April 23, 2010, 6:41 pm
I occasionally got this error while running some mpi programs. I could not find the exact reason. Even when I google it, it seems no relevant explanation of it. By investigating the codes with debugger, I notices some data in the memory are huge, when I correct them, the error information was gone too. Therefore I guess this error is related to the expression of illegal numbers. However, I would like to know more about it if someone can point out the exact reason.
April 23, 2010, 12:45 pm
DeinoMPI is an implementation of MPI-2 for Microsoft Windows. DeinoMPI is a derived work from MPICH2 provided by Argonne National Lab. By starting with the MPICH2 code base DeinoMPI inherits a stable and complete implementation of the MPI-2 standard. DeinoMPI heavily modifies the original code base and does not rely on anything from Argonne National Lab.
Continue reading ‘An MPI implementation I didn’t know before’ »
March 30, 2010, 11:23 pm
1. apt-get install mpich2 (remove other implementation to prevent trouble)
2. create .mpd.conf file under home dir ~/ , add a line “password=qwert” (replace qwert with any other texts)
3. create a hostfile or machine file (e.g. mpd.hosts) with node names listed one per line. (this file can be in any folder)
4. start mpd. There is a script shipped with mpich2 “mpdboot”. It accept some options. “mpdboot -f mpd.hosts -n #” with #<= number of nodes listed in mpd.hosts file. This command will start mpd on all the nodes if ssh or rsh is set correctly (without password).
5 use mpdtrace to verify the node names (nodes having been added to the mpd ring, ready to be used in the following mpirun calls).
6 mpirun -np N -host node1,node2 ./program (it’s not necessary to use all the nodes in the ring, but some of them is fine)