| | 1 | [[PageOutline]] |
| | 2 | = Message Passing Interface; MPI = |
| | 3 | * A standardized and portable message-passing system |
| | 4 | * Not a new language ---- Library of functions |
| | 5 | * Scales to over 100k processors |
| | 6 | * Runs on distributed memory cluster systems. |
| | 7 | * Portable |
| | 8 | * Explicit parallel programming (unlike OpenMP) |
| | 9 | * No shared variables (unlike OpenMP) |
| | 10 | |
| | 11 | |
| | 12 | == MPI’s Execution model == |
| | 13 | * All processors execute same program, but with difference data. (no fork & join) |
| | 14 | * Variable on each are all private. |
| | 15 | * Data that must be shared is explicitly send between processors. |
| | 16 | |
| | 17 | [[Image(MPIexecModel.png)]] |
| | 18 | |
| | 19 | == Distributed Memory Cluster == |
| | 20 | [[Image(CypressNodes.png, 60%)]] |
| | 21 | |
| | 22 | Data sent over Network: |
| | 23 | |
| | 24 | * Ethernet |
| | 25 | * 10M~1GBit/Sec |
| | 26 | * !InfiniBand |
| | 27 | * 100GBit~ |
| | 28 | * MPI on SMP machine |
| | 29 | * You can run MPI program on SMP (Symmetric Multiprocessing) machine. |
| | 30 | * Variables are not shared (all private). |
| | 31 | |
| | 32 | == Software Distribution == |
| | 33 | * [[http://www.open-mpi.org|OpenMPI]] |
| | 34 | * [[http://software.intel.com/en-us/intel-mpi-library|Intel MPI Library]] |
| | 35 | * [[http://www.lam-mpi.org|LAM/MPI]] |
| | 36 | * [[http://www.mpich.org|MPICH]] |
| | 37 | * [[http://mvapich.cse.ohio-state.edu/index.shtml|MVAPICH]] |
| | 38 | * [[http://msdn.microsoft.com/en-us/library/bb524831%28v=vs.85%29.aspx|MS MPI]] |
| | 39 | |
| | 40 | |
| | 41 | == Brief Tutorials for Programming Languages == |
| | 42 | * [[cypress/Programming/Cexamples/Mpi|C]] |
| | 43 | * [[cypress/Programming/CppExamples/MpiCpp|C++]] |
| | 44 | * [[cypress/Programming/FortranExamples/MpiFortran|Frotran]] |
| | 45 | |
