mexcuda中矩阵数据的传输

数据的传输以0开始,然后以列序传输

/*
 * Example of how to use the mxGPUArray API in a MEX file.  This example shows
 * how to write a MEX function that takes a gpuArray input and returns a
 * gpuArray output, e.g. B=mexFunction(A).
 *
 * Copyright 2012 The MathWorks, Inc.
 */

#include "mex.h"
#include "gpu/mxGPUArray.h"

/*
 * Device code
 */
void __global__ TimesTwo(double const * const A,
                         double * const B,
                         int const N)
{
    /* Calculate the global linear index, assuming a 1-d grid. */
    int const i = blockDim.x * blockIdx.x + threadIdx.x;
    if (i < N ) {
        if( i == 4 )
             B[i] = A[i];
        else
            B[i] = 2 * A[i];
    }
}

/*
 * Host code
 */
void mexFunction(int nlhs, mxArray *plhs[],
                 int nrhs, mxArray const *prhs[])
{
    /* Declare all variables.*/
    mxGPUArray const *A;
    mxGPUArray *B;
    double const *d_A;
    double *d_B;
    int N;
    char const * const errId = "parallel:gpu:mexGPUExample:InvalidInput";
    char const * const errMsg = "Invalid input to MEX file.";

    /* Choose a reasonably sized number of threads for the block. */
    int const threadsPerBlock = 256;
    int blocksPerGrid;

    /* Initialize the MathWorks GPU API. */
    mxInitGPU();

    /* Throw an error if the input is not a GPU array. */
    if ((nrhs!=1) || !(mxIsGPUArray(prhs[0]))) {
        mexErrMsgIdAndTxt(errId, errMsg);
    }

    A = mxGPUCreateFromMxArray(prhs[0]);

    /*
     * Verify that A really is a double array before extracting the pointer.
     */
    if (mxGPUGetClassID(A) != mxDOUBLE_CLASS) {
        mexErrMsgIdAndTxt(errId, errMsg);
    }

    /*
     * Now that we have verified the data type, extract a pointer to the input
     * data on the device.
     */
    d_A = (double const *)(mxGPUGetDataReadOnly(A));

    /* Create a GPUArray to hold the result and get its underlying pointer. */
    B = mxGPUCreateGPUArray(mxGPUGetNumberOfDimensions(A),
                            mxGPUGetDimensions(A),
                            mxGPUGetClassID(A),
                            mxGPUGetComplexity(A),
                            MX_GPU_DO_NOT_INITIALIZE);
    d_B = (double *)(mxGPUGetData(B));

    /*
     * Call the kernel using the CUDA runtime API. We are using a 1-d grid here,
     * and it would be possible for the number of elements to be too large for
     * the grid. For this example we are not guarding against this possibility.
     */
    N = (int)(mxGPUGetNumberOfElements(A));
    blocksPerGrid = (N + threadsPerBlock - 1) / threadsPerBlock;
    TimesTwo<<<blocksPerGrid, threadsPerBlock>>>(d_A, d_B, N);

    /* Wrap the result up as a MATLAB gpuArray for return. */
    plhs[0] = mxGPUCreateMxArrayOnGPU(B);

    /*
     * The mxGPUArray pointers are host-side structures that refer to device
     * data. These must be destroyed before leaving the MEX function.
     */
    mxGPUDestroyGPUArray(A);
    mxGPUDestroyGPUArray(B);
}

>> y = mexGPUExample(x)


y =


     2     1     2     2
     2     2     2     2
     2     2     2     2
     2     2     2     2