CodeML memory problem

[Simon Gunkel]

I'm running CodeML to generate an in.BV for later use with MCMCTree, using the method described for using the approximate likelihood for AA sequences. MCMCTree generates the required tmp1.* files and I call CodeML.

The output I get is:

AAML in paml version 4.7, January 2013

ns = 45         ls = 1679451

Reading sequences, sequential format..

Reading seq #45: Nasonia_vitripennis

Sequences read..

1679451 site patterns read, 2183214 sites

Counting frequencies..

     7920 bytes for distance

537424320 bytes for conP

 53742432 bytes for fhK

  5000000 bytes for space

TREE #  1

   1   27.931702

   2   23.639547

   3   22.976141

   4   22.860883

   5   22.852264

   6   22.852178

   7   22.852166

3268818560 bytes for conP, adjusted

Error: oom conP.

So, this is a memory issue, but I do have more than 6GB of free memory, which even with the non adjusted value for conP should be enough.

Is there some way to resolve this or possibly a workaround (the BLEs and Hessian matrices should be additive, right? So could I just cut the alignment into shorter sections and then generate a combined in.BV from the individual out.BVs?)

Paml version is 4.7, I'm using Win 7

[Ziheng]

You have a large dataset.
Yes, oom is my short form error message for out of memory.
I think the program may be comfortable with 2GB of ram, but you are asking for 3GB. If you are happy to send em the three files compressed into an archive, (the sequence alignment, tree, and control file), I'll have a look.
Best,
Ziheng