I successfully compiled an mpi version of hyphy and it appears to run but now I'm having some trouble getting the GABranch batch files to run properly.  hyphy seems to prompt twice for some questions, or perhaps is printing the text strings and waiting for user response out of order.  

I ran hyphy with this command:
openmpirun -np 8 /Users/mike/apps/hyphyjan2008/HYPHY_Source/HYPHYMPI BASEPATH=/Users/mike/apps/hyphyjan2008/HYPHY_Source /Users/mike/GABranchFiles/ModelSelectorBranchLocal.bf

and then hyphy printed the list of Genetic Codes, and waited for user input, but did not ask to enter a code:

Code:

1
 Please choose an option (or press q to cancel selection):
1
/Users/mike/GABranchFiles/Codon file to analyze::
/Users/mike/laeatest/laea.fasta
/Users/mike/GABranchFiles/Codon file to analyze::
/Users/mike/laeatest/laea.fasta
/Users/mike/apps/hyphyjan2008/HYPHY_Source/TemplateBatchFiles/Please select a tree file for the data::
This doesn't seem to be a valid Newick string file. Can't find the opening parenthesis.
Had:Error:MPI Node:0
Node names should begin with a letter, a number, or an underscore. Had:>?Bofuc20320 Bofuc20320
----------
Current BL Command:Build Tree givenTree from string treeString
 

The third prompt seems to want a tree file, even though it asks for the codon file a second time.  Then it prints "Please select a tree file for the data" but does not wait for user input.  It prints the error shown above and then quits.  

The HYPHYMP version seems to work ok for me on the command line.  At least I have used it for phylogeny reconstruction using the standard analyses.  

Hi Sergei - Thanks for the tip.  I have it running now on the example lysozyme data set.  How does the population size affect the genetic algorithm?  I suppose that a smaller population would be more likely to stop at a suboptimal model?
Cheers
Mike

Dear Mike,

Were you piping the input in? If you were, then what happened is either

(1) There were not enough input lines (fewer than the number of prompts made by GABranch...)

(2) One of the inputs for multiple choice lists was rejected. If that is the case, HyPhy will re-prompt up to 10 times and could then consume later lines of input.

I would run the same analysis through the MP build (not MPI), where stdin and stdout work properly, and make sure that all the prompts are accounted for. If that doesn't work, I'll post an example input file for you to try.

Cheers,
Sergei

Hi Sergei,

I try to recompile the HYPHY by using the sh build.sh MPI, but I failed because the source files don't include the mpi.h.  I manual checked the source and don't find the mpi.h.  I download HyPhy at the yestoday moring. It should be the latest.  Can you check it? Thanks very much.

zhuocheng


+-----------------------------------------------------------+
|Building a single-threaded HYPHYKernelMPI for MPI          |
+-----------------------------------------------------------+
COMPILER=g++, gcc
COMPILER_FLAGS= -w -c -fsigned-char -fast -fpermissive -I/Users/zchou/HyPhy/HYPHY_Source/Source -I/Users/zchou/HyPhy/HYPHY_Source/Source/SQLite -D INTPTR_TYPE=long -D __UNIX__ -D __HYPHYMPI__
Building HYNetInterface.cpp
In file included from HYNetInterface.h:13,
                from HYNetInterface.cpp:9:
batchlan.h:361:18: error: mpi.h: No such file or directory