MPI message passing interface
Download as PPTX, PDF3 likes1,813 views
This document discusses implementing a parallel merge sort algorithm using MPI (Message Passing Interface). It describes the background of MPI and how it can be used for communication between processes. It provides details on the dataset used, MPI functions for initialization, communication between processes, and summarizes the results which show a decrease in runtime when increasing the number of processors.
MPI message passing interface
- 1. MPI - MESSAGE PASSING INTERFACE An approach for parallel algorithms COMPUTER SCIENCE DEPARTMENT PROF. AKHTAR RASOOL UNDERTHE GUIDANCE OF
- 2. TABLE OF CONTENT Parallel Merge Sort using MPI Introduction Background MPI Code implementation Result About Us Goal and Vision Target Programming language Cluster MPI library Dataset What is MPI Data type and syntax Communication Mode Features Merge sort implementation Flow of code Process flow Single processor Multiprocessor Summery
- 3. About Us
- 4. Our Team Nishaant Sharma 131112226 Kartik 131112265 Mohit Raghuvanshi 131112232 Prabhash Prakash 131112241
- 5. “Analysis of RunningTime complexity of Sorting Algorithm (Merge Sort) on parallel processing environment (preferably on a cluster with various processors) and single node Using MPI” Goal 05
- 6. Installation Implementation Result Install mpich2 on Ubantu Implement merge sort algorithm parallel Change in time when we increse in the number of processor VISION Write your relevant text here 06
- 7. Background Programming language Base of project good knowledge of “C” for implementing code C for implementing the algorithm Python for generating the data set Faker library in python 11
- 8. Background Cluster Master and Slave Node Master Node Slave Node 11
- 9. Background MPI Message passing interface MPI is a proposed standard message-passing interface. It is a library Specification, not a language. The programs that users can write in Fortran 77 andC are compiled with Ordinary compilers and linked with the MPI library. 11
- 10. Dataset Python Script For generating dataset We use faker module for generating the dataset of census. Some basic api of faker fake.name() for generating fake name fake.email() for generating fake email fake.ean(length=13) for unique id of 13 digit fake.job() for creating fake job Many other api 11
- 11. Dataset Python Script For generating dataset Dataset attribute Id Name Phone Number Salary Email 11
- 13. What is MPI MPI Message Passing Interface A message-passing library specifications: Extended message-passing model Not a language or compiler specification Not a specific implementation or product For parallel computers, clusters, and heterogeneous networks. Designed to permit the development of parallel software libraries. Designed to provide access to advanced parallel hardware for End users Library writers Tool developers 13
- 14. Communication Modes Based on the type of send: Synchronous: Completes once the acknowledgement is received by the sender. Buffered send: completes immediately, unless if an error occurs. Standard send: completes once the message has been sent, which may or may not imply that the message has arrived at its destination. Ready send: completes immediately, if the receiver is ready for the message it will get it, otherwise the message is dropped silently.
- 15. DataTypes The following data types are supported by MPI: Predefined data types that are corresponding to data types from the programming language. Arrays. Sub blocks of a matrix User defined data structure. A set of predefined data types
- 16. API of MPI #include “mpi.h” provides basic MPI definitions and types. MPI_Init starts MPI MPI_Finalize exits MPI Note that all non-MPI routines are local; thus “printf” run on each process
- 17. MPI_INIT Initializing MPI The initialization routine MPI_INIT is the first MPI routine called. MPI_INIT is called once. int MPI_INIT( int *argc, char **argv );
- 18. MPI_Finalize After a program has finished using the MPI library, it must call MPI_Finalize in order to clean up all MPI state. Once this routine is called, no MPI routine may be called. MPI_finalize is genrally last command in the program. MPI_finalize()
- 19. MPI_Comm_rank It updates the rank of processes in the variable world_rank. Determines the rank of the calling process in the communicator int MPI_Comm_rank(MPI_COMM_WORLD, &wordld_rank);
- 20. MPI_Comm_size Determines the size of the group associated with a communicator. int MPI_Comm_size(MPI_COMM comm, int *size);
- 21. MPI_Scatter Sends data from one process to all other processes in a communicator. . int MPI_Scatter(const void *sendbuf, int sendcount, MPI_Datatype sendtype,void *recvbuf, int recvcount, MPI_Datatype recvtype, int root, MPI_Comm comm);
- 22. MPI_Scatter
- 23. MPI_Gather MPI_Gather is opposite of MPI_Scatter, it gather all the data from other processes to single (Master) process. . int MPI_Gather(const void *sendbuf, intsendcount, MPI_Datatype sendtype, void *recvbuf, int recvcount, MPI_Datatype recvtype, int root, MPI_Comm comm)
- 24. MPI_Gather
- 25. MPI_Bcast Broadcasts a message from the process with rank "root" to all other processes of the communicator. int MPI_Bcast( void *buffer, int count, MPI_Datatype datatype, int root, MPI_Comm comm )
- 26. MPI_Bcast
- 27. Parallel Merge Sort Algorithm
- 28. Result No. of Process Time elapsed in reading input file(ms) Time elapsed to sort the data(ms) Time in writing output file(ms) 1 3982.326 8528.112 5839.587 2 4050.897 6878.000 5401.234 4 8145.740 12073.895 11083.125 5 10178.689 14361.952 13087.155
- 29. Conclusion In this project we focused on using MPI as a message passing interface implementation on LINUX platform. The effect of parallel processes number and also the number of cores on the performance of parallel merge sort algorithms has been theoretically and experimentally studied.
- 30. ThankYou