2 * This file is part of the GROMACS molecular simulation package.
4 * Copyright (c) 2020, 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.
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.
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.
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.
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.
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.
37 * Runners for tests of CUDA types compatibility.
39 * \author Artem Zhmurov <zhmurov@gmail.com>
43 #include "typecasts_runner.h"
47 #include "gromacs/gpu_utils/cudautils.cuh"
48 #include "gromacs/gpu_utils/devicebuffer.h"
49 #include "gromacs/gpu_utils/typecasts.cuh"
50 #include "gromacs/hardware/device_information.h"
51 #include "gromacs/utility/exceptions.h"
52 #include "gromacs/utility/stringutil.h"
60 /* \brief Perform a component-wise conversion of the float3 vector back to RVec format.
62 * This is needed to pass the data back to the CPU testing code for comparison with the initial input.
64 * \param[out] rVecOutput Output data in RVec format for the output.
65 * \param[in] float3Output Output data in float3 format.
66 * \param[in] numElements Size of the data buffers.
68 void inline saveFloat3InRVecFormat(std::vector<gmx::RVec>& rVecOutput, const float3* float3Output, int numElements)
70 for (int i = 0; i < numElements; i++)
72 rVecOutput[i][XX] = float3Output[i].x;
73 rVecOutput[i][YY] = float3Output[i].y;
74 rVecOutput[i][ZZ] = float3Output[i].z;
78 void convertRVecToFloat3OnHost(std::vector<gmx::RVec>& rVecOutput, const std::vector<gmx::RVec>& rVecInput)
80 const int numElements = rVecInput.size();
82 float3* dataFloat3 = asFloat3(const_cast<RVec*>(rVecInput.data()));
84 saveFloat3InRVecFormat(rVecOutput, dataFloat3, numElements);
87 //! Number of CUDA threads in a block.
88 constexpr static int c_threadsPerBlock = 256;
90 /*! \brief GPU kernel to perform type conversion on the device.
92 * \param[out] gm_float3Output Buffer to write the output into.
93 * \param[in] gm_rVecInput Input data in RVec format.
94 * \param[in] size Size of the data buffers.
97 static __global__ void convertRVecToFloat3OnDevice_kernel(DeviceBuffer<float3> gm_float3Output,
98 DeviceBuffer<RVec> gm_rVecInput,
101 int threadIndex = blockIdx.x * blockDim.x + threadIdx.x;
102 if (threadIndex < size)
104 gm_float3Output[threadIndex] = asFloat3(gm_rVecInput)[threadIndex];
108 void convertRVecToFloat3OnDevice(std::vector<gmx::RVec>& h_rVecOutput,
109 const std::vector<gmx::RVec>& h_rVecInput,
110 const TestDevice* testDevice)
112 const DeviceContext& deviceContext = testDevice->deviceContext();
113 const DeviceStream& deviceStream = testDevice->deviceStream();
115 setActiveDevice(testDevice->deviceInfo());
117 const int numElements = h_rVecInput.size();
119 DeviceBuffer<RVec> d_rVecInput;
120 allocateDeviceBuffer(&d_rVecInput, numElements, deviceContext);
121 copyToDeviceBuffer(&d_rVecInput, h_rVecInput.data(), 0, numElements, deviceStream,
122 GpuApiCallBehavior::Sync, nullptr);
124 DeviceBuffer<float3> d_float3Output;
125 allocateDeviceBuffer(&d_float3Output, numElements * DIM, deviceContext);
127 std::vector<float3> h_float3Output(numElements);
129 KernelLaunchConfig kernelLaunchConfig;
130 kernelLaunchConfig.gridSize[0] = (numElements + c_threadsPerBlock - 1) / c_threadsPerBlock;
131 kernelLaunchConfig.blockSize[0] = c_threadsPerBlock;
132 kernelLaunchConfig.blockSize[1] = 1;
133 kernelLaunchConfig.blockSize[2] = 1;
134 kernelLaunchConfig.sharedMemorySize = 0;
136 auto kernelPtr = convertRVecToFloat3OnDevice_kernel;
137 const auto kernelArgs = prepareGpuKernelArguments(kernelPtr, kernelLaunchConfig,
138 &d_float3Output, &d_rVecInput, &numElements);
139 launchGpuKernel(kernelPtr, kernelLaunchConfig, deviceStream, nullptr,
140 "convertRVecToFloat3OnDevice_kernel", kernelArgs);
142 copyFromDeviceBuffer(h_float3Output.data(), &d_float3Output, 0, numElements, deviceStream,
143 GpuApiCallBehavior::Sync, nullptr);
145 saveFloat3InRVecFormat(h_rVecOutput, h_float3Output.data(), numElements);
147 freeDeviceBuffer(&d_rVecInput);
148 freeDeviceBuffer(&d_float3Output);