(git:374b731)
Loading...
Searching...
No Matches
grid_dgemm_context.c
Go to the documentation of this file.
1/*----------------------------------------------------------------------------*/
2/* CP2K: A general program to perform molecular dynamics simulations */
3/* Copyright 2000-2024 CP2K developers group <https://cp2k.org> */
4/* */
5/* SPDX-License-Identifier: BSD-3-Clause */
6/*----------------------------------------------------------------------------*/
7
8#include <math.h>
9#include <omp.h>
10#include <stdio.h>
11#include <stdlib.h>
12#include <string.h>
13
14#include "../common/grid_library.h"
17#include "grid_dgemm_context.h"
21#include "grid_dgemm_utils.h"
22
23void return_dh(void *const ptr, const int level, double *const dh) {
24 grid_context *const ctx = (grid_context *)ptr;
25
26 assert(ctx->checksum == ctx_checksum);
27 dh[0] = ctx->grid[level].dh[0][0];
28 dh[1] = ctx->grid[level].dh[0][1];
29 dh[2] = ctx->grid[level].dh[0][2];
30 dh[3] = ctx->grid[level].dh[1][0];
31 dh[4] = ctx->grid[level].dh[1][1];
32 dh[5] = ctx->grid[level].dh[1][2];
33 dh[6] = ctx->grid[level].dh[2][0];
34 dh[7] = ctx->grid[level].dh[2][1];
35 dh[8] = ctx->grid[level].dh[2][2];
36}
37
38void return_dh_inv(void *const ptr, const int level, double *const dh_inv) {
39 grid_context *const ctx = (grid_context *)ptr;
40
41 assert(ctx->checksum == ctx_checksum);
42 dh_inv[0] = ctx->grid[level].dh_inv[0][0];
43 dh_inv[1] = ctx->grid[level].dh_inv[0][1];
44 dh_inv[2] = ctx->grid[level].dh_inv[0][2];
45 dh_inv[3] = ctx->grid[level].dh_inv[1][0];
46 dh_inv[4] = ctx->grid[level].dh_inv[1][1];
47 dh_inv[5] = ctx->grid[level].dh_inv[1][2];
48 dh_inv[6] = ctx->grid[level].dh_inv[2][0];
49 dh_inv[7] = ctx->grid[level].dh_inv[2][1];
50 dh_inv[8] = ctx->grid[level].dh_inv[2][2];
51}
52
53int return_num_devs(void *const ptr) {
54 grid_context *const ctx = (grid_context *)ptr;
55 assert(ctx->checksum == ctx_checksum);
56
57 return ctx->number_of_devices;
58}
59
60int return_device_id(void *const ptr, const int device) {
61 grid_context *const ctx = (grid_context *)ptr;
62 assert(ctx->checksum == ctx_checksum);
63
64 return ctx->device_id[device];
65}
66
67int is_grid_orthorhombic(void *const ptr) {
68 grid_context *const ctx = (grid_context *)ptr;
69 assert(ctx->checksum == ctx_checksum);
70 return ctx->orthorhombic;
71}
72
73void update_queue_length(void *const ptr, const int queue_length) {
74 grid_context *const ctx = (grid_context *)ptr;
75 assert(ctx->checksum == ctx_checksum);
76 ctx->queue_length = queue_length;
77}
78
79void update_atoms_position(const int natoms,
80 const double atoms_positions[natoms][3],
81 grid_context *data) {
82 assert(data != NULL);
83
84 if (natoms == 0)
85 return;
86
87 if (data->atom_positions == NULL) {
88 data->atom_positions = malloc(3 * natoms * sizeof(double));
89 } else {
90 if (natoms > data->natoms) {
91 data->atom_positions =
92 realloc(data->atom_positions, 3 * natoms * sizeof(double));
93 }
94 }
95
96 data->natoms = natoms;
97
98 if (data->atom_positions) {
99 for (int i = 0; i < natoms; i++) {
100 data->atom_positions[3 * i] = atoms_positions[i][0];
101 data->atom_positions[3 * i + 1] = atoms_positions[i][1];
102 data->atom_positions[3 * i + 2] = atoms_positions[i][2];
103 }
104 }
105}
106
107void update_atoms_kinds(const int natoms, const int *atoms_kinds,
108 grid_context *data) {
109 assert(data != NULL);
110
111 // data->atom_kinds is a table that give the type of a given atom.
112 if (natoms == 0)
113 return;
114
115 if (data->atom_kinds == NULL) {
116 data->atom_kinds = malloc(natoms * sizeof(int));
117 } else {
118 if ((natoms > data->natoms) && (data->natoms > 0)) {
119 data->atom_kinds = realloc(data->atom_kinds, natoms * sizeof(int));
120 }
121 }
122 // data->natoms is initialized before calling this function
123 if (data->natoms)
124 memcpy(data->atom_kinds, atoms_kinds, sizeof(int) * natoms);
125
126 for (int i = 0; i < natoms; i++) {
127 data->atom_kinds[i] -= 1;
128 }
129}
130
131void update_block_offsets(const int nblocks, const int *const block_offsets,
132 grid_context *data) {
133 assert(data != NULL);
134
135 if (nblocks == 0)
136 return;
137
138 if (data->block_offsets == NULL) {
139 data->block_offsets = malloc(nblocks * sizeof(int));
140 } else {
141 if ((nblocks > data->nblocks_total) && (data->nblocks_total > 0)) {
142 data->block_offsets = realloc(data->block_offsets, sizeof(int) * nblocks);
143 }
144 }
145
146 data->nblocks = nblocks;
147 data->nblocks_total = imax(data->nblocks_total, nblocks);
148 if (nblocks)
149 memcpy(data->block_offsets, block_offsets, nblocks * sizeof(int));
150}
151
152void update_basis_set(const int nkinds, const grid_basis_set **const basis_sets,
153 grid_context *data) {
154 if (nkinds > data->nkinds_total) {
155 if (data->basis_sets == NULL) {
156 data->basis_sets = malloc(nkinds * sizeof(grid_basis_set *));
157 } else {
158 data->basis_sets =
159 realloc(data->basis_sets, nkinds * sizeof(grid_basis_set *));
160 }
161 }
162 data->nkinds = nkinds;
163 data->nkinds_total = imax(data->nkinds_total, nkinds);
164 memcpy(data->basis_sets, basis_sets, nkinds * sizeof(grid_basis_set *));
165}
166
167void update_task_lists(const int nlevels, const int ntasks,
168 const int *const level_list, const int *const iatom_list,
169 const int *const jatom_list, const int *const iset_list,
170 const int *const jset_list, const int *const ipgf_list,
171 const int *const jpgf_list,
172 const int *const border_mask_list,
173 const int *block_num_list,
174 const double *const radius_list,
175 const double rab_list[ntasks][3], grid_context *ctx) {
176
177 assert(ctx->checksum == ctx_checksum);
178
179 if (nlevels == 0)
180 return;
181
182 if (ctx->ntasks == 0) {
183 // Count tasks per level.
184 size_t size = nlevels * sizeof(int);
185 ctx->tasks_per_level = malloc(size);
186 ctx->tasks = malloc(nlevels * sizeof(_task *));
187 /* memset(ctx->tasks, 0, nlevels * sizeof(_task *)); */
188 if (ntasks)
189 ctx->tasks[0] = malloc(ntasks * sizeof(_task));
190 else
191 ctx->tasks[0] = NULL;
192 } else {
193 if (ctx->nlevels_total < nlevels) {
194 /* save the address of the full task list. NULL when completly empty */
195 ctx->tasks = realloc(ctx->tasks, nlevels * sizeof(_task *));
196 }
197 if (ctx->ntasks_total < ntasks) {
198 ctx->tasks[0] = realloc(ctx->tasks[0], ntasks * sizeof(_task));
199 }
200 }
201
202 memset(ctx->tasks_per_level, 0, nlevels * sizeof(int));
203 ctx->nlevels = nlevels;
204 ctx->nlevels_total = imax(ctx->nlevels_total, nlevels);
205 ctx->ntasks_total = imax(ctx->ntasks_total, ntasks);
206 ctx->ntasks = ntasks;
207
208 for (int i = 0; i < ntasks; i++) {
209 ctx->tasks_per_level[level_list[i] - 1]++;
210 assert(i == 0 || level_list[i] >= level_list[i - 1]); // expect ordered list
211 }
212
213 for (int i = 1; i < ctx->nlevels; i++) {
214 ctx->tasks[i] = ctx->tasks[i - 1] + ctx->tasks_per_level[i - 1];
215 }
216
217 int prev_block_num = -1;
218 int prev_iset = -1;
219 int prev_jset = -1;
220 int prev_level = -1;
221 _task *task = ctx->tasks[0];
222 for (int i = 0; i < ntasks; i++) {
223 if (prev_level != (level_list[i] - 1)) {
224 prev_level = level_list[i] - 1;
225 prev_block_num = -1;
226 prev_iset = -1;
227 prev_jset = -1;
228 }
229 task->level = level_list[i] - 1;
230 task->iatom = iatom_list[i] - 1;
231 task->jatom = jatom_list[i] - 1;
232 task->iset = iset_list[i] - 1;
233 task->jset = jset_list[i] - 1;
234 task->ipgf = ipgf_list[i] - 1;
235 task->jpgf = jpgf_list[i] - 1;
236 task->border_mask = border_mask_list[i];
237 task->block_num = block_num_list[i] - 1;
238 task->radius = radius_list[i];
239 task->rab[0] = rab_list[i][0];
240 task->rab[1] = rab_list[i][1];
241 task->rab[2] = rab_list[i][2];
242 const int iatom = task->iatom;
243 const int jatom = task->jatom;
244 const int iset = task->iset;
245 const int jset = task->jset;
246 const int ipgf = task->ipgf;
247 const int jpgf = task->jpgf;
248 const int ikind = ctx->atom_kinds[iatom];
249 const int jkind = ctx->atom_kinds[jatom];
250 const grid_basis_set *ibasis = ctx->basis_sets[ikind];
251 const grid_basis_set *jbasis = ctx->basis_sets[jkind];
252 const int ncoseta = ncoset(ibasis->lmax[iset]);
253 const int ncosetb = ncoset(jbasis->lmax[jset]);
254
255 task->zeta[0] = ibasis->zet[iset * ibasis->maxpgf + ipgf];
256 task->zeta[1] = jbasis->zet[jset * jbasis->maxpgf + jpgf];
257
258 const double *ra = &ctx->atom_positions[3 * iatom];
259 const double zetp = task->zeta[0] + task->zeta[1];
260 const double f = task->zeta[1] / zetp;
261 const double rab2 = task->rab[0] * task->rab[0] +
262 task->rab[1] * task->rab[1] +
263 task->rab[2] * task->rab[2];
264
265 task->prefactor = exp(-task->zeta[0] * f * rab2);
266 task->zetp = zetp;
267
268 const int block_num = task->block_num;
269
270 for (int i = 0; i < 3; i++) {
271 task->ra[i] = ra[i];
272 task->rp[i] = ra[i] + f * task->rab[i];
273 task->rb[i] = ra[i] + task->rab[i];
274 }
275
276 task->lmax[0] = ibasis->lmax[iset];
277 task->lmax[1] = jbasis->lmax[jset];
278 task->lmin[0] = ibasis->lmin[iset];
279 task->lmin[1] = jbasis->lmin[jset];
280
281 if ((block_num != prev_block_num) || (iset != prev_iset) ||
282 (jset != prev_jset)) {
283 task->update_block_ = true;
284 prev_block_num = block_num;
285 prev_iset = iset;
286 prev_jset = jset;
287 } else {
288 task->update_block_ = false;
289 }
290
291 task->offset[0] = ipgf * ncoseta;
292 task->offset[1] = jpgf * ncosetb;
293 task++;
294 }
295
296 // Find largest Cartesian subblock size.
297 ctx->maxco = 0;
298 for (int i = 0; i < ctx->nkinds; i++) {
299 ctx->maxco = imax(ctx->maxco, ctx->basis_sets[i]->maxco);
300 }
301}
302
303void update_layouts(const int nlevels, const int npts_global[nlevels][3],
304 const int npts_local[nlevels][3],
305 const int shift_local[nlevels][3],
306 const int border_width[nlevels][3],
307 const double dh[nlevels][3][3],
308 const double dh_inv[nlevels][3][3], grid_context *ctx) {
309
310 assert(ctx != NULL);
311 assert(ctx->checksum == ctx_checksum);
312
313 if (ctx->layouts != NULL) {
314 free(ctx->layouts);
315 }
316
317 ctx->layouts = malloc(sizeof(_layout) * nlevels);
318
319 for (int level = 0; level < nlevels; level++) {
320 for (int i = 0; i < 3; i++) {
321 ctx->layouts[level].npts_global[i] = npts_global[level][i];
322 ctx->layouts[level].npts_local[i] = npts_local[level][i];
323 ctx->layouts[level].shift_local[i] = shift_local[level][i];
324 ctx->layouts[level].border_width[i] = border_width[level][i];
325 for (int j = 0; j < 3; j++) {
326 ctx->layouts[level].dh[i][j] = dh[level][i][j];
327 ctx->layouts[level].dh_inv[i][j] = dh_inv[level][i][j];
328 }
329 }
330 }
331}
332
333void update_grid(const int nlevels, grid_context *ctx) {
334 assert(ctx != NULL);
335 assert(ctx->checksum == ctx_checksum);
336
337 if (nlevels == 0)
338 return;
339
340 if (ctx->grid == NULL) {
341 ctx->grid = malloc(sizeof(tensor) * nlevels);
342 } else {
343 if (ctx->nlevels_total < nlevels) {
344 ctx->grid = realloc(ctx->grid, sizeof(tensor) * nlevels);
345 }
346 }
347
348 ctx->nlevels_total = imax(ctx->nlevels_total, nlevels);
349 ctx->nlevels = nlevels;
350}
351
353 const bool orthorhombic, const int ntasks, const int nlevels,
354 const int natoms, const int nkinds, const int nblocks,
355 const int *block_offsets, const double atom_positions[natoms][3],
356 const int *const atom_kinds, const grid_basis_set **const basis_sets,
357 const int *const level_list, const int *const iatom_list,
358 const int *jatom_list, const int *const iset_list,
359 const int *const jset_list, const int *const ipgf_list,
360 const int *const jpgf_list, const int *const border_mask_list,
361 const int *block_num_list, const double *const radius_list,
362 const double rab_list[ntasks][3], const int npts_global[nlevels][3],
363 const int npts_local[nlevels][3], const int shift_local[nlevels][3],
364 const int border_width[nlevels][3], const double dh[nlevels][3][3],
365 const double dh_inv[nlevels][3][3]) {
366
367 grid_context *ctx = malloc(sizeof(grid_context));
368
369 memset(ctx, 0, sizeof(grid_context));
370
371 ctx->checksum = ctx_checksum;
372 ctx->orthorhombic = orthorhombic;
373 update_block_offsets(nblocks, block_offsets, ctx);
374 update_atoms_position(natoms, atom_positions, ctx);
375 update_atoms_kinds(natoms, atom_kinds, ctx);
376 update_basis_set(nkinds, basis_sets, ctx);
377 update_task_lists(nlevels, ntasks, level_list, iatom_list, jatom_list,
378 iset_list, jset_list, ipgf_list, jpgf_list,
379 border_mask_list, block_num_list, radius_list, rab_list,
380 ctx);
381 update_layouts(nlevels, npts_global, npts_local, shift_local, border_width,
382 dh, dh_inv, ctx);
383 update_grid(nlevels, ctx);
384
385 const int max_threads = omp_get_max_threads();
386
387 ctx->handler =
388 malloc(sizeof(struct collocation_integration_ *) * max_threads);
389
390 for (int i = 0; i < max_threads; i++) {
392 }
393
395
396 return ctx;
397}
398
400 const bool orthorhombic, const int ntasks, const int nlevels,
401 const int natoms, const int nkinds, const int nblocks,
402 const int *block_offsets, const double atom_positions[natoms][3],
403 const int *const atom_kinds, const grid_basis_set **const basis_sets,
404 const int *const level_list, const int *const iatom_list,
405 const int *jatom_list, const int *const iset_list,
406 const int *const jset_list, const int *const ipgf_list,
407 const int *const jpgf_list, const int *const border_mask_list,
408 const int *block_num_list, const double *const radius_list,
409 const double rab_list[ntasks][3], const int npts_global[nlevels][3],
410 const int npts_local[nlevels][3], const int shift_local[nlevels][3],
411 const int border_width[nlevels][3], const double dh[nlevels][3][3],
412 const double dh_inv[nlevels][3][3], void *ptr) {
413
414 assert(ptr != NULL);
415 grid_context *ctx = (grid_context *)ptr;
416 assert(ctx->checksum == ctx_checksum);
417
418 ctx->orthorhombic = orthorhombic;
419 update_block_offsets(nblocks, block_offsets, ctx);
420 update_atoms_position(natoms, atom_positions, ctx);
421 update_atoms_kinds(natoms, atom_kinds, ctx);
422 update_basis_set(nkinds, basis_sets, ctx);
423 update_task_lists(nlevels, ntasks, level_list, iatom_list, jatom_list,
424 iset_list, jset_list, ipgf_list, jpgf_list,
425 border_mask_list, block_num_list, radius_list, rab_list,
426 ctx);
427 update_layouts(nlevels, npts_global, npts_local, shift_local, border_width,
428 dh, dh_inv, ctx);
429 update_grid(nlevels, ctx);
430
431 // Find largest Cartesian subblock size.
432 ctx->maxco = 0;
433 for (int i = 0; i < nkinds; i++) {
434 ctx->maxco = imax(ctx->maxco, ctx->basis_sets[i]->maxco);
435 }
436}
437
438void initialize_grid_context_on_gpu(void *ptr, const int number_of_devices,
439 const int *device_id) {
440 assert(ptr != NULL);
441 grid_context *ctx = (grid_context *)ptr;
442 assert(ctx->checksum == ctx_checksum);
443 ctx->work_on_gpu = false;
444 if (number_of_devices <= 0) {
445 return;
446 }
447
448 ctx->number_of_devices = number_of_devices;
449 ctx->queue_length = 8192;
450 if (ctx->device_id == NULL)
451 ctx->device_id = malloc(sizeof(int) * number_of_devices);
452 else
453 ctx->device_id = realloc(ctx->device_id, sizeof(int) * number_of_devices);
454
455 memcpy(ctx->device_id, device_id, sizeof(int) * number_of_devices);
456}
457
459 assert(ptr);
460 grid_context *ctx = (grid_context *)ptr;
461 assert(ctx->checksum == ctx_checksum);
462 free(ctx->block_offsets);
463 free(ctx->atom_positions);
464 free(ctx->atom_kinds);
465 free(ctx->basis_sets);
466 free(ctx->tasks[0]);
467 free(ctx->tasks);
468 free(ctx->tasks_per_level);
469 free(ctx->layouts);
470 free(ctx->grid);
471 if (ctx->device_id)
472 free(ctx->device_id);
473
474 if (ctx->handler) {
475 for (int i = 0; i < ctx->number_of_handler; i++) {
477 }
478 free(ctx->handler);
479 }
480
481 free(ctx);
482}
483
484void apply_cutoff(void *ptr) {
485 assert(ptr);
486 grid_context *ctx = (grid_context *)ptr;
487 assert(ctx->checksum == ctx_checksum);
488 ctx->apply_cutoff = true;
489}
490
492 tensor *grid, const bool orthorhombic,
493 const int grid_full_size[3], /* size of the full grid */
494 const int grid_local_size[3], /* size of the local grid block */
495 const int shift_local[3], /* coordinates of the lower coordinates of the
496 local grid window */
497 const int border_width[3], /* width of the borders */
498 const double
499 dh[3][3], /* displacement vectors of the grid (cartesian) -> (ijk) */
500 const double dh_inv[3][3], /* (ijk) -> (x,y,z) */
501 offload_buffer *grid_) {
502 memset(grid, 0, sizeof(tensor));
503 initialize_tensor_3(grid, grid_local_size[2], grid_local_size[1],
504 grid_local_size[0]);
505
506 grid->data = grid_->host_buffer;
507 grid->ld_ = grid_local_size[0];
508
509 setup_global_grid_size(grid, &grid_full_size[0]);
510
511 /* the grid is divided over several ranks or not periodic */
512 if ((grid_local_size[0] != grid_full_size[0]) ||
513 (grid_local_size[1] != grid_full_size[1]) ||
514 (grid_local_size[2] != grid_full_size[2])) {
515 setup_grid_window(grid, shift_local, border_width, 0);
516 } else {
517 grid->window_shift[0] = 0;
518 grid->window_shift[1] = 0;
519 grid->window_shift[2] = 0;
520
521 grid->window_size[0] = grid->size[0];
522 grid->window_size[1] = grid->size[1];
523 grid->window_size[2] = grid->size[2];
524 }
525
526 grid->dh[0][0] = dh[0][0];
527 grid->dh[0][1] = dh[0][1];
528 grid->dh[0][2] = dh[0][2];
529 grid->dh[1][0] = dh[1][0];
530 grid->dh[1][1] = dh[1][1];
531 grid->dh[1][2] = dh[1][2];
532 grid->dh[2][0] = dh[2][0];
533 grid->dh[2][1] = dh[2][1];
534 grid->dh[2][2] = dh[2][2];
535
536 grid->dh_inv[0][0] = dh_inv[0][0];
537 grid->dh_inv[0][1] = dh_inv[0][1];
538 grid->dh_inv[0][2] = dh_inv[0][2];
539 grid->dh_inv[1][0] = dh_inv[1][0];
540 grid->dh_inv[1][1] = dh_inv[1][1];
541 grid->dh_inv[1][2] = dh_inv[1][2];
542 grid->dh_inv[2][0] = dh_inv[2][0];
543 grid->dh_inv[2][1] = dh_inv[2][1];
544 grid->dh_inv[2][2] = dh_inv[2][2];
545
546 verify_orthogonality(dh, grid->orthogonal);
547
548 if (orthorhombic) {
549 grid->orthogonal[0] = true;
550 grid->orthogonal[1] = true;
551 grid->orthogonal[2] = true;
552 }
553}
554
555/*******************************************************************************
556 * \brief Allocates a task list for the dgemm backend.
557 * See grid_task_list.h for details.
558 ******************************************************************************/
560 const bool orthorhombic, const int ntasks, const int nlevels,
561 const int natoms, const int nkinds, const int nblocks,
562 const int block_offsets[nblocks], const double atom_positions[natoms][3],
563 const int atom_kinds[natoms], const grid_basis_set *basis_sets[nkinds],
564 const int level_list[ntasks], const int iatom_list[ntasks],
565 const int jatom_list[ntasks], const int iset_list[ntasks],
566 const int jset_list[ntasks], const int ipgf_list[ntasks],
567 const int jpgf_list[ntasks], const int border_mask_list[ntasks],
568 const int block_num_list[ntasks], const double radius_list[ntasks],
569 const double rab_list[ntasks][3], const int npts_global[nlevels][3],
570 const int npts_local[nlevels][3], const int shift_local[nlevels][3],
571 const int border_width[nlevels][3], const double dh[nlevels][3][3],
572 const double dh_inv[nlevels][3][3], grid_dgemm_task_list **task_list) {
573
574 if (*task_list == NULL) {
575 *task_list = create_grid_context_dgemm(
576 orthorhombic, ntasks, nlevels, natoms, nkinds, nblocks, block_offsets,
577 atom_positions, atom_kinds, basis_sets, level_list, iatom_list,
578 jatom_list, iset_list, jset_list, ipgf_list, jpgf_list,
579 border_mask_list, block_num_list, radius_list, rab_list, npts_global,
580 npts_local, shift_local, border_width, dh, dh_inv);
581 } else {
583 orthorhombic, ntasks, nlevels, natoms, nkinds, nblocks, block_offsets,
584 atom_positions, atom_kinds, basis_sets, level_list, iatom_list,
585 jatom_list, iset_list, jset_list, ipgf_list, jpgf_list,
586 border_mask_list, block_num_list, radius_list, rab_list, npts_global,
587 npts_local, shift_local, border_width, dh, dh_inv, *task_list);
588 }
589
591 if (config.apply_cutoff) {
592 apply_cutoff(*task_list);
593 }
594}
595
596/*******************************************************************************
597 * \brief Deallocates given task list, basis_sets have to be freed separately.
598 ******************************************************************************/
static int imax(int x, int y)
Returns the larger of two given integer (missing from the C standard)
static int max_threads
Definition dbm_library.c:24
static void const int const int const int const int const int const double const int const int const int int GRID_CONST_WHEN_COLLOCATE double GRID_CONST_WHEN_INTEGRATE double * grid
static void const int const int i
static void const int const int const int const int const int const double const int const int const int npts_local[3]
void collocate_destroy_handle(void *gaussian_handle)
struct collocation_integration_ * collocate_create_handle(void)
void update_layouts(const int nlevels, const int npts_global[nlevels][3], const int npts_local[nlevels][3], const int shift_local[nlevels][3], const int border_width[nlevels][3], const double dh[nlevels][3][3], const double dh_inv[nlevels][3][3], grid_context *ctx)
void update_grid(const int nlevels, grid_context *ctx)
void update_atoms_kinds(const int natoms, const int *atoms_kinds, grid_context *data)
void update_grid_context_dgemm(const bool orthorhombic, const int ntasks, const int nlevels, const int natoms, const int nkinds, const int nblocks, const int *block_offsets, const double atom_positions[natoms][3], const int *const atom_kinds, const grid_basis_set **const basis_sets, const int *const level_list, const int *const iatom_list, const int *jatom_list, const int *const iset_list, const int *const jset_list, const int *const ipgf_list, const int *const jpgf_list, const int *const border_mask_list, const int *block_num_list, const double *const radius_list, const double rab_list[ntasks][3], const int npts_global[nlevels][3], const int npts_local[nlevels][3], const int shift_local[nlevels][3], const int border_width[nlevels][3], const double dh[nlevels][3][3], const double dh_inv[nlevels][3][3], void *ptr)
void * create_grid_context_dgemm(const bool orthorhombic, const int ntasks, const int nlevels, const int natoms, const int nkinds, const int nblocks, const int *block_offsets, const double atom_positions[natoms][3], const int *const atom_kinds, const grid_basis_set **const basis_sets, const int *const level_list, const int *const iatom_list, const int *jatom_list, const int *const iset_list, const int *const jset_list, const int *const ipgf_list, const int *const jpgf_list, const int *const border_mask_list, const int *block_num_list, const double *const radius_list, const double rab_list[ntasks][3], const int npts_global[nlevels][3], const int npts_local[nlevels][3], const int shift_local[nlevels][3], const int border_width[nlevels][3], const double dh[nlevels][3][3], const double dh_inv[nlevels][3][3])
int is_grid_orthorhombic(void *const ptr)
void update_block_offsets(const int nblocks, const int *const block_offsets, grid_context *data)
int return_device_id(void *const ptr, const int device)
void destroy_grid_context_dgemm(void *ptr)
void return_dh(void *const ptr, const int level, double *const dh)
void update_task_lists(const int nlevels, const int ntasks, const int *const level_list, const int *const iatom_list, const int *const jatom_list, const int *const iset_list, const int *const jset_list, const int *const ipgf_list, const int *const jpgf_list, const int *const border_mask_list, const int *block_num_list, const double *const radius_list, const double rab_list[ntasks][3], grid_context *ctx)
void apply_cutoff(void *ptr)
int return_num_devs(void *const ptr)
void return_dh_inv(void *const ptr, const int level, double *const dh_inv)
void update_basis_set(const int nkinds, const grid_basis_set **const basis_sets, grid_context *data)
void grid_dgemm_create_task_list(const bool orthorhombic, const int ntasks, const int nlevels, const int natoms, const int nkinds, const int nblocks, const int block_offsets[nblocks], const double atom_positions[natoms][3], const int atom_kinds[natoms], const grid_basis_set *basis_sets[nkinds], const int level_list[ntasks], const int iatom_list[ntasks], const int jatom_list[ntasks], const int iset_list[ntasks], const int jset_list[ntasks], const int ipgf_list[ntasks], const int jpgf_list[ntasks], const int border_mask_list[ntasks], const int block_num_list[ntasks], const double radius_list[ntasks], const double rab_list[ntasks][3], const int npts_global[nlevels][3], const int npts_local[nlevels][3], const int shift_local[nlevels][3], const int border_width[nlevels][3], const double dh[nlevels][3][3], const double dh_inv[nlevels][3][3], grid_dgemm_task_list **task_list)
Allocates a task list for the dgemm backend. See grid_task_list.h for details.
void update_atoms_position(const int natoms, const double atoms_positions[natoms][3], grid_context *data)
void grid_dgemm_free_task_list(grid_dgemm_task_list *task_list)
Deallocates given task list, basis_sets have to be freed separately.
void update_queue_length(void *const ptr, const int queue_length)
void set_grid_parameters(tensor *grid, const bool orthorhombic, const int grid_full_size[3], const int grid_local_size[3], const int shift_local[3], const int border_width[3], const double dh[3][3], const double dh_inv[3][3], offload_buffer *grid_)
void initialize_grid_context_on_gpu(void *ptr, const int number_of_devices, const int *device_id)
static void setup_grid_window(tensor *const grid, const int *const shift_local, const int *const border_width, const int border_mask)
static void setup_global_grid_size(tensor *const grid, const int *const full_size)
static void initialize_tensor_3(struct tensor_ *a, int n1, int n2, int n3)
void verify_orthogonality(const double dh[3][3], bool orthogonal[3])
static grid_library_config config
grid_library_config grid_library_get_config(void)
Returns the library config.
double dh_inv[3][3]
Internal representation of a basis set.
grid_basis_set ** basis_sets
struct collocation_integration_ ** handler
Configuration of the grid library.
Internal representation of a buffer.
double * host_buffer
double dh[3][3]
double dh_inv[3][3]