1 // Copyright (c) 2017-2022, Lawrence Livermore National Security, LLC and other CEED contributors. 2 // All Rights Reserved. See the top-level LICENSE and NOTICE files for details. 3 // 4 // SPDX-License-Identifier: BSD-2-Clause 5 // 6 // This file is part of CEED: http://github.com/ceed 7 8 #include <ceed.h> 9 #include <ceed/backend.h> 10 #include <ceed/jit-tools.h> 11 #include <assert.h> 12 #include <cuda.h> 13 #include <cuda_runtime.h> 14 #include <stdbool.h> 15 #include <string.h> 16 17 #include "../cuda/ceed-cuda-common.h" 18 #include "../cuda/ceed-cuda-compile.h" 19 #include "ceed-cuda-ref.h" 20 21 //------------------------------------------------------------------------------ 22 // Destroy operator 23 //------------------------------------------------------------------------------ 24 static int CeedOperatorDestroy_Cuda(CeedOperator op) { 25 CeedOperator_Cuda *impl; 26 CeedCallBackend(CeedOperatorGetData(op, &impl)); 27 28 // Apply data 29 for (CeedInt i = 0; i < impl->numein + impl->numeout; i++) { 30 CeedCallBackend(CeedVectorDestroy(&impl->evecs[i])); 31 } 32 CeedCallBackend(CeedFree(&impl->evecs)); 33 34 for (CeedInt i = 0; i < impl->numein; i++) { 35 CeedCallBackend(CeedVectorDestroy(&impl->qvecsin[i])); 36 } 37 CeedCallBackend(CeedFree(&impl->qvecsin)); 38 39 for (CeedInt i = 0; i < impl->numeout; i++) { 40 CeedCallBackend(CeedVectorDestroy(&impl->qvecsout[i])); 41 } 42 CeedCallBackend(CeedFree(&impl->qvecsout)); 43 44 // QFunction assembly data 45 for (CeedInt i = 0; i < impl->qfnumactivein; i++) { 46 CeedCallBackend(CeedVectorDestroy(&impl->qfactivein[i])); 47 } 48 CeedCallBackend(CeedFree(&impl->qfactivein)); 49 50 // Diag data 51 if (impl->diag) { 52 Ceed ceed; 53 CeedCallBackend(CeedOperatorGetCeed(op, &ceed)); 54 CeedCallCuda(ceed, cuModuleUnload(impl->diag->module)); 55 CeedCallBackend(CeedFree(&impl->diag->h_emodein)); 56 CeedCallBackend(CeedFree(&impl->diag->h_emodeout)); 57 CeedCallCuda(ceed, cudaFree(impl->diag->d_emodein)); 58 CeedCallCuda(ceed, cudaFree(impl->diag->d_emodeout)); 59 CeedCallCuda(ceed, cudaFree(impl->diag->d_identity)); 60 CeedCallCuda(ceed, cudaFree(impl->diag->d_interpin)); 61 CeedCallCuda(ceed, cudaFree(impl->diag->d_interpout)); 62 CeedCallCuda(ceed, cudaFree(impl->diag->d_gradin)); 63 CeedCallCuda(ceed, cudaFree(impl->diag->d_gradout)); 64 CeedCallBackend(CeedElemRestrictionDestroy(&impl->diag->pbdiagrstr)); 65 CeedCallBackend(CeedVectorDestroy(&impl->diag->elemdiag)); 66 CeedCallBackend(CeedVectorDestroy(&impl->diag->pbelemdiag)); 67 } 68 CeedCallBackend(CeedFree(&impl->diag)); 69 70 if (impl->asmb) { 71 Ceed ceed; 72 CeedCallBackend(CeedOperatorGetCeed(op, &ceed)); 73 CeedCallCuda(ceed, cuModuleUnload(impl->asmb->module)); 74 CeedCallCuda(ceed, cudaFree(impl->asmb->d_B_in)); 75 CeedCallCuda(ceed, cudaFree(impl->asmb->d_B_out)); 76 } 77 CeedCallBackend(CeedFree(&impl->asmb)); 78 79 CeedCallBackend(CeedFree(&impl)); 80 return CEED_ERROR_SUCCESS; 81 } 82 83 //------------------------------------------------------------------------------ 84 // Setup infields or outfields 85 //------------------------------------------------------------------------------ 86 static int CeedOperatorSetupFields_Cuda(CeedQFunction qf, CeedOperator op, bool isinput, CeedVector *evecs, CeedVector *qvecs, CeedInt starte, 87 CeedInt numfields, CeedInt Q, CeedInt numelements) { 88 CeedInt dim, size; 89 CeedSize q_size; 90 Ceed ceed; 91 CeedCallBackend(CeedOperatorGetCeed(op, &ceed)); 92 CeedBasis basis; 93 CeedElemRestriction Erestrict; 94 CeedOperatorField *opfields; 95 CeedQFunctionField *qffields; 96 CeedVector fieldvec; 97 bool strided; 98 bool skiprestrict; 99 100 if (isinput) { 101 CeedCallBackend(CeedOperatorGetFields(op, NULL, &opfields, NULL, NULL)); 102 CeedCallBackend(CeedQFunctionGetFields(qf, NULL, &qffields, NULL, NULL)); 103 } else { 104 CeedCallBackend(CeedOperatorGetFields(op, NULL, NULL, NULL, &opfields)); 105 CeedCallBackend(CeedQFunctionGetFields(qf, NULL, NULL, NULL, &qffields)); 106 } 107 108 // Loop over fields 109 for (CeedInt i = 0; i < numfields; i++) { 110 CeedEvalMode emode; 111 CeedCallBackend(CeedQFunctionFieldGetEvalMode(qffields[i], &emode)); 112 113 strided = false; 114 skiprestrict = false; 115 if (emode != CEED_EVAL_WEIGHT) { 116 CeedCallBackend(CeedOperatorFieldGetElemRestriction(opfields[i], &Erestrict)); 117 118 // Check whether this field can skip the element restriction: 119 // must be passive input, with emode NONE, and have a strided restriction with CEED_STRIDES_BACKEND. 120 121 // First, check whether the field is input or output: 122 if (isinput) { 123 // Check for passive input: 124 CeedCallBackend(CeedOperatorFieldGetVector(opfields[i], &fieldvec)); 125 if (fieldvec != CEED_VECTOR_ACTIVE) { 126 // Check emode 127 if (emode == CEED_EVAL_NONE) { 128 // Check for strided restriction 129 CeedCallBackend(CeedElemRestrictionIsStrided(Erestrict, &strided)); 130 if (strided) { 131 // Check if vector is already in preferred backend ordering 132 CeedCallBackend(CeedElemRestrictionHasBackendStrides(Erestrict, &skiprestrict)); 133 } 134 } 135 } 136 } 137 if (skiprestrict) { 138 // We do not need an E-Vector, but will use the input field vector's data directly in the operator application. 139 evecs[i + starte] = NULL; 140 } else { 141 CeedCallBackend(CeedElemRestrictionCreateVector(Erestrict, NULL, &evecs[i + starte])); 142 } 143 } 144 145 switch (emode) { 146 case CEED_EVAL_NONE: 147 CeedCallBackend(CeedQFunctionFieldGetSize(qffields[i], &size)); 148 q_size = (CeedSize)numelements * Q * size; 149 CeedCallBackend(CeedVectorCreate(ceed, q_size, &qvecs[i])); 150 break; 151 case CEED_EVAL_INTERP: 152 CeedCallBackend(CeedQFunctionFieldGetSize(qffields[i], &size)); 153 q_size = (CeedSize)numelements * Q * size; 154 CeedCallBackend(CeedVectorCreate(ceed, q_size, &qvecs[i])); 155 break; 156 case CEED_EVAL_GRAD: 157 CeedCallBackend(CeedOperatorFieldGetBasis(opfields[i], &basis)); 158 CeedCallBackend(CeedQFunctionFieldGetSize(qffields[i], &size)); 159 CeedCallBackend(CeedBasisGetDimension(basis, &dim)); 160 q_size = (CeedSize)numelements * Q * size; 161 CeedCallBackend(CeedVectorCreate(ceed, q_size, &qvecs[i])); 162 break; 163 case CEED_EVAL_WEIGHT: // Only on input fields 164 CeedCallBackend(CeedOperatorFieldGetBasis(opfields[i], &basis)); 165 q_size = (CeedSize)numelements * Q; 166 CeedCallBackend(CeedVectorCreate(ceed, q_size, &qvecs[i])); 167 CeedCallBackend(CeedBasisApply(basis, numelements, CEED_NOTRANSPOSE, CEED_EVAL_WEIGHT, NULL, qvecs[i])); 168 break; 169 case CEED_EVAL_DIV: 170 break; // TODO: Not implemented 171 case CEED_EVAL_CURL: 172 break; // TODO: Not implemented 173 } 174 } 175 return CEED_ERROR_SUCCESS; 176 } 177 178 //------------------------------------------------------------------------------ 179 // CeedOperator needs to connect all the named fields (be they active or passive) to the named inputs and outputs of its CeedQFunction. 180 //------------------------------------------------------------------------------ 181 static int CeedOperatorSetup_Cuda(CeedOperator op) { 182 bool setupdone; 183 CeedCallBackend(CeedOperatorIsSetupDone(op, &setupdone)); 184 if (setupdone) return CEED_ERROR_SUCCESS; 185 Ceed ceed; 186 CeedCallBackend(CeedOperatorGetCeed(op, &ceed)); 187 CeedOperator_Cuda *impl; 188 CeedCallBackend(CeedOperatorGetData(op, &impl)); 189 CeedQFunction qf; 190 CeedCallBackend(CeedOperatorGetQFunction(op, &qf)); 191 CeedInt Q, numelements, numinputfields, numoutputfields; 192 CeedCallBackend(CeedOperatorGetNumQuadraturePoints(op, &Q)); 193 CeedCallBackend(CeedOperatorGetNumElements(op, &numelements)); 194 CeedOperatorField *opinputfields, *opoutputfields; 195 CeedCallBackend(CeedOperatorGetFields(op, &numinputfields, &opinputfields, &numoutputfields, &opoutputfields)); 196 CeedQFunctionField *qfinputfields, *qfoutputfields; 197 CeedCallBackend(CeedQFunctionGetFields(qf, NULL, &qfinputfields, NULL, &qfoutputfields)); 198 199 // Allocate 200 CeedCallBackend(CeedCalloc(numinputfields + numoutputfields, &impl->evecs)); 201 202 CeedCallBackend(CeedCalloc(CEED_FIELD_MAX, &impl->qvecsin)); 203 CeedCallBackend(CeedCalloc(CEED_FIELD_MAX, &impl->qvecsout)); 204 205 impl->numein = numinputfields; 206 impl->numeout = numoutputfields; 207 208 // Set up infield and outfield evecs and qvecs 209 // Infields 210 CeedCallBackend(CeedOperatorSetupFields_Cuda(qf, op, true, impl->evecs, impl->qvecsin, 0, numinputfields, Q, numelements)); 211 212 // Outfields 213 CeedCallBackend(CeedOperatorSetupFields_Cuda(qf, op, false, impl->evecs, impl->qvecsout, numinputfields, numoutputfields, Q, numelements)); 214 215 CeedCallBackend(CeedOperatorSetSetupDone(op)); 216 return CEED_ERROR_SUCCESS; 217 } 218 219 //------------------------------------------------------------------------------ 220 // Setup Operator Inputs 221 //------------------------------------------------------------------------------ 222 static inline int CeedOperatorSetupInputs_Cuda(CeedInt numinputfields, CeedQFunctionField *qfinputfields, CeedOperatorField *opinputfields, 223 CeedVector invec, const bool skipactive, CeedScalar *edata[2 * CEED_FIELD_MAX], 224 CeedOperator_Cuda *impl, CeedRequest *request) { 225 CeedEvalMode emode; 226 CeedVector vec; 227 CeedElemRestriction Erestrict; 228 229 for (CeedInt i = 0; i < numinputfields; i++) { 230 // Get input vector 231 CeedCallBackend(CeedOperatorFieldGetVector(opinputfields[i], &vec)); 232 if (vec == CEED_VECTOR_ACTIVE) { 233 if (skipactive) continue; 234 else vec = invec; 235 } 236 237 CeedCallBackend(CeedQFunctionFieldGetEvalMode(qfinputfields[i], &emode)); 238 if (emode == CEED_EVAL_WEIGHT) { // Skip 239 } else { 240 // Get input vector 241 CeedCallBackend(CeedOperatorFieldGetVector(opinputfields[i], &vec)); 242 // Get input element restriction 243 CeedCallBackend(CeedOperatorFieldGetElemRestriction(opinputfields[i], &Erestrict)); 244 if (vec == CEED_VECTOR_ACTIVE) vec = invec; 245 // Restrict, if necessary 246 if (!impl->evecs[i]) { 247 // No restriction for this field; read data directly from vec. 248 CeedCallBackend(CeedVectorGetArrayRead(vec, CEED_MEM_DEVICE, (const CeedScalar **)&edata[i])); 249 } else { 250 CeedCallBackend(CeedElemRestrictionApply(Erestrict, CEED_NOTRANSPOSE, vec, impl->evecs[i], request)); 251 // Get evec 252 CeedCallBackend(CeedVectorGetArrayRead(impl->evecs[i], CEED_MEM_DEVICE, (const CeedScalar **)&edata[i])); 253 } 254 } 255 } 256 return CEED_ERROR_SUCCESS; 257 } 258 259 //------------------------------------------------------------------------------ 260 // Input Basis Action 261 //------------------------------------------------------------------------------ 262 static inline int CeedOperatorInputBasis_Cuda(CeedInt numelements, CeedQFunctionField *qfinputfields, CeedOperatorField *opinputfields, 263 CeedInt numinputfields, const bool skipactive, CeedScalar *edata[2 * CEED_FIELD_MAX], 264 CeedOperator_Cuda *impl) { 265 CeedInt elemsize, size; 266 CeedElemRestriction Erestrict; 267 CeedEvalMode emode; 268 CeedBasis basis; 269 270 for (CeedInt i = 0; i < numinputfields; i++) { 271 // Skip active input 272 if (skipactive) { 273 CeedVector vec; 274 CeedCallBackend(CeedOperatorFieldGetVector(opinputfields[i], &vec)); 275 if (vec == CEED_VECTOR_ACTIVE) continue; 276 } 277 // Get elemsize, emode, size 278 CeedCallBackend(CeedOperatorFieldGetElemRestriction(opinputfields[i], &Erestrict)); 279 CeedCallBackend(CeedElemRestrictionGetElementSize(Erestrict, &elemsize)); 280 CeedCallBackend(CeedQFunctionFieldGetEvalMode(qfinputfields[i], &emode)); 281 CeedCallBackend(CeedQFunctionFieldGetSize(qfinputfields[i], &size)); 282 // Basis action 283 switch (emode) { 284 case CEED_EVAL_NONE: 285 CeedCallBackend(CeedVectorSetArray(impl->qvecsin[i], CEED_MEM_DEVICE, CEED_USE_POINTER, edata[i])); 286 break; 287 case CEED_EVAL_INTERP: 288 CeedCallBackend(CeedOperatorFieldGetBasis(opinputfields[i], &basis)); 289 CeedCallBackend(CeedBasisApply(basis, numelements, CEED_NOTRANSPOSE, CEED_EVAL_INTERP, impl->evecs[i], impl->qvecsin[i])); 290 break; 291 case CEED_EVAL_GRAD: 292 CeedCallBackend(CeedOperatorFieldGetBasis(opinputfields[i], &basis)); 293 CeedCallBackend(CeedBasisApply(basis, numelements, CEED_NOTRANSPOSE, CEED_EVAL_GRAD, impl->evecs[i], impl->qvecsin[i])); 294 break; 295 case CEED_EVAL_WEIGHT: 296 break; // No action 297 case CEED_EVAL_DIV: 298 break; // TODO: Not implemented 299 case CEED_EVAL_CURL: 300 break; // TODO: Not implemented 301 } 302 } 303 return CEED_ERROR_SUCCESS; 304 } 305 306 //------------------------------------------------------------------------------ 307 // Restore Input Vectors 308 //------------------------------------------------------------------------------ 309 static inline int CeedOperatorRestoreInputs_Cuda(CeedInt numinputfields, CeedQFunctionField *qfinputfields, CeedOperatorField *opinputfields, 310 const bool skipactive, CeedScalar *edata[2 * CEED_FIELD_MAX], CeedOperator_Cuda *impl) { 311 CeedEvalMode emode; 312 CeedVector vec; 313 314 for (CeedInt i = 0; i < numinputfields; i++) { 315 // Skip active input 316 if (skipactive) { 317 CeedCallBackend(CeedOperatorFieldGetVector(opinputfields[i], &vec)); 318 if (vec == CEED_VECTOR_ACTIVE) continue; 319 } 320 CeedCallBackend(CeedQFunctionFieldGetEvalMode(qfinputfields[i], &emode)); 321 if (emode == CEED_EVAL_WEIGHT) { // Skip 322 } else { 323 if (!impl->evecs[i]) { // This was a skiprestrict case 324 CeedCallBackend(CeedOperatorFieldGetVector(opinputfields[i], &vec)); 325 CeedCallBackend(CeedVectorRestoreArrayRead(vec, (const CeedScalar **)&edata[i])); 326 } else { 327 CeedCallBackend(CeedVectorRestoreArrayRead(impl->evecs[i], (const CeedScalar **)&edata[i])); 328 } 329 } 330 } 331 return CEED_ERROR_SUCCESS; 332 } 333 334 //------------------------------------------------------------------------------ 335 // Apply and add to output 336 //------------------------------------------------------------------------------ 337 static int CeedOperatorApplyAdd_Cuda(CeedOperator op, CeedVector invec, CeedVector outvec, CeedRequest *request) { 338 CeedOperator_Cuda *impl; 339 CeedCallBackend(CeedOperatorGetData(op, &impl)); 340 CeedQFunction qf; 341 CeedCallBackend(CeedOperatorGetQFunction(op, &qf)); 342 CeedInt Q, numelements, elemsize, numinputfields, numoutputfields, size; 343 CeedCallBackend(CeedOperatorGetNumQuadraturePoints(op, &Q)); 344 CeedCallBackend(CeedOperatorGetNumElements(op, &numelements)); 345 CeedOperatorField *opinputfields, *opoutputfields; 346 CeedCallBackend(CeedOperatorGetFields(op, &numinputfields, &opinputfields, &numoutputfields, &opoutputfields)); 347 CeedQFunctionField *qfinputfields, *qfoutputfields; 348 CeedCallBackend(CeedQFunctionGetFields(qf, NULL, &qfinputfields, NULL, &qfoutputfields)); 349 CeedEvalMode emode; 350 CeedVector vec; 351 CeedBasis basis; 352 CeedElemRestriction Erestrict; 353 CeedScalar *edata[2 * CEED_FIELD_MAX] = {0}; 354 355 // Setup 356 CeedCallBackend(CeedOperatorSetup_Cuda(op)); 357 358 // Input Evecs and Restriction 359 CeedCallBackend(CeedOperatorSetupInputs_Cuda(numinputfields, qfinputfields, opinputfields, invec, false, edata, impl, request)); 360 361 // Input basis apply if needed 362 CeedCallBackend(CeedOperatorInputBasis_Cuda(numelements, qfinputfields, opinputfields, numinputfields, false, edata, impl)); 363 364 // Output pointers, as necessary 365 for (CeedInt i = 0; i < numoutputfields; i++) { 366 CeedCallBackend(CeedQFunctionFieldGetEvalMode(qfoutputfields[i], &emode)); 367 if (emode == CEED_EVAL_NONE) { 368 // Set the output Q-Vector to use the E-Vector data directly. 369 CeedCallBackend(CeedVectorGetArrayWrite(impl->evecs[i + impl->numein], CEED_MEM_DEVICE, &edata[i + numinputfields])); 370 CeedCallBackend(CeedVectorSetArray(impl->qvecsout[i], CEED_MEM_DEVICE, CEED_USE_POINTER, edata[i + numinputfields])); 371 } 372 } 373 374 // Q function 375 CeedCallBackend(CeedQFunctionApply(qf, numelements * Q, impl->qvecsin, impl->qvecsout)); 376 377 // Output basis apply if needed 378 for (CeedInt i = 0; i < numoutputfields; i++) { 379 // Get elemsize, emode, size 380 CeedCallBackend(CeedOperatorFieldGetElemRestriction(opoutputfields[i], &Erestrict)); 381 CeedCallBackend(CeedElemRestrictionGetElementSize(Erestrict, &elemsize)); 382 CeedCallBackend(CeedQFunctionFieldGetEvalMode(qfoutputfields[i], &emode)); 383 CeedCallBackend(CeedQFunctionFieldGetSize(qfoutputfields[i], &size)); 384 // Basis action 385 switch (emode) { 386 case CEED_EVAL_NONE: 387 break; 388 case CEED_EVAL_INTERP: 389 CeedCallBackend(CeedOperatorFieldGetBasis(opoutputfields[i], &basis)); 390 CeedCallBackend(CeedBasisApply(basis, numelements, CEED_TRANSPOSE, CEED_EVAL_INTERP, impl->qvecsout[i], impl->evecs[i + impl->numein])); 391 break; 392 case CEED_EVAL_GRAD: 393 CeedCallBackend(CeedOperatorFieldGetBasis(opoutputfields[i], &basis)); 394 CeedCallBackend(CeedBasisApply(basis, numelements, CEED_TRANSPOSE, CEED_EVAL_GRAD, impl->qvecsout[i], impl->evecs[i + impl->numein])); 395 break; 396 // LCOV_EXCL_START 397 case CEED_EVAL_WEIGHT: { 398 Ceed ceed; 399 CeedCallBackend(CeedOperatorGetCeed(op, &ceed)); 400 return CeedError(ceed, CEED_ERROR_BACKEND, "CEED_EVAL_WEIGHT cannot be an output evaluation mode"); 401 break; // Should not occur 402 } 403 case CEED_EVAL_DIV: 404 break; // TODO: Not implemented 405 case CEED_EVAL_CURL: 406 break; // TODO: Not implemented 407 // LCOV_EXCL_STOP 408 } 409 } 410 411 // Output restriction 412 for (CeedInt i = 0; i < numoutputfields; i++) { 413 // Restore evec 414 CeedCallBackend(CeedQFunctionFieldGetEvalMode(qfoutputfields[i], &emode)); 415 if (emode == CEED_EVAL_NONE) { 416 CeedCallBackend(CeedVectorRestoreArray(impl->evecs[i + impl->numein], &edata[i + numinputfields])); 417 } 418 // Get output vector 419 CeedCallBackend(CeedOperatorFieldGetVector(opoutputfields[i], &vec)); 420 // Restrict 421 CeedCallBackend(CeedOperatorFieldGetElemRestriction(opoutputfields[i], &Erestrict)); 422 // Active 423 if (vec == CEED_VECTOR_ACTIVE) vec = outvec; 424 425 CeedCallBackend(CeedElemRestrictionApply(Erestrict, CEED_TRANSPOSE, impl->evecs[i + impl->numein], vec, request)); 426 } 427 428 // Restore input arrays 429 CeedCallBackend(CeedOperatorRestoreInputs_Cuda(numinputfields, qfinputfields, opinputfields, false, edata, impl)); 430 return CEED_ERROR_SUCCESS; 431 } 432 433 //------------------------------------------------------------------------------ 434 // Core code for assembling linear QFunction 435 //------------------------------------------------------------------------------ 436 static inline int CeedOperatorLinearAssembleQFunctionCore_Cuda(CeedOperator op, bool build_objects, CeedVector *assembled, CeedElemRestriction *rstr, 437 CeedRequest *request) { 438 CeedOperator_Cuda *impl; 439 CeedCallBackend(CeedOperatorGetData(op, &impl)); 440 CeedQFunction qf; 441 CeedCallBackend(CeedOperatorGetQFunction(op, &qf)); 442 CeedInt Q, numelements, numinputfields, numoutputfields, size; 443 CeedSize q_size; 444 CeedCallBackend(CeedOperatorGetNumQuadraturePoints(op, &Q)); 445 CeedCallBackend(CeedOperatorGetNumElements(op, &numelements)); 446 CeedOperatorField *opinputfields, *opoutputfields; 447 CeedCallBackend(CeedOperatorGetFields(op, &numinputfields, &opinputfields, &numoutputfields, &opoutputfields)); 448 CeedQFunctionField *qfinputfields, *qfoutputfields; 449 CeedCallBackend(CeedQFunctionGetFields(qf, NULL, &qfinputfields, NULL, &qfoutputfields)); 450 CeedVector vec; 451 CeedInt numactivein = impl->qfnumactivein, numactiveout = impl->qfnumactiveout; 452 CeedVector *activein = impl->qfactivein; 453 CeedScalar *a, *tmp; 454 Ceed ceed, ceedparent; 455 CeedCallBackend(CeedOperatorGetCeed(op, &ceed)); 456 CeedCallBackend(CeedGetOperatorFallbackParentCeed(ceed, &ceedparent)); 457 ceedparent = ceedparent ? ceedparent : ceed; 458 CeedScalar *edata[2 * CEED_FIELD_MAX]; 459 460 // Setup 461 CeedCallBackend(CeedOperatorSetup_Cuda(op)); 462 463 // Check for identity 464 bool identityqf; 465 CeedCallBackend(CeedQFunctionIsIdentity(qf, &identityqf)); 466 if (identityqf) { 467 // LCOV_EXCL_START 468 return CeedError(ceed, CEED_ERROR_BACKEND, "Assembling identity QFunctions not supported"); 469 // LCOV_EXCL_STOP 470 } 471 472 // Input Evecs and Restriction 473 CeedCallBackend(CeedOperatorSetupInputs_Cuda(numinputfields, qfinputfields, opinputfields, NULL, true, edata, impl, request)); 474 475 // Count number of active input fields 476 if (!numactivein) { 477 for (CeedInt i = 0; i < numinputfields; i++) { 478 // Get input vector 479 CeedCallBackend(CeedOperatorFieldGetVector(opinputfields[i], &vec)); 480 // Check if active input 481 if (vec == CEED_VECTOR_ACTIVE) { 482 CeedCallBackend(CeedQFunctionFieldGetSize(qfinputfields[i], &size)); 483 CeedCallBackend(CeedVectorSetValue(impl->qvecsin[i], 0.0)); 484 CeedCallBackend(CeedVectorGetArray(impl->qvecsin[i], CEED_MEM_DEVICE, &tmp)); 485 CeedCallBackend(CeedRealloc(numactivein + size, &activein)); 486 for (CeedInt field = 0; field < size; field++) { 487 q_size = (CeedSize)Q * numelements; 488 CeedCallBackend(CeedVectorCreate(ceed, q_size, &activein[numactivein + field])); 489 CeedCallBackend(CeedVectorSetArray(activein[numactivein + field], CEED_MEM_DEVICE, CEED_USE_POINTER, &tmp[field * Q * numelements])); 490 } 491 numactivein += size; 492 CeedCallBackend(CeedVectorRestoreArray(impl->qvecsin[i], &tmp)); 493 } 494 } 495 impl->qfnumactivein = numactivein; 496 impl->qfactivein = activein; 497 } 498 499 // Count number of active output fields 500 if (!numactiveout) { 501 for (CeedInt i = 0; i < numoutputfields; i++) { 502 // Get output vector 503 CeedCallBackend(CeedOperatorFieldGetVector(opoutputfields[i], &vec)); 504 // Check if active output 505 if (vec == CEED_VECTOR_ACTIVE) { 506 CeedCallBackend(CeedQFunctionFieldGetSize(qfoutputfields[i], &size)); 507 numactiveout += size; 508 } 509 } 510 impl->qfnumactiveout = numactiveout; 511 } 512 513 // Check sizes 514 if (!numactivein || !numactiveout) { 515 // LCOV_EXCL_START 516 return CeedError(ceed, CEED_ERROR_BACKEND, "Cannot assemble QFunction without active inputs and outputs"); 517 // LCOV_EXCL_STOP 518 } 519 520 // Build objects if needed 521 if (build_objects) { 522 // Create output restriction 523 CeedInt strides[3] = {1, numelements * Q, Q}; /* *NOPAD* */ 524 CeedCallBackend(CeedElemRestrictionCreateStrided(ceedparent, numelements, Q, numactivein * numactiveout, 525 numactivein * numactiveout * numelements * Q, strides, rstr)); 526 // Create assembled vector 527 CeedSize l_size = (CeedSize)numelements * Q * numactivein * numactiveout; 528 CeedCallBackend(CeedVectorCreate(ceedparent, l_size, assembled)); 529 } 530 CeedCallBackend(CeedVectorSetValue(*assembled, 0.0)); 531 CeedCallBackend(CeedVectorGetArray(*assembled, CEED_MEM_DEVICE, &a)); 532 533 // Input basis apply 534 CeedCallBackend(CeedOperatorInputBasis_Cuda(numelements, qfinputfields, opinputfields, numinputfields, true, edata, impl)); 535 536 // Assemble QFunction 537 for (CeedInt in = 0; in < numactivein; in++) { 538 // Set Inputs 539 CeedCallBackend(CeedVectorSetValue(activein[in], 1.0)); 540 if (numactivein > 1) { 541 CeedCallBackend(CeedVectorSetValue(activein[(in + numactivein - 1) % numactivein], 0.0)); 542 } 543 // Set Outputs 544 for (CeedInt out = 0; out < numoutputfields; out++) { 545 // Get output vector 546 CeedCallBackend(CeedOperatorFieldGetVector(opoutputfields[out], &vec)); 547 // Check if active output 548 if (vec == CEED_VECTOR_ACTIVE) { 549 CeedCallBackend(CeedVectorSetArray(impl->qvecsout[out], CEED_MEM_DEVICE, CEED_USE_POINTER, a)); 550 CeedCallBackend(CeedQFunctionFieldGetSize(qfoutputfields[out], &size)); 551 a += size * Q * numelements; // Advance the pointer by the size of the output 552 } 553 } 554 // Apply QFunction 555 CeedCallBackend(CeedQFunctionApply(qf, Q * numelements, impl->qvecsin, impl->qvecsout)); 556 } 557 558 // Un-set output Qvecs to prevent accidental overwrite of Assembled 559 for (CeedInt out = 0; out < numoutputfields; out++) { 560 // Get output vector 561 CeedCallBackend(CeedOperatorFieldGetVector(opoutputfields[out], &vec)); 562 // Check if active output 563 if (vec == CEED_VECTOR_ACTIVE) { 564 CeedCallBackend(CeedVectorTakeArray(impl->qvecsout[out], CEED_MEM_DEVICE, NULL)); 565 } 566 } 567 568 // Restore input arrays 569 CeedCallBackend(CeedOperatorRestoreInputs_Cuda(numinputfields, qfinputfields, opinputfields, true, edata, impl)); 570 571 // Restore output 572 CeedCallBackend(CeedVectorRestoreArray(*assembled, &a)); 573 574 return CEED_ERROR_SUCCESS; 575 } 576 577 //------------------------------------------------------------------------------ 578 // Assemble Linear QFunction 579 //------------------------------------------------------------------------------ 580 static int CeedOperatorLinearAssembleQFunction_Cuda(CeedOperator op, CeedVector *assembled, CeedElemRestriction *rstr, CeedRequest *request) { 581 return CeedOperatorLinearAssembleQFunctionCore_Cuda(op, true, assembled, rstr, request); 582 } 583 584 //------------------------------------------------------------------------------ 585 // Update Assembled Linear QFunction 586 //------------------------------------------------------------------------------ 587 static int CeedOperatorLinearAssembleQFunctionUpdate_Cuda(CeedOperator op, CeedVector assembled, CeedElemRestriction rstr, CeedRequest *request) { 588 return CeedOperatorLinearAssembleQFunctionCore_Cuda(op, false, &assembled, &rstr, request); 589 } 590 591 //------------------------------------------------------------------------------ 592 // Create point block restriction 593 //------------------------------------------------------------------------------ 594 static int CreatePBRestriction(CeedElemRestriction rstr, CeedElemRestriction *pbRstr) { 595 Ceed ceed; 596 CeedCallBackend(CeedElemRestrictionGetCeed(rstr, &ceed)); 597 const CeedInt *offsets; 598 CeedCallBackend(CeedElemRestrictionGetOffsets(rstr, CEED_MEM_HOST, &offsets)); 599 600 // Expand offsets 601 CeedInt nelem, ncomp, elemsize, compstride, *pbOffsets; 602 CeedSize l_size; 603 CeedCallBackend(CeedElemRestrictionGetNumElements(rstr, &nelem)); 604 CeedCallBackend(CeedElemRestrictionGetNumComponents(rstr, &ncomp)); 605 CeedCallBackend(CeedElemRestrictionGetElementSize(rstr, &elemsize)); 606 CeedCallBackend(CeedElemRestrictionGetCompStride(rstr, &compstride)); 607 CeedCallBackend(CeedElemRestrictionGetLVectorSize(rstr, &l_size)); 608 CeedInt shift = ncomp; 609 if (compstride != 1) shift *= ncomp; 610 CeedCallBackend(CeedCalloc(nelem * elemsize, &pbOffsets)); 611 for (CeedInt i = 0; i < nelem * elemsize; i++) { 612 pbOffsets[i] = offsets[i] * shift; 613 } 614 615 // Create new restriction 616 CeedCallBackend( 617 CeedElemRestrictionCreate(ceed, nelem, elemsize, ncomp * ncomp, 1, l_size * ncomp, CEED_MEM_HOST, CEED_OWN_POINTER, pbOffsets, pbRstr)); 618 619 // Cleanup 620 CeedCallBackend(CeedElemRestrictionRestoreOffsets(rstr, &offsets)); 621 622 return CEED_ERROR_SUCCESS; 623 } 624 625 //------------------------------------------------------------------------------ 626 // Assemble diagonal setup 627 //------------------------------------------------------------------------------ 628 static inline int CeedOperatorAssembleDiagonalSetup_Cuda(CeedOperator op, const bool pointBlock) { 629 Ceed ceed; 630 CeedCallBackend(CeedOperatorGetCeed(op, &ceed)); 631 CeedQFunction qf; 632 CeedCallBackend(CeedOperatorGetQFunction(op, &qf)); 633 CeedInt numinputfields, numoutputfields; 634 CeedCallBackend(CeedQFunctionGetNumArgs(qf, &numinputfields, &numoutputfields)); 635 636 // Determine active input basis 637 CeedOperatorField *opfields; 638 CeedQFunctionField *qffields; 639 CeedCallBackend(CeedOperatorGetFields(op, NULL, &opfields, NULL, NULL)); 640 CeedCallBackend(CeedQFunctionGetFields(qf, NULL, &qffields, NULL, NULL)); 641 CeedInt numemodein = 0, ncomp = 0, dim = 1; 642 CeedEvalMode *emodein = NULL; 643 CeedBasis basisin = NULL; 644 CeedElemRestriction rstrin = NULL; 645 for (CeedInt i = 0; i < numinputfields; i++) { 646 CeedVector vec; 647 CeedCallBackend(CeedOperatorFieldGetVector(opfields[i], &vec)); 648 if (vec == CEED_VECTOR_ACTIVE) { 649 CeedElemRestriction rstr; 650 CeedCallBackend(CeedOperatorFieldGetBasis(opfields[i], &basisin)); 651 CeedCallBackend(CeedBasisGetNumComponents(basisin, &ncomp)); 652 CeedCallBackend(CeedBasisGetDimension(basisin, &dim)); 653 CeedCallBackend(CeedOperatorFieldGetElemRestriction(opfields[i], &rstr)); 654 if (rstrin && rstrin != rstr) { 655 // LCOV_EXCL_START 656 return CeedError(ceed, CEED_ERROR_BACKEND, "Backend does not implement multi-field non-composite operator diagonal assembly"); 657 // LCOV_EXCL_STOP 658 } 659 rstrin = rstr; 660 CeedEvalMode emode; 661 CeedCallBackend(CeedQFunctionFieldGetEvalMode(qffields[i], &emode)); 662 switch (emode) { 663 case CEED_EVAL_NONE: 664 case CEED_EVAL_INTERP: 665 CeedCallBackend(CeedRealloc(numemodein + 1, &emodein)); 666 emodein[numemodein] = emode; 667 numemodein += 1; 668 break; 669 case CEED_EVAL_GRAD: 670 CeedCallBackend(CeedRealloc(numemodein + dim, &emodein)); 671 for (CeedInt d = 0; d < dim; d++) emodein[numemodein + d] = emode; 672 numemodein += dim; 673 break; 674 case CEED_EVAL_WEIGHT: 675 case CEED_EVAL_DIV: 676 case CEED_EVAL_CURL: 677 break; // Caught by QF Assembly 678 } 679 } 680 } 681 682 // Determine active output basis 683 CeedCallBackend(CeedOperatorGetFields(op, NULL, NULL, NULL, &opfields)); 684 CeedCallBackend(CeedQFunctionGetFields(qf, NULL, NULL, NULL, &qffields)); 685 CeedInt numemodeout = 0; 686 CeedEvalMode *emodeout = NULL; 687 CeedBasis basisout = NULL; 688 CeedElemRestriction rstrout = NULL; 689 for (CeedInt i = 0; i < numoutputfields; i++) { 690 CeedVector vec; 691 CeedCallBackend(CeedOperatorFieldGetVector(opfields[i], &vec)); 692 if (vec == CEED_VECTOR_ACTIVE) { 693 CeedElemRestriction rstr; 694 CeedCallBackend(CeedOperatorFieldGetBasis(opfields[i], &basisout)); 695 CeedCallBackend(CeedOperatorFieldGetElemRestriction(opfields[i], &rstr)); 696 if (rstrout && rstrout != rstr) { 697 // LCOV_EXCL_START 698 return CeedError(ceed, CEED_ERROR_BACKEND, "Backend does not implement multi-field non-composite operator diagonal assembly"); 699 // LCOV_EXCL_STOP 700 } 701 rstrout = rstr; 702 CeedEvalMode emode; 703 CeedCallBackend(CeedQFunctionFieldGetEvalMode(qffields[i], &emode)); 704 switch (emode) { 705 case CEED_EVAL_NONE: 706 case CEED_EVAL_INTERP: 707 CeedCallBackend(CeedRealloc(numemodeout + 1, &emodeout)); 708 emodeout[numemodeout] = emode; 709 numemodeout += 1; 710 break; 711 case CEED_EVAL_GRAD: 712 CeedCallBackend(CeedRealloc(numemodeout + dim, &emodeout)); 713 for (CeedInt d = 0; d < dim; d++) emodeout[numemodeout + d] = emode; 714 numemodeout += dim; 715 break; 716 case CEED_EVAL_WEIGHT: 717 case CEED_EVAL_DIV: 718 case CEED_EVAL_CURL: 719 break; // Caught by QF Assembly 720 } 721 } 722 } 723 724 // Operator data struct 725 CeedOperator_Cuda *impl; 726 CeedCallBackend(CeedOperatorGetData(op, &impl)); 727 CeedCallBackend(CeedCalloc(1, &impl->diag)); 728 CeedOperatorDiag_Cuda *diag = impl->diag; 729 diag->basisin = basisin; 730 diag->basisout = basisout; 731 diag->h_emodein = emodein; 732 diag->h_emodeout = emodeout; 733 diag->numemodein = numemodein; 734 diag->numemodeout = numemodeout; 735 736 // Assemble kernel 737 char *diagonal_kernel_path, *diagonal_kernel_source; 738 CeedCallBackend(CeedGetJitAbsolutePath(ceed, "ceed/jit-source/cuda/cuda-ref-operator-assemble-diagonal.h", &diagonal_kernel_path)); 739 CeedDebug256(ceed, 2, "----- Loading Diagonal Assembly Kernel Source -----\n"); 740 CeedCallBackend(CeedLoadSourceToBuffer(ceed, diagonal_kernel_path, &diagonal_kernel_source)); 741 CeedDebug256(ceed, 2, "----- Loading Diagonal Assembly Source Complete! -----\n"); 742 CeedInt nnodes, nqpts; 743 CeedCallBackend(CeedBasisGetNumNodes(basisin, &nnodes)); 744 CeedCallBackend(CeedBasisGetNumQuadraturePoints(basisin, &nqpts)); 745 diag->nnodes = nnodes; 746 CeedCallCuda(ceed, CeedCompileCuda(ceed, diagonal_kernel_source, &diag->module, 5, "NUMEMODEIN", numemodein, "NUMEMODEOUT", numemodeout, "NNODES", 747 nnodes, "NQPTS", nqpts, "NCOMP", ncomp)); 748 CeedCallCuda(ceed, CeedGetKernelCuda(ceed, diag->module, "linearDiagonal", &diag->linearDiagonal)); 749 CeedCallCuda(ceed, CeedGetKernelCuda(ceed, diag->module, "linearPointBlockDiagonal", &diag->linearPointBlock)); 750 CeedCallBackend(CeedFree(&diagonal_kernel_path)); 751 CeedCallBackend(CeedFree(&diagonal_kernel_source)); 752 753 // Basis matrices 754 const CeedInt qBytes = nqpts * sizeof(CeedScalar); 755 const CeedInt iBytes = qBytes * nnodes; 756 const CeedInt gBytes = qBytes * nnodes * dim; 757 const CeedInt eBytes = sizeof(CeedEvalMode); 758 const CeedScalar *interpin, *interpout, *gradin, *gradout; 759 760 // CEED_EVAL_NONE 761 CeedScalar *identity = NULL; 762 bool evalNone = false; 763 for (CeedInt i = 0; i < numemodein; i++) evalNone = evalNone || (emodein[i] == CEED_EVAL_NONE); 764 for (CeedInt i = 0; i < numemodeout; i++) evalNone = evalNone || (emodeout[i] == CEED_EVAL_NONE); 765 if (evalNone) { 766 CeedCallBackend(CeedCalloc(nqpts * nnodes, &identity)); 767 for (CeedInt i = 0; i < (nnodes < nqpts ? nnodes : nqpts); i++) identity[i * nnodes + i] = 1.0; 768 CeedCallCuda(ceed, cudaMalloc((void **)&diag->d_identity, iBytes)); 769 CeedCallCuda(ceed, cudaMemcpy(diag->d_identity, identity, iBytes, cudaMemcpyHostToDevice)); 770 } 771 772 // CEED_EVAL_INTERP 773 CeedCallBackend(CeedBasisGetInterp(basisin, &interpin)); 774 CeedCallCuda(ceed, cudaMalloc((void **)&diag->d_interpin, iBytes)); 775 CeedCallCuda(ceed, cudaMemcpy(diag->d_interpin, interpin, iBytes, cudaMemcpyHostToDevice)); 776 CeedCallBackend(CeedBasisGetInterp(basisout, &interpout)); 777 CeedCallCuda(ceed, cudaMalloc((void **)&diag->d_interpout, iBytes)); 778 CeedCallCuda(ceed, cudaMemcpy(diag->d_interpout, interpout, iBytes, cudaMemcpyHostToDevice)); 779 780 // CEED_EVAL_GRAD 781 CeedCallBackend(CeedBasisGetGrad(basisin, &gradin)); 782 CeedCallCuda(ceed, cudaMalloc((void **)&diag->d_gradin, gBytes)); 783 CeedCallCuda(ceed, cudaMemcpy(diag->d_gradin, gradin, gBytes, cudaMemcpyHostToDevice)); 784 CeedCallBackend(CeedBasisGetGrad(basisout, &gradout)); 785 CeedCallCuda(ceed, cudaMalloc((void **)&diag->d_gradout, gBytes)); 786 CeedCallCuda(ceed, cudaMemcpy(diag->d_gradout, gradout, gBytes, cudaMemcpyHostToDevice)); 787 788 // Arrays of emodes 789 CeedCallCuda(ceed, cudaMalloc((void **)&diag->d_emodein, numemodein * eBytes)); 790 CeedCallCuda(ceed, cudaMemcpy(diag->d_emodein, emodein, numemodein * eBytes, cudaMemcpyHostToDevice)); 791 CeedCallCuda(ceed, cudaMalloc((void **)&diag->d_emodeout, numemodeout * eBytes)); 792 CeedCallCuda(ceed, cudaMemcpy(diag->d_emodeout, emodeout, numemodeout * eBytes, cudaMemcpyHostToDevice)); 793 794 // Restriction 795 diag->diagrstr = rstrout; 796 797 return CEED_ERROR_SUCCESS; 798 } 799 800 //------------------------------------------------------------------------------ 801 // Assemble diagonal common code 802 //------------------------------------------------------------------------------ 803 static inline int CeedOperatorAssembleDiagonalCore_Cuda(CeedOperator op, CeedVector assembled, CeedRequest *request, const bool pointBlock) { 804 Ceed ceed; 805 CeedCallBackend(CeedOperatorGetCeed(op, &ceed)); 806 CeedOperator_Cuda *impl; 807 CeedCallBackend(CeedOperatorGetData(op, &impl)); 808 809 // Assemble QFunction 810 CeedVector assembledqf; 811 CeedElemRestriction rstr; 812 CeedCallBackend(CeedOperatorLinearAssembleQFunctionBuildOrUpdate(op, &assembledqf, &rstr, request)); 813 CeedCallBackend(CeedElemRestrictionDestroy(&rstr)); 814 815 // Setup 816 if (!impl->diag) { 817 CeedCallBackend(CeedOperatorAssembleDiagonalSetup_Cuda(op, pointBlock)); 818 } 819 CeedOperatorDiag_Cuda *diag = impl->diag; 820 assert(diag != NULL); 821 822 // Restriction 823 if (pointBlock && !diag->pbdiagrstr) { 824 CeedElemRestriction pbdiagrstr; 825 CeedCallBackend(CreatePBRestriction(diag->diagrstr, &pbdiagrstr)); 826 diag->pbdiagrstr = pbdiagrstr; 827 } 828 CeedElemRestriction diagrstr = pointBlock ? diag->pbdiagrstr : diag->diagrstr; 829 830 // Create diagonal vector 831 CeedVector elemdiag = pointBlock ? diag->pbelemdiag : diag->elemdiag; 832 if (!elemdiag) { 833 CeedCallBackend(CeedElemRestrictionCreateVector(diagrstr, NULL, &elemdiag)); 834 if (pointBlock) diag->pbelemdiag = elemdiag; 835 else diag->elemdiag = elemdiag; 836 } 837 CeedCallBackend(CeedVectorSetValue(elemdiag, 0.0)); 838 839 // Assemble element operator diagonals 840 CeedScalar *elemdiagarray; 841 const CeedScalar *assembledqfarray; 842 CeedCallBackend(CeedVectorGetArray(elemdiag, CEED_MEM_DEVICE, &elemdiagarray)); 843 CeedCallBackend(CeedVectorGetArrayRead(assembledqf, CEED_MEM_DEVICE, &assembledqfarray)); 844 CeedInt nelem; 845 CeedCallBackend(CeedElemRestrictionGetNumElements(diagrstr, &nelem)); 846 847 // Compute the diagonal of B^T D B 848 int elemsPerBlock = 1; 849 int grid = nelem / elemsPerBlock + ((nelem / elemsPerBlock * elemsPerBlock < nelem) ? 1 : 0); 850 void *args[] = {(void *)&nelem, &diag->d_identity, &diag->d_interpin, &diag->d_gradin, &diag->d_interpout, 851 &diag->d_gradout, &diag->d_emodein, &diag->d_emodeout, &assembledqfarray, &elemdiagarray}; 852 if (pointBlock) { 853 CeedCallBackend(CeedRunKernelDimCuda(ceed, diag->linearPointBlock, grid, diag->nnodes, 1, elemsPerBlock, args)); 854 } else { 855 CeedCallBackend(CeedRunKernelDimCuda(ceed, diag->linearDiagonal, grid, diag->nnodes, 1, elemsPerBlock, args)); 856 } 857 858 // Restore arrays 859 CeedCallBackend(CeedVectorRestoreArray(elemdiag, &elemdiagarray)); 860 CeedCallBackend(CeedVectorRestoreArrayRead(assembledqf, &assembledqfarray)); 861 862 // Assemble local operator diagonal 863 CeedCallBackend(CeedElemRestrictionApply(diagrstr, CEED_TRANSPOSE, elemdiag, assembled, request)); 864 865 // Cleanup 866 CeedCallBackend(CeedVectorDestroy(&assembledqf)); 867 868 return CEED_ERROR_SUCCESS; 869 } 870 871 //------------------------------------------------------------------------------ 872 // Assemble Linear Diagonal 873 //------------------------------------------------------------------------------ 874 static int CeedOperatorLinearAssembleAddDiagonal_Cuda(CeedOperator op, CeedVector assembled, CeedRequest *request) { 875 CeedCallBackend(CeedOperatorAssembleDiagonalCore_Cuda(op, assembled, request, false)); 876 return CEED_ERROR_SUCCESS; 877 } 878 879 //------------------------------------------------------------------------------ 880 // Assemble Linear Point Block Diagonal 881 //------------------------------------------------------------------------------ 882 static int CeedOperatorLinearAssembleAddPointBlockDiagonal_Cuda(CeedOperator op, CeedVector assembled, CeedRequest *request) { 883 CeedCallBackend(CeedOperatorAssembleDiagonalCore_Cuda(op, assembled, request, true)); 884 return CEED_ERROR_SUCCESS; 885 } 886 887 //------------------------------------------------------------------------------ 888 // Single operator assembly setup 889 //------------------------------------------------------------------------------ 890 static int CeedSingleOperatorAssembleSetup_Cuda(CeedOperator op) { 891 Ceed ceed; 892 CeedCallBackend(CeedOperatorGetCeed(op, &ceed)); 893 CeedOperator_Cuda *impl; 894 CeedCallBackend(CeedOperatorGetData(op, &impl)); 895 896 // Get intput and output fields 897 CeedInt num_input_fields, num_output_fields; 898 CeedOperatorField *input_fields; 899 CeedOperatorField *output_fields; 900 CeedCallBackend(CeedOperatorGetFields(op, &num_input_fields, &input_fields, &num_output_fields, &output_fields)); 901 902 // Determine active input basis eval mode 903 CeedQFunction qf; 904 CeedCallBackend(CeedOperatorGetQFunction(op, &qf)); 905 CeedQFunctionField *qf_fields; 906 CeedCallBackend(CeedQFunctionGetFields(qf, NULL, &qf_fields, NULL, NULL)); 907 // Note that the kernel will treat each dimension of a gradient action separately; 908 // i.e., when an active input has a CEED_EVAL_GRAD mode, num_emode_in will increment by dim. 909 // However, for the purposes of loading the B matrices, it will be treated as one mode, and we will load/copy the entire gradient matrix at once, so 910 // num_B_in_mats_to_load will be incremented by 1. 911 CeedInt num_emode_in = 0, dim = 1, num_B_in_mats_to_load = 0, size_B_in = 0; 912 CeedEvalMode *eval_mode_in = NULL; // will be of size num_B_in_mats_load 913 CeedBasis basis_in = NULL; 914 CeedInt nqpts = 0, esize = 0; 915 CeedElemRestriction rstr_in = NULL; 916 for (CeedInt i = 0; i < num_input_fields; i++) { 917 CeedVector vec; 918 CeedCallBackend(CeedOperatorFieldGetVector(input_fields[i], &vec)); 919 if (vec == CEED_VECTOR_ACTIVE) { 920 CeedCallBackend(CeedOperatorFieldGetBasis(input_fields[i], &basis_in)); 921 CeedCallBackend(CeedBasisGetDimension(basis_in, &dim)); 922 CeedCallBackend(CeedBasisGetNumQuadraturePoints(basis_in, &nqpts)); 923 CeedCallBackend(CeedOperatorFieldGetElemRestriction(input_fields[i], &rstr_in)); 924 CeedCallBackend(CeedElemRestrictionGetElementSize(rstr_in, &esize)); 925 CeedEvalMode eval_mode; 926 CeedCallBackend(CeedQFunctionFieldGetEvalMode(qf_fields[i], &eval_mode)); 927 if (eval_mode != CEED_EVAL_NONE) { 928 CeedCallBackend(CeedRealloc(num_B_in_mats_to_load + 1, &eval_mode_in)); 929 eval_mode_in[num_B_in_mats_to_load] = eval_mode; 930 num_B_in_mats_to_load += 1; 931 if (eval_mode == CEED_EVAL_GRAD) { 932 num_emode_in += dim; 933 size_B_in += dim * esize * nqpts; 934 } else { 935 num_emode_in += 1; 936 size_B_in += esize * nqpts; 937 } 938 } 939 } 940 } 941 942 // Determine active output basis; basis_out and rstr_out only used if same as input, TODO 943 CeedCallBackend(CeedQFunctionGetFields(qf, NULL, NULL, NULL, &qf_fields)); 944 CeedInt num_emode_out = 0, num_B_out_mats_to_load = 0, size_B_out = 0; 945 CeedEvalMode *eval_mode_out = NULL; 946 CeedBasis basis_out = NULL; 947 CeedElemRestriction rstr_out = NULL; 948 for (CeedInt i = 0; i < num_output_fields; i++) { 949 CeedVector vec; 950 CeedCallBackend(CeedOperatorFieldGetVector(output_fields[i], &vec)); 951 if (vec == CEED_VECTOR_ACTIVE) { 952 CeedCallBackend(CeedOperatorFieldGetBasis(output_fields[i], &basis_out)); 953 CeedCallBackend(CeedOperatorFieldGetElemRestriction(output_fields[i], &rstr_out)); 954 if (rstr_out && rstr_out != rstr_in) { 955 // LCOV_EXCL_START 956 return CeedError(ceed, CEED_ERROR_BACKEND, "Backend does not implement multi-field non-composite operator assembly"); 957 // LCOV_EXCL_STOP 958 } 959 CeedEvalMode eval_mode; 960 CeedCallBackend(CeedQFunctionFieldGetEvalMode(qf_fields[i], &eval_mode)); 961 if (eval_mode != CEED_EVAL_NONE) { 962 CeedCallBackend(CeedRealloc(num_B_out_mats_to_load + 1, &eval_mode_out)); 963 eval_mode_out[num_B_out_mats_to_load] = eval_mode; 964 num_B_out_mats_to_load += 1; 965 if (eval_mode == CEED_EVAL_GRAD) { 966 num_emode_out += dim; 967 size_B_out += dim * esize * nqpts; 968 } else { 969 num_emode_out += 1; 970 size_B_out += esize * nqpts; 971 } 972 } 973 } 974 } 975 976 if (num_emode_in == 0 || num_emode_out == 0) { 977 // LCOV_EXCL_START 978 return CeedError(ceed, CEED_ERROR_UNSUPPORTED, "Cannot assemble operator without inputs/outputs"); 979 // LCOV_EXCL_STOP 980 } 981 982 CeedInt nelem, ncomp; 983 CeedCallBackend(CeedElemRestrictionGetNumElements(rstr_in, &nelem)); 984 CeedCallBackend(CeedElemRestrictionGetNumComponents(rstr_in, &ncomp)); 985 986 CeedCallBackend(CeedCalloc(1, &impl->asmb)); 987 CeedOperatorAssemble_Cuda *asmb = impl->asmb; 988 asmb->nelem = nelem; 989 990 // Compile kernels 991 int elemsPerBlock = 1; 992 asmb->elemsPerBlock = elemsPerBlock; 993 CeedInt block_size = esize * esize * elemsPerBlock; 994 Ceed_Cuda *cuda_data; 995 CeedCallBackend(CeedGetData(ceed, &cuda_data)); 996 char *assembly_kernel_path, *assembly_kernel_source; 997 CeedCallBackend(CeedGetJitAbsolutePath(ceed, "ceed/jit-source/cuda/cuda-ref-operator-assemble.h", &assembly_kernel_path)); 998 CeedDebug256(ceed, 2, "----- Loading Assembly Kernel Source -----\n"); 999 CeedCallBackend(CeedLoadSourceToBuffer(ceed, assembly_kernel_path, &assembly_kernel_source)); 1000 CeedDebug256(ceed, 2, "----- Loading Assembly Source Complete! -----\n"); 1001 bool fallback = block_size > cuda_data->device_prop.maxThreadsPerBlock; 1002 if (fallback) { 1003 // Use fallback kernel with 1D threadblock 1004 block_size = esize * elemsPerBlock; 1005 asmb->block_size_x = esize; 1006 asmb->block_size_y = 1; 1007 } else { // Use kernel with 2D threadblock 1008 asmb->block_size_x = esize; 1009 asmb->block_size_y = esize; 1010 } 1011 CeedCallCuda(ceed, CeedCompileCuda(ceed, assembly_kernel_source, &asmb->module, 7, "NELEM", nelem, "NUMEMODEIN", num_emode_in, "NUMEMODEOUT", 1012 num_emode_out, "NQPTS", nqpts, "NNODES", esize, "BLOCK_SIZE", block_size, "NCOMP", ncomp)); 1013 CeedCallCuda(ceed, CeedGetKernelCuda(ceed, asmb->module, fallback ? "linearAssembleFallback" : "linearAssemble", &asmb->linearAssemble)); 1014 CeedCallBackend(CeedFree(&assembly_kernel_path)); 1015 CeedCallBackend(CeedFree(&assembly_kernel_source)); 1016 1017 // Build 'full' B matrices (not 1D arrays used for tensor-product matrices) 1018 const CeedScalar *interp_in, *grad_in; 1019 CeedCallBackend(CeedBasisGetInterp(basis_in, &interp_in)); 1020 CeedCallBackend(CeedBasisGetGrad(basis_in, &grad_in)); 1021 1022 // Load into B_in, in order that they will be used in eval_mode 1023 const CeedInt inBytes = size_B_in * sizeof(CeedScalar); 1024 CeedInt mat_start = 0; 1025 CeedCallCuda(ceed, cudaMalloc((void **)&asmb->d_B_in, inBytes)); 1026 for (int i = 0; i < num_B_in_mats_to_load; i++) { 1027 CeedEvalMode eval_mode = eval_mode_in[i]; 1028 if (eval_mode == CEED_EVAL_INTERP) { 1029 CeedCallCuda(ceed, cudaMemcpy(&asmb->d_B_in[mat_start], interp_in, esize * nqpts * sizeof(CeedScalar), cudaMemcpyHostToDevice)); 1030 mat_start += esize * nqpts; 1031 } else if (eval_mode == CEED_EVAL_GRAD) { 1032 CeedCallCuda(ceed, cudaMemcpy(&asmb->d_B_in[mat_start], grad_in, dim * esize * nqpts * sizeof(CeedScalar), cudaMemcpyHostToDevice)); 1033 mat_start += dim * esize * nqpts; 1034 } 1035 } 1036 1037 const CeedScalar *interp_out, *grad_out; 1038 // Note that this function currently assumes 1 basis, so this should always be true 1039 // for now 1040 if (basis_out == basis_in) { 1041 interp_out = interp_in; 1042 grad_out = grad_in; 1043 } else { 1044 CeedCallBackend(CeedBasisGetInterp(basis_out, &interp_out)); 1045 CeedCallBackend(CeedBasisGetGrad(basis_out, &grad_out)); 1046 } 1047 1048 // Load into B_out, in order that they will be used in eval_mode 1049 const CeedInt outBytes = size_B_out * sizeof(CeedScalar); 1050 mat_start = 0; 1051 CeedCallCuda(ceed, cudaMalloc((void **)&asmb->d_B_out, outBytes)); 1052 for (int i = 0; i < num_B_out_mats_to_load; i++) { 1053 CeedEvalMode eval_mode = eval_mode_out[i]; 1054 if (eval_mode == CEED_EVAL_INTERP) { 1055 CeedCallCuda(ceed, cudaMemcpy(&asmb->d_B_out[mat_start], interp_out, esize * nqpts * sizeof(CeedScalar), cudaMemcpyHostToDevice)); 1056 mat_start += esize * nqpts; 1057 } else if (eval_mode == CEED_EVAL_GRAD) { 1058 CeedCallCuda(ceed, cudaMemcpy(&asmb->d_B_out[mat_start], grad_out, dim * esize * nqpts * sizeof(CeedScalar), cudaMemcpyHostToDevice)); 1059 mat_start += dim * esize * nqpts; 1060 } 1061 } 1062 return CEED_ERROR_SUCCESS; 1063 } 1064 1065 //------------------------------------------------------------------------------ 1066 // Assemble matrix data for COO matrix of assembled operator. 1067 // The sparsity pattern is set by CeedOperatorLinearAssembleSymbolic. 1068 // 1069 // Note that this (and other assembly routines) currently assume only one active input restriction/basis per operator (could have multiple basis eval 1070 // modes). 1071 // TODO: allow multiple active input restrictions/basis objects 1072 //------------------------------------------------------------------------------ 1073 static int CeedSingleOperatorAssemble_Cuda(CeedOperator op, CeedInt offset, CeedVector values) { 1074 Ceed ceed; 1075 CeedCallBackend(CeedOperatorGetCeed(op, &ceed)); 1076 CeedOperator_Cuda *impl; 1077 CeedCallBackend(CeedOperatorGetData(op, &impl)); 1078 1079 // Setup 1080 if (!impl->asmb) { 1081 CeedCallBackend(CeedSingleOperatorAssembleSetup_Cuda(op)); 1082 assert(impl->asmb != NULL); 1083 } 1084 1085 // Assemble QFunction 1086 CeedVector assembled_qf; 1087 CeedElemRestriction rstr_q; 1088 CeedCallBackend(CeedOperatorLinearAssembleQFunctionBuildOrUpdate(op, &assembled_qf, &rstr_q, CEED_REQUEST_IMMEDIATE)); 1089 CeedCallBackend(CeedElemRestrictionDestroy(&rstr_q)); 1090 CeedScalar *values_array; 1091 CeedCallBackend(CeedVectorGetArray(values, CEED_MEM_DEVICE, &values_array)); 1092 values_array += offset; 1093 const CeedScalar *qf_array; 1094 CeedCallBackend(CeedVectorGetArrayRead(assembled_qf, CEED_MEM_DEVICE, &qf_array)); 1095 1096 // Compute B^T D B 1097 const CeedInt nelem = impl->asmb->nelem; 1098 const CeedInt elemsPerBlock = impl->asmb->elemsPerBlock; 1099 const CeedInt grid = nelem / elemsPerBlock + ((nelem / elemsPerBlock * elemsPerBlock < nelem) ? 1 : 0); 1100 void *args[] = {&impl->asmb->d_B_in, &impl->asmb->d_B_out, &qf_array, &values_array}; 1101 CeedCallBackend( 1102 CeedRunKernelDimCuda(ceed, impl->asmb->linearAssemble, grid, impl->asmb->block_size_x, impl->asmb->block_size_y, elemsPerBlock, args)); 1103 1104 // Restore arrays 1105 CeedCallBackend(CeedVectorRestoreArray(values, &values_array)); 1106 CeedCallBackend(CeedVectorRestoreArrayRead(assembled_qf, &qf_array)); 1107 1108 // Cleanup 1109 CeedCallBackend(CeedVectorDestroy(&assembled_qf)); 1110 1111 return CEED_ERROR_SUCCESS; 1112 } 1113 1114 //------------------------------------------------------------------------------ 1115 // Create operator 1116 //------------------------------------------------------------------------------ 1117 int CeedOperatorCreate_Cuda(CeedOperator op) { 1118 Ceed ceed; 1119 CeedCallBackend(CeedOperatorGetCeed(op, &ceed)); 1120 CeedOperator_Cuda *impl; 1121 1122 CeedCallBackend(CeedCalloc(1, &impl)); 1123 CeedCallBackend(CeedOperatorSetData(op, impl)); 1124 1125 CeedCallBackend(CeedSetBackendFunction(ceed, "Operator", op, "LinearAssembleQFunction", CeedOperatorLinearAssembleQFunction_Cuda)); 1126 CeedCallBackend(CeedSetBackendFunction(ceed, "Operator", op, "LinearAssembleQFunctionUpdate", CeedOperatorLinearAssembleQFunctionUpdate_Cuda)); 1127 CeedCallBackend(CeedSetBackendFunction(ceed, "Operator", op, "LinearAssembleAddDiagonal", CeedOperatorLinearAssembleAddDiagonal_Cuda)); 1128 CeedCallBackend( 1129 CeedSetBackendFunction(ceed, "Operator", op, "LinearAssembleAddPointBlockDiagonal", CeedOperatorLinearAssembleAddPointBlockDiagonal_Cuda)); 1130 CeedCallBackend(CeedSetBackendFunction(ceed, "Operator", op, "LinearAssembleSingle", CeedSingleOperatorAssemble_Cuda)); 1131 CeedCallBackend(CeedSetBackendFunction(ceed, "Operator", op, "ApplyAdd", CeedOperatorApplyAdd_Cuda)); 1132 CeedCallBackend(CeedSetBackendFunction(ceed, "Operator", op, "Destroy", CeedOperatorDestroy_Cuda)); 1133 return CEED_ERROR_SUCCESS; 1134 } 1135 1136 //------------------------------------------------------------------------------ 1137