GPU与MATLAB混合编程之书本实现代码

#include "mex.h" #include <cuda_runtime.h> #include "register3D_cuda.h" // funciton [movVol_updated, Tx, Ty, Tz] = register3D(movVol, refVol, ... // movXGrad_Org, movYGrad_Org, movZGrad_Org, ... // Tx, Ty, Tz, ... // iterLimit, regulFactor, stepSize); void mexFunction(int nlhs, mxArray *plhs[], int nrhs, const mxArray *prhs[]) { if (nrhs != 11) mexErrMsgTxt("Invaid number of input arguments"); if (nlhs != 4) mexErrMsgTxt("Invalid number of outputs"); for (int i = 0; i < nrhs; ++i) { if (!mxIsDouble(prhs[i])) { mexErrMsgTxt("input vector data type must be double"); break; } } const mwSize* size = mxGetDimensions(prhs[0]); int sx = size[0]; int sy = size[1]; int sz = size[2]; // inputs double* movVol = (double*)mxGetData(prhs[0]); double* refVol = (double*)mxGetData(prhs[1]); double* movXGrad_Org = (double*)mxGetData(prhs[2]); double* movYGrad_Org = (double*)mxGetData(prhs[3]); double* movZGrad_Org = (double*)mxGetData(prhs[4]); double* TxIn = (double*)mxGetData(prhs[5]); double* TyIn = (double*)mxGetData(prhs[6]); double* TzIn = (double*)mxGetData(prhs[7]); double iterLimit = *((double*)mxGetData(prhs[8])); double regulFactor = *((double*)mxGetData(prhs[9])); double stepSize = *((double*)mxGetData(prhs[10])); // outputs plhs[0] = mxCreateNumericArray(3, size, mxDOUBLE_CLASS, mxREAL); double* movVol_updated = (double*)mxGetData(plhs[0]); plhs[1] = mxCreateNumericArray(3, size, mxDOUBLE_CLASS, mxREAL); double* Tx = (double*)mxGetData(plhs[1]); plhs[2] = mxCreateNumericArray(3, size, mxDOUBLE_CLASS, mxREAL); double* Ty = (double*)mxGetData(plhs[2]); plhs[3] = mxCreateNumericArray(3, size, mxDOUBLE_CLASS, mxREAL); double* Tz = (double*)mxGetData(plhs[3]); // register3D cuda register3D_cuda(movVol, refVol, movXGrad_Org, movYGrad_Org, movZGrad_Org, TxIn, TyIn, TzIn, movVol_updated, Tx, Ty, Tz, sx, sy, sz, (int)iterLimit, regulFactor, stepSize); cudaError_t error = cudaGetLastError(); if (error != cudaSuccess) { mexPrintf("%s\n", cudaGetErrorString(error)); mexErrMsgTxt("CUDA failed\n"); } }