src/gromacs/ewald/pme-gpu-internal.h

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2016,2017,2018, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35
  36 /*! \internal \file
  37  *
  38  * \brief This file contains internal function definitions for performing the PME calculations on GPU.
  39  * These are not meant to be exposed outside of the PME GPU code.
  40  * As of now, their bodies are still in the common pme-gpu.cpp files.
  41  *
  42  * \author Aleksei Iupinov <a.yupinov@gmail.com>
  43  * \ingroup module_ewald
  44  */
  45
  46 #ifndef GMX_EWALD_PME_GPU_INTERNAL_H
  47 #define GMX_EWALD_PME_GPU_INTERNAL_H
  48
  49 #include "gromacs/fft/fft.h"                   // for the gmx_fft_direction enum
  50 #include "gromacs/gpu_utils/gpu_macros.h"      // for the GPU_FUNC_ macros
  51 #include "gromacs/utility/arrayref.h"
  52
  53 #include "pme-gpu-types-host.h"                     // for the inline functions accessing PmeGpu members
  54
  55 struct gmx_hw_info_t;
  56 struct gmx_gpu_opt_t;
  57 struct gmx_pme_t;                              // only used in pme_gpu_reinit
  58 struct gmx_wallclock_gpu_pme_t;
  59 struct pme_atomcomm_t;
  60 struct t_complex;
  61
  62 namespace gmx
  63 {
  64 class MDLogger;
  65 }
  66
  67 //! Type of spline data
  68 enum class PmeSplineDataType
  69 {
  70     Values,      // theta
  71     Derivatives, // dtheta
  72 };               //TODO move this into new and shiny pme.h (pme-types.h?)
  73
  74 //! PME grid dimension ordering (from major to minor)
  75 enum class GridOrdering
  76 {
  77     YZX,
  78     XYZ
  79 };
  80
  81 /*! \libinternal \brief
  82  * Returns the number of atoms per chunk in the atom charges/coordinates data layout.
  83  * Depends on CUDA-specific block sizes, needed for the atom data padding.
  84  *
  85  * \param[in] pmeGpu            The PME GPU structure.
  86  * \returns   Number of atoms in a single GPU atom data chunk.
  87  */
  88 int pme_gpu_get_atom_data_alignment(const PmeGpu *pmeGpu);
  89
  90 /*! \libinternal \brief
  91  * Returns the number of atoms per chunk in the atom spline theta/dtheta data layout.
  92  *
  93  * \param[in] pmeGpu            The PME GPU structure.
  94  * \returns   Number of atoms in a single GPU atom spline data chunk.
  95  */
  96 int pme_gpu_get_atoms_per_warp(const PmeGpu *pmeGpu);
  97
  98 /*! \libinternal \brief
  99  * Synchronizes the current computation, waiting for the GPU kernels/transfers to finish.
 100  *
 101  * \param[in] pmeGpu            The PME GPU structure.
 102  */
 103 GPU_FUNC_QUALIFIER void pme_gpu_synchronize(const PmeGpu *GPU_FUNC_ARGUMENT(pmeGpu)) GPU_FUNC_TERM
 104
 105 /*! \libinternal \brief
 106  * Allocates the fixed size energy and virial buffer both on GPU and CPU.
 107  *
 108  * \param[in] pmeGpu            The PME GPU structure.
 109  */
 110 void pme_gpu_alloc_energy_virial(const PmeGpu *pmeGpu);
 111
 112 /*! \libinternal \brief
 113  * Frees the energy and virial memory both on GPU and CPU.
 114  *
 115  * \param[in] pmeGpu            The PME GPU structure.
 116  */
 117 void pme_gpu_free_energy_virial(PmeGpu *pmeGpu);
 118
 119 /*! \libinternal \brief
 120  * Clears the energy and virial memory on GPU with 0.
 121  * Should be called at the end of PME computation which returned energy/virial.
 122  *
 123  * \param[in] pmeGpu            The PME GPU structure.
 124  */
 125 void pme_gpu_clear_energy_virial(const PmeGpu *pmeGpu);
 126
 127 /*! \libinternal \brief
 128  * Reallocates and copies the pre-computed B-spline values to the GPU.
 129  *
 130  * \param[in] pmeGpu             The PME GPU structure.
 131  */
 132 void pme_gpu_realloc_and_copy_bspline_values(const PmeGpu *pmeGpu);
 133
 134 /*! \libinternal \brief
 135  * Frees the pre-computed B-spline values on the GPU (and the transfer CPU buffers).
 136  *
 137  * \param[in] pmeGpu             The PME GPU structure.
 138  */
 139 void pme_gpu_free_bspline_values(const PmeGpu *pmeGpu);
 140
 141 /*! \libinternal \brief
 142  * Reallocates the GPU buffer for the PME forces.
 143  *
 144  * \param[in] pmeGpu             The PME GPU structure.
 145  */
 146 void pme_gpu_realloc_forces(PmeGpu *pmeGpu);
 147
 148 /*! \libinternal \brief
 149  * Frees the GPU buffer for the PME forces.
 150  *
 151  * \param[in] pmeGpu             The PME GPU structure.
 152  */
 153 void pme_gpu_free_forces(const PmeGpu *pmeGpu);
 154
 155 /*! \libinternal \brief
 156  * Copies the forces from the CPU buffer to the GPU (to reduce them with the PME GPU gathered forces).
 157  * To be called e.g. after the bonded calculations.
 158  *
 159  * \param[in] pmeGpu             The PME GPU structure.
 160  */
 161 void pme_gpu_copy_input_forces(PmeGpu *pmeGpu);
 162
 163 /*! \libinternal \brief
 164  * Copies the forces from the GPU to the CPU buffer. To be called after the gathering stage.
 165  *
 166  * \param[in] pmeGpu             The PME GPU structure.
 167  */
 168 void pme_gpu_copy_output_forces(PmeGpu *pmeGpu);
 169
 170 /*! \libinternal \brief
 171  * Checks whether work in the PME GPU stream has completed.
 172  *
 173  * \param[in] pmeGpu            The PME GPU structure.
 174  *
 175  * \returns                     True if work in the PME stream has completed.
 176  */
 177 bool pme_gpu_stream_query(const PmeGpu *pmeGpu);
 178
 179 /*! \libinternal \brief
 180  * Reallocates the input coordinates buffer on the GPU (and clears the padded part if needed).
 181  *
 182  * \param[in] pmeGpu            The PME GPU structure.
 183  *
 184  * Needs to be called on every DD step/in the beginning.
 185  */
 186 void pme_gpu_realloc_coordinates(const PmeGpu *pmeGpu);
 187
 188 /*! \libinternal \brief
 189  * Copies the input coordinates from the CPU buffer onto the GPU.
 190  *
 191  * \param[in] pmeGpu            The PME GPU structure.
 192  * \param[in] h_coordinates     Input coordinates (XYZ rvec array).
 193  *
 194  * Needs to be called for every PME computation. The coordinates are then used in the spline calculation.
 195  */
 196 GPU_FUNC_QUALIFIER void pme_gpu_copy_input_coordinates(const PmeGpu *GPU_FUNC_ARGUMENT(pmeGpu),
 197                                                        const rvec   *GPU_FUNC_ARGUMENT(h_coordinates)) GPU_FUNC_TERM
 198
 199 /*! \libinternal \brief
 200  * Frees the coordinates on the GPU.
 201  *
 202  * \param[in] pmeGpu            The PME GPU structure.
 203  */
 204 void pme_gpu_free_coordinates(const PmeGpu *pmeGpu);
 205
 206 /*! \libinternal \brief
 207  * Reallocates the buffer on the GPU and copies the charges/coefficients from the CPU buffer.
 208  * Clears the padded part if needed.
 209  *
 210  * \param[in] pmeGpu            The PME GPU structure.
 211  * \param[in] h_coefficients    The input atom charges/coefficients.
 212  *
 213  * Does not need to be done for every PME computation, only whenever the local charges change.
 214  * (So, in the beginning of the run, or on DD step).
 215  */
 216 void pme_gpu_realloc_and_copy_input_coefficients(const PmeGpu    *pmeGpu,
 217                                                  const float     *h_coefficients);
 218
 219 /*! \libinternal \brief
 220  * Frees the charges/coefficients on the GPU.
 221  *
 222  * \param[in] pmeGpu             The PME GPU structure.
 223  */
 224 void pme_gpu_free_coefficients(const PmeGpu *pmeGpu);
 225
 226 /*! \libinternal \brief
 227  * Reallocates the buffers on the GPU and the host for the atoms spline data.
 228  *
 229  * \param[in] pmeGpu            The PME GPU structure.
 230  */
 231 void pme_gpu_realloc_spline_data(const PmeGpu *pmeGpu);
 232
 233 /*! \libinternal \brief
 234  * Frees the buffers on the GPU for the atoms spline data.
 235  *
 236  * \param[in] pmeGpu            The PME GPU structure.
 237  */
 238 void pme_gpu_free_spline_data(const PmeGpu *pmeGpu);
 239
 240 /*! \libinternal \brief
 241  * Reallocates the buffers on the GPU and the host for the particle gridline indices.
 242  *
 243  * \param[in] pmeGpu            The PME GPU structure.
 244  */
 245 void pme_gpu_realloc_grid_indices(const PmeGpu *pmeGpu);
 246
 247 /*! \libinternal \brief
 248  * Frees the buffer on the GPU for the particle gridline indices.
 249  *
 250  * \param[in] pmeGpu            The PME GPU structure.
 251  */
 252 void pme_gpu_free_grid_indices(const PmeGpu *pmeGpu);
 253
 254 /*! \libinternal \brief
 255  * Reallocates the real space grid and the complex reciprocal grid (if needed) on the GPU.
 256  *
 257  * \param[in] pmeGpu            The PME GPU structure.
 258  */
 259 void pme_gpu_realloc_grids(PmeGpu *pmeGpu);
 260
 261 /*! \libinternal \brief
 262  * Frees the real space grid and the complex reciprocal grid (if needed) on the GPU.
 263  *
 264  * \param[in] pmeGpu            The PME GPU structure.
 265  */
 266 void pme_gpu_free_grids(const PmeGpu *pmeGpu);
 267
 268 /*! \libinternal \brief
 269  * Clears the real space grid on the GPU.
 270  * Should be called at the end of each computation.
 271  *
 272  * \param[in] pmeGpu            The PME GPU structure.
 273  */
 274 void pme_gpu_clear_grids(const PmeGpu *pmeGpu);
 275
 276 /*! \libinternal \brief
 277  * Reallocates and copies the pre-computed fractional coordinates' shifts to the GPU.
 278  *
 279  * \param[in] pmeGpu            The PME GPU structure.
 280  */
 281 void pme_gpu_realloc_and_copy_fract_shifts(PmeGpu *pmeGpu);
 282
 283 /*! \libinternal \brief
 284  * Frees the pre-computed fractional coordinates' shifts on the GPU.
 285  *
 286  * \param[in] pmeGpu            The PME GPU structure.
 287  */
 288 void pme_gpu_free_fract_shifts(const PmeGpu *pmeGpu);
 289
 290 /*! \libinternal \brief
 291  * Copies the input real-space grid from the host to the GPU.
 292  *
 293  * \param[in] pmeGpu   The PME GPU structure.
 294  * \param[in] h_grid   The host-side grid buffer.
 295  */
 296 void pme_gpu_copy_input_gather_grid(const PmeGpu *pmeGpu,
 297                                     float        *h_grid);
 298
 299 /*! \libinternal \brief
 300  * Copies the output real-space grid from the GPU to the host.
 301  *
 302  * \param[in] pmeGpu   The PME GPU structure.
 303  * \param[out] h_grid  The host-side grid buffer.
 304  */
 305 void pme_gpu_copy_output_spread_grid(const PmeGpu *pmeGpu,
 306                                      float        *h_grid);
 307
 308 /*! \libinternal \brief
 309  * Copies the spread output spline data and gridline indices from the GPU to the host.
 310  *
 311  * \param[in] pmeGpu   The PME GPU structure.
 312  */
 313 void pme_gpu_copy_output_spread_atom_data(const PmeGpu *pmeGpu);
 314
 315 /*! \libinternal \brief
 316  * Copies the gather input spline data and gridline indices from the host to the GPU.
 317  *
 318  * \param[in] pmeGpu   The PME GPU structure.
 319  */
 320 void pme_gpu_copy_input_gather_atom_data(const PmeGpu *pmeGpu);
 321
 322 /*! \libinternal \brief
 323  * Waits for the grid copying to the host-side buffer after spreading to finish.
 324  *
 325  * \param[in] pmeGpu  The PME GPU structure.
 326  */
 327 void pme_gpu_sync_spread_grid(const PmeGpu *pmeGpu);
 328
 329 /*! \libinternal \brief
 330  * Does the one-time GPU-framework specific PME initialization.
 331  * For CUDA, the PME stream is created with the highest priority.
 332  *
 333  * \param[in] pmeGpu  The PME GPU structure.
 334  */
 335 void pme_gpu_init_internal(PmeGpu *pmeGpu);
 336
 337 /*! \libinternal \brief
 338  * Destroys the PME GPU-framework specific data.
 339  * Should be called last in the PME GPU destructor.
 340  *
 341  * \param[in] pmeGpu  The PME GPU structure.
 342  */
 343 void pme_gpu_destroy_specific(const PmeGpu *pmeGpu);
 344
 345 /*! \libinternal \brief
 346  * Initializes the CUDA FFT structures.
 347  *
 348  * \param[in] pmeGpu  The PME GPU structure.
 349  */
 350 void pme_gpu_reinit_3dfft(const PmeGpu *pmeGpu);
 351
 352 /*! \libinternal \brief
 353  * Destroys the CUDA FFT structures.
 354  *
 355  * \param[in] pmeGpu  The PME GPU structure.
 356  */
 357 void pme_gpu_destroy_3dfft(const PmeGpu *pmeGpu);
 358
 359 /* Several GPU event-based timing functions that live in pme-gpu-timings.cpp */
 360
 361 /*! \libinternal \brief
 362  * Finalizes all the active PME GPU stage timings for the current computation. Should be called at the end of every computation.
 363  *
 364  * \param[in] pmeGpu         The PME GPU structure.
 365  */
 366 void pme_gpu_update_timings(const PmeGpu *pmeGpu);
 367
 368 /*! \libinternal \brief
 369  * Updates the internal list of active PME GPU stages (if timings are enabled).
 370  *
 371  * \param[in] pmeGpu         The PME GPU data structure.
 372  */
 373 void pme_gpu_reinit_timings(const PmeGpu *pmeGpu);
 374
 375 /*! \brief
 376  * Resets the PME GPU timings. To be called at the reset MD step.
 377  *
 378  * \param[in] pmeGpu         The PME GPU structure.
 379  */
 380 void pme_gpu_reset_timings(const PmeGpu *pmeGpu);
 381
 382 /*! \libinternal \brief
 383  * Copies the PME GPU timings to the gmx_wallclock_gpu_t structure (for log output). To be called at the run end.
 384  *
 385  * \param[in] pmeGpu         The PME GPU structure.
 386  * \param[in] timings        The gmx_wallclock_gpu_pme_t structure.
 387  */
 388 void pme_gpu_get_timings(const PmeGpu            *pmeGpu,
 389                          gmx_wallclock_gpu_pme_t *timings);
 390
 391 /* The PME stages themselves */
 392
 393 /*! \libinternal \brief
 394  * A GPU spline computation and charge spreading function.
 395  *
 396  * \param[in]  pmeGpu          The PME GPU structure.
 397  * \param[in]  gridIndex       Index of the PME grid - unused, assumed to be 0.
 398  * \param[out] h_grid          The host-side grid buffer (used only if the result of the spread is expected on the host,
 399  *                             e.g. testing or host-side FFT)
 400  * \param[in]  computeSplines  Should the computation of spline parameters and gridline indices be performed.
 401  * \param[in]  spreadCharges   Should the charges/coefficients be spread on the grid.
 402  */
 403 CUDA_FUNC_QUALIFIER void pme_gpu_spread(const PmeGpu    *CUDA_FUNC_ARGUMENT(pmeGpu),
 404                                         int              CUDA_FUNC_ARGUMENT(gridIndex),
 405                                         real            *CUDA_FUNC_ARGUMENT(h_grid),
 406                                         bool             CUDA_FUNC_ARGUMENT(computeSplines),
 407                                         bool             CUDA_FUNC_ARGUMENT(spreadCharges)) CUDA_FUNC_TERM
 408
 409 /*! \libinternal \brief
 410  * 3D FFT R2C/C2R routine.
 411  *
 412  * \param[in]  pmeGpu          The PME GPU structure.
 413  * \param[in]  direction       Transform direction (real-to-complex or complex-to-real)
 414  * \param[in]  gridIndex       Index of the PME grid - unused, assumed to be 0.
 415  */
 416 void pme_gpu_3dfft(const PmeGpu          *pmeGpu,
 417                    enum gmx_fft_direction direction,
 418                    const int              gridIndex);
 419
 420 /*! \libinternal \brief
 421  * A GPU Fourier space solving function.
 422  *
 423  * \param[in]     pmeGpu                  The PME GPU structure.
 424  * \param[in,out] h_grid                  The host-side input and output Fourier grid buffer (used only with testing or host-side FFT)
 425  * \param[in]     gridOrdering            Specifies the dimenion ordering of the complex grid. TODO: store this information?
 426  * \param[in]     computeEnergyAndVirial  Tells if the energy and virial computation should also be performed.
 427  */
 428 CUDA_FUNC_QUALIFIER void pme_gpu_solve(const PmeGpu    *CUDA_FUNC_ARGUMENT(pmeGpu),
 429                                        t_complex       *CUDA_FUNC_ARGUMENT(h_grid),
 430                                        GridOrdering     CUDA_FUNC_ARGUMENT(gridOrdering),
 431                                        bool             CUDA_FUNC_ARGUMENT(computeEnergyAndVirial)) CUDA_FUNC_TERM
 432
 433 /*! \libinternal \brief
 434  * A GPU force gathering function.
 435  *
 436  * \param[in]     pmeGpu           The PME GPU structure.
 437  * \param[in]     forceTreatment   Tells how data in h_forces should be treated.
 438  *                                 TODO: determine efficiency/balance of host/device-side reductions.
 439  * \param[in]     h_grid           The host-side grid buffer (used only in testing mode)
 440  */
 441 CUDA_FUNC_QUALIFIER void pme_gpu_gather(PmeGpu                *CUDA_FUNC_ARGUMENT(pmeGpu),
 442                                         PmeForceOutputHandling CUDA_FUNC_ARGUMENT(forceTreatment),
 443                                         const float           *CUDA_FUNC_ARGUMENT(h_grid)
 444                                         ) CUDA_FUNC_TERM
 445
 446
 447 /* The inlined convenience PME GPU status getters */
 448
 449 /*! \libinternal \brief
 450  * Tells if PME runs on multiple GPUs with the decomposition.
 451  *
 452  * \param[in] pmeGpu         The PME GPU structure.
 453  * \returns                  True if PME runs on multiple GPUs, false otherwise.
 454  */
 455 inline bool pme_gpu_uses_dd(const PmeGpu *pmeGpu)
 456 {
 457     return !pmeGpu->settings.useDecomposition;
 458 }
 459
 460 /*! \libinternal \brief
 461  * Tells if PME performs the gathering stage on GPU.
 462  *
 463  * \param[in] pmeGpu         The PME GPU structure.
 464  * \returns                  True if the gathering is performed on GPU, false otherwise.
 465  */
 466 inline bool pme_gpu_performs_gather(const PmeGpu *pmeGpu)
 467 {
 468     return pmeGpu->settings.performGPUGather;
 469 }
 470
 471 /*! \libinternal \brief
 472  * Tells if PME performs the FFT stages on GPU.
 473  *
 474  * \param[in] pmeGpu         The PME GPU structure.
 475  * \returns                  True if FFT is performed on GPU, false otherwise.
 476  */
 477 inline bool pme_gpu_performs_FFT(const PmeGpu *pmeGpu)
 478 {
 479     return pmeGpu->settings.performGPUFFT;
 480 }
 481
 482 /*! \libinternal \brief
 483  * Tells if PME performs the grid (un-)wrapping on GPU.
 484  *
 485  * \param[in] pmeGpu         The PME GPU structure.
 486  * \returns                  True if (un-)wrapping is performed on GPU, false otherwise.
 487  */
 488 inline bool pme_gpu_performs_wrapping(const PmeGpu *pmeGpu)
 489 {
 490     return pmeGpu->settings.useDecomposition;
 491 }
 492
 493 /*! \libinternal \brief
 494  * Tells if PME performs the grid solving on GPU.
 495  *
 496  * \param[in] pmeGpu         The PME GPU structure.
 497  * \returns                  True if solving is performed on GPU, false otherwise.
 498  */
 499 inline bool pme_gpu_performs_solve(const PmeGpu *pmeGpu)
 500 {
 501     return pmeGpu->settings.performGPUSolve;
 502 }
 503
 504 /*! \libinternal \brief
 505  * Enables or disables the testing mode.
 506  * Testing mode only implies copying all the outputs, even the intermediate ones, to the host,
 507  * and also makes the copies synchronous.
 508  *
 509  * \param[in] pmeGpu             The PME GPU structure.
 510  * \param[in] testing            Should the testing mode be enabled, or disabled.
 511  */
 512 inline void pme_gpu_set_testing(PmeGpu *pmeGpu, bool testing)
 513 {
 514     if (pmeGpu)
 515     {
 516         pmeGpu->settings.copyAllOutputs = testing;
 517         pmeGpu->settings.transferKind   = testing ? GpuApiCallBehavior::Sync : GpuApiCallBehavior::Async;
 518     }
 519 }
 520
 521 /*! \libinternal \brief
 522  * Tells if PME is in the testing mode.
 523  *
 524  * \param[in] pmeGpu             The PME GPU structure.
 525  * \returns                      true if testing mode is enabled, false otherwise.
 526  */
 527 inline bool pme_gpu_is_testing(const PmeGpu *pmeGpu)
 528 {
 529     return pmeGpu->settings.copyAllOutputs;
 530 }
 531
 532 /* A block of C++ functions that live in pme-gpu-internal.cpp */
 533
 534 /*! \libinternal \brief
 535  * Returns the GPU gathering staging forces buffer.
 536  *
 537  * \param[in] pmeGpu             The PME GPU structure.
 538  * \returns                      The input/output forces.
 539  */
 540 GPU_FUNC_QUALIFIER gmx::ArrayRef<gmx::RVec> pme_gpu_get_forces(PmeGpu *GPU_FUNC_ARGUMENT(pmeGpu)) GPU_FUNC_TERM_WITH_RETURN(gmx::EmptyArrayRef())
 541
 542 /*! \libinternal \brief
 543  * Returns the output virial and energy of the PME solving.
 544  *
 545  * \param[in] pmeGpu             The PME GPU structure.
 546  * \param[out] energy            The output energy.
 547  * \param[out] virial            The output virial matrix.
 548  */
 549 GPU_FUNC_QUALIFIER void pme_gpu_get_energy_virial(const PmeGpu *GPU_FUNC_ARGUMENT(pmeGpu),
 550                                                   real *GPU_FUNC_ARGUMENT(energy),
 551                                                   matrix GPU_FUNC_ARGUMENT(virial)) GPU_FUNC_TERM
 552
 553 /*! \libinternal \brief
 554  * Updates the unit cell parameters. Does not check if update is necessary - that is done in pme_gpu_prepare_computation().
 555  *
 556  * \param[in] pmeGpu         The PME GPU structure.
 557  * \param[in] box            The unit cell box.
 558  */
 559 GPU_FUNC_QUALIFIER void pme_gpu_update_input_box(PmeGpu *GPU_FUNC_ARGUMENT(pmeGpu),
 560                                                  const matrix GPU_FUNC_ARGUMENT(box)) GPU_FUNC_TERM
 561
 562 /*! \libinternal \brief
 563  * Finishes the PME GPU computation, waiting for the output forces and/or energy/virial to be copied to the host.
 564  * If forces were computed, they will have arrived at the external host buffer provided to gather.
 565  * If virial/energy were computed, they will have arrived into the internal staging buffer
 566  * (even though that should have already happened before even launching the gather).
 567  * Finally, cudaEvent_t based GPU timers get updated if enabled. They also need stream synchronization for correctness.
 568  * Additionally, device-side buffers are cleared asynchronously for the next computation.
 569  *
 570  * \param[in] pmeGpu         The PME GPU structure.
 571  */
 572 void pme_gpu_finish_computation(const PmeGpu *pmeGpu);
 573
 574 //! A binary enum for spline data layout transformation
 575 enum class PmeLayoutTransform
 576 {
 577     GpuToHost,
 578     HostToGpu
 579 };
 580
 581 /*! \libinternal \brief
 582  * Rearranges the atom spline data between the GPU and host layouts.
 583  * Only used for test purposes so far, likely to be horribly slow.
 584  *
 585  * \param[in]  pmeGpu     The PME GPU structure.
 586  * \param[out] atc        The PME CPU atom data structure (with a single-threaded layout).
 587  * \param[in]  type       The spline data type (values or derivatives).
 588  * \param[in]  dimIndex   Dimension index.
 589  * \param[in]  transform  Layout transform type
 590  */
 591 GPU_FUNC_QUALIFIER void pme_gpu_transform_spline_atom_data(const PmeGpu *GPU_FUNC_ARGUMENT(pmeGpu),
 592                                                            const pme_atomcomm_t *GPU_FUNC_ARGUMENT(atc),
 593                                                            PmeSplineDataType GPU_FUNC_ARGUMENT(type),
 594                                                            int GPU_FUNC_ARGUMENT(dimIndex),
 595                                                            PmeLayoutTransform GPU_FUNC_ARGUMENT(transform)) GPU_FUNC_TERM
 596
 597 /*! \libinternal \brief
 598  * Gets a unique index to an element in a spline parameter buffer (theta/dtheta),
 599  * which is laid out for GPU spread/gather kernels. The index is wrt the execution block,
 600  * in range(0, atomsPerBlock * order * DIM).
 601  * This is a wrapper, only used in unit tests.
 602  * \param[in] order            PME order
 603  * \param[in] splineIndex      Spline contribution index (from 0 to \p order - 1)
 604  * \param[in] dimIndex         Dimension index (from 0 to 2)
 605  * \param[in] warpIndex        Warp index wrt the block.
 606  * \param[in] atomWarpIndex    Atom index wrt the warp (0 or 1).
 607  *
 608  * \returns Index into theta or dtheta array using GPU layout.
 609  */
 610 int getSplineParamFullIndex(int order,
 611                             int splineIndex,
 612                             int dimIndex,
 613                             int warpIndex,
 614                             int atomWarpIndex);
 615
 616 /*! \libinternal \brief
 617  * Get the normal/padded grid dimensions of the real-space PME grid on GPU. Only used in tests.
 618  *
 619  * \param[in] pmeGpu             The PME GPU structure.
 620  * \param[out] gridSize          Pointer to the grid dimensions to fill in.
 621  * \param[out] paddedGridSize    Pointer to the padded grid dimensions to fill in.
 622  */
 623 GPU_FUNC_QUALIFIER void pme_gpu_get_real_grid_sizes(const PmeGpu *GPU_FUNC_ARGUMENT(pmeGpu),
 624                                                     gmx::IVec *GPU_FUNC_ARGUMENT(gridSize),
 625                                                     gmx::IVec *GPU_FUNC_ARGUMENT(paddedGridSize)) GPU_FUNC_TERM
 626
 627 /*! \libinternal \brief
 628  * (Re-)initializes the PME GPU data at the beginning of the run or on DLB.
 629  *
 630  * \param[in,out] pme             The PME structure.
 631  * \param[in,out] gpuInfo         The GPU information structure.
 632  * \param[in]     pmeGpuProgram   The PME GPU program data
 633  * \throws gmx::NotImplementedError if this generally valid PME structure is not valid for GPU runs.
 634  */
 635 GPU_FUNC_QUALIFIER void pme_gpu_reinit(gmx_pme_t *GPU_FUNC_ARGUMENT(pme),
 636                                        gmx_device_info_t *GPU_FUNC_ARGUMENT(gpuInfo),
 637                                        PmeGpuProgramHandle GPU_FUNC_ARGUMENT(pmeGpuProgram)) GPU_FUNC_TERM
 638
 639 /*! \libinternal \brief
 640  * Destroys the PME GPU data at the end of the run.
 641  *
 642  * \param[in] pmeGpu     The PME GPU structure.
 643  */
 644 GPU_FUNC_QUALIFIER void pme_gpu_destroy(PmeGpu *GPU_FUNC_ARGUMENT(pmeGpu)) GPU_FUNC_TERM
 645
 646 /*! \libinternal \brief
 647  * Reallocates the local atoms data (charges, coordinates, etc.). Copies the charges to the GPU.
 648  *
 649  * \param[in] pmeGpu    The PME GPU structure.
 650  * \param[in] nAtoms    The number of particles.
 651  * \param[in] charges   The pointer to the host-side array of particle charges.
 652  *
 653  * This is a function that should only be called in the beginning of the run and on domain decomposition.
 654  * Should be called before the pme_gpu_set_io_ranges.
 655  */
 656 GPU_FUNC_QUALIFIER void pme_gpu_reinit_atoms(PmeGpu *GPU_FUNC_ARGUMENT(pmeGpu),
 657                                              const int         GPU_FUNC_ARGUMENT(nAtoms),
 658                                              const real       *GPU_FUNC_ARGUMENT(charges)) GPU_FUNC_TERM
 659
 660 /*! \brief \libinternal
 661  * The PME GPU reinitialization function that is called both at the end of any PME computation and on any load balancing.
 662  *
 663  * This clears the device-side working buffers in preparation for new computation.
 664  *
 665  * \param[in] pmeGpu            The PME GPU structure.
 666  */
 667 void pme_gpu_reinit_computation(const PmeGpu *pmeGpu);
 668
 669
 670 #endif