Mercurial > hg > Members > Moririn

comparison src/parallel_execution/CUDAtwice.cbc @ 314:1839586f5b41

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
pthread CUDA test
author Shinji KONO <kono@ie.u-ryukyu.ac.jp>
date Wed, 15 Feb 2017 12:34:19 +0900
parents 7dd5a7d52a67
children 54d203daf06b
comparison
equal deleted inserted replaced
313:4addbc7469ee 314:1839586f5b41
10 // Worker *worker = context->worker; 10 // Worker *worker = context->worker;
11 // CUDAWorker* cudaWorker = (CUDAWorker*)worker->worker; 11 // CUDAWorker* cudaWorker = (CUDAWorker*)worker->worker;
12 // memory allocate 12 // memory allocate
13 CUdeviceptr devA; 13 CUdeviceptr devA;
14 CUdeviceptr devLoopCounter; 14 CUdeviceptr devLoopCounter;
15 printf("CUdA Exe 1\n");
15 16
16 checkCudaErrors(cuMemAlloc(&devA, array->size)); 17 checkCudaErrors(cuMemAlloc(&devA, array->size));
17 checkCudaErrors(cuMemAlloc(&devLoopCounter, sizeof(LoopCounter))); 18 checkCudaErrors(cuMemAlloc(&devLoopCounter, sizeof(LoopCounter)));
18 19
19 //twiceカーネルが定義されてなければそれをロードする 20 //twiceカーネルが定義されてなければそれをロードする
20 checkCudaErrors(cuModuleLoad(&context->module, "CUDAtwice.ptx")); 21 checkCudaErrors(cuModuleLoad(&context->module, "CUDAtwice.ptx"));
21 checkCudaErrors(cuModuleGetFunction(&context->function, context->module, "twice")); 22 checkCudaErrors(cuModuleGetFunction(&context->function, context->module, "twice"));
23 printf("CUdA Exe 2\n");
22 24
23 //入力のDataGearをGPUにbuffer経由で送る 25 //入力のDataGearをGPUにbuffer経由で送る
24 // Synchronous data transfer(host to device) 26 // Synchronous data transfer(host to device)
25 checkCudaErrors(cuMemcpyHtoD(devLoopCounter, loopCounter, sizeof(LoopCounter))); 27 checkCudaErrors(cuMemcpyHtoD(devLoopCounter, loopCounter, sizeof(LoopCounter)));
26 checkCudaErrors(cuMemcpyHtoD(devA, array->array, array->size)); 28 checkCudaErrors(cuMemcpyHtoD(devA, array->array, array->size));
51 loopCounter->i = 0; 53 loopCounter->i = 0;
52 goto meta(workerContext, workerContext->next); 54 goto meta(workerContext, workerContext->next);
53 } 55 }
54 56
55 __code CUDAtwice_stub(struct Context* context) { 57 __code CUDAtwice_stub(struct Context* context) {
58 printf("CUdAtwice stub\n");
56 struct LoopCounter* loopCounter = &context->data[context->dataNum]->LoopCounter; 59 struct LoopCounter* loopCounter = &context->data[context->dataNum]->LoopCounter;
57 struct Array* array = &context->data[context->dataNum+1]->Array; 60 struct Array* array = &context->data[context->dataNum+1]->Array;
58 CUDAExec(context,array,loopCounter); 61 CUDAExec(context,array,loopCounter);
59 62
60 //continuationにそってGPUworkerに戻る 63 //continuationにそってGPUworkerに戻る