performdocking.cpp 54.2 KB
Newer Older
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
/*

OCLADock, an OpenCL implementation of AutoDock 4.2 running a Lamarckian Genetic Algorithm
Copyright (C) 2017 TU Darmstadt, Embedded Systems and Applications Group, Germany. All rights reserved.

AutoDock is a Trade Mark of the Scripps Research Institute.

This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License
as published by the Free Software Foundation; either version 2
of the License, or (at your option) any later version.

This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
GNU General Public License for more details.

You should have received a copy of the GNU General Public License
along with this program; if not, write to the Free Software
Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301, USA.

*/


25
26
27
28
29
30
31
32
33
34
#ifndef _WIN32
#define STRINGIZE2(s) #s
#define STRINGIZE(s)	STRINGIZE2(s)
#define KRNL_FILE STRINGIZE(KRNL_SOURCE)
#define KRNL_FOLDER STRINGIZE(KRNL_DIRECTORY)
#define KRNL_COMMON STRINGIZE(KCMN_DIRECTORY)
#define KRNL1 STRINGIZE(K1)
#define KRNL2 STRINGIZE(K2)
#define KRNL3 STRINGIZE(K3)
#define KRNL4 STRINGIZE(K4)
35
#define KRNL5 STRINGIZE(K5)
36
37
38
39
40
41
42
43
44

#else
#define KRNL_FILE KRNL_SOURCE
#define KRNL_FOLDER KRNL_DIRECTORY
#define KRNL_COMMON KCMN_DIRECTORY
#define KRNL1 K1
#define KRNL2 K2
#define KRNL3 K3
#define KRNL4 K4
45
#define KRNL5 K5
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
#endif

#define INC " -I " KRNL_FOLDER " -I " KRNL_COMMON

#if defined (N16WI)
	#define KNWI " -DN16WI "
#elif defined (N32WI)
	#define KNWI " -DN32WI "
#elif defined (N64WI)
	#define KNWI " -DN64WI "
#elif defined (N128WI)
	#define KNWI " -DN128WI "
#elif defined (N256WI)
		#define KNWI " -DN256WI "
#else
	#define KNWI	" -DN64WI "
#endif

#if defined (REPRO)
	#define REP " -DREPRO "
#else
	#define REP " "
#endif

70

71
72
73
#define KGDB_AMD 		" -g -O0 "
#define KGDB_INTEL	" -g -s " KRNL_FILE

74

75
76
77
78
79
80
81
82
83
84
#if defined (DOCK_DEBUG)
	#if defined (CPU_DEVICE)
		#define KGDB KGDB_INTEL
	#elif defined (GPU_DEVICE)
		#define KGDB KGDB_AMD
	#endif
#else
	#define KGDB " "
#endif

85

86
87
88
89
90
91
#define OPT_PROG INC KNWI REP KGDB

#include "performdocking.h"
#include "stringify.h"

int docking_with_gpu(const Gridinfo*   mygrid,
92
	         /*const*/ float*      cpu_floatgrids,
93
                           Dockpars*   mypars,
94
95
96
97
		     const Liganddata* myligand_init,
		     const int*        argc,
		           char**      argv,
		           clock_t     clock_start_program)
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
/* The function performs the docking algorithm and generates the corresponding result files.
parameter mygrid:
		describes the grid
		filled with get_gridinfo()
parameter cpu_floatgrids:
		points to the memory region containing the grids
		filled with get_gridvalues_f()
parameter mypars:
		describes the docking parameters
		filled with get_commandpars()
parameter myligand_init:
		describes the ligands
		filled with get_liganddata()
parameters argc and argv:
		are the corresponding command line arguments parameter clock_start_program:
		contains the state of the clock tick counter at the beginning of the program
filled with clock() */
{
// =======================================================================
// OpenCL Host Setup
// =======================================================================
	cl_platform_id* platform_id;
	cl_device_id* device_id;
	cl_context context;
	cl_command_queue command_queue;

#ifdef _WIN32
	const char *filename = KRNL_FILE;
	printf("\n%-40s %-40s\n", "Kernel source file: ", filename);  fflush(stdout);
#else
	printf("\n%-40s %-40s\n", "Kernel source used for development: ", "./device/calcenergy.cl");  fflush(stdout);
	printf(  "%-40s %-40s\n", "Kernel string used for building: ",    "./host/inc/stringify.h");  fflush(stdout);
#endif

	const char* options_program = OPT_PROG;
	printf("%-40s %-40s\n", "Kernel compilation flags: ", options_program); fflush(stdout);

	cl_kernel kernel1; const char *name_k1 = KRNL1;
	size_t kernel1_gxsize, kernel1_lxsize;

	cl_kernel kernel2; const char *name_k2 = KRNL2;
	size_t kernel2_gxsize, kernel2_lxsize;

	cl_kernel kernel3; const char *name_k3 = KRNL3;
	size_t kernel3_gxsize, kernel3_lxsize;

	cl_kernel kernel4; const char *name_k4 = KRNL4;
	size_t kernel4_gxsize, kernel4_lxsize;

147
148
149
	cl_kernel kernel5; const char *name_k5 = KRNL5;
	size_t kernel5_gxsize, kernel5_lxsize;

150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
	cl_uint platformCount;
	cl_uint deviceCount;

	// Times
	cl_ulong time_start_kernel;
	cl_ulong time_end_kernel;

	// Get all available platforms
	if (getPlatforms(&platform_id,&platformCount) != 0) return 1;

	// Get all devices of first platform
	if (getDevices(platform_id[0],platformCount,&device_id,&deviceCount) != 0) return 1;

	// Create context from first platform
	if (createContext(platform_id[0],1,device_id,&context) != 0) return 1;

	// Create command queue for first device
	if (createCommandQueue(context,device_id[0],&command_queue) != 0) return 1;

	// Create program and kernel from source
#ifdef _WIN32
	if (ImportSource(filename, name_k1, device_id, context, options_program, &kernel1) != 0) return 1;
	if (ImportSource(filename, name_k2, device_id, context, options_program, &kernel2) != 0) return 1;
	if (ImportSource(filename, name_k3, device_id, context, options_program, &kernel3) != 0) return 1;
	if (ImportSource(filename, name_k4, device_id, context, options_program, &kernel4) != 0) return 1;
175
	if (ImportSource(filename, name_k5, device_id, context, options_program, &kernel5) != 0) return 1;
176
177
178
179
180
#else
	if (ImportSource(calcenergy_ocl, name_k1, device_id, context, options_program, &kernel1) != 0) return 1;
	if (ImportSource(calcenergy_ocl, name_k2, device_id, context, options_program, &kernel2) != 0) return 1;
	if (ImportSource(calcenergy_ocl, name_k3, device_id, context, options_program, &kernel3) != 0) return 1;
	if (ImportSource(calcenergy_ocl, name_k4, device_id, context, options_program, &kernel4) != 0) return 1;
181
	if (ImportSource(calcenergy_ocl, name_k5, device_id, context, options_program, &kernel5) != 0) return 1;
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
#endif

// End of OpenCL Host Setup
// =======================================================================

	Liganddata myligand_reference;

	float* cpu_init_populations;
	float* cpu_final_populations;
	float* cpu_energies;
	Ligandresult* cpu_result_ligands;
	unsigned int* cpu_prng_seeds;
	int* cpu_evals_of_runs;
	float* cpu_ref_ori_angles;

	Dockparameters dockpars;
	size_t size_floatgrids;
	size_t size_populations;
	size_t size_energies;
	size_t size_prng_seeds;
	size_t size_evals_of_runs;

	int threadsPerBlock;
	int blocksPerGridForEachEntity;
	int blocksPerGridForEachRun;
	int blocksPerGridForEachLSEntity;
208
	int blocksPerGridForEachGradMinimizerEntity;
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283

	unsigned long run_cnt;	/* int run_cnt; */
	int generation_cnt;
	int i;
	double progress;

	int curr_progress_cnt;
	int new_progress_cnt;

	clock_t clock_start_docking;
	clock_t	clock_stop_docking;
	clock_t clock_stop_program_before_clustering;

	//setting number of blocks and threads
	threadsPerBlock = NUM_OF_THREADS_PER_BLOCK;
	blocksPerGridForEachEntity = mypars->pop_size * mypars->num_of_runs;
	blocksPerGridForEachRun = mypars->num_of_runs;

	//allocating CPU memory for initial populations
	size_populations = mypars->num_of_runs * mypars->pop_size * GENOTYPE_LENGTH_IN_GLOBMEM*sizeof(float);
	cpu_init_populations = (float*) malloc(size_populations);
	memset(cpu_init_populations, 0, size_populations);

	//allocating CPU memory for results
	size_energies = mypars->pop_size * mypars->num_of_runs * sizeof(float);
	cpu_energies = (float*) malloc(size_energies);
	cpu_result_ligands = (Ligandresult*) malloc(sizeof(Ligandresult)*(mypars->num_of_runs));
	cpu_final_populations = cpu_init_populations;

	//allocating memory in CPU for reference orientation angles
	cpu_ref_ori_angles = (float*) malloc(mypars->num_of_runs*3*sizeof(float));

	//generating initial populations and random orientation angles of reference ligand
	//(ligand will be moved to origo and scaled as well)
	myligand_reference = *myligand_init;
	gen_initpop_and_reflig(mypars, cpu_init_populations, cpu_ref_ori_angles, &myligand_reference, mygrid);

	//allocating memory in CPU for pseudorandom number generator seeds and
	//generating them (seed for each thread during GA)
	size_prng_seeds = blocksPerGridForEachEntity * threadsPerBlock * sizeof(unsigned int);
	cpu_prng_seeds = (unsigned int*) malloc(size_prng_seeds);

	genseed(time(NULL));	//initializing seed generator

	for (i=0; i<blocksPerGridForEachEntity*threadsPerBlock; i++)
#if defined (REPRO)
		cpu_prng_seeds[i] = 1u;
#else
		cpu_prng_seeds[i] = genseed(0u);
#endif

	//allocating memory in CPU for evaluation counters
	size_evals_of_runs = mypars->num_of_runs*sizeof(int);
	cpu_evals_of_runs = (int*) malloc(size_evals_of_runs);
	memset(cpu_evals_of_runs, 0, size_evals_of_runs);

	//preparing the constant data fields for the GPU
	// ----------------------------------------------------------------------
	// The original function does CUDA calls initializing const Kernel data.
	// We create a struct to hold those constants
	// and return them <here> (<here> = where prepare_const_fields_for_gpu() is called),
	// so we can send them to Kernels from <here>, instead of from calcenergy.cpp as originally.
	// ----------------------------------------------------------------------
	// Constant struct
	kernelconstant KerConst;

	if (prepare_const_fields_for_gpu(&myligand_reference, mypars, cpu_ref_ori_angles, &KerConst) == 1)
		return 1;

	// Constant data holding struct data
	// Created because structs containing array
	// are not supported as OpenCL kernel args
  cl_mem mem_atom_charges_const;
  cl_mem mem_atom_types_const;
  cl_mem mem_intraE_contributors_const;
lvs's avatar
lvs committed
284
285
  cl_mem mem_reqm_const;
  cl_mem mem_reqm_hbond_const;
286
287
288
289
290
291
292
293
294
295
296
  cl_mem mem_VWpars_AC_const;
  cl_mem mem_VWpars_BD_const;
  cl_mem mem_dspars_S_const;
  cl_mem mem_dspars_V_const;
  cl_mem mem_rotlist_const;
  cl_mem mem_ref_coords_x_const;
  cl_mem mem_ref_coords_y_const;
  cl_mem mem_ref_coords_z_const;
  cl_mem mem_rotbonds_moving_vectors_const;
  cl_mem mem_rotbonds_unit_vectors_const;
  cl_mem mem_ref_orientation_quats_const;
297
298
299
  cl_mem mem_rotbonds_const;
  cl_mem mem_rotbonds_atoms_const;
  cl_mem mem_num_rotating_atoms_per_rotbond_const;
300
301
302
303
304
305
306

	// These constants are allocated in global memory since
	// there is a limited number of constants that can be passed
	// as arguments to kernel
  mallocBufferObject(context,CL_MEM_READ_ONLY,MAX_NUM_OF_ATOMS*sizeof(float),                         &mem_atom_charges_const);
  mallocBufferObject(context,CL_MEM_READ_ONLY,MAX_NUM_OF_ATOMS*sizeof(char),                          &mem_atom_types_const);
  mallocBufferObject(context,CL_MEM_READ_ONLY,3*MAX_INTRAE_CONTRIBUTORS*sizeof(char),                 &mem_intraE_contributors_const);
lvs's avatar
lvs committed
307
308
  mallocBufferObject(context,CL_MEM_READ_ONLY,ATYPE_NUM*sizeof(float),&mem_reqm_const);
  mallocBufferObject(context,CL_MEM_READ_ONLY,ATYPE_NUM*sizeof(float),&mem_reqm_hbond_const);
309
310
311
312
313
314
315
316
317
318
319
  mallocBufferObject(context,CL_MEM_READ_ONLY,MAX_NUM_OF_ATYPES*MAX_NUM_OF_ATYPES*sizeof(float),      &mem_VWpars_AC_const);
  mallocBufferObject(context,CL_MEM_READ_ONLY,MAX_NUM_OF_ATYPES*MAX_NUM_OF_ATYPES*sizeof(float),      &mem_VWpars_BD_const);
  mallocBufferObject(context,CL_MEM_READ_ONLY,MAX_NUM_OF_ATYPES*sizeof(float),                        &mem_dspars_S_const);
  mallocBufferObject(context,CL_MEM_READ_ONLY,MAX_NUM_OF_ATYPES*sizeof(float),                        &mem_dspars_V_const);
  mallocBufferObject(context,CL_MEM_READ_ONLY,MAX_NUM_OF_ROTATIONS*sizeof(int),                       &mem_rotlist_const);
  mallocBufferObject(context,CL_MEM_READ_ONLY,MAX_NUM_OF_ATOMS*sizeof(float),                         &mem_ref_coords_x_const);
  mallocBufferObject(context,CL_MEM_READ_ONLY,MAX_NUM_OF_ATOMS*sizeof(float),                         &mem_ref_coords_y_const);
  mallocBufferObject(context,CL_MEM_READ_ONLY,MAX_NUM_OF_ATOMS*sizeof(float),                         &mem_ref_coords_z_const);
  mallocBufferObject(context,CL_MEM_READ_ONLY,3*MAX_NUM_OF_ROTBONDS*sizeof(float),                    &mem_rotbonds_moving_vectors_const);
  mallocBufferObject(context,CL_MEM_READ_ONLY,3*MAX_NUM_OF_ROTBONDS*sizeof(float),                    &mem_rotbonds_unit_vectors_const);
  mallocBufferObject(context,CL_MEM_READ_ONLY,4*MAX_NUM_OF_RUNS*sizeof(float),                        &mem_ref_orientation_quats_const);
320
321
322
  mallocBufferObject(context,CL_MEM_READ_ONLY,2*MAX_NUM_OF_ROTBONDS*sizeof(int),                      &mem_rotbonds_const);
  mallocBufferObject(context,CL_MEM_READ_ONLY,MAX_NUM_OF_ATOMS*MAX_NUM_OF_ROTBONDS*sizeof(int),      &mem_rotbonds_atoms_const);
  mallocBufferObject(context,CL_MEM_READ_ONLY,MAX_NUM_OF_ROTBONDS*sizeof(int),      		&mem_num_rotating_atoms_per_rotbond_const);
323
324
325
326

  memcopyBufferObjectToDevice(command_queue,mem_atom_charges_const,         	&KerConst.atom_charges_const,           MAX_NUM_OF_ATOMS*sizeof(float));
  memcopyBufferObjectToDevice(command_queue,mem_atom_types_const,           	&KerConst.atom_types_const,             MAX_NUM_OF_ATOMS*sizeof(char));
  memcopyBufferObjectToDevice(command_queue,mem_intraE_contributors_const,  	&KerConst.intraE_contributors_const,    3*MAX_INTRAE_CONTRIBUTORS*sizeof(char));
lvs's avatar
lvs committed
327
328
  memcopyBufferObjectToDevice(command_queue,mem_reqm_const,         		&KerConst.reqm_const,           	ATYPE_NUM*sizeof(float));
  memcopyBufferObjectToDevice(command_queue,mem_reqm_hbond_const,         	&KerConst.reqm_hbond_const,           	ATYPE_NUM*sizeof(float));
329
330
331
332
333
334
335
336
337
338
339
  memcopyBufferObjectToDevice(command_queue,mem_VWpars_AC_const,            	&KerConst.VWpars_AC_const,              MAX_NUM_OF_ATYPES*MAX_NUM_OF_ATYPES*sizeof(float));
  memcopyBufferObjectToDevice(command_queue,mem_VWpars_BD_const,            	&KerConst.VWpars_BD_const,              MAX_NUM_OF_ATYPES*MAX_NUM_OF_ATYPES*sizeof(float));
  memcopyBufferObjectToDevice(command_queue,mem_dspars_S_const,             	&KerConst.dspars_S_const,               MAX_NUM_OF_ATYPES*sizeof(float));
  memcopyBufferObjectToDevice(command_queue,mem_dspars_V_const,             	&KerConst.dspars_V_const,               MAX_NUM_OF_ATYPES*sizeof(float));
  memcopyBufferObjectToDevice(command_queue,mem_rotlist_const,              	&KerConst.rotlist_const,                MAX_NUM_OF_ROTATIONS*sizeof(int));
  memcopyBufferObjectToDevice(command_queue,mem_ref_coords_x_const,         	&KerConst.ref_coords_x_const,           MAX_NUM_OF_ATOMS*sizeof(float));
  memcopyBufferObjectToDevice(command_queue,mem_ref_coords_y_const,         	&KerConst.ref_coords_y_const,           MAX_NUM_OF_ATOMS*sizeof(float));
  memcopyBufferObjectToDevice(command_queue,mem_ref_coords_z_const,         	&KerConst.ref_coords_z_const,           MAX_NUM_OF_ATOMS*sizeof(float));
  memcopyBufferObjectToDevice(command_queue,mem_rotbonds_moving_vectors_const,&KerConst.rotbonds_moving_vectors_const,3*MAX_NUM_OF_ROTBONDS*sizeof(float));
  memcopyBufferObjectToDevice(command_queue,mem_rotbonds_unit_vectors_const,  &KerConst.rotbonds_unit_vectors_const,  3*MAX_NUM_OF_ROTBONDS*sizeof(float));
  memcopyBufferObjectToDevice(command_queue,mem_ref_orientation_quats_const,  &KerConst.ref_orientation_quats_const,  4*MAX_NUM_OF_RUNS*sizeof(float));
340
341
342
  memcopyBufferObjectToDevice(command_queue,mem_rotbonds_const,  	      &KerConst.rotbonds,  2*MAX_NUM_OF_ROTBONDS*sizeof(int));
  memcopyBufferObjectToDevice(command_queue,mem_rotbonds_atoms_const,  	      &KerConst.rotbonds_atoms,  MAX_NUM_OF_ATOMS*MAX_NUM_OF_ROTBONDS*sizeof(int));
  memcopyBufferObjectToDevice(command_queue,mem_num_rotating_atoms_per_rotbond_const, &KerConst.num_rotating_atoms_per_rotbond, MAX_NUM_OF_ROTBONDS*sizeof(int));
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
	// ----------------------------------------------------------------------

 	//allocating GPU memory for populations, floatgirds,
	//energies, evaluation counters and random number generator states
	size_floatgrids = (sizeof(float)) * (mygrid->num_of_atypes+2) * (mygrid->size_xyz[0]) * (mygrid->size_xyz[1]) * (mygrid->size_xyz[2]);

	cl_mem mem_dockpars_fgrids;
	cl_mem mem_dockpars_conformations_current;
	cl_mem mem_dockpars_energies_current;
	cl_mem mem_dockpars_conformations_next;
	cl_mem mem_dockpars_energies_next;
	cl_mem mem_dockpars_evals_of_new_entities;
	cl_mem mem_gpu_evals_of_runs;
	cl_mem mem_dockpars_prng_states;

	mallocBufferObject(context,CL_MEM_READ_ONLY,size_floatgrids,         				&mem_dockpars_fgrids);
	mallocBufferObject(context,CL_MEM_READ_ONLY,size_populations,        				&mem_dockpars_conformations_current);
	mallocBufferObject(context,CL_MEM_READ_WRITE,size_energies,           			&mem_dockpars_energies_current);
	mallocBufferObject(context,CL_MEM_READ_WRITE,size_populations,        			&mem_dockpars_conformations_next);
	mallocBufferObject(context,CL_MEM_READ_WRITE,size_energies,    	      			&mem_dockpars_energies_next);
	mallocBufferObject(context,CL_MEM_READ_WRITE,mypars->pop_size*mypars->num_of_runs*sizeof(int), 	&mem_dockpars_evals_of_new_entities);

	// -------- Replacing with memory maps! ------------
#if defined (MAPPED_COPY)
	mallocBufferObject(context,CL_MEM_READ_WRITE | CL_MEM_ALLOC_HOST_PTR ,size_evals_of_runs,	  			&mem_gpu_evals_of_runs);
#else
	mallocBufferObject(context,CL_MEM_READ_WRITE,size_evals_of_runs,	  			&mem_gpu_evals_of_runs);
#endif
	// -------- Replacing with memory maps! ------------

	mallocBufferObject(context,CL_MEM_READ_WRITE,size_prng_seeds,  	      				&mem_dockpars_prng_states);

	memcopyBufferObjectToDevice(command_queue,mem_dockpars_fgrids,                /*(void *)*/ cpu_floatgrids,  size_floatgrids);
 	memcopyBufferObjectToDevice(command_queue,mem_dockpars_conformations_current, cpu_init_populations, 				size_populations);
	memcopyBufferObjectToDevice(command_queue,mem_gpu_evals_of_runs, 							cpu_evals_of_runs, 	 					size_evals_of_runs);
	memcopyBufferObjectToDevice(command_queue,mem_dockpars_prng_states,     			cpu_prng_seeds,      					size_prng_seeds);

	//preparing parameter struct
	dockpars.num_of_atoms  = ((char)  myligand_reference.num_of_atoms);
	dockpars.num_of_atypes = ((char)  myligand_reference.num_of_atypes);
	dockpars.num_of_intraE_contributors = ((int) myligand_reference.num_of_intraE_contributors);
	dockpars.gridsize_x    = ((char)  mygrid->size_xyz[0]);
	dockpars.gridsize_y    = ((char)  mygrid->size_xyz[1]);
	dockpars.gridsize_z    = ((char)  mygrid->size_xyz[2]);
	dockpars.grid_spacing  = ((float) mygrid->spacing);
	dockpars.rotbondlist_length = ((int) NUM_OF_THREADS_PER_BLOCK*(myligand_reference.num_of_rotcyc));
	dockpars.coeff_elec    = ((float) mypars->coeffs.scaled_AD4_coeff_elec);
	dockpars.coeff_desolv  = ((float) mypars->coeffs.AD4_coeff_desolv);
	dockpars.pop_size      = mypars->pop_size;
	dockpars.num_of_genes  = myligand_reference.num_of_rotbonds + 6;
	dockpars.tournament_rate = mypars->tournament_rate;
	dockpars.crossover_rate  = mypars->crossover_rate;
	dockpars.mutation_rate   = mypars->mutation_rate;
	dockpars.abs_max_dang    = mypars->abs_max_dang;
	dockpars.abs_max_dmov    = mypars->abs_max_dmov;
	dockpars.lsearch_rate    = mypars->lsearch_rate;
	dockpars.num_of_lsentities = (unsigned int) (mypars->lsearch_rate/100.0*mypars->pop_size + 0.5);
	dockpars.rho_lower_bound   = mypars->rho_lower_bound;
	dockpars.base_dmov_mul_sqrt3 = mypars->base_dmov_mul_sqrt3;
	dockpars.base_dang_mul_sqrt3 = mypars->base_dang_mul_sqrt3;
	dockpars.cons_limit        = (unsigned int) mypars->cons_limit;
	dockpars.max_num_of_iters  = (unsigned int) mypars->max_num_of_iters;
	dockpars.qasp = mypars->qasp;
lvs's avatar
lvs committed
406
407
	dockpars.smooth = mypars->smooth;

408
409
410
	unsigned int g2 = dockpars.gridsize_x * dockpars.gridsize_y;
	unsigned int g3 = dockpars.gridsize_x * dockpars.gridsize_y * dockpars.gridsize_z;

lvs's avatar
lvs committed
411
	printf("Local-search chosen method is: %s\n", (strcmp(mypars->ls_method, "sw") == 0)?"Solis-Wets":
412
413
414
						      (strcmp(mypars->ls_method, "sd") == 0)?"Steepest descent": "Unknown");


415
416
	blocksPerGridForEachLSEntity = dockpars.num_of_lsentities*mypars->num_of_runs;

Leonardo Solis's avatar
Leonardo Solis committed
417
	/*
Leonardo Solis's avatar
Leonardo Solis committed
418
	printf("dockpars.num_of_intraE_contributors:%u\n", dockpars.num_of_intraE_contributors);
419
	printf("dockpars.rotbondlist_length:%u\n", dockpars.rotbondlist_length);
Leonardo Solis's avatar
Leonardo Solis committed
420
	*/
421
422
423
424
425

	// Experimental TSRI gradient-based minimizer kernel argument
	// Setup here (temporarily?) the gradient and associated parameters.
	// This should be ultimately configurable by the user as program exec. flags.
	Gradientparameters gradientpars;
Leonardo Solis's avatar
Leonardo Solis committed
426
	gradientpars.max_num_of_iters = 300;
427
428

/*
Leonardo Solis's avatar
Leonardo Solis committed
429
430
	gradientpars.max_num_of_consec_fails = (unsigned int)(0.05 * gradientpars.max_num_of_iters);

431
	gradientpars.alpha = 0.000001f; 
432
433
434

	// Set minimum values for input conformation (translation genes as x, y, z)
	for (unsigned int gene_cnt=0; gene_cnt<3; gene_cnt++) {
435
			gradientpars.conformation_min_perturbation[0] = 1e-5; // FIXME: set right min values max .1/.375
436
437
438
439
	}

	// Set minimum values for input conformation (rotation genes as Shoemake genes)
	for (unsigned int gene_cnt=3; gene_cnt<6; gene_cnt++) {
440
			gradientpars.conformation_min_perturbation[0] = 0.0005f; //max 0.01
441
442
443
444
	}

	// Set minimum values for input conformations (torsion genes as angles)
	for (unsigned int gene_cnt=6; gene_cnt<(dockpars.num_of_genes - 6); gene_cnt++) {
445
			gradientpars.conformation_min_perturbation[0] = 0.1f; // FIXME: set right min values max 1.0
446
447
448
449
450
451
452
	}

	cl_mem mem_gradpars_conformation_min_perturbation;
	size_t size_conformation_min_perturbation = ACTUAL_GENOTYPE_LENGTH*sizeof(float);

	mallocBufferObject(context, CL_MEM_READ_ONLY, size_conformation_min_perturbation, &mem_gradpars_conformation_min_perturbation);
	memcopyBufferObjectToDevice(command_queue, mem_gradpars_conformation_min_perturbation, gradientpars.conformation_min_perturbation, size_conformation_min_perturbation);
453
454
*/

455

Leonardo Solis's avatar
Leonardo Solis committed
456
///*
457
	// Initially, the number of entities that undergo gradient-minimization,
458
	// by default, it is the same as the number of entities that undergo gradient-based minimizer
459
	blocksPerGridForEachGradMinimizerEntity = dockpars.num_of_lsentities*mypars->num_of_runs;
Leonardo Solis's avatar
Leonardo Solis committed
460
//*/
461

Leonardo Solis's avatar
Leonardo Solis committed
462
/*
463
464
	// test, only one entity per reach run, undergoes gradient minimization
	blocksPerGridForEachGradMinimizerEntity = mypars->num_of_runs;
Leonardo Solis's avatar
Leonardo Solis committed
465
*/
466

467
468
	clock_start_docking = clock();

469
470
471



472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
	//print progress bar
#ifndef DOCK_DEBUG
	printf("\nExecuting docking runs:\n");
	printf("        20%%        40%%       60%%       80%%       100%%\n");
	printf("---------+---------+---------+---------+---------+\n");
#else
	printf("\n");
#endif
	curr_progress_cnt = 0;

#ifdef DOCK_DEBUG
	// Main while-loop iterarion counter
	unsigned int ite_cnt = 0;
#endif

Leonardo Solis's avatar
Leonardo Solis committed
487
	/*
Leonardo Solis's avatar
Leonardo Solis committed
488
489
490
491
492
493
494
495
496
497
498
499
	// Addded for printing intracontributor_pairs (autodockdevpy)
	for (unsigned int intrapair_cnt=0; 
			  intrapair_cnt<dockpars.num_of_intraE_contributors;
			  intrapair_cnt++) {
		if (intrapair_cnt == 0) {
			printf("%-10s %-10s %-10s\n", "#pair", "#atom1", "#atom2");
		}

		printf ("%-10u %-10u %-10u\n", intrapair_cnt,
					    KerConst.intraE_contributors_const[3*intrapair_cnt],
					    KerConst.intraE_contributors_const[3*intrapair_cnt+1]);
	}
Leonardo Solis's avatar
Leonardo Solis committed
500
	*/
Leonardo Solis's avatar
Leonardo Solis committed
501
502
503
504
505
506






507
508
509
510
511
512
513
// Kernel1
  setKernelArg(kernel1,0, sizeof(dockpars.num_of_atoms),                  &dockpars.num_of_atoms);
  setKernelArg(kernel1,1, sizeof(dockpars.num_of_atypes),                 &dockpars.num_of_atypes);
  setKernelArg(kernel1,2, sizeof(dockpars.num_of_intraE_contributors),    &dockpars.num_of_intraE_contributors);
  setKernelArg(kernel1,3, sizeof(dockpars.gridsize_x),                    &dockpars.gridsize_x);
  setKernelArg(kernel1,4, sizeof(dockpars.gridsize_y),                    &dockpars.gridsize_y);
  setKernelArg(kernel1,5, sizeof(dockpars.gridsize_z),                    &dockpars.gridsize_z);
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
  setKernelArg(kernel1,6, sizeof(g2),                    		  &g2);
  setKernelArg(kernel1,7, sizeof(g3),                    		  &g3);
  setKernelArg(kernel1,8, sizeof(dockpars.grid_spacing),                  &dockpars.grid_spacing);
  setKernelArg(kernel1,9, sizeof(mem_dockpars_fgrids),                    &mem_dockpars_fgrids);
  setKernelArg(kernel1,10,sizeof(dockpars.rotbondlist_length),            &dockpars.rotbondlist_length);
  setKernelArg(kernel1,11,sizeof(dockpars.coeff_elec),                    &dockpars.coeff_elec);
  setKernelArg(kernel1,12,sizeof(dockpars.coeff_desolv),                  &dockpars.coeff_desolv);
  setKernelArg(kernel1,13,sizeof(mem_dockpars_conformations_current),     &mem_dockpars_conformations_current);
  setKernelArg(kernel1,14,sizeof(mem_dockpars_energies_current),          &mem_dockpars_energies_current);
  setKernelArg(kernel1,15,sizeof(mem_dockpars_evals_of_new_entities),     &mem_dockpars_evals_of_new_entities);
  setKernelArg(kernel1,16,sizeof(dockpars.pop_size),                      &dockpars.pop_size);
  setKernelArg(kernel1,17,sizeof(dockpars.qasp),                          &dockpars.qasp);
  setKernelArg(kernel1,18,sizeof(mem_atom_charges_const),                 &mem_atom_charges_const);
  setKernelArg(kernel1,19,sizeof(mem_atom_types_const),                   &mem_atom_types_const);
  setKernelArg(kernel1,20,sizeof(mem_intraE_contributors_const),          &mem_intraE_contributors_const);
lvs's avatar
lvs committed
529
530
531
532
533
534
535
536
537
538
539
540
541
542
  setKernelArg(kernel1,21,sizeof(dockpars.smooth),                        &dockpars.smooth);
  setKernelArg(kernel1,22,sizeof(mem_reqm_const),                         &mem_reqm_const);
  setKernelArg(kernel1,23,sizeof(mem_reqm_hbond_const),                   &mem_reqm_hbond_const);
  setKernelArg(kernel1,24/*21*/,sizeof(mem_VWpars_AC_const),                    &mem_VWpars_AC_const);
  setKernelArg(kernel1,25/*22*/,sizeof(mem_VWpars_BD_const),                    &mem_VWpars_BD_const);
  setKernelArg(kernel1,26/*23*/,sizeof(mem_dspars_S_const),                     &mem_dspars_S_const);
  setKernelArg(kernel1,27/*24*/,sizeof(mem_dspars_V_const),                     &mem_dspars_V_const);
  setKernelArg(kernel1,28/*25*/,sizeof(mem_rotlist_const),                      &mem_rotlist_const);
  setKernelArg(kernel1,29/*26*/,sizeof(mem_ref_coords_x_const),                 &mem_ref_coords_x_const);
  setKernelArg(kernel1,30/*27*/,sizeof(mem_ref_coords_y_const),                 &mem_ref_coords_y_const);
  setKernelArg(kernel1,31/*28*/,sizeof(mem_ref_coords_z_const),                 &mem_ref_coords_z_const);
  setKernelArg(kernel1,32/*29*/,sizeof(mem_rotbonds_moving_vectors_const),      &mem_rotbonds_moving_vectors_const);
  setKernelArg(kernel1,33/*30*/,sizeof(mem_rotbonds_unit_vectors_const),        &mem_rotbonds_unit_vectors_const);
  setKernelArg(kernel1,34/*31*/,sizeof(mem_ref_orientation_quats_const),        &mem_ref_orientation_quats_const);
543
544
545
546
547
548
549
550
551
  kernel1_gxsize = blocksPerGridForEachEntity * threadsPerBlock;
  kernel1_lxsize = threadsPerBlock;
#ifdef DOCK_DEBUG
	printf("%-25s %10s %8u %10s %4u\n", "Kernel1: ", "gSize: ", kernel1_gxsize, "lSize: ", kernel1_lxsize); fflush(stdout);
#endif
// End of Kernel1

// Kernel2
//  setKernelArg(kernel2,0,sizeof(mypars->pop_size),       		&mypars->pop_size);
552
  setKernelArg(kernel2,0,sizeof(dockpars.pop_size),       		  &dockpars.pop_size);
553
554
  setKernelArg(kernel2,1,sizeof(mem_dockpars_evals_of_new_entities),      &mem_dockpars_evals_of_new_entities);
  setKernelArg(kernel2,2,sizeof(mem_gpu_evals_of_runs),                   &mem_gpu_evals_of_runs);
555
  kernel2_gxsize = blocksPerGridForEachRun * threadsPerBlock;
556
557
558
559
560
561
562
  kernel2_lxsize = threadsPerBlock;
#ifdef DOCK_DEBUG
	printf("%-25s %10s %8u %10s %4u\n", "Kernel2: ", "gSize: ", kernel2_gxsize, "lSize: ",  kernel2_lxsize); fflush(stdout);
#endif
// End of Kernel2

// Kernel4
563
  setKernelArg(kernel4,0, sizeof(dockpars.num_of_atoms),                  &dockpars.num_of_atoms);
564
565
  setKernelArg(kernel4,1, sizeof(dockpars.num_of_atypes),                 &dockpars.num_of_atypes);
  setKernelArg(kernel4,2, sizeof(dockpars.num_of_intraE_contributors),    &dockpars.num_of_intraE_contributors);
566
  setKernelArg(kernel4,3, sizeof(dockpars.gridsize_x),                    &dockpars.gridsize_x);
567
568
  setKernelArg(kernel4,4, sizeof(dockpars.gridsize_y),                    &dockpars.gridsize_y);
  setKernelArg(kernel4,5, sizeof(dockpars.gridsize_z),                    &dockpars.gridsize_z);
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
  setKernelArg(kernel4,6, sizeof(g2),                    		  &g2);
  setKernelArg(kernel4,7, sizeof(g3),                    		  &g3);
  setKernelArg(kernel4,8, sizeof(dockpars.grid_spacing),                  &dockpars.grid_spacing);
  setKernelArg(kernel4,9, sizeof(mem_dockpars_fgrids),                    &mem_dockpars_fgrids);
  setKernelArg(kernel4,10,sizeof(dockpars.rotbondlist_length),            &dockpars.rotbondlist_length);
  setKernelArg(kernel4,11,sizeof(dockpars.coeff_elec),                    &dockpars.coeff_elec);
  setKernelArg(kernel4,12,sizeof(dockpars.coeff_desolv),                  &dockpars.coeff_desolv);
  setKernelArg(kernel4,13,sizeof(mem_dockpars_conformations_current),     &mem_dockpars_conformations_current);
  setKernelArg(kernel4,14,sizeof(mem_dockpars_energies_current),          &mem_dockpars_energies_current);
  setKernelArg(kernel4,15,sizeof(mem_dockpars_conformations_next),        &mem_dockpars_conformations_next);
  setKernelArg(kernel4,16,sizeof(mem_dockpars_energies_next),             &mem_dockpars_energies_next);
  setKernelArg(kernel4,17,sizeof(mem_dockpars_evals_of_new_entities),     &mem_dockpars_evals_of_new_entities);
  setKernelArg(kernel4,18,sizeof(mem_dockpars_prng_states),               &mem_dockpars_prng_states);
  setKernelArg(kernel4,19,sizeof(dockpars.pop_size),                      &dockpars.pop_size);
  setKernelArg(kernel4,20,sizeof(dockpars.num_of_genes),                  &dockpars.num_of_genes);
  setKernelArg(kernel4,21,sizeof(dockpars.tournament_rate),               &dockpars.tournament_rate);
  setKernelArg(kernel4,22,sizeof(dockpars.crossover_rate),                &dockpars.crossover_rate);
  setKernelArg(kernel4,23,sizeof(dockpars.mutation_rate),                 &dockpars.mutation_rate);
  setKernelArg(kernel4,24,sizeof(dockpars.abs_max_dmov),                  &dockpars.abs_max_dmov);
  setKernelArg(kernel4,25,sizeof(dockpars.abs_max_dang),                  &dockpars.abs_max_dang);
  setKernelArg(kernel4,26,sizeof(dockpars.qasp),                          &dockpars.qasp);
  setKernelArg(kernel4,27,sizeof(mem_atom_charges_const),                 &mem_atom_charges_const);
  setKernelArg(kernel4,28,sizeof(mem_atom_types_const),                   &mem_atom_types_const);
  setKernelArg(kernel4,29,sizeof(mem_intraE_contributors_const),          &mem_intraE_contributors_const);
lvs's avatar
lvs committed
593
594
595
596
597
598
599
600
601
602
603
604
605
606
  setKernelArg(kernel4,30,sizeof(dockpars.smooth),                        &dockpars.smooth);
  setKernelArg(kernel4,31,sizeof(mem_reqm_const),                         &mem_reqm_const);
  setKernelArg(kernel4,32,sizeof(mem_reqm_hbond_const),                   &mem_reqm_hbond_const);
  setKernelArg(kernel4,33/*30*/,sizeof(mem_VWpars_AC_const),                    &mem_VWpars_AC_const);
  setKernelArg(kernel4,34/*31*/,sizeof(mem_VWpars_BD_const),                    &mem_VWpars_BD_const);
  setKernelArg(kernel4,35/*32*/,sizeof(mem_dspars_S_const),                     &mem_dspars_S_const);
  setKernelArg(kernel4,36/*33*/,sizeof(mem_dspars_V_const),                     &mem_dspars_V_const);
  setKernelArg(kernel4,37/*34*/,sizeof(mem_rotlist_const),                      &mem_rotlist_const);
  setKernelArg(kernel4,38/*35*/,sizeof(mem_ref_coords_x_const),                 &mem_ref_coords_x_const);
  setKernelArg(kernel4,39/*36*/,sizeof(mem_ref_coords_y_const),                 &mem_ref_coords_y_const);
  setKernelArg(kernel4,40/*37*/,sizeof(mem_ref_coords_z_const),                 &mem_ref_coords_z_const);
  setKernelArg(kernel4,41/*38*/,sizeof(mem_rotbonds_moving_vectors_const),      &mem_rotbonds_moving_vectors_const);
  setKernelArg(kernel4,42/*39*/,sizeof(mem_rotbonds_unit_vectors_const),        &mem_rotbonds_unit_vectors_const);
  setKernelArg(kernel4,43/*40*/,sizeof(mem_ref_orientation_quats_const),        &mem_ref_orientation_quats_const);
607
  kernel4_gxsize = blocksPerGridForEachEntity * threadsPerBlock;
608
609
610
611
612
613
  kernel4_lxsize = threadsPerBlock;
#ifdef DOCK_DEBUG
	printf("%-25s %10s %8u %10s %4u\n", "K_GENETIC_GENERATION: ", "gSize: ",  kernel4_gxsize, "lSize: ", kernel4_lxsize); fflush(stdout);
#endif
// End of Kernel4

614

615
/*
616
#if !defined (GRADIENT_ENABLED)
617
618
*/
if (strcmp(mypars->ls_method, "sw") == 0) {
619
// Kernel3
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
  setKernelArg(kernel3,0, sizeof(dockpars.num_of_atoms),                  &dockpars.num_of_atoms);
  setKernelArg(kernel3,1, sizeof(dockpars.num_of_atypes),                 &dockpars.num_of_atypes);
  setKernelArg(kernel3,2, sizeof(dockpars.num_of_intraE_contributors),    &dockpars.num_of_intraE_contributors);
  setKernelArg(kernel3,3, sizeof(dockpars.gridsize_x),                    &dockpars.gridsize_x);
  setKernelArg(kernel3,4, sizeof(dockpars.gridsize_y),                    &dockpars.gridsize_y);
  setKernelArg(kernel3,5, sizeof(dockpars.gridsize_z),                    &dockpars.gridsize_z);
  setKernelArg(kernel3,6, sizeof(g2),                    		  &g2);
  setKernelArg(kernel3,7, sizeof(g3),                    		  &g3);
  setKernelArg(kernel3,8, sizeof(dockpars.grid_spacing),                  &dockpars.grid_spacing);
  setKernelArg(kernel3,9, sizeof(mem_dockpars_fgrids),                    &mem_dockpars_fgrids);
  setKernelArg(kernel3,10,sizeof(dockpars.rotbondlist_length),            &dockpars.rotbondlist_length);
  setKernelArg(kernel3,11,sizeof(dockpars.coeff_elec),                    &dockpars.coeff_elec);
  setKernelArg(kernel3,12,sizeof(dockpars.coeff_desolv),                  &dockpars.coeff_desolv);
  setKernelArg(kernel3,13,sizeof(mem_dockpars_conformations_next),        &mem_dockpars_conformations_next);
  setKernelArg(kernel3,14,sizeof(mem_dockpars_energies_next),             &mem_dockpars_energies_next);
  setKernelArg(kernel3,15,sizeof(mem_dockpars_evals_of_new_entities),     &mem_dockpars_evals_of_new_entities);
  setKernelArg(kernel3,16,sizeof(mem_dockpars_prng_states),               &mem_dockpars_prng_states);
  setKernelArg(kernel3,17,sizeof(dockpars.pop_size),                      &dockpars.pop_size);
  setKernelArg(kernel3,18,sizeof(dockpars.num_of_genes),                  &dockpars.num_of_genes);
  setKernelArg(kernel3,19,sizeof(dockpars.lsearch_rate),                  &dockpars.lsearch_rate);
  setKernelArg(kernel3,20,sizeof(dockpars.num_of_lsentities),             &dockpars.num_of_lsentities);
  setKernelArg(kernel3,21,sizeof(dockpars.rho_lower_bound),               &dockpars.rho_lower_bound);
  setKernelArg(kernel3,22,sizeof(dockpars.base_dmov_mul_sqrt3),           &dockpars.base_dmov_mul_sqrt3);
  setKernelArg(kernel3,23,sizeof(dockpars.base_dang_mul_sqrt3),           &dockpars.base_dang_mul_sqrt3);
  setKernelArg(kernel3,24,sizeof(dockpars.cons_limit),                    &dockpars.cons_limit);
  setKernelArg(kernel3,25,sizeof(dockpars.max_num_of_iters),              &dockpars.max_num_of_iters);
  setKernelArg(kernel3,26,sizeof(dockpars.qasp),                          &dockpars.qasp);
  setKernelArg(kernel3,27,sizeof(mem_atom_charges_const),                 &mem_atom_charges_const);
  setKernelArg(kernel3,28,sizeof(mem_atom_types_const),                   &mem_atom_types_const);
  setKernelArg(kernel3,29,sizeof(mem_intraE_contributors_const),          &mem_intraE_contributors_const);
lvs's avatar
lvs committed
650
651
652
653
654
655
656
657
658
659
660
661
662
663
  setKernelArg(kernel3,30,sizeof(dockpars.smooth),                        &dockpars.smooth);
  setKernelArg(kernel3,31,sizeof(mem_reqm_const),                         &mem_reqm_const);
  setKernelArg(kernel3,32,sizeof(mem_reqm_hbond_const),                   &mem_reqm_hbond_const);
  setKernelArg(kernel3,33/*30*/,sizeof(mem_VWpars_AC_const),                    &mem_VWpars_AC_const);
  setKernelArg(kernel3,34/*31*/,sizeof(mem_VWpars_BD_const),                    &mem_VWpars_BD_const);
  setKernelArg(kernel3,35/*32*/,sizeof(mem_dspars_S_const),                     &mem_dspars_S_const);
  setKernelArg(kernel3,36/*33*/,sizeof(mem_dspars_V_const),                     &mem_dspars_V_const);
  setKernelArg(kernel3,37/*34*/,sizeof(mem_rotlist_const),                      &mem_rotlist_const);
  setKernelArg(kernel3,38/*35*/,sizeof(mem_ref_coords_x_const),                 &mem_ref_coords_x_const);
  setKernelArg(kernel3,39/*36*/,sizeof(mem_ref_coords_y_const),                 &mem_ref_coords_y_const);
  setKernelArg(kernel3,40/*37*/,sizeof(mem_ref_coords_z_const),                 &mem_ref_coords_z_const);
  setKernelArg(kernel3,41/*38*/,sizeof(mem_rotbonds_moving_vectors_const),      &mem_rotbonds_moving_vectors_const);
  setKernelArg(kernel3,42/*39*/,sizeof(mem_rotbonds_unit_vectors_const),        &mem_rotbonds_unit_vectors_const);
  setKernelArg(kernel3,43/*40*/,sizeof(mem_ref_orientation_quats_const),        &mem_ref_orientation_quats_const);
664
665
  kernel3_gxsize = blocksPerGridForEachLSEntity * threadsPerBlock;
  kernel3_lxsize = threadsPerBlock;
666
667
668
  #ifdef DOCK_DEBUG
  printf("%-25s %10s %8u %10s %4u\n", "K_LOCAL_SEARCH: ", "gSize: ", kernel3_gxsize, "lSize: ", kernel3_lxsize); fflush(stdout);
  #endif
669
// End of Kernel3
670
671
} else if (strcmp(mypars->ls_method, "sd") == 0) {
/*
672
#else
673
674
*/

675
676
677
678
679
680
681
// Kernel5
  setKernelArg(kernel5,0, sizeof(dockpars.num_of_atoms),                   &dockpars.num_of_atoms);
  setKernelArg(kernel5,1, sizeof(dockpars.num_of_atypes),                  &dockpars.num_of_atypes);
  setKernelArg(kernel5,2, sizeof(dockpars.num_of_intraE_contributors),     &dockpars.num_of_intraE_contributors);
  setKernelArg(kernel5,3, sizeof(dockpars.gridsize_x),                     &dockpars.gridsize_x);
  setKernelArg(kernel5,4, sizeof(dockpars.gridsize_y),                     &dockpars.gridsize_y);
  setKernelArg(kernel5,5, sizeof(dockpars.gridsize_z),                     &dockpars.gridsize_z);
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
  setKernelArg(kernel5,6, sizeof(g2),                    		   &g2);
  setKernelArg(kernel5,7, sizeof(g3),                    		   &g3);
  setKernelArg(kernel5,8, sizeof(dockpars.grid_spacing),                   &dockpars.grid_spacing);
  setKernelArg(kernel5,9, sizeof(mem_dockpars_fgrids),                     &mem_dockpars_fgrids);
  setKernelArg(kernel5,10,sizeof(dockpars.rotbondlist_length),             &dockpars.rotbondlist_length);
  setKernelArg(kernel5,11,sizeof(dockpars.coeff_elec),                     &dockpars.coeff_elec);
  setKernelArg(kernel5,12,sizeof(dockpars.coeff_desolv),                   &dockpars.coeff_desolv);
  setKernelArg(kernel5,13,sizeof(mem_dockpars_conformations_next),         &mem_dockpars_conformations_next);
  setKernelArg(kernel5,14,sizeof(mem_dockpars_energies_next),              &mem_dockpars_energies_next);
  setKernelArg(kernel5,15,sizeof(mem_dockpars_evals_of_new_entities),      &mem_dockpars_evals_of_new_entities);
  setKernelArg(kernel5,16,sizeof(mem_dockpars_prng_states),                &mem_dockpars_prng_states);
  setKernelArg(kernel5,17,sizeof(dockpars.pop_size),                       &dockpars.pop_size);
  setKernelArg(kernel5,18,sizeof(dockpars.num_of_genes),                   &dockpars.num_of_genes);
  setKernelArg(kernel5,19,sizeof(dockpars.lsearch_rate),                   &dockpars.lsearch_rate);
  setKernelArg(kernel5,20,sizeof(dockpars.num_of_lsentities),              &dockpars.num_of_lsentities);
  setKernelArg(kernel5,21,sizeof(dockpars.qasp),                           &dockpars.qasp);
  setKernelArg(kernel5,22,sizeof(mem_atom_charges_const),                  &mem_atom_charges_const);
  setKernelArg(kernel5,23,sizeof(mem_atom_types_const),                    &mem_atom_types_const);
  setKernelArg(kernel5,24,sizeof(mem_intraE_contributors_const),           &mem_intraE_contributors_const);
lvs's avatar
lvs committed
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
  setKernelArg(kernel5,25,sizeof(dockpars.smooth),                        &dockpars.smooth);
  setKernelArg(kernel5,26,sizeof(mem_reqm_const),                         &mem_reqm_const);
  setKernelArg(kernel5,27,sizeof(mem_reqm_hbond_const),                   &mem_reqm_hbond_const);
  setKernelArg(kernel5,28/*25*/,sizeof(mem_VWpars_AC_const),                     &mem_VWpars_AC_const);
  setKernelArg(kernel5,29/*26*/,sizeof(mem_VWpars_BD_const),                     &mem_VWpars_BD_const);
  setKernelArg(kernel5,30/*27*/,sizeof(mem_dspars_S_const),                      &mem_dspars_S_const);
  setKernelArg(kernel5,31/*28*/,sizeof(mem_dspars_V_const),                      &mem_dspars_V_const);
  setKernelArg(kernel5,32/*29*/,sizeof(mem_rotlist_const),                       &mem_rotlist_const);
  setKernelArg(kernel5,33/*30*/,sizeof(mem_ref_coords_x_const),                  &mem_ref_coords_x_const);
  setKernelArg(kernel5,34/*31*/,sizeof(mem_ref_coords_y_const),                  &mem_ref_coords_y_const);
  setKernelArg(kernel5,35/*32*/,sizeof(mem_ref_coords_z_const),                  &mem_ref_coords_z_const);
  setKernelArg(kernel5,36/*33*/,sizeof(mem_rotbonds_moving_vectors_const),       &mem_rotbonds_moving_vectors_const);
  setKernelArg(kernel5,37/*34*/,sizeof(mem_rotbonds_unit_vectors_const),         &mem_rotbonds_unit_vectors_const);
  setKernelArg(kernel5,38/*35*/,sizeof(mem_ref_orientation_quats_const),         &mem_ref_orientation_quats_const);
  setKernelArg(kernel5,39/*36*/,sizeof(mem_rotbonds_const),         		   &mem_rotbonds_const);
  setKernelArg(kernel5,40/*37*/,sizeof(mem_rotbonds_atoms_const),   		   &mem_rotbonds_atoms_const);
  setKernelArg(kernel5,41/*38*/,sizeof(mem_num_rotating_atoms_per_rotbond_const),&mem_num_rotating_atoms_per_rotbond_const);
718
  // Specific gradient-minimizer args
lvs's avatar
lvs committed
719
  setKernelArg(kernel5,42/*39*/,sizeof(gradientpars.max_num_of_iters),           &gradientpars.max_num_of_iters);
720
	/*
Leonardo Solis's avatar
Leonardo Solis committed
721
  setKernelArg(kernel5,40,sizeof(gradientpars.max_num_of_consec_fails),    &gradientpars.max_num_of_consec_fails);
Leonardo Solis's avatar
Leonardo Solis committed
722
723
  setKernelArg(kernel5,39,sizeof(gradientpars.alpha),            	   &gradientpars.alpha);
  setKernelArg(kernel5,40,sizeof(mem_gradpars_conformation_min_perturbation),              &mem_gradpars_conformation_min_perturbation);
724
	*/
725
726
727
  kernel5_gxsize = blocksPerGridForEachGradMinimizerEntity * threadsPerBlock;
  kernel5_lxsize = threadsPerBlock;
#ifdef DOCK_DEBUG
728
729
730
	printf("%-25s %10s %8u %10s %4u\n", "K_GRAD_MINIMIZER: ", "gSize: ", kernel5_gxsize, "lSize: ", kernel5_lxsize); fflush(stdout);
#endif
// End of Kernel5
731

732
733
734
735
/*
#endif
*/
}
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750















751
// Kernel1
752
753
754
	#ifdef DOCK_DEBUG
		printf("Start Kernel1 ... ");fflush(stdout);
	#endif
755
	runKernel1D(command_queue,kernel1,kernel1_gxsize,kernel1_lxsize,&time_start_kernel,&time_end_kernel);
756
757
758
	#ifdef DOCK_DEBUG
		printf(" ... Finish Kernel1\n");fflush(stdout);
	#endif
759
760
761
// End of Kernel1

// Kernel2
762
763
764
	#ifdef DOCK_DEBUG
		printf("Start Kernel2 ... ");fflush(stdout);
	#endif
765
	runKernel1D(command_queue,kernel2,kernel2_gxsize,kernel2_lxsize,&time_start_kernel,&time_end_kernel);
766
767
768
	#ifdef DOCK_DEBUG
		printf(" ... Finish Kernel2\n");fflush(stdout);
	#endif
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
// End of Kernel2
	// ===============================================================================


	// -------- Replacing with memory maps! ------------
#if defined (MAPPED_COPY)
	int* map_cpu_evals_of_runs;
	map_cpu_evals_of_runs = (int*) memMap(command_queue, mem_gpu_evals_of_runs, CL_MAP_READ, size_evals_of_runs);
#else
	memcopyBufferObjectFromDevice(command_queue,cpu_evals_of_runs,mem_gpu_evals_of_runs,size_evals_of_runs);
#endif
	// -------- Replacing with memory maps! ------------


	generation_cnt = 1;


	// -------- Replacing with memory maps! ------------
#if defined (MAPPED_COPY)
	while ((progress = check_progress(map_cpu_evals_of_runs, generation_cnt, mypars->num_of_energy_evals, mypars->num_of_generations, mypars->num_of_runs)) < 100.0)
#else
	while ((progress = check_progress(cpu_evals_of_runs, generation_cnt, mypars->num_of_energy_evals, mypars->num_of_generations, mypars->num_of_runs)) < 100.0)
#endif
	// -------- Replacing with memory maps! ------------

	{
#ifdef DOCK_DEBUG
    ite_cnt++;
    printf("\nIteration # %u\n", ite_cnt);
    fflush(stdout);
#endif

	 //update progress bar (bar length is 50)
	 new_progress_cnt = (int) (progress/2.0+0.5);
	 if (new_progress_cnt > 50)
	 	new_progress_cnt = 50;

	 while (curr_progress_cnt < new_progress_cnt) {
		curr_progress_cnt++;
#ifndef DOCK_DEBUG
		printf("*");
#endif
		fflush(stdout);
	}

// Kernel4
	#ifdef DOCK_DEBUG
		printf("%-25s", "K_GENETIC_GENERATION: ");fflush(stdout);
	#endif
		runKernel1D(command_queue,kernel4,kernel4_gxsize,kernel4_lxsize,&time_start_kernel,&time_end_kernel);
	#ifdef DOCK_DEBUG
		printf("%15s", " ... Finished\n");fflush(stdout);
	#endif
// End of Kernel4

824
825
	if (strcmp(mypars->ls_method, "sw") == 0) {
/*
826
#if !defined (GRADIENT_ENABLED)
827
*/
828
829
830
831
832
833
834
835
836
// Kernel3
	#ifdef DOCK_DEBUG
		printf("%-25s", "K_LOCAL_SEARCH: ");fflush(stdout);
	#endif
		runKernel1D(command_queue,kernel3,kernel3_gxsize,kernel3_lxsize,&time_start_kernel,&time_end_kernel);
	#ifdef DOCK_DEBUG
		printf("%15s" ," ... Finished\n");fflush(stdout);
	#endif
// End of Kernel3
837
838

/*
839
#else
840
841
842
843
*/
	} else if (strcmp(mypars->ls_method, "sd") == 0) {


844
845
846
847
848
849
850
851
852
// Kernel5
	#ifdef DOCK_DEBUG
		printf("%-25s", "K_GRAD_MINIMIZER: ");fflush(stdout);
	#endif
		runKernel1D(command_queue,kernel5,kernel5_gxsize,kernel5_lxsize,&time_start_kernel,&time_end_kernel);
	#ifdef DOCK_DEBUG
		printf("%15s" ," ... Finished\n");fflush(stdout);
	#endif
// End of Kernel5
853
854

/*
855
#endif
856
857
*/
	}
858
859

// Kernel2
860
861
862
	#ifdef DOCK_DEBUG
		printf("Start Kernel2 ... ");fflush(stdout);
	#endif
863
		runKernel1D(command_queue,kernel2,kernel2_gxsize,kernel2_lxsize,&time_start_kernel,&time_end_kernel);
864
865
866
	#ifdef DOCK_DEBUG
		printf(" ... Finish Kernel2\n");fflush(stdout);
	#endif
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
// End of Kernel2
		// ===============================================================================


		// -------- Replacing with memory maps! ------------
#if defined (MAPPED_COPY)
		map_cpu_evals_of_runs = (int*) memMap(command_queue, mem_gpu_evals_of_runs, CL_MAP_READ, size_evals_of_runs);
#else
		memcopyBufferObjectFromDevice(command_queue,cpu_evals_of_runs,mem_gpu_evals_of_runs,size_evals_of_runs);
#endif
		// -------- Replacing with memory maps! ------------

		generation_cnt++;

		// ----------------------------------------------------------------------
		// ORIGINAL APPROACH: switching conformation and energy pointers
		// CURRENT APPROACH:  copy data from one buffer to another, pointers are kept the same
		// IMPROVED CURRENT APPROACH
		// Kernel arguments are changed on every iteration
		// No copy from dev glob memory to dev glob memory occurs
		// Use generation_cnt as it evolves with the main loop
		// No need to use tempfloat
		// No performance improvement wrt to "CURRENT APPROACH"

		// Kernel args exchange regions they point to
		// But never two args point to the same region of dev memory
		// NO ALIASING -> use restrict in Kernel
		if (generation_cnt % 2 == 0) {
			// Kernel 4
896
897
898
899
			setKernelArg(kernel4,13,sizeof(mem_dockpars_conformations_next),                &mem_dockpars_conformations_next);
			setKernelArg(kernel4,14,sizeof(mem_dockpars_energies_next),                     &mem_dockpars_energies_next);
      			setKernelArg(kernel4,15,sizeof(mem_dockpars_conformations_current),             &mem_dockpars_conformations_current);
			setKernelArg(kernel4,16,sizeof(mem_dockpars_energies_current),                  &mem_dockpars_energies_current);
900
/*
901
#if !defined (GRADIENT_ENABLED)
902
903
*/
	if (strcmp(mypars->ls_method, "sw") == 0) {
904
			// Kernel 3
905
906
     			setKernelArg(kernel3,13,sizeof(mem_dockpars_conformations_current),             &mem_dockpars_conformations_current);
      			setKernelArg(kernel3,14,sizeof(mem_dockpars_energies_current),                  &mem_dockpars_energies_current);
907
/*
908
#else
909
910
*/
	} else if (strcmp(mypars->ls_method, "sd") == 0) {
911
			// Kernel 5
912
913
     			setKernelArg(kernel5,13,sizeof(mem_dockpars_conformations_current),             &mem_dockpars_conformations_current);
      			setKernelArg(kernel5,14,sizeof(mem_dockpars_energies_current),                  &mem_dockpars_energies_current);
914
/*
915
#endif
916
917
*/
	}
918
919
920
		}
		else { // In this configuration, the program starts
			// Kernel 4
921
922
923
924
			setKernelArg(kernel4,13,sizeof(mem_dockpars_conformations_current),             &mem_dockpars_conformations_current);
			setKernelArg(kernel4,14,sizeof(mem_dockpars_energies_current),                  &mem_dockpars_energies_current);
      			setKernelArg(kernel4,15,sizeof(mem_dockpars_conformations_next),                &mem_dockpars_conformations_next);
			setKernelArg(kernel4,16,sizeof(mem_dockpars_energies_next),                     &mem_dockpars_energies_next);
925

926
927
		if (strcmp(mypars->ls_method, "sw") == 0) {
/*
928
#if !defined (GRADIENT_ENABLED)
929
*/
930
			// Kernel 3
931
932
			setKernelArg(kernel3,13,sizeof(mem_dockpars_conformations_next),                &mem_dockpars_conformations_next);
      			setKernelArg(kernel3,14,sizeof(mem_dockpars_energies_next),                     &mem_dockpars_energies_next);
933
/*
934
#else
935
936
*/
		} else if (strcmp(mypars->ls_method, "sd") == 0) {
937
			// Kernel 5
938
939
			setKernelArg(kernel5,13,sizeof(mem_dockpars_conformations_next),                &mem_dockpars_conformations_next);
      			setKernelArg(kernel5,14,sizeof(mem_dockpars_energies_next),                     &mem_dockpars_energies_next);
940
/*
941
#endif
942
943
*/
		}
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
		}
		// ----------------------------------------------------------------------

#ifdef DOCK_DEBUG
        printf("Progress %.3f %%\n", progress);
        fflush(stdout);
#endif
	} // End of while-loop


	// -------- Replacing with memory maps! ------------
#if defined (MAPPED_COPY)
	unmemMap(command_queue,mem_gpu_evals_of_runs,map_cpu_evals_of_runs);
#endif
	// -------- Replacing with memory maps! ------------


	clock_stop_docking = clock();

963
	//update progress bar (bar length is 50)mem_num_of_rotatingatoms_per_rotbond_const
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
	while (curr_progress_cnt < 50) {
		curr_progress_cnt++;
		printf("*");
		fflush(stdout);
	}

	printf("\n\n");

	// ===============================================================================
	// OCLADock modification
	// http://www.cc.gatech.edu/~vetter/keeneland/tutorial-2012-02-20/08-opencl.pdf
	// ===============================================================================

	//processing results
	memcopyBufferObjectFromDevice(command_queue,cpu_final_populations,mem_dockpars_conformations_current,size_populations);
	memcopyBufferObjectFromDevice(command_queue,cpu_energies,mem_dockpars_energies_current,size_energies);

#if defined (DOCK_DEBUG)
	for (int cnt_pop=0;cnt_pop<size_populations/sizeof(float);cnt_pop++)
		printf("total_num_pop: %u, cpu_final_populations[%u]: %f\n",(unsigned int)(size_populations/sizeof(float)),cnt_pop,cpu_final_populations[cnt_pop]);
984

985
986
	for (int cnt_pop=0;cnt_pop<size_energies/sizeof(float);cnt_pop++)
		printf("total_num_energies: %u, cpu_energies[%u]: %f\n",    (unsigned int)(size_energies/sizeof(float)),cnt_pop,cpu_energies[cnt_pop]);
987
#endif
988

989
990
991
992
993
994
995
996

	// ===============================================================================


	for (run_cnt=0; run_cnt < mypars->num_of_runs; run_cnt++)
	{
		arrange_result(cpu_final_populations+run_cnt*mypars->pop_size*GENOTYPE_LENGTH_IN_GLOBMEM, cpu_energies+run_cnt*mypars->pop_size, mypars->pop_size);

Leonardo Solis's avatar
Leonardo Solis committed
997
998
999
1000
		make_resfiles(cpu_final_populations+run_cnt*mypars->pop_size*GENOTYPE_LENGTH_IN_GLOBMEM, 
			      cpu_energies+run_cnt*mypars->pop_size, &myligand_reference,
			      myligand_init, 
			      mypars, 
For faster browsing, not all history is shown. View entire blame