VS2008 + cuda5.5 프로젝트 생성

예전에는 복잡하게 룰을 추가해야 했는데
2011/01/04 - [Programming/openCL / CUDA] - Visual Studio 2008 에서 CUDA 프로젝트 만들기

기본 템플렛이 추가 되어 있어서 편리하게 만들수 있다.

그런데.. 기본적으로 5개 짜리 더하는 예제가 추가되는건가?

view source

print?

001	#include "cuda_runtime.h"

002	#include "device_launch_parameters.h"

003

004	#include < stdio.h >

005

006	cudaError_t addWithCuda(int *c, const int *a, const int *b, unsigned int size);

007

008	__global__ void addKernel(int *c, const int *a, const int *b)

009

{

010	int i = threadIdx.x;

011	c[i] = a[i] + b[i];

012

}

013

014	int main()

015

{

016	const int arraySize = 5;

017	const int a[arraySize] = { 1, 2, 3, 4, 5 };

018	const int b[arraySize] = { 10, 20, 30, 40, 50 };

019	int c[arraySize] = { 0 };

020

021	// Add vectors in parallel.

022	cudaError_t cudaStatus = addWithCuda(c, a, b, arraySize);

023	if (cudaStatus != cudaSuccess) {

024	fprintf(stderr, "addWithCuda failed!");

025

return 1;

026

}

027

028	printf("{1,2,3,4,5} + {10,20,30,40,50} = {%d,%d,%d,%d,%d}\n",

029	c[0], c[1], c[2], c[3], c[4]);

030

031	// cudaDeviceReset must be called before exiting in order for profiling and

032	// tracing tools such as Nsight and Visual Profiler to show complete traces.

033	cudaStatus = cudaDeviceReset();

034	if (cudaStatus != cudaSuccess) {

035	fprintf(stderr, "cudaDeviceReset failed!");

036

return 1;

037

}

038

039

return 0;

040

}

041

042	// Helper function for using CUDA to add vectors in parallel.

043	cudaError_t addWithCuda(int *c, const int *a, const int *b, unsigned int size)

044

{

045	int *dev_a = 0;

046	int *dev_b = 0;

047	int *dev_c = 0;

048	cudaError_t cudaStatus;

049

050	// Choose which GPU to run on, change this on a multi-GPU system.

051	cudaStatus = cudaSetDevice(0);

052	if (cudaStatus != cudaSuccess) {

053	fprintf(stderr, "cudaSetDevice failed!  Do you have a CUDA-capable GPU installed?");

054	goto Error;

055

}

056

057	// Allocate GPU buffers for three vectors (two input, one output)    .

058	cudaStatus = cudaMalloc((void**)&dev_c, size * sizeof(int));

059	if (cudaStatus != cudaSuccess) {

060	fprintf(stderr, "cudaMalloc failed!");

061	goto Error;

062

}

063

064	cudaStatus = cudaMalloc((void**)&dev_a, size * sizeof(int));

065	if (cudaStatus != cudaSuccess) {

066	fprintf(stderr, "cudaMalloc failed!");

067	goto Error;

068

}

069

070	cudaStatus = cudaMalloc((void**)&dev_b, size * sizeof(int));

071	if (cudaStatus != cudaSuccess) {

072	fprintf(stderr, "cudaMalloc failed!");

073	goto Error;

074

}

075

076	// Copy input vectors from host memory to GPU buffers.

077	cudaStatus = cudaMemcpy(dev_a, a, size * sizeof(int), cudaMemcpyHostToDevice);

078	if (cudaStatus != cudaSuccess) {

079	fprintf(stderr, "cudaMemcpy failed!");

080	goto Error;

081

}

082

083	cudaStatus = cudaMemcpy(dev_b, b, size * sizeof(int), cudaMemcpyHostToDevice);

084	if (cudaStatus != cudaSuccess) {

085	fprintf(stderr, "cudaMemcpy failed!");

086	goto Error;

087

}

088

089	// Launch a kernel on the GPU with one thread for each element.

090	addKernel<<<1, size>>>(dev_c, dev_a, dev_b);

091

092	// Check for any errors launching the kernel

093	cudaStatus = cudaGetLastError();

094	if (cudaStatus != cudaSuccess) {

095	fprintf(stderr, "addKernel launch failed: %s\n", cudaGetErrorString(cudaStatus));

096	goto Error;

097

}

098

099	// cudaDeviceSynchronize waits for the kernel to finish, and returns

100	// any errors encountered during the launch.

101	cudaStatus = cudaDeviceSynchronize();

102	if (cudaStatus != cudaSuccess) {

103	fprintf(stderr, "cudaDeviceSynchronize returned error code %d after launching addKernel!\n", cudaStatus);

104	goto Error;

105

}

106

107	// Copy output vector from GPU buffer to host memory.

108	cudaStatus = cudaMemcpy(c, dev_c, size * sizeof(int), cudaMemcpyDeviceToHost);

109	if (cudaStatus != cudaSuccess) {

110	fprintf(stderr, "cudaMemcpy failed!");

111	goto Error;

112

}

113

114

Error:

115	cudaFree(dev_c);

116	cudaFree(dev_a);

117	cudaFree(dev_b);

118

119	return cudaStatus;

120

}