problems with poy 5.1.1 (only unix&win versions)

[jaakko....@gmail.com]

Hi, made some experiments using MacOSX version of poy 5.1.1 successfully but using the same file & commands with unix & win versions led to the error message: Memory allocation problem in cube. However, the memory usage at any point of the analyses was well below the defined limit (according to the person in charge of the mainframe used). Two versions of compiler used but that didn´t help. The command file used in the latest experiment was:

(* 4xcp_genomes *)

read(chromosome:("/Users/admin_jhyvonen/Desktop/cp_genomes612/poy5test/4xcp_genomes.txt"))

transform(tcm:(1,1),gap_opening:1)

transform(chromosome:(circular:true))

transform(chromosome:(locus_inversion:100,locus_indel:(10,0.9)))

transform(chromosome:(annotate:(mauve,25.0,0.3,0.005,0.25)))

transform(fixed_states:("4xcp_mauveout3",ignore_polymorphism))

build()

swap()

select()

report("4xcp3",diagnosis)

report("4xcp_consensustree3",graphconsensus)

report ("4xcp3.time", timer:"total time")

transform(fixed_states:("4xcp_mauveout3",ignore_polymorphism))

exit()

The data file composed of 4 angiosperm chloroplast genomes (ca. 150 kb each) downloaded from GenBank.

Any and all your hints how to solve this problem would be greatly appreciated, all the best, jaakko

[Nicholas Lucaroni]

That error happens when trying to allocate memory for a three sequence alignment (thus the use of 'cube' to denote the dimensions of the matrix being allocated), so the memory usage would not spike since it fails before it is being used. Are the Unix or Windows machines running 32bit processors or environment? Regardless, I need to dig into somethings to verify where in the script it happens, but we should be able to avoid even doing 3d alignment in this analysis -- you are using fixed state characters. I'm not 100% familiar with all the components of the Mauve aligner though, and would defer to Ward or Lin to add some context here.

--
Nicholas Lucaroni

[Ward C Wheeler]

[Back from Travel]  Yeah--for sure this should never have been doing 3d.  Fixed states shouldn't do that.

Since I got a different error.  I'll look at that code.  as I remember, there were some 

inconsistencies in word size use in the C-module.

W

--
You received this message because you are subscribed to the Google Groups "POY - Phylogenetic Analysis Software" group.
To unsubscribe from this group and stop receiving emails from it, send an email to poy4+uns...@googlegroups.com.
To post to this group, send email to po...@googlegroups.com.
Visit this group at http://groups.google.com/group/poy4.
For more options, visit https://groups.google.com/d/optout.

Ward Wheeler

Division of Invertebrate Zoology

American Museum of Natural History

Central Park West at 79th Street

New York, NY 10024-5192

USA

whe...@amnh.org

http://www.amnh.org/our-research/staff-directory/ward-wheeler

http://www.amnh.org/our-research/computational-sciences/research/projects/systematic-biology/poy/download

http://www.wiley.com/WileyCDA/WileyTitle/productCd-047067170X.html

[Jaakko Hyvönen]

Hi Ward, many thanks in advance, keep me updated. I tried once more with the mainframe version, now removing the superfluous second transform command at the end of the script. Unfortunately still the same error message. On the laptop running now 215000+ seconds, thus mac version seems to be totally OK in this respect, all the best, jaakko

[Ward C Wheeler]

Jaakko,

Do you have the "parmap" version?  If you have multiple cores and enough ram, this 

form of parallelism could help your analysis.

W

Ward Wheeler

Division of Invertebrate Zoology

American Museum of Natural History

Central Park West at 79th Street

New York, NY 10024-5192

USA

http://www.amnh.org/our-research/staff-directory/ward-wheeler

http://www.amnh.org/our-research/computational-sciences/research/projects/systematic-biology/poy/download

http://eu.wiley.com/WileyCDA/WileyTitle/productCd-047067170X.html

[Nick Lucaroni]

parmap: http://rdicosmo.github.io/parmap/

This should be installed via ocamlfind. It also can be installed via OPAM (if that's the way you've installed ocaml). then in the poy configure add --enable-parmap and this allows for the following command to operate...

set(parmap:X)

where X are the number of parallel processes to run.  You should use this before transforming fixedstates and reverse it after (set(parmap:1)) . Having parmap on during the search will slow down your characters but helps in others (dynamic-likelihood). 

[Jaakko Hyvönen]

Hi Nick and Ward, I have only one core but I´ll forward this info to a colleague planning to perform our analysis on mac desktop. Don´t know the specs but it might well have multiple cores, many thanks, all the best, jaakko

[Jaakko Hyvönen]

ah, got things mixed up, even my laptop DOES have two cores in its sole processor and thus installing parmap even on this one makes sense. many thanks for the hint, all the best, jaakko