Re: [AMBER-Developers] Compilation of AMBER Tools

From: Jason Swails <>
Date: Thu, 9 Jun 2011 21:30:40 -0600

On Thu, Jun 9, 2011 at 7:54 PM, Ross Walker <> wrote:

> > gcc_select is MacPorts only, and since the newest version of
> > gcc_select, its
> > behavior has changed. (Instead of using "gcc_select mp-gcc45" you need
> > to
> > use "port select gcc mp-gcc45"). I think it's fine to drop support for
> > old
> > GNU compilers when supporting them becomes inconvenient since they're
> > open
> > source, and updating is typically as easy as "port install gcc45" or
> > "apt-get install gcc45". For the commercial compilers we should
> > probably
> > work a little harder to maintain compatibility...
> Assuming you have ROOT which on a large number of machines is NOT the case
> and the person that has root is VERY resistant to installing something new.
> End result a lot of additional effort and bureaucracy that just drives
> people crazy.

Yes, but you can always install gnu compilers in directories *you* have
access to. Not to mention that many centralized machines have versions of
*some* commercial compiler (i.e. all teragrid machines, all local clusters
at UF; so all compute clusters that I've dealt with), which I've argued is
*ok* to make sure we support.

Moreover, RHEL has a gcc44 that comes installed currently, no? It's just
that the default compilers are 4.1.something... In any case, my argument
was that I think we're OK to drop support for GCC 4.1.2. It's hardly the
biggest hoop we're making users jump through.

My last point: I believe pmemd still works with GNU 4.1.2, so you can still
install that bit if you want to on one of those centralized machines (even
though commercial compilers are suggested).

All the best,

> My 5c... (the smallest denomination of coin I still use).

Who still uses real money in the plastic age?

All the best,

Jason M. Swails
Quantum Theory Project,
University of Florida
Ph.D. Candidate
AMBER-Developers mailing list
Received on Thu Jun 09 2011 - 21:00:02 PDT
Custom Search