1 |
Well... I'm presently working on and OpenMPI ebuild that follows the belowmentionned path specs to be able to slot the MPI implementation as per: |
2 |
$implementation/$version/$compiler/ |
3 |
|
4 |
But I don't think this structure will change wether we use modules or eselect (we still need a separate place to store the mutually exclusive packages). But DO poke me if I am going in the wrong direction with this so that I don't waste too much time in the wrong direction. |
5 |
|
6 |
Thanks, |
7 |
|
8 |
Eric |
9 |
|
10 |
PS: couldn't eselect also work for a normal user...and just set the proper environment variable also....??...it's just a BASH script so we should be able to do it the "eselect" way ;) |
11 |
|
12 |
Le 12 Février 2006 18:33, Donnie Berkholz a écrit : |
13 |
> Andrew D. Fant wrote: |
14 |
> > I'll dig around and forward the ebuild to the list later tonight or |
15 |
> > tomorrow. At the risk of going off on a rant, I think that the |
16 |
> > /etc/env.d model is a core part of the problem. If two users want to |
17 |
> > use two different mpi implementations, will eselect allow them to both |
18 |
> > use them at the same time, or will it have to be one or the other at |
19 |
> > once? As I see it, something will have to be tweaking $LD_LIBRARY_PATH |
20 |
> > for each user's environment, |
21 |
> |
22 |
> Yep, that's the plan. Some things just won't work as well because they |
23 |
> don't really respect variables, like the includes. |
24 |
> |
25 |
> Something like this: |
26 |
> |
27 |
> MPI_IMPLEMENTATION="$implementation" |
28 |
> PATH="/usr/lib/mpi/$implementation/$version/$compiler/bin:$PATH" |
29 |
> LD_LIBRARY_PATH="/usr/lib/mpi/$implementation/$version/$compiler/lib:$LD_LIBRARY_PATH" |
30 |
> CPPFLAGS="-I/usr/lib/mpi/$implementation/$version/$compiler/include:$CPPFLAGS" |
31 |
> MANPATH="/usr/lib/mpi/$implementation/$version/$compiler/share/man:$MANPATH" |
32 |
> INFOPATH="/usr/lib/mpi/$implementation/$version/$compiler/share/info:$INFOPATH" |
33 |
> |
34 |
> > or /etc/ld.so.cache will have to be changed |
35 |
> > for each user's login to reflect their prefered mpi environment. |
36 |
> |
37 |
> That totally screws with multiple simultaneous logins. Environment |
38 |
> variables will have to be the way to go. |
39 |
> |
40 |
> > $PATH |
41 |
> > and $MANPATH can be handled by judicious checks in /etc/csh.cshrc and |
42 |
> > /etc/profile, but since /etc/env.d insists on manipulating the shared |
43 |
> > object cache directly, this has the potential to get really ugly without |
44 |
> > some cooperation from the core system team. I don't think that the |
45 |
> > cluster herd can fix it on it's own. |
46 |
> |
47 |
> I think we can, but I could certainly be proven wrong on this. |
48 |
> |
49 |
> > I would love to be proven wrong and will gladly defer to anyone who does |
50 |
> > so, I just know that I've been trying to outthink this problem for a |
51 |
> > year or so on our cluster and haven't come up with a sane solution for a |
52 |
> > true multi-user environment. |
53 |
> |
54 |
> I would love to find a way that works too. =) |
55 |
> |
56 |
> Thanks, |
57 |
> Donnie |
58 |
> |
59 |
> |
60 |
|
61 |
-- |
62 |
Eric Thibodeau |
63 |
Neural Bucket Solutions Inc. |
64 |
T. (514) 736-1436 |
65 |
C. (514) 710-0517 |
66 |
|
67 |
-- |
68 |
gentoo-cluster@g.o mailing list |