int bY(const int iD, int *piF){
//Y -> A.A
int bSucces, iM, iF;
bSucces = bA(iD,&iM) && bA(iM,&iF);
*piF = (bSucces) ? iF:iD;
return bSucces;
}
int bS(const int iD, int *piF){
//S->AB
int bSucces, iM, iF;
bSucces = bA(iD,&iM) && bB(iM,&iF);
*piF = (bSucces) ? iF:iD;
return bSucces;
}
int main(){
std::cout << "Pointer: starting unit test\n";
// int
Pointer<int> a(new int(100));
Pointer<int> b(new int(200));
if(a.count() != 1) std::cout << "Pointer: unit test 1.0 failing\n";
if(b.count() != 1) std::cout << "Pointer: unit test 1.1 failing\n";
if(*a != 100) std::cout << "Pointer: unit test 1.2 failing\n";
if(*b != 200) std::cout << "Pointer: unit test 1.3 failing\n";
Pointer<int> c(a);
if(a.count() != 2) std::cout << "Pointer: unit test 1.4 failing\n";
if(c.count() != 2) std::cout << "Pointer: unit test 1.5 failing\n";
if(*a != 100) std::cout << "Pointer: unit test 1.6 failing\n";
if(*c != 100) std::cout << "Pointer: unit test 1.7 failing\n";
if(a.get() != c.get()) std::cout << "Pointer: unit test 1.8 failing\n";
IPointer<int> *d = new Pointer<int>(a);
if(a.count() != 3) std::cout << "Pointer: unit test 1.9 failing\n";
if(**d !=100) std::cout << "Pointer: unit test 1.10 failing\n";
delete d;
if(a.count() != 2) std::cout << "Pointer: unit test 1.11 failing\n";
c = b;
if(a.count() != 1) std::cout << "Pointer: unit test 1.12 failing\n";
if(b.count() != 2) std::cout << "Pointer: unit test 1.13 failing\n";
if(c.count() != 2) std::cout << "Pointer: unit test 1.14 failing\n";
if(*a != 100) std::cout << "Pointer: unit test 1.15 failing\n";
if(*b != 200) std::cout << "Pointer: unit test 1.16 failing\n";
if(*c != 200) std::cout << "Pointer: unit test 1.17 failing\n";
if(b.get() != c.get()) std::cout << "Pointer: unit test 1.18 failing\n";
a = b;
if(a.count() != 3) std::cout << "Pointer: unit test 1.19 failing\n";
if(b.count() != 3) std::cout << "Pointer: unit test 1.20 failing\n";
if(c.count() != 3) std::cout << "Pointer: unit test 1.21 failing\n";
if(*a != 200) std::cout << "Pointer: unit test 1.22 failing\n";
if(*b != 200) std::cout << "Pointer: unit test 1.23 failing\n";
if(*c != 200) std::cout << "Pointer: unit test 1.24 failing\n";
if(b.get() != c.get()) std::cout << "Pointer: unit test 1.25 failing\n";
if(b.get() != a.get()) std::cout << "Pointer: unit test 1.25 failing\n";
// A
Pointer<A> aA(new A(10));
Pointer<A> bA(new A(20));
if(aA.count() != 1) std::cout << "Pointer: unit test 2.0 failing\n";
if(bA.count() != 1) std::cout << "Pointer: unit test 2.1 failing\n";
if(foo(*aA) != 10) std::cout << "Pointer: unit test 2.2 failing\n";
if(foo(*bA) != 20) std::cout << "Pointer: unit test 2.3 failing\n";
if(aA->get() != 10) std::cout << "Pointer:unit test 2.4 failing\n";
if(bA->get() != 20) std::cout << "Pointer::unit test 2.5 failing\n";
Pointer<A> cA(aA);
if(aA.count() != 2) std::cout << "Pointer: unit test 2.6 failing\n";
if(cA.count() != 2) std::cout << "Pointer: unit test 2.7 failing\n";
if(foo(*aA) != 10) std::cout << "Pointer: unit test 2.8 failing\n";
if(foo(*cA) != 10) std::cout << "Pointer: unit test 2.9 failing\n";
if(aA->get() != 10) std::cout << "Pointer:unit test 2.10 failing\n";
if(cA->get() != 10) std::cout << "Pointer::unit test 2.11 failing\n";
if(aA.get() != cA.get()) std::cout << "Pointer: unit test 2.12 failing\n";
IPointer<A> *dA = new Pointer<A>(aA);
if(aA.count() != 3) std::cout << "Pointer: unit test 2.13 failing\n";
if(foo(**dA) != 10) std::cout << "Pointer: unit test 2.14 failing\n";
if((*dA)->get() != 10) std::cout << "Pointer: unit test 2.15 failing\n";
delete dA;
if(aA.count() != 2) std::cout << "Pointer: unit test 2.16 failing\n";
cA = bA;
if(aA.count() != 1) std::cout << "Pointer: unit test 2.17 failing\n";
if(bA.count() != 2) std::cout << "Pointer: unit test 2.18 failing\n";
if(cA.count() != 2) std::cout << "Pointer: unit test 2.19 failing\n";
if(foo(*aA) != 10) std::cout << "Pointer: unit test 2.20 failing\n";
if(foo(*bA) != 20) std::cout << "Pointer: unit test 2.21 failing\n";
if(foo(*cA) != 20) std::cout << "Pointer: unit test 2.22 failing\n";
if(aA->get() != 10) std::cout << "Pointer: unit test 2.23 failing\n";
if(bA->get() != 20) std::cout << "Pointer: unit test 2.24 failing\n";
if(cA->get() != 20) std::cout << "Pointer: unit test 2.25 failing\n";
if(bA.get() != cA.get()) std::cout << "Pointer: unit test 2.26 failing\n";
aA = bA;
if(aA.count() != 3) std::cout << "Pointer: unit test 2.27 failing\n";
if(bA.count() != 3) std::cout << "Pointer: unit test 2.28 failing\n";
if(cA.count() != 3) std::cout << "Pointer: unit test 2.29 failing\n";
if(foo(*aA) != 20) std::cout << "Pointer: unit test 2.30 failing\n";
if(foo(*bA) != 20) std::cout << "Pointer: unit test 2.31 failing\n";
if(foo(*cA) != 20) std::cout << "Pointer: unit test 2.32 failing\n";
if(aA->get() != 20) std::cout << "Pointer: unit test 2.33 failing\n";
if(bA->get() != 20) std::cout << "Pointer: unit test 2.34 failing\n";
if(cA->get() != 20) std::cout << "Pointer: unit test 2.35 failing\n";
if(bA.get() != cA.get()) std::cout << "Pointer: unit test 2.36 failing\n";
if(bA.get() != aA.get()) std::cout << "Pointer: unit test 2.37 failing\n";
std::cout << "Pointer: unit test complete\n";
return 0;
}
int main(int argc, char** argv)
{
if(argc != 4){
std::cout << "Error! Three arguments m, n and p are needed!" << std::endl;
return 1;
}
int m = atoi(argv[1]);
int n = atoi(argv[2]);
int p = atoi(argv[3]);
double elapseTime, elapseTime_single;
int myrank, numprocs;
MPI_Status status;
MPI_Init(&argc, &argv); // 并行开始
//std::cout << "Start to computing..." << std::endl;
MPI_Comm_size(MPI_COMM_WORLD, &numprocs); // 获取进程数
MPI_Comm_rank(MPI_COMM_WORLD, &myrank); // 获取本进城ID号
if(numprocs < 2){
std::cout << "Error! There must be more than two processors." << std::endl;
return 1;
}
int bm = m / numprocs;
int bn = n / numprocs;
SCMatrix A(1), B(1), C(1), C_true(1);
SCMatrix bA(bm,p), bB_send(bn,p), bB_recv(bn,p), bC(bm, bn), bC_send(bm, n);
if(myrank == 0){
SCMatrix A1(m,p), B1(n,p), C1(m,n), C_true1(m,n);
A = A1; B = B1; C = C1; C_true = C_true1;
A.init_with_RV();
B.init_with_RV();
}
MPI_Barrier(MPI_COMM_WORLD);
//A.print(); B.print();
// start to timing
clock_t start_time = clock();
//clock_t t_temp_start = clock();
MPI_Scatter(&A, bm * p, MPI_DOUBLE, &bA, bm * p, MPI_DOUBLE, 0, MPI_COMM_WORLD);
MPI_Scatter(&B, bn * p, MPI_DOUBLE, &bB_recv, bn * p, MPI_DOUBLE, 0, MPI_COMM_WORLD);
int sendTo = (myrank + 1) % numprocs;
int recvFrom = (myrank - 1 + numprocs) % numprocs;
int circle = 0;
bB_send = bB_recv;
do{
bC = bA * bB_recv.transpose();
int blocks_col = (myrank - circle + numprocs) % numprocs;
for(int i=0; i<bm; i++){
for(int j=0; j<bn; j++){
bC_send(i, blocks_col*bn + j) = bC(i, j);
}
}
if(myrank % 2 == 0){
bB_send = bB_recv;
MPI_Ssend(&bB_send, bn*p, MPI_DOUBLE, sendTo, circle, MPI_COMM_WORLD);
MPI_Recv(&bB_recv, bn*p, MPI_DOUBLE, recvFrom, circle, MPI_COMM_WORLD, &status);
}else{
MPI_Recv(&bB_recv, bn*p, MPI_DOUBLE, recvFrom, circle, MPI_COMM_WORLD, &status);
MPI_Ssend(&bB_send, bn*p, MPI_DOUBLE, sendTo, circle, MPI_COMM_WORLD);
bB_send = bB_recv;
}
circle++;
}while(circle < numprocs);
MPI_Barrier(MPI_COMM_WORLD);
MPI_Gather(&bC_send, bm * n, MPI_DOUBLE, &C, bm * n, MPI_DOUBLE, 0, MPI_COMM_WORLD);
if(myrank == 0){
int remainAStartId = bm * numprocs;
int remainBStartId = bn * numprocs;
for(int i=remainAStartId; i<m; i++){
for(int j=0; j<n; j++){
double temp=0;
for(int k=0; k<p; k++){
temp += A(i,k) * B(j,k);
}
C(i,j) = temp;
}
}
for(int i=0; i<remainAStartId; i++){
for(int j=remainBStartId; j<n; j++){
double temp = 0;
for(int k=0; k<p; k++){
temp += A(i,k)*B(j,k);
}
C(i,j) = temp;
}
}
}
// end timing
clock_t end_time = clock();
elapseTime = (double)(end_time-start_time) / CLOCKS_PER_SEC;
if(myrank == 0){
//std::cout << "[P_0] m = " << m << ", np = " << numprocs << ", time cost: " << elapseTime << std::endl;
std::cout << "[P_0] Totally cost " << elapseTime << " seconds in parallel procedure." << std::endl;
start_time = clock();
C_true = A * B.transpose();
end_time = clock();
elapseTime_single = double(end_time-start_time) / CLOCKS_PER_SEC;
std::cout << "[P_0] Totally cost " << elapseTime_single << " seconds with one processor." << std::endl;
bool b = (C==C_true);
if(b){
std::cout << "[P_0] Congradulations! The two results are equal." << std::endl;
}else{
std::cout << "[P_0] Bad news! The two results do not match." << std::endl;
A.print();
B.print();
C.print();
C_true.print();
}
}
MPI_Finalize(); // 并行结束
return 0;
}
