// COMP/XJCO3221 Parallel Computation Coursework 2: MPI / Distributed Memory Parallelism.
// Includes.
Copyright By PowCoder代写 加微信 powcoder
// Standard includes.
#include
#include
#include
#include
// The MPI library.
#include
// The include file for this coursework. Do not change the routines in “cwk2_extra.h”, or avoid calling them,
// as theyt will be changed with different versions for the purposes of assessment.
#include “cwk2_extra.h”
int main( int argc, char *argv[] )
// Initialisation.
// Start MPI and get the rank of this process.
int numprocs, rank;
MPI_Init( &argc, &argv );
MPI_Comm_size( MPI_COMM_WORLD, &numprocs );
MPI_Comm_rank( MPI_COMM_WORLD, &rank );
// The matrix is of size NxN, and the vectors are of size N, where N is specified on the command line,
// so need to first get N. The function ‘getNFromCommandLine’ does this, and also checks that N is divisible
// by the number of processes. If it returns -1, something went wrong and MPI_Finalize() has already been called on all ranks.
int N = getNFromCommandLine( argc, argv, rank, numprocs ); // Do NOT modify this call or the function in cwk2_extra.h
if( N==-1 ) return EXIT_FAILURE;
// The number of matrix rows per process. Note we have checked by this point that N is divisible by numprocs.
int rowsPerProc = N / numprocs;
// All processes use the full vector x, i.e. the one that is being multiplied. Allocate memory for it – no error check for simplicity.
float *x = (float*) malloc( N*sizeof(float) );
// All processes also have an equal number of matrix rows, and a portion of the solution vector, so allocate memory for these as well.
float *A_perProc = (float*) malloc( N * rowsPerProc * sizeof(float) );
float *b_perProc = (float*) malloc( rowsPerProc * sizeof(float) );
// Allocate the full matrix and the full solution vector on rank 0 only. Also initialise A and x with values.
float *A = NULL, *b = NULL;
if( rank==0 ) initialiseMatrixAndSolution(&A,&b,x,N); // Do NOT modify this call or the function in cwk2_extra.h
// Start the timing.
double startTime = MPI_Wtime();
// Perform matrix-vector multiplication in parallel.
// Your solution should go here.
// Check the answer on rank 0 in serial. Also output the result of the timing.
if( rank==0 )
// Time for the parallel computation.
printf( “Parallel calculation took a total time %g s.\n”, MPI_Wtime() – startTime );
// Call the check routine “checkAgainstSerial()” in cwk2_extra.h, which prints messages explaining if the parallel calculation
// matches the serial one, to within floating-point precision.
checkAgainstSerial(A,x,b,N); // Do NOT modify this call or the function in cwk2_extra.h
// Finalise.
// Free up all memory allocated for x, A_perProc and b_perProc, and (on rank 0 only) A and b.
freeAllMemory(rank,A,b,x,A_perProc,b_perProc);
// Tell MPI to free up resources.
MPI_Finalize();
程序代写 CS代考 加微信: powcoder QQ: 1823890830 Email: powcoder@163.com