I don't agree to stop those blaming. Actually there are countless of
problems of various computational tool in this stupid OS. I am using it
simply because someone stupidly installed Ubuntu for a GPU workstation.
Actually, I don't believe the /usr/bin/mpirun is the only mpirun in my
path since I've already set it up in the shell. Such configuration works
both in my Redhat and SUSE, but not Ubuntu.
Although Ross state that GTX590 may not work with mpirun evoking
multiple GPU for single job running, but it works fine in both Redhat
and SUSE. I can use
mpirun -np 2 pemed.cuda.MPI .....
but it doesn't work in Ubuntu!!!, Even the administrator have problem to
compile it in MPI.
the ACEMD is another GPU software, which can also accelerate Amber. It
also doesn't work in Ubuntu, but work fine in Redhat and SUSE.
Someone use DiscoveryStudio also found that it have troubles running
jobs in Ubuntu, but works very well under Redhat and SUSE.
Schrodinger software often crashed in Ubuntu, we never observe this in
either Redhat or SUSE.
A good OS is definitely important for computational work. I don't
believe Ubuntu can be classified a good type, at least not for
computational work.
Albert
On 03/19/2013 11:28 AM, Jan-Philip Gehrcke wrote:
> You should stop blaming Ubuntu and start using things properly:-)
>
> From 'man which' on an Ubuntu system: `which` is "searching the PATH
> for executable files matching the names of the arguments."
>
> So, if /usr/bin/mpirun is the first and only mpirun in your PATH, it
> makes perfectly sense that this is the one returned by which. According
> to this man page `which` does not evaluate any aliases.
>
> What you have to appreciate is that `which` itself is not specified by
> POSIX, i.e. it could be that it behaves differently on different Linux
> distributions (although I doubt that `which` on Redhat evaluates
> aliases, but that's not the point here).
>
> The corresponding POXIX-specified command is `command -v`. It will give
> the same result on any POSIX-compliant system.
>
> Cheers,
>
> Jan-Philip
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Tue Mar 19 2013 - 04:00:03 PDT
