Difference between revisions of "Multi 2"
(Created page with " Model na programirane s MPI ==компилиране == ==стартиране == == . ==") |
|||
(One intermediate revision by the same user not shown) | |||
Line 1: | Line 1: | ||
+ | Model na programirane s MPI | ||
+ | |||
+ | == Достп до клъстера == | ||
+ | |||
+ | == Достъп до БлуеГене/P == | ||
− | + | ==Използване на MPI под Windows == | |
+ | |||
+ | ==Какво е MPI == | ||
+ | |||
+ | ==стартиране и създаване на приложения == | ||
+ | |||
+ | |||
+ | == MPICH2 == | ||
+ | |||
+ | най-използвани библиотеки | ||
+ | |||
+ | == Примери == | ||
+ | 1. Compile your OpenMP application | ||
+ | g++ -fopenmp -o <output_file> <source_files> | ||
+ | |||
+ | 2. Compile your MPI application | ||
+ | mpicxx -o <output_file> <source_files> | ||
+ | |||
+ | 3. Compiling your MPI + OpenMP application | ||
+ | mpicxx -fopenmp -o <output_file> <source_files> | ||
+ | |||
+ | 4. Compile your MPI application with MPE support. This will generate a clog2 when you start the | ||
+ | executable | ||
+ | mpicxx -mpe=mpilog -o <output_file> <source_files> | ||
− | + | 5. Execute a MPI program | |
+ | mpiexec -n <number of nodes> ./<executable> <input parameters> | ||
− | + | 6. Login to one of the two 8 core Xeon processors | |
+ | ssh wn9 | ||
+ | ssh wn10 | ||
+ | 7. Target a specific node with the MPI application. The file in the below command must have the syntax <IP | ||
+ | address>:<number of MPI processes>. The IP address can be between 192.168.0.1 - 192.168.0.10. Computers from 192.168.0.1 | ||
+ | to 192.168.0.8 have two cores and the two remaining have 8 cores each. | ||
+ | mpiexec -machinefile <file name> -n <number of nodes> ./<executable> <input parameters> | ||
− | + | 8. nohup | |
+ | nohup <command> > out.txt 2>&1 & |
Latest revision as of 08:46, 11 April 2012
Model na programirane s MPI
Contents
Достп до клъстера
Достъп до БлуеГене/P
Използване на MPI под Windows
Какво е MPI
стартиране и създаване на приложения
MPICH2
най-използвани библиотеки
Примери
1. Compile your OpenMP application g++ -fopenmp -o <output_file> <source_files>
2. Compile your MPI application mpicxx -o <output_file> <source_files>
3. Compiling your MPI + OpenMP application mpicxx -fopenmp -o <output_file> <source_files>
4. Compile your MPI application with MPE support. This will generate a clog2 when you start the executable mpicxx -mpe=mpilog -o <output_file> <source_files>
5. Execute a MPI program mpiexec -n <number of nodes> ./<executable> <input parameters>
6. Login to one of the two 8 core Xeon processors ssh wn9 ssh wn10
7. Target a specific node with the MPI application. The file in the below command must have the syntax <IP address>:<number of MPI processes>. The IP address can be between 192.168.0.1 - 192.168.0.10. Computers from 192.168.0.1 to 192.168.0.8 have two cores and the two remaining have 8 cores each. mpiexec -machinefile <file name> -n <number of nodes> ./<executable> <input parameters>
8. nohup nohup <command> > out.txt 2>&1 &