Warning: file_get_contents(/data/phpspider/zhask/data//catemap/6/cplusplus/135.json): failed to open stream: No such file or directory in /data/phpspider/zhask/libs/function.php on line 167

Warning: Invalid argument supplied for foreach() in /data/phpspider/zhask/libs/tag.function.php on line 1116

Notice: Undefined index: in /data/phpspider/zhask/libs/function.php on line 180

Warning: array_chunk() expects parameter 1 to be array, null given in /data/phpspider/zhask/libs/function.php on line 181
C++ Boost mpi挂起_C++_Boost_Mpi - Fatal编程技术网

C++ Boost mpi挂起

C++ Boost mpi挂起,c++,boost,mpi,C++,Boost,Mpi,我有一个简单的异步消息传递程序,如下所示。我在两台有16个进程的电脑上运行它。这些过程通过torrus作为4x4矩阵连接。因此,在主函数的开头,您将看到谁是流程的邻居。我想做的是实现一个检查点算法。因此,当i=5或10(假设i是时间)时,每个进程都会向其所有邻居发送一条消息。当一个进程收到消息时,它会为同一进程启动一个新的irecv。但如果我10岁,它不会启动irecv,因为在那之后不会收到任何消息。在程序结束时,所有进程都等待未接收的消息 /* Demo_01_Main.cpp */ #in

我有一个简单的异步消息传递程序,如下所示。我在两台有16个进程的电脑上运行它。这些过程通过torrus作为4x4矩阵连接。因此,在主函数的开头,您将看到谁是流程的邻居。我想做的是实现一个检查点算法。因此,当i=5或10(假设i是时间)时,每个进程都会向其所有邻居发送一条消息。当一个进程收到消息时,它会为同一进程启动一个新的irecv。但如果我10岁,它不会启动irecv,因为在那之后不会收到任何消息。在程序结束时,所有进程都等待未接收的消息

/* Demo_01_Main.cpp */

#include <boost/mpi.hpp>
#include <boost/serialization/string.hpp>
#include <string>
#include <iostream>
#include <fstream>
#include <map>

using namespace std;

class Packet{
    friend class boost::serialization::access;
private:
    int receiver;
    int sender;
    int data;

public:
    Packet(){
        receiver = 0;
        sender = 0;
        data = 0;
    }

    Packet(int receiver, int sender, int data){
        this->receiver = receiver;
        this->sender = sender;
        this->data = data;
    }

    ~Packet(){}

    int getData() {
        return data;
    }

    void setData(int data) {
        this->data = data;
    }

    int getReceiver() {
        return receiver;
    }

    void setReceiver(int receiver) {
        this->receiver = receiver;
    }

    int getSender() {
        return sender;
    }

    void setSender(int sender) {
        this->sender = sender;
    }

    template<class Archive>
    void serialize(Archive& ar, const unsigned int version) {
        ar & receiver;
        ar & sender;
        ar & data;
    }

    string toString(){
        stringstream ss;
        ss << "Packet = [Data: " << data << ", Receiver: " << receiver << ", Sender: " << sender << "]";
        return ss.str();
    }
};

int rank;

void log(string str){
    ofstream outfile;
    stringstream logFileName;
    logFileName << "log_" << rank << ".txt";
    outfile.open(logFileName.str().c_str(), std::ios_base::app);
    outfile << str;
    outfile.close();
}

int main(int argc, char* argv[]){
    map<int, boost::mpi::request> mpiReceiveRequest;
    map<int, boost::mpi::request> mpiSendRequest;
    map<int, Packet *> receivedData;
    vector<int> neighbors;

    boost::mpi::environment env(argc, argv);
    boost::mpi::communicator world;

    rank = world.rank();

    if(rank == 0){
        neighbors.push_back(1);
        neighbors.push_back(3);
        neighbors.push_back(4);
        neighbors.push_back(5);
        neighbors.push_back(7);
        neighbors.push_back(12);
        neighbors.push_back(13);
        neighbors.push_back(15);
    }
    else if(rank == 1){
        neighbors.push_back(0);
        neighbors.push_back(2);
        neighbors.push_back(4);
        neighbors.push_back(5);
        neighbors.push_back(6);
        neighbors.push_back(12);
        neighbors.push_back(13);
        neighbors.push_back(14);
    }
    else if(rank == 2){
        neighbors.push_back(1);
        neighbors.push_back(3);
        neighbors.push_back(5);
        neighbors.push_back(6);
        neighbors.push_back(7);
        neighbors.push_back(13);
        neighbors.push_back(14);
        neighbors.push_back(15);
    }
    else if(rank == 3){
        neighbors.push_back(0);
        neighbors.push_back(2);
        neighbors.push_back(4);
        neighbors.push_back(6);
        neighbors.push_back(7);
        neighbors.push_back(12);
        neighbors.push_back(14);
        neighbors.push_back(15);
    }
    else if(rank == 4){
        neighbors.push_back(0);
        neighbors.push_back(1);
        neighbors.push_back(3);
        neighbors.push_back(5);
        neighbors.push_back(7);
        neighbors.push_back(8);
        neighbors.push_back(9);
        neighbors.push_back(11);
    }
    else if(rank == 5){
        neighbors.push_back(0);
        neighbors.push_back(1);
        neighbors.push_back(2);
        neighbors.push_back(4);
        neighbors.push_back(6);
        neighbors.push_back(8);
        neighbors.push_back(9);
        neighbors.push_back(10);
    }
    else if(rank == 6){
        neighbors.push_back(1);
        neighbors.push_back(2);
        neighbors.push_back(3);
        neighbors.push_back(5);
        neighbors.push_back(7);
        neighbors.push_back(9);
        neighbors.push_back(10);
        neighbors.push_back(11);
    }
    else if(rank == 7){
        neighbors.push_back(0);
        neighbors.push_back(2);
        neighbors.push_back(3);
        neighbors.push_back(4);
        neighbors.push_back(6);
        neighbors.push_back(8);
        neighbors.push_back(10);
        neighbors.push_back(11);
    }
    else if(rank == 8){
        neighbors.push_back(4);
        neighbors.push_back(5);
        neighbors.push_back(7);
        neighbors.push_back(9);
        neighbors.push_back(11);
        neighbors.push_back(12);
        neighbors.push_back(13);
        neighbors.push_back(15);
    }
    else if(rank == 9){
        neighbors.push_back(4);
        neighbors.push_back(5);
        neighbors.push_back(6);
        neighbors.push_back(8);
        neighbors.push_back(10);
        neighbors.push_back(12);
        neighbors.push_back(13);
        neighbors.push_back(14);
    }
    else if(rank == 10){
        neighbors.push_back(5);
        neighbors.push_back(6);
        neighbors.push_back(7);
        neighbors.push_back(9);
        neighbors.push_back(11);
        neighbors.push_back(13);
        neighbors.push_back(14);
        neighbors.push_back(15);
    }
    else if(rank == 11){
        neighbors.push_back(4);
        neighbors.push_back(6);
        neighbors.push_back(7);
        neighbors.push_back(8);
        neighbors.push_back(10);
        neighbors.push_back(12);
        neighbors.push_back(14);
        neighbors.push_back(15);
    }
    else if(rank == 12){
        neighbors.push_back(0);
        neighbors.push_back(1);
        neighbors.push_back(3);
        neighbors.push_back(8);
        neighbors.push_back(9);
        neighbors.push_back(11);
        neighbors.push_back(13);
        neighbors.push_back(15);
    }
    else if(rank == 13){
        neighbors.push_back(0);
        neighbors.push_back(1);
        neighbors.push_back(2);
        neighbors.push_back(8);
        neighbors.push_back(9);
        neighbors.push_back(10);
        neighbors.push_back(12);
        neighbors.push_back(14);
    }
    else if(rank == 14){
        neighbors.push_back(1);
        neighbors.push_back(2);
        neighbors.push_back(3);
        neighbors.push_back(9);
        neighbors.push_back(10);
        neighbors.push_back(11);
        neighbors.push_back(13);
        neighbors.push_back(15);
    }
    else if(rank == 15){
        neighbors.push_back(0);
        neighbors.push_back(2);
        neighbors.push_back(3);
        neighbors.push_back(8);
        neighbors.push_back(10);
        neighbors.push_back(11);
        neighbors.push_back(12);
        neighbors.push_back(14);
    }

    for(int i=0; i<8; i++){
        Packet * packet = new Packet();
        receivedData[neighbors[i]] = packet;
        mpiReceiveRequest[neighbors[i]] = world.irecv(neighbors[i], 100, *packet);
    }


    for(int i=1; i<=10; i++){
        if(i%5 == 0){    // Checkpoint time
            for(int j=0; j<8; j++){
                Packet * p = new Packet(neighbors[j], rank, i);
                mpiSendRequest[neighbors[j]] = world.isend(neighbors[j], 100, *p);
                log("Sending: ");
                log(p->toString());
                log("\n");
            }
        }

        for(int j=0; j<8; j++){
            if(mpiReceiveRequest[neighbors[j]].test()){
                Packet * p = receivedData[neighbors[j]];
                log("Received: ");
                log(receivedData[neighbors[j]]->toString());
                log("\n");

                if(p->getData() != 10){
                    Packet * packet = new Packet();
                    receivedData[neighbors[j]] = packet;
                    mpiReceiveRequest[neighbors[j]] = world.irecv(neighbors[j], 100, *packet);
                }
            }
        }
    }

    for(int i=0; i<8; i++){
        stringstream ss;
        ss << "  Wait from: " << neighbors[i] << endl;
        log(ss.str());
        mpiReceiveRequest[neighbors[i]].wait();

        log("Received: ");
        log(receivedData[neighbors[i]]->toString());
        log("\n");
    }

    stringstream ss;
    ss << rank << " is done" << endl;
    log(ss.str());

    return 0;
}

我想不出问题出在哪里。

问题出在为请求调用wait命令时,该请求在之前经过测试并已通过。事实上,我意识到这也是我前面一个问题的原因:

因此,我修改了我的代码如下,现在它运行:

/* ...
 else if(rank == 15){
    neighbors.push_back(0);
    neighbors.push_back(2);
    neighbors.push_back(3);
    neighbors.push_back(8);
    neighbors.push_back(10);
    neighbors.push_back(11);
    neighbors.push_back(12);
    neighbors.push_back(14);
}
...After assigning neighbors
*/

for(int i=0; i<8; i++){
    Packet * packet = new Packet();
    receivedData[neighbors[i]] = packet;
    mpiReceiveRequest[i] = world.irecv(neighbors[i], 100, *packet);
}

vector<int> completed;
for(int i=1; i<=10; i++){
    if(i%5 == 0){    // Checkpoint time
        for(int j=0; j<8; j++){
            Packet * p = new Packet(neighbors[j], rank, i);
            mpiSendRequest.push_back(world.isend(neighbors[j], 100, *p));
            log("Sending: ");
            log(p->toString());
            log("\n");
        }
    }

    for(int j=0; j<8; j++){
        vector<int>::iterator it = completed.begin();
        bool passed = false;

        while(it != completed.end()){
            if(*it == j){
                passed = true;
                break;
            }
            it++;
        }

        if(!passed){
            if(mpiReceiveRequest[j].test()){
                completed.push_back(j);

                Packet * p = receivedData[neighbors[j]];
                if(p->getData() != 10){
                    Packet * packet = new Packet();
                    receivedData[neighbors[j]] = packet;
                    mpiReceiveRequest[j] = world.irecv(neighbors[j], 100, *packet);
                    completed.pop_back();
                }
            }
        }
    }
}

vector<boost::mpi::request> reqs;
for(int i=0; i<8; i++){
    vector<int>::iterator it = completed.begin();
    bool passed = false;

    while(it != completed.end()){
        if(*it == i){
            passed = true;
            break;
        }
        it++;
    }

    if(!passed){
        mpiReceiveRequest[i].wait();
        reqs.push_back(mpiReceiveRequest[i]);
    }
}

stringstream ss;
ss << rank << " is done" << endl;
log(ss.str());

return 0;
} // End of main
/*。。。
否则如果(秩==15){
邻居。推回(0);
邻居。推回(2);
邻居。推回(3);
邻居。推回(8);
邻居。推回(10);
邻居。推回(11);
邻居。推回(12);
邻居。推回(14);
}
…在分配邻居之后
*/

对于(int i=0;i而言,问题在于调用wait命令来处理之前测试并通过的请求。实际上,我意识到这也是我前面一个问题的原因:

因此,我修改了我的代码如下,现在它运行:

/* ...
 else if(rank == 15){
    neighbors.push_back(0);
    neighbors.push_back(2);
    neighbors.push_back(3);
    neighbors.push_back(8);
    neighbors.push_back(10);
    neighbors.push_back(11);
    neighbors.push_back(12);
    neighbors.push_back(14);
}
...After assigning neighbors
*/

for(int i=0; i<8; i++){
    Packet * packet = new Packet();
    receivedData[neighbors[i]] = packet;
    mpiReceiveRequest[i] = world.irecv(neighbors[i], 100, *packet);
}

vector<int> completed;
for(int i=1; i<=10; i++){
    if(i%5 == 0){    // Checkpoint time
        for(int j=0; j<8; j++){
            Packet * p = new Packet(neighbors[j], rank, i);
            mpiSendRequest.push_back(world.isend(neighbors[j], 100, *p));
            log("Sending: ");
            log(p->toString());
            log("\n");
        }
    }

    for(int j=0; j<8; j++){
        vector<int>::iterator it = completed.begin();
        bool passed = false;

        while(it != completed.end()){
            if(*it == j){
                passed = true;
                break;
            }
            it++;
        }

        if(!passed){
            if(mpiReceiveRequest[j].test()){
                completed.push_back(j);

                Packet * p = receivedData[neighbors[j]];
                if(p->getData() != 10){
                    Packet * packet = new Packet();
                    receivedData[neighbors[j]] = packet;
                    mpiReceiveRequest[j] = world.irecv(neighbors[j], 100, *packet);
                    completed.pop_back();
                }
            }
        }
    }
}

vector<boost::mpi::request> reqs;
for(int i=0; i<8; i++){
    vector<int>::iterator it = completed.begin();
    bool passed = false;

    while(it != completed.end()){
        if(*it == i){
            passed = true;
            break;
        }
        it++;
    }

    if(!passed){
        mpiReceiveRequest[i].wait();
        reqs.push_back(mpiReceiveRequest[i]);
    }
}

stringstream ss;
ss << rank << " is done" << endl;
log(ss.str());

return 0;
} // End of main
/*。。。
否则如果(秩==15){
邻居。推回(0);
邻居。推回(2);
邻居。推回(3);
邻居。推回(8);
邻居。推回(10);
邻居。推回(11);
邻居。推回(12);
邻居。推回(14);
}
…在分配邻居之后
*/

对于(int i=0;i通常,我不理解您的代码。但显而易见的是,您从不等待非阻塞发送请求,因此可能有些消息从未真正发送过,而您的代码块正在等待那些从未发送过的消息的非阻塞接收。您如何启动?我尝试了
mpirun-np 16./test
和类似程序,但所有进程都会中止。@HristoIliev我也尝试了send命令而不是isend,但结果是一样的。@sehe我也是这样启动这个程序的。老实说,我根本不懂你的代码。但很明显,你从不等待非阻塞的send请求请注意,可能有些消息从未真正发送过,而您的代码块正在等待那些从未发送过的消息的非阻塞接收。如何启动?我尝试了
mpirun-np 16./test
和类似的方法,但所有进程都会中止。@hristoilev我也尝试了send命令而不是isend,但结果是一样的。@我也是这样开始这个程序的。它根本不适合我