bioem_cuda.cu 27.2 KB
Newer Older
Pilar Cossio's avatar
License  
Pilar Cossio committed
1
/* ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
2
3
4
5
   < BioEM software for Bayesian inference of Electron Microscopy images>
   Copyright (C) 2016 Pilar Cossio, David Rohr, Fabio Baruffa, Markus Rampp, 
        Volker Lindenstruth and Gerhard Hummer.
   Max Planck Institute of Biophysics, Frankfurt, Germany.
6

7
   See license statement for terms of distribution.
Pilar Cossio's avatar
License  
Pilar Cossio committed
8
9
10

   ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++*/

11
12
13
14
15
16
17
18
19
20
21
#define BIOEM_GPUCODE

#if defined(_WIN32)
#include <windows.h>
#endif

#include <iostream>
using namespace std;

#include "bioem_cuda_internal.h"
#include "bioem_algorithm.h"
Pilar Cossio's avatar
Pilar Cossio committed
22
//#include "helper_cuda.h"
23

24
25
26
27
28
29
30
#define checkCudaErrors(error) \
{ \
	if ((error) != cudaSuccess) \
	{ \
		printf("CUDA Error %d / %s (%s: %d)\n", error, cudaGetErrorString(error), __FILE__, __LINE__); \
		exit(1); \
	} \
31
32
}

David Rohr's avatar
David Rohr committed
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
static const char *cufftGetErrorStrung(cufftResult error)
{
    switch (error)
    {
        case CUFFT_SUCCESS:
            return "CUFFT_SUCCESS";

        case CUFFT_INVALID_PLAN:
            return "CUFFT_INVALID_PLAN";

        case CUFFT_ALLOC_FAILED:
            return "CUFFT_ALLOC_FAILED";

        case CUFFT_INVALID_TYPE:
            return "CUFFT_INVALID_TYPE";

        case CUFFT_INVALID_VALUE:
            return "CUFFT_INVALID_VALUE";

        case CUFFT_INTERNAL_ERROR:
            return "CUFFT_INTERNAL_ERROR";

        case CUFFT_EXEC_FAILED:
            return "CUFFT_EXEC_FAILED";

        case CUFFT_SETUP_FAILED:
            return "CUFFT_SETUP_FAILED";

        case CUFFT_INVALID_SIZE:
            return "CUFFT_INVALID_SIZE";

        case CUFFT_UNALIGNED_DATA:
            return "CUFFT_UNALIGNED_DATA";
    }
    return "UNKNOWN";
}

Luka Stanisic's avatar
Luka Stanisic committed
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
/* Handing CUDA Driver errors */

#define cuErrorCheck(call) \
  do { \
    CUresult __error__; \
    if ((__error__ = (call)) != CUDA_SUCCESS) { \
      printf("CUDA Driver Error %d / %s (%s %d)\n", __error__, cuGetError(__error__),__FILE__, __LINE__); \
      return __error__; \
    } \
  } while (false)

static const char * cuGetError(CUresult result) {
  switch (result) {
    case CUDA_SUCCESS:                              return "No errors";
    case CUDA_ERROR_INVALID_VALUE:                  return "Invalid value";
    case CUDA_ERROR_OUT_OF_MEMORY:                  return "Out of memory";
    case CUDA_ERROR_NOT_INITIALIZED:                return "Driver not initialized";
    case CUDA_ERROR_DEINITIALIZED:                  return "Driver deinitialized";
    case CUDA_ERROR_PROFILER_DISABLED:              return "Profiler disabled";
    case CUDA_ERROR_PROFILER_NOT_INITIALIZED:       return "Profiler not initialized";
    case CUDA_ERROR_PROFILER_ALREADY_STARTED:       return "Profiler already started";
    case CUDA_ERROR_PROFILER_ALREADY_STOPPED:       return "Profiler already stopped";
    case CUDA_ERROR_NO_DEVICE:                      return "No CUDA-capable device available";
    case CUDA_ERROR_INVALID_DEVICE:                 return "Invalid device";
    case CUDA_ERROR_INVALID_IMAGE:                  return "Invalid kernel image";
    case CUDA_ERROR_INVALID_CONTEXT:                return "Invalid context";
    case CUDA_ERROR_CONTEXT_ALREADY_CURRENT:        return "Context already current";
    case CUDA_ERROR_MAP_FAILED:                     return "Map failed";
    case CUDA_ERROR_UNMAP_FAILED:                   return "Unmap failed";
    case CUDA_ERROR_ARRAY_IS_MAPPED:                return "Array is mapped";
    case CUDA_ERROR_ALREADY_MAPPED:                 return "Already mapped";
    case CUDA_ERROR_NO_BINARY_FOR_GPU:              return "No binary for GPU";
    case CUDA_ERROR_ALREADY_ACQUIRED:               return "Already acquired";
    case CUDA_ERROR_NOT_MAPPED:                     return "Not mapped";
    case CUDA_ERROR_NOT_MAPPED_AS_ARRAY:            return "Not mapped as array";
    case CUDA_ERROR_NOT_MAPPED_AS_POINTER:          return "Not mapped as pointer";
    case CUDA_ERROR_ECC_UNCORRECTABLE:              return "Uncorrectable ECC error";
    case CUDA_ERROR_UNSUPPORTED_LIMIT:              return "Unsupported CUlimit";
    case CUDA_ERROR_CONTEXT_ALREADY_IN_USE:         return "Context already in use";
    case CUDA_ERROR_INVALID_SOURCE:                 return "Invalid source";
    case CUDA_ERROR_FILE_NOT_FOUND:                 return "File not found";
    case CUDA_ERROR_SHARED_OBJECT_SYMBOL_NOT_FOUND: return "Shared object symbol not found";
    case CUDA_ERROR_SHARED_OBJECT_INIT_FAILED:      return "Shared object initialization failed";
    case CUDA_ERROR_OPERATING_SYSTEM:               return "Operating System call failed";
    case CUDA_ERROR_INVALID_HANDLE:                 return "Invalid handle";
    case CUDA_ERROR_NOT_FOUND:                      return "Not found";
    case CUDA_ERROR_NOT_READY:                      return "CUDA not ready";
    case CUDA_ERROR_LAUNCH_FAILED:                  return "Launch failed";
    case CUDA_ERROR_LAUNCH_OUT_OF_RESOURCES:        return "Launch exceeded resources";
    case CUDA_ERROR_LAUNCH_TIMEOUT:                 return "Launch exceeded timeout";
    case CUDA_ERROR_LAUNCH_INCOMPATIBLE_TEXTURING:  return "Launch with incompatible texturing";
    case CUDA_ERROR_PEER_ACCESS_ALREADY_ENABLED:    return "Peer access already enabled";
    case CUDA_ERROR_PEER_ACCESS_NOT_ENABLED:        return "Peer access not enabled";
    case CUDA_ERROR_PRIMARY_CONTEXT_ACTIVE:         return "Primary context active";
    case CUDA_ERROR_CONTEXT_IS_DESTROYED:           return "Context is destroyed";
    case CUDA_ERROR_ASSERT:                         return "Device assert failed";
    case CUDA_ERROR_TOO_MANY_PEERS:                 return "Too many peers";
    case CUDA_ERROR_HOST_MEMORY_ALREADY_REGISTERED: return "Host memory already registered";
    case CUDA_ERROR_HOST_MEMORY_NOT_REGISTERED:     return "Host memory not registered";
    case CUDA_ERROR_UNKNOWN:                        return "Unknown error";
    default:                                        return "Unknown error code";
  }
}

134
135
136
137
bioem_cuda::bioem_cuda()
{
	deviceInitialized = 0;
	GPUAlgo = getenv("GPUALGO") == NULL ? 2 : atoi(getenv("GPUALGO"));
138
139
	GPUAsync = getenv("GPUASYNC") == NULL ? 1 : atoi(getenv("GPUASYNC"));
	GPUWorkload = getenv("GPUWORKLOAD") == NULL ? 100 : atoi(getenv("GPUWORKLOAD"));
Luka Stanisic's avatar
Luka Stanisic committed
140
	if (GPUWorkload == -1) GPUWorkload = 100;
141
	GPUDualStream = getenv("GPUDUALSTREAM") == NULL ? 1 : atoi(getenv("GPUDUALSTREAM"));
142
143
144
145
146
147
148
}

bioem_cuda::~bioem_cuda()
{
	deviceExit();
}

149
150
151
__global__ void compareRefMap_kernel(const int iOrient, const int iConv,  const myfloat_t amp, const myfloat_t pha, const myfloat_t env, const myfloat_t sumC,
                                                const myfloat_t sumsquareC, const myfloat_t* pMap, bioem_Probability pProb, 
						const bioem_param_device param, const bioem_RefMap_Mod RefMap, const int cent_x, const int cent_y, const int maxRef)
152
153
{
	const int iRefMap = myBlockIdxX * myBlockDimX + myThreadIdxX;
154
	if (iRefMap < maxRef)
155
	{
156
		compareRefMap<0>(iRefMap, iOrient, iConv, amp, pha, env, sumC, sumsquareC, pMap, pProb, param, RefMap, cent_x, cent_y);
157
158
159
	}
}

Pilar Cossio's avatar
Pilar Cossio committed
160
__global__ void compareRefMapShifted_kernel(const int iOrient, const int iConv, const myfloat_t amp, const myfloat_t pha, const myfloat_t env, const myfloat_t sumC, const myfloat_t sumsquareC, const myfloat_t* pMap, bioem_Probability pProb, const bioem_param_device param, const bioem_RefMap_Mod RefMap, const int maxRef)
161
162
{
	const int iRefMap = myBlockIdxX * myBlockDimX + myThreadIdxX;
163
	if (iRefMap < maxRef)
164
	{
165
		compareRefMapShifted<1>(iRefMap, iOrient, iConv, amp, pha, env, sumC, sumsquareC, pMap, pProb, param, RefMap);
166
167
168
	}
}

169
170
171
172
173
174
__global__ void cudaZeroMem(void* ptr, size_t size)
{
	int* myptr = (int*) ptr;
	int mysize = size / sizeof(int);
	int myid = myBlockDimX * myBlockIdxX + myThreadIdxX;
	int mygrid = myBlockDimX * myGridDimX;
175
	for (int i = myid; i < mysize; i += mygrid) myptr[i] = 0;
176
177
}

Pilar Cossio's avatar
Pilar Cossio committed
178
__global__ void compareRefMapLoopShifts_kernel(const int iOrient, const int iConv, const myfloat_t amp, const myfloat_t pha, const myfloat_t env, const myfloat_t sumC, const myfloat_t sumsquareC, const myfloat_t* pMap, bioem_Probability pProb, const bioem_param_device param, const bioem_RefMap RefMap, const int blockoffset, const int nShifts, const int nShiftBits, const int maxRef)
179
180
181
182
183
184
185
186
187
{
	const size_t myid = (myBlockIdxX + blockoffset) * myBlockDimX + myThreadIdxX;
	const int iRefMap = myid >> (nShiftBits << 1);
	const int myRef = myThreadIdxX >> (nShiftBits << 1);
	const int myShiftIdx = (myid >> nShiftBits) & (nShifts - 1);
	const int myShiftIdy = myid & (nShifts - 1);
	const int myShift = myid & (nShifts * nShifts - 1);
	const int cent_x = myShiftIdx * param.GridSpaceCenter - param.maxDisplaceCenter;
	const int cent_y = myShiftIdy * param.GridSpaceCenter - param.maxDisplaceCenter;
188

189
	const bool threadActive = myShiftIdx < nShifts && myShiftIdy < nShifts && iRefMap < maxRef;
190

Pilar Cossio's avatar
Pilar Cossio committed
191
	compareRefMap<2>(iRefMap, iOrient, iConv, amp, pha, env, sumC, sumsquareC, pMap, pProb, param, RefMap, cent_x, cent_y, myShift, nShifts * nShifts, myRef, threadActive);
192
193
}

194
__global__ void multComplexMap(const mycomplex_t* convmap, const mycomplex_t* refmap, mycuComplex_t* out, const int NumberPixelsTotal, const int MapSize, const int NumberMaps, const int Offset)
195
196
{
	if (myBlockIdxX >= NumberMaps) return;
197
	const mycuComplex_t* myin = (mycuComplex_t*) &refmap[(myBlockIdxX + Offset) * MapSize];
198
	const mycuComplex_t* myconv = (mycuComplex_t*) convmap;
199
	mycuComplex_t* myout = &out[myBlockIdxX * MapSize];
200
	for(int i = myThreadIdxX; i < NumberPixelsTotal; i += myBlockDimX)
201
	{
202
203
204
205
206
207
208
		mycuComplex_t val;
		const mycuComplex_t conv = myconv[i];
		const mycuComplex_t in = myin[i];

		val.x = conv.x * in.x + conv.y * in.y;
		val.y = conv.y * in.x - conv.x * in.y;
		myout[i] = val;
209
210
211
	}
}

212
__global__ void cuDoRefMapsFFT(const int iOrient, const int iConv, const myfloat_t amp, const myfloat_t pha, const myfloat_t env, const myfloat_t* lCC, const myfloat_t sumC, const myfloat_t sumsquareC, bioem_Probability pProb, const bioem_param_device param, const bioem_RefMap RefMap, const int maxRef, const int Offset)
213
{
214
	if (myBlockIdxX * myBlockDimX + myThreadIdxX >= maxRef) return;
215
216
	const int iRefMap = myBlockIdxX * myBlockDimX + myThreadIdxX + Offset;
	const myfloat_t* mylCC = &lCC[(myBlockIdxX * myBlockDimX + myThreadIdxX) * param.NumberPixels * param.NumberPixels];
217
	doRefMapFFT(iRefMap, iOrient, iConv, amp, pha, env, mylCC, sumC, sumsquareC, pProb, param, RefMap);
218
219
}

220
221
222
223
224
225
226
227
228
229
230
231
232
template <class T> static inline T divup(T num, T divider) {return((num + divider - 1) / divider);}
static inline bool IsPowerOf2(int x) {return ((x > 0) && ((x & (x - 1)) == 0));}
#if defined(_WIN32)
static inline int ilog2 (int value)
{
	DWORD index;
	_BitScanReverse (&index, value);
	return(value);
}
#else
static inline int ilog2(int value) {return 31 - __builtin_clz(value);}
#endif

233
int bioem_cuda::compareRefMaps(int iOrient, int iConv, myfloat_t amp, myfloat_t pha, myfloat_t env, const myfloat_t* conv_map, mycomplex_t* localmultFFT, myfloat_t sumC, myfloat_t sumsquareC, const int startMap)
234
{
235
236
237
238
239
	if (startMap)
	{
		cout << "Error startMap not implemented for GPU Code\n";
		exit(1);
	}
Luka Stanisic's avatar
Luka Stanisic committed
240
241
242
243
244
245
246
#ifdef DEBUG_GPU
	float time;
	cudaEvent_t start, stop;
	checkCudaErrors(cudaEventCreate(&start));
	checkCudaErrors(cudaEventCreate(&stop));
	checkCudaErrors(cudaEventRecord(start, 0));
#endif
247
248
249
250
	if (GPUAsync)
	{
		checkCudaErrors(cudaEventSynchronize(cudaEvent[iConv & 1]));
	}
Luka Stanisic's avatar
Luka Stanisic committed
251
252
253
254
255
256
257
#ifdef DEBUG_GPU
	checkCudaErrors(cudaEventRecord(stop, 0));
	checkCudaErrors(cudaEventSynchronize(stop));
	checkCudaErrors(cudaEventElapsedTime(&time, start, stop));
	printf("\t\t\tGPU: time to synch projections %1.6f sec\n", time/1000);
	checkCudaErrors(cudaEventRecord(start, 0));
#endif
258
	if (FFTAlgo)
259
	{
260
		memcpy(&pConvMapFFT_Host[(iConv & 1) * param.FFTMapSize], localmultFFT, param.FFTMapSize * sizeof(mycomplex_t));
261
		checkCudaErrors(cudaMemcpyAsync(&pConvMapFFT[(iConv & 1) * param.FFTMapSize], &pConvMapFFT_Host[(iConv & 1) * param.FFTMapSize], param.FFTMapSize * sizeof(mycomplex_t), cudaMemcpyHostToDevice, cudaStream[GPUAsync ? 2 : 0]));
Luka Stanisic's avatar
Luka Stanisic committed
262
263
264
265
266
267
268
#ifdef DEBUG_GPU
		checkCudaErrors(cudaEventRecord(stop, 0));
		checkCudaErrors(cudaEventSynchronize(stop));
		checkCudaErrors(cudaEventElapsedTime(&time, start, stop));
		printf("\t\t\tGPU: time for memcpy %1.6f sec\n", time/1000);
		checkCudaErrors(cudaEventRecord(start, 0));
#endif
269
270
271
272
273
		if (GPUAsync)
		{
			checkCudaErrors(cudaEventRecord(cudaEvent[2], cudaStream[2]));
			checkCudaErrors(cudaStreamWaitEvent(cudaStream[0], cudaEvent[2], 0));
		}
274
		if (GPUDualStream)
275
		{
276
277
278
279
280
281
			checkCudaErrors(cudaEventRecord(cudaFFTEvent[0], cudaStream[0]));
			checkCudaErrors(cudaStreamWaitEvent(cudaStream[1], cudaFFTEvent[0], 0));
		}
		for (int i = 0, j = 0; i < maxRef; i += CUDA_FFTS_AT_ONCE, j++)
		{
			if (!GPUDualStream) j = 0;
282
			const int num = min(CUDA_FFTS_AT_ONCE, maxRef - i);
283
284
			multComplexMap<<<num, CUDA_THREAD_COUNT, 0, cudaStream[j & 1]>>>(&pConvMapFFT[(iConv & 1) * param.FFTMapSize], pRefMapsFFT, pFFTtmp2[j & 1], param.param_device.NumberPixels * param.param_device.NumberFFTPixels1D, param.FFTMapSize, num, i);
			cufftResult err = mycufftExecC2R(i + CUDA_FFTS_AT_ONCE > maxRef ? plan[1][j & 1] : plan[0][j & 1], pFFTtmp2[j & 1], pFFTtmp[j & 1]);
David Rohr's avatar
David Rohr committed
285
			if (err != CUFFT_SUCCESS)
286
			{
David Rohr's avatar
David Rohr committed
287
				cout << "Error running CUFFT " << cufftGetErrorStrung(err) << "\n";
288
289
				exit(1);
			}
290
			cuDoRefMapsFFT<<<divup(num, CUDA_THREAD_COUNT), CUDA_THREAD_COUNT, 0, cudaStream[j & 1]>>>(iOrient, iConv,  amp, pha, env, pFFTtmp[j & 1], sumC, sumsquareC, pProb_device, param.param_device, *gpumap, num, i);
291
		}
Luka Stanisic's avatar
Luka Stanisic committed
292
		checkCudaErrors(cudaPeekAtLastError());
293
294
295
296
297
		if (GPUDualStream)
		{
			checkCudaErrors(cudaEventRecord(cudaFFTEvent[1], cudaStream[1]));
			checkCudaErrors(cudaStreamWaitEvent(cudaStream[0], cudaFFTEvent[1], 0));
		}
298
299
300
	}
	else
	{
301
		checkCudaErrors(cudaMemcpyAsync(pConvMap_device[iConv & 1], conv_map, sizeof(myfloat_t) * RefMap.refMapSize, cudaMemcpyHostToDevice, cudaStream[0]));
Luka Stanisic's avatar
Luka Stanisic committed
302
303
304
305
306
307
308
#ifdef DEBUG_GPU
		checkCudaErrors(cudaEventRecord(stop, 0));
		checkCudaErrors(cudaEventSynchronize(stop));
		checkCudaErrors(cudaEventElapsedTime(&time, start, stop));
		printf("\t\t\tGPU: time for memcpy %1.6f sec\n", time/1000);
		checkCudaErrors(cudaEventRecord(start, 0) );
#endif
309
		if (GPUAlgo == 2) //Loop over shifts
310
		{
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
			const int nShifts = 2 * param.param_device.maxDisplaceCenter / param.param_device.GridSpaceCenter + 1;
			if (!IsPowerOf2(nShifts))
			{
				cout << "Invalid number of displacements, no power of two\n";
				exit(1);
			}
			if (CUDA_THREAD_COUNT % (nShifts * nShifts))
			{
				cout << "CUDA Thread count (" << CUDA_THREAD_COUNT << ") is no multiple of number of shifts (" << (nShifts * nShifts) << ")\n";
				exit(1);
			}
			if (nShifts > CUDA_MAX_SHIFT_REDUCE)
			{
				cout << "Too many displacements for CUDA reduction\n";
				exit(1);
			}
			const int nShiftBits = ilog2(nShifts);
			size_t totalBlocks = divup((size_t) maxRef * (size_t) nShifts * (size_t) nShifts, (size_t) CUDA_THREAD_COUNT);
			size_t nBlocks = CUDA_BLOCK_COUNT;
330
			for (size_t i = 0; i < totalBlocks; i += nBlocks)
331
			{
Pilar Cossio's avatar
Pilar Cossio committed
332
				compareRefMapLoopShifts_kernel<<<min(nBlocks, totalBlocks - i), CUDA_THREAD_COUNT, (CUDA_THREAD_COUNT * 2 + CUDA_THREAD_COUNT / (nShifts * nShifts) * 4) * sizeof(myfloat_t), cudaStream[0] >>> (iOrient, iConv, amp, pha, env, sumC, sumsquareC, pConvMap_device[iConv & 1], pProb_device, param.param_device, *gpumap, i, nShifts, nShiftBits, maxRef);
333
			}
334
		}
335
		else if (GPUAlgo == 1) //Split shifts in multiple kernels
336
		{
337
			for (int cent_x = -param.param_device.maxDisplaceCenter; cent_x <= param.param_device.maxDisplaceCenter; cent_x = cent_x + param.param_device.GridSpaceCenter)
338
			{
339
				for (int cent_y = -param.param_device.maxDisplaceCenter; cent_y <= param.param_device.maxDisplaceCenter; cent_y = cent_y + param.param_device.GridSpaceCenter)
340
				{
Pilar Cossio's avatar
Pilar Cossio committed
341
					compareRefMap_kernel<<<divup(maxRef, CUDA_THREAD_COUNT), CUDA_THREAD_COUNT, 0, cudaStream[0]>>> (iOrient, iConv, amp, pha, env, sumC, sumsquareC, pConvMap_device[iConv & 1], pProb_device, param.param_device, *pRefMap_device_Mod, cent_x, cent_y, maxRef);
342
343
				}
			}
344
		}
345
		else if (GPUAlgo == 0) //All shifts in one kernel
346
		{
347
			compareRefMapShifted_kernel<<<divup(maxRef, CUDA_THREAD_COUNT), CUDA_THREAD_COUNT, 0, cudaStream[0]>>> (iOrient, iConv, amp, pha, env, sumC, sumsquareC, pConvMap_device[iConv & 1], pProb_device, param.param_device, *pRefMap_device_Mod, maxRef);
348
		}
349
		else
350
		{
351
352
			cout << "Invalid GPU Algorithm selected\n";
			exit(1);
353
		}
354
	}
Luka Stanisic's avatar
Luka Stanisic committed
355
356
357
358
359
360
361
#ifdef DEBUG_GPU
	checkCudaErrors(cudaEventRecord(stop, 0));
	checkCudaErrors(cudaEventSynchronize(stop));
	checkCudaErrors(cudaEventElapsedTime(&time, start, stop));
	printf("\t\t\tGPU: time to run CUDA %1.6f sec\n", time/1000);
	checkCudaErrors(cudaEventRecord(start, 0));
#endif
362
363
	if (GPUWorkload < 100)
	{
364
		bioem::compareRefMaps(iOrient, iConv, amp, pha, env, conv_map, localmultFFT, sumC, sumsquareC, maxRef);
365
	}
Luka Stanisic's avatar
Luka Stanisic committed
366
367
368
369
370
371
#ifdef DEBUG_GPU
	checkCudaErrors(cudaEventRecord(stop, 0));
	checkCudaErrors(cudaEventSynchronize(stop));
	checkCudaErrors(cudaEventElapsedTime(&time, start, stop));
	printf("\t\t\tGPU: time to run OMP %1.6f sec\n", time/1000);
#endif
372
373
	if (GPUAsync)
	{
374
		checkCudaErrors(cudaEventRecord(cudaEvent[iConv & 1], cudaStream[0]));
375
	}
376
377
	else
	{
378
		checkCudaErrors(cudaStreamSynchronize(cudaStream[0]));
379
380
381
382
	}
	return(0);
}

David Rohr's avatar
David Rohr committed
383
384
385
int bioem_cuda::selectCudaDevice()
{
	int count;
386
	int bestDevice = 0;
David Rohr's avatar
David Rohr committed
387
	cudaDeviceProp deviceProp;
388
389
390
391
392

	/* Initializing CUDA driver API */
	cuErrorCheck(cuInit(0));

	/* Get number of available CUDA devices */
David Rohr's avatar
David Rohr committed
393
394
395
396
397
398
	checkCudaErrors(cudaGetDeviceCount(&count));
	if (count == 0)
	{
		printf("No CUDA device detected\n");
		return(1);
	}
399
400
401
402

	/* Find the best GPU */
	long long int bestDeviceSpeed = -1, deviceSpeed = -1;
	for (int i = 0; i < count; i++)
David Rohr's avatar
David Rohr committed
403
	{
404
405
		cudaGetDeviceProperties(&deviceProp, i);
		deviceSpeed = (long long int) deviceProp.multiProcessorCount * (long long int) deviceProp.clockRate * (long long int) deviceProp.warpSize;
David Rohr's avatar
David Rohr committed
406
407
408
409
410
411
		if (deviceSpeed > bestDeviceSpeed)
		{
			bestDevice = i;
			bestDeviceSpeed = deviceSpeed;
		}
	}
412
413

	/* Get user-specified GPU choice */
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
	if (getenv("GPUDEVICE"))
	{
		int device = atoi(getenv("GPUDEVICE"));
		if (device > count)
		{
			printf("Invalid CUDA device specified, max device number is %d\n", count);
			exit(1);
		}
#ifdef WITH_MPI
		if (device == -1)
		{
			device = mpi_rank % count;
		}
#endif
		if (device < 0)
		{
			printf("Negative CUDA device specified: %d, invalid!\n", device);
431
			exit(1);
432
433
434
		}
		bestDevice = device;
	}
David Rohr's avatar
David Rohr committed
435

436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
	/* Set CUDA processes to appropriate devices */
	cudaGetDeviceProperties(&deviceProp, bestDevice);
	if (deviceProp.computeMode == 0)
	{
		checkCudaErrors(cudaSetDevice(bestDevice));
	}
	else
	{
		if (DebugOutput >= 1)
		{
			printf("CUDA device %d is not set in DEFAULT mode, make sure that CUDA processes are pinned as planned!\n", bestDevice);
			printf("Pinning process %d to CUDA device %d\n", mpi_rank, bestDevice);
		}
		checkCudaErrors(cudaSetDevice(bestDevice));
		/* This synchronization is needed in order to detect bogus silent errors from cudaSetDevice call */
		checkCudaErrors(cudaDeviceSynchronize());
	}
David Rohr's avatar
David Rohr committed
453

454
	/* Debugging information about CUDA devices used by the current process */
David Rohr's avatar
David Rohr committed
455
	if (DebugOutput >= 3)
David Rohr's avatar
David Rohr committed
456
	{
David Rohr's avatar
David Rohr committed
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
		printf("Using CUDA Device %s with Properties:\n", deviceProp.name);
		printf("totalGlobalMem = %lld\n", (unsigned long long int) deviceProp.totalGlobalMem);
		printf("sharedMemPerBlock = %lld\n", (unsigned long long int) deviceProp.sharedMemPerBlock);
		printf("regsPerBlock = %d\n", deviceProp.regsPerBlock);
		printf("warpSize = %d\n", deviceProp.warpSize);
		printf("memPitch = %lld\n", (unsigned long long int) deviceProp.memPitch);
		printf("maxThreadsPerBlock = %d\n", deviceProp.maxThreadsPerBlock);
		printf("maxThreadsDim = %d %d %d\n", deviceProp.maxThreadsDim[0], deviceProp.maxThreadsDim[1], deviceProp.maxThreadsDim[2]);
		printf("maxGridSize = %d %d %d\n", deviceProp.maxGridSize[0], deviceProp.maxGridSize[1], deviceProp.maxGridSize[2]);
		printf("totalConstMem = %lld\n", (unsigned long long int) deviceProp.totalConstMem);
		printf("major = %d\n", deviceProp.major);
		printf("minor = %d\n", deviceProp.minor);
		printf("clockRate = %d\n", deviceProp.clockRate);
		printf("memoryClockRate = %d\n", deviceProp.memoryClockRate);
		printf("multiProcessorCount = %d\n", deviceProp.multiProcessorCount);
		printf("textureAlignment = %lld\n", (unsigned long long int) deviceProp.textureAlignment);
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
		printf("computeMode = %d\n", deviceProp.computeMode);
#if CUDA_VERSION > 3010
		size_t free, total;
#else
		unsigned int free, total;
#endif
		if (deviceProp.computeMode == 0)
		{
			CUdevice tmpDevice;
			cuErrorCheck(cuDeviceGet(&tmpDevice, bestDevice));
			CUcontext tmpContext;
			cuErrorCheck(cuCtxCreate(&tmpContext, 0, tmpDevice));
			cuErrorCheck(cuMemGetInfo(&free, &total));
			cuErrorCheck(cuCtxDestroy(tmpContext));
		}
		else
		{
			cuErrorCheck(cuMemGetInfo(&free, &total));
		}
		printf("free memory = %lld; total memory = %lld\n", free, total);
David Rohr's avatar
David Rohr committed
493
	}
494

David Rohr's avatar
David Rohr committed
495
496
	if (DebugOutput >= 1)
	{
David Rohr's avatar
David Rohr committed
497
		printf("BioEM for CUDA initialized (MPI Rank %d), %d GPUs found, using GPU %d\n", mpi_rank, count, bestDevice);
David Rohr's avatar
David Rohr committed
498
	}
499

David Rohr's avatar
David Rohr committed
500
501
502
	return(0);
}

503
504
505
int bioem_cuda::deviceInit()
{
	deviceExit();
David Rohr's avatar
David Rohr committed
506
	
507
	selectCudaDevice();
508

509
510
	if (FFTAlgo) GPUAlgo = 2;

511
512
513
514
515
	gpumap = new bioem_RefMap;
	memcpy(gpumap, &RefMap, sizeof(bioem_RefMap));
	if (FFTAlgo == 0)
	{
		checkCudaErrors(cudaMalloc(&maps, sizeof(myfloat_t) * RefMap.ntotRefMap * RefMap.refMapSize));
516
517
518
519
520
521
522
523
524
525
526
527
528

		if (GPUAlgo == 0 || GPUAlgo == 1)
		{
			pRefMap_device_Mod = (bioem_RefMap_Mod*) gpumap;
			bioem_RefMap_Mod* RefMapGPU = new bioem_RefMap_Mod;
			RefMapGPU->init(RefMap);
			checkCudaErrors(cudaMemcpy(maps, RefMapGPU->maps, sizeof(myfloat_t) * RefMap.ntotRefMap * RefMap.refMapSize, cudaMemcpyHostToDevice));
			delete RefMapGPU;
		}
		else
		{
			checkCudaErrors(cudaMemcpy(maps, RefMap.maps, sizeof(myfloat_t) * RefMap.ntotRefMap * RefMap.refMapSize, cudaMemcpyHostToDevice));
		}
529
530
531
532
533
534
535
536
537
	}
	checkCudaErrors(cudaMalloc(&sum, sizeof(myfloat_t) * RefMap.ntotRefMap));
	checkCudaErrors(cudaMemcpy(sum, RefMap.sum_RefMap, sizeof(myfloat_t) * RefMap.ntotRefMap, cudaMemcpyHostToDevice));
	checkCudaErrors(cudaMalloc(&sumsquare, sizeof(myfloat_t) * RefMap.ntotRefMap));
	checkCudaErrors(cudaMemcpy(sumsquare, RefMap.sumsquare_RefMap, sizeof(myfloat_t) * RefMap.ntotRefMap, cudaMemcpyHostToDevice));
	gpumap->maps = maps;
	gpumap->sum_RefMap = sum;
	gpumap->sumsquare_RefMap = sumsquare;

538
	checkCudaErrors(cudaMalloc(&pProb_memory, pProb_device.get_size(RefMap.ntotRefMap, param.nTotGridAngles, param.nTotCC, param.param_device.writeAngles, param.param_device.writeCC)));
539

540
	for (int i = 0; i < 2; i++)
541
	{
542
		checkCudaErrors(cudaStreamCreate(&cudaStream[i]));
543
		checkCudaErrors(cudaEventCreate(&cudaEvent[i]));
544
		checkCudaErrors(cudaEventCreate(&cudaFFTEvent[i]));
545
		checkCudaErrors(cudaMalloc(&pConvMap_device[i], sizeof(myfloat_t) * RefMap.refMapSize));
546
	}
547
548
549
550
551
	if (GPUAsync)
	{
		checkCudaErrors(cudaStreamCreate(&cudaStream[2]));
		checkCudaErrors(cudaEventCreate(&cudaEvent[2]));
	}
552

553
554
	if (FFTAlgo)
	{
555
		checkCudaErrors(cudaMalloc(&pRefMapsFFT, RefMap.ntotRefMap * param.FFTMapSize * sizeof(mycomplex_t)));
556
557
558
559
		checkCudaErrors(cudaMalloc(&pFFTtmp2[0], CUDA_FFTS_AT_ONCE * param.FFTMapSize * 2 * sizeof(mycomplex_t)));
		checkCudaErrors(cudaMalloc(&pFFTtmp[0], CUDA_FFTS_AT_ONCE * param.param_device.NumberPixels * param.param_device.NumberPixels * 2 * sizeof(myfloat_t)));
		pFFTtmp2[1] = pFFTtmp2[0] + CUDA_FFTS_AT_ONCE * param.FFTMapSize;
		pFFTtmp[1] = pFFTtmp[0] + CUDA_FFTS_AT_ONCE * param.param_device.NumberPixels * param.param_device.NumberPixels;
560
		checkCudaErrors(cudaMalloc(&pConvMapFFT, param.FFTMapSize * sizeof(mycomplex_t) * 2));
561
		checkCudaErrors(cudaHostAlloc(&pConvMapFFT_Host, param.FFTMapSize * sizeof(mycomplex_t) * 2, 0));
562
		checkCudaErrors(cudaMemcpy(pRefMapsFFT, RefMap.RefMapsFFT, RefMap.ntotRefMap * param.FFTMapSize * sizeof(mycomplex_t), cudaMemcpyHostToDevice));
563
564
	}

565
566
567
568
569
570
571
	deviceInitialized = 1;
	return(0);
}

int bioem_cuda::deviceExit()
{
	if (deviceInitialized == 0) return(0);
572

573

David Rohr's avatar
David Rohr committed
574
	cudaFree(pProb_memory);
575
576
	cudaFree(sum);
	cudaFree(sumsquare);
577
	for (int i = 0; i < 2; i++)
578
	{
579
		cudaStreamDestroy(cudaStream[i]);
580
		cudaEventDestroy(cudaEvent[i]);
581
		cudaEventDestroy(cudaFFTEvent[i]);
582
		cudaFree(pConvMap_device[i]);
583
	}
584
585
586
587
	if (FFTAlgo)
	{
		cudaFree(pRefMapsFFT);
		cudaFree(pConvMapFFT);
588
		cudaFreeHost(pConvMapFFT_Host);
589
590
		cudaFree(pFFTtmp[0]);
		cudaFree(pFFTtmp2[0]);
591
	}
592
593
594
595
596
597
598
599
	else
	{
		cudaFree(maps);
	}
	if (GPUAlgo == 0 || GPUAlgo == 1)
	{
		cudaFree(pRefMap_device_Mod);
	}
600
601
602
603
604
605
	if (GPUAsync)
	{
		cudaStreamDestroy(cudaStream[2]);
		cudaEventDestroy(cudaEvent[2]);
	}

606
	delete gpumap;
607
	cudaThreadExit();
608

609
610
611
612
613
614
	deviceInitialized = 0;
	return(0);
}

int bioem_cuda::deviceStartRun()
{
David Rohr's avatar
David Rohr committed
615
616
617
618
619
620
621
	if (GPUWorkload >= 100)
	{
		maxRef = RefMap.ntotRefMap;
		pProb_host = &pProb;
	}
	else
	{
622
		maxRef = RefMap.ntotRefMap == 1 ? (size_t) RefMap.ntotRefMap : (size_t) RefMap.ntotRefMap * (size_t) GPUWorkload / 100;
David Rohr's avatar
David Rohr committed
623
		pProb_host = new bioem_Probability;
624
		pProb_host->init(maxRef, param.nTotGridAngles, param.nTotCC, *this);
David Rohr's avatar
David Rohr committed
625
626
		pProb_host->copyFrom(&pProb, *this);
	}
627

David Rohr's avatar
David Rohr committed
628
629
630
	pProb_device = *pProb_host;
	pProb_device.ptr = pProb_memory;
	pProb_device.set_pointers();
631
	checkCudaErrors(cudaMemcpyAsync(pProb_device.ptr, pProb_host->ptr, pProb_host->get_size(maxRef, param.nTotGridAngles, param.nTotCC, param.param_device.writeAngles, param.param_device.writeCC), cudaMemcpyHostToDevice, cudaStream[0]));
632
633
634

	if (FFTAlgo)
	{
635
		for (int j = 0;j < 2;j++)
636
		{
637
			for (int i = 0; i < 2; i++)
638
			{
639
				if (i && maxRef % CUDA_FFTS_AT_ONCE == 0) continue;
640
641
642
643
644
645
				int n[2] = {param.param_device.NumberPixels, param.param_device.NumberPixels};
				if (cufftPlanMany(&plan[i][j], 2, n, NULL, 1, param.FFTMapSize, NULL, 1, 0, MY_CUFFT_C2R, i ? (maxRef % CUDA_FFTS_AT_ONCE) : CUDA_FFTS_AT_ONCE) != CUFFT_SUCCESS)
				{
					cout << "Error planning CUFFT\n";
					exit(1);
				}
646
			        if (cufftSetCompatibilityMode(plan[i][j], CUFFT_COMPATIBILITY_FFTW_PADDING) != CUFFT_SUCCESS)
647
648
649
650
651
652
653
654
655
				{
					cout << "Error planning CUFFT compatibility\n";
					exit(1);
				}
				if (cufftSetStream(plan[i][j], cudaStream[j]) != CUFFT_SUCCESS)
				{
					cout << "Error setting CUFFT stream\n";
					exit(1);
				}
656
			}
657
			if (!GPUDualStream) break;
658
659
		}
	}
660
661
662
663
664
	return(0);
}

int bioem_cuda::deviceFinishRun()
{
665
	if (GPUAsync) cudaStreamSynchronize(cudaStream[0]);
666
	checkCudaErrors(cudaMemcpyAsync(pProb_host->ptr, pProb_device.ptr, pProb_host->get_size(maxRef, param.nTotGridAngles, param.nTotCC, param.param_device.writeAngles, param.param_device.writeCC), cudaMemcpyDeviceToHost, cudaStream[0]));
667

668
669
	if (FFTAlgo)
	{
670
671
		for (int j = 0;j < 2;j++)
		{
672
673
674
675
676
			for (int i = 0; i < 2; i++)
			{
				if (i && maxRef % CUDA_FFTS_AT_ONCE == 0) continue;
				cufftDestroy(plan[i][j]);
			}
677
678
			if (!GPUDualStream) break;
		}
679
	}
David Rohr's avatar
David Rohr committed
680
681
682
683
	cudaThreadSynchronize();
	if (GPUWorkload < 100)
	{
		pProb.copyFrom(pProb_host, *this);
684
		free_device_host(pProb_host->ptr);
David Rohr's avatar
David Rohr committed
685
686
		delete[] pProb_host;
	}
687

688
689
690
	return(0);
}

691
692
693
694
695
696
697
698
699
700
701
702
void* bioem_cuda::malloc_device_host(size_t size)
{
	void* ptr;
	checkCudaErrors(cudaHostAlloc(&ptr, size, 0));
	return(ptr);
}

void bioem_cuda::free_device_host(void* ptr)
{
	cudaFreeHost(ptr);
}

Luka Stanisic's avatar
Luka Stanisic committed
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
void bioem_cuda::rebalance(int workload)
{
	if ((workload < 0) || (workload > 100) || (workload == GPUWorkload)) return;

	deviceFinishRun();

	if (DebugOutput >= 2)
	{
	  printf("\t\tSetting GPU workload to %d%% (rank %d)\n", workload, mpi_rank);
	}

	GPUWorkload = workload;
	maxRef = (size_t) RefMap.ntotRefMap * (size_t) GPUWorkload / 100;

	deviceStartRun();
}

720
721
bioem* bioem_cuda_create()
{
David Rohr's avatar
David Rohr committed
722
723
724
725
726
727
728
729
730
	int count;
	
	if (cudaGetDeviceCount(&count) != cudaSuccess) count = 0;
	if (count == 0)
	{
		printf("No CUDA device available, using fallback to CPU version\n");
		return new bioem;
	}

731
732
	return new bioem_cuda;
}