fastDNAml on Solaris 10 x86 with mpi
By jeevan on Oct 31, 2006
I have been using this app called fastDNAml, written by Gary J. Olsen, a very useful bio-app that finds out the maximum likelyhood phylogenetic trees from nucleotide sequences. I am interested in the parallel version of the same app, developed by the Indiana University. You can download the same here.To make it work on Solaris 10 x86 mpi 1.2 . The only tweaks to be done are.
- un-tar the source package.
- Switch to the source directory.
- You would want to make changes to Makefile.LINUX.
- Open the Makefile.LINUX , and change
- gcc path.
- set the MPI_ROOT appropriately.
- make the MPICC variable point to the $MPI_ROOT/bin/mpicc
- Build the project using this command #make -f Makefile.LINUX mpi .
- Create a file called machines that looks like this. I have done so cause I need to run four process of fastDNAmlp mpi_foreman, mpi_worker, mpi_fastDNAml, mpi_fastDNAml_mon.
- For the mpirun to pick this up we need to use a tool to change the file and assign each of these nodes a process to run on.
- run this command 'fastDNAml_1.2.2p/testdata/mkp4pg fastDNAml_1.2.2p/src/machines 1 mpi_dnaml_mon ~/dev/fastDNAml_1.2.2p/src' > myProc.pg
- this command generates a proc group file that looks like this.
- the arguments it takes are
- the machines file we generated.
- the number of processor in each node.
- the first process to be run.
- the absolute path of the place where all the executables.
- mpirun -v -p4pg fastDNAml_1.2.2p/myProc.pg fastDNAml_1.2.2p/src/mpi_dnaml_mon -d4 -nmpi56 -s fastDNAml_1.2.2p/testdata/test56.phy
- It should take about 12 minutes to complete this on an Ultra 20 with 1 GB of RAM.
The folks at Indiana have written a nice help file that covers all options, please refer to this fastDNAml_1.2.2p/docs/fastDNAml_1.2.2.txt . I have uploaded it here for your benefit.
Myself and Prem Kumar L (my colleague and friend), have compiled these steps together.
powered by performancing firefox