1241a4b83SYohann // Copyright (c) 2017-2018, Lawrence Livermore National Security, LLC. 2241a4b83SYohann // Produced at the Lawrence Livermore National Laboratory. LLNL-CODE-734707. 3241a4b83SYohann // All Rights reserved. See files LICENSE and NOTICE for details. 4241a4b83SYohann // 5241a4b83SYohann // This file is part of CEED, a collection of benchmarks, miniapps, software 6241a4b83SYohann // libraries and APIs for efficient high-order finite element and spectral 7241a4b83SYohann // element discretizations for exascale applications. For more information and 8241a4b83SYohann // source code availability see http://github.com/ceed. 9241a4b83SYohann // 10241a4b83SYohann // The CEED research is supported by the Exascale Computing Project 17-SC-20-SC, 11241a4b83SYohann // a collaborative effort of two U.S. Department of Energy organizations (Office 12241a4b83SYohann // of Science and the National Nuclear Security Administration) responsible for 13241a4b83SYohann // the planning and preparation of a capable exascale ecosystem, including 14241a4b83SYohann // software, applications, hardware, advanced system engineering and early 15241a4b83SYohann // testbed platforms, in support of the nation's exascale computing imperative. 16241a4b83SYohann 17241a4b83SYohann #include <ceed-backend.h> 18241a4b83SYohann #include <string.h> 19241a4b83SYohann #include <stdarg.h> 20241a4b83SYohann #include <nvrtc.h> 21241a4b83SYohann #include <cuda.h> 22241a4b83SYohann #include <cuda_runtime.h> 23241a4b83SYohann #include "ceed-cuda-gen.h" 24241a4b83SYohann 25ab213215SJeremy L Thompson //------------------------------------------------------------------------------ 26ab213215SJeremy L Thompson // Backend init 27ab213215SJeremy L Thompson //------------------------------------------------------------------------------ 28241a4b83SYohann static int CeedInit_Cuda_gen(const char *resource, Ceed ceed) { 29241a4b83SYohann int ierr; 30241a4b83SYohann const int nrc = 9; // number of characters in resource 31241a4b83SYohann if (strncmp(resource, "/gpu/cuda/gen", nrc)) 32241a4b83SYohann return CeedError(ceed, 1, "Cuda backend cannot use resource: %s", resource); 33241a4b83SYohann 34241a4b83SYohann Ceed ceedshared; 35241a4b83SYohann CeedInit("/gpu/cuda/shared", &ceedshared); 36241a4b83SYohann ierr = CeedSetDelegate(ceed, ceedshared); CeedChk(ierr); 37241a4b83SYohann 38241a4b83SYohann Ceed_Cuda_gen *data; 39241a4b83SYohann ierr = CeedCalloc(1,&data); CeedChk(ierr); 40241a4b83SYohann ierr = CeedSetData(ceed,(void *)&data); CeedChk(ierr); 41abfaacbbSSander Arens ierr = CeedCudaInit(ceed, resource, nrc); CeedChk(ierr); 42abfaacbbSSander Arens 43b1d74153SJeremy L Thompson const char fallbackresource[] = "/gpu/cuda/ref"; 44ccaff030SJeremy L Thompson ierr = CeedSetOperatorFallbackResource(ceed, fallbackresource); CeedChk(ierr); 45ccaff030SJeremy L Thompson 46241a4b83SYohann ierr = CeedSetBackendFunction(ceed, "Ceed", ceed, "QFunctionCreate", 47241a4b83SYohann CeedQFunctionCreate_Cuda_gen); CeedChk(ierr); 48241a4b83SYohann ierr = CeedSetBackendFunction(ceed, "Ceed", ceed, "OperatorCreate", 49241a4b83SYohann CeedOperatorCreate_Cuda_gen); CeedChk(ierr); 50*73b3ccafSJeremy L Thompson ierr = CeedSetBackendFunction(ceed, "Ceed", ceed, "Destroy", 51*73b3ccafSJeremy L Thompson CeedDestroy_Cuda); CeedChk(ierr); 52241a4b83SYohann return 0; 53241a4b83SYohann } 54241a4b83SYohann 55ab213215SJeremy L Thompson //------------------------------------------------------------------------------ 56ab213215SJeremy L Thompson // Register backend 57ab213215SJeremy L Thompson //------------------------------------------------------------------------------ 58241a4b83SYohann __attribute__((constructor)) 59241a4b83SYohann static void Register(void) { 60241a4b83SYohann CeedRegister("/gpu/cuda/gen", CeedInit_Cuda_gen, 40); 61241a4b83SYohann } 62ab213215SJeremy L Thompson //------------------------------------------------------------------------------ 63