1 // Copyright (c) 2017-2024, 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 <cuda.h> 12 #include <cuda_runtime.h> 13 14 #include "../cuda/ceed-cuda-common.h" 15 #include "../cuda/ceed-cuda-compile.h" 16 #include "ceed-cuda-ref.h" 17 18 //------------------------------------------------------------------------------ 19 // Basis apply - tensor 20 //------------------------------------------------------------------------------ 21 int CeedBasisApply_Cuda(CeedBasis basis, const CeedInt num_elem, CeedTransposeMode t_mode, CeedEvalMode eval_mode, CeedVector u, CeedVector v) { 22 Ceed ceed; 23 CeedInt Q_1d, dim; 24 const CeedInt is_transpose = t_mode == CEED_TRANSPOSE; 25 const int max_block_size = 32; 26 const CeedScalar *d_u; 27 CeedScalar *d_v; 28 CeedBasis_Cuda *data; 29 30 CeedCallBackend(CeedBasisGetCeed(basis, &ceed)); 31 CeedCallBackend(CeedBasisGetData(basis, &data)); 32 33 // Get read/write access to u, v 34 if (u != CEED_VECTOR_NONE) CeedCallBackend(CeedVectorGetArrayRead(u, CEED_MEM_DEVICE, &d_u)); 35 else CeedCheck(eval_mode == CEED_EVAL_WEIGHT, ceed, CEED_ERROR_BACKEND, "An input vector is required for this CeedEvalMode"); 36 CeedCallBackend(CeedVectorGetArrayWrite(v, CEED_MEM_DEVICE, &d_v)); 37 38 // Clear v for transpose operation 39 if (is_transpose) { 40 CeedSize length; 41 42 CeedCallBackend(CeedVectorGetLength(v, &length)); 43 CeedCallCuda(ceed, cudaMemset(d_v, 0, length * sizeof(CeedScalar))); 44 } 45 CeedCallBackend(CeedBasisGetNumQuadraturePoints1D(basis, &Q_1d)); 46 CeedCallBackend(CeedBasisGetDimension(basis, &dim)); 47 48 // Basis action 49 switch (eval_mode) { 50 case CEED_EVAL_INTERP: { 51 void *interp_args[] = {(void *)&num_elem, (void *)&is_transpose, &data->d_interp_1d, &d_u, &d_v}; 52 const CeedInt block_size = CeedIntMin(CeedIntPow(Q_1d, dim), max_block_size); 53 54 CeedCallBackend(CeedRunKernel_Cuda(ceed, data->Interp, num_elem, block_size, interp_args)); 55 } break; 56 case CEED_EVAL_GRAD: { 57 void *grad_args[] = {(void *)&num_elem, (void *)&is_transpose, &data->d_interp_1d, &data->d_grad_1d, &d_u, &d_v}; 58 const CeedInt block_size = max_block_size; 59 60 CeedCallBackend(CeedRunKernel_Cuda(ceed, data->Grad, num_elem, block_size, grad_args)); 61 } break; 62 case CEED_EVAL_WEIGHT: { 63 CeedCheck(data->d_q_weight_1d, ceed, CEED_ERROR_BACKEND, "%s not supported; q_weights_1d not set", CeedEvalModes[eval_mode]); 64 void *weight_args[] = {(void *)&num_elem, (void *)&data->d_q_weight_1d, &d_v}; 65 const int block_size_x = Q_1d; 66 const int block_size_y = dim >= 2 ? Q_1d : 1; 67 68 CeedCallBackend(CeedRunKernelDim_Cuda(ceed, data->Weight, num_elem, block_size_x, block_size_y, 1, weight_args)); 69 } break; 70 case CEED_EVAL_NONE: /* handled separately below */ 71 break; 72 // LCOV_EXCL_START 73 case CEED_EVAL_DIV: 74 case CEED_EVAL_CURL: 75 return CeedError(ceed, CEED_ERROR_BACKEND, "%s not supported", CeedEvalModes[eval_mode]); 76 // LCOV_EXCL_STOP 77 } 78 79 // Restore vectors, cover CEED_EVAL_NONE 80 CeedCallBackend(CeedVectorRestoreArray(v, &d_v)); 81 if (eval_mode == CEED_EVAL_NONE) CeedCallBackend(CeedVectorSetArray(v, CEED_MEM_DEVICE, CEED_COPY_VALUES, (CeedScalar *)d_u)); 82 if (eval_mode != CEED_EVAL_WEIGHT) CeedCallBackend(CeedVectorRestoreArrayRead(u, &d_u)); 83 return CEED_ERROR_SUCCESS; 84 } 85 86 //------------------------------------------------------------------------------ 87 // Basis apply - tensor AtPoints 88 //------------------------------------------------------------------------------ 89 int CeedBasisApplyAtPoints_Cuda(CeedBasis basis, const CeedInt num_elem, const CeedInt *num_points, CeedTransposeMode t_mode, CeedEvalMode eval_mode, 90 CeedVector x_ref, CeedVector u, CeedVector v) { 91 Ceed ceed; 92 CeedInt Q_1d, dim, max_num_points = num_points[0]; 93 const CeedInt is_transpose = t_mode == CEED_TRANSPOSE; 94 const int max_block_size = 32; 95 const CeedScalar *d_x, *d_u; 96 CeedScalar *d_v; 97 CeedBasis_Cuda *data; 98 99 CeedCallBackend(CeedBasisGetCeed(basis, &ceed)); 100 CeedCallBackend(CeedBasisGetData(basis, &data)); 101 CeedCallBackend(CeedBasisGetNumQuadraturePoints1D(basis, &Q_1d)); 102 CeedCallBackend(CeedBasisGetDimension(basis, &dim)); 103 104 // Check uniform number of points per elem 105 for (CeedInt i = 1; i < num_elem; i++) { 106 CeedCheck(max_num_points == num_points[i], ceed, CEED_ERROR_BACKEND, 107 "BasisApplyAtPoints only supported for the same number of points in each element"); 108 } 109 110 // Weight handled separately 111 if (eval_mode == CEED_EVAL_WEIGHT) { 112 CeedCall(CeedVectorSetValue(v, 1.0)); 113 return CEED_ERROR_SUCCESS; 114 } 115 116 // Build kernels if needed 117 if (data->num_points != max_num_points) { 118 CeedInt P_1d; 119 120 CeedCallBackend(CeedBasisGetNumNodes1D(basis, &P_1d)); 121 data->num_points = max_num_points; 122 123 // -- Create interp matrix to Chebyshev coefficients 124 if (!data->d_chebyshev_interp_1d) { 125 CeedSize interp_bytes; 126 CeedScalar *chebyshev_interp_1d; 127 128 interp_bytes = P_1d * Q_1d * sizeof(CeedScalar); 129 CeedCallBackend(CeedCalloc(P_1d * Q_1d, &chebyshev_interp_1d)); 130 CeedCall(CeedBasisGetChebyshevInterp1D(basis, chebyshev_interp_1d)); 131 CeedCallCuda(ceed, cudaMalloc((void **)&data->d_chebyshev_interp_1d, interp_bytes)); 132 CeedCallCuda(ceed, cudaMemcpy(data->d_chebyshev_interp_1d, chebyshev_interp_1d, interp_bytes, cudaMemcpyHostToDevice)); 133 CeedCallBackend(CeedFree(&chebyshev_interp_1d)); 134 } 135 136 // -- Compile kernels 137 char *basis_kernel_source; 138 const char *basis_kernel_path; 139 CeedInt num_comp; 140 141 if (data->moduleAtPoints) CeedCallCuda(ceed, cuModuleUnload(data->moduleAtPoints)); 142 CeedCallBackend(CeedBasisGetNumComponents(basis, &num_comp)); 143 CeedCallBackend(CeedGetJitAbsolutePath(ceed, "ceed/jit-source/cuda/cuda-ref-basis-tensor-at-points.h", &basis_kernel_path)); 144 CeedDebug256(ceed, CEED_DEBUG_COLOR_SUCCESS, "----- Loading Basis Kernel Source -----\n"); 145 CeedCallBackend(CeedLoadSourceToBuffer(ceed, basis_kernel_path, &basis_kernel_source)); 146 CeedDebug256(ceed, CEED_DEBUG_COLOR_SUCCESS, "----- Loading Basis Kernel Source Complete! -----\n"); 147 CeedCallBackend(CeedCompile_Cuda(ceed, basis_kernel_source, &data->moduleAtPoints, 9, "BASIS_Q_1D", Q_1d, "BASIS_P_1D", P_1d, "BASIS_BUF_LEN", 148 Q_1d * CeedIntPow(Q_1d > P_1d ? Q_1d : P_1d, dim - 1), "BASIS_DIM", dim, "BASIS_NUM_COMP", num_comp, 149 "BASIS_NUM_NODES", CeedIntPow(P_1d, dim), "BASIS_NUM_QPTS", CeedIntPow(Q_1d, dim), "BASIS_NUM_PTS", 150 max_num_points, "POINTS_BUFF_LEN", CeedIntPow(Q_1d, dim - 1))); 151 CeedCallBackend(CeedGetKernel_Cuda(ceed, data->moduleAtPoints, "InterpAtPoints", &data->InterpAtPoints)); 152 CeedCallBackend(CeedGetKernel_Cuda(ceed, data->moduleAtPoints, "GradAtPoints", &data->GradAtPoints)); 153 CeedCallBackend(CeedFree(&basis_kernel_path)); 154 CeedCallBackend(CeedFree(&basis_kernel_source)); 155 } 156 157 // Get read/write access to u, v 158 CeedCallBackend(CeedVectorGetArrayRead(x_ref, CEED_MEM_DEVICE, &d_x)); 159 if (u != CEED_VECTOR_NONE) CeedCallBackend(CeedVectorGetArrayRead(u, CEED_MEM_DEVICE, &d_u)); 160 else CeedCheck(eval_mode == CEED_EVAL_WEIGHT, ceed, CEED_ERROR_BACKEND, "An input vector is required for this CeedEvalMode"); 161 CeedCallBackend(CeedVectorGetArrayWrite(v, CEED_MEM_DEVICE, &d_v)); 162 163 // Clear v for transpose operation 164 if (is_transpose) { 165 CeedSize length; 166 167 CeedCallBackend(CeedVectorGetLength(v, &length)); 168 CeedCallCuda(ceed, cudaMemset(d_v, 0, length * sizeof(CeedScalar))); 169 } 170 171 // Basis action 172 switch (eval_mode) { 173 case CEED_EVAL_INTERP: { 174 void *interp_args[] = {(void *)&num_elem, (void *)&is_transpose, &data->d_chebyshev_interp_1d, &d_x, &d_u, &d_v}; 175 const CeedInt block_size = CeedIntMin(CeedIntPow(Q_1d, dim), max_block_size); 176 177 CeedCallBackend(CeedRunKernel_Cuda(ceed, data->InterpAtPoints, num_elem, block_size, interp_args)); 178 } break; 179 case CEED_EVAL_GRAD: { 180 void *grad_args[] = {(void *)&num_elem, (void *)&is_transpose, &data->d_chebyshev_interp_1d, &d_x, &d_u, &d_v}; 181 const CeedInt block_size = CeedIntMin(CeedIntPow(Q_1d, dim), max_block_size); 182 183 CeedCallBackend(CeedRunKernel_Cuda(ceed, data->GradAtPoints, num_elem, block_size, grad_args)); 184 } break; 185 case CEED_EVAL_WEIGHT: 186 case CEED_EVAL_NONE: /* handled separately below */ 187 break; 188 // LCOV_EXCL_START 189 case CEED_EVAL_DIV: 190 case CEED_EVAL_CURL: 191 return CeedError(ceed, CEED_ERROR_BACKEND, "%s not supported", CeedEvalModes[eval_mode]); 192 // LCOV_EXCL_STOP 193 } 194 195 // Restore vectors, cover CEED_EVAL_NONE 196 CeedCallBackend(CeedVectorRestoreArrayRead(x_ref, &d_x)); 197 CeedCallBackend(CeedVectorRestoreArray(v, &d_v)); 198 if (eval_mode == CEED_EVAL_NONE) CeedCallBackend(CeedVectorSetArray(v, CEED_MEM_DEVICE, CEED_COPY_VALUES, (CeedScalar *)d_u)); 199 if (eval_mode != CEED_EVAL_WEIGHT) CeedCallBackend(CeedVectorRestoreArrayRead(u, &d_u)); 200 return CEED_ERROR_SUCCESS; 201 } 202 203 //------------------------------------------------------------------------------ 204 // Basis apply - non-tensor 205 //------------------------------------------------------------------------------ 206 int CeedBasisApplyNonTensor_Cuda(CeedBasis basis, const CeedInt num_elem, CeedTransposeMode t_mode, CeedEvalMode eval_mode, CeedVector u, 207 CeedVector v) { 208 Ceed ceed; 209 CeedInt num_nodes, num_qpts; 210 const CeedInt is_transpose = t_mode == CEED_TRANSPOSE; 211 const int elems_per_block = 1; 212 const int grid = CeedDivUpInt(num_elem, elems_per_block); 213 const CeedScalar *d_u; 214 CeedScalar *d_v; 215 CeedBasisNonTensor_Cuda *data; 216 217 CeedCallBackend(CeedBasisGetCeed(basis, &ceed)); 218 CeedCallBackend(CeedBasisGetData(basis, &data)); 219 CeedCallBackend(CeedBasisGetNumQuadraturePoints(basis, &num_qpts)); 220 CeedCallBackend(CeedBasisGetNumNodes(basis, &num_nodes)); 221 222 // Get read/write access to u, v 223 if (u != CEED_VECTOR_NONE) CeedCallBackend(CeedVectorGetArrayRead(u, CEED_MEM_DEVICE, &d_u)); 224 else CeedCheck(eval_mode == CEED_EVAL_WEIGHT, ceed, CEED_ERROR_BACKEND, "An input vector is required for this CeedEvalMode"); 225 CeedCallBackend(CeedVectorGetArrayWrite(v, CEED_MEM_DEVICE, &d_v)); 226 227 // Clear v for transpose operation 228 if (is_transpose) { 229 CeedSize length; 230 231 CeedCallBackend(CeedVectorGetLength(v, &length)); 232 CeedCallCuda(ceed, cudaMemset(d_v, 0, length * sizeof(CeedScalar))); 233 } 234 235 // Apply basis operation 236 switch (eval_mode) { 237 case CEED_EVAL_INTERP: { 238 void *interp_args[] = {(void *)&num_elem, &data->d_interp, &d_u, &d_v}; 239 const int block_size_x = is_transpose ? num_nodes : num_qpts; 240 241 if (is_transpose) { 242 CeedCallBackend(CeedRunKernelDim_Cuda(ceed, data->InterpTranspose, grid, block_size_x, 1, elems_per_block, interp_args)); 243 } else { 244 CeedCallBackend(CeedRunKernelDim_Cuda(ceed, data->Interp, grid, block_size_x, 1, elems_per_block, interp_args)); 245 } 246 } break; 247 case CEED_EVAL_GRAD: { 248 void *grad_args[] = {(void *)&num_elem, &data->d_grad, &d_u, &d_v}; 249 const int block_size_x = is_transpose ? num_nodes : num_qpts; 250 251 if (is_transpose) { 252 CeedCallBackend(CeedRunKernelDim_Cuda(ceed, data->DerivTranspose, grid, block_size_x, 1, elems_per_block, grad_args)); 253 } else { 254 CeedCallBackend(CeedRunKernelDim_Cuda(ceed, data->Deriv, grid, block_size_x, 1, elems_per_block, grad_args)); 255 } 256 } break; 257 case CEED_EVAL_DIV: { 258 void *div_args[] = {(void *)&num_elem, &data->d_div, &d_u, &d_v}; 259 const int block_size_x = is_transpose ? num_nodes : num_qpts; 260 261 if (is_transpose) { 262 CeedCallBackend(CeedRunKernelDim_Cuda(ceed, data->DerivTranspose, grid, block_size_x, 1, elems_per_block, div_args)); 263 } else { 264 CeedCallBackend(CeedRunKernelDim_Cuda(ceed, data->Deriv, grid, block_size_x, 1, elems_per_block, div_args)); 265 } 266 } break; 267 case CEED_EVAL_CURL: { 268 void *curl_args[] = {(void *)&num_elem, &data->d_curl, &d_u, &d_v}; 269 const int block_size_x = is_transpose ? num_nodes : num_qpts; 270 271 if (is_transpose) { 272 CeedCallBackend(CeedRunKernelDim_Cuda(ceed, data->DerivTranspose, grid, block_size_x, 1, elems_per_block, curl_args)); 273 } else { 274 CeedCallBackend(CeedRunKernelDim_Cuda(ceed, data->Deriv, grid, block_size_x, 1, elems_per_block, curl_args)); 275 } 276 } break; 277 case CEED_EVAL_WEIGHT: { 278 CeedCheck(data->d_q_weight, ceed, CEED_ERROR_BACKEND, "%s not supported; q_weights not set", CeedEvalModes[eval_mode]); 279 void *weight_args[] = {(void *)&num_elem, (void *)&data->d_q_weight, &d_v}; 280 281 CeedCallBackend(CeedRunKernelDim_Cuda(ceed, data->Weight, grid, num_qpts, 1, elems_per_block, weight_args)); 282 } break; 283 case CEED_EVAL_NONE: /* handled separately below */ 284 break; 285 } 286 287 // Restore vectors, cover CEED_EVAL_NONE 288 CeedCallBackend(CeedVectorRestoreArray(v, &d_v)); 289 if (eval_mode == CEED_EVAL_NONE) CeedCallBackend(CeedVectorSetArray(v, CEED_MEM_DEVICE, CEED_COPY_VALUES, (CeedScalar *)d_u)); 290 if (eval_mode != CEED_EVAL_WEIGHT) CeedCallBackend(CeedVectorRestoreArrayRead(u, &d_u)); 291 return CEED_ERROR_SUCCESS; 292 } 293 294 //------------------------------------------------------------------------------ 295 // Destroy tensor basis 296 //------------------------------------------------------------------------------ 297 static int CeedBasisDestroy_Cuda(CeedBasis basis) { 298 Ceed ceed; 299 CeedBasis_Cuda *data; 300 301 CeedCallBackend(CeedBasisGetCeed(basis, &ceed)); 302 CeedCallBackend(CeedBasisGetData(basis, &data)); 303 CeedCallCuda(ceed, cuModuleUnload(data->module)); 304 if (data->moduleAtPoints) CeedCallCuda(ceed, cuModuleUnload(data->moduleAtPoints)); 305 if (data->d_q_weight_1d) CeedCallCuda(ceed, cudaFree(data->d_q_weight_1d)); 306 CeedCallCuda(ceed, cudaFree(data->d_interp_1d)); 307 CeedCallCuda(ceed, cudaFree(data->d_grad_1d)); 308 CeedCallCuda(ceed, cudaFree(data->d_chebyshev_interp_1d)); 309 CeedCallBackend(CeedFree(&data)); 310 return CEED_ERROR_SUCCESS; 311 } 312 313 //------------------------------------------------------------------------------ 314 // Destroy non-tensor basis 315 //------------------------------------------------------------------------------ 316 static int CeedBasisDestroyNonTensor_Cuda(CeedBasis basis) { 317 Ceed ceed; 318 CeedBasisNonTensor_Cuda *data; 319 320 CeedCallBackend(CeedBasisGetCeed(basis, &ceed)); 321 CeedCallBackend(CeedBasisGetData(basis, &data)); 322 CeedCallCuda(ceed, cuModuleUnload(data->module)); 323 if (data->d_q_weight) CeedCallCuda(ceed, cudaFree(data->d_q_weight)); 324 CeedCallCuda(ceed, cudaFree(data->d_interp)); 325 CeedCallCuda(ceed, cudaFree(data->d_grad)); 326 CeedCallCuda(ceed, cudaFree(data->d_div)); 327 CeedCallCuda(ceed, cudaFree(data->d_curl)); 328 CeedCallBackend(CeedFree(&data)); 329 return CEED_ERROR_SUCCESS; 330 } 331 332 //------------------------------------------------------------------------------ 333 // Create tensor 334 //------------------------------------------------------------------------------ 335 int CeedBasisCreateTensorH1_Cuda(CeedInt dim, CeedInt P_1d, CeedInt Q_1d, const CeedScalar *interp_1d, const CeedScalar *grad_1d, 336 const CeedScalar *q_ref_1d, const CeedScalar *q_weight_1d, CeedBasis basis) { 337 Ceed ceed; 338 char *basis_kernel_source; 339 const char *basis_kernel_path; 340 CeedInt num_comp; 341 const CeedInt q_bytes = Q_1d * sizeof(CeedScalar); 342 const CeedInt interp_bytes = q_bytes * P_1d; 343 CeedBasis_Cuda *data; 344 345 CeedCallBackend(CeedBasisGetCeed(basis, &ceed)); 346 CeedCallBackend(CeedCalloc(1, &data)); 347 348 // Copy data to GPU 349 if (q_weight_1d) { 350 CeedCallCuda(ceed, cudaMalloc((void **)&data->d_q_weight_1d, q_bytes)); 351 CeedCallCuda(ceed, cudaMemcpy(data->d_q_weight_1d, q_weight_1d, q_bytes, cudaMemcpyHostToDevice)); 352 } 353 CeedCallCuda(ceed, cudaMalloc((void **)&data->d_interp_1d, interp_bytes)); 354 CeedCallCuda(ceed, cudaMemcpy(data->d_interp_1d, interp_1d, interp_bytes, cudaMemcpyHostToDevice)); 355 CeedCallCuda(ceed, cudaMalloc((void **)&data->d_grad_1d, interp_bytes)); 356 CeedCallCuda(ceed, cudaMemcpy(data->d_grad_1d, grad_1d, interp_bytes, cudaMemcpyHostToDevice)); 357 358 // Compile basis kernels 359 CeedCallBackend(CeedBasisGetNumComponents(basis, &num_comp)); 360 CeedCallBackend(CeedGetJitAbsolutePath(ceed, "ceed/jit-source/cuda/cuda-ref-basis-tensor.h", &basis_kernel_path)); 361 CeedDebug256(ceed, CEED_DEBUG_COLOR_SUCCESS, "----- Loading Basis Kernel Source -----\n"); 362 CeedCallBackend(CeedLoadSourceToBuffer(ceed, basis_kernel_path, &basis_kernel_source)); 363 CeedDebug256(ceed, CEED_DEBUG_COLOR_SUCCESS, "----- Loading Basis Kernel Source Complete! -----\n"); 364 CeedCallBackend(CeedCompile_Cuda(ceed, basis_kernel_source, &data->module, 7, "BASIS_Q_1D", Q_1d, "BASIS_P_1D", P_1d, "BASIS_BUF_LEN", 365 Q_1d * CeedIntPow(Q_1d > P_1d ? Q_1d : P_1d, dim - 1), "BASIS_DIM", dim, "BASIS_NUM_COMP", num_comp, 366 "BASIS_NUM_NODES", CeedIntPow(P_1d, dim), "BASIS_NUM_QPTS", CeedIntPow(Q_1d, dim))); 367 CeedCallBackend(CeedGetKernel_Cuda(ceed, data->module, "Interp", &data->Interp)); 368 CeedCallBackend(CeedGetKernel_Cuda(ceed, data->module, "Grad", &data->Grad)); 369 CeedCallBackend(CeedGetKernel_Cuda(ceed, data->module, "Weight", &data->Weight)); 370 CeedCallBackend(CeedFree(&basis_kernel_path)); 371 CeedCallBackend(CeedFree(&basis_kernel_source)); 372 373 CeedCallBackend(CeedBasisSetData(basis, data)); 374 375 // Register backend functions 376 CeedCallBackend(CeedSetBackendFunction(ceed, "Basis", basis, "Apply", CeedBasisApply_Cuda)); 377 CeedCallBackend(CeedSetBackendFunction(ceed, "Basis", basis, "ApplyAtPoints", CeedBasisApplyAtPoints_Cuda)); 378 CeedCallBackend(CeedSetBackendFunction(ceed, "Basis", basis, "Destroy", CeedBasisDestroy_Cuda)); 379 return CEED_ERROR_SUCCESS; 380 } 381 382 //------------------------------------------------------------------------------ 383 // Create non-tensor H^1 384 //------------------------------------------------------------------------------ 385 int CeedBasisCreateH1_Cuda(CeedElemTopology topo, CeedInt dim, CeedInt num_nodes, CeedInt num_qpts, const CeedScalar *interp, const CeedScalar *grad, 386 const CeedScalar *q_ref, const CeedScalar *q_weight, CeedBasis basis) { 387 Ceed ceed; 388 char *basis_kernel_source; 389 const char *basis_kernel_path; 390 CeedInt num_comp, q_comp_interp, q_comp_grad; 391 const CeedInt q_bytes = num_qpts * sizeof(CeedScalar); 392 CeedBasisNonTensor_Cuda *data; 393 394 CeedCallBackend(CeedBasisGetCeed(basis, &ceed)); 395 CeedCallBackend(CeedCalloc(1, &data)); 396 397 // Copy basis data to GPU 398 CeedCallBackend(CeedBasisGetNumQuadratureComponents(basis, CEED_EVAL_INTERP, &q_comp_interp)); 399 CeedCallBackend(CeedBasisGetNumQuadratureComponents(basis, CEED_EVAL_GRAD, &q_comp_grad)); 400 if (q_weight) { 401 CeedCallCuda(ceed, cudaMalloc((void **)&data->d_q_weight, q_bytes)); 402 CeedCallCuda(ceed, cudaMemcpy(data->d_q_weight, q_weight, q_bytes, cudaMemcpyHostToDevice)); 403 } 404 if (interp) { 405 const CeedInt interp_bytes = q_bytes * num_nodes * q_comp_interp; 406 407 CeedCallCuda(ceed, cudaMalloc((void **)&data->d_interp, interp_bytes)); 408 CeedCallCuda(ceed, cudaMemcpy(data->d_interp, interp, interp_bytes, cudaMemcpyHostToDevice)); 409 } 410 if (grad) { 411 const CeedInt grad_bytes = q_bytes * num_nodes * q_comp_grad; 412 413 CeedCallCuda(ceed, cudaMalloc((void **)&data->d_grad, grad_bytes)); 414 CeedCallCuda(ceed, cudaMemcpy(data->d_grad, grad, grad_bytes, cudaMemcpyHostToDevice)); 415 } 416 417 // Compile basis kernels 418 CeedCallBackend(CeedBasisGetNumComponents(basis, &num_comp)); 419 CeedCallBackend(CeedGetJitAbsolutePath(ceed, "ceed/jit-source/cuda/cuda-ref-basis-nontensor.h", &basis_kernel_path)); 420 CeedDebug256(ceed, CEED_DEBUG_COLOR_SUCCESS, "----- Loading Basis Kernel Source -----\n"); 421 CeedCallBackend(CeedLoadSourceToBuffer(ceed, basis_kernel_path, &basis_kernel_source)); 422 CeedDebug256(ceed, CEED_DEBUG_COLOR_SUCCESS, "----- Loading Basis Kernel Source Complete! -----\n"); 423 CeedCallBackend(CeedCompile_Cuda(ceed, basis_kernel_source, &data->module, 5, "BASIS_Q", num_qpts, "BASIS_P", num_nodes, "BASIS_Q_COMP_INTERP", 424 q_comp_interp, "BASIS_Q_COMP_DERIV", q_comp_grad, "BASIS_NUM_COMP", num_comp)); 425 CeedCallBackend(CeedGetKernel_Cuda(ceed, data->module, "Interp", &data->Interp)); 426 CeedCallBackend(CeedGetKernel_Cuda(ceed, data->module, "InterpTranspose", &data->InterpTranspose)); 427 CeedCallBackend(CeedGetKernel_Cuda(ceed, data->module, "Deriv", &data->Deriv)); 428 CeedCallBackend(CeedGetKernel_Cuda(ceed, data->module, "DerivTranspose", &data->DerivTranspose)); 429 CeedCallBackend(CeedGetKernel_Cuda(ceed, data->module, "Weight", &data->Weight)); 430 CeedCallBackend(CeedFree(&basis_kernel_path)); 431 CeedCallBackend(CeedFree(&basis_kernel_source)); 432 433 CeedCallBackend(CeedBasisSetData(basis, data)); 434 435 // Register backend functions 436 CeedCallBackend(CeedSetBackendFunction(ceed, "Basis", basis, "Apply", CeedBasisApplyNonTensor_Cuda)); 437 CeedCallBackend(CeedSetBackendFunction(ceed, "Basis", basis, "Destroy", CeedBasisDestroyNonTensor_Cuda)); 438 return CEED_ERROR_SUCCESS; 439 } 440 441 //------------------------------------------------------------------------------ 442 // Create non-tensor H(div) 443 //------------------------------------------------------------------------------ 444 int CeedBasisCreateHdiv_Cuda(CeedElemTopology topo, CeedInt dim, CeedInt num_nodes, CeedInt num_qpts, const CeedScalar *interp, const CeedScalar *div, 445 const CeedScalar *q_ref, const CeedScalar *q_weight, CeedBasis basis) { 446 Ceed ceed; 447 char *basis_kernel_source; 448 const char *basis_kernel_path; 449 CeedInt num_comp, q_comp_interp, q_comp_div; 450 const CeedInt q_bytes = num_qpts * sizeof(CeedScalar); 451 CeedBasisNonTensor_Cuda *data; 452 453 CeedCallBackend(CeedBasisGetCeed(basis, &ceed)); 454 CeedCallBackend(CeedCalloc(1, &data)); 455 456 // Copy basis data to GPU 457 CeedCallBackend(CeedBasisGetNumQuadratureComponents(basis, CEED_EVAL_INTERP, &q_comp_interp)); 458 CeedCallBackend(CeedBasisGetNumQuadratureComponents(basis, CEED_EVAL_DIV, &q_comp_div)); 459 if (q_weight) { 460 CeedCallCuda(ceed, cudaMalloc((void **)&data->d_q_weight, q_bytes)); 461 CeedCallCuda(ceed, cudaMemcpy(data->d_q_weight, q_weight, q_bytes, cudaMemcpyHostToDevice)); 462 } 463 if (interp) { 464 const CeedInt interp_bytes = q_bytes * num_nodes * q_comp_interp; 465 466 CeedCallCuda(ceed, cudaMalloc((void **)&data->d_interp, interp_bytes)); 467 CeedCallCuda(ceed, cudaMemcpy(data->d_interp, interp, interp_bytes, cudaMemcpyHostToDevice)); 468 } 469 if (div) { 470 const CeedInt div_bytes = q_bytes * num_nodes * q_comp_div; 471 472 CeedCallCuda(ceed, cudaMalloc((void **)&data->d_div, div_bytes)); 473 CeedCallCuda(ceed, cudaMemcpy(data->d_div, div, div_bytes, cudaMemcpyHostToDevice)); 474 } 475 476 // Compile basis kernels 477 CeedCallBackend(CeedBasisGetNumComponents(basis, &num_comp)); 478 CeedCallBackend(CeedGetJitAbsolutePath(ceed, "ceed/jit-source/cuda/cuda-ref-basis-nontensor.h", &basis_kernel_path)); 479 CeedDebug256(ceed, CEED_DEBUG_COLOR_SUCCESS, "----- Loading Basis Kernel Source -----\n"); 480 CeedCallBackend(CeedLoadSourceToBuffer(ceed, basis_kernel_path, &basis_kernel_source)); 481 CeedDebug256(ceed, CEED_DEBUG_COLOR_SUCCESS, "----- Loading Basis Kernel Source Complete! -----\n"); 482 CeedCallBackend(CeedCompile_Cuda(ceed, basis_kernel_source, &data->module, 5, "BASIS_Q", num_qpts, "BASIS_P", num_nodes, "BASIS_Q_COMP_INTERP", 483 q_comp_interp, "BASIS_Q_COMP_DERIV", q_comp_div, "BASIS_NUM_COMP", num_comp)); 484 CeedCallBackend(CeedGetKernel_Cuda(ceed, data->module, "Interp", &data->Interp)); 485 CeedCallBackend(CeedGetKernel_Cuda(ceed, data->module, "InterpTranspose", &data->InterpTranspose)); 486 CeedCallBackend(CeedGetKernel_Cuda(ceed, data->module, "Deriv", &data->Deriv)); 487 CeedCallBackend(CeedGetKernel_Cuda(ceed, data->module, "DerivTranspose", &data->DerivTranspose)); 488 CeedCallBackend(CeedGetKernel_Cuda(ceed, data->module, "Weight", &data->Weight)); 489 CeedCallBackend(CeedFree(&basis_kernel_path)); 490 CeedCallBackend(CeedFree(&basis_kernel_source)); 491 492 CeedCallBackend(CeedBasisSetData(basis, data)); 493 494 // Register backend functions 495 CeedCallBackend(CeedSetBackendFunction(ceed, "Basis", basis, "Apply", CeedBasisApplyNonTensor_Cuda)); 496 CeedCallBackend(CeedSetBackendFunction(ceed, "Basis", basis, "Destroy", CeedBasisDestroyNonTensor_Cuda)); 497 return CEED_ERROR_SUCCESS; 498 } 499 500 //------------------------------------------------------------------------------ 501 // Create non-tensor H(curl) 502 //------------------------------------------------------------------------------ 503 int CeedBasisCreateHcurl_Cuda(CeedElemTopology topo, CeedInt dim, CeedInt num_nodes, CeedInt num_qpts, const CeedScalar *interp, 504 const CeedScalar *curl, const CeedScalar *q_ref, const CeedScalar *q_weight, CeedBasis basis) { 505 Ceed ceed; 506 char *basis_kernel_source; 507 const char *basis_kernel_path; 508 CeedInt num_comp, q_comp_interp, q_comp_curl; 509 const CeedInt q_bytes = num_qpts * sizeof(CeedScalar); 510 CeedBasisNonTensor_Cuda *data; 511 512 CeedCallBackend(CeedBasisGetCeed(basis, &ceed)); 513 CeedCallBackend(CeedCalloc(1, &data)); 514 515 // Copy basis data to GPU 516 CeedCallBackend(CeedBasisGetNumQuadratureComponents(basis, CEED_EVAL_INTERP, &q_comp_interp)); 517 CeedCallBackend(CeedBasisGetNumQuadratureComponents(basis, CEED_EVAL_CURL, &q_comp_curl)); 518 if (q_weight) { 519 CeedCallCuda(ceed, cudaMalloc((void **)&data->d_q_weight, q_bytes)); 520 CeedCallCuda(ceed, cudaMemcpy(data->d_q_weight, q_weight, q_bytes, cudaMemcpyHostToDevice)); 521 } 522 if (interp) { 523 const CeedInt interp_bytes = q_bytes * num_nodes * q_comp_interp; 524 525 CeedCallCuda(ceed, cudaMalloc((void **)&data->d_interp, interp_bytes)); 526 CeedCallCuda(ceed, cudaMemcpy(data->d_interp, interp, interp_bytes, cudaMemcpyHostToDevice)); 527 } 528 if (curl) { 529 const CeedInt curl_bytes = q_bytes * num_nodes * q_comp_curl; 530 531 CeedCallCuda(ceed, cudaMalloc((void **)&data->d_curl, curl_bytes)); 532 CeedCallCuda(ceed, cudaMemcpy(data->d_curl, curl, curl_bytes, cudaMemcpyHostToDevice)); 533 } 534 535 // Compile basis kernels 536 CeedCallBackend(CeedBasisGetNumComponents(basis, &num_comp)); 537 CeedCallBackend(CeedGetJitAbsolutePath(ceed, "ceed/jit-source/cuda/cuda-ref-basis-nontensor.h", &basis_kernel_path)); 538 CeedDebug256(ceed, CEED_DEBUG_COLOR_SUCCESS, "----- Loading Basis Kernel Source -----\n"); 539 CeedCallBackend(CeedLoadSourceToBuffer(ceed, basis_kernel_path, &basis_kernel_source)); 540 CeedDebug256(ceed, CEED_DEBUG_COLOR_SUCCESS, "----- Loading Basis Kernel Source Complete! -----\n"); 541 CeedCallBackend(CeedCompile_Cuda(ceed, basis_kernel_source, &data->module, 5, "BASIS_Q", num_qpts, "BASIS_P", num_nodes, "BASIS_Q_COMP_INTERP", 542 q_comp_interp, "BASIS_Q_COMP_DERIV", q_comp_curl, "BASIS_NUM_COMP", num_comp)); 543 CeedCallBackend(CeedGetKernel_Cuda(ceed, data->module, "Interp", &data->Interp)); 544 CeedCallBackend(CeedGetKernel_Cuda(ceed, data->module, "InterpTranspose", &data->InterpTranspose)); 545 CeedCallBackend(CeedGetKernel_Cuda(ceed, data->module, "Deriv", &data->Deriv)); 546 CeedCallBackend(CeedGetKernel_Cuda(ceed, data->module, "DerivTranspose", &data->DerivTranspose)); 547 CeedCallBackend(CeedGetKernel_Cuda(ceed, data->module, "Weight", &data->Weight)); 548 CeedCallBackend(CeedFree(&basis_kernel_path)); 549 CeedCallBackend(CeedFree(&basis_kernel_source)); 550 551 CeedCallBackend(CeedBasisSetData(basis, data)); 552 553 // Register backend functions 554 CeedCallBackend(CeedSetBackendFunction(ceed, "Basis", basis, "Apply", CeedBasisApplyNonTensor_Cuda)); 555 CeedCallBackend(CeedSetBackendFunction(ceed, "Basis", basis, "Destroy", CeedBasisDestroyNonTensor_Cuda)); 556 return CEED_ERROR_SUCCESS; 557 } 558 559 //------------------------------------------------------------------------------ 560