Index of /examples/mpi/tutorials/legacy/intro2mpi/F77

Name	Last modified	Size

Parent Directory		-
myinput	2013-11-22 16:10	4
Readme	2013-11-23 22:40	343
batch_scc	2013-11-23 11:50	1.3K
example1.f	2013-11-22 08:54	1.5K
sample.jcf	2013-11-22 08:54	1.5K
make.scc	2013-11-22 16:13	1.7K
batch_bgl	2013-11-22 08:54	1.8K
make.bgl	2013-11-22 08:54	2.5K
readme.html.detail	2013-11-25 09:15	2.6K
example1_4.f	2013-11-22 08:54	2.8K
example1_5.f	2013-11-22 08:54	3.2K
example1_1.f	2013-11-22 08:54	3.3K
example1_2.f	2013-11-22 08:54	3.4K
example1_52.f	2013-11-25 11:32	3.4K
example1_3.f	2013-11-22 08:54	3.7K

Multiprocessing by Message Passing MPI Tutorial Example Programs

This tutorial came with 6 examples. They are all based on the numerical integration of a cosine function over the range of [a, b]. A mid-point rule is used to perform the integration.

Example 1. Performs the integration serially. MPI is not used.
Example 1_1. Performs the same integral in parallel, with help from six most fundamental MPI utility functions/subroutines: MPI_Init, MPI_Comm_rank, MPI_Comm_size, MPI_Send, MPI_Recv, and MPI_Finalize.
Example 1_2. A more careful review of the previous algorithm exposes the fact that it is in fact a redundancy to send its share of local integral, my_int, to itself. By skipping over this, the deadlocking potential is eliminated.
Example 1_3. The use of MPI_ANY_SOURCE, essentially a wild card, enables messages sent to processor 0 (from all other processes) handled on a first come first served basis. This makes receiving of incoming messages more efficient.
Example1_4. In place of the MPI_Send and MPI_Recv pair of point-to-point communication functions, a collective communication function, MPI_Gather, will be used instead.
Example1_5. The simplest, and perhaps most efficient, parallel implementation for this numerical integration example is achieved with the use of the reduction operation function, MPI_Reduce.

Program Compilation

You can compile the set of examples for the Shared Computing Cluster (SCC) or the IBM Bluegene.

scc% make -f make.scc
lee% make -f make.bgl

Program Execution

scc% example1
scc% mpirun -np 4 example1_2

Notes

The examples are developed for the Linux system. However, they all follow the ANSI standard and hence should work for all operating systems.
To make compiling of an MPI-enabled code more convenient, a set of MPI wrapper scripts, such as mpicc, can be used in place of, say, gcc. The wrapper script includes the necessary MPI libraries and header files paths . On the SCC, these wrappers use the GNU compilers to build your executables. If Portland Group compilers are preferred, performs this before compilation
```
scc1:~ % setenv MPI_COMPILER pgi

scc1:~ % mpicc -o mycode mycode.c
```

Contact Info

Kadin Tseng (kadin@bu.edu)

References

Multiprocessing by Message Passing MPI Tutorial

Dates

Created: November 24, 2013
Modified: