int main (int argc, char *argv[]) {
/*! structure to store all simulation parameters
*/
Config mcxconfig; /** mcxconfig: structure to store all simulation parameters */
GPUInfo *gpuinfo=NULL; /** gpuinfo: structure to store GPU information */
unsigned int activedev=0; /** activedev: count of total active GPUs to be used */
/**
To start an MCX simulation, we first create a simulation configuration and
set all elements to its default settings.
*/
mcx_initcfg(&mcxconfig);
/**
Then, we parse the full command line parameters and set user specified settings
*/
mcx_parsecmd(argc,argv,&mcxconfig);
/** The next step, we identify gpu number and query all GPU info */
if(!(activedev=mcx_list_gpu(&mcxconfig,&gpuinfo))){
mcx_error(-1,"No GPU device found\n",__FILE__,__LINE__);
}
#ifdef _OPENMP
/**
Now we are ready to launch one thread for each involked GPU to run the simulation
*/
omp_set_num_threads(activedev);
#pragma omp parallel
{
#endif
/**
This line runs the main MCX simulation for each GPU inside each thread
*/
mcx_run_simulation(&mcxconfig,gpuinfo);
#ifdef _OPENMP
}
#endif
/**
Once simulation is complete, we clean up the allocated memory in config and gpuinfo, and exit
*/
mcx_cleargpuinfo(&gpuinfo);
mcx_clearcfg(&mcxconfig);
return 0;
}
void mexFunction(int nlhs, mxArray *plhs[], int nrhs, const mxArray *prhs[]){
Config cfg;
GPUInfo *gpuinfo=NULL;
mxArray *tmp;
int ifield, jstruct;
int ncfg, nfields;
int fielddim[4];
int activedev=0;
int errorflag=0;
int threadid=0;
const char *outputtag[]={"data"};
const char *datastruct[]={"data","stat"};
const char *statstruct[]={"runtime","nphoton","energytot","energyabs","normalizer","workload"};
const char *gpuinfotag[]={"name","id","devcount","major","minor","globalmem",
"constmem","sharedmem","regcount","clock","sm","core",
"autoblock","autothread","maxgate"};
if (nrhs==0){
mcxlab_usage();
return;
}
if(nrhs==1 && mxIsChar(prhs[0])){
char shortcmd[MAX_SESSION_LENGTH];
mxGetString(prhs[0], shortcmd, MAX_SESSION_LENGTH);
shortcmd[MAX_SESSION_LENGTH-1]='\0';
if(strcmp(shortcmd,"gpuinfo")==0){
mcx_initcfg(&cfg);
cfg.isgpuinfo=3;
if(!(activedev=mcx_list_gpu(&cfg,&gpuinfo))){
mexWarnMsgTxt("no active GPU device found");
}
plhs[0] = mxCreateStructMatrix(gpuinfo[0].devcount,1,15,gpuinfotag);
for(int i=0;i<gpuinfo[0].devcount;i++){
mxSetField(plhs[0],i,"name",mxCreateString(gpuinfo[i].name));
SET_GPU_INFO(plhs[0],i,id);
SET_GPU_INFO(plhs[0],i,devcount);
SET_GPU_INFO(plhs[0],i,major);
SET_GPU_INFO(plhs[0],i,minor);
SET_GPU_INFO(plhs[0],i,globalmem);
SET_GPU_INFO(plhs[0],i,constmem);
SET_GPU_INFO(plhs[0],i,sharedmem);
SET_GPU_INFO(plhs[0],i,regcount);
SET_GPU_INFO(plhs[0],i,clock);
SET_GPU_INFO(plhs[0],i,sm);
SET_GPU_INFO(plhs[0],i,core);
SET_GPU_INFO(plhs[0],i,autoblock);
SET_GPU_INFO(plhs[0],i,autothread);
SET_GPU_INFO(plhs[0],i,maxgate);
}
mcx_cleargpuinfo(&gpuinfo);
mcx_clearcfg(&cfg);
}
return;
}
printf("Launching MCXLAB - Monte Carlo eXtreme for MATLAB & GNU Octave ...\n");
if (!mxIsStruct(prhs[0]))
mexErrMsgTxt("Input must be a structure.");
nfields = mxGetNumberOfFields(prhs[0]);
ncfg = mxGetNumberOfElements(prhs[0]);
if(nlhs>=1)
plhs[0] = mxCreateStructMatrix(ncfg,1,2,datastruct);
if(nlhs>=2)
plhs[1] = mxCreateStructMatrix(ncfg,1,1,outputtag);
if(nlhs>=3)
plhs[2] = mxCreateStructMatrix(ncfg,1,1,outputtag);
if(nlhs>=4)
plhs[3] = mxCreateStructMatrix(ncfg,1,1,outputtag);
if(nlhs>=5)
plhs[4] = mxCreateStructMatrix(ncfg,1,1,outputtag);
for (jstruct = 0; jstruct < ncfg; jstruct++) { /* how many configs */
try{
printf("Running simulations for configuration #%d ...\n", jstruct+1);
mcx_initcfg(&cfg);
for (ifield = 0; ifield < nfields; ifield++) { /* how many input struct fields */
tmp = mxGetFieldByNumber(prhs[0], jstruct, ifield);
if (tmp == NULL) {
continue;
}
mcx_set_field(prhs[0],tmp,ifield,&cfg);
}
#ifndef MATLAB_MEX_FILE
mexEvalString("fflush(stdout);");
#else
mexEvalString("drawnow;");
#endif
cfg.issave2pt=(nlhs>=1);
cfg.issavedet=(nlhs>=2);
cfg.issaveseed=(nlhs>=4);
#if defined(USE_MT_RAND)
cfg.issaveseed=0;
#endif
if(cfg.vol==NULL || cfg.medianum==0){
mexErrMsgTxt("You must define 'vol' and 'prop' field.");
}
if(!(activedev=mcx_list_gpu(&cfg,&gpuinfo))){
mexErrMsgTxt("No active GPU device found");
}
if(nlhs>=1){
int fieldlen=cfg.dim.x*cfg.dim.y*cfg.dim.z*(int)((cfg.tend-cfg.tstart)/cfg.tstep+0.5);
cfg.exportfield = (float*)calloc(fieldlen,sizeof(float));
}
if(nlhs>=2){
cfg.exportdetected=(float*)malloc((cfg.medianum+1)*cfg.maxdetphoton*sizeof(float));
}
if(nlhs>=4){
cfg.seeddata=malloc(cfg.maxdetphoton*sizeof(float)*RAND_BUF_LEN);
}
if(nlhs>=5){
cfg.exportdebugdata=(float*)malloc(cfg.maxjumpdebug*sizeof(float)*MCX_DEBUG_REC_LEN);
}
mcx_validate_config(&cfg);
#ifdef _OPENMP
omp_set_num_threads(activedev);
#pragma omp parallel shared(errorflag)
{
threadid=omp_get_thread_num();
#endif
try{
mcx_run_simulation(&cfg,gpuinfo);
}catch(const char *err){
mexPrintf("Error from thread (%d): %s\n",threadid,err);
errorflag++;
}catch(const std::exception &err){
mexPrintf("C++ Error from thread (%d): %s\n",threadid,err.what());
errorflag++;
}catch(...){
mexPrintf("Unknown Exception from thread (%d)",threadid);
errorflag++;
}
#ifdef _OPENMP
}
#endif
if(errorflag)
mexErrMsgTxt("MCXLAB Terminated due to an exception!");
if(nlhs>=5){
fielddim[0]=MCX_DEBUG_REC_LEN; fielddim[1]=cfg.debugdatalen; // his.savedphoton is for one repetition, should correct
fielddim[2]=0; fielddim[3]=0;
mxSetFieldByNumber(plhs[4],jstruct,0, mxCreateNumericArray(2,fielddim,mxSINGLE_CLASS,mxREAL));
if(cfg.debuglevel & MCX_DEBUG_MOVE)
memcpy((float*)mxGetPr(mxGetFieldByNumber(plhs[4],jstruct,0)),cfg.exportdebugdata,fielddim[0]*fielddim[1]*sizeof(float));
if(cfg.exportdebugdata)
free(cfg.exportdebugdata);
cfg.exportdebugdata=NULL;
}
if(nlhs>=4){
fielddim[0]=(cfg.issaveseed>0)*RAND_BUF_LEN*sizeof(float); fielddim[1]=cfg.detectedcount; // his.savedphoton is for one repetition, should correct
fielddim[2]=0; fielddim[3]=0;
mxSetFieldByNumber(plhs[3],jstruct,0, mxCreateNumericArray(2,fielddim,mxUINT8_CLASS,mxREAL));
memcpy((unsigned char*)mxGetPr(mxGetFieldByNumber(plhs[3],jstruct,0)),cfg.seeddata,fielddim[0]*fielddim[1]);
free(cfg.seeddata);
cfg.seeddata=NULL;
}
if(nlhs>=3){
fielddim[0]=cfg.dim.x; fielddim[1]=cfg.dim.y;
fielddim[2]=cfg.dim.z; fielddim[3]=0;
if(cfg.vol){
mxSetFieldByNumber(plhs[2],jstruct,0, mxCreateNumericArray(3,fielddim,mxUINT8_CLASS,mxREAL));
memcpy((unsigned char*)mxGetPr(mxGetFieldByNumber(plhs[2],jstruct,0)),cfg.vol,
fielddim[0]*fielddim[1]*fielddim[2]*sizeof(unsigned char));
}
}
if(nlhs>=2){
fielddim[0]=(cfg.medianum+1); fielddim[1]=cfg.detectedcount;
fielddim[2]=0; fielddim[3]=0;
if(cfg.detectedcount>0){
mxSetFieldByNumber(plhs[1],jstruct,0, mxCreateNumericArray(2,fielddim,mxSINGLE_CLASS,mxREAL));
memcpy((float*)mxGetPr(mxGetFieldByNumber(plhs[1],jstruct,0)),cfg.exportdetected,
fielddim[0]*fielddim[1]*sizeof(float));
}
free(cfg.exportdetected);
cfg.exportdetected=NULL;
}
if(nlhs>=1){
fielddim[0]=cfg.dim.x; fielddim[1]=cfg.dim.y;
fielddim[2]=cfg.dim.z; fielddim[3]=(int)((cfg.tend-cfg.tstart)/cfg.tstep+0.5);
mxSetFieldByNumber(plhs[0],jstruct,0, mxCreateNumericArray(4,fielddim,mxSINGLE_CLASS,mxREAL));
memcpy((float*)mxGetPr(mxGetFieldByNumber(plhs[0],jstruct,0)),cfg.exportfield,
fielddim[0]*fielddim[1]*fielddim[2]*fielddim[3]*sizeof(float));
free(cfg.exportfield);
cfg.exportfield=NULL;
mxArray *stat=mxCreateStructMatrix(1,1,6,statstruct);
mxArray *val = mxCreateDoubleMatrix(1,1,mxREAL);
*mxGetPr(val) = cfg.runtime;
mxSetFieldByNumber(stat,0,0, val);
val = mxCreateDoubleMatrix(1,1,mxREAL);
*mxGetPr(val) = cfg.nphoton;
mxSetFieldByNumber(stat,0,1, val);
val = mxCreateDoubleMatrix(1,1,mxREAL);
*mxGetPr(val) = cfg.energytot;
mxSetFieldByNumber(stat,0,2, val);
val = mxCreateDoubleMatrix(1,1,mxREAL);
*mxGetPr(val) = cfg.energyabs;
mxSetFieldByNumber(stat,0,3, val);
val = mxCreateDoubleMatrix(1,1,mxREAL);
*mxGetPr(val) = cfg.normalizer;
mxSetFieldByNumber(stat,0,4, val);
val = mxCreateDoubleMatrix(1,activedev,mxREAL);
for(int i=0;i<activedev;i++)
*(mxGetPr(val)+i) = cfg.workload[i];
mxSetFieldByNumber(stat,0,5, val);
mxSetFieldByNumber(plhs[0],jstruct,1, stat);
}
}catch(const char *err){
mexPrintf("Error: %s\n",err);
}catch(const std::exception &err){
mexPrintf("C++ Error: %s\n",err.what());
}catch(...){
mexPrintf("Unknown Exception");
}
if(detps)
free(detps);
mcx_cleargpuinfo(&gpuinfo);
mcx_clearcfg(&cfg);
}
return;
}