | 0 | fact * log10(dfReal * dfReal + dfImag * dfImag); |
1836 | |
|
1837 | 0 | GDALCopyWords(&dfPixVal, GDT_Float64, 0, |
1838 | 0 | static_cast<GByte *>(pData) + |
1839 | 0 | static_cast<GSpacing>(nLineSpace) * iLine + |
1840 | 0 | iCol * nPixelSpace, |
1841 | 0 | eBufType, nPixelSpace, 1); |
1842 | 0 | } |
1843 | 0 | } |
1844 | 0 | } |
1845 | 0 | else |
1846 | 0 | { |
1847 | | /* ---- Set pixels ---- */ |
1848 | 0 | size_t ii = 0; |
1849 | 0 | for (int iLine = 0; iLine < nYSize; ++iLine) |
1850 | 0 | { |
1851 | 0 | for (int iCol = 0; iCol < nXSize; ++iCol, ++ii) |
1852 | 0 | { |
1853 | | // Source raster pixels may be obtained with GetSrcVal macro. |
1854 | 0 | const double dfSrcVal = GetSrcVal(papoSources[0], eSrcType, ii); |
1855 | 0 | const double dfPixVal = |
1856 | 0 | bHasNoData && IsNoData(dfSrcVal, dfNoData) |
1857 | 0 | ? dfNoData |
1858 | 0 | : fact * std::log10(std::abs(dfSrcVal)); |
1859 | |
|
1860 | 0 | GDALCopyWords(&dfPixVal, GDT_Float64, 0, |
1861 | 0 | static_cast<GByte *>(pData) + |
1862 | 0 | static_cast<GSpacing>(nLineSpace) * iLine + |
1863 | 0 | iCol * nPixelSpace, |
1864 | 0 | eBufType, nPixelSpace, 1); |
1865 | 0 | } |
1866 | 0 | } |
1867 | 0 | } |
1868 | | |
1869 | | /* ---- Return success ---- */ |
1870 | 0 | return CE_None; |
1871 | 0 | } // Log10PixelFuncHelper |
1872 | | |
1873 | | static const char pszLog10PixelFuncMetadata[] = |
1874 | | "<PixelFunctionArgumentsList>" |
1875 | | " <Argument type='builtin' value='NoData' optional='true'/>" |
1876 | | "</PixelFunctionArgumentsList>"; |
1877 | | |
1878 | | static CPLErr Log10PixelFunc(void **papoSources, int nSources, void *pData, |
1879 | | int nXSize, int nYSize, GDALDataType eSrcType, |
1880 | | GDALDataType eBufType, int nPixelSpace, |
1881 | | int nLineSpace, CSLConstList papszArgs) |
1882 | 0 | { |
1883 | 0 | return Log10PixelFuncHelper(papoSources, nSources, pData, nXSize, nYSize, |
1884 | 0 | eSrcType, eBufType, nPixelSpace, nLineSpace, |
1885 | 0 | papszArgs, 1.0); |
1886 | 0 | } // Log10PixelFunc |
1887 | | |
1888 | | static const char pszDBPixelFuncMetadata[] = |
1889 | | "<PixelFunctionArgumentsList>" |
1890 | | " <Argument name='fact' description='Factor' type='double' " |
1891 | | "default='20.0' />" |
1892 | | " <Argument type='builtin' value='NoData' optional='true' />" |
1893 | | "</PixelFunctionArgumentsList>"; |
1894 | | |
1895 | | static CPLErr DBPixelFunc(void **papoSources, int nSources, void *pData, |
1896 | | int nXSize, int nYSize, GDALDataType eSrcType, |
1897 | | GDALDataType eBufType, int nPixelSpace, |
1898 | | int nLineSpace, CSLConstList papszArgs) |
1899 | 0 | { |
1900 | 0 | double dfFact = 20.; |
1901 | 0 | if (FetchDoubleArg(papszArgs, "fact", &dfFact, &dfFact) != CE_None) |
1902 | 0 | return CE_Failure; |
1903 | | |
1904 | 0 | return Log10PixelFuncHelper(papoSources, nSources, pData, nXSize, nYSize, |
1905 | 0 | eSrcType, eBufType, nPixelSpace, nLineSpace, |
1906 | 0 | papszArgs, dfFact); |
1907 | 0 | } // DBPixelFunc |
1908 | | |
1909 | | static CPLErr ExpPixelFuncHelper(void **papoSources, int nSources, void *pData, |
1910 | | int nXSize, int nYSize, GDALDataType eSrcType, |
1911 | | GDALDataType eBufType, int nPixelSpace, |
1912 | | int nLineSpace, CSLConstList papszArgs, |
1913 | | double base, double fact) |
1914 | 0 | { |
1915 | | /* ---- Init ---- */ |
1916 | 0 | if (nSources != 1) |
1917 | 0 | return CE_Failure; |
1918 | 0 | if (GDALDataTypeIsComplex(eSrcType)) |
1919 | 0 | return CE_Failure; |
1920 | | |
1921 | 0 | double dfNoData{0}; |
1922 | 0 | const bool bHasNoData = CSLFindName(papszArgs, "NoData") != -1; |
1923 | 0 | if (bHasNoData && FetchDoubleArg(papszArgs, "NoData", &dfNoData) != CE_None) |
1924 | 0 | return CE_Failure; |
1925 | | |
1926 | | /* ---- Set pixels ---- */ |
1927 | 0 | size_t ii = 0; |
1928 | 0 | for (int iLine = 0; iLine < nYSize; ++iLine) |
1929 | 0 | { |
1930 | 0 | for (int iCol = 0; iCol < nXSize; ++iCol, ++ii) |
1931 | 0 | { |
1932 | | // Source raster pixels may be obtained with GetSrcVal macro. |
1933 | 0 | const double dfVal = GetSrcVal(papoSources[0], eSrcType, ii); |
1934 | 0 | const double dfPixVal = bHasNoData && IsNoData(dfVal, dfNoData) |
1935 | 0 | ? dfNoData |
1936 | 0 | : pow(base, dfVal * fact); |
1937 | |
|
1938 | 0 | GDALCopyWords(&dfPixVal, GDT_Float64, 0, |
1939 | 0 | static_cast<GByte *>(pData) + |
1940 | 0 | static_cast<GSpacing>(nLineSpace) * iLine + |
1941 | 0 | iCol * nPixelSpace, |
1942 | 0 | eBufType, nPixelSpace, 1); |
1943 | 0 | } |
1944 | 0 | } |
1945 | | |
1946 | | /* ---- Return success ---- */ |
1947 | 0 | return CE_None; |
1948 | 0 | } // ExpPixelFuncHelper |
1949 | | |
1950 | | static const char pszExpPixelFuncMetadata[] = |
1951 | | "<PixelFunctionArgumentsList>" |
1952 | | " <Argument name='base' description='Base' type='double' " |
1953 | | "default='2.7182818284590452353602874713526624' />" |
1954 | | " <Argument name='fact' description='Factor' type='double' default='1' />" |
1955 | | " <Argument type='builtin' value='NoData' optional='true' />" |
1956 | | "</PixelFunctionArgumentsList>"; |
1957 | | |
1958 | | static CPLErr ExpPixelFunc(void **papoSources, int nSources, void *pData, |
1959 | | int nXSize, int nYSize, GDALDataType eSrcType, |
1960 | | GDALDataType eBufType, int nPixelSpace, |
1961 | | int nLineSpace, CSLConstList papszArgs) |
1962 | 0 | { |
1963 | 0 | double dfBase = 2.7182818284590452353602874713526624; |
1964 | 0 | double dfFact = 1.; |
1965 | |
|
1966 | 0 | if (FetchDoubleArg(papszArgs, "base", &dfBase, &dfBase) != CE_None) |
1967 | 0 | return CE_Failure; |
1968 | | |
1969 | 0 | if (FetchDoubleArg(papszArgs, "fact", &dfFact, &dfFact) != CE_None) |
1970 | 0 | return CE_Failure; |
1971 | | |
1972 | 0 | return ExpPixelFuncHelper(papoSources, nSources, pData, nXSize, nYSize, |
1973 | 0 | eSrcType, eBufType, nPixelSpace, nLineSpace, |
1974 | 0 | papszArgs, dfBase, dfFact); |
1975 | 0 | } // ExpPixelFunc |
1976 | | |
1977 | | static CPLErr dB2AmpPixelFunc(void **papoSources, int nSources, void *pData, |
1978 | | int nXSize, int nYSize, GDALDataType eSrcType, |
1979 | | GDALDataType eBufType, int nPixelSpace, |
1980 | | int nLineSpace) |
1981 | 0 | { |
1982 | 0 | return ExpPixelFuncHelper(papoSources, nSources, pData, nXSize, nYSize, |
1983 | 0 | eSrcType, eBufType, nPixelSpace, nLineSpace, |
1984 | 0 | nullptr, 10.0, 1. / 20); |
1985 | 0 | } // dB2AmpPixelFunc |
1986 | | |
1987 | | static CPLErr dB2PowPixelFunc(void **papoSources, int nSources, void *pData, |
1988 | | int nXSize, int nYSize, GDALDataType eSrcType, |
1989 | | GDALDataType eBufType, int nPixelSpace, |
1990 | | int nLineSpace) |
1991 | 0 | { |
1992 | 0 | return ExpPixelFuncHelper(papoSources, nSources, pData, nXSize, nYSize, |
1993 | 0 | eSrcType, eBufType, nPixelSpace, nLineSpace, |
1994 | 0 | nullptr, 10.0, 1. / 10); |
1995 | 0 | } // dB2PowPixelFunc |
1996 | | |
1997 | | static const char pszPowPixelFuncMetadata[] = |
1998 | | "<PixelFunctionArgumentsList>" |
1999 | | " <Argument name='power' description='Exponent' type='double' " |
2000 | | "mandatory='1' />" |
2001 | | " <Argument type='builtin' value='NoData' optional='true' />" |
2002 | | "</PixelFunctionArgumentsList>"; |
2003 | | |
2004 | | static CPLErr PowPixelFunc(void **papoSources, int nSources, void *pData, |
2005 | | int nXSize, int nYSize, GDALDataType eSrcType, |
2006 | | GDALDataType eBufType, int nPixelSpace, |
2007 | | int nLineSpace, CSLConstList papszArgs) |
2008 | 0 | { |
2009 | | /* ---- Init ---- */ |
2010 | 0 | if (nSources != 1) |
2011 | 0 | return CE_Failure; |
2012 | 0 | if (GDALDataTypeIsComplex(eSrcType)) |
2013 | 0 | return CE_Failure; |
2014 | | |
2015 | 0 | double power; |
2016 | 0 | if (FetchDoubleArg(papszArgs, "power", &power) != CE_None) |
2017 | 0 | return CE_Failure; |
2018 | | |
2019 | 0 | double dfNoData{0}; |
2020 | 0 | const bool bHasNoData = CSLFindName(papszArgs, "NoData") != -1; |
2021 | 0 | if (bHasNoData && FetchDoubleArg(papszArgs, "NoData", &dfNoData) != CE_None) |
2022 | 0 | return CE_Failure; |
2023 | | |
2024 | | /* ---- Set pixels ---- */ |
2025 | 0 | size_t ii = 0; |
2026 | 0 | for (int iLine = 0; iLine < nYSize; ++iLine) |
2027 | 0 | { |
2028 | 0 | for (int iCol = 0; iCol < nXSize; ++iCol, ++ii) |
2029 | 0 | { |
2030 | 0 | const double dfVal = GetSrcVal(papoSources[0], eSrcType, ii); |
2031 | |
|
2032 | 0 | const double dfPixVal = bHasNoData && IsNoData(dfVal, dfNoData) |
2033 | 0 | ? dfNoData |
2034 | 0 | : std::pow(dfVal, power); |
2035 | |
|
2036 | 0 | GDALCopyWords(&dfPixVal, GDT_Float64, 0, |
2037 | 0 | static_cast<GByte *>(pData) + |
2038 | 0 | static_cast<GSpacing>(nLineSpace) * iLine + |
2039 | 0 | iCol * nPixelSpace, |
2040 | 0 | eBufType, nPixelSpace, 1); |
2041 | 0 | } |
2042 | 0 | } |
2043 | | |
2044 | | /* ---- Return success ---- */ |
2045 | 0 | return CE_None; |
2046 | 0 | } |
2047 | | |
2048 | | // Given nt intervals spaced by dt and beginning at t0, return the index of |
2049 | | // the lower bound of the interval that should be used to |
2050 | | // interpolate/extrapolate a value for t. |
2051 | | static std::size_t intervalLeft(double t0, double dt, std::size_t nt, double t) |
2052 | 0 | { |
2053 | 0 | if (t < t0) |
2054 | 0 | { |
2055 | 0 | return 0; |
2056 | 0 | } |
2057 | | |
2058 | 0 | std::size_t n = static_cast<std::size_t>((t - t0) / dt); |
2059 | |
|
2060 | 0 | if (n >= nt - 1) |
2061 | 0 | { |
2062 | 0 | return nt - 2; |
2063 | 0 | } |
2064 | | |
2065 | 0 | return n; |
2066 | 0 | } |
2067 | | |
2068 | | static double InterpolateLinear(double dfX0, double dfX1, double dfY0, |
2069 | | double dfY1, double dfX) |
2070 | 0 | { |
2071 | 0 | return dfY0 + (dfX - dfX0) * (dfY1 - dfY0) / (dfX1 - dfX0); |
2072 | 0 | } |
2073 | | |
2074 | | static double InterpolateExponential(double dfX0, double dfX1, double dfY0, |
2075 | | double dfY1, double dfX) |
2076 | 0 | { |
2077 | 0 | const double r = std::log(dfY1 / dfY0) / (dfX1 - dfX0); |
2078 | 0 | return dfY0 * std::exp(r * (dfX - dfX0)); |
2079 | 0 | } |
2080 | | |
2081 | | static const char pszInterpolatePixelFuncMetadata[] = |
2082 | | "<PixelFunctionArgumentsList>" |
2083 | | " <Argument name='t0' description='t0' type='double' mandatory='1' />" |
2084 | | " <Argument name='dt' description='dt' type='double' mandatory='1' />" |
2085 | | " <Argument name='t' description='t' type='double' mandatory='1' />" |
2086 | | " <Argument type='builtin' value='NoData' optional='true' />" |
2087 | | "</PixelFunctionArgumentsList>"; |
2088 | | |
2089 | | template <decltype(InterpolateLinear) InterpolationFunction> |
2090 | | CPLErr InterpolatePixelFunc(void **papoSources, int nSources, void *pData, |
2091 | | int nXSize, int nYSize, GDALDataType eSrcType, |
2092 | | GDALDataType eBufType, int nPixelSpace, |
2093 | | int nLineSpace, CSLConstList papszArgs) |
2094 | 0 | { |
2095 | | /* ---- Init ---- */ |
2096 | 0 | if (GDALDataTypeIsComplex(eSrcType)) |
2097 | 0 | return CE_Failure; |
2098 | | |
2099 | 0 | double dfT0; |
2100 | 0 | if (FetchDoubleArg(papszArgs, "t0", &dfT0) == CE_Failure) |
2101 | 0 | return CE_Failure; |
2102 | | |
2103 | 0 | double dfT; |
2104 | 0 | if (FetchDoubleArg(papszArgs, "t", &dfT) == CE_Failure) |
2105 | 0 | return CE_Failure; |
2106 | | |
2107 | 0 | double dfDt; |
2108 | 0 | if (FetchDoubleArg(papszArgs, "dt", &dfDt) == CE_Failure) |
2109 | 0 | return CE_Failure; |
2110 | | |
2111 | 0 | double dfNoData{0}; |
2112 | 0 | const bool bHasNoData = CSLFindName(papszArgs, "NoData") != -1; |
2113 | 0 | if (bHasNoData && FetchDoubleArg(papszArgs, "NoData", &dfNoData) != CE_None) |
2114 | 0 | return CE_Failure; |
2115 | | |
2116 | 0 | if (nSources < 2) |
2117 | 0 | { |
2118 | 0 | CPLError(CE_Failure, CPLE_AppDefined, |
2119 | 0 | "At least two sources required for interpolation."); |
2120 | 0 | return CE_Failure; |
2121 | 0 | } |
2122 | | |
2123 | 0 | if (dfT == 0 || !std::isfinite(dfT)) |
2124 | 0 | { |
2125 | 0 | CPLError(CE_Failure, CPLE_AppDefined, "dt must be finite and non-zero"); |
2126 | 0 | return CE_Failure; |
2127 | 0 | } |
2128 | | |
2129 | 0 | const auto i0 = intervalLeft(dfT0, dfDt, nSources, dfT); |
2130 | 0 | const auto i1 = i0 + 1; |
2131 | 0 | const double dfX0 = dfT0 + static_cast<double>(i0) * dfDt; |
2132 | 0 | const double dfX1 = dfT0 + static_cast<double>(i0 + 1) * dfDt; |
2133 | | |
2134 | | /* ---- Set pixels ---- */ |
2135 | 0 | size_t ii = 0; |
2136 | 0 | for (int iLine = 0; iLine < nYSize; ++iLine) |
2137 | 0 | { |
2138 | 0 | for (int iCol = 0; iCol < nXSize; ++iCol, ++ii) |
2139 | 0 | { |
2140 | 0 | const double dfY0 = GetSrcVal(papoSources[i0], eSrcType, ii); |
2141 | 0 | const double dfY1 = GetSrcVal(papoSources[i1], eSrcType, ii); |
2142 | |
|
2143 | 0 | double dfPixVal = dfNoData; |
2144 | 0 | if (dfT == dfX0) |
2145 | 0 | dfPixVal = dfY0; |
2146 | 0 | else if (dfT == dfX1) |
2147 | 0 | dfPixVal = dfY1; |
2148 | 0 | else if (!bHasNoData || |
2149 | 0 | (!IsNoData(dfY0, dfNoData) && !IsNoData(dfY1, dfNoData))) |
2150 | 0 | dfPixVal = InterpolationFunction(dfX0, dfX1, dfY0, dfY1, dfT); |
2151 | |
|
2152 | 0 | GDALCopyWords(&dfPixVal, GDT_Float64, 0, |
2153 | 0 | static_cast<GByte *>(pData) + |
2154 | 0 | static_cast<GSpacing>(nLineSpace) * iLine + |
2155 | 0 | iCol * nPixelSpace, |
2156 | 0 | eBufType, nPixelSpace, 1); |
2157 | 0 | } |
2158 | 0 | } |
2159 | | |
2160 | | /* ---- Return success ---- */ |
2161 | 0 | return CE_None; |
2162 | 0 | } Unexecuted instantiation: pixelfunctions.cpp:CPLErr InterpolatePixelFunc<&(InterpolateLinear(double, double, double, double, double))>(void**, int, void*, int, int, GDALDataType, GDALDataType, int, int, char const* const*) Unexecuted instantiation: pixelfunctions.cpp:CPLErr InterpolatePixelFunc<&(InterpolateExponential(double, double, double, double, double))>(void**, int, void*, int, int, GDALDataType, GDALDataType, int, int, char const* const*) |
2163 | | |
2164 | | static const char pszReplaceNoDataPixelFuncMetadata[] = |
2165 | | "<PixelFunctionArgumentsList>" |
2166 | | " <Argument type='builtin' value='NoData' />" |
2167 | | " <Argument name='to' type='double' description='New NoData value to be " |
2168 | | "replaced' default='nan' />" |
2169 | | "</PixelFunctionArgumentsList>"; |
2170 | | |
2171 | | static CPLErr ReplaceNoDataPixelFunc(void **papoSources, int nSources, |
2172 | | void *pData, int nXSize, int nYSize, |
2173 | | GDALDataType eSrcType, |
2174 | | GDALDataType eBufType, int nPixelSpace, |
2175 | | int nLineSpace, CSLConstList papszArgs) |
2176 | 0 | { |
2177 | | /* ---- Init ---- */ |
2178 | 0 | if (nSources != 1) |
2179 | 0 | return CE_Failure; |
2180 | 0 | if (GDALDataTypeIsComplex(eSrcType)) |
2181 | 0 | { |
2182 | 0 | CPLError(CE_Failure, CPLE_AppDefined, |
2183 | 0 | "replace_nodata cannot convert complex data types"); |
2184 | 0 | return CE_Failure; |
2185 | 0 | } |
2186 | | |
2187 | 0 | double dfOldNoData, dfNewNoData = NAN; |
2188 | 0 | if (FetchDoubleArg(papszArgs, "NoData", &dfOldNoData) != CE_None) |
2189 | 0 | return CE_Failure; |
2190 | 0 | if (FetchDoubleArg(papszArgs, "to", &dfNewNoData, &dfNewNoData) != CE_None) |
2191 | 0 | return CE_Failure; |
2192 | | |
2193 | 0 | if (!GDALDataTypeIsFloating(eBufType) && std::isnan(dfNewNoData)) |
2194 | 0 | { |
2195 | 0 | CPLError(CE_Failure, CPLE_AppDefined, |
2196 | 0 | "Using nan requires a floating point type output buffer"); |
2197 | 0 | return CE_Failure; |
2198 | 0 | } |
2199 | | |
2200 | | /* ---- Set pixels ---- */ |
2201 | 0 | size_t ii = 0; |
2202 | 0 | for (int iLine = 0; iLine < nYSize; ++iLine) |
2203 | 0 | { |
2204 | 0 | for (int iCol = 0; iCol < nXSize; ++iCol, ++ii) |
2205 | 0 | { |
2206 | 0 | double dfPixVal = GetSrcVal(papoSources[0], eSrcType, ii); |
2207 | 0 | if (dfPixVal == dfOldNoData || std::isnan(dfPixVal)) |
2208 | 0 | dfPixVal = dfNewNoData; |
2209 | |
|
2210 | 0 | GDALCopyWords(&dfPixVal, GDT_Float64, 0, |
2211 | 0 | static_cast<GByte *>(pData) + |
2212 | 0 | static_cast<GSpacing>(nLineSpace) * iLine + |
2213 | 0 | iCol * nPixelSpace, |
2214 | 0 | eBufType, nPixelSpace, 1); |
2215 | 0 | } |
2216 | 0 | } |
2217 | | |
2218 | | /* ---- Return success ---- */ |
2219 | 0 | return CE_None; |
2220 | 0 | } |
2221 | | |
2222 | | static const char pszScalePixelFuncMetadata[] = |
2223 | | "<PixelFunctionArgumentsList>" |
2224 | | " <Argument type='builtin' value='offset' />" |
2225 | | " <Argument type='builtin' value='scale' />" |
2226 | | " <Argument type='builtin' value='NoData' optional='true' />" |
2227 | | "</PixelFunctionArgumentsList>"; |
2228 | | |
2229 | | static CPLErr ScalePixelFunc(void **papoSources, int nSources, void *pData, |
2230 | | int nXSize, int nYSize, GDALDataType eSrcType, |
2231 | | GDALDataType eBufType, int nPixelSpace, |
2232 | | int nLineSpace, CSLConstList papszArgs) |
2233 | 0 | { |
2234 | | /* ---- Init ---- */ |
2235 | 0 | if (nSources != 1) |
2236 | 0 | return CE_Failure; |
2237 | 0 | if (GDALDataTypeIsComplex(eSrcType)) |
2238 | 0 | { |
2239 | 0 | CPLError(CE_Failure, CPLE_AppDefined, |
2240 | 0 | "scale cannot by applied to complex data types"); |
2241 | 0 | return CE_Failure; |
2242 | 0 | } |
2243 | | |
2244 | 0 | double dfNoData{0}; |
2245 | 0 | const bool bHasNoData = CSLFindName(papszArgs, "NoData") != -1; |
2246 | 0 | if (bHasNoData && FetchDoubleArg(papszArgs, "NoData", &dfNoData) != CE_None) |
2247 | 0 | return CE_Failure; |
2248 | | |
2249 | 0 | double dfScale, dfOffset; |
2250 | 0 | if (FetchDoubleArg(papszArgs, "scale", &dfScale) != CE_None) |
2251 | 0 | return CE_Failure; |
2252 | 0 | if (FetchDoubleArg(papszArgs, "offset", &dfOffset) != CE_None) |
2253 | 0 | return CE_Failure; |
2254 | | |
2255 | | /* ---- Set pixels ---- */ |
2256 | 0 | size_t ii = 0; |
2257 | 0 | for (int iLine = 0; iLine < nYSize; ++iLine) |
2258 | 0 | { |
2259 | 0 | for (int iCol = 0; iCol < nXSize; ++iCol, ++ii) |
2260 | 0 | { |
2261 | 0 | const double dfVal = GetSrcVal(papoSources[0], eSrcType, ii); |
2262 | |
|
2263 | 0 | const double dfPixVal = bHasNoData && IsNoData(dfVal, dfNoData) |
2264 | 0 | ? dfNoData |
2265 | 0 | : dfVal * dfScale + dfOffset; |
2266 | |
|
2267 | 0 | GDALCopyWords(&dfPixVal, GDT_Float64, 0, |
2268 | 0 | static_cast<GByte *>(pData) + |
2269 | 0 | static_cast<GSpacing>(nLineSpace) * iLine + |
2270 | 0 | iCol * nPixelSpace, |
2271 | 0 | eBufType, nPixelSpace, 1); |
2272 | 0 | } |
2273 | 0 | } |
2274 | | |
2275 | | /* ---- Return success ---- */ |
2276 | 0 | return CE_None; |
2277 | 0 | } |
2278 | | |
2279 | | static const char pszNormDiffPixelFuncMetadata[] = |
2280 | | "<PixelFunctionArgumentsList>" |
2281 | | " <Argument type='builtin' value='NoData' optional='true' />" |
2282 | | "</PixelFunctionArgumentsList>"; |
2283 | | |
2284 | | static CPLErr NormDiffPixelFunc(void **papoSources, int nSources, void *pData, |
2285 | | int nXSize, int nYSize, GDALDataType eSrcType, |
2286 | | GDALDataType eBufType, int nPixelSpace, |
2287 | | int nLineSpace, CSLConstList papszArgs) |
2288 | 0 | { |
2289 | | /* ---- Init ---- */ |
2290 | 0 | if (nSources != 2) |
2291 | 0 | return CE_Failure; |
2292 | | |
2293 | 0 | if (GDALDataTypeIsComplex(eSrcType)) |
2294 | 0 | { |
2295 | 0 | CPLError(CE_Failure, CPLE_AppDefined, |
2296 | 0 | "norm_diff cannot by applied to complex data types"); |
2297 | 0 | return CE_Failure; |
2298 | 0 | } |
2299 | | |
2300 | 0 | double dfNoData{0}; |
2301 | 0 | const bool bHasNoData = CSLFindName(papszArgs, "NoData") != -1; |
2302 | 0 | if (bHasNoData && FetchDoubleArg(papszArgs, "NoData", &dfNoData) != CE_None) |
2303 | 0 | return CE_Failure; |
2304 | | |
2305 | | /* ---- Set pixels ---- */ |
2306 | 0 | size_t ii = 0; |
2307 | 0 | for (int iLine = 0; iLine < nYSize; ++iLine) |
2308 | 0 | { |
2309 | 0 | for (int iCol = 0; iCol < nXSize; ++iCol, ++ii) |
2310 | 0 | { |
2311 | 0 | const double dfLeftVal = GetSrcVal(papoSources[0], eSrcType, ii); |
2312 | 0 | const double dfRightVal = GetSrcVal(papoSources[1], eSrcType, ii); |
2313 | |
|
2314 | 0 | double dfPixVal = dfNoData; |
2315 | |
|
2316 | 0 | if (!bHasNoData || (!IsNoData(dfLeftVal, dfNoData) && |
2317 | 0 | !IsNoData(dfRightVal, dfNoData))) |
2318 | 0 | { |
2319 | 0 | const double dfDenom = (dfLeftVal + dfRightVal); |
2320 | | // coverity[divide_by_zero] |
2321 | 0 | dfPixVal = |
2322 | 0 | dfDenom == 0 |
2323 | 0 | ? std::numeric_limits<double>::infinity() |
2324 | 0 | : (dfLeftVal - dfRightVal) / |
2325 | | #ifdef __COVERITY__ |
2326 | | (dfDenom + std::numeric_limits<double>::min()) |
2327 | | #else |
2328 | 0 | dfDenom |
2329 | 0 | #endif |
2330 | 0 | ; |
2331 | 0 | } |
2332 | |
|
2333 | 0 | GDALCopyWords(&dfPixVal, GDT_Float64, 0, |
2334 | 0 | static_cast<GByte *>(pData) + |
2335 | 0 | static_cast<GSpacing>(nLineSpace) * iLine + |
2336 | 0 | iCol * nPixelSpace, |
2337 | 0 | eBufType, nPixelSpace, 1); |
2338 | 0 | } |
2339 | 0 | } |
2340 | | |
2341 | | /* ---- Return success ---- */ |
2342 | 0 | return CE_None; |
2343 | 0 | } // NormDiffPixelFunc |
2344 | | |
2345 | | /************************************************************************/ |
2346 | | /* pszMinMaxFuncMetadataNodata */ |
2347 | | /************************************************************************/ |
2348 | | |
2349 | | static const char pszMinMaxFuncMetadataNodata[] = |
2350 | | "<PixelFunctionArgumentsList>" |
2351 | | " <Argument name='k' description='Optional constant term' type='double' " |
2352 | | "default='nan' />" |
2353 | | " <Argument type='builtin' value='NoData' optional='true' />" |
2354 | | " <Argument name='propagateNoData' description='Whether the output value " |
2355 | | "should be NoData as as soon as one source is NoData' type='boolean' " |
2356 | | "default='false' />" |
2357 | | "</PixelFunctionArgumentsList>"; |
2358 | | |
2359 | | template <class Comparator> |
2360 | | static CPLErr MinOrMaxPixelFunc(double dfK, void **papoSources, int nSources, |
2361 | | void *pData, int nXSize, int nYSize, |
2362 | | GDALDataType eSrcType, GDALDataType eBufType, |
2363 | | int nPixelSpace, int nLineSpace, |
2364 | | CSLConstList papszArgs) |
2365 | 0 | { |
2366 | | /* ---- Init ---- */ |
2367 | 0 | if (GDALDataTypeIsComplex(eSrcType)) |
2368 | 0 | { |
2369 | 0 | CPLError(CE_Failure, CPLE_AppDefined, |
2370 | 0 | "Complex data type not supported for min/max()."); |
2371 | 0 | return CE_Failure; |
2372 | 0 | } |
2373 | | |
2374 | 0 | double dfNoData = std::numeric_limits<double>::quiet_NaN(); |
2375 | 0 | if (FetchDoubleArg(papszArgs, "NoData", &dfNoData, &dfNoData) != CE_None) |
2376 | 0 | return CE_Failure; |
2377 | 0 | const bool bPropagateNoData = CPLTestBool( |
2378 | 0 | CSLFetchNameValueDef(papszArgs, "propagateNoData", "false")); |
2379 | | |
2380 | | /* ---- Set pixels ---- */ |
2381 | 0 | size_t ii = 0; |
2382 | 0 | for (int iLine = 0; iLine < nYSize; ++iLine) |
2383 | 0 | { |
2384 | 0 | for (int iCol = 0; iCol < nXSize; ++iCol, ++ii) |
2385 | 0 | { |
2386 | 0 | double dfRes = std::numeric_limits<double>::quiet_NaN(); |
2387 | |
|
2388 | 0 | for (int iSrc = 0; iSrc < nSources; ++iSrc) |
2389 | 0 | { |
2390 | 0 | const double dfVal = GetSrcVal(papoSources[iSrc], eSrcType, ii); |
2391 | |
|
2392 | 0 | if (std::isnan(dfVal) || dfVal == dfNoData) |
2393 | 0 | { |
2394 | 0 | if (bPropagateNoData) |
2395 | 0 | { |
2396 | 0 | dfRes = dfNoData; |
2397 | 0 | break; |
2398 | 0 | } |
2399 | 0 | } |
2400 | 0 | else if (Comparator::compare(dfVal, dfRes)) |
2401 | 0 | { |
2402 | 0 | dfRes = dfVal; |
2403 | 0 | } |
2404 | 0 | } |
2405 | |
|
2406 | 0 | if (std::isnan(dfRes)) |
2407 | 0 | { |
2408 | 0 | if (!bPropagateNoData) |
2409 | 0 | dfRes = dfNoData; |
2410 | 0 | } |
2411 | 0 | else if (!std::isnan(dfK) && Comparator::compare(dfK, dfRes)) |
2412 | 0 | { |
2413 | 0 | dfRes = dfK; |
2414 | 0 | } |
2415 | |
|
2416 | 0 | GDALCopyWords(&dfRes, GDT_Float64, 0, |
2417 | 0 | static_cast<GByte *>(pData) + |
2418 | 0 | static_cast<GSpacing>(nLineSpace) * iLine + |
2419 | 0 | iCol * nPixelSpace, |
2420 | 0 | eBufType, nPixelSpace, 1); |
2421 | 0 | } |
2422 | 0 | } |
2423 | | |
2424 | | /* ---- Return success ---- */ |
2425 | 0 | return CE_None; |
2426 | 0 | } /* MinOrMaxPixelFunc */ Unexecuted instantiation: pixelfunctions.cpp:CPLErr MinOrMaxPixelFunc<MinPixelFunc(void**, int, void*, int, int, GDALDataType, GDALDataType, int, int, char const* const*)::Comparator>(double, void**, int, void*, int, int, GDALDataType, GDALDataType, int, int, char const* const*) Unexecuted instantiation: pixelfunctions.cpp:CPLErr MinOrMaxPixelFunc<MaxPixelFunc(void**, int, void*, int, int, GDALDataType, GDALDataType, int, int, char const* const*)::Comparator>(double, void**, int, void*, int, int, GDALDataType, GDALDataType, int, int, char const* const*) |
2427 | | |
2428 | | #ifdef USE_SSE2 |
2429 | | |
2430 | | template <class T, class SSEWrapper> |
2431 | | static void OptimizedMinOrMaxSSE2(const void *const *papoSources, int nSources, |
2432 | | void *pData, int nXSize, int nYSize, |
2433 | | int nLineSpace) |
2434 | 0 | { |
2435 | 0 | assert(nSources >= 1); |
2436 | 0 | constexpr int VALUES_PER_REG = |
2437 | 0 | static_cast<int>(sizeof(typename SSEWrapper::Vec) / sizeof(T)); |
2438 | 0 | for (int iLine = 0; iLine < nYSize; ++iLine) |
2439 | 0 | { |
2440 | 0 | T *CPL_RESTRICT pDest = |
2441 | 0 | reinterpret_cast<T *>(static_cast<GByte *>(pData) + |
2442 | 0 | static_cast<GSpacing>(nLineSpace) * iLine); |
2443 | 0 | const size_t iOffsetLine = static_cast<size_t>(iLine) * nXSize; |
2444 | 0 | int iCol = 0; |
2445 | 0 | for (; iCol < nXSize - (2 * VALUES_PER_REG - 1); |
2446 | 0 | iCol += 2 * VALUES_PER_REG) |
2447 | 0 | { |
2448 | 0 | auto reg0 = SSEWrapper::LoadU( |
2449 | 0 | static_cast<const T * CPL_RESTRICT>(papoSources[0]) + |
2450 | 0 | iOffsetLine + iCol); |
2451 | 0 | auto reg1 = SSEWrapper::LoadU( |
2452 | 0 | static_cast<const T * CPL_RESTRICT>(papoSources[0]) + |
2453 | 0 | iOffsetLine + iCol + VALUES_PER_REG); |
2454 | 0 | for (int iSrc = 1; iSrc < nSources; ++iSrc) |
2455 | 0 | { |
2456 | 0 | reg0 = SSEWrapper::MinOrMax( |
2457 | 0 | reg0, SSEWrapper::LoadU(static_cast<const T * CPL_RESTRICT>( |
2458 | 0 | papoSources[iSrc]) + |
2459 | 0 | iOffsetLine + iCol)); |
2460 | 0 | reg1 = SSEWrapper::MinOrMax( |
2461 | 0 | reg1, |
2462 | 0 | SSEWrapper::LoadU( |
2463 | 0 | static_cast<const T * CPL_RESTRICT>(papoSources[iSrc]) + |
2464 | 0 | iOffsetLine + iCol + VALUES_PER_REG)); |
2465 | 0 | } |
2466 | 0 | SSEWrapper::StoreU(pDest + iCol, reg0); |
2467 | 0 | SSEWrapper::StoreU(pDest + iCol + VALUES_PER_REG, reg1); |
2468 | 0 | } |
2469 | 0 | for (; iCol < nXSize; ++iCol) |
2470 | 0 | { |
2471 | 0 | T v = static_cast<const T * CPL_RESTRICT>( |
2472 | 0 | papoSources[0])[iOffsetLine + iCol]; |
2473 | 0 | for (int iSrc = 1; iSrc < nSources; ++iSrc) |
2474 | 0 | { |
2475 | 0 | v = SSEWrapper::MinOrMax( |
2476 | 0 | v, static_cast<const T * CPL_RESTRICT>( |
2477 | 0 | papoSources[iSrc])[iOffsetLine + iCol]); |
2478 | 0 | } |
2479 | 0 | pDest[iCol] = v; |
2480 | 0 | } |
2481 | 0 | } |
2482 | 0 | } Unexecuted instantiation: pixelfunctions.cpp:void OptimizedMinOrMaxSSE2<unsigned char, (anonymous namespace)::SSEWrapperMinByte>(void const* const*, int, void*, int, int, int) Unexecuted instantiation: pixelfunctions.cpp:void OptimizedMinOrMaxSSE2<unsigned short, (anonymous namespace)::SSEWrapperMinUInt16>(void const* const*, int, void*, int, int, int) Unexecuted instantiation: pixelfunctions.cpp:void OptimizedMinOrMaxSSE2<short, (anonymous namespace)::SSEWrapperMinInt16>(void const* const*, int, void*, int, int, int) Unexecuted instantiation: pixelfunctions.cpp:void OptimizedMinOrMaxSSE2<float, (anonymous namespace)::SSEWrapperMinFloat>(void const* const*, int, void*, int, int, int) Unexecuted instantiation: pixelfunctions.cpp:void OptimizedMinOrMaxSSE2<double, (anonymous namespace)::SSEWrapperMinDouble>(void const* const*, int, void*, int, int, int) Unexecuted instantiation: pixelfunctions.cpp:void OptimizedMinOrMaxSSE2<unsigned char, (anonymous namespace)::SSEWrapperMaxByte>(void const* const*, int, void*, int, int, int) Unexecuted instantiation: pixelfunctions.cpp:void OptimizedMinOrMaxSSE2<unsigned short, (anonymous namespace)::SSEWrapperMaxUInt16>(void const* const*, int, void*, int, int, int) Unexecuted instantiation: pixelfunctions.cpp:void OptimizedMinOrMaxSSE2<short, (anonymous namespace)::SSEWrapperMaxInt16>(void const* const*, int, void*, int, int, int) Unexecuted instantiation: pixelfunctions.cpp:void OptimizedMinOrMaxSSE2<float, (anonymous namespace)::SSEWrapperMaxFloat>(void const* const*, int, void*, int, int, int) Unexecuted instantiation: pixelfunctions.cpp:void OptimizedMinOrMaxSSE2<double, (anonymous namespace)::SSEWrapperMaxDouble>(void const* const*, int, void*, int, int, int) |
2483 | | |
2484 | | // clang-format off |
2485 | | namespace |
2486 | | { |
2487 | | struct SSEWrapperMinByte |
2488 | | { |
2489 | | using T = uint8_t; |
2490 | | typedef __m128i Vec; |
2491 | | |
2492 | 0 | static inline Vec LoadU(const T *x) { return _mm_loadu_si128(reinterpret_cast<const Vec*>(x)); } |
2493 | 0 | static inline void StoreU(T *x, Vec y) { _mm_storeu_si128(reinterpret_cast<Vec*>(x), y); } |
2494 | 0 | static inline Vec MinOrMax(Vec x, Vec y) { return _mm_min_epu8(x, y); } |
2495 | 0 | static inline T MinOrMax(T x, T y) { return std::min(x, y); } |
2496 | | }; |
2497 | | |
2498 | | struct SSEWrapperMaxByte |
2499 | | { |
2500 | | using T = uint8_t; |
2501 | | typedef __m128i Vec; |
2502 | | |
2503 | 0 | static inline Vec LoadU(const T *x) { return _mm_loadu_si128(reinterpret_cast<const Vec*>(x)); } |
2504 | 0 | static inline void StoreU(T *x, Vec y) { _mm_storeu_si128(reinterpret_cast<Vec*>(x), y); } |
2505 | 0 | static inline Vec MinOrMax(Vec x, Vec y) { return _mm_max_epu8(x, y); } |
2506 | 0 | static inline T MinOrMax(T x, T y) { return std::max(x, y); } |
2507 | | }; |
2508 | | |
2509 | | struct SSEWrapperMinUInt16 |
2510 | | { |
2511 | | using T = uint16_t; |
2512 | | typedef __m128i Vec; |
2513 | | |
2514 | 0 | static inline Vec LoadU(const T *x) { return _mm_loadu_si128(reinterpret_cast<const Vec*>(x)); } |
2515 | 0 | static inline void StoreU(T *x, Vec y) { _mm_storeu_si128(reinterpret_cast<Vec*>(x), y); } |
2516 | | #if defined(__SSE4_1__) || defined(USE_NEON_OPTIMIZATIONS) |
2517 | | static inline Vec MinOrMax(Vec x, Vec y) { return _mm_min_epu16(x, y); } |
2518 | | #else |
2519 | 0 | static inline Vec MinOrMax(Vec x, Vec y) { return |
2520 | 0 | _mm_add_epi16( |
2521 | 0 | _mm_min_epi16( |
2522 | 0 | _mm_add_epi16(x, _mm_set1_epi16(-32768)), |
2523 | 0 | _mm_add_epi16(y, _mm_set1_epi16(-32768))), |
2524 | 0 | _mm_set1_epi16(-32768)); } |
2525 | | #endif |
2526 | 0 | static inline T MinOrMax(T x, T y) { return std::min(x, y); } |
2527 | | }; |
2528 | | |
2529 | | struct SSEWrapperMaxUInt16 |
2530 | | { |
2531 | | using T = uint16_t; |
2532 | | typedef __m128i Vec; |
2533 | | |
2534 | 0 | static inline Vec LoadU(const T *x) { return _mm_loadu_si128(reinterpret_cast<const Vec*>(x)); } |
2535 | 0 | static inline void StoreU(T *x, Vec y) { _mm_storeu_si128(reinterpret_cast<Vec*>(x), y); } |
2536 | | #if defined(__SSE4_1__) || defined(USE_NEON_OPTIMIZATIONS) |
2537 | | static inline Vec MinOrMax(Vec x, Vec y) { return _mm_max_epu16(x, y); } |
2538 | | #else |
2539 | 0 | static inline Vec MinOrMax(Vec x, Vec y) { return |
2540 | 0 | _mm_add_epi16( |
2541 | 0 | _mm_max_epi16( |
2542 | 0 | _mm_add_epi16(x, _mm_set1_epi16(-32768)), |
2543 | 0 | _mm_add_epi16(y, _mm_set1_epi16(-32768))), |
2544 | 0 | _mm_set1_epi16(-32768)); } |
2545 | | #endif |
2546 | 0 | static inline T MinOrMax(T x, T y) { return std::max(x, y); } |
2547 | | }; |
2548 | | |
2549 | | struct SSEWrapperMinInt16 |
2550 | | { |
2551 | | using T = int16_t; |
2552 | | typedef __m128i Vec; |
2553 | | |
2554 | 0 | static inline Vec LoadU(const T *x) { return _mm_loadu_si128(reinterpret_cast<const Vec*>(x)); } |
2555 | 0 | static inline void StoreU(T *x, Vec y) { _mm_storeu_si128(reinterpret_cast<Vec*>(x), y); } |
2556 | 0 | static inline Vec MinOrMax(Vec x, Vec y) { return _mm_min_epi16(x, y); } |
2557 | 0 | static inline T MinOrMax(T x, T y) { return std::min(x, y); } |
2558 | | }; |
2559 | | |
2560 | | struct SSEWrapperMaxInt16 |
2561 | | { |
2562 | | using T = int16_t; |
2563 | | typedef __m128i Vec; |
2564 | | |
2565 | 0 | static inline Vec LoadU(const T *x) { return _mm_loadu_si128(reinterpret_cast<const Vec*>(x)); } |
2566 | 0 | static inline void StoreU(T *x, Vec y) { _mm_storeu_si128(reinterpret_cast<Vec*>(x), y); } |
2567 | 0 | static inline Vec MinOrMax(Vec x, Vec y) { return _mm_max_epi16(x, y); } |
2568 | 0 | static inline T MinOrMax(T x, T y) { return std::max(x, y); } |
2569 | | }; |
2570 | | |
2571 | | struct SSEWrapperMinFloat |
2572 | | { |
2573 | | using T = float; |
2574 | | typedef __m128 Vec; |
2575 | | |
2576 | 0 | static inline Vec LoadU(const T *x) { return _mm_loadu_ps(x); } |
2577 | 0 | static inline void StoreU(T *x, Vec y) { _mm_storeu_ps(x, y); } |
2578 | 0 | static inline Vec MinOrMax(Vec x, Vec y) { return _mm_min_ps(x, y); } |
2579 | 0 | static inline T MinOrMax(T x, T y) { return std::min(x, y); } |
2580 | | }; |
2581 | | |
2582 | | struct SSEWrapperMaxFloat |
2583 | | { |
2584 | | using T = float; |
2585 | | typedef __m128 Vec; |
2586 | | |
2587 | 0 | static inline Vec LoadU(const T *x) { return _mm_loadu_ps(x); } |
2588 | 0 | static inline void StoreU(T *x, Vec y) { _mm_storeu_ps(x, y); } |
2589 | 0 | static inline Vec MinOrMax(Vec x, Vec y) { return _mm_max_ps(x, y); } |
2590 | 0 | static inline T MinOrMax(T x, T y) { return std::max(x, y); } |
2591 | | }; |
2592 | | |
2593 | | struct SSEWrapperMinDouble |
2594 | | { |
2595 | | using T = double; |
2596 | | typedef __m128d Vec; |
2597 | | |
2598 | 0 | static inline Vec LoadU(const T *x) { return _mm_loadu_pd(x); } |
2599 | 0 | static inline void StoreU(T *x, Vec y) { _mm_storeu_pd(x, y); } |
2600 | 0 | static inline Vec MinOrMax(Vec x, Vec y) { return _mm_min_pd(x, y); } |
2601 | 0 | static inline T MinOrMax(T x, T y) { return std::min(x, y); } |
2602 | | }; |
2603 | | |
2604 | | struct SSEWrapperMaxDouble |
2605 | | { |
2606 | | using T = double; |
2607 | | typedef __m128d Vec; |
2608 | | |
2609 | 0 | static inline Vec LoadU(const T *x) { return _mm_loadu_pd(x); } |
2610 | 0 | static inline void StoreU(T *x, Vec y) { _mm_storeu_pd(x, y); } |
2611 | 0 | static inline Vec MinOrMax(Vec x, Vec y) { return _mm_max_pd(x, y); } |
2612 | 0 | static inline T MinOrMax(T x, T y) { return std::max(x, y); } |
2613 | | }; |
2614 | | |
2615 | | } // namespace |
2616 | | |
2617 | | // clang-format on |
2618 | | |
2619 | | #endif // USE_SSE2 |
2620 | | |
2621 | | static CPLErr MinPixelFunc(void **papoSources, int nSources, void *pData, |
2622 | | int nXSize, int nYSize, GDALDataType eSrcType, |
2623 | | GDALDataType eBufType, int nPixelSpace, |
2624 | | int nLineSpace, CSLConstList papszArgs) |
2625 | 0 | { |
2626 | 0 | struct Comparator |
2627 | 0 | { |
2628 | 0 | static bool compare(double x, double resVal) |
2629 | 0 | { |
2630 | | // Written this way to deal with resVal being NaN |
2631 | 0 | return !(x >= resVal); |
2632 | 0 | } |
2633 | 0 | }; |
2634 | |
|
2635 | 0 | double dfK = std::numeric_limits<double>::quiet_NaN(); |
2636 | 0 | if (FetchDoubleArg(papszArgs, "k", &dfK, &dfK) != CE_None) |
2637 | 0 | return CE_Failure; |
2638 | | |
2639 | 0 | #ifdef USE_SSE2 |
2640 | 0 | const bool bHasNoData = CSLFindName(papszArgs, "NoData") != -1; |
2641 | 0 | if (std::isnan(dfK) && nSources > 0 && !bHasNoData && |
2642 | 0 | eSrcType == eBufType && |
2643 | 0 | nPixelSpace == GDALGetDataTypeSizeBytes(eSrcType)) |
2644 | 0 | { |
2645 | 0 | if (eSrcType == GDT_Byte) |
2646 | 0 | { |
2647 | 0 | OptimizedMinOrMaxSSE2<uint8_t, SSEWrapperMinByte>( |
2648 | 0 | papoSources, nSources, pData, nXSize, nYSize, nLineSpace); |
2649 | 0 | return CE_None; |
2650 | 0 | } |
2651 | 0 | else if (eSrcType == GDT_UInt16) |
2652 | 0 | { |
2653 | 0 | OptimizedMinOrMaxSSE2<uint16_t, SSEWrapperMinUInt16>( |
2654 | 0 | papoSources, nSources, pData, nXSize, nYSize, nLineSpace); |
2655 | 0 | return CE_None; |
2656 | 0 | } |
2657 | 0 | else if (eSrcType == GDT_Int16) |
2658 | 0 | { |
2659 | 0 | OptimizedMinOrMaxSSE2<int16_t, SSEWrapperMinInt16>( |
2660 | 0 | papoSources, nSources, pData, nXSize, nYSize, nLineSpace); |
2661 | 0 | return CE_None; |
2662 | 0 | } |
2663 | 0 | else if (eSrcType == GDT_Float32) |
2664 | 0 | { |
2665 | 0 | OptimizedMinOrMaxSSE2<float, SSEWrapperMinFloat>( |
2666 | 0 | papoSources, nSources, pData, nXSize, nYSize, nLineSpace); |
2667 | 0 | return CE_None; |
2668 | 0 | } |
2669 | 0 | else if (eSrcType == GDT_Float64) |
2670 | 0 | { |
2671 | 0 | OptimizedMinOrMaxSSE2<double, SSEWrapperMinDouble>( |
2672 | 0 | papoSources, nSources, pData, nXSize, nYSize, nLineSpace); |
2673 | 0 | return CE_None; |
2674 | 0 | } |
2675 | 0 | } |
2676 | 0 | #endif |
2677 | | |
2678 | 0 | return MinOrMaxPixelFunc<Comparator>(dfK, papoSources, nSources, pData, |
2679 | 0 | nXSize, nYSize, eSrcType, eBufType, |
2680 | 0 | nPixelSpace, nLineSpace, papszArgs); |
2681 | 0 | } |
2682 | | |
2683 | | static CPLErr MaxPixelFunc(void **papoSources, int nSources, void *pData, |
2684 | | int nXSize, int nYSize, GDALDataType eSrcType, |
2685 | | GDALDataType eBufType, int nPixelSpace, |
2686 | | int nLineSpace, CSLConstList papszArgs) |
2687 | 0 | { |
2688 | 0 | struct Comparator |
2689 | 0 | { |
2690 | 0 | static bool compare(double x, double resVal) |
2691 | 0 | { |
2692 | | // Written this way to deal with resVal being NaN |
2693 | 0 | return !(x <= resVal); |
2694 | 0 | } |
2695 | 0 | }; |
2696 | |
|
2697 | 0 | double dfK = std::numeric_limits<double>::quiet_NaN(); |
2698 | 0 | if (FetchDoubleArg(papszArgs, "k", &dfK, &dfK) != CE_None) |
2699 | 0 | return CE_Failure; |
2700 | | |
2701 | 0 | #ifdef USE_SSE2 |
2702 | 0 | const bool bHasNoData = CSLFindName(papszArgs, "NoData") != -1; |
2703 | 0 | if (std::isnan(dfK) && nSources > 0 && !bHasNoData && |
2704 | 0 | eSrcType == eBufType && |
2705 | 0 | nPixelSpace == GDALGetDataTypeSizeBytes(eSrcType)) |
2706 | 0 | { |
2707 | 0 | if (eSrcType == GDT_Byte) |
2708 | 0 | { |
2709 | 0 | OptimizedMinOrMaxSSE2<uint8_t, SSEWrapperMaxByte>( |
2710 | 0 | papoSources, nSources, pData, nXSize, nYSize, nLineSpace); |
2711 | 0 | return CE_None; |
2712 | 0 | } |
2713 | 0 | else if (eSrcType == GDT_UInt16) |
2714 | 0 | { |
2715 | 0 | OptimizedMinOrMaxSSE2<uint16_t, SSEWrapperMaxUInt16>( |
2716 | 0 | papoSources, nSources, pData, nXSize, nYSize, nLineSpace); |
2717 | 0 | return CE_None; |
2718 | 0 | } |
2719 | 0 | else if (eSrcType == GDT_Int16) |
2720 | 0 | { |
2721 | 0 | OptimizedMinOrMaxSSE2<int16_t, SSEWrapperMaxInt16>( |
2722 | 0 | papoSources, nSources, pData, nXSize, nYSize, nLineSpace); |
2723 | 0 | return CE_None; |
2724 | 0 | } |
2725 | 0 | else if (eSrcType == GDT_Float32) |
2726 | 0 | { |
2727 | 0 | OptimizedMinOrMaxSSE2<float, SSEWrapperMaxFloat>( |
2728 | 0 | papoSources, nSources, pData, nXSize, nYSize, nLineSpace); |
2729 | 0 | return CE_None; |
2730 | 0 | } |
2731 | 0 | else if (eSrcType == GDT_Float64) |
2732 | 0 | { |
2733 | 0 | OptimizedMinOrMaxSSE2<double, SSEWrapperMaxDouble>( |
2734 | 0 | papoSources, nSources, pData, nXSize, nYSize, nLineSpace); |
2735 | 0 | return CE_None; |
2736 | 0 | } |
2737 | 0 | } |
2738 | 0 | #endif |
2739 | | |
2740 | 0 | return MinOrMaxPixelFunc<Comparator>(dfK, papoSources, nSources, pData, |
2741 | 0 | nXSize, nYSize, eSrcType, eBufType, |
2742 | 0 | nPixelSpace, nLineSpace, papszArgs); |
2743 | 0 | } |
2744 | | |
2745 | | static const char pszExprPixelFuncMetadata[] = |
2746 | | "<PixelFunctionArgumentsList>" |
2747 | | " <Argument name='expression' " |
2748 | | " description='Expression to be evaluated' " |
2749 | | " type='string'></Argument>" |
2750 | | " <Argument name='dialect' " |
2751 | | " description='Expression dialect' " |
2752 | | " type='string-select'" |
2753 | | " default='muparser'>" |
2754 | | " <Value>exprtk</Value>" |
2755 | | " <Value>muparser</Value>" |
2756 | | " </Argument>" |
2757 | | " <Argument type='builtin' value='source_names' />" |
2758 | | " <Argument type='builtin' value='xoff' />" |
2759 | | " <Argument type='builtin' value='yoff' />" |
2760 | | " <Argument type='builtin' value='geotransform' />" |
2761 | | "</PixelFunctionArgumentsList>"; |
2762 | | |
2763 | | static CPLErr ExprPixelFunc(void **papoSources, int nSources, void *pData, |
2764 | | int nXSize, int nYSize, GDALDataType eSrcType, |
2765 | | GDALDataType eBufType, int nPixelSpace, |
2766 | | int nLineSpace, CSLConstList papszArgs) |
2767 | 0 | { |
2768 | | /* ---- Init ---- */ |
2769 | |
|
2770 | 0 | if (GDALDataTypeIsComplex(eSrcType)) |
2771 | 0 | { |
2772 | 0 | CPLError(CE_Failure, CPLE_AppDefined, |
2773 | 0 | "expression cannot by applied to complex data types"); |
2774 | 0 | return CE_Failure; |
2775 | 0 | } |
2776 | | |
2777 | 0 | std::unique_ptr<gdal::MathExpression> poExpression; |
2778 | |
|
2779 | 0 | const char *pszExpression = CSLFetchNameValue(papszArgs, "expression"); |
2780 | |
|
2781 | 0 | const char *pszSourceNames = CSLFetchNameValue(papszArgs, "source_names"); |
2782 | 0 | const CPLStringList aosSourceNames( |
2783 | 0 | CSLTokenizeString2(pszSourceNames, "|", 0)); |
2784 | 0 | if (aosSourceNames.size() != nSources) |
2785 | 0 | { |
2786 | 0 | CPLError(CE_Failure, CPLE_AppDefined, |
2787 | 0 | "The source_names variable passed to ExprPixelFunc() has %d " |
2788 | 0 | "values, whereas %d were expected. An invalid variable name " |
2789 | 0 | "has likely been used", |
2790 | 0 | aosSourceNames.size(), nSources); |
2791 | 0 | return CE_Failure; |
2792 | 0 | } |
2793 | | |
2794 | 0 | std::vector<double> adfValuesForPixel(nSources); |
2795 | |
|
2796 | 0 | const char *pszDialect = CSLFetchNameValue(papszArgs, "dialect"); |
2797 | 0 | if (!pszDialect) |
2798 | 0 | { |
2799 | 0 | pszDialect = "muparser"; |
2800 | 0 | } |
2801 | |
|
2802 | 0 | poExpression = gdal::MathExpression::Create(pszExpression, pszDialect); |
2803 | | |
2804 | | // cppcheck-suppress knownConditionTrueFalse |
2805 | 0 | if (!poExpression) |
2806 | 0 | { |
2807 | 0 | return CE_Failure; |
2808 | 0 | } |
2809 | | |
2810 | 0 | int nXOff = 0; |
2811 | 0 | int nYOff = 0; |
2812 | 0 | std::array<double, 6> gt; |
2813 | 0 | double dfCenterX = 0; |
2814 | 0 | double dfCenterY = 0; |
2815 | |
|
2816 | 0 | bool includeCenterCoords = false; |
2817 | 0 | if (strstr(pszExpression, "_CENTER_X_") || |
2818 | 0 | strstr(pszExpression, "_CENTER_Y_")) |
2819 | 0 | { |
2820 | 0 | includeCenterCoords = true; |
2821 | |
|
2822 | 0 | const char *pszXOff = CSLFetchNameValue(papszArgs, "xoff"); |
2823 | 0 | nXOff = std::atoi(pszXOff); |
2824 | |
|
2825 | 0 | const char *pszYOff = CSLFetchNameValue(papszArgs, "yoff"); |
2826 | 0 | nYOff = std::atoi(pszYOff); |
2827 | |
|
2828 | 0 | const char *pszGT = CSLFetchNameValue(papszArgs, "geotransform"); |
2829 | 0 | if (pszGT == nullptr) |
2830 | 0 | { |
2831 | 0 | CPLError(CE_Failure, CPLE_AppDefined, |
2832 | 0 | "To use _CENTER_X_ or _CENTER_Y_ in an expression, " |
2833 | 0 | "VRTDataset must have a <GeoTransform> element."); |
2834 | 0 | return CE_Failure; |
2835 | 0 | } |
2836 | | |
2837 | 0 | CPLStringList aosGeoTransform( |
2838 | 0 | CSLTokenizeString2(pszGT, ",", CSLT_HONOURSTRINGS)); |
2839 | 0 | if (aosGeoTransform.size() != 6) |
2840 | 0 | { |
2841 | 0 | CPLError(CE_Failure, CPLE_AppDefined, |
2842 | 0 | "Invalid GeoTransform argument"); |
2843 | 0 | return CE_Failure; |
2844 | 0 | } |
2845 | 0 | for (int i = 0; i < 6; i++) |
2846 | 0 | { |
2847 | 0 | gt[i] = CPLAtof(aosGeoTransform[i]); |
2848 | 0 | } |
2849 | 0 | } |
2850 | | |
2851 | 0 | { |
2852 | 0 | int iSource = 0; |
2853 | 0 | for (const auto &osName : aosSourceNames) |
2854 | 0 | { |
2855 | 0 | poExpression->RegisterVariable(osName, |
2856 | 0 | &adfValuesForPixel[iSource++]); |
2857 | 0 | } |
2858 | 0 | } |
2859 | |
|
2860 | 0 | if (includeCenterCoords) |
2861 | 0 | { |
2862 | 0 | poExpression->RegisterVariable("_CENTER_X_", &dfCenterX); |
2863 | 0 | poExpression->RegisterVariable("_CENTER_Y_", &dfCenterY); |
2864 | 0 | } |
2865 | |
|
2866 | 0 | if (strstr(pszExpression, "BANDS")) |
2867 | 0 | { |
2868 | 0 | poExpression->RegisterVector("BANDS", &adfValuesForPixel); |
2869 | 0 | } |
2870 | |
|
2871 | 0 | std::unique_ptr<double, VSIFreeReleaser> padfResults( |
2872 | 0 | static_cast<double *>(VSI_MALLOC2_VERBOSE(nXSize, sizeof(double)))); |
2873 | 0 | if (!padfResults) |
2874 | 0 | return CE_Failure; |
2875 | | |
2876 | | /* ---- Set pixels ---- */ |
2877 | 0 | size_t ii = 0; |
2878 | 0 | for (int iLine = 0; iLine < nYSize; ++iLine) |
2879 | 0 | { |
2880 | 0 | for (int iCol = 0; iCol < nXSize; ++iCol, ++ii) |
2881 | 0 | { |
2882 | 0 | for (int iSrc = 0; iSrc < nSources; iSrc++) |
2883 | 0 | { |
2884 | | // cppcheck-suppress unreadVariable |
2885 | 0 | adfValuesForPixel[iSrc] = |
2886 | 0 | GetSrcVal(papoSources[iSrc], eSrcType, ii); |
2887 | 0 | } |
2888 | |
|
2889 | 0 | if (includeCenterCoords) |
2890 | 0 | { |
2891 | | // Add 0.5 to pixel / line to move from pixel corner to cell center |
2892 | 0 | GDALApplyGeoTransform(gt.data(), |
2893 | 0 | static_cast<double>(iCol + nXOff) + 0.5, |
2894 | 0 | static_cast<double>(iLine + nYOff) + 0.5, |
2895 | 0 | &dfCenterX, &dfCenterY); |
2896 | 0 | } |
2897 | |
|
2898 | 0 | if (auto eErr = poExpression->Evaluate(); eErr != CE_None) |
2899 | 0 | { |
2900 | 0 | return CE_Failure; |
2901 | 0 | } |
2902 | 0 | else |
2903 | 0 | { |
2904 | 0 | padfResults.get()[iCol] = poExpression->Results()[0]; |
2905 | 0 | } |
2906 | 0 | } |
2907 | | |
2908 | 0 | GDALCopyWords(padfResults.get(), GDT_Float64, sizeof(double), |
2909 | 0 | static_cast<GByte *>(pData) + |
2910 | 0 | static_cast<GSpacing>(nLineSpace) * iLine, |
2911 | 0 | eBufType, nPixelSpace, nXSize); |
2912 | 0 | } |
2913 | | |
2914 | | /* ---- Return success ---- */ |
2915 | 0 | return CE_None; |
2916 | 0 | } // ExprPixelFunc |
2917 | | |
2918 | | static const char pszReclassifyPixelFuncMetadata[] = |
2919 | | "<PixelFunctionArgumentsList>" |
2920 | | " <Argument name='mapping' " |
2921 | | " description='Lookup table for mapping, in format " |
2922 | | "from=to,from=to' " |
2923 | | " type='string'></Argument>" |
2924 | | " <Argument type='builtin' value='NoData' optional='true' />" |
2925 | | "</PixelFunctionArgumentsList>"; |
2926 | | |
2927 | | static CPLErr ReclassifyPixelFunc(void **papoSources, int nSources, void *pData, |
2928 | | int nXSize, int nYSize, GDALDataType eSrcType, |
2929 | | GDALDataType eBufType, int nPixelSpace, |
2930 | | int nLineSpace, CSLConstList papszArgs) |
2931 | 0 | { |
2932 | 0 | if (GDALDataTypeIsComplex(eSrcType)) |
2933 | 0 | { |
2934 | 0 | CPLError(CE_Failure, CPLE_AppDefined, |
2935 | 0 | "reclassify cannot by applied to complex data types"); |
2936 | 0 | return CE_Failure; |
2937 | 0 | } |
2938 | | |
2939 | 0 | if (nSources != 1) |
2940 | 0 | { |
2941 | 0 | CPLError(CE_Failure, CPLE_AppDefined, |
2942 | 0 | "reclassify only be applied to a single source at a time"); |
2943 | 0 | return CE_Failure; |
2944 | 0 | } |
2945 | 0 | std::optional<double> noDataValue{}; |
2946 | |
|
2947 | 0 | const char *pszNoData = CSLFetchNameValue(papszArgs, "NoData"); |
2948 | 0 | if (pszNoData != nullptr) |
2949 | 0 | { |
2950 | 0 | noDataValue = CPLAtof(pszNoData); |
2951 | 0 | } |
2952 | |
|
2953 | 0 | const char *pszMappings = CSLFetchNameValue(papszArgs, "mapping"); |
2954 | 0 | if (pszMappings == nullptr) |
2955 | 0 | { |
2956 | 0 | CPLError(CE_Failure, CPLE_AppDefined, |
2957 | 0 | "reclassify must be called with 'mapping' argument"); |
2958 | 0 | return CE_Failure; |
2959 | 0 | } |
2960 | | |
2961 | 0 | gdal::Reclassifier oReclassifier; |
2962 | 0 | if (auto eErr = oReclassifier.Init(pszMappings, noDataValue, eBufType); |
2963 | 0 | eErr != CE_None) |
2964 | 0 | { |
2965 | 0 | return eErr; |
2966 | 0 | } |
2967 | | |
2968 | 0 | std::unique_ptr<double, VSIFreeReleaser> padfResults( |
2969 | 0 | static_cast<double *>(VSI_MALLOC2_VERBOSE(nXSize, sizeof(double)))); |
2970 | 0 | if (!padfResults) |
2971 | 0 | return CE_Failure; |
2972 | | |
2973 | 0 | size_t ii = 0; |
2974 | 0 | bool bSuccess = false; |
2975 | 0 | for (int iLine = 0; iLine < nYSize; ++iLine) |
2976 | 0 | { |
2977 | 0 | for (int iCol = 0; iCol < nXSize; ++iCol, ++ii) |
2978 | 0 | { |
2979 | 0 | double srcVal = GetSrcVal(papoSources[0], eSrcType, ii); |
2980 | 0 | padfResults.get()[iCol] = |
2981 | 0 | oReclassifier.Reclassify(srcVal, bSuccess); |
2982 | 0 | if (!bSuccess) |
2983 | 0 | { |
2984 | 0 | CPLError(CE_Failure, CPLE_AppDefined, |
2985 | 0 | "Encountered value %g with no specified mapping", |
2986 | 0 | srcVal); |
2987 | 0 | return CE_Failure; |
2988 | 0 | } |
2989 | 0 | } |
2990 | | |
2991 | 0 | GDALCopyWords(padfResults.get(), GDT_Float64, sizeof(double), |
2992 | 0 | static_cast<GByte *>(pData) + |
2993 | 0 | static_cast<GSpacing>(nLineSpace) * iLine, |
2994 | 0 | eBufType, nPixelSpace, nXSize); |
2995 | 0 | } |
2996 | | |
2997 | 0 | return CE_None; |
2998 | 0 | } // ReclassifyPixelFunc |
2999 | | |
3000 | | struct MeanKernel |
3001 | | { |
3002 | | static constexpr const char *pszName = "mean"; |
3003 | | |
3004 | | double dfMean = 0; |
3005 | | int nValidSources = 0; |
3006 | | |
3007 | | void Reset() |
3008 | 0 | { |
3009 | 0 | dfMean = 0; |
3010 | 0 | nValidSources = 0; |
3011 | 0 | } |
3012 | | |
3013 | | static CPLErr ProcessArguments(CSLConstList) |
3014 | 0 | { |
3015 | 0 | return CE_None; |
3016 | 0 | } |
3017 | | |
3018 | | void ProcessPixel(double dfVal) |
3019 | 0 | { |
3020 | 0 | ++nValidSources; |
3021 | |
|
3022 | 0 | if (CPL_UNLIKELY(std::isinf(dfVal))) |
3023 | 0 | { |
3024 | 0 | if (nValidSources == 1) |
3025 | 0 | { |
3026 | 0 | dfMean = dfVal; |
3027 | 0 | } |
3028 | 0 | else if (dfVal == -dfMean) |
3029 | 0 | { |
3030 | 0 | dfMean = std::numeric_limits<double>::quiet_NaN(); |
3031 | 0 | } |
3032 | 0 | } |
3033 | 0 | else if (CPL_UNLIKELY(std::isinf(dfMean))) |
3034 | 0 | { |
3035 | 0 | if (!std::isfinite(dfVal)) |
3036 | 0 | { |
3037 | 0 | dfMean = std::numeric_limits<double>::quiet_NaN(); |
3038 | 0 | } |
3039 | 0 | } |
3040 | 0 | else |
3041 | 0 | { |
3042 | 0 | const double delta = dfVal - dfMean; |
3043 | 0 | if (CPL_UNLIKELY(std::isinf(delta))) |
3044 | 0 | dfMean += dfVal / nValidSources - dfMean / nValidSources; |
3045 | 0 | else |
3046 | 0 | dfMean += delta / nValidSources; |
3047 | 0 | } |
3048 | 0 | } |
3049 | | |
3050 | | bool HasValue() const |
3051 | 0 | { |
3052 | 0 | return nValidSources > 0; |
3053 | 0 | } |
3054 | | |
3055 | | double GetValue() const |
3056 | 0 | { |
3057 | 0 | return dfMean; |
3058 | 0 | } |
3059 | | }; |
3060 | | |
3061 | | struct GeoMeanKernel |
3062 | | { |
3063 | | static constexpr const char *pszName = "geometric_mean"; |
3064 | | |
3065 | | double dfProduct = 1; |
3066 | | int nValidSources = 0; |
3067 | | |
3068 | | void Reset() |
3069 | 0 | { |
3070 | 0 | dfProduct = 1; |
3071 | 0 | nValidSources = 0; |
3072 | 0 | } |
3073 | | |
3074 | | static CPLErr ProcessArguments(CSLConstList) |
3075 | 0 | { |
3076 | 0 | return CE_None; |
3077 | 0 | } |
3078 | | |
3079 | | void ProcessPixel(double dfVal) |
3080 | 0 | { |
3081 | 0 | dfProduct *= dfVal; |
3082 | 0 | nValidSources++; |
3083 | 0 | } |
3084 | | |
3085 | | bool HasValue() const |
3086 | 0 | { |
3087 | 0 | return nValidSources > 0; |
3088 | 0 | } |
3089 | | |
3090 | | double GetValue() const |
3091 | 0 | { |
3092 | 0 | return std::pow(dfProduct, 1.0 / nValidSources); |
3093 | 0 | } |
3094 | | }; |
3095 | | |
3096 | | struct HarmonicMeanKernel |
3097 | | { |
3098 | | static constexpr const char *pszName = "harmonic_mean"; |
3099 | | |
3100 | | double dfDenom = 0; |
3101 | | int nValidSources = 0; |
3102 | | bool bValueIsZero = false; |
3103 | | bool bPropagateZero = false; |
3104 | | |
3105 | | void Reset() |
3106 | 0 | { |
3107 | 0 | dfDenom = 0; |
3108 | 0 | nValidSources = 0; |
3109 | 0 | bValueIsZero = false; |
3110 | 0 | } |
3111 | | |
3112 | | void ProcessPixel(double dfVal) |
3113 | 0 | { |
3114 | 0 | if (dfVal == 0) |
3115 | 0 | { |
3116 | 0 | bValueIsZero = true; |
3117 | 0 | } |
3118 | 0 | else |
3119 | 0 | { |
3120 | 0 | dfDenom += 1 / dfVal; |
3121 | 0 | } |
3122 | 0 | nValidSources++; |
3123 | 0 | } |
3124 | | |
3125 | | CPLErr ProcessArguments(CSLConstList papszArgs) |
3126 | 0 | { |
3127 | 0 | bPropagateZero = |
3128 | 0 | CPLTestBool(CSLFetchNameValueDef(papszArgs, "propagateZero", "0")); |
3129 | 0 | return CE_None; |
3130 | 0 | } |
3131 | | |
3132 | | bool HasValue() const |
3133 | 0 | { |
3134 | 0 | return dfDenom > 0 && (bPropagateZero || !bValueIsZero); |
3135 | 0 | } |
3136 | | |
3137 | | double GetValue() const |
3138 | 0 | { |
3139 | 0 | if (bPropagateZero && bValueIsZero) |
3140 | 0 | { |
3141 | 0 | return 0; |
3142 | 0 | } |
3143 | 0 | return static_cast<double>(nValidSources) / dfDenom; |
3144 | 0 | } |
3145 | | }; |
3146 | | |
3147 | | struct MedianKernel |
3148 | | { |
3149 | | static constexpr const char *pszName = "median"; |
3150 | | |
3151 | | mutable std::vector<double> values{}; |
3152 | | |
3153 | | void Reset() |
3154 | 0 | { |
3155 | 0 | values.clear(); |
3156 | 0 | } |
3157 | | |
3158 | | static CPLErr ProcessArguments(CSLConstList) |
3159 | 0 | { |
3160 | 0 | return CE_None; |
3161 | 0 | } |
3162 | | |
3163 | | void ProcessPixel(double dfVal) |
3164 | 0 | { |
3165 | 0 | if (!std::isnan(dfVal)) |
3166 | 0 | { |
3167 | 0 | values.push_back(dfVal); |
3168 | 0 | } |
3169 | 0 | } |
3170 | | |
3171 | | bool HasValue() const |
3172 | 0 | { |
3173 | 0 | return !values.empty(); |
3174 | 0 | } |
3175 | | |
3176 | | double GetValue() const |
3177 | 0 | { |
3178 | 0 | std::sort(values.begin(), values.end()); |
3179 | 0 | if (values.size() % 2 == 0) |
3180 | 0 | { |
3181 | 0 | return 0.5 * |
3182 | 0 | (values[values.size() / 2 - 1] + values[values.size() / 2]); |
3183 | 0 | } |
3184 | | |
3185 | 0 | return values[values.size() / 2]; |
3186 | 0 | } |
3187 | | }; |
3188 | | |
3189 | | struct ModeKernel |
3190 | | { |
3191 | | static constexpr const char *pszName = "mode"; |
3192 | | |
3193 | | std::map<double, size_t> counts{}; |
3194 | | std::size_t nanCount{0}; |
3195 | | double dfMax = std::numeric_limits<double>::quiet_NaN(); |
3196 | | decltype(counts.begin()) oMax = counts.end(); |
3197 | | |
3198 | | void Reset() |
3199 | 0 | { |
3200 | 0 | nanCount = 0; |
3201 | 0 | counts.clear(); |
3202 | 0 | oMax = counts.end(); |
3203 | 0 | } |
3204 | | |
3205 | | static CPLErr ProcessArguments(CSLConstList) |
3206 | 0 | { |
3207 | 0 | return CE_None; |
3208 | 0 | } |
3209 | | |
3210 | | void ProcessPixel(double dfVal) |
3211 | 0 | { |
3212 | 0 | if (std::isnan(dfVal)) |
3213 | 0 | { |
3214 | 0 | nanCount += 1; |
3215 | 0 | return; |
3216 | 0 | } |
3217 | | |
3218 | | // if dfVal is NaN, try_emplace will return an entry for a different key! |
3219 | 0 | auto [it, inserted] = counts.try_emplace(dfVal, 0); |
3220 | |
|
3221 | 0 | it->second += 1; |
3222 | |
|
3223 | 0 | if (oMax == counts.end() || it->second > oMax->second) |
3224 | 0 | { |
3225 | 0 | oMax = it; |
3226 | 0 | } |
3227 | 0 | } |
3228 | | |
3229 | | bool HasValue() const |
3230 | 0 | { |
3231 | 0 | return nanCount > 0 || oMax != counts.end(); |
3232 | 0 | } |
3233 | | |
3234 | | double GetValue() const |
3235 | 0 | { |
3236 | 0 | double ret = std::numeric_limits<double>::quiet_NaN(); |
3237 | 0 | if (oMax != counts.end()) |
3238 | 0 | { |
3239 | 0 | const size_t nCount = oMax->second; |
3240 | 0 | if (nCount > nanCount) |
3241 | 0 | ret = oMax->first; |
3242 | 0 | } |
3243 | 0 | return ret; |
3244 | 0 | } |
3245 | | }; |
3246 | | |
3247 | | static const char pszBasicPixelFuncMetadata[] = |
3248 | | "<PixelFunctionArgumentsList>" |
3249 | | " <Argument type='builtin' value='NoData' optional='true' />" |
3250 | | " <Argument name='propagateNoData' description='Whether the output value " |
3251 | | "should be NoData as as soon as one source is NoData' type='boolean' " |
3252 | | "default='false' />" |
3253 | | "</PixelFunctionArgumentsList>"; |
3254 | | |
3255 | | #if defined(USE_SSE2) && !defined(USE_NEON_OPTIMIZATIONS) |
3256 | | inline __m128i packus_epi32(__m128i low, __m128i high) |
3257 | 0 | { |
3258 | | #if __SSE4_1__ |
3259 | | return _mm_packus_epi32(low, high); // Pack uint32 to uint16 |
3260 | | #else |
3261 | 0 | low = _mm_add_epi32(low, _mm_set1_epi32(-32768)); |
3262 | 0 | high = _mm_add_epi32(high, _mm_set1_epi32(-32768)); |
3263 | 0 | return _mm_sub_epi16(_mm_packs_epi32(low, high), _mm_set1_epi16(-32768)); |
3264 | 0 | #endif |
3265 | 0 | } |
3266 | | #endif |
3267 | | |
3268 | | #ifdef USE_SSE2 |
3269 | | |
3270 | | template <class T, class SSEWrapper> |
3271 | | static void OptimizedMeanFloatSSE2(const void *const *papoSources, int nSources, |
3272 | | void *pData, int nXSize, int nYSize, |
3273 | | int nLineSpace) |
3274 | 0 | { |
3275 | 0 | assert(nSources >= 1); |
3276 | 0 | constexpr int VALUES_PER_REG = |
3277 | 0 | static_cast<int>(sizeof(typename SSEWrapper::Vec) / sizeof(T)); |
3278 | 0 | const T invSources = static_cast<T>(1.0) / static_cast<T>(nSources); |
3279 | 0 | const auto invSourcesSSE = SSEWrapper::Set1(invSources); |
3280 | 0 | const auto signMaskSSE = SSEWrapper::Set1(static_cast<T>(-0.0)); |
3281 | 0 | const auto infSSE = SSEWrapper::Set1(std::numeric_limits<T>::infinity()); |
3282 | 0 | for (int iLine = 0; iLine < nYSize; ++iLine) |
3283 | 0 | { |
3284 | 0 | T *CPL_RESTRICT pDest = |
3285 | 0 | reinterpret_cast<T *>(static_cast<GByte *>(pData) + |
3286 | 0 | static_cast<GSpacing>(nLineSpace) * iLine); |
3287 | 0 | const size_t iOffsetLine = static_cast<size_t>(iLine) * nXSize; |
3288 | 0 | int iCol = 0; |
3289 | 0 | for (; iCol < nXSize - (2 * VALUES_PER_REG - 1); |
3290 | 0 | iCol += 2 * VALUES_PER_REG) |
3291 | 0 | { |
3292 | 0 | auto reg0 = SSEWrapper::LoadU( |
3293 | 0 | static_cast<const T * CPL_RESTRICT>(papoSources[0]) + |
3294 | 0 | iOffsetLine + iCol); |
3295 | 0 | auto reg1 = SSEWrapper::LoadU( |
3296 | 0 | static_cast<const T * CPL_RESTRICT>(papoSources[0]) + |
3297 | 0 | iOffsetLine + iCol + VALUES_PER_REG); |
3298 | 0 | for (int iSrc = 1; iSrc < nSources; ++iSrc) |
3299 | 0 | { |
3300 | 0 | const auto inputVal0 = SSEWrapper::LoadU( |
3301 | 0 | static_cast<const T * CPL_RESTRICT>(papoSources[iSrc]) + |
3302 | 0 | iOffsetLine + iCol); |
3303 | 0 | const auto inputVal1 = SSEWrapper::LoadU( |
3304 | 0 | static_cast<const T * CPL_RESTRICT>(papoSources[iSrc]) + |
3305 | 0 | iOffsetLine + iCol + VALUES_PER_REG); |
3306 | 0 | reg0 = SSEWrapper::Add(reg0, inputVal0); |
3307 | 0 | reg1 = SSEWrapper::Add(reg1, inputVal1); |
3308 | 0 | } |
3309 | 0 | reg0 = SSEWrapper::Mul(reg0, invSourcesSSE); |
3310 | 0 | reg1 = SSEWrapper::Mul(reg1, invSourcesSSE); |
3311 | | |
3312 | | // Detect infinity that could happen when summing huge |
3313 | | // values |
3314 | 0 | if (SSEWrapper::MoveMask(SSEWrapper::Or( |
3315 | 0 | SSEWrapper::CmpEq(SSEWrapper::AndNot(signMaskSSE, reg0), |
3316 | 0 | infSSE), |
3317 | 0 | SSEWrapper::CmpEq(SSEWrapper::AndNot(signMaskSSE, reg1), |
3318 | 0 | infSSE)))) |
3319 | 0 | { |
3320 | 0 | break; |
3321 | 0 | } |
3322 | | |
3323 | 0 | SSEWrapper::StoreU(pDest + iCol, reg0); |
3324 | 0 | SSEWrapper::StoreU(pDest + iCol + VALUES_PER_REG, reg1); |
3325 | 0 | } |
3326 | | |
3327 | | // Use numerically stable mean computation |
3328 | 0 | for (; iCol < nXSize; ++iCol) |
3329 | 0 | { |
3330 | 0 | T mean = static_cast<const T * CPL_RESTRICT>( |
3331 | 0 | papoSources[0])[iOffsetLine + iCol]; |
3332 | 0 | if (nSources >= 2) |
3333 | 0 | { |
3334 | 0 | T new_val = static_cast<const T * CPL_RESTRICT>( |
3335 | 0 | papoSources[1])[iOffsetLine + iCol]; |
3336 | 0 | if (CPL_UNLIKELY(std::isinf(new_val))) |
3337 | 0 | { |
3338 | 0 | if (new_val == -mean) |
3339 | 0 | { |
3340 | 0 | pDest[iCol] = std::numeric_limits<T>::quiet_NaN(); |
3341 | 0 | continue; |
3342 | 0 | } |
3343 | 0 | } |
3344 | 0 | else if (CPL_UNLIKELY(std::isinf(mean))) |
3345 | 0 | { |
3346 | 0 | if (!std::isfinite(new_val)) |
3347 | 0 | { |
3348 | 0 | pDest[iCol] = std::numeric_limits<T>::quiet_NaN(); |
3349 | 0 | continue; |
3350 | 0 | } |
3351 | 0 | } |
3352 | 0 | else |
3353 | 0 | { |
3354 | 0 | const T delta = new_val - mean; |
3355 | 0 | if (CPL_UNLIKELY(std::isinf(delta))) |
3356 | 0 | mean += new_val * static_cast<T>(0.5) - |
3357 | 0 | mean * static_cast<T>(0.5); |
3358 | 0 | else |
3359 | 0 | mean += delta * static_cast<T>(0.5); |
3360 | 0 | } |
3361 | | |
3362 | 0 | for (int iSrc = 2; iSrc < nSources; ++iSrc) |
3363 | 0 | { |
3364 | 0 | new_val = static_cast<const T * CPL_RESTRICT>( |
3365 | 0 | papoSources[iSrc])[iOffsetLine + iCol]; |
3366 | 0 | if (CPL_UNLIKELY(std::isinf(new_val))) |
3367 | 0 | { |
3368 | 0 | if (new_val == -mean) |
3369 | 0 | { |
3370 | 0 | mean = std::numeric_limits<T>::quiet_NaN(); |
3371 | 0 | break; |
3372 | 0 | } |
3373 | 0 | } |
3374 | 0 | else if (CPL_UNLIKELY(std::isinf(mean))) |
3375 | 0 | { |
3376 | 0 | if (!std::isfinite(new_val)) |
3377 | 0 | { |
3378 | 0 | mean = std::numeric_limits<T>::quiet_NaN(); |
3379 | 0 | break; |
3380 | 0 | } |
3381 | 0 | } |
3382 | 0 | else |
3383 | 0 | { |
3384 | 0 | const T delta = new_val - mean; |
3385 | 0 | if (CPL_UNLIKELY(std::isinf(delta))) |
3386 | 0 | mean += new_val / static_cast<T>(iSrc + 1) - |
3387 | 0 | mean / static_cast<T>(iSrc + 1); |
3388 | 0 | else |
3389 | 0 | mean += delta / static_cast<T>(iSrc + 1); |
3390 | 0 | } |
3391 | 0 | } |
3392 | 0 | } |
3393 | 0 | pDest[iCol] = mean; |
3394 | 0 | } |
3395 | 0 | } |
3396 | 0 | } Unexecuted instantiation: pixelfunctions.cpp:void OptimizedMeanFloatSSE2<float, (anonymous namespace)::SSEWrapperFloat>(void const* const*, int, void*, int, int, int) Unexecuted instantiation: pixelfunctions.cpp:void OptimizedMeanFloatSSE2<double, (anonymous namespace)::SSEWrapperDouble>(void const* const*, int, void*, int, int, int) |
3397 | | |
3398 | | // clang-format off |
3399 | | namespace |
3400 | | { |
3401 | | #ifdef __AVX2__ |
3402 | | struct SSEWrapperFloat |
3403 | | { |
3404 | | typedef __m256 Vec; |
3405 | | |
3406 | | static inline Vec Set1(float x) { return _mm256_set1_ps(x); } |
3407 | | static inline Vec LoadU(const float *x) { return _mm256_loadu_ps(x); } |
3408 | | static inline void StoreU(float *x, Vec y) { _mm256_storeu_ps(x, y); } |
3409 | | static inline Vec Add(Vec x, Vec y) { return _mm256_add_ps(x, y); } |
3410 | | static inline Vec Mul(Vec x, Vec y) { return _mm256_mul_ps(x, y); } |
3411 | | static inline Vec Or(Vec x, Vec y) { return _mm256_or_ps(x, y); } |
3412 | | static inline Vec AndNot(Vec x, Vec y) { return _mm256_andnot_ps(x, y); } |
3413 | | static inline Vec CmpEq(Vec x, Vec y) { return _mm256_cmp_ps(x, y, _CMP_EQ_OQ); } |
3414 | | static inline int MoveMask(Vec x) { return _mm256_movemask_ps(x); } |
3415 | | }; |
3416 | | |
3417 | | struct SSEWrapperDouble |
3418 | | { |
3419 | | typedef __m256d Vec; |
3420 | | |
3421 | | static inline Vec Set1(double x) { return _mm256_set1_pd(x); } |
3422 | | static inline Vec LoadU(const double *x) { return _mm256_loadu_pd(x); } |
3423 | | static inline void StoreU(double *x, Vec y) { _mm256_storeu_pd(x, y); } |
3424 | | static inline Vec Add(Vec x, Vec y) { return _mm256_add_pd(x, y); } |
3425 | | static inline Vec Mul(Vec x, Vec y) { return _mm256_mul_pd(x, y); } |
3426 | | static inline Vec Or(Vec x, Vec y) { return _mm256_or_pd(x, y); } |
3427 | | static inline Vec AndNot(Vec x, Vec y) { return _mm256_andnot_pd(x, y); } |
3428 | | static inline Vec CmpEq(Vec x, Vec y) { return _mm256_cmp_pd(x, y, _CMP_EQ_OQ); } |
3429 | | static inline int MoveMask(Vec x) { return _mm256_movemask_pd(x); } |
3430 | | }; |
3431 | | |
3432 | | #else |
3433 | | |
3434 | | struct SSEWrapperFloat |
3435 | | { |
3436 | | typedef __m128 Vec; |
3437 | | |
3438 | 0 | static inline Vec Set1(float x) { return _mm_set1_ps(x); } |
3439 | 0 | static inline Vec LoadU(const float *x) { return _mm_loadu_ps(x); } |
3440 | 0 | static inline void StoreU(float *x, Vec y) { _mm_storeu_ps(x, y); } |
3441 | 0 | static inline Vec Add(Vec x, Vec y) { return _mm_add_ps(x, y); } |
3442 | 0 | static inline Vec Mul(Vec x, Vec y) { return _mm_mul_ps(x, y); } |
3443 | 0 | static inline Vec Or(Vec x, Vec y) { return _mm_or_ps(x, y); } |
3444 | 0 | static inline Vec AndNot(Vec x, Vec y) { return _mm_andnot_ps(x, y); } |
3445 | 0 | static inline Vec CmpEq(Vec x, Vec y) { return _mm_cmpeq_ps(x, y); } |
3446 | 0 | static inline int MoveMask(Vec x) { return _mm_movemask_ps(x); } |
3447 | | }; |
3448 | | |
3449 | | struct SSEWrapperDouble |
3450 | | { |
3451 | | typedef __m128d Vec; |
3452 | | |
3453 | 0 | static inline Vec Set1(double x) { return _mm_set1_pd(x); } |
3454 | 0 | static inline Vec LoadU(const double *x) { return _mm_loadu_pd(x); } |
3455 | 0 | static inline void StoreU(double *x, Vec y) { _mm_storeu_pd(x, y); } |
3456 | 0 | static inline Vec Add(Vec x, Vec y) { return _mm_add_pd(x, y); } |
3457 | 0 | static inline Vec Mul(Vec x, Vec y) { return _mm_mul_pd(x, y); } |
3458 | 0 | static inline Vec Or(Vec x, Vec y) { return _mm_or_pd(x, y); } |
3459 | 0 | static inline Vec AndNot(Vec x, Vec y) { return _mm_andnot_pd(x, y); } |
3460 | 0 | static inline Vec CmpEq(Vec x, Vec y) { return _mm_cmpeq_pd(x, y); } |
3461 | 0 | static inline int MoveMask(Vec x) { return _mm_movemask_pd(x); } |
3462 | | }; |
3463 | | #endif |
3464 | | } // namespace |
3465 | | |
3466 | | // clang-format on |
3467 | | |
3468 | | #endif // USE_SSE2 |
3469 | | |
3470 | | template <typename Kernel> |
3471 | | static CPLErr BasicPixelFunc(void **papoSources, int nSources, void *pData, |
3472 | | int nXSize, int nYSize, GDALDataType eSrcType, |
3473 | | GDALDataType eBufType, int nPixelSpace, |
3474 | | int nLineSpace, CSLConstList papszArgs) |
3475 | 0 | { |
3476 | | /* ---- Init ---- */ |
3477 | 0 | Kernel oKernel; |
3478 | |
|
3479 | 0 | if (GDALDataTypeIsComplex(eSrcType)) |
3480 | 0 | { |
3481 | 0 | CPLError(CE_Failure, CPLE_AppDefined, |
3482 | 0 | "Complex data types not supported by %s", oKernel.pszName); |
3483 | 0 | return CE_Failure; |
3484 | 0 | } |
3485 | | |
3486 | 0 | double dfNoData{0}; |
3487 | 0 | const bool bHasNoData = CSLFindName(papszArgs, "NoData") != -1; |
3488 | 0 | if (bHasNoData && FetchDoubleArg(papszArgs, "NoData", &dfNoData) != CE_None) |
3489 | 0 | return CE_Failure; |
3490 | | |
3491 | 0 | const bool bPropagateNoData = CPLTestBool( |
3492 | 0 | CSLFetchNameValueDef(papszArgs, "propagateNoData", "false")); |
3493 | |
|
3494 | 0 | if (oKernel.ProcessArguments(papszArgs) == CE_Failure) |
3495 | 0 | { |
3496 | 0 | return CE_Failure; |
3497 | 0 | } |
3498 | | |
3499 | 0 | #if defined(USE_SSE2) && !defined(USE_NEON_OPTIMIZATIONS) |
3500 | | if constexpr (std::is_same_v<Kernel, MeanKernel>) |
3501 | 0 | { |
3502 | 0 | if (!bHasNoData && eSrcType == GDT_Byte && eBufType == GDT_Byte && |
3503 | 0 | nPixelSpace == 1 && |
3504 | | // We use signed int16 to accumulate |
3505 | 0 | nSources <= std::numeric_limits<int16_t>::max() / |
3506 | 0 | std::numeric_limits<uint8_t>::max()) |
3507 | 0 | { |
3508 | 0 | using T = uint8_t; |
3509 | 0 | constexpr int VALUES_PER_REG = 16; |
3510 | 0 | if (nSources == 2) |
3511 | 0 | { |
3512 | 0 | for (int iLine = 0; iLine < nYSize; ++iLine) |
3513 | 0 | { |
3514 | 0 | T *CPL_RESTRICT pDest = reinterpret_cast<T *>( |
3515 | 0 | static_cast<GByte *>(pData) + |
3516 | 0 | static_cast<GSpacing>(nLineSpace) * iLine); |
3517 | 0 | const size_t iOffsetLine = |
3518 | 0 | static_cast<size_t>(iLine) * nXSize; |
3519 | 0 | int iCol = 0; |
3520 | 0 | for (; iCol < nXSize - (VALUES_PER_REG - 1); |
3521 | 0 | iCol += VALUES_PER_REG) |
3522 | 0 | { |
3523 | 0 | const __m128i inputVal0 = |
3524 | 0 | _mm_loadu_si128(reinterpret_cast<const __m128i *>( |
3525 | 0 | static_cast<const T * CPL_RESTRICT>( |
3526 | 0 | papoSources[0]) + |
3527 | 0 | iOffsetLine + iCol)); |
3528 | 0 | const __m128i inputVal1 = |
3529 | 0 | _mm_loadu_si128(reinterpret_cast<const __m128i *>( |
3530 | 0 | static_cast<const T * CPL_RESTRICT>( |
3531 | 0 | papoSources[1]) + |
3532 | 0 | iOffsetLine + iCol)); |
3533 | 0 | _mm_storeu_si128( |
3534 | 0 | reinterpret_cast<__m128i *>(pDest + iCol), |
3535 | 0 | _mm_avg_epu8(inputVal0, inputVal1)); |
3536 | 0 | } |
3537 | 0 | for (; iCol < nXSize; ++iCol) |
3538 | 0 | { |
3539 | 0 | uint32_t acc = 1 + |
3540 | 0 | static_cast<const T * CPL_RESTRICT>( |
3541 | 0 | papoSources[0])[iOffsetLine + iCol] + |
3542 | 0 | static_cast<const T * CPL_RESTRICT>( |
3543 | 0 | papoSources[1])[iOffsetLine + iCol]; |
3544 | 0 | pDest[iCol] = static_cast<T>(acc / 2); |
3545 | 0 | } |
3546 | 0 | } |
3547 | 0 | } |
3548 | 0 | else |
3549 | 0 | { |
3550 | 0 | libdivide::divider<uint16_t> fast_d( |
3551 | 0 | static_cast<uint16_t>(nSources)); |
3552 | 0 | const auto halfConstant = |
3553 | 0 | _mm_set1_epi16(static_cast<int16_t>(nSources / 2)); |
3554 | 0 | for (int iLine = 0; iLine < nYSize; ++iLine) |
3555 | 0 | { |
3556 | 0 | T *CPL_RESTRICT pDest = |
3557 | 0 | static_cast<GByte *>(pData) + |
3558 | 0 | static_cast<GSpacing>(nLineSpace) * iLine; |
3559 | 0 | const size_t iOffsetLine = |
3560 | 0 | static_cast<size_t>(iLine) * nXSize; |
3561 | 0 | int iCol = 0; |
3562 | 0 | for (; iCol < nXSize - (VALUES_PER_REG - 1); |
3563 | 0 | iCol += VALUES_PER_REG) |
3564 | 0 | { |
3565 | 0 | __m128i reg0 = halfConstant; |
3566 | 0 | __m128i reg1 = halfConstant; |
3567 | 0 | for (int iSrc = 0; iSrc < nSources; ++iSrc) |
3568 | 0 | { |
3569 | 0 | const __m128i inputVal = _mm_loadu_si128( |
3570 | 0 | reinterpret_cast<const __m128i *>( |
3571 | 0 | static_cast<const T * CPL_RESTRICT>( |
3572 | 0 | papoSources[iSrc]) + |
3573 | 0 | iOffsetLine + iCol)); |
3574 | 0 | reg0 = _mm_add_epi16( |
3575 | 0 | reg0, _mm_unpacklo_epi8(inputVal, |
3576 | 0 | _mm_setzero_si128())); |
3577 | 0 | reg1 = _mm_add_epi16( |
3578 | 0 | reg1, _mm_unpackhi_epi8(inputVal, |
3579 | 0 | _mm_setzero_si128())); |
3580 | 0 | } |
3581 | 0 | reg0 /= fast_d; |
3582 | 0 | reg1 /= fast_d; |
3583 | 0 | _mm_storeu_si128( |
3584 | 0 | reinterpret_cast<__m128i *>(pDest + iCol), |
3585 | 0 | _mm_packus_epi16(reg0, reg1)); |
3586 | 0 | } |
3587 | 0 | for (; iCol < nXSize; ++iCol) |
3588 | 0 | { |
3589 | 0 | uint32_t acc = nSources / 2; |
3590 | 0 | for (int iSrc = 0; iSrc < nSources; ++iSrc) |
3591 | 0 | { |
3592 | 0 | acc += static_cast<const T * CPL_RESTRICT>( |
3593 | 0 | papoSources[iSrc])[iOffsetLine + iCol]; |
3594 | 0 | } |
3595 | 0 | pDest[iCol] = static_cast<T>(acc / nSources); |
3596 | 0 | } |
3597 | 0 | } |
3598 | 0 | } |
3599 | 0 | return CE_None; |
3600 | 0 | } |
3601 | | |
3602 | 0 | if (!bHasNoData && eSrcType == GDT_Byte && eBufType == GDT_Byte && |
3603 | 0 | nPixelSpace == 1 && |
3604 | | // We use signed int32 to accumulate |
3605 | 0 | nSources <= std::numeric_limits<int32_t>::max() / |
3606 | 0 | std::numeric_limits<uint8_t>::max()) |
3607 | 0 | { |
3608 | 0 | using T = uint8_t; |
3609 | 0 | constexpr int VALUES_PER_REG = 16; |
3610 | 0 | libdivide::divider<uint32_t> fast_d(nSources); |
3611 | 0 | const auto halfConstant = _mm_set1_epi32(nSources / 2); |
3612 | 0 | for (int iLine = 0; iLine < nYSize; ++iLine) |
3613 | 0 | { |
3614 | 0 | T *CPL_RESTRICT pDest = |
3615 | 0 | static_cast<GByte *>(pData) + |
3616 | 0 | static_cast<GSpacing>(nLineSpace) * iLine; |
3617 | 0 | const size_t iOffsetLine = static_cast<size_t>(iLine) * nXSize; |
3618 | 0 | int iCol = 0; |
3619 | 0 | for (; iCol < nXSize - (VALUES_PER_REG - 1); |
3620 | 0 | iCol += VALUES_PER_REG) |
3621 | 0 | { |
3622 | 0 | __m128i reg0 = halfConstant; |
3623 | 0 | __m128i reg1 = halfConstant; |
3624 | 0 | __m128i reg2 = halfConstant; |
3625 | 0 | __m128i reg3 = halfConstant; |
3626 | 0 | for (int iSrc = 0; iSrc < nSources; ++iSrc) |
3627 | 0 | { |
3628 | 0 | const __m128i inputVal = |
3629 | 0 | _mm_loadu_si128(reinterpret_cast<const __m128i *>( |
3630 | 0 | static_cast<const T * CPL_RESTRICT>( |
3631 | 0 | papoSources[iSrc]) + |
3632 | 0 | iOffsetLine + iCol)); |
3633 | 0 | const __m128i low = |
3634 | 0 | _mm_unpacklo_epi8(inputVal, _mm_setzero_si128()); |
3635 | 0 | const __m128i high = |
3636 | 0 | _mm_unpackhi_epi8(inputVal, _mm_setzero_si128()); |
3637 | 0 | reg0 = _mm_add_epi32( |
3638 | 0 | reg0, _mm_unpacklo_epi16(low, _mm_setzero_si128())); |
3639 | 0 | reg1 = _mm_add_epi32( |
3640 | 0 | reg1, _mm_unpackhi_epi16(low, _mm_setzero_si128())); |
3641 | 0 | reg2 = _mm_add_epi32( |
3642 | 0 | reg2, |
3643 | 0 | _mm_unpacklo_epi16(high, _mm_setzero_si128())); |
3644 | 0 | reg3 = _mm_add_epi32( |
3645 | 0 | reg3, |
3646 | 0 | _mm_unpackhi_epi16(high, _mm_setzero_si128())); |
3647 | 0 | } |
3648 | 0 | reg0 /= fast_d; |
3649 | 0 | reg1 /= fast_d; |
3650 | 0 | reg2 /= fast_d; |
3651 | 0 | reg3 /= fast_d; |
3652 | 0 | _mm_storeu_si128( |
3653 | 0 | reinterpret_cast<__m128i *>(pDest + iCol), |
3654 | 0 | _mm_packus_epi16(packus_epi32(reg0, reg1), |
3655 | 0 | packus_epi32(reg2, reg3))); |
3656 | 0 | } |
3657 | 0 | for (; iCol < nXSize; ++iCol) |
3658 | 0 | { |
3659 | 0 | uint32_t acc = nSources / 2; |
3660 | 0 | for (int iSrc = 0; iSrc < nSources; ++iSrc) |
3661 | 0 | { |
3662 | 0 | acc += static_cast<const T * CPL_RESTRICT>( |
3663 | 0 | papoSources[iSrc])[iOffsetLine + iCol]; |
3664 | 0 | } |
3665 | 0 | pDest[iCol] = static_cast<T>(acc / nSources); |
3666 | 0 | } |
3667 | 0 | } |
3668 | 0 | return CE_None; |
3669 | 0 | } |
3670 | | |
3671 | 0 | if (!bHasNoData && eSrcType == GDT_UInt16 && eBufType == GDT_UInt16 && |
3672 | 0 | nPixelSpace == 2 && |
3673 | 0 | nSources <= std::numeric_limits<int32_t>::max() / |
3674 | 0 | std::numeric_limits<uint16_t>::max()) |
3675 | 0 | { |
3676 | 0 | libdivide::divider<uint32_t> fast_d(nSources); |
3677 | 0 | using T = uint16_t; |
3678 | 0 | const auto halfConstant = _mm_set1_epi32(nSources / 2); |
3679 | 0 | constexpr int VALUES_PER_REG = 8; |
3680 | 0 | for (int iLine = 0; iLine < nYSize; ++iLine) |
3681 | 0 | { |
3682 | 0 | T *CPL_RESTRICT pDest = reinterpret_cast<T *>( |
3683 | 0 | static_cast<GByte *>(pData) + |
3684 | 0 | static_cast<GSpacing>(nLineSpace) * iLine); |
3685 | 0 | const size_t iOffsetLine = static_cast<size_t>(iLine) * nXSize; |
3686 | 0 | int iCol = 0; |
3687 | 0 | for (; iCol < nXSize - (VALUES_PER_REG - 1); |
3688 | 0 | iCol += VALUES_PER_REG) |
3689 | 0 | { |
3690 | 0 | __m128i reg0 = halfConstant; |
3691 | 0 | __m128i reg1 = halfConstant; |
3692 | 0 | for (int iSrc = 0; iSrc < nSources; ++iSrc) |
3693 | 0 | { |
3694 | 0 | const __m128i inputVal = |
3695 | 0 | _mm_loadu_si128(reinterpret_cast<const __m128i *>( |
3696 | 0 | static_cast<const T * CPL_RESTRICT>( |
3697 | 0 | papoSources[iSrc]) + |
3698 | 0 | iOffsetLine + iCol)); |
3699 | 0 | reg0 = _mm_add_epi32( |
3700 | 0 | reg0, |
3701 | 0 | _mm_unpacklo_epi16(inputVal, _mm_setzero_si128())); |
3702 | 0 | reg1 = _mm_add_epi32( |
3703 | 0 | reg1, |
3704 | 0 | _mm_unpackhi_epi16(inputVal, _mm_setzero_si128())); |
3705 | 0 | } |
3706 | 0 | reg0 /= fast_d; |
3707 | 0 | reg1 /= fast_d; |
3708 | 0 | _mm_storeu_si128(reinterpret_cast<__m128i *>(pDest + iCol), |
3709 | 0 | packus_epi32(reg0, reg1)); |
3710 | 0 | } |
3711 | 0 | for (; iCol < nXSize; ++iCol) |
3712 | 0 | { |
3713 | 0 | uint32_t acc = nSources / 2; |
3714 | 0 | for (int iSrc = 0; iSrc < nSources; ++iSrc) |
3715 | 0 | { |
3716 | 0 | acc += static_cast<const T * CPL_RESTRICT>( |
3717 | 0 | papoSources[iSrc])[iOffsetLine + iCol]; |
3718 | 0 | } |
3719 | 0 | pDest[iCol] = static_cast<T>(acc / nSources); |
3720 | 0 | } |
3721 | 0 | } |
3722 | 0 | return CE_None; |
3723 | 0 | } |
3724 | | |
3725 | 0 | if (!bHasNoData && eSrcType == GDT_Int16 && eBufType == GDT_Int16 && |
3726 | 0 | nPixelSpace == 2 && |
3727 | 0 | nSources <= std::numeric_limits<int32_t>::max() / |
3728 | 0 | std::numeric_limits<uint16_t>::max()) |
3729 | 0 | { |
3730 | 0 | libdivide::divider<uint32_t> fast_d(nSources); |
3731 | 0 | using T = int16_t; |
3732 | 0 | const auto halfConstant = _mm_set1_epi32(nSources / 2); |
3733 | 0 | const auto shift = _mm_set1_epi16(std::numeric_limits<T>::min()); |
3734 | 0 | constexpr int VALUES_PER_REG = 8; |
3735 | 0 | for (int iLine = 0; iLine < nYSize; ++iLine) |
3736 | 0 | { |
3737 | 0 | T *CPL_RESTRICT pDest = reinterpret_cast<T *>( |
3738 | 0 | static_cast<GByte *>(pData) + |
3739 | 0 | static_cast<GSpacing>(nLineSpace) * iLine); |
3740 | 0 | const size_t iOffsetLine = static_cast<size_t>(iLine) * nXSize; |
3741 | 0 | int iCol = 0; |
3742 | 0 | for (; iCol < nXSize - (VALUES_PER_REG - 1); |
3743 | 0 | iCol += VALUES_PER_REG) |
3744 | 0 | { |
3745 | 0 | __m128i reg0 = halfConstant; |
3746 | 0 | __m128i reg1 = halfConstant; |
3747 | 0 | for (int iSrc = 0; iSrc < nSources; ++iSrc) |
3748 | 0 | { |
3749 | | // Shift input values by 32768 to get unsigned values |
3750 | 0 | const __m128i inputVal = _mm_add_epi16( |
3751 | 0 | _mm_loadu_si128(reinterpret_cast<const __m128i *>( |
3752 | 0 | static_cast<const T * CPL_RESTRICT>( |
3753 | 0 | papoSources[iSrc]) + |
3754 | 0 | iOffsetLine + iCol)), |
3755 | 0 | shift); |
3756 | 0 | reg0 = _mm_add_epi32( |
3757 | 0 | reg0, |
3758 | 0 | _mm_unpacklo_epi16(inputVal, _mm_setzero_si128())); |
3759 | 0 | reg1 = _mm_add_epi32( |
3760 | 0 | reg1, |
3761 | 0 | _mm_unpackhi_epi16(inputVal, _mm_setzero_si128())); |
3762 | 0 | } |
3763 | 0 | reg0 /= fast_d; |
3764 | 0 | reg1 /= fast_d; |
3765 | 0 | _mm_storeu_si128( |
3766 | 0 | reinterpret_cast<__m128i *>(pDest + iCol), |
3767 | 0 | _mm_add_epi16(packus_epi32(reg0, reg1), shift)); |
3768 | 0 | } |
3769 | 0 | for (; iCol < nXSize; ++iCol) |
3770 | 0 | { |
3771 | 0 | int32_t acc = (-std::numeric_limits<T>::min()) * nSources + |
3772 | 0 | nSources / 2; |
3773 | 0 | for (int iSrc = 0; iSrc < nSources; ++iSrc) |
3774 | 0 | { |
3775 | 0 | acc += static_cast<const T * CPL_RESTRICT>( |
3776 | 0 | papoSources[iSrc])[iOffsetLine + iCol]; |
3777 | 0 | } |
3778 | 0 | pDest[iCol] = static_cast<T>(acc / nSources + |
3779 | 0 | std::numeric_limits<T>::min()); |
3780 | 0 | } |
3781 | 0 | } |
3782 | 0 | return CE_None; |
3783 | 0 | } |
3784 | 0 | } |
3785 | 0 | #endif // defined(USE_SSE2) && !defined(USE_NEON_OPTIMIZATIONS) |
3786 | | |
3787 | 0 | #if defined(USE_SSE2) |
3788 | | if constexpr (std::is_same_v<Kernel, MeanKernel>) |
3789 | 0 | { |
3790 | 0 | if (!bHasNoData && eSrcType == GDT_Float32 && eBufType == GDT_Float32 && |
3791 | 0 | nPixelSpace == 4 && nSources > 0) |
3792 | 0 | { |
3793 | 0 | OptimizedMeanFloatSSE2<float, SSEWrapperFloat>( |
3794 | 0 | papoSources, nSources, pData, nXSize, nYSize, nLineSpace); |
3795 | 0 | return CE_None; |
3796 | 0 | } |
3797 | | |
3798 | 0 | if (!bHasNoData && eSrcType == GDT_Float64 && eBufType == GDT_Float64 && |
3799 | 0 | nPixelSpace == 8 && nSources > 0) |
3800 | 0 | { |
3801 | 0 | OptimizedMeanFloatSSE2<double, SSEWrapperDouble>( |
3802 | 0 | papoSources, nSources, pData, nXSize, nYSize, nLineSpace); |
3803 | 0 | return CE_None; |
3804 | 0 | } |
3805 | 0 | } |
3806 | 0 | #endif // USE_SSE2 |
3807 | | |
3808 | | /* ---- Set pixels ---- */ |
3809 | 0 | size_t ii = 0; |
3810 | 0 | for (int iLine = 0; iLine < nYSize; ++iLine) |
3811 | 0 | { |
3812 | 0 | for (int iCol = 0; iCol < nXSize; ++iCol, ++ii) |
3813 | 0 | { |
3814 | 0 | oKernel.Reset(); |
3815 | 0 | bool bWriteNoData = false; |
3816 | |
|
3817 | 0 | for (int iSrc = 0; iSrc < nSources; ++iSrc) |
3818 | 0 | { |
3819 | 0 | const double dfVal = GetSrcVal(papoSources[iSrc], eSrcType, ii); |
3820 | |
|
3821 | 0 | if (bHasNoData && IsNoData(dfVal, dfNoData)) |
3822 | 0 | { |
3823 | 0 | if (bPropagateNoData) |
3824 | 0 | { |
3825 | 0 | bWriteNoData = true; |
3826 | 0 | break; |
3827 | 0 | } |
3828 | 0 | } |
3829 | 0 | else |
3830 | 0 | { |
3831 | 0 | oKernel.ProcessPixel(dfVal); |
3832 | 0 | } |
3833 | 0 | } |
3834 | |
|
3835 | 0 | double dfPixVal{dfNoData}; |
3836 | 0 | if (!bWriteNoData && oKernel.HasValue()) |
3837 | 0 | { |
3838 | 0 | dfPixVal = oKernel.GetValue(); |
3839 | 0 | } |
3840 | |
|
3841 | 0 | GDALCopyWords(&dfPixVal, GDT_Float64, 0, |
3842 | 0 | static_cast<GByte *>(pData) + |
3843 | 0 | static_cast<GSpacing>(nLineSpace) * iLine + |
3844 | 0 | iCol * nPixelSpace, |
3845 | 0 | eBufType, nPixelSpace, 1); |
3846 | 0 | } |
3847 | 0 | } |
3848 | | |
3849 | | /* ---- Return success ---- */ |
3850 | 0 | return CE_None; |
3851 | 0 | } // BasicPixelFunc Unexecuted instantiation: pixelfunctions.cpp:CPLErr BasicPixelFunc<MeanKernel>(void**, int, void*, int, int, GDALDataType, GDALDataType, int, int, char const* const*) Unexecuted instantiation: pixelfunctions.cpp:CPLErr BasicPixelFunc<GeoMeanKernel>(void**, int, void*, int, int, GDALDataType, GDALDataType, int, int, char const* const*) Unexecuted instantiation: pixelfunctions.cpp:CPLErr BasicPixelFunc<HarmonicMeanKernel>(void**, int, void*, int, int, GDALDataType, GDALDataType, int, int, char const* const*) Unexecuted instantiation: pixelfunctions.cpp:CPLErr BasicPixelFunc<MedianKernel>(void**, int, void*, int, int, GDALDataType, GDALDataType, int, int, char const* const*) Unexecuted instantiation: pixelfunctions.cpp:CPLErr BasicPixelFunc<ModeKernel>(void**, int, void*, int, int, GDALDataType, GDALDataType, int, int, char const* const*) |
3852 | | |
3853 | | /************************************************************************/ |
3854 | | /* GDALRegisterDefaultPixelFunc() */ |
3855 | | /************************************************************************/ |
3856 | | |
3857 | | /** |
3858 | | * This adds a default set of pixel functions to the global list of |
3859 | | * available pixel functions for derived bands: |
3860 | | * |
3861 | | * - "real": extract real part from a single raster band (just a copy if the |
3862 | | * input is non-complex) |
3863 | | * - "imag": extract imaginary part from a single raster band (0 for |
3864 | | * non-complex) |
3865 | | * - "complex": make a complex band merging two bands used as real and |
3866 | | * imag values |
3867 | | * - "polar": make a complex band using input bands for amplitude and |
3868 | | * phase values (b1 * exp( j * b2 )) |
3869 | | * - "mod": extract module from a single raster band (real or complex) |
3870 | | * - "phase": extract phase from a single raster band [-PI,PI] (0 or PI for |
3871 | | non-complex) |
3872 | | * - "conj": computes the complex conjugate of a single raster band (just a |
3873 | | * copy if the input is non-complex) |
3874 | | * - "sum": sum 2 or more raster bands |
3875 | | * - "diff": computes the difference between 2 raster bands (b1 - b2) |
3876 | | * - "mul": multiply 2 or more raster bands |
3877 | | * - "div": divide one raster band by another (b1 / b2). |
3878 | | * - "min": minimum value of 2 or more raster bands |
3879 | | * - "max": maximum value of 2 or more raster bands |
3880 | | * - "norm_diff": computes the normalized difference between two raster bands: |
3881 | | * ``(b1 - b2)/(b1 + b2)``. |
3882 | | * - "cmul": multiply the first band for the complex conjugate of the second |
3883 | | * - "inv": inverse (1./x). |
3884 | | * - "intensity": computes the intensity Re(x*conj(x)) of a single raster band |
3885 | | * (real or complex) |
3886 | | * - "sqrt": perform the square root of a single raster band (real only) |
3887 | | * - "log10": compute the logarithm (base 10) of the abs of a single raster |
3888 | | * band (real or complex): log10( abs( x ) ) |
3889 | | * - "dB": perform conversion to dB of the abs of a single raster |
3890 | | * band (real or complex): 20. * log10( abs( x ) ). |
3891 | | * Note: the optional fact parameter can be set to 10. to get the |
3892 | | * alternative formula: 10. * log10( abs( x ) ) |
3893 | | * - "exp": computes the exponential of each element in the input band ``x`` |
3894 | | * (of real values): ``e ^ x``. |
3895 | | * The function also accepts two optional parameters: ``base`` and |
3896 | | ``fact`` |
3897 | | * that allow to compute the generalized formula: ``base ^ ( fact * |
3898 | | x)``. |
3899 | | * Note: this function is the recommended one to perform conversion |
3900 | | * form logarithmic scale (dB): `` 10. ^ (x / 20.)``, in this case |
3901 | | * ``base = 10.`` and ``fact = 1./20`` |
3902 | | * - "dB2amp": perform scale conversion from logarithmic to linear |
3903 | | * (amplitude) (i.e. 10 ^ ( x / 20 ) ) of a single raster |
3904 | | * band (real only). |
3905 | | * Deprecated in GDAL v3.5. Please use the ``exp`` pixel function |
3906 | | with |
3907 | | * ``base = 10.`` and ``fact = 0.05`` i.e. ``1./20`` |
3908 | | * - "dB2pow": perform scale conversion from logarithmic to linear |
3909 | | * (power) (i.e. 10 ^ ( x / 10 ) ) of a single raster |
3910 | | * band (real only) |
3911 | | * Deprecated in GDAL v3.5. Please use the ``exp`` pixel function |
3912 | | with |
3913 | | * ``base = 10.`` and ``fact = 0.1`` i.e. ``1./10`` |
3914 | | * - "pow": raise a single raster band to a constant power |
3915 | | * - "interpolate_linear": interpolate values between two raster bands |
3916 | | * using linear interpolation |
3917 | | * - "interpolate_exp": interpolate values between two raster bands using |
3918 | | * exponential interpolation |
3919 | | * - "scale": Apply the RasterBand metadata values of "offset" and "scale" |
3920 | | * - "reclassify": Reclassify values matching ranges in a table |
3921 | | * - "nan": Convert incoming NoData values to IEEE 754 nan |
3922 | | * |
3923 | | * @see GDALAddDerivedBandPixelFunc |
3924 | | * |
3925 | | * @return CE_None |
3926 | | */ |
3927 | | CPLErr GDALRegisterDefaultPixelFunc() |
3928 | 0 | { |
3929 | 0 | GDALAddDerivedBandPixelFunc("real", RealPixelFunc); |
3930 | 0 | GDALAddDerivedBandPixelFunc("imag", ImagPixelFunc); |
3931 | 0 | GDALAddDerivedBandPixelFunc("complex", ComplexPixelFunc); |
3932 | 0 | GDALAddDerivedBandPixelFuncWithArgs("polar", PolarPixelFunc, |
3933 | 0 | pszPolarPixelFuncMetadata); |
3934 | 0 | GDALAddDerivedBandPixelFunc("mod", ModulePixelFunc); |
3935 | 0 | GDALAddDerivedBandPixelFunc("phase", PhasePixelFunc); |
3936 | 0 | GDALAddDerivedBandPixelFunc("conj", ConjPixelFunc); |
3937 | 0 | GDALAddDerivedBandPixelFuncWithArgs("sum", SumPixelFunc, |
3938 | 0 | pszSumPixelFuncMetadata); |
3939 | 0 | GDALAddDerivedBandPixelFuncWithArgs("diff", DiffPixelFunc, |
3940 | 0 | pszDiffPixelFuncMetadata); |
3941 | 0 | GDALAddDerivedBandPixelFuncWithArgs("mul", MulPixelFunc, |
3942 | 0 | pszMulPixelFuncMetadata); |
3943 | 0 | GDALAddDerivedBandPixelFuncWithArgs("div", DivPixelFunc, |
3944 | 0 | pszDivPixelFuncMetadata); |
3945 | 0 | GDALAddDerivedBandPixelFunc("cmul", CMulPixelFunc); |
3946 | 0 | GDALAddDerivedBandPixelFuncWithArgs("inv", InvPixelFunc, |
3947 | 0 | pszInvPixelFuncMetadata); |
3948 | 0 | GDALAddDerivedBandPixelFunc("intensity", IntensityPixelFunc); |
3949 | 0 | GDALAddDerivedBandPixelFuncWithArgs("sqrt", SqrtPixelFunc, |
3950 | 0 | pszSqrtPixelFuncMetadata); |
3951 | 0 | GDALAddDerivedBandPixelFuncWithArgs("log10", Log10PixelFunc, |
3952 | 0 | pszLog10PixelFuncMetadata); |
3953 | 0 | GDALAddDerivedBandPixelFuncWithArgs("dB", DBPixelFunc, |
3954 | 0 | pszDBPixelFuncMetadata); |
3955 | 0 | GDALAddDerivedBandPixelFuncWithArgs("exp", ExpPixelFunc, |
3956 | 0 | pszExpPixelFuncMetadata); |
3957 | 0 | GDALAddDerivedBandPixelFunc("dB2amp", |
3958 | 0 | dB2AmpPixelFunc); // deprecated in v3.5 |
3959 | 0 | GDALAddDerivedBandPixelFunc("dB2pow", |
3960 | 0 | dB2PowPixelFunc); // deprecated in v3.5 |
3961 | 0 | GDALAddDerivedBandPixelFuncWithArgs("pow", PowPixelFunc, |
3962 | 0 | pszPowPixelFuncMetadata); |
3963 | 0 | GDALAddDerivedBandPixelFuncWithArgs("interpolate_linear", |
3964 | 0 | InterpolatePixelFunc<InterpolateLinear>, |
3965 | 0 | pszInterpolatePixelFuncMetadata); |
3966 | 0 | GDALAddDerivedBandPixelFuncWithArgs( |
3967 | 0 | "interpolate_exp", InterpolatePixelFunc<InterpolateExponential>, |
3968 | 0 | pszInterpolatePixelFuncMetadata); |
3969 | 0 | GDALAddDerivedBandPixelFuncWithArgs("replace_nodata", |
3970 | 0 | ReplaceNoDataPixelFunc, |
3971 | 0 | pszReplaceNoDataPixelFuncMetadata); |
3972 | 0 | GDALAddDerivedBandPixelFuncWithArgs("scale", ScalePixelFunc, |
3973 | 0 | pszScalePixelFuncMetadata); |
3974 | 0 | GDALAddDerivedBandPixelFuncWithArgs("norm_diff", NormDiffPixelFunc, |
3975 | 0 | pszNormDiffPixelFuncMetadata); |
3976 | 0 | GDALAddDerivedBandPixelFuncWithArgs("min", MinPixelFunc, |
3977 | 0 | pszMinMaxFuncMetadataNodata); |
3978 | 0 | GDALAddDerivedBandPixelFuncWithArgs("max", MaxPixelFunc, |
3979 | 0 | pszMinMaxFuncMetadataNodata); |
3980 | 0 | GDALAddDerivedBandPixelFuncWithArgs("expression", ExprPixelFunc, |
3981 | 0 | pszExprPixelFuncMetadata); |
3982 | 0 | GDALAddDerivedBandPixelFuncWithArgs("reclassify", ReclassifyPixelFunc, |
3983 | 0 | pszReclassifyPixelFuncMetadata); |
3984 | 0 | GDALAddDerivedBandPixelFuncWithArgs("mean", BasicPixelFunc<MeanKernel>, |
3985 | 0 | pszBasicPixelFuncMetadata); |
3986 | 0 | GDALAddDerivedBandPixelFuncWithArgs("geometric_mean", |
3987 | 0 | BasicPixelFunc<GeoMeanKernel>, |
3988 | 0 | pszBasicPixelFuncMetadata); |
3989 | 0 | GDALAddDerivedBandPixelFuncWithArgs("harmonic_mean", |
3990 | 0 | BasicPixelFunc<HarmonicMeanKernel>, |
3991 | 0 | pszBasicPixelFuncMetadata); |
3992 | 0 | GDALAddDerivedBandPixelFuncWithArgs("median", BasicPixelFunc<MedianKernel>, |
3993 | 0 | pszBasicPixelFuncMetadata); |
3994 | 0 | GDALAddDerivedBandPixelFuncWithArgs("mode", BasicPixelFunc<ModeKernel>, |
3995 | 0 | pszBasicPixelFuncMetadata); |
3996 | 0 | return CE_None; |
3997 | 0 | } |