Blame view

stim/envi/binary.h 26.1 KB
1f2ca8ec   David Mayerich   added files for b...
1
2
  
  //make sure that this header file is only loaded once
180d7f3a   David Mayerich   added binary file...
3
4
5
  #ifndef RTS_BINARY_H
  #define RTS_BINARY_H
  
e8eb202f   David Mayerich   added a new ENVI ...
6
  #include "../envi/envi_header.h"
4dcc9392   heziqi   Ziqi's first chan...
7
  #include "../math/vector.h"
180d7f3a   David Mayerich   added binary file...
8
9
  #include <fstream>
  #include <sys/stat.h>
15a31763   David Mayerich   set up buffer est...
10
  #include <cstring>
5435830b   David Mayerich   linux chrono fixes
11
  #include <chrono>
008ae990   David Mayerich   implemented Windo...
12
13
  #ifdef _WIN32
  #include <Windows.h>
15a31763   David Mayerich   set up buffer est...
14
15
  #else
  #include <unistd.h>
008ae990   David Mayerich   implemented Windo...
16
17
  #endif
  
8a86bd56   David Mayerich   changed rts names...
18
  namespace stim{
180d7f3a   David Mayerich   added binary file...
19
  
6db250f5   David Mayerich   adaptive streamin...
20
21
22
23
  /// This class calculates the optimal setting for independent parameter b (batch size) for
  ///		minimizing the dependent parameter bps (bytes per second)
  class stream_optimizer{
  protected:
0a575bb1   David Mayerich   updated optimizat...
24
  	size_t Bps[2];							//bytes per second for the previous batch
6db250f5   David Mayerich   adaptive streamin...
25
26
27
  	size_t interval_B;					//number of bytes processed this interval
  	size_t interval_ms;					//number of milliseconds spent in the current interval
  	size_t n[2];							//current batch size (in bytes)
0a575bb1   David Mayerich   updated optimizat...
28
  	size_t h;							//spacing used for finite difference calculations
6db250f5   David Mayerich   adaptive streamin...
29
30
31
  	size_t dn;							//delta value (in bytes) for setting the batch size (minimum change in batch parameter)
  	size_t maxn;						//maximum value for the batch size
  
0a575bb1   David Mayerich   updated optimizat...
32
33
  	double alpha;						//alpha value controls the factor of the gradient that is used to calculate the next point (speed of convergence)
  
6db250f5   David Mayerich   adaptive streamin...
34
35
36
37
38
39
  	bool sample_step;					//calculating the derivative (this class alternates between calculating dBps and B)
  	bool forward_diff;					//evaluate the derivative using forward differences
  
  	size_t window_ms;					//size of the interval (in milliseconds) integrated to get a reliable bps value
  
  	// This function rounds x to the nearest value within dB
0a575bb1   David Mayerich   updated optimizat...
40
41
42
43
44
  	size_t round_limit(double n0){
  		if(n0 < 0) return dn;					//if n0 is less than zero, return the lowest possible n
  
  		size_t new_n = (size_t)(n0 + 0.5);		//now n0 must be positive, so round it to the nearest integer
  		if(new_n > maxn) new_n = maxn;			//limit the returned size of x to within the specified bounds
6db250f5   David Mayerich   adaptive streamin...
45
  
0a575bb1   David Mayerich   updated optimizat...
46
  		size_t lowest = new_n / dn;
6db250f5   David Mayerich   adaptive streamin...
47
  		size_t highest = lowest + dn;
0a575bb1   David Mayerich   updated optimizat...
48
  		size_t diff[2] = {new_n - lowest, highest - new_n};	//calculate the two differences
6db250f5   David Mayerich   adaptive streamin...
49
50
51
52
53
54
55
56
  		if(diff[0] < diff[1])
  			return lowest;
  		return highest;
  	}
  
  public:
  
  	//constructor initializes a stream optimizer
c6251f8b   David Mayerich   allows user to de...
57
  	stream_optimizer(size_t min_batch_size, size_t max_batch_size, double a = 0.003, size_t probe_step = 5, size_t window = 2000){
0a575bb1   David Mayerich   updated optimizat...
58
59
  		//Bps = 0;						//initialize to zero bytes per second processed
  		Bps[0] = Bps[1] = 0;			//initialize the bits per second to 0
6db250f5   David Mayerich   adaptive streamin...
60
61
62
63
  		interval_B = 0;					//zero bytes have been processed at initialization
  		interval_ms = 0;				//no time has been spent on the batch so far
  		dn = min_batch_size;			//set the minimum batch size as the minimum change in batch size
  		maxn = max_batch_size;			//set the maximum batch size
0a575bb1   David Mayerich   updated optimizat...
64
  		n[0] = max_batch_size;			//set B
8933935f   David Mayerich   updated VERBOSE i...
65
  		h = (max_batch_size / min_batch_size) / probe_step * dn;
0a575bb1   David Mayerich   updated optimizat...
66
67
68
69
  		std::cout<<"h = "<<h<<std::endl;
  		if(h < dn) h = dn;
  		alpha = a;
  		//n[0] = round_limit( (max_batch_size - min_batch_size)/2 );
6db250f5   David Mayerich   adaptive streamin...
70
71
72
73
74
  		window_ms = window;		//minimum integration interval (for getting a reliable bps measure)
  		sample_step = true;					//the first step is to calculate the derivative
  		forward_diff = true;			//start with the forward difference (since we start at the maximum batch size)
  	}
  
c6251f8b   David Mayerich   allows user to de...
75
  	size_t update(size_t bytes_processed, size_t ms_spent, size_t& data_rate, bool VERBOSE = false){
0a575bb1   David Mayerich   updated optimizat...
76
77
  		interval_B += bytes_processed;		//increment the number of bytes processed
  		interval_ms += ms_spent;			//increment the number of milliseconds spent processing
c6251f8b   David Mayerich   allows user to de...
78
  		data_rate = interval_B / interval_ms;
0a575bb1   David Mayerich   updated optimizat...
79
80
81
82
83
84
  
  		//if we have sufficient information to evaluate the optimization function at this point
  		if(interval_ms < window_ms){					//if insufficient time has passed to get a reliable Bps measurement
  			return n[0];
  		}
  		else{											//if we have collected enough information for a reliable Bps estimate
c6251f8b   David Mayerich   allows user to de...
85
  			
0a575bb1   David Mayerich   updated optimizat...
86
  			if(Bps[0] == 0){							//if n[0] hasn't been evaluated yet, this is the first step
c6251f8b   David Mayerich   allows user to de...
87
  				Bps[0] = data_rate;						//set the initial Bps value
0a575bb1   David Mayerich   updated optimizat...
88
  				n[1] = n[0] - h;						//set the position of the next sample point
c6251f8b   David Mayerich   allows user to de...
89
90
  				if(VERBOSE)
  					std::cout<<"Bps value at n = "<<n[0]<<" is "<<Bps[0]<<" Bps, probing n = "<<n[1]<<std::endl;
0a575bb1   David Mayerich   updated optimizat...
91
92
93
  				return n[1];							//return the probe point
  			}
  			else{
c6251f8b   David Mayerich   allows user to de...
94
  				Bps[1] = data_rate;						//set the Bps for the current point (n[1])
0a575bb1   David Mayerich   updated optimizat...
95
96
97
98
99
100
101
102
103
  
  				double Bps_p;							//allocate a variable for the derivative
  				//calculate the derivative
  				if(n[0] < n[1]){						//if the current point is less than the previous one (probably the most common)
  					Bps_p = ((double)Bps[1] - (double)Bps[0]) / (double)h;		//calculate the derivative using the forward finite difference
  				}
  				else{
  					Bps_p = ((double)Bps[0] - (double)Bps[1]) / (double)h;		//calculate the derivative using the backward finite difference
  				}
c6251f8b   David Mayerich   allows user to de...
104
105
  				if(VERBOSE)
  					std::cout<<"     probed n = "<<n[1]<<" with "<<Bps[1]<<" Bps, gradient = "<<Bps_p<<" Bps"<<std::endl;
0a575bb1   David Mayerich   updated optimizat...
106
107
108
109
110
111
112
113
  
  				double new_n_precise = n[0] + alpha * Bps_p;			//calculate the next point (snap to closest integer)
  				size_t new_n_nearest = round_limit(new_n_precise);		//calculate the next point (given batch parameters)
  
  				if(new_n_nearest == n[0]){								//if the newest point is the same as the original point
  					Bps[0] = Bps[1];									//update the Bps
  					//if(n[0] == dn) n[1] = n[0] + h;					//if we're on the left edge, probe forward
  					//else n[1] = n[0] - h;								//otherwise probe backwards
c6251f8b   David Mayerich   allows user to de...
114
115
  					if(VERBOSE)
  						std::cout<<"     staying at n = "<<n[0]<<" for now"<<std::endl;
0a575bb1   David Mayerich   updated optimizat...
116
117
118
119
120
121
122
123
  					//return n[1];										//return the probe point
  
  					Bps[0] = 0;											//reset the Bps for the current point
  					return n[0];										//return the current point for a re-calculation
  				}
  				else{													//if the newest point is different from the original point
  					n[0] = new_n_nearest;								//move to the new point
  					Bps[0] = 0;											//set the Bps to zero (point hasn't been tested)
c6251f8b   David Mayerich   allows user to de...
124
125
  					if(VERBOSE)
  						std::cout<<"     moving to n = "<<n[0]<<std::endl;
0a575bb1   David Mayerich   updated optimizat...
126
127
128
129
130
131
132
  					return n[0];										//return the new point
  				}
  			}
  		}
  	}
  
  	/*// this function updates the optimizer, given the number of bytes processed in an interval and time spent processing
6db250f5   David Mayerich   adaptive streamin...
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
  	size_t update(size_t bytes_processed, size_t ms_spent){
  		interval_B += bytes_processed;		//increment the number of bytes processed
  		interval_ms += ms_spent;			//increment the number of milliseconds spent processing
  
  		//if we have sufficient information to evaluate the optimization function at this point
  		if(interval_ms >= window_ms){					//if sufficient time has passed to get a reliable Bps measurement
  			size_t new_Bps = interval_B / interval_ms;	//calculate the current Bps
  
  			if(sample_step){							//if this is a sample step, collect the information for Bps = f(n0)
  				Bps = new_Bps;							//set the Bps to the evaluated value
  				n[1] = n[0] - dn;								//reduce the batch size by one delta to take a second sample
  				if(n[1] == 0){							//if the resulting batch size is zero
  					n[1] = 2*dn;						//we're at the left edge: set the new sample point to 2*dn
  				}
  
  				interval_B = interval_ms = 0;			//start a new interval at the new sample point
  				sample_step = false;					//next step will calculate the new batch size via optimization
  				return n[1];								//return the new batch size
  			}
  			else{								//if we have sufficient information to evaluate the derivative and optimize
  				double f = (double)new_Bps;				//we have evaluated the function at this location
  				double fprime;
  				if(n[1] < n[0] ){									//if the new point is less than the previous point (usually the case)
  					fprime = (double)(Bps - new_Bps) / (double)dn;	//calculate the forward difference
  				}
  				else{												//if the new point is larger (only happens at the minimum limit)
  					fprime = (double)(new_Bps - Bps) / (double)dn;	//calculate the backward difference
  				}
  				size_t bestn = n[1] - (size_t)(f / fprime);			//calculate the best value for B using Newton's method
  				n[0] = round_limit( (size_t)bestn );						//set the new dependent point
  				sample_step = true;									//the next step will be a sample step
  			}
  
  		}
  		if(sample_step) return n[0];
  		return n[1];										//insufficient information, keep the same batch size
0a575bb1   David Mayerich   updated optimizat...
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
  	}*/
  
  	/*size_t update(size_t bytes_processed, size_t ms_spent){
  		interval_B += bytes_processed;		//increment the number of bytes processed
  		interval_ms += ms_spent;			//increment the number of milliseconds spent processing
  
  		//if( Bps[0] == 0 ){				//if the left boundary hasn't been processed
  
  
  		//if we have sufficient information to evaluate the optimization function at this point
  		if(interval_ms >= window_ms){
  			size_t new_Bps = interval_B / interval_ms;	//calculate the current Bps
  
  			if(Bps[0] == 0)							//if the left interval Bps hasn't been calculated
  				Bps[0] = interval_B / interval_ms;	//that is the interval being processed
  			else
  				Bps[1] = interval_B / interval_ms;	//otherwise the right interval is being processed
  
  			if(Bps[0] != 0 && Bps[1] != 0){			//if both intervals have been processed
  
  
  		}
  	}*/
6db250f5   David Mayerich   adaptive streamin...
192
  
c6251f8b   David Mayerich   allows user to de...
193
194
  	/*size_t update(size_t bytes_processed, size_t ms_spent, size_t& data_rate, bool VERBOSE){
  		size_t time = update(bytes_processed, ms_spent, VERBOSE);
0a575bb1   David Mayerich   updated optimizat...
195
  		data_rate = Bps[0];
6db250f5   David Mayerich   adaptive streamin...
196
  		return time;
c6251f8b   David Mayerich   allows user to de...
197
  	}*/
6db250f5   David Mayerich   adaptive streamin...
198
199
  };
  
33d7d3cf   David Mayerich   documentation for...
200
201
202
203
204
205
206
  /** This class manages the streaming of large multidimensional binary files.
   *  Generally these are hyperspectral files with 2 spatial and 1 spectral dimension. However, this class supports
   *  other dimensions via the template parameter D.
   *
   *  @param T is the data type used to store data to disk (generally float or double)
   *  @param D is the dimension of the data (default 3)
   */
180d7f3a   David Mayerich   added binary file...
207
208
209
210
  template< typename T, unsigned int D = 3 >
  class binary{
  
  protected:
180d7f3a   David Mayerich   added binary file...
211
212
213
  	std::fstream file;		//file stream used for reading and writing
  	std::string name;		//file name
  
63fc1df8   David Mayerich   added an inverse ...
214
  	unsigned long long R[D];		//resolution
9d3ba0b1   David Mayerich   added stim::hsi a...
215
  	unsigned long long header;	//header size (in bytes)
f4c5d71b   David Mayerich   started working w...
216
  	unsigned char* mask;	//pointer to a character array: 0 = background, 1 = foreground (or valid data)
180d7f3a   David Mayerich   added binary file...
217
  
63fc1df8   David Mayerich   added an inverse ...
218
  	double progress;		//stores the progress on the current operation (accessible using a thread)
c6251f8b   David Mayerich   allows user to de...
219
  	size_t data_rate;		//data rate (currently in Bps)
ee4dea28   David Mayerich   fixed errors in c...
220
  
008ae990   David Mayerich   implemented Windo...
221
  	size_t buffer_size;		//available memory for processing large files
ee4dea28   David Mayerich   fixed errors in c...
222
  
33d7d3cf   David Mayerich   documentation for...
223
  	/// Private initialization function used to set default parameters in the data structure.
180d7f3a   David Mayerich   added binary file...
224
  	void init(){
15a31763   David Mayerich   set up buffer est...
225
  		std::memset(R, 0, sizeof(unsigned long long) * D);		//initialize the resolution to zero
008ae990   David Mayerich   implemented Windo...
226
  		header = 0;											//initialize the header size to zero
f4c5d71b   David Mayerich   started working w...
227
  		mask = NULL;
cbce396e   David Mayerich   added support for...
228
  
c6251f8b   David Mayerich   allows user to de...
229
230
231
  		progress = 0;										//initialize the progress for any algorithm to zero
  		data_rate = 0;										//initialize the data rate to zero
  		set_buffer_frac();										//set the maximum buffer size to the default
63fc1df8   David Mayerich   added an inverse ...
232
233
  	}
  
33d7d3cf   David Mayerich   documentation for...
234
  	/// Private helper function that returns the size of the file on disk using system functions.
b0fee590   heziqi   convert long int ...
235
  	long long int get_file_size(){
193ff00e   David Mayerich   dealth with preci...
236
  #ifdef _WIN32
b0fee590   heziqi   convert long int ...
237
238
  		struct _stat64 results;
  		if(_stat64(name.c_str(), &results) == 0)
180d7f3a   David Mayerich   added binary file...
239
  			return results.st_size;
193ff00e   David Mayerich   dealth with preci...
240
241
242
243
244
  #else
  		struct stat results;
  		if(stat(name.c_str(), &results) == 0)
  			return results.st_size;
  #endif
180d7f3a   David Mayerich   added binary file...
245
246
247
  		else return 0;
  	}
  
33d7d3cf   David Mayerich   documentation for...
248
  	/// Private helper function that tests to make sure that the calculated data size specified by the structure is the same as the data size on disk.
180d7f3a   David Mayerich   added binary file...
249
  	bool test_file_size(){
b0fee590   heziqi   convert long int ...
250
  		long long int npts = 1;				//initialize the number of data points to 1
87da9adc   David Mayerich   fixed issues with...
251
252
  		for(unsigned int i = 0; i<D; i++)	//iterate over each dimension
  			npts *= R[i];					//compute the total number of data points in the file
b0fee590   heziqi   convert long int ...
253
  		long long int datasize = npts * sizeof(T);//multiply the sum by the size of the template parameter
87da9adc   David Mayerich   fixed issues with...
254
255
  
  		if(datasize + header == get_file_size()) return true;	//if the byte size matches the file size, we're golden
180d7f3a   David Mayerich   added binary file...
256
257
258
259
  		else return false;					//otherwise return an error
  
  	}
  
f92397d2   David Mayerich   fixed a Win32 bug...
260
261
262
263
264
265
  	/// Private helper function that resets the file pointer to the beginning of the data
  
  	void reset(){
  		file.seekg(header, std::ios_base::beg);
  	}
  
33d7d3cf   David Mayerich   documentation for...
266
267
268
  	/// Private helper file that opens a specified binary file.
  
  	/// @param filename is the name of the binary file to stream
180d7f3a   David Mayerich   added binary file...
269
270
271
272
  	bool open_file(std::string filename){
  		//open the file as binary for reading and writing
  		file.open(filename.c_str(), std::ios::in | std::ios::out | std::ios::binary);
  
ba72c5b0   David Mayerich   added OBJ class
273
274
  		//if the file isn't open, the user may only have read access
  		if(!file.is_open()){
3d0b6243   David Mayerich   fixed hsiproc bug...
275
  			std::cout<<"class STIM::BINARY - failed to open file, trying for read only"<<std::endl;
ba72c5b0   David Mayerich   added OBJ class
276
277
278
279
280
281
282
  			file.open(filename.c_str(), std::ios::in | std::ios::binary);
  			if(!file.is_open()){
  				std::cout<<"               still unable to load the file"<<std::endl;
  				return false;
  			}
  		}
  
180d7f3a   David Mayerich   added binary file...
283
284
285
286
287
288
  		//if the file is successful
  		if(file){
  			name = filename;		//set the name
  			if(test_file_size())	//test the file size
  				return true;
  		}
f92397d2   David Mayerich   fixed a Win32 bug...
289
  
180d7f3a   David Mayerich   added binary file...
290
291
292
  		return false;
  	}
  
724ec347   David Mayerich   simplified the EN...
293
294
  
  
f92397d2   David Mayerich   fixed a Win32 bug...
295
296
  
  
180d7f3a   David Mayerich   added binary file...
297
298
  public:
  
008ae990   David Mayerich   implemented Windo...
299
300
301
302
303
  	//default constructor
  	binary(){
  		init();
  	}
  
9d3ba0b1   David Mayerich   added stim::hsi a...
304
  	double get_progress(){
63fc1df8   David Mayerich   added an inverse ...
305
  		return progress;
cbce396e   David Mayerich   added support for...
306
307
  	}
  
63fc1df8   David Mayerich   added an inverse ...
308
309
  	void reset_progress(){
  		progress = 0;
cbce396e   David Mayerich   added support for...
310
311
  	}
  
c6251f8b   David Mayerich   allows user to de...
312
313
314
315
  	size_t get_data_rate(){
  		return data_rate;
  	}
  
008ae990   David Mayerich   implemented Windo...
316
  	//specify the maximum fraction of available memory that this class will use for buffering
c6251f8b   David Mayerich   allows user to de...
317
  	void set_buffer_frac(double mem_frac = 0.5){				//default to 50%
008ae990   David Mayerich   implemented Windo...
318
319
320
321
322
  #ifdef _WIN32
  		MEMORYSTATUSEX statex;
  		statex.dwLength = sizeof (statex);
  		GlobalMemoryStatusEx (&statex);
  		buffer_size = (size_t)(statex.ullAvailPhys * mem_frac);
15a31763   David Mayerich   set up buffer est...
323
324
325
326
  #else
  		size_t pages = sysconf(_SC_PHYS_PAGES);
  		size_t page_size = sysconf(_SC_PAGE_SIZE);
  		buffer_size = (size_t)(pages * page_size *  mem_frac);
008ae990   David Mayerich   implemented Windo...
327
328
329
  #endif
  	}
  
c6251f8b   David Mayerich   allows user to de...
330
331
332
333
  	void set_buffer_raw(size_t bytes){
  		buffer_size = bytes;
  	}
  
33d7d3cf   David Mayerich   documentation for...
334
335
336
337
338
  	/// Open a binary file for streaming.
  
  	/// @param filename is the name of the binary file
  	/// @param r is a STIM vector specifying the size of the binary file along each dimension
  	/// @param h is the length (in bytes) of any header file (default zero)
9d3ba0b1   David Mayerich   added stim::hsi a...
339
  	bool open(std::string filename, vec<unsigned long long> r, unsigned long long h = 0){
180d7f3a   David Mayerich   added binary file...
340
  
9d3ba0b1   David Mayerich   added stim::hsi a...
341
  		for(unsigned long long i = 0; i < D; i++)		//set the dimensions of the binary file object
180d7f3a   David Mayerich   added binary file...
342
343
  			R[i] = r[i];
  
87da9adc   David Mayerich   fixed issues with...
344
345
346
  		header = h;				//save the header size
  
  		if(!open_file(filename)) return false;	//open the binary file
180d7f3a   David Mayerich   added binary file...
347
  
f92397d2   David Mayerich   fixed a Win32 bug...
348
349
  		//reset();
  
180d7f3a   David Mayerich   added binary file...
350
351
352
  		return test_file_size();
  	}
  
33d7d3cf   David Mayerich   documentation for...
353
354
355
356
357
  	/// Creates a new binary file for streaming
  
  	/// @param filename is the name of the binary file to be created
  	/// @param r is a STIM vector specifying the size of the file along each dimension
  	/// @offset specifies how many bytes to offset the file (used to leave room for a header)
9d3ba0b1   David Mayerich   added stim::hsi a...
358
  	bool create(std::string filename, vec<unsigned long long> r, unsigned long long offset = 0){
f47168a2   heziqi   Ziqi added create...
359
360
361
362
363
  
  		std::ofstream target(filename.c_str(), std::ios::binary);
  
  		//initialize binary file
  		T p = 0;
9d3ba0b1   David Mayerich   added stim::hsi a...
364
  		for(unsigned long long i =0; i < r[0] * r[1] * r[2]; i++){
f47168a2   heziqi   Ziqi added create...
365
366
367
  			target.write((char*)(&p), sizeof(T));
  		}
  
9d3ba0b1   David Mayerich   added stim::hsi a...
368
  		for(unsigned long long i = 0; i < D; i++)		//set the dimensions of the binary file object
f47168a2   heziqi   Ziqi added create...
369
370
371
372
373
374
375
376
377
  			R[i] = r[i];
  
  		header = offset;				//save the header size
  
  		if(!open_file(filename)) return false;	//open the binary file
  
  		return test_file_size();
  	}
  
33d7d3cf   David Mayerich   documentation for...
378
379
380
381
  	/// Writes a single page of data to disk. A page consists of a sequence of data of size R[0] * R[1] * ... * R[D-1].
  
  	/// @param p is a pointer to the data to be written
  	/// @param page is the page number (index of the highest-numbered dimension)
9d3ba0b1   David Mayerich   added stim::hsi a...
382
  	bool write_page( T * p, unsigned long long page){
f47168a2   heziqi   Ziqi added create...
383
384
385
386
387
388
  
  		if(p == NULL){
  			std::cout<<"ERROR: unable to write into file, empty pointer"<<std::endl;
  			exit(1);
  		}
  
f92397d2   David Mayerich   fixed a Win32 bug...
389
  		file.seekg(R[1] * R[0] * page * sizeof(T) + header, std::ios::beg);   //seek to the desired location on disk
8157c392   David Mayerich   added parser and ...
390
  		file.write((char *)p, R[0] * R[1] * sizeof(T));				 //write binary data
f47168a2   heziqi   Ziqi added create...
391
392
393
394
  
  		return true;
  	}
  
33d7d3cf   David Mayerich   documentation for...
395
396
397
398
  	/// Reads a page from disk. A page consists of a sequence of data of size R[0] * R[1] * ... * R[D-1].
  
  	/// @param p is a pointer to pre-allocated memory equal to the page size
  	/// @param page is the index of the page
9d3ba0b1   David Mayerich   added stim::hsi a...
399
  	bool read_page( T * p, unsigned long long page, bool PROGRESS = false){
63fc1df8   David Mayerich   added an inverse ...
400
401
  
  		if(PROGRESS) progress = 0;
c7da85e3   heziqi   he,ziqi add metho...
402
  
5d5ce39a   heziqi   Ziqi convert the ...
403
  		if (page >= R[2]){										//make sure the bank number is right
2ad0ce4f   David Mayerich   syntactic edits b...
404
  			std::cout<<"ERROR: page out of range"<<std::endl;
5d5ce39a   heziqi   Ziqi convert the ...
405
  			return false;
c7da85e3   heziqi   he,ziqi add metho...
406
407
  		}
  
f92397d2   David Mayerich   fixed a Win32 bug...
408
  		file.seekg(R[1] * R[0] * page * sizeof(T) + header, std::ios::beg);   //write into memory from the binary file
2ad0ce4f   David Mayerich   syntactic edits b...
409
  		file.read((char *)p, R[0] * R[1] * sizeof(T));
63fc1df8   David Mayerich   added an inverse ...
410
  		if(PROGRESS) progress = 100;
5d5ce39a   heziqi   Ziqi convert the ...
411
  		return true;
c7da85e3   heziqi   he,ziqi add metho...
412
413
  	}
  
724ec347   David Mayerich   simplified the EN...
414
415
416
417
418
419
420
  
  
  	///Reads a line Z (slowest dimension) for a given XY value
  
  	/// @param p is a pointer to pre-allocated memory equal to the line size R[2]
  	/// @param x is the x coordinate
  	/// @param y is the y coordinate
735a2a24   David Mayerich   started testing o...
421
  	void read_line_2( T* p, size_t n, bool PROGRESS = false){
9d3ba0b1   David Mayerich   added stim::hsi a...
422
  		unsigned long long i;
724ec347   David Mayerich   simplified the EN...
423
  
63fc1df8   David Mayerich   added an inverse ...
424
425
  		if(PROGRESS) progress = 0;
  
735a2a24   David Mayerich   started testing o...
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
  		if ( n > R[0] * R[1]){							//make sure the sample and line number is right
  			std::cout<<"ERROR: sample or line out of range in "<<__FILE__<<" (line "<<__LINE__<<")"<<std::endl;
  			exit(1);
  		}
  
  		file.seekg(n * sizeof(T), std::ios::beg);           //point to the certain sample and line
  		for (i = 0; i < R[2]; i++){							//for each band
  			file.read((char *)(p + i), sizeof(T));
  			file.seekg((R[1] * R[0] - 1) * sizeof(T), std::ios::cur);    //go to the next band
  			if(PROGRESS) progress = (double)i / (double)R[2] * 100;
  		}
  		if(PROGRESS) progress = 100;
  	}
  	void read_line_2( T * p, unsigned long long x, unsigned long long y, bool PROGRESS = false){
  		read_line_2(p, y * R[0] + x, PROGRESS);
  		/*unsigned long long i;
  
  		if(PROGRESS) progress = 0;
  
724ec347   David Mayerich   simplified the EN...
445
446
447
448
449
450
451
452
453
454
  		if ( x >= R[0] || y >= R[1]){							//make sure the sample and line number is right
  			std::cout<<"ERROR: sample or line out of range"<<std::endl;
  			return false;
  		}
  
  		file.seekg((x + y * R[0]) * sizeof(T), std::ios::beg);           //point to the certain sample and line
  		for (i = 0; i < R[2]; i++)
  		{
  			file.read((char *)(p + i), sizeof(T));
  			file.seekg((R[1] * R[0] - 1) * sizeof(T), std::ios::cur);    //go to the next band
63fc1df8   David Mayerich   added an inverse ...
455
  			if(PROGRESS) progress = (double)i / (double)R[2] * 100;
724ec347   David Mayerich   simplified the EN...
456
  		}
63fc1df8   David Mayerich   added an inverse ...
457
  		if(PROGRESS) progress = 100;
724ec347   David Mayerich   simplified the EN...
458
  
735a2a24   David Mayerich   started testing o...
459
  		return true;*/
724ec347   David Mayerich   simplified the EN...
460
461
462
463
464
465
466
  	}
  
  	///Reads a line X (fastest dimension) for a given YZ value
  
  	/// @param p is a pointer to pre-allocated memory equal to the line size R[2]
  	/// @param x is the y coordinate
  	/// @param y is the z coordinate
9d3ba0b1   David Mayerich   added stim::hsi a...
467
  	bool read_line_0(T * p, unsigned long long y, unsigned long long z, bool PROGRESS = false){
724ec347   David Mayerich   simplified the EN...
468
469
  		//test to make sure the specified value is within range
  		if( y >= R[1] || z >= R[2] ){
735a2a24   David Mayerich   started testing o...
470
  			std::cout<<"ERROR: sample ("<<y<<", "<<z<<") out of range in "<<__FILE__<<" (line "<<__LINE__<<")"<<std::endl;
724ec347   David Mayerich   simplified the EN...
471
472
473
474
475
  			return false;
  		}
  
  		file.seekg((z * R[0] * R[1] + y * R[0]) * sizeof(T), std::ios::beg);	//seek to the start of the line
  		file.read((char *)p, sizeof(T) * R[0]);									//read the line
63fc1df8   David Mayerich   added an inverse ...
476
  		if(PROGRESS) progress = 100;
724ec347   David Mayerich   simplified the EN...
477
478
479
480
481
482
483
484
  		return true;
  	}
  
  	///Reads a line Y (second fastest dimension) for a given XZ value
  
  	/// @param p is a pointer to pre-allocated memory equal to the line size R[2]
  	/// @param x is the y coordinate
  	/// @param z is the z coordinate
9d3ba0b1   David Mayerich   added stim::hsi a...
485
  	bool read_line_1(T * p, unsigned long long x, unsigned long long z, bool PROGRESS = false){
63fc1df8   David Mayerich   added an inverse ...
486
  		if(PROGRESS) progress = 0;
724ec347   David Mayerich   simplified the EN...
487
488
  		//test to make sure the specified value is within range
  		if( x >= R[0] || z >= R[2] ){
735a2a24   David Mayerich   started testing o...
489
  			std::cout<<"ERROR: sample or line out of range in "<<__FILE__<<" (line "<<__LINE__<<")"<<std::endl;
724ec347   David Mayerich   simplified the EN...
490
491
492
493
  			return false;
  		}
  
  		file.seekg((z * R[0] * R[1] + x) * sizeof(T), std::ios::beg);           //seek to the start of the line
9d3ba0b1   David Mayerich   added stim::hsi a...
494
  		for (unsigned long long i = 0; i < R[1]; i++){									//for each pixel in the line
724ec347   David Mayerich   simplified the EN...
495
496
  			file.read((char *)(p + i), sizeof(T));					//read the pixel
  			file.seekg((R[0] - 1) * sizeof(T), std::ios::cur);		//seek to the next pixel in the line
63fc1df8   David Mayerich   added an inverse ...
497
  			if(PROGRESS) progress = (double)i / (double)R[1] * 100;
724ec347   David Mayerich   simplified the EN...
498
  		}
63fc1df8   David Mayerich   added an inverse ...
499
  		if(PROGRESS) progress = 100;
724ec347   David Mayerich   simplified the EN...
500
501
502
  		return true;
  	}
  
27b826a8   David Mayerich   added a spherical...
503
504
505
506
  	/// Reads a plane given a coordinate along the 0-axis (YZ plane)
  
  	/// @param p is a pointer to pre-allocated memory of size R[1] * R[2] * sizeof(T)
  	/// @param n is the 0-axis coordinate used to retrieve the plane
9d3ba0b1   David Mayerich   added stim::hsi a...
507
  	bool read_plane_0(T* p, unsigned long long n, bool PROGRESS = false){
63fc1df8   David Mayerich   added an inverse ...
508
  		if(PROGRESS) progress = 0;
724ec347   David Mayerich   simplified the EN...
509
  		if (n >= R[0]){										//make sure the number is within the possible range
735a2a24   David Mayerich   started testing o...
510
  			std::cout<<"ERROR: sample or line out of range in "<<__FILE__<<" (line "<<__LINE__<<")"<<std::endl;
724ec347   David Mayerich   simplified the EN...
511
512
  			return false;
  		}
9d3ba0b1   David Mayerich   added stim::hsi a...
513
  		unsigned long long jump = (R[0] - 1) * sizeof(T);		//number of bytes to skip between samples
724ec347   David Mayerich   simplified the EN...
514
515
516
517
  
  		//seek to the start of the plane
  		file.seekg(n * sizeof(T), std::ios::beg);
  
9d3ba0b1   David Mayerich   added stim::hsi a...
518
519
  		unsigned long long N = R[1] * R[2];
  		for(unsigned long long i = 0; i<N; i++){
724ec347   David Mayerich   simplified the EN...
520
521
  			file.read((char*)(p+i), sizeof(T));
  			file.seekg(jump, std::ios::cur);
9d3ba0b1   David Mayerich   added stim::hsi a...
522
  			if(PROGRESS) progress = (double)(i+1) / N * 100;
724ec347   David Mayerich   simplified the EN...
523
524
  		}
  
724ec347   David Mayerich   simplified the EN...
525
526
527
528
529
  		return true;
  
  
  	}
  
27b826a8   David Mayerich   added a spherical...
530
531
532
533
  	/// Reads a plane given a coordinate along the 1-axis (XZ plane)
  
  	/// @param p is a pointer to pre-allocated memory of size R[0] * R[2] * sizeof(T)
  	/// @param n is the 1-axis coordinate used to retrieve the plane
9d3ba0b1   David Mayerich   added stim::hsi a...
534
  	bool read_plane_1(T* p, unsigned long long n, bool PROGRESS = false){
63fc1df8   David Mayerich   added an inverse ...
535
  		if(PROGRESS) progress = 0;
9d3ba0b1   David Mayerich   added stim::hsi a...
536
537
  		unsigned long long L = R[0] * sizeof(T);		//caculate the number of bytes in a sample line
  		unsigned long long jump = R[0] * (R[1] - 1) * sizeof(T);
724ec347   David Mayerich   simplified the EN...
538
539
540
541
542
543
544
  
  		if (n >= R[1]){										//make sure the bank number is right
  			std::cout<<"ERROR read_plane_1: page out of range"<<std::endl;
  			return false;
  		}
  
  		file.seekg(R[0] * n * sizeof(T), std::ios::beg);
9d3ba0b1   David Mayerich   added stim::hsi a...
545
  		for (unsigned long long i = 0; i < R[2]; i++){
63fc1df8   David Mayerich   added an inverse ...
546
  			if(PROGRESS) progress = (double)i / R[2] * 100;
724ec347   David Mayerich   simplified the EN...
547
548
549
550
551
  			file.read((char *)(p + i * R[0]), L);
  			file.seekg( jump, std::ios::cur);
  			std::cout<<i<<"    ";
  		}
  
63fc1df8   David Mayerich   added an inverse ...
552
  		if(PROGRESS) progress = 100;
724ec347   David Mayerich   simplified the EN...
553
554
555
  		return true;
  	}
  
27b826a8   David Mayerich   added a spherical...
556
557
558
559
  	/// Reads a plane given a coordinate along the 2-axis (XY plane)
  
  	/// @param p is a pointer to pre-allocated memory of size R[0] * R[1] * sizeof(T)
  	/// @param n is the 2-axis coordinate used to retrieve the plane
9d3ba0b1   David Mayerich   added stim::hsi a...
560
  	bool read_plane_2(T* p, unsigned long long n, bool PROGRESS = false){
63fc1df8   David Mayerich   added an inverse ...
561
  		return read_page(p, n, PROGRESS);
724ec347   David Mayerich   simplified the EN...
562
563
  	}
  
27b826a8   David Mayerich   added a spherical...
564
565
566
567
  	/// Reads a single pixel, treating the entire data set as a linear array
  
  	/// @param p is a pointer to pre-allocated memory of size sizeof(T)
  	/// @param i is the index to the pixel using linear indexing
9d3ba0b1   David Mayerich   added stim::hsi a...
568
  	bool read_pixel(T* p, unsigned long long i){
724ec347   David Mayerich   simplified the EN...
569
570
571
572
573
574
575
576
577
578
  		if(i >= R[0] * R[1] * R[2]){
  			std::cout<<"ERROR read_pixel: n is out of range"<<std::endl;
  			return false;
  		}
  
  		file.seekg(i * sizeof(T), std::ios::cur);
  		file.read((char*)p, sizeof(T));
  
  	}
  
27b826a8   David Mayerich   added a spherical...
579
580
581
582
583
584
  	/// Reads a single pixel, given an x, y, z coordinate
  
  	/// @param p is a pointer to pre-allocated memory of size sizeof(T)
  	/// @param x is the x (0) axis coordinate
  	/// @param y is the y (1) axis coordinate
  	/// @param z is the z (2) axis coordinate
9d3ba0b1   David Mayerich   added stim::hsi a...
585
  	bool read_pixel(T* p, unsigned long long x, unsigned long long y, unsigned long long z){
724ec347   David Mayerich   simplified the EN...
586
587
588
589
590
591
  
  		if(x < 0 || x >= R[0] || y < 0 || y >= R[1] || z < 0 || z > R[2]){
  			std::cout<<"ERROR read_pixel: (x,y,z) is out of range"<<std::endl;
  			return false;
  		}
  
9d3ba0b1   David Mayerich   added stim::hsi a...
592
  		unsigned long long i = z * R[0] * R[1] + y * R[0] + z;
724ec347   David Mayerich   simplified the EN...
593
594
595
  		return read_pixel(p, i);
  	}
  
8b899c24   David Mayerich   fixed asynchronou...
596
  	/// Reads a block specified by an (x, y, z) position and size using the largest possible contiguous reads
6db250f5   David Mayerich   adaptive streamin...
597
598
  	size_t read(T* dest, size_t x, size_t y, size_t z, size_t sx, size_t sy, size_t sz){
  		auto t0 = std::chrono::high_resolution_clock::now();
8b899c24   David Mayerich   fixed asynchronou...
599
600
601
602
603
604
605
606
607
  		size_t size_bytes = sx * sy * sz * sizeof(T);					//size of the block to read in bytes
  
  		size_t start = z * R[0] * R[1] + y * R[0] + x;						//calculate the start postion
  		size_t start_bytes = start * sizeof(T);							//start position in bytes
  		file.seekg(start * sizeof(T), std::ios::beg);					//seek to the start position
  
  		
  		if(sx == R[0] && sy == R[1]){				//if sx and sy result in a contiguous volume along z
  			file.read((char*)dest, size_bytes);			//read the block in one pass
8b899c24   David Mayerich   fixed asynchronou...
608
  		}
6db250f5   David Mayerich   adaptive streamin...
609
  		else if(sx == R[0]){												//if sx is contiguous, read each z-axis slice can be read in one pass
8b899c24   David Mayerich   fixed asynchronou...
610
611
612
613
614
615
616
  			size_t jump_bytes = (R[1] - sy) * R[0] * sizeof(T);		//jump between each slice
  			size_t slice_bytes = sx * sy * sizeof(T);				//size of the slice to be read
  			for(size_t zi = 0; zi < sz; zi++){						//for each z-axis slice
  				file.read((char*)dest, slice_bytes);						//read the slice
  				dest += sx * sy;									//move the destination pointer to the next slice
  				file.seekg(jump_bytes, std::ios::cur);				//skip to the next slice in the file
  			}
8b899c24   David Mayerich   fixed asynchronou...
617
  		}
6db250f5   David Mayerich   adaptive streamin...
618
619
620
621
622
623
624
625
626
627
628
629
630
631
  		else{
  			//in this case, x is not contiguous so the volume must be read line-by-line
  			size_t jump_x_bytes = (R[0] - sx) * sizeof(T);				//number of bytes skipped in the x direction
  			size_t jump_y_bytes = (R[1] - sy) * R[0] * sizeof(T) + jump_x_bytes;	//number of bytes skipped between slices
  			size_t line_bytes = sx * sizeof(T);							//size of the line to be read
  			size_t zi, yi;
  			for(zi = 0; zi < sz; zi++){									//for each slice
  				file.read((char*)dest, line_bytes);							//read the first line
  				for(yi = 1; yi < sy; yi++){								//read each additional line
  					dest += sx;											//move the pointer in the destination block to the next line
  					file.seekg(jump_x_bytes, std::ios::cur);			//skip to the next line in the file
  					file.read((char*)dest, line_bytes);						//read the line to the destination block
  				}
  				file.seekg(jump_y_bytes, std::ios::cur);				//skip to the beginning of the next slice
8b899c24   David Mayerich   fixed asynchronou...
632
  			}
8b899c24   David Mayerich   fixed asynchronou...
633
  		}
6db250f5   David Mayerich   adaptive streamin...
634
635
  		auto t1 = std::chrono::high_resolution_clock::now();
  		return std::chrono::duration_cast<std::chrono::milliseconds>(t1-t0).count();
8b899c24   David Mayerich   fixed asynchronou...
636
637
  	}
  
cf5b4c92   David Mayerich   simplified asynch...
638
639
  	// permutes a block of data from the current interleave to the interleave specified (re-arranged dimensions to the order specified by [d0, d1, d2])
  
6db250f5   David Mayerich   adaptive streamin...
640
641
  	size_t permute(T* dest, T* src, size_t sx, size_t sy, size_t sz, size_t d0, size_t d1, size_t d2){
  		auto t0 = std::chrono::high_resolution_clock::now();
cf5b4c92   David Mayerich   simplified asynch...
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
  		size_t d[3] = {d0, d1, d2};
  		size_t s[3] = {sx, sy, sz};
  		size_t p[3];// = {x, y, z};
  		
  		if(d[0] == 0 && d[1] == 1 && d[2] == 2){
  			//this isn't actually a permute - just copy the data
  			memcpy(dest, src, sizeof(T) * sx * sy * sz);
  		}
  		else if(d[0] == 0){						//the individual lines are contiguous, so you can memcpy line-by-line
  			size_t y, z;
  			size_t src_idx, dest_idx;
  			size_t x_bytes = sizeof(T) * sx;
  			for(z = 0; z < sz; z++){
  				p[2] = z;
  				for(y = 0; y < sy; y++){
  					p[1] = y;
  					src_idx = z * sx * sy + y * sx;
  					dest_idx = p[d[2]] * s[d[0]] * s[d[1]] + p[d[1]] * s[d[0]];
cf5b4c92   David Mayerich   simplified asynch...
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
  					memcpy(dest + dest_idx, src + src_idx, x_bytes);
  				}
  			}
  		}
  		else{									//loop through every damn point
  			size_t x, y, z;
  			size_t src_idx, dest_idx;
  			size_t src_z, src_y;
  			for(z = 0; z < sz; z++){
  				p[2] = z;
  				src_z = z * sx * sy;
  				for(y = 0; y < sy; y++){
  					p[1] = y;
  					src_y = src_z + y * sx;
  					for(x = 0; x < sx; x++){
  						p[0] = x;
  						src_idx = src_y + x;
  						dest_idx = p[d[2]] * s[d[0]] * s[d[1]] + p[d[1]] * s[d[0]] + p[d[0]];
  						dest[dest_idx] = src[src_idx];
  					}
  				}
  			}
  		}
6db250f5   David Mayerich   adaptive streamin...
683
684
  		auto t1 = std::chrono::high_resolution_clock::now();
  		return std::chrono::duration_cast<std::chrono::milliseconds>(t1-t0).count();
cf5b4c92   David Mayerich   simplified asynch...
685
686
  	}
  
180d7f3a   David Mayerich   added binary file...
687
688
689
690
691
  };
  
  }
  
  #endif