Did you read my previous email? Jeff On Tue, Oct 15, 2013 at 5:16 PM, Julio Henrique <[email protected]>wrote:

I have mpich-3.0.4 compiled with # ./configure --prefix=/opt/mpich --enable-shared --enable-f95 --enable-threads=runtime --enable-romio --enable-nemesis-shm-collectives --with-pvfs2=/opt/pvfs2 --with-file-system="pvfs2+nfs" --with-thread-package=posix --with-device=ch3:nemesis --with-java=/etc/java-7-openjdk # make # make install

and PVFS2 on Whezzy 7.2 and works fine. 01 server and 08 nodes.

However, I want to install Siesta and my doubt is: I need to install Siesta on all nodes AND on server or just on server?

Thank's. Julio.

------------------------------ From: [email protected] Date: Tue, 15 Oct 2013 15:55:44 -0500 To: [email protected] Subject: Re: [mpich-discuss] mpich - siesta - cluster

If you use a static binary, you don't need the libraries to live on the compute nodes. If you use a binary with dynamic linkage (run 'ldd binary.x' to see what the dependencies are), then you need those paths to resolve everywhere the binary launches.

If you have a shared filesystem, just build everything there.

Jeff

On Tue, Oct 15, 2013 at 3:52 PM, Julio Henrique <[email protected]>wrote:

Does anyone know if I need to install the Siesta (ab initio molecular dynamics simulations ) only on the server or on all nodes in a cluster? Or just a libraries (mpich, blacs, scalapck, etc) on nodes and the libraries and Siesta on the server? Julio.

_______________________________________________ discuss mailing list [email protected] To manage subscription options or unsubscribe: https://lists.mpich.org/mailman/listinfo/discuss

-- Jeff Hammond [email protected]

_______________________________________________ discuss mailing list [email protected] To manage subscription options or unsubscribe: https://lists.mpich.org/mailman/listinfo/discuss

_______________________________________________ discuss mailing list [email protected] To manage subscription options or unsubscribe: https://lists.mpich.org/mailman/listinfo/discuss

-- Jeff Hammond [email protected]