CUDA中模板类和模板函数

以向量加法为例，包含三个文件：kernel.h,kernel.cu,test.cpp

kernel.h:

 

#ifndef __KERNEL_H_
#define __KERNEL_H_
extern "C" void runtest();
#endif

kernel.cu:

#include "kernel.h"
#include "cuda_runtime.h"
#include "device_launch_parameters.h"

#include 

template 
class operate {
public:
    cudaError_t addWithCuda(T *c, const T *a, const T *b, unsigned int size);
};

template 
void __global__ addKernel1(T *c, const T *a, const T *b)
{
    int i = threadIdx.x;
    c[i] = a[i] + b[i];
}


template 
cudaError_t operate::addWithCuda(T *c, const T *a, const T *b, unsigned int size)
{
    T *dev_a = 0;
    T *dev_b = 0;
    T *dev_c = 0;
    cudaError_t cudaStatus;

    // Choose which GPU to run on, change this on a multi-GPU system.
    cudaStatus = cudaSetDevice(0);
    if (cudaStatus != cudaSuccess) {
        fprintf(stderr, "cudaSetDevice failed!  Do you have a CUDA-capable GPU installed?");
        goto Error;
    }

    // Allocate GPU buffers for three vectors (two input, one output)    .
    cudaStatus = cudaMalloc((void**)&dev_c, size * sizeof(T));
    if (cudaStatus != cudaSuccess) {
        fprintf(stderr, "cudaMalloc failed!");
        goto Error;
    }

    cudaStatus = cudaMalloc((void**)&dev_a, size * sizeof(T));
    if (cudaStatus != cudaSuccess) {
        fprintf(stderr, "cudaMalloc failed!");
        goto Error;
    }

    cudaStatus = cudaMalloc((void**)&dev_b, size * sizeof(T));
    if (cudaStatus != cudaSuccess) {
        fprintf(stderr, "cudaMalloc failed!");
        goto Error;
    }

    // Copy input vectors from host memory to GPU buffers.
    cudaStatus = cudaMemcpy(dev_a, a, size * sizeof(T), cudaMemcpyHostToDevice);
    if (cudaStatus != cudaSuccess) {
        fprintf(stderr, "cudaMemcpy failed!");
        goto Error;
    }

    cudaStatus = cudaMemcpy(dev_b, b, size * sizeof(T), cudaMemcpyHostToDevice);
    if (cudaStatus != cudaSuccess) {
        fprintf(stderr, "cudaMemcpy failed!");
        goto Error;
    }

    // Launch a kernel on the GPU with one thread for each element.
    addKernel1<<<1, size>>>(dev_c, dev_a, dev_b);

    // Check for any errors launching the kernel
    cudaStatus = cudaGetLastError();
    if (cudaStatus != cudaSuccess) {
        fprintf(stderr, "addKernel launch failed: %s\n", cudaGetErrorString(cudaStatus));
        goto Error;
    }
    
    // cudaDeviceSynchronize waits for the kernel to finish, and returns
    // any errors encountered during the launch.
    cudaStatus = cudaDeviceSynchronize();
    if (cudaStatus != cudaSuccess) {
        fprintf(stderr, "cudaDeviceSynchronize returned error code %d after launching addKernel!\n", cudaStatus);
        goto Error;
    }

    // Copy output vector from GPU buffer to host memory.
    cudaStatus = cudaMemcpy(c, dev_c, size * sizeof(T), cudaMemcpyDeviceToHost);
    if (cudaStatus != cudaSuccess) {
        fprintf(stderr, "cudaMemcpy failed!");
        goto Error;
    }

Error:
    cudaFree(dev_c);
    cudaFree(dev_a);
    cudaFree(dev_b);
    
    return cudaStatus;
}

extern "C" void runtest()
{
    const int arraySize = 5;
    const double a_d[arraySize] = { 1.1, 2.2, 3.3, 4.4, 5.5 };
    const double b_d[arraySize] = { 10.1, 20.1, 30.1, 40.1, 50.1 };
    double c_d[arraySize] = { 0 };

    // Add vectors in parallel.
	operate op;
	cudaError_t cudaStatus = op.addWithCuda(c_d, a_d, b_d, arraySize);
    if (cudaStatus != cudaSuccess) {
        fprintf(stderr, "addWithCuda failed!");
        return;
	}

    printf("{1.1,2.2,3.3,4.4,5.5} + {10.1,20.1,30.1,40.1,50.1} = {%f,%f,%f,%f,%f}\n",
        c_d[0], c_d[1], c_d[2], c_d[3], c_d[4]);

    // cudaDeviceReset must be called before exiting in order for profiling and
    // tracing tools such as Nsight and Visual Profiler to show complete traces.
    cudaStatus = cudaDeviceReset();
    if (cudaStatus != cudaSuccess) {
        fprintf(stderr, "cudaDeviceReset failed!");
        return;
    }
}

 

test.cpp:

#include "kernel.h"
#include 
int main()
{
    runtest();
    char a=getchar();
    return 0;
}