1 // Copyright (c) 2017-2022, Lawrence Livermore National Security, LLC and other 2 // CEED contributors. All Rights Reserved. See the top-level LICENSE and NOTICE 3 // files for details. 4 // 5 // SPDX-License-Identifier: BSD-2-Clause 6 // 7 // This file is part of CEED: http://github.com/ceed 8 9 #include <ceed/backend.h> 10 #include <ceed/ceed.h> 11 12 #include <cassert> 13 #include <string> 14 #include <sycl/sycl.hpp> 15 16 #include "../sycl/ceed-sycl-compile.hpp" 17 #include "ceed-sycl-ref.hpp" 18 19 class CeedOperatorSyclLinearDiagonal; 20 class CeedOperatorSyclLinearAssemble; 21 class CeedOperatorSyclLinearAssembleFallback; 22 23 //------------------------------------------------------------------------------ 24 // Get Basis Emode Pointer 25 //------------------------------------------------------------------------------ 26 void CeedOperatorGetBasisPointer_Sycl(const CeedScalar **basis_ptr, CeedEvalMode e_mode, const CeedScalar *identity, const CeedScalar *interp, 27 const CeedScalar *grad) { 28 switch (e_mode) { 29 case CEED_EVAL_NONE: 30 *basis_ptr = identity; 31 break; 32 case CEED_EVAL_INTERP: 33 *basis_ptr = interp; 34 break; 35 case CEED_EVAL_GRAD: 36 *basis_ptr = grad; 37 break; 38 case CEED_EVAL_WEIGHT: 39 case CEED_EVAL_DIV: 40 case CEED_EVAL_CURL: 41 break; // Caught by QF Assembly 42 } 43 } 44 45 //------------------------------------------------------------------------------ 46 // Destroy operator 47 //------------------------------------------------------------------------------ 48 static int CeedOperatorDestroy_Sycl(CeedOperator op) { 49 Ceed ceed; 50 Ceed_Sycl *sycl_data; 51 CeedOperator_Sycl *impl; 52 53 CeedCallBackend(CeedOperatorGetData(op, &impl)); 54 CeedCallBackend(CeedOperatorGetCeed(op, &ceed)); 55 CeedCallBackend(CeedGetData(ceed, &sycl_data)); 56 57 // Apply data 58 for (CeedInt i = 0; i < impl->num_e_in + impl->num_e_out; i++) { 59 CeedCallBackend(CeedVectorDestroy(&impl->e_vecs[i])); 60 } 61 CeedCallBackend(CeedFree(&impl->e_vecs)); 62 63 for (CeedInt i = 0; i < impl->num_e_in; i++) { 64 CeedCallBackend(CeedVectorDestroy(&impl->q_vecs_in[i])); 65 } 66 CeedCallBackend(CeedFree(&impl->q_vecs_in)); 67 68 for (CeedInt i = 0; i < impl->num_e_out; i++) { 69 CeedCallBackend(CeedVectorDestroy(&impl->q_vecs_out[i])); 70 } 71 CeedCallBackend(CeedFree(&impl->q_vecs_out)); 72 73 // QFunction assembly data 74 for (CeedInt i = 0; i < impl->num_active_in; i++) { 75 CeedCallBackend(CeedVectorDestroy(&impl->qf_active_in[i])); 76 } 77 CeedCallBackend(CeedFree(&impl->qf_active_in)); 78 79 // Diag data 80 if (impl->diag) { 81 CeedCallBackend(CeedFree(&impl->diag->h_e_mode_in)); 82 CeedCallBackend(CeedFree(&impl->diag->h_e_mode_out)); 83 84 CeedCallSycl(ceed, sycl_data->sycl_queue.wait_and_throw()); 85 CeedCallSycl(ceed, sycl::free(impl->diag->d_e_mode_in, sycl_data->sycl_context)); 86 CeedCallSycl(ceed, sycl::free(impl->diag->d_e_mode_out, sycl_data->sycl_context)); 87 CeedCallSycl(ceed, sycl::free(impl->diag->d_identity, sycl_data->sycl_context)); 88 CeedCallSycl(ceed, sycl::free(impl->diag->d_interp_in, sycl_data->sycl_context)); 89 CeedCallSycl(ceed, sycl::free(impl->diag->d_interp_out, sycl_data->sycl_context)); 90 CeedCallSycl(ceed, sycl::free(impl->diag->d_grad_in, sycl_data->sycl_context)); 91 CeedCallSycl(ceed, sycl::free(impl->diag->d_grad_out, sycl_data->sycl_context)); 92 CeedCallBackend(CeedElemRestrictionDestroy(&impl->diag->point_block_diag_rstr)); 93 94 CeedCallBackend(CeedVectorDestroy(&impl->diag->elem_diag)); 95 CeedCallBackend(CeedVectorDestroy(&impl->diag->point_block_elem_diag)); 96 } 97 CeedCallBackend(CeedFree(&impl->diag)); 98 99 if (impl->asmb) { 100 CeedCallSycl(ceed, sycl_data->sycl_queue.wait_and_throw()); 101 CeedCallSycl(ceed, sycl::free(impl->asmb->d_B_in, sycl_data->sycl_context)); 102 CeedCallSycl(ceed, sycl::free(impl->asmb->d_B_out, sycl_data->sycl_context)); 103 } 104 CeedCallBackend(CeedFree(&impl->asmb)); 105 106 CeedCallBackend(CeedFree(&impl)); 107 return CEED_ERROR_SUCCESS; 108 } 109 110 //------------------------------------------------------------------------------ 111 // Setup infields or outfields 112 //------------------------------------------------------------------------------ 113 static int CeedOperatorSetupFields_Sycl(CeedQFunction qf, CeedOperator op, bool is_input, CeedVector *e_vecs, CeedVector *q_vecs, CeedInt start_e, 114 CeedInt num_fields, CeedInt Q, CeedInt num_elem) { 115 Ceed ceed; 116 CeedSize q_size; 117 bool is_strided, skip_restriction; 118 CeedInt dim, size; 119 CeedOperatorField *op_fields; 120 CeedQFunctionField *qf_fields; 121 122 CeedCallBackend(CeedOperatorGetCeed(op, &ceed)); 123 if (is_input) { 124 CeedCallBackend(CeedOperatorGetFields(op, NULL, &op_fields, NULL, NULL)); 125 CeedCallBackend(CeedQFunctionGetFields(qf, NULL, &qf_fields, NULL, NULL)); 126 } else { 127 CeedCallBackend(CeedOperatorGetFields(op, NULL, NULL, NULL, &op_fields)); 128 CeedCallBackend(CeedQFunctionGetFields(qf, NULL, NULL, NULL, &qf_fields)); 129 } 130 131 // Loop over fields 132 for (CeedInt i = 0; i < num_fields; i++) { 133 CeedEvalMode e_mode; 134 CeedVector vec; 135 CeedElemRestriction rstr; 136 CeedBasis basis; 137 138 CeedCallBackend(CeedQFunctionFieldGetEvalMode(qf_fields[i], &e_mode)); 139 140 is_strided = false; 141 skip_restriction = false; 142 if (e_mode != CEED_EVAL_WEIGHT) { 143 CeedCallBackend(CeedOperatorFieldGetElemRestriction(op_fields[i], &rstr)); 144 145 // Check whether this field can skip the element restriction: 146 // must be passive input, with e_mode NONE, and have a strided restriction with CEED_STRIDES_BACKEND. 147 148 // First, check whether the field is input or output: 149 if (is_input) { 150 // Check for passive input: 151 CeedCallBackend(CeedOperatorFieldGetVector(op_fields[i], &vec)); 152 if (vec != CEED_VECTOR_ACTIVE) { 153 // Check e_mode 154 if (e_mode == CEED_EVAL_NONE) { 155 // Check for is_strided restriction 156 CeedCallBackend(CeedElemRestrictionIsStrided(rstr, &is_strided)); 157 if (is_strided) { 158 // Check if vector is already in preferred backend ordering 159 CeedCallBackend(CeedElemRestrictionHasBackendStrides(rstr, &skip_restriction)); 160 } 161 } 162 } 163 } 164 if (skip_restriction) { 165 // We do not need an E-Vector, but will use the input field vector's data directly in the operator application 166 e_vecs[i + start_e] = NULL; 167 } else { 168 CeedCallBackend(CeedElemRestrictionCreateVector(rstr, NULL, &e_vecs[i + start_e])); 169 } 170 } 171 172 switch (e_mode) { 173 case CEED_EVAL_NONE: 174 CeedCallBackend(CeedQFunctionFieldGetSize(qf_fields[i], &size)); 175 q_size = (CeedSize)num_elem * Q * size; 176 CeedCallBackend(CeedVectorCreate(ceed, q_size, &q_vecs[i])); 177 break; 178 case CEED_EVAL_INTERP: 179 CeedCallBackend(CeedQFunctionFieldGetSize(qf_fields[i], &size)); 180 q_size = (CeedSize)num_elem * Q * size; 181 CeedCallBackend(CeedVectorCreate(ceed, q_size, &q_vecs[i])); 182 break; 183 case CEED_EVAL_GRAD: 184 CeedCallBackend(CeedOperatorFieldGetBasis(op_fields[i], &basis)); 185 CeedCallBackend(CeedQFunctionFieldGetSize(qf_fields[i], &size)); 186 CeedCallBackend(CeedBasisGetDimension(basis, &dim)); 187 q_size = (CeedSize)num_elem * Q * size; 188 CeedCallBackend(CeedVectorCreate(ceed, q_size, &q_vecs[i])); 189 break; 190 case CEED_EVAL_WEIGHT: // Only on input fields 191 CeedCallBackend(CeedOperatorFieldGetBasis(op_fields[i], &basis)); 192 q_size = (CeedSize)num_elem * Q; 193 CeedCallBackend(CeedVectorCreate(ceed, q_size, &q_vecs[i])); 194 CeedCallBackend(CeedBasisApply(basis, num_elem, CEED_NOTRANSPOSE, CEED_EVAL_WEIGHT, NULL, q_vecs[i])); 195 break; 196 case CEED_EVAL_DIV: 197 break; // TODO: Not implemented 198 case CEED_EVAL_CURL: 199 break; // TODO: Not implemented 200 } 201 } 202 return CEED_ERROR_SUCCESS; 203 } 204 205 //------------------------------------------------------------------------------ 206 // CeedOperator needs to connect all the named fields (be they active or 207 // passive) to the named inputs and outputs of its CeedQFunction. 208 //------------------------------------------------------------------------------ 209 static int CeedOperatorSetup_Sycl(CeedOperator op) { 210 Ceed ceed; 211 bool is_setup_done; 212 CeedInt Q, num_elem, num_input_fields, num_output_fields; 213 CeedQFunctionField *qf_input_fields, *qf_output_fields; 214 CeedQFunction qf; 215 CeedOperatorField *op_input_fields, *op_output_fields; 216 CeedOperator_Sycl *impl; 217 218 CeedCallBackend(CeedOperatorIsSetupDone(op, &is_setup_done)); 219 if (is_setup_done) return CEED_ERROR_SUCCESS; 220 221 CeedCallBackend(CeedOperatorGetCeed(op, &ceed)); 222 CeedCallBackend(CeedOperatorGetData(op, &impl)); 223 CeedCallBackend(CeedOperatorGetQFunction(op, &qf)); 224 CeedCallBackend(CeedOperatorGetNumQuadraturePoints(op, &Q)); 225 CeedCallBackend(CeedOperatorGetNumElements(op, &num_elem)); 226 CeedCallBackend(CeedOperatorGetFields(op, &num_input_fields, &op_input_fields, &num_output_fields, &op_output_fields)); 227 CeedCallBackend(CeedQFunctionGetFields(qf, NULL, &qf_input_fields, NULL, &qf_output_fields)); 228 229 // Allocate 230 CeedCallBackend(CeedCalloc(num_input_fields + num_output_fields, &impl->e_vecs)); 231 232 CeedCallBackend(CeedCalloc(CEED_FIELD_MAX, &impl->q_vecs_in)); 233 CeedCallBackend(CeedCalloc(CEED_FIELD_MAX, &impl->q_vecs_out)); 234 235 impl->num_e_in = num_input_fields; 236 impl->num_e_out = num_output_fields; 237 238 // Set up infield and outfield e_vecs and q_vecs 239 // Infields 240 CeedCallBackend(CeedOperatorSetupFields_Sycl(qf, op, true, impl->e_vecs, impl->q_vecs_in, 0, num_input_fields, Q, num_elem)); 241 // Outfields 242 CeedCallBackend(CeedOperatorSetupFields_Sycl(qf, op, false, impl->e_vecs, impl->q_vecs_out, num_input_fields, num_output_fields, Q, num_elem)); 243 244 CeedCallBackend(CeedOperatorSetSetupDone(op)); 245 return CEED_ERROR_SUCCESS; 246 } 247 248 //------------------------------------------------------------------------------ 249 // Setup Operator Inputs 250 //------------------------------------------------------------------------------ 251 static inline int CeedOperatorSetupInputs_Sycl(CeedInt num_input_fields, CeedQFunctionField *qf_input_fields, CeedOperatorField *op_input_fields, 252 CeedVector in_vec, const bool skip_active, CeedScalar *e_data[2 * CEED_FIELD_MAX], 253 CeedOperator_Sycl *impl, CeedRequest *request) { 254 for (CeedInt i = 0; i < num_input_fields; i++) { 255 CeedEvalMode e_mode; 256 CeedVector vec; 257 CeedElemRestriction rstr; 258 259 // Get input vector 260 CeedCallBackend(CeedOperatorFieldGetVector(op_input_fields[i], &vec)); 261 if (vec == CEED_VECTOR_ACTIVE) { 262 if (skip_active) continue; 263 else vec = in_vec; 264 } 265 266 CeedCallBackend(CeedQFunctionFieldGetEvalMode(qf_input_fields[i], &e_mode)); 267 if (e_mode == CEED_EVAL_WEIGHT) { // Skip 268 } else { 269 // Get input vector 270 CeedCallBackend(CeedOperatorFieldGetVector(op_input_fields[i], &vec)); 271 // Get input element restriction 272 CeedCallBackend(CeedOperatorFieldGetElemRestriction(op_input_fields[i], &rstr)); 273 if (vec == CEED_VECTOR_ACTIVE) vec = in_vec; 274 // Restrict, if necessary 275 if (!impl->e_vecs[i]) { 276 // No restriction for this field; read data directly from vec. 277 CeedCallBackend(CeedVectorGetArrayRead(vec, CEED_MEM_DEVICE, (const CeedScalar **)&e_data[i])); 278 } else { 279 CeedCallBackend(CeedElemRestrictionApply(rstr, CEED_NOTRANSPOSE, vec, impl->e_vecs[i], request)); 280 // Get evec 281 CeedCallBackend(CeedVectorGetArrayRead(impl->e_vecs[i], CEED_MEM_DEVICE, (const CeedScalar **)&e_data[i])); 282 } 283 } 284 } 285 return CEED_ERROR_SUCCESS; 286 } 287 288 //------------------------------------------------------------------------------ 289 // Input Basis Action 290 //------------------------------------------------------------------------------ 291 static inline int CeedOperatorInputBasis_Sycl(CeedInt num_elem, CeedQFunctionField *qf_input_fields, CeedOperatorField *op_input_fields, 292 CeedInt num_input_fields, const bool skip_active, CeedScalar *e_data[2 * CEED_FIELD_MAX], 293 CeedOperator_Sycl *impl) { 294 for (CeedInt i = 0; i < num_input_fields; i++) { 295 CeedInt elem_size, size; 296 CeedElemRestriction rstr; 297 CeedEvalMode e_mode; 298 CeedBasis basis; 299 300 // Skip active input 301 if (skip_active) { 302 CeedVector vec; 303 304 CeedCallBackend(CeedOperatorFieldGetVector(op_input_fields[i], &vec)); 305 if (vec == CEED_VECTOR_ACTIVE) continue; 306 } 307 // Get elem_size, e_mode, size 308 CeedCallBackend(CeedOperatorFieldGetElemRestriction(op_input_fields[i], &rstr)); 309 CeedCallBackend(CeedElemRestrictionGetElementSize(rstr, &elem_size)); 310 CeedCallBackend(CeedQFunctionFieldGetEvalMode(qf_input_fields[i], &e_mode)); 311 CeedCallBackend(CeedQFunctionFieldGetSize(qf_input_fields[i], &size)); 312 // Basis action 313 switch (e_mode) { 314 case CEED_EVAL_NONE: 315 CeedCallBackend(CeedVectorSetArray(impl->q_vecs_in[i], CEED_MEM_DEVICE, CEED_USE_POINTER, e_data[i])); 316 break; 317 case CEED_EVAL_INTERP: 318 CeedCallBackend(CeedOperatorFieldGetBasis(op_input_fields[i], &basis)); 319 CeedCallBackend(CeedBasisApply(basis, num_elem, CEED_NOTRANSPOSE, CEED_EVAL_INTERP, impl->e_vecs[i], impl->q_vecs_in[i])); 320 break; 321 case CEED_EVAL_GRAD: 322 CeedCallBackend(CeedOperatorFieldGetBasis(op_input_fields[i], &basis)); 323 CeedCallBackend(CeedBasisApply(basis, num_elem, CEED_NOTRANSPOSE, CEED_EVAL_GRAD, impl->e_vecs[i], impl->q_vecs_in[i])); 324 break; 325 case CEED_EVAL_WEIGHT: 326 break; // No action 327 case CEED_EVAL_DIV: 328 break; // TODO: Not implemented 329 case CEED_EVAL_CURL: 330 break; // TODO: Not implemented 331 } 332 } 333 return CEED_ERROR_SUCCESS; 334 } 335 336 //------------------------------------------------------------------------------ 337 // Restore Input Vectors 338 //------------------------------------------------------------------------------ 339 static inline int CeedOperatorRestoreInputs_Sycl(CeedInt num_input_fields, CeedQFunctionField *qf_input_fields, CeedOperatorField *op_input_fields, 340 const bool skip_active, CeedScalar *e_data[2 * CEED_FIELD_MAX], CeedOperator_Sycl *impl) { 341 for (CeedInt i = 0; i < num_input_fields; i++) { 342 CeedEvalMode e_mode; 343 CeedVector vec; 344 345 // Skip active input 346 if (skip_active) { 347 CeedCallBackend(CeedOperatorFieldGetVector(op_input_fields[i], &vec)); 348 if (vec == CEED_VECTOR_ACTIVE) continue; 349 } 350 CeedCallBackend(CeedQFunctionFieldGetEvalMode(qf_input_fields[i], &e_mode)); 351 if (e_mode == CEED_EVAL_WEIGHT) { // Skip 352 } else { 353 if (!impl->e_vecs[i]) { // This was a skip_restriction case 354 CeedCallBackend(CeedOperatorFieldGetVector(op_input_fields[i], &vec)); 355 CeedCallBackend(CeedVectorRestoreArrayRead(vec, (const CeedScalar **)&e_data[i])); 356 } else { 357 CeedCallBackend(CeedVectorRestoreArrayRead(impl->e_vecs[i], (const CeedScalar **)&e_data[i])); 358 } 359 } 360 } 361 return CEED_ERROR_SUCCESS; 362 } 363 364 //------------------------------------------------------------------------------ 365 // Apply and add to output 366 //------------------------------------------------------------------------------ 367 static int CeedOperatorApplyAdd_Sycl(CeedOperator op, CeedVector in_vec, CeedVector out_vec, CeedRequest *request) { 368 CeedInt Q, num_elem, elem_size, num_input_fields, num_output_fields, size; 369 CeedEvalMode e_mode; 370 CeedScalar *e_data[2 * CEED_FIELD_MAX] = {0}; 371 CeedQFunctionField *qf_input_fields, *qf_output_fields; 372 CeedQFunction qf; 373 CeedOperatorField *op_input_fields, *op_output_fields; 374 CeedOperator_Sycl *impl; 375 376 CeedCallBackend(CeedOperatorGetData(op, &impl)); 377 CeedCallBackend(CeedOperatorGetQFunction(op, &qf)); 378 CeedCallBackend(CeedOperatorGetNumQuadraturePoints(op, &Q)); 379 CeedCallBackend(CeedOperatorGetNumElements(op, &num_elem)); 380 CeedCallBackend(CeedOperatorGetFields(op, &num_input_fields, &op_input_fields, &num_output_fields, &op_output_fields)); 381 CeedCallBackend(CeedQFunctionGetFields(qf, NULL, &qf_input_fields, NULL, &qf_output_fields)); 382 383 // Setup 384 CeedCallBackend(CeedOperatorSetup_Sycl(op)); 385 386 // Input Evecs and Restriction 387 CeedCallBackend(CeedOperatorSetupInputs_Sycl(num_input_fields, qf_input_fields, op_input_fields, in_vec, false, e_data, impl, request)); 388 389 // Input basis apply if needed 390 CeedCallBackend(CeedOperatorInputBasis_Sycl(num_elem, qf_input_fields, op_input_fields, num_input_fields, false, e_data, impl)); 391 392 // Output pointers, as necessary 393 for (CeedInt i = 0; i < num_output_fields; i++) { 394 CeedCallBackend(CeedQFunctionFieldGetEvalMode(qf_output_fields[i], &e_mode)); 395 if (e_mode == CEED_EVAL_NONE) { 396 // Set the output Q-Vector to use the E-Vector data directly 397 CeedCallBackend(CeedVectorGetArrayWrite(impl->e_vecs[i + impl->num_e_in], CEED_MEM_DEVICE, &e_data[i + num_input_fields])); 398 CeedCallBackend(CeedVectorSetArray(impl->q_vecs_out[i], CEED_MEM_DEVICE, CEED_USE_POINTER, e_data[i + num_input_fields])); 399 } 400 } 401 402 // Q function 403 CeedCallBackend(CeedQFunctionApply(qf, num_elem * Q, impl->q_vecs_in, impl->q_vecs_out)); 404 405 // Output basis apply if needed 406 for (CeedInt i = 0; i < num_output_fields; i++) { 407 CeedElemRestriction rstr; 408 CeedBasis basis; 409 410 // Get elem_size, e_mode, size 411 CeedCallBackend(CeedOperatorFieldGetElemRestriction(op_output_fields[i], &rstr)); 412 CeedCallBackend(CeedElemRestrictionGetElementSize(rstr, &elem_size)); 413 CeedCallBackend(CeedQFunctionFieldGetEvalMode(qf_output_fields[i], &e_mode)); 414 CeedCallBackend(CeedQFunctionFieldGetSize(qf_output_fields[i], &size)); 415 // Basis action 416 switch (e_mode) { 417 case CEED_EVAL_NONE: 418 break; 419 case CEED_EVAL_INTERP: 420 CeedCallBackend(CeedOperatorFieldGetBasis(op_output_fields[i], &basis)); 421 CeedCallBackend(CeedBasisApply(basis, num_elem, CEED_TRANSPOSE, CEED_EVAL_INTERP, impl->q_vecs_out[i], impl->e_vecs[i + impl->num_e_in])); 422 break; 423 case CEED_EVAL_GRAD: 424 CeedCallBackend(CeedOperatorFieldGetBasis(op_output_fields[i], &basis)); 425 CeedCallBackend(CeedBasisApply(basis, num_elem, CEED_TRANSPOSE, CEED_EVAL_GRAD, impl->q_vecs_out[i], impl->e_vecs[i + impl->num_e_in])); 426 break; 427 // LCOV_EXCL_START 428 case CEED_EVAL_WEIGHT: 429 Ceed ceed; 430 431 CeedCallBackend(CeedOperatorGetCeed(op, &ceed)); 432 return CeedError(ceed, CEED_ERROR_BACKEND, "CEED_EVAL_WEIGHT cannot be an output evaluation mode"); 433 break; // Should not occur 434 case CEED_EVAL_DIV: 435 case CEED_EVAL_CURL: { 436 Ceed ceed; 437 438 CeedCallBackend(CeedOperatorGetCeed(op, &ceed)); 439 return CeedError(ceed, CEED_ERROR_BACKEND, "%s not supported", CeedEvalModes[eval_mode]); 440 break; // Should not occur 441 } 442 // LCOV_EXCL_STOP 443 } 444 } 445 446 // Output restriction 447 for (CeedInt i = 0; i < num_output_fields; i++) { 448 CeedVector vec; 449 CeedElemRestriction rstr; 450 451 // Restore evec 452 CeedCallBackend(CeedQFunctionFieldGetEvalMode(qf_output_fields[i], &e_mode)); 453 if (e_mode == CEED_EVAL_NONE) { 454 CeedCallBackend(CeedVectorRestoreArray(impl->e_vecs[i + impl->num_e_in], &e_data[i + num_input_fields])); 455 } 456 // Get output vector 457 CeedCallBackend(CeedOperatorFieldGetVector(op_output_fields[i], &vec)); 458 // Restrict 459 CeedCallBackend(CeedOperatorFieldGetElemRestriction(op_output_fields[i], &rstr)); 460 // Active 461 if (vec == CEED_VECTOR_ACTIVE) vec = out_vec; 462 463 CeedCallBackend(CeedElemRestrictionApply(rstr, CEED_TRANSPOSE, impl->e_vecs[i + impl->num_e_in], vec, request)); 464 } 465 466 // Restore input arrays 467 CeedCallBackend(CeedOperatorRestoreInputs_Sycl(num_input_fields, qf_input_fields, op_input_fields, false, e_data, impl)); 468 return CEED_ERROR_SUCCESS; 469 } 470 471 //------------------------------------------------------------------------------ 472 // Core code for assembling linear QFunction 473 //------------------------------------------------------------------------------ 474 static inline int CeedOperatorLinearAssembleQFunctionCore_Sycl(CeedOperator op, bool build_objects, CeedVector *assembled, CeedElemRestriction *rstr, 475 CeedRequest *request) { 476 Ceed ceed, ceed_parent; 477 CeedSize q_size; 478 CeedInt num_active_in, num_active_out, Q, num_elem, num_input_fields, num_output_fields, size; 479 CeedScalar *assembled_array, *e_data[2 * CEED_FIELD_MAX] = {NULL}; 480 CeedVector *active_in; 481 CeedQFunctionField *qf_input_fields, *qf_output_fields; 482 CeedQFunction qf; 483 CeedOperatorField *op_input_fields, *op_output_fields; 484 CeedOperator_Sycl *impl; 485 486 CeedCallBackend(CeedOperatorGetCeed(op, &ceed)); 487 CeedCallBackend(CeedOperatorGetFallbackParentCeed(op, &ceed_parent)); 488 CeedCallBackend(CeedOperatorGetData(op, &impl)); 489 CeedCallBackend(CeedOperatorGetQFunction(op, &qf)); 490 CeedCallBackend(CeedOperatorGetNumQuadraturePoints(op, &Q)); 491 CeedCallBackend(CeedOperatorGetNumElements(op, &num_elem)); 492 CeedCallBackend(CeedQFunctionGetFields(qf, NULL, &qf_input_fields, NULL, &qf_output_fields)); 493 CeedCallBackend(CeedOperatorGetFields(op, &num_input_fields, &op_input_fields, &num_output_fields, &op_output_fields)); 494 active_in = impl->qf_active_in; 495 num_active_in = impl->num_active_in, num_active_out = impl->num_active_out; 496 497 // Setup 498 CeedCallBackend(CeedOperatorSetup_Sycl(op)); 499 500 // Input Evecs and Restriction 501 CeedCallBackend(CeedOperatorSetupInputs_Sycl(num_input_fields, qf_input_fields, op_input_fields, NULL, true, e_data, impl, request)); 502 503 // Count number of active input fields 504 if (!num_active_in) { 505 for (CeedInt i = 0; i < num_input_fields; i++) { 506 CeedScalar *q_vec_array; 507 CeedVector vec; 508 509 // Get input vector 510 CeedCallBackend(CeedOperatorFieldGetVector(op_input_fields[i], &vec)); 511 // Check if active input 512 if (vec == CEED_VECTOR_ACTIVE) { 513 CeedCallBackend(CeedQFunctionFieldGetSize(qf_input_fields[i], &size)); 514 CeedCallBackend(CeedVectorSetValue(impl->q_vecs_in[i], 0.0)); 515 CeedCallBackend(CeedVectorGetArray(impl->q_vecs_in[i], CEED_MEM_DEVICE, &q_vec_array)); 516 CeedCallBackend(CeedRealloc(num_active_in + size, &active_in)); 517 for (CeedInt field = 0; field < size; field++) { 518 q_size = (CeedSize)Q * num_elem; 519 CeedCallBackend(CeedVectorCreate(ceed, q_size, &active_in[num_active_in + field])); 520 CeedCallBackend( 521 CeedVectorSetArray(active_in[num_active_in + field], CEED_MEM_DEVICE, CEED_USE_POINTER, &q_vec_array[field * Q * num_elem])); 522 } 523 num_active_in += size; 524 CeedCallBackend(CeedVectorRestoreArray(impl->q_vecs_in[i], &q_vec_array)); 525 } 526 } 527 impl->num_active_in = num_active_in; 528 impl->qf_active_in = active_in; 529 } 530 531 // Count number of active output fields 532 if (!num_active_out) { 533 for (CeedInt i = 0; i < num_output_fields; i++) { 534 CeedVector vec; 535 536 // Get output vector 537 CeedCallBackend(CeedOperatorFieldGetVector(op_output_fields[i], &vec)); 538 // Check if active output 539 if (vec == CEED_VECTOR_ACTIVE) { 540 CeedCallBackend(CeedQFunctionFieldGetSize(qf_output_fields[i], &size)); 541 num_active_out += size; 542 } 543 } 544 impl->num_active_out = num_active_out; 545 } 546 547 // Check sizes 548 CeedCheck(num_active_in > 0 && num_active_out > 0, ceed, CEED_ERROR_BACKEND, "Cannot assemble QFunction without active inputs and outputs"); 549 550 // Build objects if needed 551 if (build_objects) { 552 CeedSize l_size = (CeedSize)num_elem * Q * num_active_in * num_active_out; 553 CeedInt strides[3] = {1, num_elem * Q, Q}; /* *NOPAD* */ 554 555 // Create output restriction 556 CeedCallBackend(CeedElemRestrictionCreateStrided(ceed_parent, num_elem, Q, num_active_in * num_active_out, 557 num_active_in * num_active_out * num_elem * Q, strides, rstr)); 558 // Create assembled vector 559 CeedCallBackend(CeedVectorCreate(ceed_parent, l_size, assembled)); 560 } 561 CeedCallBackend(CeedVectorSetValue(*assembled, 0.0)); 562 CeedCallBackend(CeedVectorGetArray(*assembled, CEED_MEM_DEVICE, &assembled_array)); 563 564 // Input basis apply 565 CeedCallBackend(CeedOperatorInputBasis_Sycl(num_elem, qf_input_fields, op_input_fields, num_input_fields, true, e_data, impl)); 566 567 // Assemble QFunction 568 for (CeedInt in = 0; in < num_active_in; in++) { 569 // Set Inputs 570 CeedCallBackend(CeedVectorSetValue(active_in[in], 1.0)); 571 if (num_active_in > 1) { 572 CeedCallBackend(CeedVectorSetValue(active_in[(in + num_active_in - 1) % num_active_in], 0.0)); 573 } 574 // Set Outputs 575 for (CeedInt out = 0; out < num_output_fields; out++) { 576 CeedVector vec; 577 578 // Get output vector 579 CeedCallBackend(CeedOperatorFieldGetVector(op_output_fields[out], &vec)); 580 // Check if active output 581 if (vec == CEED_VECTOR_ACTIVE) { 582 CeedCallBackend(CeedVectorSetArray(impl->q_vecs_out[out], CEED_MEM_DEVICE, CEED_USE_POINTER, assembled_array)); 583 CeedCallBackend(CeedQFunctionFieldGetSize(qf_output_fields[out], &size)); 584 assembled_array += size * Q * num_elem; // Advance the pointer by the size of the output 585 } 586 } 587 // Apply QFunction 588 CeedCallBackend(CeedQFunctionApply(qf, Q * num_elem, impl->q_vecs_in, impl->q_vecs_out)); 589 } 590 591 // Un-set output Qvecs to prevent accidental overwrite of Assembled 592 for (CeedInt out = 0; out < num_output_fields; out++) { 593 CeedVector vec; 594 595 // Get output vector 596 CeedCallBackend(CeedOperatorFieldGetVector(op_output_fields[out], &vec)); 597 // Check if active output 598 if (vec == CEED_VECTOR_ACTIVE) { 599 CeedCallBackend(CeedVectorTakeArray(impl->q_vecs_out[out], CEED_MEM_DEVICE, NULL)); 600 } 601 } 602 603 // Restore input arrays 604 CeedCallBackend(CeedOperatorRestoreInputs_Sycl(num_input_fields, qf_input_fields, op_input_fields, true, e_data, impl)); 605 606 // Restore output 607 CeedCallBackend(CeedVectorRestoreArray(*assembled, &assembled_array)); 608 return CEED_ERROR_SUCCESS; 609 } 610 611 //------------------------------------------------------------------------------ 612 // Assemble Linear QFunction 613 //------------------------------------------------------------------------------ 614 static int CeedOperatorLinearAssembleQFunction_Sycl(CeedOperator op, CeedVector *assembled, CeedElemRestriction *rstr, CeedRequest *request) { 615 return CeedOperatorLinearAssembleQFunctionCore_Sycl(op, true, assembled, rstr, request); 616 } 617 618 //------------------------------------------------------------------------------ 619 // Update Assembled Linear QFunction 620 //------------------------------------------------------------------------------ 621 static int CeedOperatorLinearAssembleQFunctionUpdate_Sycl(CeedOperator op, CeedVector assembled, CeedElemRestriction rstr, CeedRequest *request) { 622 return CeedOperatorLinearAssembleQFunctionCore_Sycl(op, false, &assembled, &rstr, request); 623 } 624 625 //------------------------------------------------------------------------------ 626 // Assemble diagonal setup 627 //------------------------------------------------------------------------------ 628 static inline int CeedOperatorAssembleDiagonalSetup_Sycl(CeedOperator op) { 629 Ceed ceed; 630 Ceed_Sycl *sycl_data; 631 CeedInt num_input_fields, num_output_fields, num_e_mode_in = 0, num_comp = 0, dim = 1, num_e_mode_out = 0; 632 CeedEvalMode *e_mode_in = NULL, *e_mode_out = NULL; 633 CeedBasis basis_in = NULL, basis_out = NULL; 634 CeedElemRestriction rstr_in = NULL, rstr_out = NULL; 635 CeedQFunctionField *qf_fields; 636 CeedQFunction qf; 637 CeedOperatorField *op_fields; 638 CeedOperator_Sycl *impl; 639 640 CeedCallBackend(CeedOperatorGetCeed(op, &ceed)); 641 CeedCallBackend(CeedOperatorGetQFunction(op, &qf)); 642 CeedCallBackend(CeedQFunctionGetNumArgs(qf, &num_input_fields, &num_output_fields)); 643 644 // Determine active input basis 645 CeedCallBackend(CeedOperatorGetFields(op, NULL, &op_fields, NULL, NULL)); 646 CeedCallBackend(CeedQFunctionGetFields(qf, NULL, &qf_fields, NULL, NULL)); 647 for (CeedInt i = 0; i < num_input_fields; i++) { 648 CeedVector vec; 649 650 CeedCallBackend(CeedOperatorFieldGetVector(op_fields[i], &vec)); 651 if (vec == CEED_VECTOR_ACTIVE) { 652 CeedEvalMode e_mode; 653 CeedElemRestriction rstr; 654 655 CeedCallBackend(CeedOperatorFieldGetBasis(op_fields[i], &basis_in)); 656 CeedCallBackend(CeedBasisGetNumComponents(basis_in, &num_comp)); 657 CeedCallBackend(CeedBasisGetDimension(basis_in, &dim)); 658 CeedCallBackend(CeedOperatorFieldGetElemRestriction(op_fields[i], &rstr)); 659 CeedCheck(!rstr_in || rstr_in == rstr, ceed, CEED_ERROR_BACKEND, 660 "Backend does not implement multi-field non-composite operator diagonal assembly"); 661 rstr_in = rstr; 662 CeedCallBackend(CeedQFunctionFieldGetEvalMode(qf_fields[i], &e_mode)); 663 switch (e_mode) { 664 case CEED_EVAL_NONE: 665 case CEED_EVAL_INTERP: 666 CeedCallBackend(CeedRealloc(num_e_mode_in + 1, &e_mode_in)); 667 e_mode_in[num_e_mode_in] = e_mode; 668 num_e_mode_in += 1; 669 break; 670 case CEED_EVAL_GRAD: 671 CeedCallBackend(CeedRealloc(num_e_mode_in + dim, &e_mode_in)); 672 for (CeedInt d = 0; d < dim; d++) e_mode_in[num_e_mode_in + d] = e_mode; 673 num_e_mode_in += dim; 674 break; 675 case CEED_EVAL_WEIGHT: 676 case CEED_EVAL_DIV: 677 case CEED_EVAL_CURL: 678 break; // Caught by QF Assembly 679 } 680 } 681 } 682 683 // Determine active output basis 684 CeedCallBackend(CeedOperatorGetFields(op, NULL, NULL, NULL, &op_fields)); 685 CeedCallBackend(CeedQFunctionGetFields(qf, NULL, NULL, NULL, &qf_fields)); 686 for (CeedInt i = 0; i < num_output_fields; i++) { 687 CeedVector vec; 688 689 CeedCallBackend(CeedOperatorFieldGetVector(op_fields[i], &vec)); 690 if (vec == CEED_VECTOR_ACTIVE) { 691 CeedEvalMode e_mode; 692 CeedElemRestriction rstr; 693 694 CeedCallBackend(CeedOperatorFieldGetBasis(op_fields[i], &basis_out)); 695 CeedCallBackend(CeedOperatorFieldGetElemRestriction(op_fields[i], &rstr)); 696 CeedCheck(!rstr_out || rstr_out == rstr, ceed, CEED_ERROR_BACKEND, 697 "Backend does not implement multi-field non-composite operator diagonal assembly"); 698 rstr_out = rstr; 699 CeedCallBackend(CeedQFunctionFieldGetEvalMode(qf_fields[i], &e_mode)); 700 switch (e_mode) { 701 case CEED_EVAL_NONE: 702 case CEED_EVAL_INTERP: 703 CeedCallBackend(CeedRealloc(num_e_mode_out + 1, &e_mode_out)); 704 e_mode_out[num_e_mode_out] = e_mode; 705 num_e_mode_out += 1; 706 break; 707 case CEED_EVAL_GRAD: 708 CeedCallBackend(CeedRealloc(num_e_mode_out + dim, &e_mode_out)); 709 for (CeedInt d = 0; d < dim; d++) e_mode_out[num_e_mode_out + d] = e_mode; 710 num_e_mode_out += dim; 711 break; 712 case CEED_EVAL_WEIGHT: 713 case CEED_EVAL_DIV: 714 case CEED_EVAL_CURL: 715 break; // Caught by QF Assembly 716 } 717 } 718 } 719 720 // Operator data struct 721 CeedCallBackend(CeedOperatorGetData(op, &impl)); 722 CeedCallBackend(CeedGetData(ceed, &sycl_data)); 723 CeedCallBackend(CeedCalloc(1, &impl->diag)); 724 CeedOperatorDiag_Sycl *diag = impl->diag; 725 726 diag->basis_in = basis_in; 727 diag->basis_out = basis_out; 728 diag->h_e_mode_in = e_mode_in; 729 diag->h_e_mode_out = e_mode_out; 730 diag->num_e_mode_in = num_e_mode_in; 731 diag->num_e_mode_out = num_e_mode_out; 732 733 // Kernel parameters 734 CeedInt num_nodes, num_qpts; 735 CeedCallBackend(CeedBasisGetNumNodes(basis_in, &num_nodes)); 736 CeedCallBackend(CeedBasisGetNumQuadraturePoints(basis_in, &num_qpts)); 737 diag->num_nodes = num_nodes; 738 diag->num_qpts = num_qpts; 739 diag->num_comp = num_comp; 740 741 // Basis matrices 742 const CeedInt i_len = num_qpts * num_nodes; 743 const CeedInt g_len = num_qpts * num_nodes * dim; 744 const CeedScalar *interp_in, *interp_out, *grad_in, *grad_out; 745 746 // CEED_EVAL_NONE 747 CeedScalar *identity = NULL; 748 bool has_eval_none = false; 749 for (CeedInt i = 0; i < num_e_mode_in; i++) has_eval_none = has_eval_none || (e_mode_in[i] == CEED_EVAL_NONE); 750 for (CeedInt i = 0; i < num_e_mode_out; i++) has_eval_none = has_eval_none || (e_mode_out[i] == CEED_EVAL_NONE); 751 752 // Order queue 753 sycl::event e = sycl_data->sycl_queue.ext_oneapi_submit_barrier(); 754 755 std::vector<sycl::event> copy_events; 756 if (has_eval_none) { 757 CeedCallBackend(CeedCalloc(num_qpts * num_nodes, &identity)); 758 for (CeedSize i = 0; i < (num_nodes < num_qpts ? num_nodes : num_qpts); i++) identity[i * num_nodes + i] = 1.0; 759 CeedCallSycl(ceed, diag->d_identity = sycl::malloc_device<CeedScalar>(i_len, sycl_data->sycl_device, sycl_data->sycl_context)); 760 sycl::event identity_copy = sycl_data->sycl_queue.copy<CeedScalar>(identity, diag->d_identity, i_len, {e}); 761 copy_events.push_back(identity_copy); 762 } 763 764 // CEED_EVAL_INTERP 765 CeedCallBackend(CeedBasisGetInterp(basis_in, &interp_in)); 766 CeedCallSycl(ceed, diag->d_interp_in = sycl::malloc_device<CeedScalar>(i_len, sycl_data->sycl_device, sycl_data->sycl_context)); 767 sycl::event interp_in_copy = sycl_data->sycl_queue.copy<CeedScalar>(interp_in, diag->d_interp_in, i_len, {e}); 768 copy_events.push_back(interp_in_copy); 769 770 CeedCallBackend(CeedBasisGetInterp(basis_out, &interp_out)); 771 CeedCallSycl(ceed, diag->d_interp_out = sycl::malloc_device<CeedScalar>(i_len, sycl_data->sycl_device, sycl_data->sycl_context)); 772 sycl::event interp_out_copy = sycl_data->sycl_queue.copy<CeedScalar>(interp_out, diag->d_interp_out, i_len, {e}); 773 copy_events.push_back(interp_out_copy); 774 775 // CEED_EVAL_GRAD 776 CeedCallBackend(CeedBasisGetGrad(basis_in, &grad_in)); 777 CeedCallSycl(ceed, diag->d_grad_in = sycl::malloc_device<CeedScalar>(g_len, sycl_data->sycl_device, sycl_data->sycl_context)); 778 sycl::event grad_in_copy = sycl_data->sycl_queue.copy<CeedScalar>(grad_in, diag->d_grad_in, g_len, {e}); 779 copy_events.push_back(grad_in_copy); 780 781 CeedCallBackend(CeedBasisGetGrad(basis_out, &grad_out)); 782 CeedCallSycl(ceed, diag->d_grad_out = sycl::malloc_device<CeedScalar>(g_len, sycl_data->sycl_device, sycl_data->sycl_context)); 783 sycl::event grad_out_copy = sycl_data->sycl_queue.copy<CeedScalar>(grad_out, diag->d_grad_out, g_len, {e}); 784 copy_events.push_back(grad_out_copy); 785 786 // Arrays of e_modes 787 CeedCallSycl(ceed, diag->d_e_mode_in = sycl::malloc_device<CeedEvalMode>(num_e_mode_in, sycl_data->sycl_device, sycl_data->sycl_context)); 788 sycl::event e_mode_in_copy = sycl_data->sycl_queue.copy<CeedEvalMode>(e_mode_in, diag->d_e_mode_in, num_e_mode_in, {e}); 789 copy_events.push_back(e_mode_in_copy); 790 791 CeedCallSycl(ceed, diag->d_e_mode_out = sycl::malloc_device<CeedEvalMode>(num_e_mode_out, sycl_data->sycl_device, sycl_data->sycl_context)); 792 sycl::event e_mode_out_copy = sycl_data->sycl_queue.copy<CeedEvalMode>(e_mode_out, diag->d_e_mode_out, num_e_mode_out, {e}); 793 copy_events.push_back(e_mode_out_copy); 794 795 // Restriction 796 diag->diag_rstr = rstr_out; 797 798 // Wait for all copies to complete and handle exceptions 799 CeedCallSycl(ceed, sycl::event::wait_and_throw(copy_events)); 800 return CEED_ERROR_SUCCESS; 801 } 802 803 //------------------------------------------------------------------------------ 804 // Kernel for diagonal assembly 805 //------------------------------------------------------------------------------ 806 static int CeedOperatorLinearDiagonal_Sycl(sycl::queue &sycl_queue, const bool is_point_block, const CeedInt num_elem, 807 const CeedOperatorDiag_Sycl *diag, const CeedScalar *assembled_qf_array, CeedScalar *elem_diag_array) { 808 const CeedSize num_nodes = diag->num_nodes; 809 const CeedSize num_qpts = diag->num_qpts; 810 const CeedSize num_comp = diag->num_comp; 811 const CeedSize num_e_mode_in = diag->num_e_mode_in; 812 const CeedSize num_e_mode_out = diag->num_e_mode_out; 813 const CeedScalar *identity = diag->d_identity; 814 const CeedScalar *interp_in = diag->d_interp_in; 815 const CeedScalar *grad_in = diag->d_grad_in; 816 const CeedScalar *interp_out = diag->d_interp_out; 817 const CeedScalar *grad_out = diag->d_grad_out; 818 const CeedEvalMode *e_mode_in = diag->d_e_mode_in; 819 const CeedEvalMode *e_mode_out = diag->d_e_mode_out; 820 821 sycl::range<1> kernel_range(num_elem * num_nodes); 822 823 // Order queue 824 sycl::event e = sycl_queue.ext_oneapi_submit_barrier(); 825 sycl_queue.parallel_for<CeedOperatorSyclLinearDiagonal>(kernel_range, {e}, [=](sycl::id<1> idx) { 826 const CeedInt tid = idx % num_nodes; 827 const CeedInt e = idx / num_nodes; 828 829 // Compute the diagonal of B^T D B 830 // Each element 831 CeedInt d_out = -1; 832 // Each basis eval mode pair 833 for (CeedSize e_out = 0; e_out < num_e_mode_out; e_out++) { 834 const CeedScalar *bt = NULL; 835 836 if (e_mode_out[e_out] == CEED_EVAL_GRAD) ++d_out; 837 CeedOperatorGetBasisPointer_Sycl(&bt, e_mode_out[e_out], identity, interp_out, &grad_out[d_out * num_qpts * num_nodes]); 838 CeedInt d_in = -1; 839 840 for (CeedSize e_in = 0; e_in < num_e_mode_in; e_in++) { 841 const CeedScalar *b = NULL; 842 843 if (e_mode_in[e_in] == CEED_EVAL_GRAD) ++d_in; 844 CeedOperatorGetBasisPointer_Sycl(&b, e_mode_in[e_in], identity, interp_in, &grad_in[d_in * num_qpts * num_nodes]); 845 // Each component 846 for (CeedSize comp_out = 0; comp_out < num_comp; comp_out++) { 847 // Each qpoint/node pair 848 if (is_point_block) { 849 // Point Block Diagonal 850 for (CeedInt comp_in = 0; comp_in < num_comp; comp_in++) { 851 CeedScalar e_value = 0.0; 852 853 for (CeedSize q = 0; q < num_qpts; q++) { 854 const CeedScalar qf_value = 855 assembled_qf_array[((((e_in * num_comp + comp_in) * num_e_mode_out + e_out) * num_comp + comp_out) * num_elem + e) * num_qpts + 856 q]; 857 858 e_value += bt[q * num_nodes + tid] * qf_value * b[q * num_nodes + tid]; 859 } 860 elem_diag_array[((comp_out * num_comp + comp_in) * num_elem + e) * num_nodes + tid] += e_value; 861 } 862 } else { 863 // Diagonal Only 864 CeedScalar e_value = 0.0; 865 866 for (CeedSize q = 0; q < num_qpts; q++) { 867 const CeedScalar qf_value = 868 assembled_qf_array[((((e_in * num_comp + comp_out) * num_e_mode_out + e_out) * num_comp + comp_out) * num_elem + e) * num_qpts + q]; 869 e_value += bt[q * num_nodes + tid] * qf_value * b[q * num_nodes + tid]; 870 } 871 elem_diag_array[(comp_out * num_elem + e) * num_nodes + tid] += e_value; 872 } 873 } 874 } 875 } 876 }); 877 return CEED_ERROR_SUCCESS; 878 } 879 880 //------------------------------------------------------------------------------ 881 // Assemble diagonal common code 882 //------------------------------------------------------------------------------ 883 static inline int CeedOperatorAssembleDiagonalCore_Sycl(CeedOperator op, CeedVector assembled, CeedRequest *request, const bool is_point_block) { 884 Ceed ceed; 885 Ceed_Sycl *sycl_data; 886 CeedInt num_elem; 887 CeedScalar *elem_diag_array; 888 const CeedScalar *assembled_qf_array; 889 CeedVector assembled_qf = NULL; 890 CeedElemRestriction rstr = NULL; 891 CeedOperator_Sycl *impl; 892 893 CeedCallBackend(CeedOperatorGetCeed(op, &ceed)); 894 CeedCallBackend(CeedOperatorGetData(op, &impl)); 895 CeedCallBackend(CeedGetData(ceed, &sycl_data)); 896 897 // Assemble QFunction 898 CeedCallBackend(CeedOperatorLinearAssembleQFunctionBuildOrUpdate(op, &assembled_qf, &rstr, request)); 899 CeedCallBackend(CeedElemRestrictionDestroy(&rstr)); 900 901 // Setup 902 if (!impl->diag) { 903 CeedCallBackend(CeedOperatorAssembleDiagonalSetup_Sycl(op)); 904 } 905 CeedOperatorDiag_Sycl *diag = impl->diag; 906 907 assert(diag != NULL); 908 909 // Restriction 910 if (is_point_block && !diag->point_block_diag_rstr) { 911 CeedCallBackend(CeedOperatorCreateActivePointBlockRestriction(diag->diag_rstr, &diag->point_block_diag_rstr)); 912 } 913 CeedElemRestriction diag_rstr = is_point_block ? diag->point_block_diag_rstr : diag->diag_rstr; 914 915 // Create diagonal vector 916 CeedVector elem_diag = is_point_block ? diag->point_block_elem_diag : diag->elem_diag; 917 918 if (!elem_diag) { 919 CeedCallBackend(CeedElemRestrictionCreateVector(diag_rstr, NULL, &elem_diag)); 920 if (is_point_block) diag->point_block_elem_diag = elem_diag; 921 else diag->elem_diag = elem_diag; 922 } 923 CeedCallBackend(CeedVectorSetValue(elem_diag, 0.0)); 924 925 // Assemble element operator diagonals 926 CeedCallBackend(CeedVectorGetArray(elem_diag, CEED_MEM_DEVICE, &elem_diag_array)); 927 CeedCallBackend(CeedVectorGetArrayRead(assembled_qf, CEED_MEM_DEVICE, &assembled_qf_array)); 928 CeedCallBackend(CeedElemRestrictionGetNumElements(diag_rstr, &num_elem)); 929 930 // Compute the diagonal of B^T D B 931 CeedCallBackend(CeedOperatorLinearDiagonal_Sycl(sycl_data->sycl_queue, is_point_block, num_elem, diag, assembled_qf_array, elem_diag_array)); 932 933 // Wait for queue to complete and handle exceptions 934 sycl_data->sycl_queue.wait_and_throw(); 935 936 // Restore arrays 937 CeedCallBackend(CeedVectorRestoreArray(elem_diag, &elem_diag_array)); 938 CeedCallBackend(CeedVectorRestoreArrayRead(assembled_qf, &assembled_qf_array)); 939 940 // Assemble local operator diagonal 941 CeedCallBackend(CeedElemRestrictionApply(diag_rstr, CEED_TRANSPOSE, elem_diag, assembled, request)); 942 943 // Cleanup 944 CeedCallBackend(CeedVectorDestroy(&assembled_qf)); 945 return CEED_ERROR_SUCCESS; 946 } 947 948 //------------------------------------------------------------------------------ 949 // Assemble Linear Diagonal 950 //------------------------------------------------------------------------------ 951 static int CeedOperatorLinearAssembleAddDiagonal_Sycl(CeedOperator op, CeedVector assembled, CeedRequest *request) { 952 CeedCallBackend(CeedOperatorAssembleDiagonalCore_Sycl(op, assembled, request, false)); 953 return CEED_ERROR_SUCCESS; 954 } 955 956 //------------------------------------------------------------------------------ 957 // Assemble Linear Point Block Diagonal 958 //------------------------------------------------------------------------------ 959 static int CeedOperatorLinearAssembleAddPointBlockDiagonal_Sycl(CeedOperator op, CeedVector assembled, CeedRequest *request) { 960 CeedCallBackend(CeedOperatorAssembleDiagonalCore_Sycl(op, assembled, request, true)); 961 return CEED_ERROR_SUCCESS; 962 } 963 964 //------------------------------------------------------------------------------ 965 // Single operator assembly setup 966 //------------------------------------------------------------------------------ 967 static int CeedSingleOperatorAssembleSetup_Sycl(CeedOperator op) { 968 Ceed ceed; 969 CeedInt num_input_fields, num_output_fields, num_e_mode_in = 0, dim = 1, num_B_in_mats_to_load = 0, size_B_in = 0, num_e_mode_out = 0, 970 num_B_out_mats_to_load = 0, size_B_out = 0, num_qpts = 0, elem_size = 0, num_elem, num_comp, 971 mat_start = 0; 972 CeedEvalMode *eval_mode_in = NULL, *eval_mode_out = NULL; 973 const CeedScalar *interp_in, *grad_in; 974 CeedElemRestriction rstr_in = NULL, rstr_out = NULL; 975 CeedBasis basis_in = NULL, basis_out = NULL; 976 CeedQFunctionField *qf_fields; 977 CeedQFunction qf; 978 CeedOperatorField *input_fields, *output_fields; 979 CeedOperator_Sycl *impl; 980 981 CeedCallBackend(CeedOperatorGetCeed(op, &ceed)); 982 CeedCallBackend(CeedOperatorGetData(op, &impl)); 983 984 // Get input and output fields 985 CeedCallBackend(CeedOperatorGetFields(op, &num_input_fields, &input_fields, &num_output_fields, &output_fields)); 986 987 // Determine active input basis eval mode 988 CeedCallBackend(CeedOperatorGetQFunction(op, &qf)); 989 CeedCallBackend(CeedQFunctionGetFields(qf, NULL, &qf_fields, NULL, NULL)); 990 // Note that the kernel will treat each dimension of a gradient action separately; 991 // i.e., when an active input has a CEED_EVAL_GRAD mode, num_ e_mode_in will increment by dim. 992 // However, for the purposes of load_ing the B matrices, it will be treated as one mode, and we will load/copy the entire gradient matrix at once, 993 // so num_B_in_mats_to_load will be incremented by 1. 994 for (CeedInt i = 0; i < num_input_fields; i++) { 995 CeedEvalMode eval_mode; 996 CeedVector vec; 997 998 CeedCallBackend(CeedOperatorFieldGetVector(input_fields[i], &vec)); 999 if (vec == CEED_VECTOR_ACTIVE) { 1000 CeedCallBackend(CeedOperatorFieldGetBasis(input_fields[i], &basis_in)); 1001 CeedCallBackend(CeedBasisGetDimension(basis_in, &dim)); 1002 CeedCallBackend(CeedBasisGetNumQuadraturePoints(basis_in, &num_qpts)); 1003 CeedCallBackend(CeedOperatorFieldGetElemRestriction(input_fields[i], &rstr_in)); 1004 CeedCallBackend(CeedElemRestrictionGetElementSize(rstr_in, &elem_size)); 1005 CeedCallBackend(CeedQFunctionFieldGetEvalMode(qf_fields[i], &eval_mode)); 1006 if (eval_mode != CEED_EVAL_NONE) { 1007 CeedCallBackend(CeedRealloc(num_B_in_mats_to_load + 1, &eval_mode_in)); 1008 eval_mode_in[num_B_in_mats_to_load] = eval_mode; 1009 num_B_in_mats_to_load += 1; 1010 if (eval_mode == CEED_EVAL_GRAD) { 1011 num_e_mode_in += dim; 1012 size_B_in += dim * elem_size * num_qpts; 1013 } else { 1014 num_e_mode_in += 1; 1015 size_B_in += elem_size * num_qpts; 1016 } 1017 } 1018 } 1019 } 1020 1021 // Determine active output basis; basis_out and rstr_out only used if same as input, TODO 1022 CeedCallBackend(CeedQFunctionGetFields(qf, NULL, NULL, NULL, &qf_fields)); 1023 for (CeedInt i = 0; i < num_output_fields; i++) { 1024 CeedEvalMode eval_mode; 1025 CeedVector vec; 1026 1027 CeedCallBackend(CeedOperatorFieldGetVector(output_fields[i], &vec)); 1028 if (vec == CEED_VECTOR_ACTIVE) { 1029 CeedCallBackend(CeedOperatorFieldGetBasis(output_fields[i], &basis_out)); 1030 CeedCallBackend(CeedOperatorFieldGetElemRestriction(output_fields[i], &rstr_out)); 1031 CeedCheck(!rstr_out || rstr_out == rstr_in, ceed, CEED_ERROR_BACKEND, "Backend does not implement multi-field non-composite operator assembly"); 1032 CeedCallBackend(CeedQFunctionFieldGetEvalMode(qf_fields[i], &eval_mode)); 1033 if (eval_mode != CEED_EVAL_NONE) { 1034 CeedCallBackend(CeedRealloc(num_B_out_mats_to_load + 1, &eval_mode_out)); 1035 eval_mode_out[num_B_out_mats_to_load] = eval_mode; 1036 num_B_out_mats_to_load += 1; 1037 if (eval_mode == CEED_EVAL_GRAD) { 1038 num_e_mode_out += dim; 1039 size_B_out += dim * elem_size * num_qpts; 1040 } else { 1041 num_e_mode_out += 1; 1042 size_B_out += elem_size * num_qpts; 1043 } 1044 } 1045 } 1046 } 1047 CeedCheck(num_e_mode_in > 0 && num_e_mode_out > 0, ceed, CEED_ERROR_UNSUPPORTED, "Cannot assemble operator without inputs/outputs"); 1048 1049 CeedCallBackend(CeedElemRestrictionGetNumElements(rstr_in, &num_elem)); 1050 CeedCallBackend(CeedElemRestrictionGetNumComponents(rstr_in, &num_comp)); 1051 1052 CeedCallBackend(CeedCalloc(1, &impl->asmb)); 1053 CeedOperatorAssemble_Sycl *asmb = impl->asmb; 1054 asmb->num_elem = num_elem; 1055 1056 Ceed_Sycl *sycl_data; 1057 CeedCallBackend(CeedGetData(ceed, &sycl_data)); 1058 1059 // Kernel setup 1060 int elems_per_block = 1; 1061 asmb->elems_per_block = elems_per_block; 1062 asmb->block_size_x = elem_size; 1063 asmb->block_size_y = elem_size; 1064 asmb->num_e_mode_in = num_e_mode_in; 1065 asmb->num_e_mode_out = num_e_mode_out; 1066 asmb->num_qpts = num_qpts; 1067 asmb->num_nodes = elem_size; 1068 asmb->block_size = elem_size * elem_size * elems_per_block; 1069 asmb->num_comp = num_comp; 1070 1071 // Build 'full' B matrices (not 1D arrays used for tensor-product matrices 1072 CeedCallBackend(CeedBasisGetInterp(basis_in, &interp_in)); 1073 CeedCallBackend(CeedBasisGetGrad(basis_in, &grad_in)); 1074 1075 // Load into B_in, in order that they will be used in eval_mode 1076 CeedCallSycl(ceed, asmb->d_B_in = sycl::malloc_device<CeedScalar>(size_B_in, sycl_data->sycl_device, sycl_data->sycl_context)); 1077 for (int i = 0; i < num_B_in_mats_to_load; i++) { 1078 CeedEvalMode eval_mode = eval_mode_in[i]; 1079 1080 if (eval_mode == CEED_EVAL_INTERP) { 1081 // Order queue 1082 sycl::event e = sycl_data->sycl_queue.ext_oneapi_submit_barrier(); 1083 sycl_data->sycl_queue.copy<CeedScalar>(interp_in, &asmb->d_B_in[mat_start], elem_size * num_qpts, {e}); 1084 mat_start += elem_size * num_qpts; 1085 } else if (eval_mode == CEED_EVAL_GRAD) { 1086 // Order queue 1087 sycl::event e = sycl_data->sycl_queue.ext_oneapi_submit_barrier(); 1088 sycl_data->sycl_queue.copy<CeedScalar>(grad_in, &asmb->d_B_in[mat_start], dim * elem_size * num_qpts, {e}); 1089 mat_start += dim * elem_size * num_qpts; 1090 } 1091 } 1092 1093 const CeedScalar *interp_out, *grad_out; 1094 // Note that this function currently assumes 1 basis, so this should always be true 1095 // for now 1096 if (basis_out == basis_in) { 1097 interp_out = interp_in; 1098 grad_out = grad_in; 1099 } else { 1100 CeedCallBackend(CeedBasisGetInterp(basis_out, &interp_out)); 1101 CeedCallBackend(CeedBasisGetGrad(basis_out, &grad_out)); 1102 } 1103 1104 // Load into B_out, in order that they will be used in eval_mode 1105 mat_start = 0; 1106 CeedCallSycl(ceed, asmb->d_B_out = sycl::malloc_device<CeedScalar>(size_B_out, sycl_data->sycl_device, sycl_data->sycl_context)); 1107 for (int i = 0; i < num_B_out_mats_to_load; i++) { 1108 CeedEvalMode eval_mode = eval_mode_out[i]; 1109 1110 if (eval_mode == CEED_EVAL_INTERP) { 1111 // Order queue 1112 sycl::event e = sycl_data->sycl_queue.ext_oneapi_submit_barrier(); 1113 sycl_data->sycl_queue.copy<CeedScalar>(interp_out, &asmb->d_B_out[mat_start], elem_size * num_qpts, {e}); 1114 mat_start += elem_size * num_qpts; 1115 } else if (eval_mode == CEED_EVAL_GRAD) { 1116 // Order queue 1117 sycl::event e = sycl_data->sycl_queue.ext_oneapi_submit_barrier(); 1118 sycl_data->sycl_queue.copy<CeedScalar>(grad_out, &asmb->d_B_out[mat_start], dim * elem_size * num_qpts, {e}); 1119 mat_start += dim * elem_size * num_qpts; 1120 } 1121 } 1122 return CEED_ERROR_SUCCESS; 1123 } 1124 1125 //------------------------------------------------------------------------------ 1126 // Matrix assembly kernel for low-order elements (3D thread block) 1127 //------------------------------------------------------------------------------ 1128 static int CeedOperatorLinearAssemble_Sycl(sycl::queue &sycl_queue, const CeedOperator_Sycl *impl, const CeedScalar *qf_array, 1129 CeedScalar *values_array) { 1130 // This kernels assumes B_in and B_out have the same number of quadrature points and basis points. 1131 // TODO: expand to more general cases 1132 CeedOperatorAssemble_Sycl *asmb = impl->asmb; 1133 const CeedInt num_elem = asmb->num_elem; 1134 const CeedSize num_nodes = asmb->num_nodes; 1135 const CeedSize num_comp = asmb->num_comp; 1136 const CeedSize num_qpts = asmb->num_qpts; 1137 const CeedSize num_e_mode_in = asmb->num_e_mode_in; 1138 const CeedSize num_e_mode_out = asmb->num_e_mode_out; 1139 1140 // Strides for final output ordering, determined by the reference (inference) implementation of the symbolic assembly, slowest --> fastest: element, 1141 // comp_in, comp_out, node_row, node_col 1142 const CeedSize comp_out_stride = num_nodes * num_nodes; 1143 const CeedSize comp_in_stride = comp_out_stride * num_comp; 1144 const CeedSize e_stride = comp_in_stride * num_comp; 1145 // Strides for QF array, slowest --> fastest: e_mode_in, comp_in, e_mode_out, comp_out, elem, qpt 1146 const CeedSize q_e_stride = num_qpts; 1147 const CeedSize q_comp_out_stride = num_elem * q_e_stride; 1148 const CeedSize q_e_mode_out_stride = q_comp_out_stride * num_comp; 1149 const CeedSize q_comp_in_stride = q_e_mode_out_stride * num_e_mode_out; 1150 const CeedSize q_e_mode_in_stride = q_comp_in_stride * num_comp; 1151 1152 CeedScalar *B_in, *B_out; 1153 B_in = asmb->d_B_in; 1154 B_out = asmb->d_B_out; 1155 const CeedInt block_size_x = asmb->block_size_x; 1156 const CeedInt block_size_y = asmb->block_size_y; 1157 1158 sycl::range<3> kernel_range(num_elem, block_size_y, block_size_x); 1159 1160 // Order queue 1161 sycl::event e = sycl_queue.ext_oneapi_submit_barrier(); 1162 sycl_queue.parallel_for<CeedOperatorSyclLinearAssemble>(kernel_range, {e}, [=](sycl::id<3> idx) { 1163 const int e = idx.get(0); // Element index 1164 const int l = idx.get(1); // The output column index of each B^TDB operation 1165 const int i = idx.get(2); // The output row index of each B^TDB operation 1166 // such that we have (Bout^T)_ij D_jk Bin_kl = C_il 1167 for (CeedSize comp_in = 0; comp_in < num_comp; comp_in++) { 1168 for (CeedSize comp_out = 0; comp_out < num_comp; comp_out++) { 1169 CeedScalar result = 0.0; 1170 CeedSize qf_index_comp = q_comp_in_stride * comp_in + q_comp_out_stride * comp_out + q_e_stride * e; 1171 1172 for (CeedSize e_mode_in = 0; e_mode_in < num_e_mode_in; e_mode_in++) { 1173 CeedSize b_in_index = e_mode_in * num_qpts * num_nodes; 1174 1175 for (CeedSize e_mode_out = 0; e_mode_out < num_e_mode_out; e_mode_out++) { 1176 CeedSize b_out_index = e_mode_out * num_qpts * num_nodes; 1177 CeedSize qf_index = qf_index_comp + q_e_mode_out_stride * e_mode_out + q_e_mode_in_stride * e_mode_in; 1178 1179 // Perform the B^T D B operation for this 'chunk' of D (the qf_array) 1180 for (CeedSize j = 0; j < num_qpts; j++) { 1181 result += B_out[b_out_index + j * num_nodes + i] * qf_array[qf_index + j] * B_in[b_in_index + j * num_nodes + l]; 1182 } 1183 } // end of e_mode_out 1184 } // end of e_mode_in 1185 CeedSize val_index = comp_in_stride * comp_in + comp_out_stride * comp_out + e_stride * e + num_nodes * i + l; 1186 1187 values_array[val_index] = result; 1188 } // end of out component 1189 } // end of in component 1190 }); 1191 return CEED_ERROR_SUCCESS; 1192 } 1193 1194 //------------------------------------------------------------------------------ 1195 // Fallback kernel for larger orders (1D thread block) 1196 //------------------------------------------------------------------------------ 1197 /* 1198 static int CeedOperatorLinearAssembleFallback_Sycl(sycl::queue &sycl_queue, const CeedOperator_Sycl *impl, const CeedScalar *qf_array, 1199 CeedScalar *values_array) { 1200 // This kernel assumes B_in and B_out have the same number of quadrature points and basis points. 1201 // TODO: expand to more general cases 1202 CeedOperatorAssemble_Sycl *asmb = impl->asmb; 1203 const CeedInt num_elem = asmb->num_elem; 1204 const CeedInt num_nodes = asmb->num_nodes; 1205 const CeedInt num_comp = asmb->num_comp; 1206 const CeedInt num_qpts = asmb->num_qpts; 1207 const CeedInt num_e_mode_in = asmb->num_e_mode_in; 1208 const CeedInt num_e_mode_out = asmb->num_e_mode_out; 1209 1210 // Strides for final output ordering, determined by the reference (interface) implementation of the symbolic assembly, slowest --> fastest: elememt, 1211 // comp_in, comp_out, node_row, node_col 1212 const CeedInt comp_out_stride = num_nodes * num_nodes; 1213 const CeedInt comp_in_stride = comp_out_stride * num_comp; 1214 const CeedInt e_stride = comp_in_stride * num_comp; 1215 // Strides for QF array, slowest --> fastest: e_mode_in, comp_in, e_mode_out, comp_out, elem, qpt 1216 const CeedInt q_e_stride = num_qpts; 1217 const CeedInt q_comp_out_stride = num_elem * q_e_stride; 1218 const CeedInt q_e_mode_out_stride = q_comp_out_stride * num_comp; 1219 const CeedInt q_comp_in_stride = q_e_mode_out_stride * num_e_mode_out; 1220 const CeedInt q_e_mode_in_stride = q_comp_in_stride * num_comp; 1221 1222 CeedScalar *B_in, *B_out; 1223 B_in = asmb->d_B_in; 1224 B_out = asmb->d_B_out; 1225 const CeedInt elems_per_block = asmb->elems_per_block; 1226 const CeedInt block_size_x = asmb->block_size_x; 1227 const CeedInt block_size_y = asmb->block_size_y; // This will be 1 for the fallback kernel 1228 1229 const CeedInt grid = num_elem / elems_per_block + ((num_elem / elems_per_block * elems_per_block < num_elem) ? 1 : 0); 1230 sycl::range<3> local_range(block_size_x, block_size_y, elems_per_block); 1231 sycl::range<3> global_range(grid * block_size_x, block_size_y, elems_per_block); 1232 sycl::nd_range<3> kernel_range(global_range, local_range); 1233 1234 sycl_queue.parallel_for<CeedOperatorSyclLinearAssembleFallback>(kernel_range, [=](sycl::nd_item<3> work_item) { 1235 const CeedInt blockIdx = work_item.get_group(0); 1236 const CeedInt gridDimx = work_item.get_group_range(0); 1237 const CeedInt threadIdx = work_item.get_local_id(0); 1238 const CeedInt threadIdz = work_item.get_local_id(2); 1239 const CeedInt blockDimz = work_item.get_local_range(2); 1240 1241 const int l = threadIdx; // The output column index of each B^TDB operation 1242 // such that we have (Bout^T)_ij D_jk Bin_kl = C_il 1243 for (CeedInt e = blockIdx * blockDimz + threadIdz; e < num_elem; e += gridDimx * blockDimz) { 1244 for (CeedInt comp_in = 0; comp_in < num_comp; comp_in++) { 1245 for (CeedInt comp_out = 0; comp_out < num_comp; comp_out++) { 1246 for (CeedInt i = 0; i < num_nodes; i++) { 1247 CeedScalar result = 0.0; 1248 CeedInt qf_index_comp = q_comp_in_stride * comp_in + q_comp_out_stride * comp_out + q_e_stride * e; 1249 for (CeedInt e_mode_in = 0; e_mode_in < num_e_mode_in; e_mode_in++) { 1250 CeedInt b_in_index = e_mode_in * num_qpts * num_nodes; 1251 for (CeedInt e_mode_out = 0; e_mode_out < num_e_mode_out; e_mode_out++) { 1252 CeedInt b_out_index = e_mode_out * num_qpts * num_nodes; 1253 CeedInt qf_index = qf_index_comp + q_e_mode_out_stride * e_mode_out + q_e_mode_in_stride * e_mode_in; 1254 // Perform the B^T D B operation for this 'chunk' of D (the qf_array) 1255 for (CeedInt j = 0; j < num_qpts; j++) { 1256 result += B_out[b_out_index + j * num_nodes + i] * qf_array[qf_index + j] * B_in[b_in_index + j * num_nodes + l]; 1257 } 1258 } // end of e_mode_out 1259 } // end of e_mode_in 1260 CeedInt val_index = comp_in_stride * comp_in + comp_out_stride * comp_out + e_stride * e + num_nodes * i + l; 1261 values_array[val_index] = result; 1262 } // end of loop over element node index, i 1263 } // end of out component 1264 } // end of in component 1265 } // end of element loop 1266 }); 1267 return CEED_ERROR_SUCCESS; 1268 }*/ 1269 1270 //------------------------------------------------------------------------------ 1271 // Assemble matrix data for COO matrix of assembled operator. 1272 // The sparsity pattern is set by CeedOperatorLinearAssembleSymbolic. 1273 // 1274 // Note that this (and other assembly routines) currently assume only one active 1275 // input restriction/basis per operator (could have multiple basis eval modes). 1276 // TODO: allow multiple active input restrictions/basis objects 1277 //------------------------------------------------------------------------------ 1278 static int CeedSingleOperatorAssemble_Sycl(CeedOperator op, CeedInt offset, CeedVector values) { 1279 Ceed ceed; 1280 Ceed_Sycl *sycl_data; 1281 CeedScalar *values_array; 1282 const CeedScalar *qf_array; 1283 CeedVector assembled_qf = NULL; 1284 CeedElemRestriction rstr_q = NULL; 1285 CeedOperator_Sycl *impl; 1286 1287 CeedCallBackend(CeedOperatorGetCeed(op, &ceed)); 1288 CeedCallBackend(CeedOperatorGetData(op, &impl)); 1289 CeedCallBackend(CeedGetData(ceed, &sycl_data)); 1290 1291 // Setup 1292 if (!impl->asmb) { 1293 CeedCallBackend(CeedSingleOperatorAssembleSetup_Sycl(op)); 1294 assert(impl->asmb != NULL); 1295 } 1296 1297 // Assemble QFunction 1298 CeedCallBackend(CeedOperatorLinearAssembleQFunctionBuildOrUpdate(op, &assembled_qf, &rstr_q, CEED_REQUEST_IMMEDIATE)); 1299 CeedCallBackend(CeedElemRestrictionDestroy(&rstr_q)); 1300 CeedCallBackend(CeedVectorGetArrayWrite(values, CEED_MEM_DEVICE, &values_array)); 1301 values_array += offset; 1302 CeedCallBackend(CeedVectorGetArrayRead(assembled_qf, CEED_MEM_DEVICE, &qf_array)); 1303 1304 // Compute B^T D B 1305 CeedCallBackend(CeedOperatorLinearAssemble_Sycl(sycl_data->sycl_queue, impl, qf_array, values_array)); 1306 1307 // Wait for kernels to be completed 1308 // Kris: Review if this is necessary -- enqueing an async barrier may be sufficient 1309 sycl_data->sycl_queue.wait_and_throw(); 1310 1311 // Restore arrays 1312 CeedCallBackend(CeedVectorRestoreArray(values, &values_array)); 1313 CeedCallBackend(CeedVectorRestoreArrayRead(assembled_qf, &qf_array)); 1314 1315 // Cleanup 1316 CeedCallBackend(CeedVectorDestroy(&assembled_qf)); 1317 return CEED_ERROR_SUCCESS; 1318 } 1319 1320 //------------------------------------------------------------------------------ 1321 // Create operator 1322 //------------------------------------------------------------------------------ 1323 int CeedOperatorCreate_Sycl(CeedOperator op) { 1324 Ceed ceed; 1325 CeedOperator_Sycl *impl; 1326 1327 CeedCallBackend(CeedOperatorGetCeed(op, &ceed)); 1328 1329 CeedCallBackend(CeedCalloc(1, &impl)); 1330 CeedCallBackend(CeedOperatorSetData(op, impl)); 1331 1332 CeedCallBackend(CeedSetBackendFunctionCpp(ceed, "Operator", op, "LinearAssembleQFunction", CeedOperatorLinearAssembleQFunction_Sycl)); 1333 CeedCallBackend(CeedSetBackendFunctionCpp(ceed, "Operator", op, "LinearAssembleQFunctionUpdate", CeedOperatorLinearAssembleQFunctionUpdate_Sycl)); 1334 CeedCallBackend(CeedSetBackendFunctionCpp(ceed, "Operator", op, "LinearAssembleAddDiagonal", CeedOperatorLinearAssembleAddDiagonal_Sycl)); 1335 CeedCallBackend( 1336 CeedSetBackendFunctionCpp(ceed, "Operator", op, "LinearAssembleAddPointBlockDiagonal", CeedOperatorLinearAssembleAddPointBlockDiagonal_Sycl)); 1337 CeedCallBackend(CeedSetBackendFunctionCpp(ceed, "Operator", op, "LinearAssembleSingle", CeedSingleOperatorAssemble_Sycl)); 1338 CeedCallBackend(CeedSetBackendFunctionCpp(ceed, "Operator", op, "ApplyAdd", CeedOperatorApplyAdd_Sycl)); 1339 CeedCallBackend(CeedSetBackendFunctionCpp(ceed, "Operator", op, "Destroy", CeedOperatorDestroy_Sycl)); 1340 return CEED_ERROR_SUCCESS; 1341 } 1342 1343 //------------------------------------------------------------------------------ 1344