多核编程A.pptVIP

* * * * Benchmarking parallel performance Measure the performance of a parallel application. How? Measuring the number of seconds that elapse from the time we initiate execution until the program terminates. double MPI_Wtime(void) Returns the numbers of seconds that have elapsed since some point of time in the past. double MPI_Wtick(void) Returns the precision of the result returned by MPI_Wtime. * Benchmarking parallel performance MPI_Barrier int MPI_Barrier(MPI_Comm comm) comm: indicate in which communicator the processes will participate the barrier synchronization. Function of MPI_Barrier is…. * For example Send and receive operation * For example Compute pi * For example 0 1 4 * For example Compute pi * For example Matrix Multiplication * * where A is an n x l matrix and B is an l x m matrix. * * * * * Summary MPI is a Library. Six foundational functions of MPI. collective communication. MPI communication Model. Fell free to contact me via yjf@ for any questions or suggestions. And Welcome to Wuhan University! Thanks! * Inherent concurrency. * Non-blocking Synchronous Send int MPI_Issend (void *buf, int count, MPI_Datatype datatype, int dest, int tag, MPI_Comm comm, MPI_Request *request) IN = provided by programmer, OUT = set by routine buf: starting address of message buffer (IN) count: number of elements in message (IN) datatype: type of elements in message (IN) dest: rank of destination task in communicator comm (IN) tag: message tag (IN) comm: communicator (IN) request: identifies a communication event (OUT)  * Non-blocking Ready Send int MPI_Irsend (void *buf, int count, MPI_Datatype datatype, int dest, int tag, MPI_Comm comm, MPI_Request *request) * Non-blocking Buffered Send int MPI_Ibsend (void *buf, int count, MPI_Datatype datatype, int dest, int tag, MPI_Comm comm, MPI_Request *request) * Non-blocking Standard Send int MPI_Isend (void *buf, int count, MPI_Datatype datatype, int dest, int tag, MPI_Comm comm, MPI_Request *request) * No