1 // Copyright (c) 2017-2018, Lawrence Livermore National Security, LLC. 2 // Produced at the Lawrence Livermore National Laboratory. LLNL-CODE-734707. 3 // All Rights reserved. See files LICENSE and NOTICE for details. 4 // 5 // This file is part of CEED, a collection of benchmarks, miniapps, software 6 // libraries and APIs for efficient high-order finite element and spectral 7 // element discretizations for exascale applications. For more information and 8 // source code availability see http://github.com/ceed. 9 // 10 // The CEED research is supported by the Exascale Computing Project 17-SC-20-SC, 11 // a collaborative effort of two U.S. Department of Energy organizations (Office 12 // of Science and the National Nuclear Security Administration) responsible for 13 // the planning and preparation of a capable exascale ecosystem, including 14 // software, applications, hardware, advanced system engineering and early 15 // testbed platforms, in support of the nation's exascale computing imperative. 16 17 #include <ceed-backend.h> 18 #include <string.h> 19 #include <stdio.h> 20 #include "../cuda/ceed-cuda.h" 21 #include "ceed-cuda-gen.h" 22 23 static int CeedQFunctionApply_Cuda_gen(CeedQFunction qf, CeedInt Q, 24 CeedVector *U, CeedVector *V) { 25 int ierr; 26 Ceed ceed; 27 ierr = CeedQFunctionGetCeed(qf, &ceed); CeedChk(ierr); 28 return CeedError(ceed, 1, "Backend does not implement QFunctionApply"); 29 } 30 31 static int CeedQFunctionDestroy_Cuda_gen(CeedQFunction qf) { 32 int ierr; 33 CeedQFunction_Cuda_gen *data; 34 ierr = CeedQFunctionGetData(qf, (void *)&data); CeedChk(ierr); 35 Ceed ceed; 36 ierr = CeedQFunctionGetCeed(qf, &ceed); CeedChk(ierr); 37 38 ierr = cudaFree(data->d_c); CeedChk_Cu(ceed, ierr); 39 40 ierr = CeedFree(&data); CeedChk(ierr); 41 42 return 0; 43 } 44 45 static int loadCudaFunction(CeedQFunction qf, char *c_src_file) { 46 int ierr; 47 Ceed ceed; 48 CeedQFunctionGetCeed(qf, &ceed); 49 char *cuda_file; 50 ierr = CeedCalloc(CUDA_MAX_PATH, &cuda_file); CeedChk(ierr); 51 memcpy(cuda_file, c_src_file, strlen(c_src_file)); 52 const char *last_dot = strrchr(cuda_file, '.'); 53 if (!last_dot) 54 return CeedError(ceed, 1, "Cannot find file's extension!"); 55 const size_t cuda_path_len = last_dot - cuda_file; 56 strcpy(&cuda_file[cuda_path_len], ".h"); 57 //******************* 58 FILE *fp; 59 long lSize; 60 char *buffer; 61 62 fp = fopen ( cuda_file, "rb" ); 63 if( !fp ) CeedError(ceed, 1, "Couldn't open the Cuda file for the QFunction."); 64 65 fseek( fp, 0L, SEEK_END); 66 lSize = ftell( fp ); 67 rewind( fp ); 68 69 /* allocate memory for entire content */ 70 ierr = CeedCalloc( lSize+1, &buffer ); CeedChk(ierr); 71 72 /* copy the file into the buffer */ 73 if( 1!=fread( buffer, lSize, 1, fp) ) { 74 fclose(fp); 75 CeedFree(&buffer); 76 CeedError(ceed, 1, "Couldn't read the Cuda file for the QFunction."); 77 } 78 79 //FIXME: the magic number 16 should be defined somewhere... 80 char *fields_string = 81 "typedef struct { const CeedScalar* inputs[16]; CeedScalar* outputs[16]; } Fields_Cuda_gen;"; 82 char *source = (char *) malloc(1 + strlen(fields_string)+ strlen(buffer) ); 83 strcpy(source, fields_string); 84 strcat(source, buffer); 85 86 //******************** 87 CeedQFunction_Cuda_gen *data; 88 ierr = CeedQFunctionGetData(qf, (void *)&data); CeedChk(ierr); 89 data->qFunctionSource = buffer; 90 91 //******************** 92 fclose(fp); 93 94 return 0; 95 } 96 97 int CeedQFunctionCreate_Cuda_gen(CeedQFunction qf) { 98 int ierr; 99 Ceed ceed; 100 CeedQFunctionGetCeed(qf, &ceed); 101 CeedQFunction_Cuda_gen *data; 102 ierr = CeedCalloc(1,&data); CeedChk(ierr); 103 ierr = CeedQFunctionSetData(qf, (void *)&data); CeedChk(ierr); 104 size_t ctxsize; 105 ierr = CeedQFunctionGetContextSize(qf, &ctxsize); CeedChk(ierr); 106 ierr = cudaMalloc(&data->d_c, ctxsize); CeedChk_Cu(ceed, ierr); 107 108 char *source; 109 ierr = CeedQFunctionGetSourcePath(qf, &source); CeedChk(ierr); 110 const char *funname = strrchr(source, ':') + 1; 111 data->qFunctionName = (char *)funname; 112 const int filenamelen = funname - source; 113 char filename[filenamelen]; 114 memcpy(filename, source, filenamelen - 1); 115 filename[filenamelen - 1] = '\0'; 116 ierr = loadCudaFunction(qf, filename); CeedChk(ierr); 117 118 ierr = CeedSetBackendFunction(ceed, "QFunction", qf, "Apply", 119 CeedQFunctionApply_Cuda_gen); CeedChk(ierr); 120 ierr = CeedSetBackendFunction(ceed, "QFunction", qf, "Destroy", 121 CeedQFunctionDestroy_Cuda_gen); CeedChk(ierr); 122 return 0; 123 } 124