[gmx-users] maximum number of atoms

[Erik Lindahl]

hugo verli wrote:

>Hi,
>
>what is the maximum number of atoms that can be simulated in gromacs? I
>apologize if this information is in the manual, but I was not able to find it.
>
>Thanks in advance,
>
>Hugo.
>

Hi Hugo,
 
We are using signed (vs. unsigned) integers in a couple of places, so I 
think you are currently limited to 2^31-1=2,147,483,647 atoms. The only 
reason we are not using 64 bit integers is that it would hurt 
performance on x86, powerpc, and other 32 bit platforms.



Since you usually need 20-30 bytes of storage per atom it is possible 
that we hit the roof slightly earlier if there are places in the code 
where we first calculate the amount of memory we need and then allocate 
it. Let us know if that's a problem and I promise we'll look into it ;-)


You might have noticed that the pdb and gro input files can't handle 
atom numbers larger than 9999 and 99999, respectively. For this reason 
we just skip the actual number, and use the linecount when reading. When 
we write files we just restart from 0 or 1 after (9)9999.

Cheers,

Erik