Didn't get to answer this yet, here's my view on this...
The way I get it, the particular flavor you pick is tied to the system
you're installing CUDA on.
So, if you're on a Ubuntu system, you need to pick the ubuntu packages,
if you're on RHEL, pick rhel, etc.
Usually, you handle something like this with a version suffix in the
CUDA easyconfig file, and hence you'll end up with a CUDA module like
"CUDA/4.2.9-ubuntu10.04" .
If you want to avoid dragging along the suffix for everything you build
using that CUDA version+flavor, maybe defining a new toolchain would be
wise.
You could have a versionsuffix for the toolchain (a brief one, e.g. only
the first two letters of the CUDA flavor) to indicate the CUDA flavor
included in it.
Including CUDA in your compiler toolchain is something you'll need to do
anything anyway, to handle all the environment variables, compiler
commands and library paths that are specific to CUDA...
That would yield a toolchain like "goalfc/1.2.0-ub", or something like that.
TL;DR: We don't have a standard way of handling something like this,
except for setting a version suffix and dragging it along everywhere.
K.
On 01/14/2013 04:43 PM, Jens Timmerman wrote:
On Mon, 2013-01-14 at 16:12 +0100, Fotis Georgatos wrote:
Hoi Jens,
I mean something that we do not fix already with sething the preload
path and include paths?
I doubt that it can be fixed in this way, since Debian (& Ubuntu?),
do not seem to provide for libexpat.so.0
But easybuild is providing libexpat, so if you put it as a dependency it
will be in LD_LIBRARY_PATH and found by ldd (and others)
Otherwise I would op for just picking one, and making sure we can
install it on any/every platform...
I would rather have some kind of facility to tell "use whichever 4.2 CUDA"
is available on a given system or, automatically produce a default 4.2
modulefile.
(yeah, manually I can always do this).
The tricky part here is, because I want to insert CUDA as dependency for NAMD
but,
The point is that we want a way to install cuda on a system where nvidia
is not providing packages for (as far as this is possible, and/or
allowed by the cuda license)
I believe that as long as cuda doesn't try to do load any kernel modules
(which will only work if you're root anyway, so I think this is not the
case?) we can install it and all of it's dependencies in a os agnostic
way...
Other might have other opinions however...
Regards
Jens Timmerman
