| 1 |
// ------------------------------------------------------------------------ |
| 2 |
// |
| 3 |
// SPDX-License-Identifier: LGPL-2.1-or-later |
| 4 |
// Copyright (C) 2017 - 2025 by the deal.II authors |
| 5 |
// |
| 6 |
// This file is part of the deal.II library. |
| 7 |
// |
| 8 |
// Part of the source code is dual licensed under Apache-2.0 WITH |
| 9 |
// LLVM-exception OR LGPL-2.1-or-later. Detailed license information |
| 10 |
// governing the source code and code contributions can be found in |
| 11 |
// LICENSE.md and CONTRIBUTING.md at the top level directory of deal.II. |
| 12 |
// |
| 13 |
// ------------------------------------------------------------------------ |
| 14 |
|
| 15 |
|
| 16 |
#ifndef dealii__tensor_product_kernels_h |
| 17 |
#define dealii__tensor_product_kernels_h |
| 18 |
|
| 19 |
#include <deal.II/base/config.h> |
| 20 |
|
| 21 |
#include <deal.II/base/memory_space.h> |
| 22 |
#include <deal.II/base/utilities.h> |
| 23 |
|
| 24 |
#include <Kokkos_Core.hpp> |
| 25 |
|
| 26 |
|
| 27 |
DEAL_II_NAMESPACE_OPEN |
| 28 |
|
| 29 |
|
| 30 |
namespace Portable |
| 31 |
{ |
| 32 |
namespace internal |
| 33 |
{ |
| 34 |
/** |
| 35 |
* In this namespace, the evaluator routines that evaluate the tensor |
| 36 |
* products are implemented. |
| 37 |
*/ |
| 38 |
// TODO: for now only the general variant is implemented |
| 39 |
enum EvaluatorVariant |
| 40 |
{ |
| 41 |
evaluate_general, |
| 42 |
evaluate_symmetric, |
| 43 |
evaluate_evenodd |
| 44 |
}; |
| 45 |
|
| 46 |
|
| 47 |
|
| 48 |
/** |
| 49 |
* Helper function that copies or adds the first N entries of src to |
| 50 |
* dst, depending on the template argument "add". |
| 51 |
*/ |
| 52 |
template <bool add, typename ViewTypeIn, typename ViewTypeOut> |
| 53 |
DEAL_II_HOST_DEVICE void |
| 54 |
populate_view( |
| 55 |
const Kokkos::TeamPolicy< |
| 56 |
MemorySpace::Default::kokkos_space::execution_space>::member_type |
| 57 |
&team_member, |
| 58 |
ViewTypeOut dst, |
| 59 |
const ViewTypeIn src, |
| 60 |
const int N) |
| 61 |
{ |
| 62 |
Assert(dst.size() >= static_cast<unsigned int>(N), ExcInternalError()); |
| 63 |
Assert(src.size() >= static_cast<unsigned int>(N), ExcInternalError()); |
| 64 |
Kokkos::parallel_for(Kokkos::TeamVectorRange(team_member, N), |
| 65 |
[&](const int i) { |
| 66 |
if constexpr (add) |
| 67 |
Kokkos::atomic_add(&dst(i), src(i)); |
| 68 |
else |
| 69 |
dst(i) = src(i); |
| 70 |
}); |
| 71 |
|
| 72 |
team_member.team_barrier(); |
| 73 |
} |
| 74 |
|
| 75 |
|
| 76 |
|
| 77 |
#if DEAL_II_KOKKOS_VERSION_GTE(4, 0, 0) |
| 78 |
/** |
| 79 |
* Helper function for apply() in 1D |
| 80 |
*/ |
| 81 |
template <int n_rows, |
| 82 |
int n_columns, |
| 83 |
int direction, |
| 84 |
typename Number, |
| 85 |
bool contract_over_rows, |
| 86 |
bool add, |
| 87 |
typename ViewTypeIn, |
| 88 |
typename ViewTypeOut> |
| 89 |
DEAL_II_HOST_DEVICE void |
| 90 |
apply_1d(const Kokkos::TeamPolicy< |
| 91 |
MemorySpace::Default::kokkos_space::execution_space>::member_type |
| 92 |
&team_member, |
| 93 |
const Kokkos::View<Number *, MemorySpace::Default::kokkos_space> |
| 94 |
shape_data, |
| 95 |
const ViewTypeIn in, |
| 96 |
ViewTypeOut out) |
| 97 |
{ |
| 98 |
constexpr int Nk = (contract_over_rows ? n_rows : n_columns), |
| 99 |
Nq = (contract_over_rows ? n_columns : n_rows); |
| 100 |
|
| 101 |
Assert(shape_data.size() == n_rows * n_columns, ExcInternalError()); |
| 102 |
Assert(in.size() >= Nk, ExcInternalError()); |
| 103 |
Assert(out.size() >= Nq, ExcInternalError()); |
| 104 |
|
| 105 |
Kokkos::parallel_for(Kokkos::TeamThreadRange(team_member, Nq), |
| 106 |
[&](const int q) { |
| 107 |
Number sum = 0; |
| 108 |
for (int k = 0; k < Nk; ++k) |
| 109 |
{ |
| 110 |
const int shape_idx = |
| 111 |
(contract_over_rows ? q + k * Nq : |
| 112 |
k + q * Nk); |
| 113 |
sum += shape_data(shape_idx) * in(k); |
| 114 |
} |
| 115 |
|
| 116 |
if constexpr (add) |
| 117 |
Kokkos::atomic_add(&out(q), sum); |
| 118 |
else |
| 119 |
out(q) = sum; |
| 120 |
}); |
| 121 |
|
| 122 |
team_member.team_barrier(); |
| 123 |
} |
| 124 |
|
| 125 |
|
| 126 |
|
| 127 |
/** |
| 128 |
* Helper function for apply() in 2D |
| 129 |
*/ |
| 130 |
template <int n_rows, |
| 131 |
int n_columns, |
| 132 |
int direction, |
| 133 |
typename Number, |
| 134 |
bool contract_over_rows, |
| 135 |
bool add, |
| 136 |
typename ViewTypeIn, |
| 137 |
typename ViewTypeOut> |
| 138 |
DEAL_II_HOST_DEVICE void |
| 139 |
apply_2d(const Kokkos::TeamPolicy< |
| 140 |
MemorySpace::Default::kokkos_space::execution_space>::member_type |
| 141 |
&team_member, |
| 142 |
const Kokkos::View<Number *, MemorySpace::Default::kokkos_space> |
| 143 |
shape_data, |
| 144 |
const ViewTypeIn in, |
| 145 |
ViewTypeOut out) |
| 146 |
{ |
| 147 |
using TeamType = Kokkos::TeamPolicy< |
| 148 |
MemorySpace::Default::kokkos_space::execution_space>::member_type; |
| 149 |
|
| 150 |
// Sizes of the input and output vectors: |
| 151 |
// ----------------------------------------------------------- |
| 152 |
// direction | contract_over_rows | !contract_over_rows |
| 153 |
// ----------------------------------------------------------- |
| 154 |
// 0 | m x m -> n x m | n x m -> m x m |
| 155 |
// ----------------------------------------------------------- |
| 156 |
// 1 | n x m -> n x n | n x n -> n x m |
| 157 |
// ----------------------------------------------------------- |
| 158 |
// |
| 159 |
// Directions of the cycle indices: |
| 160 |
// ----------------------------- |
| 161 |
// direction | j | q/k |
| 162 |
// ----------------------------- |
| 163 |
// 0 | 1 | 0 |
| 164 |
// ----------------------------- |
| 165 |
// 1 | 0 | 1 |
| 166 |
// ----------------------------- |
| 167 |
constexpr int Nj = (direction < 1 ? n_rows : n_columns), |
| 168 |
Nk = (contract_over_rows ? n_rows : n_columns), |
| 169 |
Nq = (contract_over_rows ? n_columns : n_rows); |
| 170 |
|
| 171 |
Assert(shape_data.size() == n_rows * n_columns, ExcInternalError()); |
| 172 |
Assert(in.size() >= Nj * Nk, ExcInternalError()); |
| 173 |
Assert(out.size() >= Nj * Nq, ExcInternalError()); |
| 174 |
|
| 175 |
auto thread_policy = |
| 176 |
Kokkos::TeamThreadMDRange<Kokkos::Rank<2>, TeamType>(team_member, |
| 177 |
Nj, |
| 178 |
Nq); |
| 179 |
Kokkos::parallel_for(thread_policy, [&](const int j, const int q) { |
| 180 |
const int base_shape = contract_over_rows ? q : q * n_columns; |
| 181 |
const int stride_shape = contract_over_rows ? n_columns : 1; |
| 182 |
|
| 183 |
const int base_in = (direction == 0 ? j * Nk : j); |
| 184 |
const int stride_in = Utilities::pow(n_columns, direction); |
| 185 |
|
| 186 |
Number sum = shape_data(base_shape) * in(base_in); |
| 187 |
for (int k = 1; k < Nk; ++k) |
| 188 |
sum += shape_data(base_shape + k * stride_shape) * |
| 189 |
in(base_in + k * stride_in); |
| 190 |
|
| 191 |
const int index_out = (direction == 0 ? j * Nq + q : j + q * Nj); |
| 192 |
|
| 193 |
if constexpr (add) |
| 194 |
Kokkos::atomic_add(&out(index_out), sum); |
| 195 |
else |
| 196 |
out(index_out) = sum; |
| 197 |
}); |
| 198 |
|
| 199 |
team_member.team_barrier(); |
| 200 |
} |
| 201 |
|
| 202 |
|
| 203 |
|
| 204 |
/** |
| 205 |
* Helper function for apply() in 3D |
| 206 |
*/ |
| 207 |
template <int n_rows, |
| 208 |
int n_columns, |
| 209 |
int direction, |
| 210 |
typename Number, |
| 211 |
bool contract_over_rows, |
| 212 |
bool add, |
| 213 |
typename ViewTypeIn, |
| 214 |
typename ViewTypeOut> |
| 215 |
DEAL_II_HOST_DEVICE void |
| 216 |
apply_3d(const Kokkos::TeamPolicy< |
| 217 |
MemorySpace::Default::kokkos_space::execution_space>::member_type |
| 218 |
&team_member, |
| 219 |
const Kokkos::View<Number *, MemorySpace::Default::kokkos_space> |
| 220 |
shape_data, |
| 221 |
const ViewTypeIn in, |
| 222 |
ViewTypeOut out) |
| 223 |
{ |
| 224 |
using TeamType = Kokkos::TeamPolicy< |
| 225 |
MemorySpace::Default::kokkos_space::execution_space>::member_type; |
| 226 |
|
| 227 |
// Sizes of the input and output vectors: |
| 228 |
// ------------------------------------------------------------------ |
| 229 |
// direction | contract_over_rows | !contract_over_rows |
| 230 |
// ------------------------------------------------------------------ |
| 231 |
// 0 | m x m x m -> n x m x m | n x m x m -> m x m x m |
| 232 |
// ------------------------------------------------------------------ |
| 233 |
// 1 | n x m x m -> n x n x m | n x n x m -> n x m x m |
| 234 |
// ------------------------------------------------------------------ |
| 235 |
// 2 | n x n x m -> n x n x n | n x n x n -> n x n x m |
| 236 |
// ------------------------------------------------------------------ |
| 237 |
// |
| 238 |
// Directions of the cycle indices: |
| 239 |
// ------------------------------------- |
| 240 |
// direction | i | j | q/k |
| 241 |
// ------------------------------------- |
| 242 |
// 0 | 2 | 1 | 0 |
| 243 |
// ------------------------------------- |
| 244 |
// 1 | 0 | 2 | 1 |
| 245 |
// ------------------------------------- |
| 246 |
// 2 | 1 | 0 | 2 |
| 247 |
// ------------------------------------- |
| 248 |
constexpr int Ni = (direction < 1 ? n_rows : n_columns), |
| 249 |
Nj = (direction < 2 ? n_rows : n_columns), |
| 250 |
Nk = (contract_over_rows ? n_rows : n_columns), |
| 251 |
Nq = (contract_over_rows ? n_columns : n_rows); |
| 252 |
|
| 253 |
Assert(shape_data.size() == n_rows * n_columns, ExcInternalError()); |
| 254 |
Assert(in.size() >= Ni * Nj * Nk, ExcInternalError()); |
| 255 |
Assert(out.size() >= Ni * Nj * Nq, ExcInternalError()); |
| 256 |
|
| 257 |
auto thread_policy = Kokkos::TeamThreadMDRange<Kokkos::Rank<3>, TeamType>( |
| 258 |
team_member, Ni, Nj, Nq); |
| 259 |
Kokkos::parallel_for( |
| 260 |
thread_policy, [&](const int i, const int j, const int q) { |
| 261 |
const int base_shape = contract_over_rows ? q : q * n_columns; |
| 262 |
const int stride_shape = contract_over_rows ? n_columns : 1; |
| 263 |
|
| 264 |
const int base_in = |
| 265 |
(direction == 0 ? (i * Nj + j) * Nk : |
| 266 |
(direction == 1 ? i + j * Ni * Nk : i * Nj + j)); |
| 267 |
const int stride_in = Utilities::pow(n_columns, direction); |
| 268 |
|
| 269 |
Number sum = shape_data(base_shape) * in(base_in); |
| 270 |
for (int k = 1; k < Nk; ++k) |
| 271 |
sum += shape_data(base_shape + k * stride_shape) * |
| 272 |
in(base_in + k * stride_in); |
| 273 |
|
| 274 |
const int index_out = |
| 275 |
(direction == 0 ? (i * Nj + j) * Nq + q : |
| 276 |
(direction == 1 ? i + (j * Nq + q) * Ni : |
| 277 |
(i + q * Ni) * Nj + j)); |
| 278 |
|
| 279 |
if constexpr (add) |
| 280 |
Kokkos::atomic_add(&out(index_out), sum); |
| 281 |
else |
| 282 |
out(index_out) = sum; |
| 283 |
}); |
| 284 |
|
| 285 |
team_member.team_barrier(); |
| 286 |
} |
| 287 |
#endif |
| 288 |
|
| 289 |
|
| 290 |
|
| 291 |
template <int dim, |
| 292 |
int n_rows, |
| 293 |
int n_columns, |
| 294 |
typename Number, |
| 295 |
int direction, |
| 296 |
bool contract_over_rows, |
| 297 |
bool add, |
| 298 |
typename ViewTypeIn, |
| 299 |
typename ViewTypeOut> |
| 300 |
DEAL_II_HOST_DEVICE void |
| 301 |
apply(const Kokkos::TeamPolicy< |
| 302 |
MemorySpace::Default::kokkos_space::execution_space>::member_type |
| 303 |
&team_member, |
| 304 |
const Kokkos::View<Number *, MemorySpace::Default::kokkos_space> |
| 305 |
shape_data, |
| 306 |
const ViewTypeIn in, |
| 307 |
ViewTypeOut out) |
| 308 |
{ |
| 309 |
#if DEAL_II_KOKKOS_VERSION_GTE(4, 0, 0) |
| 310 |
if constexpr (dim == 1) |
| 311 |
apply_1d<n_rows, n_columns, direction, Number, contract_over_rows, add>( |
| 312 |
team_member, shape_data, in, out); |
| 313 |
if constexpr (dim == 2) |
| 314 |
apply_2d<n_rows, n_columns, direction, Number, contract_over_rows, add>( |
| 315 |
team_member, shape_data, in, out); |
| 316 |
if constexpr (dim == 3) |
| 317 |
apply_3d<n_rows, n_columns, direction, Number, contract_over_rows, add>( |
| 318 |
team_member, shape_data, in, out); |
| 319 |
#else |
| 320 |
// I: [0, m^{dim - direction - 1}) |
| 321 |
// J: [0, n^direction) |
| 322 |
constexpr int NI = Utilities::pow(n_rows, dim - direction - 1); |
| 323 |
constexpr int NJ = Utilities::pow(n_columns, direction); |
| 324 |
|
| 325 |
constexpr int Nk = contract_over_rows ? n_rows : n_columns; |
| 326 |
constexpr int Nq = contract_over_rows ? n_columns : n_rows; |
| 327 |
|
| 328 |
Assert(shape_data.size() == n_rows * n_columns, ExcInternalError()); |
| 329 |
Assert(in.size() >= NI * NJ * Nk, ExcInternalError()); |
| 330 |
Assert(out.size() >= NI * NJ * Nq, ExcInternalError()); |
| 331 |
|
| 332 |
constexpr int N = NI * NJ * Nq; |
| 333 |
constexpr int stride = Utilities::pow(n_columns, direction); |
| 334 |
|
| 335 |
Kokkos::parallel_for( |
| 336 |
Kokkos::TeamThreadRange(team_member, N), [&](const int index_out) { |
| 337 |
// index_in = (I Nk + k) n^direction + J |
| 338 |
// index_out = (I Nq + q) n^direction + J |
| 339 |
const int q = (index_out / stride) % Nq; |
| 340 |
const int I = (index_out / stride) / Nq; |
| 341 |
const int J = index_out % stride; |
| 342 |
|
| 343 |
const int base_shape = contract_over_rows ? q : q * n_columns; |
| 344 |
const int stride_shape = contract_over_rows ? n_columns : 1; |
| 345 |
const int base_in = I * Nk * stride + J; |
| 346 |
|
| 347 |
Number sum = shape_data(base_shape) * in(base_in); |
| 348 |
for (int k = 1; k < Nk; ++k) |
| 349 |
{ |
| 350 |
const int index_in = (I * Nk + k) * stride + J; |
| 351 |
sum += shape_data(base_shape + k * stride_shape) * in(index_in); |
| 352 |
} |
| 353 |
|
| 354 |
if constexpr (add) |
| 355 |
Kokkos::atomic_add(&out(index_out), sum); |
| 356 |
else |
| 357 |
out(index_out) = sum; |
| 358 |
}); |
| 359 |
|
| 360 |
team_member.team_barrier(); |
| 361 |
#endif |
| 362 |
} |
| 363 |
|
| 364 |
|
| 365 |
|
| 366 |
/** |
| 367 |
* Generic evaluator framework. |
| 368 |
*/ |
| 369 |
template <EvaluatorVariant variant, |
| 370 |
int dim, |
| 371 |
int n_rows, |
| 372 |
int n_columns, |
| 373 |
typename Number> |
| 374 |
struct EvaluatorTensorProduct |
| 375 |
{}; |
| 376 |
|
| 377 |
|
| 378 |
|
| 379 |
/** |
| 380 |
* Internal evaluator for 1d-3d shape function using the tensor product form |
| 381 |
* of the basis functions. |
| 382 |
*/ |
| 383 |
template <int dim, int n_rows, int n_columns, typename Number> |
| 384 |
struct EvaluatorTensorProduct<evaluate_general, |
| 385 |
dim, |
| 386 |
n_rows, |
| 387 |
n_columns, |
| 388 |
Number> |
| 389 |
{ |
| 390 |
public: |
| 391 |
using TeamHandle = Kokkos::TeamPolicy< |
| 392 |
MemorySpace::Default::kokkos_space::execution_space>::member_type; |
| 393 |
|
| 394 |
using SharedView = Kokkos::View<Number *, |
| 395 |
MemorySpace::Default::kokkos_space:: |
| 396 |
execution_space::scratch_memory_space, |
| 397 |
Kokkos::MemoryTraits<Kokkos::Unmanaged>>; |
| 398 |
|
| 399 |
DEAL_II_HOST_DEVICE |
| 400 |
EvaluatorTensorProduct( |
| 401 |
const TeamHandle &team_member, |
| 402 |
Kokkos::View<Number *, MemorySpace::Default::kokkos_space> shape_values, |
| 403 |
Kokkos::View<Number *, MemorySpace::Default::kokkos_space> |
| 404 |
shape_gradients, |
| 405 |
Kokkos::View<Number *, MemorySpace::Default::kokkos_space> |
| 406 |
co_shape_gradients, |
| 407 |
SharedView temp); |
| 408 |
|
| 409 |
/** |
| 410 |
* Evaluate/integrate the values of a finite element function at the |
| 411 |
* quadrature points for a given @p direction. |
| 412 |
*/ |
| 413 |
template <int direction, |
| 414 |
bool dof_to_quad, |
| 415 |
bool add, |
| 416 |
bool in_place, |
| 417 |
typename ViewTypeIn, |
| 418 |
typename ViewTypeOut> |
| 419 |
DEAL_II_HOST_DEVICE void |
| 420 |
values(const ViewTypeIn in, ViewTypeOut out) const; |
| 421 |
|
| 422 |
/** |
| 423 |
* Evaluate/integrate the gradient of a finite element function at the |
| 424 |
* quadrature points for a given @p direction. |
| 425 |
*/ |
| 426 |
template <int direction, |
| 427 |
bool dof_to_quad, |
| 428 |
bool add, |
| 429 |
bool in_place, |
| 430 |
typename ViewTypeIn, |
| 431 |
typename ViewTypeOut> |
| 432 |
DEAL_II_HOST_DEVICE void |
| 433 |
gradients(const ViewTypeIn in, ViewTypeOut out) const; |
| 434 |
|
| 435 |
/** |
| 436 |
* Evaluate the gradient of a finite element function at the quadrature |
| 437 |
* points for a given @p direction for collocation methods. |
| 438 |
*/ |
| 439 |
template <int direction, |
| 440 |
bool dof_to_quad, |
| 441 |
bool add, |
| 442 |
bool in_place, |
| 443 |
typename ViewTypeIn, |
| 444 |
typename ViewTypeOut> |
| 445 |
DEAL_II_HOST_DEVICE void |
| 446 |
co_gradients(const ViewTypeIn in, ViewTypeOut out) const; |
| 447 |
|
| 448 |
/** |
| 449 |
* TeamPolicy handle. |
| 450 |
*/ |
| 451 |
const TeamHandle &team_member; |
| 452 |
|
| 453 |
/** |
| 454 |
* Values of the shape functions. |
| 455 |
*/ |
| 456 |
Kokkos::View<Number *, MemorySpace::Default::kokkos_space> shape_values; |
| 457 |
|
| 458 |
/** |
| 459 |
* Values of the shape function gradients. |
| 460 |
*/ |
| 461 |
Kokkos::View<Number *, MemorySpace::Default::kokkos_space> |
| 462 |
shape_gradients; |
| 463 |
|
| 464 |
/** |
| 465 |
* Values of the shape function gradients for collocation methods. |
| 466 |
*/ |
| 467 |
Kokkos::View<Number *, MemorySpace::Default::kokkos_space> |
| 468 |
co_shape_gradients; |
| 469 |
|
| 470 |
/** |
| 471 |
* Temporary storage for in-place evaluations. |
| 472 |
*/ |
| 473 |
SharedView temp; |
| 474 |
}; |
| 475 |
|
| 476 |
|
| 477 |
|
| 478 |
template <int dim, int n_rows, int n_columns, typename Number> |
| 479 |
DEAL_II_HOST_DEVICE |
| 480 |
EvaluatorTensorProduct<evaluate_general, dim, n_rows, n_columns, Number>:: |
| 481 |
EvaluatorTensorProduct( |
| 482 |
const TeamHandle &team_member, |
| 483 |
Kokkos::View<Number *, MemorySpace::Default::kokkos_space> shape_values, |
| 484 |
Kokkos::View<Number *, MemorySpace::Default::kokkos_space> |
| 485 |
shape_gradients, |
| 486 |
Kokkos::View<Number *, MemorySpace::Default::kokkos_space> |
| 487 |
co_shape_gradients, |
| 488 |
SharedView temp) |
| 489 |
: team_member(team_member) |
| 490 |
, shape_values(shape_values) |
| 491 |
, shape_gradients(shape_gradients) |
| 492 |
, co_shape_gradients(co_shape_gradients) |
| 493 |
, temp(temp) |
| 494 |
{} |
| 495 |
|
| 496 |
|
| 497 |
|
| 498 |
template <int dim, int n_rows, int n_columns, typename Number> |
| 499 |
template <int direction, |
| 500 |
bool dof_to_quad, |
| 501 |
bool add, |
| 502 |
bool in_place, |
| 503 |
typename ViewTypeIn, |
| 504 |
typename ViewTypeOut> |
| 505 |
DEAL_II_HOST_DEVICE void |
| 506 |
EvaluatorTensorProduct<evaluate_general, dim, n_rows, n_columns, Number>:: |
| 507 |
values(const ViewTypeIn in, ViewTypeOut out) const |
| 508 |
{ |
| 509 |
if constexpr (in_place) |
| 510 |
{ |
| 511 |
apply<dim, n_rows, n_columns, Number, direction, dof_to_quad, false>( |
| 512 |
team_member, shape_values, in, temp); |
| 513 |
|
| 514 |
populate_view<add>(team_member, out, temp, out.extent(0)); |
| 515 |
} |
| 516 |
else |
| 517 |
apply<dim, n_rows, n_columns, Number, direction, dof_to_quad, add>( |
| 518 |
team_member, shape_values, in, out); |
| 519 |
} |
| 520 |
|
| 521 |
|
| 522 |
|
| 523 |
template <int dim, int n_rows, int n_columns, typename Number> |
| 524 |
template <int direction, |
| 525 |
bool dof_to_quad, |
| 526 |
bool add, |
| 527 |
bool in_place, |
| 528 |
typename ViewTypeIn, |
| 529 |
typename ViewTypeOut> |
| 530 |
DEAL_II_HOST_DEVICE void |
| 531 |
EvaluatorTensorProduct<evaluate_general, dim, n_rows, n_columns, Number>:: |
| 532 |
gradients(const ViewTypeIn in, ViewTypeOut out) const |
| 533 |
{ |
| 534 |
if constexpr (in_place) |
| 535 |
{ |
| 536 |
apply<dim, n_rows, n_columns, Number, direction, dof_to_quad, false>( |
| 537 |
team_member, shape_gradients, in, temp); |
| 538 |
|
| 539 |
populate_view<add>(team_member, out, temp, out.extent(0)); |
| 540 |
} |
| 541 |
else |
| 542 |
apply<dim, n_rows, n_columns, Number, direction, dof_to_quad, add>( |
| 543 |
team_member, shape_gradients, in, out); |
| 544 |
} |
| 545 |
|
| 546 |
|
| 547 |
|
| 548 |
template <int dim, int n_rows, int n_columns, typename Number> |
| 549 |
template <int direction, |
| 550 |
bool dof_to_quad, |
| 551 |
bool add, |
| 552 |
bool in_place, |
| 553 |
typename ViewTypeIn, |
| 554 |
typename ViewTypeOut> |
| 555 |
DEAL_II_HOST_DEVICE void |
| 556 |
EvaluatorTensorProduct<evaluate_general, dim, n_rows, n_columns, Number>:: |
| 557 |
co_gradients(const ViewTypeIn in, ViewTypeOut out) const |
| 558 |
{ |
| 559 |
if constexpr (in_place) |
| 560 |
{ |
| 561 |
apply<dim, |
| 562 |
n_columns, |
| 563 |
n_columns, |
| 564 |
Number, |
| 565 |
direction, |
| 566 |
dof_to_quad, |
| 567 |
false>(team_member, co_shape_gradients, in, temp); |
| 568 |
|
| 569 |
populate_view<add>(team_member, out, temp, out.extent(0)); |
| 570 |
} |
| 571 |
else |
| 572 |
apply<dim, n_columns, n_columns, Number, direction, dof_to_quad, add>( |
| 573 |
team_member, co_shape_gradients, in, out); |
| 574 |
} |
| 575 |
} // namespace internal |
| 576 |
} // namespace Portable |
| 577 |
|
| 578 |
DEAL_II_NAMESPACE_CLOSE |
| 579 |
|
| 580 |
#endif |