I have hard time analyzing my data and it normally takes me 13 hours to analyze it. The data is a 14Gb Hijing root file…

I was looking at the tutorials … icore.html

But none of these tutorial deals with a simple “for loop” multithreading. I just want to loop over the data, and save the results in a vector for analyzing it later with some functions, and plot it to the histogram. (It’s only the loop that take all computing time )

Right now my code produce a map< double,vector<double> >, which contains max 2000 keys and 10 to 20 “double" values for each keys, and it is analyzed in different functions. And all the functions produce 5 histograms, which is plottet on top of each other.

I have uploaded the code, and I have marked the part that should be multithreaded.

Hope someone have a solution thanks.
ep.cpp (8.38 KB)


I understand you want to want to parallelise this portion of the code*.
Now, it is not clear what the MCP class is and what the MCP_Intersection function actually does but what I think it’s on your plate is:

  • to figure our what is the input and the output of the calculations in the body of the loop
  • figure out if there are side effects in the afore mentioned class and method in order to decide for a multithreaded or multiprocess approach
  • match the outcome of the analysis of the previous two points to one of the cases illustrated in the tutorials, for example understanding if a multithreaded approach is possible (no side effects) or a multiprocess approach is needed.

Note that a reformulation of the calculations in the body of the for loop might be needed in order to fit the map or map/reduce paradigms proposed by the TProcPool class.

Sorry, no magic when it comes to expressing parallelism in real life.


                    for (unsigned int i = 0; i < position_array_quartz_T0A.size(); i++){ // Search intersection for each detectors
                        // For T0C
                        MCP_detector_T0C = new MCP(nameC, Mcp_dimension, position_array_quartz_T0C[i], Quartz_Dimension, number_of_sub_quartz,space_between_the_sub_quartz,Reflective_index,$
                        MCP_detector_T0C->MCP_Intersection(in, IP, MCP_detector_T0C, Intersection_plot_T0C, Output, pore_positions, particles_intersection_mcp,0);

                        // For T0A
                        MCP_detector_T0A = new MCP(nameA, Mcp_dimension, position_array_quartz_T0A[i], Quartz_Dimension, number_of_sub_quartz,space_between_the_sub_quartz,Reflective_index,$
                        MCP_detector_T0A->MCP_Intersection(in, IP, MCP_detector_T0A, Intersection_plot_T0A, Output, pore_positions, particles_intersection_mcp,0);
                        delete MCP_detector_T0C;
                        delete MCP_detector_T0A;}

Hi Danilo,

Thank you I will try to reformulate the code so it fits TProcPool class and hope it works.