About your machine

You use a supercomputer SGI CRAY Origin2000. It has 256 processors, each clock speed is 400MHz and the size of the memory locally connected to the each processor is 1GB. The primary cache is 32KB two-way set associative on chip and the secondary cache is 8MB two-way set associative. Each team can use 64 processors and 64 GB memory for the contest at maximum.

Though the memory processors of the supercomputer are logically shared, you had better think that each memory processor is distributed when you write a program in C languge with MPI libraries.

The host name of the supercomputer is mpp.cc.titech.ac.jp.

The operating system running on the supercomputer is a IRIX, which is a dialect of 64bit UNIX. We give you a short introduction of UNIX on this page.

Parallel Programming with MPI

You compile a MPI program with the following command: 
  mpp% cc -64 program.c -lmpi
then run the execution code: 
  mpp% mpirun -np 4 ./a.out
Here 'a.out' is a default name of the execution code. The option '-np' specifies the number of processors on which the same MPI program, program.c here, will runs. MPI program assings a rank, which value is an integer, to each processor. The root processor has the rank 0. The rest of the processors have posivitive rank starting at 1. So you control parallel execution with these ranks.

We also give MPI examples in this page.