what(): std::bad_alloc

[Tristan Lefebure]

Dear all,
I am getting an error message that looks different than the
one that was reported few days ago.

------------------
[tristan@babylon 4bucky] bucky --opt-space *.t.in
Bayesian Untangling of Concordance Knots (applied to yeast
and other organisms)
BUCKy version 1.4.0, 28 June 2010
Copyright (C) 2006-2010 by Bret Larget and Cecile Ane

This is free software; see the source for copying
conditions. There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A
PARTICULAR PURPOSE.

Screen output written to file run1.out
Program initiated at Thu Dec 2 08:35:05 2010

Reading in summary files....
0 10 20 30 40 50 60 70 80 90 100
+----+----+----+----+----+----+----+----+----+----+
****************************************************
....done.
Number of genes sequenced for each taxon:
1 sde 1072
2 spyb 1072
3 spy9 1072
4 spy6 1072
5 spy4 1072
6 spy3 1072
7 spyd 1072
8 spy5 1072
9 spy8 1072
10 spy1 1072
11 spyc 1072
12 spya 1072
13 spy7 1072
14 spy2 1072
15 seq 1072
16 sca 1072
Read 1072 genes with a total of 37845666 different sampled
tree topologies
Writing input file names to file run1.input....done.
Sorting trees by average posterior probability....done.
Initializing random number generator....done.
Initializing gene information....terminate called after
throwing an instance of 'std::bad_alloc'
what(): std::bad_alloc
Aborted
---------------

I ran it an a 20GB memory server, and I don't think I ran
out of memory. Any idea?

Thanks!

--
Tristan Lefebure

Darwin C
Université Lyon 1

+33 4 72 44 79 54

[Cecile Ane]

Hi Tristan,
I guess this error also comes from a memory issue ('bad_alloc'). 37.8
million different tree topologies is big... The program will need to
store a matrix with this number of rows, unless the opt-space option is
used. Did you try using this option? It should be particularly useful in
cases with a very large number of trees, just like what you have.
Cecile.

[Tristan Lefebure]

Thanks Cecile. That's indeed a memory issue. In the example below I was already using the opt-space option. Should I try to reduce the tree topology space by reducing the number of samples per genes, or the number of genes, ... , or is Bucky more suited for smaller data-sets (my example is 1072genesx16taxa)? 

[Cecile Ane]

Yes, BUCKy is best used --and reliable-- on a smaller number of taxa.
The number of genes is usually not a problem.
Cecile.