Difference between revisions of "MPI"

From SourceWiki
Jump to navigation Jump to search
Line 22: Line 22:
 
The triple (address, count, datatype) provides a pattern.
 
The triple (address, count, datatype) provides a pattern.
  
Tags.  Can use these as a form of filtering: junk mail, bin; handwritten & perfumed, open now!; bill, open later!  Exercise, what happens if the tags don't match?
+
Tags.  Can use these as a form of filtering: junk mail, bin; handwritten & perfumed, open now!; bill, open later!   
  
 
==Synchronisation, Blocking and the role of Buffers==
 
==Synchronisation, Blocking and the role of Buffers==
Line 31: Line 31:
 
MPI_Recv() only returns when the message has been received, however.  Hence the term blocking.  
 
MPI_Recv() only returns when the message has been received, however.  Hence the term blocking.  
  
===A Common Bug===
+
==A Common Bug==
  
 
If all processes are waiting to receive prior to sending, then we will have deadlock.
 
If all processes are waiting to receive prior to sending, then we will have deadlock.
 +
 +
==Exercises==
 +
 +
What happens if the tags don't match?  (ans. deadlock)
  
 
=Non-Blocking Communication=
 
=Non-Blocking Communication=

Revision as of 14:28, 19 July 2010

MPI: Message passing for distributed memory computing

Introduction

MPI-1[ref]--the first incarnation of the standard--arrived in 1994 in response to the need for a portable means to program the growing number of distributed memory computers appearing in the marketplace. MPI stands for Message Passing Interface, and as its name suggests, it is an API, rather than a new programming language. At the time of writing, MPI can be used in C, C++, Fortran-77 and Fortran-90/95 programs. We will see that MPI-1 contained little on the topic I/O. This was rectified in 1997 with the arrival of MPI-2[ref], which contained the MPI-IO standard (supporting parallel I/O) along with additional functionality to support the dynamic creation of processes and also one-sided communication models.

We can extend Flynn's original Taxonomy[ref] with the acronym SPMD--Single Program Multiple Data. This emphasises the fact that using e.g. MPI, we can write single programs that will execute on computers comprised of multiple compute elements, each with its own--not shared--memory space.

Hello World

The quintessential start. Programs in C, Fortran-77 and Fortran-90.

MPI_Init() and MPI_Finalise(). MPI_Comm_Size() and MPI_Comm_rank(), which use the communicator MPI_COMM_WORLD.

These programs assume that all processes can write to the screen. This is not a safe assumption.

Send and Receive

The triple (address, count, datatype) provides a pattern.

Tags. Can use these as a form of filtering: junk mail, bin; handwritten & perfumed, open now!; bill, open later!

Synchronisation, Blocking and the role of Buffers

Independent 'compute elements' Synchronised communication requires that both sender and receiver are ready. Through the introduction of a buffer, a sender can deposit a message before the receiver is ready. MPI_Recv() only returns when the message has been received, however. Hence the term blocking.

A Common Bug

If all processes are waiting to receive prior to sending, then we will have deadlock.

Exercises

What happens if the tags don't match? (ans. deadlock)

Non-Blocking Communication

Latency Hiding

Collective Communications

Load Balancing