LAM/MPI logo

LAM/MPI General User's Mailing List Archives

  |   Home   |   Download   |   Documentation   |   FAQ   |   all just in this list

From: Tim Prince (timothyprince_at_[hidden])
Date: 2006-05-18 15:03:07


Andrew Friedley <afriedle_at_[hidden]> wrote: Beniamino Sciacca wrote:
> Hi community!
> this is my question:
> I've two pc (centrino duo and AMD 64 X2), and i use molecular mechanics
> software (gromacs).
> I need to use both my two pc to simulate faster. The kernel in the pc is
> the same (2.6....).
> The pc are connected by an hub (10Mbps).
> How can i run simulation in parallel with both the pc?
> LAM is installed in both the pc, and during the compilation of gromacs i
> enabled LAM; so before starting "mdrun" i type "lamboot", and then
> mdrun.......
> What must I do to tell mdrun to use both the machines?
> thanks very much for your help
>
> Beniamino Sciacca
<snip>

IIRC, gromacs is very communications-intensive. I wouldn't be surprised
if gromacs ran slower on your two nodes over a 10mbps hub than it did on
just one of the machines. As Brian suggested, you should really get
something better. If the two machines are all you need connected, a
crossover cable might be sufficient as well.
  How can the kernel be the same on a 32-bit and a 64-bit machine? Do you mean you are running a 32-bit OS on both? The 64-bit OS could be OK, but you will have to run the same 32-bit lam and application on both.
  Even if they are running kernels derived from the same source, a 32-bit dual core laptop with limited RAM and limited potential for scaling performance from 1 to 2 cores is likely no match for clustering with a 64-bit machine with sufficient RAM, though I have heard no reports about MPI performance on 64-bit laptops.