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

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 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 #ifndef GMX_EWALD_PME_GPU_UTILS_H
  36 #define GMX_EWALD_PME_GPU_UTILS_H
  37
  38 /*! \internal \file
  39  * \brief This file defines the small PME GPU inline host/device functions.
  40  *
  41  * \author Aleksei Iupinov <a.yupinov@gmail.com>
  42  * \ingroup module_ewald
  43  */
  44
  45 #include "config.h"
  46
  47 #include <cassert>
  48
  49 #include "pme-gpu-constants.h"
  50
  51 //! A macro for inline GPU functions.
  52 #if GMX_GPU == GMX_GPU_CUDA
  53 #define INLINE_EVERYWHERE __host__ __device__ __forceinline__
  54 #else
  55 #define INLINE_EVERYWHERE inline
  56 #endif
  57
  58 /*! \internal \brief
  59  * Gets a base of the unique index to an element in a spline parameter buffer (theta/dtheta),
  60  * which is laid out for GPU spread/gather kernels. The base only corresponds to the atom index within the execution block.
  61  * Feed the result into getSplineParamIndex() to get a full index.
  62  * TODO: it's likely that both parameters can be just replaced with a single atom index, as they are derived from it.
  63  * Do that, verifying that the generated code is not bloated, and/or revise the spline indexing scheme.
  64  * Removing warp dependency would also be nice (and would probably coincide with removing PME_SPREADGATHER_ATOMS_PER_WARP).
  65  *
  66  * \tparam order               PME order
  67  * \param[in] warpIndex        Warp index wrt the block.
  68  * \param[in] atomWarpIndex    Atom index wrt the warp (from 0 to PME_SPREADGATHER_ATOMS_PER_WARP - 1).
  69  *
  70  * \returns Index into theta or dtheta array using GPU layout.
  71  */
  72 template <int order>
  73 int INLINE_EVERYWHERE getSplineParamIndexBase(int warpIndex, int atomWarpIndex)
  74 {
  75     assert((atomWarpIndex >= 0) && (atomWarpIndex < PME_SPREADGATHER_ATOMS_PER_WARP));
  76     const int dimIndex    = 0;
  77     const int splineIndex = 0;
  78     // The zeroes are here to preserve the full index formula for reference
  79     return (((splineIndex + order * warpIndex) * DIM + dimIndex) * PME_SPREADGATHER_ATOMS_PER_WARP + atomWarpIndex);
  80 }
  81
  82 /*! \internal \brief
  83  * Gets a unique index to an element in a spline parameter buffer (theta/dtheta),
  84  * which is laid out for GPU spread/gather kernels. The index is wrt to the execution block,
  85  * in range(0, atomsPerBlock * order * DIM).
  86  * This function consumes result of getSplineParamIndexBase() and adjusts it for \p dimIndex and \p splineIndex.
  87  *
  88  * \tparam order               PME order
  89  * \param[in] paramIndexBase   Must be result of getSplineParamIndexBase().
  90  * \param[in] dimIndex         Dimension index (from 0 to 2)
  91  * \param[in] splineIndex      Spline contribution index (from 0 to \p order - 1)
  92  *
  93  * \returns Index into theta or dtheta array using GPU layout.
  94  */
  95 template <int order>
  96 int INLINE_EVERYWHERE getSplineParamIndex(int paramIndexBase, int dimIndex, int splineIndex)
  97 {
  98     assert((dimIndex >= XX) && (dimIndex < DIM));
  99     assert((splineIndex >= 0) && (splineIndex < order));
 100     return (paramIndexBase + (splineIndex * DIM + dimIndex) * PME_SPREADGATHER_ATOMS_PER_WARP);
 101 }
 102
 103 #endif