[gentoo-science] Fireball

[gentoo-science] Fireball

[Alessandro Yoshi Polliotti]

Has anyone successfully compiled/run this software under gentoo? Is there an 
ebuild (I found none in bugzilla)?

http://www.physics.byu.edu/research/lewis/fireball/

Given the apparent number of headaches it's causing to other people I'd like 
to have as much preliminary information as possible from fellow gentooers :)

TIA

BTW: I'm new to the list :)

-- 
"Because while the truncheon may be used in lieu of conversation, words will 
always retain their power. Words offer the means to meaning, and for those 
who will listen, the annunciation of truth."
V for Vendetta
-- 
gentoo-science@gentoo.org mailing list

RE: [gentoo-science] Fireball

[Nelson, David (ED, PAR&D)]

> -----Original Message-----
> From: Alessandro Yoshi Polliotti [mailto:polliotti@bo.imm.cnr.it]
> Sent: 04 January 2007 08:21
> To: gentoo-science@lists.gentoo.org
> Subject: [gentoo-science] Fireball
> 
> 
> Has anyone successfully compiled/run this software under 
> gentoo? Is there an 
> ebuild (I found none in bugzilla)?
> 
> http://www.physics.byu.edu/research/lewis/fireball/
> 
> Given the apparent number of headaches it's causing to other 
> people I'd like 
> to have as much preliminary information as possible from 
> fellow gentooers :)
> 
> TIA
> 

>From the looks of it you need to specifically request the software. They may not be too happy if you wanted to distribute it as an ebuild but you could make it one of those F-ebuilds (can't remember the proper term - the ones where you have to fetch the files yourself rather than via wget).

David

Note: These views are my own, advice is provided with no guarantee of success. I do not represent anyone else in any emails I send to this list.

-- 
gentoo-science@gentoo.org mailing list

Re: [gentoo-science] Fireball

[Alessandro Yoshi Polliotti]

On Thursday 04 January 2007 14:03, Nelson, David (ED, PAR&D) wrote:
> From the looks of it you need to specifically request the software. They
> may not be too happy if you wanted to distribute it as an ebuild but you
> could make it one of those F-ebuilds (can't remember the proper term - the
> ones where you have to fetch the files yourself rather than via wget).

That was my idea, to be honest I don't even know if they have proper release 
numbering. The package is 1gb (uncompressed) with a lot of stuff in it, I 
still have to wade through it.
Since i have read of many problems regarding this program I'm interested in 
any tip that could make my life easier, including a working ebuild ;)

Regards

-- 
"Because while the truncheon may be used in lieu of conversation, words will 
always retain their power. Words offer the means to meaning, and for those 
who will listen, the annunciation of truth."
V for Vendetta
-- 
gentoo-science@gentoo.org mailing list

[gentoo-science] MPI and multicore + lib question

[Alessandro Yoshi Polliotti]

I hope I'm not asking anything too obvious but I wasn't able to find relevant 
information online.

I have installed apbs enabling mpi, it works but it detects only one cpu in my 
core duo system. Strictly speaking it's correct, but not quite what I 
expected. I would like to know more about this behaviour, is it due to my 
misconfiguration of apbs, mpi, kernel or other factors?

Besides this, trying to make apbs detect my two cores I tried also mpich2 and 
lam-mpi as mpi implementations. Using both of them and recompiling maloc apbs 
will complain that maloc has not been compiled with mpi support, and will 
keep mpi out of the build process. maloc however doesn't print any warning.

Thanks in advance

-- 
"Because while the truncheon may be used in lieu of conversation, words will 
always retain their power. Words offer the means to meaning, and for those 
who will listen, the annunciation of truth."
V for Vendetta
-- 
gentoo-science@gentoo.org mailing list

Re: [gentoo-science] MPI and multicore + lib question

[Markus Dittrich]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On Thu, 11 Jan 2007, Alessandro Yoshi Polliotti wrote:
> I hope I'm not asking anything too obvious but I wasn't able to find relevant
> information online.
>
> I have installed apbs enabling mpi, it works but it detects only one cpu in my
> core duo system. Strictly speaking it's correct, but not quite what I
> expected. I would like to know more about this behaviour, is it due to my
> misconfiguration of apbs, mpi, kernel or other factors?
>
> Besides this, trying to make apbs detect my two cores I tried also mpich2 and
> lam-mpi as mpi implementations. Using both of them and recompiling maloc apbs
> will complain that maloc has not been compiled with mpi support, and will
> keep mpi out of the build process. maloc however doesn't print any warning.
>

Hi Alessandro,

If I interpret your question correctly it looks
like you did not use mpirun to start your parallel
apbs job, which is what you need to do. 
Have a look at mpirun's man page which should
get you started.

Best,
Markus


- -- 
Markus Dittrich (markusle)
Gentoo Linux Developer
Scientific applications
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.6 (GNU/Linux)

iD8DBQFFpvJ2xlRwCwb7k40RAnRSAJ0XoCU67qxODwOFSr7MgbAtTHmwGwCfW8F/
zEdE3Ik3SgDILbMVcVOyByI=
=iVxE
-----END PGP SIGNATURE-----
-- 
gentoo-science@gentoo.org mailing list

Re: [gentoo-science] MPI and multicore + lib question

[Benjamin Sobotta]

On Thursday 11 January 2007 16:46, Alessandro Yoshi Polliotti wrote:
> I hope I'm not asking anything too obvious but I wasn't able to find
> relevant information online.
>
> I have installed apbs enabling mpi, it works but it detects only one cpu in
> my core duo system. Strictly speaking it's correct, but not quite what I
> expected. I would like to know more about this behaviour, is it due to my
> misconfiguration of apbs, mpi, kernel or other factors?
>
> Besides this, trying to make apbs detect my two cores I tried also mpich2
> and lam-mpi as mpi implementations. Using both of them and recompiling
> maloc apbs will complain that maloc has not been compiled with mpi support,
> and will keep mpi out of the build process. maloc however doesn't print any
> warning.
>
> Thanks in advance
>
> --
> "Because while the truncheon may be used in lieu of conversation, words
> will always retain their power. Words offer the means to meaning, and for
> those who will listen, the annunciation of truth."
> V for Vendetta

Hey!

This may be slightly OT, and I dont know apbs, but are you sure that MPI is 
what you need? To my knowledge, MPI is for distributed computing, i.e. when 
you have several nodes to do calculating work.

As far as I understand you only have one computer. Wouldn't some shared memory 
parallel implementation like OpenMP make more sense? You might want to check 
if your application is multithreaded. If it is, and you're just using one 
machine it should already use both cores available in your system. No need 
for MPI in this case.

Cheers,

Ben




-- 
gentoo-science@gentoo.org mailing list

Re: [gentoo-science] MPI and multicore + lib question

[Alessandro Yoshi Polliotti]

On Friday 12 January 2007 03:29, Markus Dittrich wrote:
> If I interpret your question correctly it looks
> like you did not use mpirun to start your parallel
> apbs job, which is what you need to do.
> Have a look at mpirun's man page which should
> get you started.

Bingo :)

On Monday 15 January 2007 03:03, Benjamin Sobotta wrote:
> This may be slightly OT, and I dont know apbs, but are you sure that MPI is
> what you need? To my knowledge, MPI is for distributed computing, i.e. when
> you have several nodes to do calculating work.
>
> As far as I understand you only have one computer. Wouldn't some shared
> memory parallel implementation like OpenMP make more sense?

You are right, however I'm probably going to setup a cluster and this is part 
of an evaluation to determine the advantages and problems I could have.
Honestly right now I know near nothing about mpi and I'd like to understand 
how it works before I get a request to create a production environment based 
on it. ;)
That's why I'm trying to make it work on my laptop where I have full control 
of the system.

-- 
"Because while the truncheon may be used in lieu of conversation, words will 
always retain their power. Words offer the means to meaning, and for those 
who will listen, the annunciation of truth."
V for Vendetta
-- 
gentoo-science@gentoo.org mailing list

[gentoo-science] mpb: dependencies require two mpi implementations

[Alessandro Yoshi Polliotti]

These two libs need different mpi implementations, and both are needed for 
mpb:

[ebuild  N    ] sci-libs/hdf5-1.6.4  USE="mpi ssl 
zlib -debug -hlapi -static -szip -threads" 4,748 kB
needs mpich2

[ebuild  NS   ] sci-libs/fftw-2.1.5-r2  USE="mpi" 1,227 kB
needs lam-mpi

Needless to say, the two implementations block each other.
I assume that both of them are quite computationally intensive since they have 
mpi support. The question is: which one of the libraries should I compile 
with mpi support, given that they will be used with mpb?

ps: thanks for helping me solve the previous problems.

-- 
"Because while the truncheon may be used in lieu of conversation, words will 
always retain their power. Words offer the means to meaning, and for those 
who will listen, the annunciation of truth."
V for Vendetta
-- 
gentoo-science@gentoo.org mailing list

Re: [gentoo-science] mpb: dependencies require two mpi implementations

[Donnie Berkholz]

Alessandro Yoshi Polliotti wrote:
> These two libs need different mpi implementations, and both are needed for 
> mpb:
> 
> [ebuild  N    ] sci-libs/hdf5-1.6.4  USE="mpi ssl 
> zlib -debug -hlapi -static -szip -threads" 4,748 kB
> needs mpich2
> 
> [ebuild  NS   ] sci-libs/fftw-2.1.5-r2  USE="mpi" 1,227 kB
> needs lam-mpi
> 
> Needless to say, the two implementations block each other.
> I assume that both of them are quite computationally intensive since they have 
> mpi support. The question is: which one of the libraries should I compile 
> with mpi support, given that they will be used with mpb?
> 
> ps: thanks for helping me solve the previous problems.

Use hdf5 1.6.5, it allows any MPI implementation.

Thanks,
Donnie
-- 
gentoo-science@gentoo.org mailing list