src/gromacs/gpu_utils/oclutils.cpp

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2014,2015,2016,2017,2018 by the GROMACS development team.
   5  * Copyright (c) 2019,2020, by the GROMACS development team, led by
   6  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   7  * and including many others, as listed in the AUTHORS file in the
   8  * top-level source directory and at http://www.gromacs.org.
   9  *
  10  * GROMACS is free software; you can redistribute it and/or
  11  * modify it under the terms of the GNU Lesser General Public License
  12  * as published by the Free Software Foundation; either version 2.1
  13  * of the License, or (at your option) any later version.
  14  *
  15  * GROMACS is distributed in the hope that it will be useful,
  16  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  17  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  18  * Lesser General Public License for more details.
  19  *
  20  * You should have received a copy of the GNU Lesser General Public
  21  * License along with GROMACS; if not, see
  22  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  23  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  24  *
  25  * If you want to redistribute modifications to GROMACS, please
  26  * consider that scientific software is very special. Version
  27  * control is crucial - bugs must be traceable. We will be happy to
  28  * consider code for inclusion in the official distribution, but
  29  * derived work must not be called official GROMACS. Details are found
  30  * in the README & COPYING files - if they are missing, get the
  31  * official version at http://www.gromacs.org.
  32  *
  33  * To help us fund GROMACS development, we humbly ask that you cite
  34  * the research papers on the package. Check out http://www.gromacs.org.
  35  */
  36 /*! \internal \file
  37  *  \brief Define utility routines for OpenCL
  38  *
  39  *  \author Anca Hamuraru <anca@streamcomputing.eu>
  40  */
  41 #include "gmxpre.h"
  42
  43 #include "oclutils.h"
  44
  45 #include <stdlib.h>
  46
  47 #include <cassert>
  48 #include <cstdio>
  49
  50 #include <string>
  51
  52 #include "gromacs/gpu_utils/gpu_utils.h"
  53 #include "gromacs/utility/fatalerror.h"
  54 #include "gromacs/utility/smalloc.h"
  55
  56 /*! \brief \brief Allocates nbytes of host memory. Use ocl_free to free memory allocated with this function.
  57  *
  58  *  \todo
  59  *  This function should allocate page-locked memory to help reduce D2H and H2D
  60  *  transfer times, similar with pmalloc from pmalloc_cuda.cu.
  61  *
  62  * \param[in,out]    h_ptr   Pointer where to store the address of the newly allocated buffer.
  63  * \param[in]        nbytes  Size in bytes of the buffer to be allocated.
  64  */
  65 void pmalloc(void** h_ptr, size_t nbytes)
  66 {
  67     /* Need a temporary type whose size is 1 byte, so that the
  68      * implementation of snew_aligned can cope without issuing
  69      * warnings. */
  70     char** temporary = reinterpret_cast<char**>(h_ptr);
  71
  72     /* 16-byte alignment is required by the neighbour-searching code,
  73      * because it uses four-wide SIMD for bounding-box calculation.
  74      * However, when we organize using page-locked memory for
  75      * device-host transfers, it will probably need to be aligned to a
  76      * 4kb page, like CUDA does. */
  77     snew_aligned(*temporary, nbytes, 16);
  78 }
  79
  80 /*! \brief Frees memory allocated with pmalloc.
  81  *
  82  * \param[in]    h_ptr   Buffer allocated with pmalloc that needs to be freed.
  83  */
  84 void pfree(void* h_ptr)
  85 {
  86
  87     if (h_ptr)
  88     {
  89         sfree_aligned(h_ptr);
  90     }
  91 }
  92
  93 /*! \brief Convert error code to diagnostic string */
  94 std::string ocl_get_error_string(cl_int error)
  95 {
  96     switch (error)
  97     {
  98         // run-time and JIT compiler errors
  99         case 0: return "CL_SUCCESS";
 100         case -1: return "CL_DEVICE_NOT_FOUND";
 101         case -2: return "CL_DEVICE_NOT_AVAILABLE";
 102         case -3: return "CL_COMPILER_NOT_AVAILABLE";
 103         case -4: return "CL_MEM_OBJECT_ALLOCATION_FAILURE";
 104         case -5: return "CL_OUT_OF_RESOURCES";
 105         case -6: return "CL_OUT_OF_HOST_MEMORY";
 106         case -7: return "CL_PROFILING_INFO_NOT_AVAILABLE";
 107         case -8: return "CL_MEM_COPY_OVERLAP";
 108         case -9: return "CL_IMAGE_FORMAT_MISMATCH";
 109         case -10: return "CL_IMAGE_FORMAT_NOT_SUPPORTED";
 110         case -11: return "CL_BUILD_PROGRAM_FAILURE";
 111         case -12: return "CL_MAP_FAILURE";
 112         case -13: return "CL_MISALIGNED_SUB_BUFFER_OFFSET";
 113         case -14: return "CL_EXEC_STATUS_ERROR_FOR_EVENTS_IN_WAIT_LIST";
 114         case -15: return "CL_COMPILE_PROGRAM_FAILURE";
 115         case -16: return "CL_LINKER_NOT_AVAILABLE";
 116         case -17: return "CL_LINK_PROGRAM_FAILURE";
 117         case -18: return "CL_DEVICE_PARTITION_FAILED";
 118         case -19: return "CL_KERNEL_ARG_INFO_NOT_AVAILABLE";
 119
 120         // compile-time errors
 121         case -30: return "CL_INVALID_VALUE";
 122         case -31: return "CL_INVALID_DEVICE_TYPE";
 123         case -32: return "CL_INVALID_PLATFORM";
 124         case -33: return "CL_INVALID_DEVICE";
 125         case -34: return "CL_INVALID_CONTEXT";
 126         case -35: return "CL_INVALID_QUEUE_PROPERTIES";
 127         case -36: return "CL_INVALID_COMMAND_QUEUE";
 128         case -37: return "CL_INVALID_HOST_PTR";
 129         case -38: return "CL_INVALID_MEM_OBJECT";
 130         case -39: return "CL_INVALID_IMAGE_FORMAT_DESCRIPTOR";
 131         case -40: return "CL_INVALID_IMAGE_SIZE";
 132         case -41: return "CL_INVALID_SAMPLER";
 133         case -42: return "CL_INVALID_BINARY";
 134         case -43: return "CL_INVALID_BUILD_OPTIONS";
 135         case -44: return "CL_INVALID_PROGRAM";
 136         case -45: return "CL_INVALID_PROGRAM_EXECUTABLE";
 137         case -46: return "CL_INVALID_KERNEL_NAME";
 138         case -47: return "CL_INVALID_KERNEL_DEFINITION";
 139         case -48: return "CL_INVALID_KERNEL";
 140         case -49: return "CL_INVALID_ARG_INDEX";
 141         case -50: return "CL_INVALID_ARG_VALUE";
 142         case -51: return "CL_INVALID_ARG_SIZE";
 143         case -52: return "CL_INVALID_KERNEL_ARGS";
 144         case -53: return "CL_INVALID_WORK_DIMENSION";
 145         case -54: return "CL_INVALID_WORK_GROUP_SIZE";
 146         case -55: return "CL_INVALID_WORK_ITEM_SIZE";
 147         case -56: return "CL_INVALID_GLOBAL_OFFSET";
 148         case -57: return "CL_INVALID_EVENT_WAIT_LIST";
 149         case -58: return "CL_INVALID_EVENT";
 150         case -59: return "CL_INVALID_OPERATION";
 151         case -60: return "CL_INVALID_GL_OBJECT";
 152         case -61: return "CL_INVALID_BUFFER_SIZE";
 153         case -62: return "CL_INVALID_MIP_LEVEL";
 154         case -63: return "CL_INVALID_GLOBAL_WORK_SIZE";
 155         case -64: return "CL_INVALID_PROPERTY";
 156         case -65: return "CL_INVALID_IMAGE_DESCRIPTOR";
 157         case -66: return "CL_INVALID_COMPILER_OPTIONS";
 158         case -67: return "CL_INVALID_LINKER_OPTIONS";
 159         case -68: return "CL_INVALID_DEVICE_PARTITION_COUNT";
 160
 161         // extension errors
 162         case -1000: return "CL_INVALID_GL_SHAREGROUP_REFERENCE_KHR";
 163         case -1001: return "CL_PLATFORM_NOT_FOUND_KHR";
 164         case -1002: return "CL_INVALID_D3D10_DEVICE_KHR";
 165         case -1003: return "CL_INVALID_D3D10_RESOURCE_KHR";
 166         case -1004: return "CL_D3D10_RESOURCE_ALREADY_ACQUIRED_KHR";
 167         case -1005: return "CL_D3D10_RESOURCE_NOT_ACQUIRED_KHR";
 168         default: return "Unknown OpenCL error: " + std::to_string(static_cast<int32_t>(error));
 169     }
 170 }