[gmx-users] Negative number of RMSD calculations left in g_cluster

[David van der Spoel]

On 2010-03-26 22.52, Bob Johnson wrote:
> Hello everyone,
> I am attempting to use g_cluster on a large data set (about 69000)
> frames. I'm on a machine that has 48 GB of RAM, so I shouldn't be
> running out of memory. As g_cluster runs, it counts down the number of
> RMSD calculations left. Normally this number is positive (as it should
> be). Instead of stopping at 0, however, the calculation keeps on going
> and displays a negative number of RMSD calculations that remain. This
> negative keeps on getting more and more negative until eventually the
> machine runs out of memory and I get a seg fault.
>
> I have run this same exact calculation skip 2 (35000 frames instead of
> 69000) and it works fine. What is the reason for the negative number
> of RMSD calculations and how can it be fixed?
> Thanks,
> Bob
>
69000^2 comparisons is actually more than 4 bilion, the max number in an 
integer is 2 billion, therefore it wraps round. The memory needed for 
this operation would be at least 19 Gb, if it only stores a matrix of 
all pairwise RMSD. I don't know exactly how the program works but it may 
be less efficient, thereby eating up all your memory, and most likely it 
would take days to build up the entire matrix as well.

-- 
David van der Spoel, Ph.D., Professor of Biology
Dept. of Cell & Molec. Biol., Uppsala University.
Box 596, 75124 Uppsala, Sweden. Phone:	+46184714205.
spoel at xray.bmc.uu.se    http://folding.bmc.uu.se