MPI_Gather 收集双重向量std::vector＜std::vector＜double＞＞

#include <mpi.h>
#include<vector>
#include<iostream>

void t13(int argumentcount, char* argumentvector[]) {
	int procRank, procSize;
	int namelen;

	char processor_name[MPI_MAX_PROCESSOR_NAME];
	MPI_Init(&argumentcount, &argumentvector);
	MPI_Comm_rank(MPI_COMM_WORLD, &procRank);
	MPI_Comm_size(MPI_COMM_WORLD, &procSize);
	MPI_Get_processor_name(processor_name, &namelen);
	MPI_Status status;

	int nSamples = 100, numRP = 3;
	int nPart = nSamples;
	if (procSize >= 1)
		nPart = nSamples / procSize + 1;
	std::cout << "nPart=" << nPart << std::endl;

	std::vector<std::vector<double>> report_params_values_all_procs(numRP);
	for (size_t c = 0; c < report_params_values_all_procs.size(); ++c) {
		report_params_values_all_procs[c].resize(nPart*procSize);
	}
	

	std::vector<std::vector<double>> report_params_values_one_proc(numRP);
	for (size_t c = 0; c < report_params_values_one_proc.size(); ++c) {
		report_params_values_one_proc[c].resize(nPart);
	}

	for (int s = 0; s < nPart; ++s) {
		for (size_t c = 0; c < numRP; ++c) {		
			report_params_values_one_proc[c][s] = procRank + s + c*0.01;
		}
	}

	for (size_t c = 0; c < numRP; ++c)
		MPI_Gather(report_params_values_one_proc[c].data(),                 
                   report_params_values_one_proc[c].size()*sizeof(double), 
                   MPI_BYTE,
		           report_params_values_all_procs[c].data(),                                                                 
                   report_params_values_one_proc[c].size()*sizeof(double),
                   MPI_BYTE, 0, MPI_COMM_WORLD);

	if (procRank == 0) {
		for (int s = 0; s < nPart*procSize; ++s) {
			std::cout << "id=" << s << ", ";
			for (size_t c = 0; c < numRP; ++c) {
				std::cout << report_params_values_all_procs[c][s] << ", ";
			}
			std::cout << std::endl;
		}
	}
}

MPI_Gather收集双重向量std::vector<std::vector<double>>，在这里是在内层数据收集的，把各个进程的内层向量收集到进程0

编译后在命令行执行mpiexec.exe -n 4 ./t13.exe ，这个指令开了4个进程