2 * This file is part of the GROMACS molecular simulation package.
4 * Copyright (c) 2012,2014,2015,2016,2017,2018,2019, 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.
35 #ifndef GMX_GPU_UTILS_CUDAUTILS_CUH
36 #define GMX_GPU_UTILS_CUDAUTILS_CUH
43 #include "gromacs/gpu_utils/gputraits.cuh"
44 #include "gromacs/math/vec.h"
45 #include "gromacs/math/vectypes.h"
46 #include "gromacs/utility/exceptions.h"
47 #include "gromacs/utility/fatalerror.h"
48 #include "gromacs/utility/gmxassert.h"
49 #include "gromacs/utility/stringutil.h"
56 /*! \brief Helper function to ensure no pending error silently
57 * disrupts error handling.
59 * Asserts in a debug build if an unhandled error is present. Issues a
60 * warning at run time otherwise.
62 * \todo This is similar to CU_CHECK_PREV_ERR, which should be
65 static inline void ensureNoPendingCudaError(const char* errorMessage)
67 // Ensure there is no pending error that would otherwise affect
68 // the behaviour of future error handling.
69 cudaError_t stat = cudaGetLastError();
70 if (stat == cudaSuccess)
75 // If we would find an error in a release build, we do not know
76 // what is appropriate to do about it, so assert only for debug
78 auto fullMessage = formatString(
79 "%s An unhandled error from a previous CUDA operation was detected. %s: %s",
80 errorMessage, cudaGetErrorName(stat), cudaGetErrorString(stat));
81 GMX_ASSERT(stat == cudaSuccess, fullMessage.c_str());
82 // TODO When we evolve a better logging framework, use that
83 // for release-build error reporting.
84 gmx_warning("%s", fullMessage.c_str());
90 enum class GpuApiCallBehavior;
92 /* TODO error checking needs to be rewritten. We have 2 types of error checks needed
93 based on where they occur in the code:
94 - non performance-critical: these errors are unsafe to be ignored and must be
95 _always_ checked for, e.g. initializations
96 - performance critical: handling errors might hurt performance so care need to be taken
97 when/if we should check for them at all, e.g. in cu_upload_X. However, we should be
98 able to turn the check for these errors on!
100 Probably we'll need two sets of the macros below...
103 #define CHECK_CUDA_ERRORS
105 #ifdef CHECK_CUDA_ERRORS
107 /*! Check for CUDA error on the return status of a CUDA RT API call. */
108 # define CU_RET_ERR(status, msg) \
111 if (status != cudaSuccess) \
113 gmx_fatal(FARGS, "%s: %s\n", msg, cudaGetErrorString(status)); \
117 /*! Check for any previously occurred uncaught CUDA error. */
118 # define CU_CHECK_PREV_ERR() \
121 cudaError_t _CU_CHECK_PREV_ERR_status = cudaGetLastError(); \
122 if (_CU_CHECK_PREV_ERR_status != cudaSuccess) \
125 "Just caught a previously occurred CUDA error (%s), will try to " \
127 cudaGetErrorString(_CU_CHECK_PREV_ERR_status)); \
131 #else /* CHECK_CUDA_ERRORS */
133 # define CU_RET_ERR(status, msg) \
137 # define CU_CHECK_PREV_ERR() \
142 #endif /* CHECK_CUDA_ERRORS */
144 /*! \brief CUDA device information.
146 * The CUDA device information is queried and set at detection and contains
147 * both information about the device/hardware returned by the runtime as well
148 * as additional data like support status.
150 struct gmx_device_info_t
152 int id; /* id of the CUDA device */
153 cudaDeviceProp prop; /* CUDA device properties */
154 int stat; /* result of the device check */
157 /*! Launches synchronous or asynchronous device to host memory copy.
159 * The copy is launched in stream s or if not specified, in stream 0.
161 int cu_copy_D2H(void* h_dest, void* d_src, size_t bytes, GpuApiCallBehavior transferKind, cudaStream_t /*s = nullptr*/);
163 /*! Launches synchronous host to device memory copy in stream 0. */
164 int cu_copy_D2H_sync(void* /*h_dest*/, void* /*d_src*/, size_t /*bytes*/);
166 /*! Launches asynchronous host to device memory copy in stream s. */
167 int cu_copy_D2H_async(void* /*h_dest*/, void* /*d_src*/, size_t /*bytes*/, cudaStream_t /*s = nullptr*/);
169 /*! Launches synchronous or asynchronous host to device memory copy.
171 * The copy is launched in stream s or if not specified, in stream 0.
173 int cu_copy_H2D(void* d_dest, const void* h_src, size_t bytes, GpuApiCallBehavior transferKind, cudaStream_t /*s = nullptr*/);
175 /*! Launches synchronous host to device memory copy. */
176 int cu_copy_H2D_sync(void* /*d_dest*/, const void* /*h_src*/, size_t /*bytes*/);
178 /*! Launches asynchronous host to device memory copy in stream s. */
179 int cu_copy_H2D_async(void* /*d_dest*/, const void* /*h_src*/, size_t /*bytes*/, cudaStream_t /*s = nullptr*/);
181 // TODO: the 2 functions below are pretty much a constructor/destructor of a simple
182 // GPU table object. There is also almost self-contained fetchFromParamLookupTable()
183 // in cuda_kernel_utils.cuh. They could all live in a separate class/struct file.
185 /*! \brief Initialize parameter lookup table.
187 * Initializes device memory, copies data from host and binds
188 * a texture to allocated device memory to be used for parameter lookup.
190 * \tparam[in] T Raw data type
191 * \param[out] d_ptr device pointer to the memory to be allocated
192 * \param[out] texObj texture object to be initialized
193 * \param[in] h_ptr pointer to the host memory to be uploaded to the device
194 * \param[in] numElem number of elements in the h_ptr
197 void initParamLookupTable(T*& d_ptr, cudaTextureObject_t& texObj, const T* h_ptr, int numElem);
199 // Add extern declarations so each translation unit understands that
200 // there will be a definition provided.
201 extern template void initParamLookupTable<int>(int*&, cudaTextureObject_t&, const int*, int);
202 extern template void initParamLookupTable<float>(float*&, cudaTextureObject_t&, const float*, int);
204 /*! \brief Destroy parameter lookup table.
206 * Unbinds texture object, deallocates device memory.
208 * \tparam[in] T Raw data type
209 * \param[in] d_ptr Device pointer to the memory to be deallocated
210 * \param[in] texObj Texture object to be deinitialized
213 void destroyParamLookupTable(T* d_ptr, cudaTextureObject_t texObj);
215 // Add extern declarations so each translation unit understands that
216 // there will be a definition provided.
217 extern template void destroyParamLookupTable<int>(int*, cudaTextureObject_t);
218 extern template void destroyParamLookupTable<float>(float*, cudaTextureObject_t);
220 /*! \brief Add a triplets stored in a float3 to an rvec variable.
222 * \param[out] a Rvec to increment
223 * \param[in] b Float triplet to increment with.
225 static inline void rvec_inc(rvec a, const float3 b)
227 rvec tmp = { b.x, b.y, b.z };
231 /*! \brief Wait for all taks in stream \p s to complete.
233 * \param[in] s stream to synchronize with
235 static inline void gpuStreamSynchronize(cudaStream_t s)
237 cudaError_t stat = cudaStreamSynchronize(s);
238 CU_RET_ERR(stat, "cudaStreamSynchronize failed");
241 /*! \brief Returns true if all tasks in \p s have completed.
243 * \param[in] s stream to check
245 * \returns True if all tasks enqueued in the stream \p s (at the time of this call) have completed.
247 static inline bool haveStreamTasksCompleted(cudaStream_t s)
249 cudaError_t stat = cudaStreamQuery(s);
251 if (stat == cudaErrorNotReady)
253 // work is still in progress in the stream
257 GMX_ASSERT(stat != cudaErrorInvalidResourceHandle, "Stream idnetifier not valid");
259 // cudaSuccess and cudaErrorNotReady are the expected return values
260 CU_RET_ERR(stat, "Unexpected cudaStreamQuery failure");
262 GMX_ASSERT(stat == cudaSuccess,
263 "Values other than cudaSuccess should have been explicitly handled");
268 /* Kernel launch helpers */
271 * A function for setting up a single CUDA kernel argument.
272 * This is the tail of the compile-time recursive function below.
273 * It has to be seen by the compiler first.
275 * \tparam totalArgsCount Number of the kernel arguments
276 * \tparam KernelPtr Kernel function handle type
277 * \param[in] argIndex Index of the current argument
279 template<size_t totalArgsCount, typename KernelPtr>
280 void prepareGpuKernelArgument(KernelPtr /*kernel*/,
281 std::array<void*, totalArgsCount>* /* kernelArgsPtr */,
282 size_t gmx_used_in_debug argIndex)
284 GMX_ASSERT(argIndex == totalArgsCount, "Tail expansion");
288 * Compile-time recursive function for setting up a single CUDA kernel argument.
289 * This function copies a kernel argument pointer \p argPtr into \p kernelArgsPtr,
290 * and calls itself on the next argument, eventually calling the tail function above.
292 * \tparam CurrentArg Type of the current argument
293 * \tparam RemainingArgs Types of remaining arguments after the current one
294 * \tparam totalArgsCount Number of the kernel arguments
295 * \tparam KernelPtr Kernel function handle type
296 * \param[in] kernel Kernel function handle
297 * \param[in,out] kernelArgsPtr Pointer to the argument array to be filled in
298 * \param[in] argIndex Index of the current argument
299 * \param[in] argPtr Pointer to the current argument
300 * \param[in] otherArgsPtrs Pack of pointers to arguments remaining to process after the current one
302 template<typename CurrentArg, typename... RemainingArgs, size_t totalArgsCount, typename KernelPtr>
303 void prepareGpuKernelArgument(KernelPtr kernel,
304 std::array<void*, totalArgsCount>* kernelArgsPtr,
306 const CurrentArg* argPtr,
307 const RemainingArgs*... otherArgsPtrs)
309 (*kernelArgsPtr)[argIndex] = (void*)argPtr;
310 prepareGpuKernelArgument(kernel, kernelArgsPtr, argIndex + 1, otherArgsPtrs...);
314 * A wrapper function for setting up all the CUDA kernel arguments.
315 * Calls the recursive functions above.
317 * \tparam KernelPtr Kernel function handle type
318 * \tparam Args Types of all the kernel arguments
319 * \param[in] kernel Kernel function handle
320 * \param[in] argsPtrs Pointers to all the kernel arguments
321 * \returns A prepared parameter pack to be used with launchGpuKernel() as the last argument.
323 template<typename KernelPtr, typename... Args>
324 std::array<void*, sizeof...(Args)> prepareGpuKernelArguments(KernelPtr kernel,
325 const KernelLaunchConfig& /*config */,
326 const Args*... argsPtrs)
328 std::array<void*, sizeof...(Args)> kernelArgs;
329 prepareGpuKernelArgument(kernel, &kernelArgs, 0, argsPtrs...);
333 /*! \brief Launches the CUDA kernel and handles the errors.
335 * \tparam Args Types of all the kernel arguments
336 * \param[in] kernel Kernel function handle
337 * \param[in] config Kernel configuration for launching
338 * \param[in] kernelName Human readable kernel description, for error handling only
339 * \param[in] kernelArgs Array of the pointers to the kernel arguments, prepared by
340 * prepareGpuKernelArguments() \throws gmx::InternalError on kernel launch failure
342 template<typename... Args>
343 void launchGpuKernel(void (*kernel)(Args...),
344 const KernelLaunchConfig& config,
345 CommandEvent* /*timingEvent */,
346 const char* kernelName,
347 const std::array<void*, sizeof...(Args)>& kernelArgs)
349 dim3 blockSize(config.blockSize[0], config.blockSize[1], config.blockSize[2]);
350 dim3 gridSize(config.gridSize[0], config.gridSize[1], config.gridSize[2]);
351 cudaLaunchKernel((void*)kernel, gridSize, blockSize, const_cast<void**>(kernelArgs.data()),
352 config.sharedMemorySize, config.stream);
354 cudaError_t status = cudaGetLastError();
355 if (cudaSuccess != status)
357 const std::string errorMessage =
358 "GPU kernel (" + std::string(kernelName)
359 + ") failed to launch: " + std::string(cudaGetErrorString(status));
360 GMX_THROW(gmx::InternalError(errorMessage));