[9fans] MPI

[9fans] MPI

[hugo rivera]

Hi,
I'm about to start a course that goes by the title of "Computational
Physics", and as I was having a look at the items that we are going to
cover, and saw that there's an "Introduction to parallel computing and
parallel programming with Message Passing Interface (MPI)". Some of
you 9fans may be familiar with this protocol, any comments on it? can
you recommend a book on this topic (not MPI, but concurrent stuff in
general)? how plan9 solves this problem, if at all? Any feedback is
welcome ☺
If you consider this message as noise, please, disregard it.
Saludos,

-- 
Hugo

Re: [9fans] MPI

[ron minnich]

If you want to know the current state of the art at the high end you
need to understand MPI.

It's not great, in fact it's awful, but it does the job for now.

ron

Re: [9fans] MPI

[Andrew Smart]

On Wed, 2010-03-31 at 16:57 +0200, hugo rivera wrote:
> I'm about to start a course that goes by the title of "Computational
> Physics", and as I was having a look at the items that we are going to
> cover, and saw that there's an "Introduction to parallel computing and
> parallel programming with Message Passing Interface (MPI)". Some of
> you 9fans may be familiar with this protocol, any comments on it? can
> you recommend a book on this topic (not MPI, but concurrent stuff in
> general)?
Don't know of any books.  Read the book the instructor recommends.  Here
are a couple good sites with interesting articles on parallel
programming in general:

http://www.devx.com/SpecialReports/Door/40893
http://www.drdobbs.com/high-performance-computing/

Having access to an operational cluster to test things on while you are
learning helps.  In your class you will likely be given access to a
cluster.

Happy coding!

-Andrew Smart

Re: [9fans] MPI

[EBo]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 1878 bytes --]



> I'm about to start a course that goes by the title of "Computational
> Physics", and as I was having a look at the items that we are going to
> cover, and saw that there's an "Introduction to parallel computing and
> parallel programming with Message Passing Interface (MPI)". Some of
> you 9fans may be familiar with this protocol, any comments on it? can
> you recommend a book on this topic (not MPI, but concurrent stuff in
> general)? how plan9 solves this problem, if at all? Any feedback is
> welcome ☺

the open source implementations of MPI (mpich1, mpich2, openmpi) all have
little examples to play with and get started -- and also test the
installation.  I would recommend to start by playing with those.

It has been years since I had taken the initial intros to MPI/PVM, but at the
time the supercomputer center had some nice tutorials and online
documentation.  So, I would search a number of these installations.

Other than that, you might want to download some of the models which use MPI
and possibly play with them.  Depending on the size of the codebase this might
scare you off a bit, but I actually find playing with the GCM WRF and RegCM3
rather enlightening once I got past the initial frustrations...  On second
thought I never got over my frustrations, but this brings me to a pet peve --
if you are looking into modeling and are not a computer scientist by training,
PLEASE, PLEASE, PLEASE at LEAST sit in a course on algorithms and data
structures, and possibly another one in software engineering.  For all the
theoretical physicists, ecologists, biologists, sociologists, and economists I
have worked with, I have yet to meet one which have had more than a practicum
in FORTRAN and/or C/C++.  If you already have a CS background, you come to
understand why some of the code still makes me twitch...

Re: [9fans] MPI

[hiro]

wrong list

Re: [9fans] MPI

[David Leimbach]

[-- Attachment #1: Type: text/plain, Size: 269 bytes --]

Wrong answer... He was also asking about any libraries or technologies that
Plan 9 might have, no one has answered that part of the question yet as far
as I know.

Dave

On Wed, Mar 31, 2010 at 11:37 AM, hiro <23hiro@googlemail.com> wrote:

> wrong list
>
>

[-- Attachment #2: Type: text/html, Size: 522 bytes --]

Re: [9fans] MPI

[andrey mirtchovski]

there was a proof-of-concept implementation of MPI for Plan 9 at LANL
which relied on primitives specific to the OS in order to implement
the communication (mounted remote file servers and namespaces). The
cool thing about it was that it allowed one to write MPI programs
using shell scripts. The funding for the project didn't materialize
and it wasn't pursued further.

i don't have any copies of the two technical reports that discussed it
(only references), nor do i have the code anymore, so i'm not sure how
useful this whole email is, sorry.

Re: [9fans] MPI

[EBo]

> > wrong list
>
> Wrong answer... He was also asking about any libraries or technologies that
> Plan 9 might have, no one has answered that part of the question yet as far
> as I know.

ahhh... I missed that.  I do not have all my lit integrated into bibtex yet,
so here are some URL's and citations that might be of interest:

  http://www.usenix.org/events/usenix07/posters/vanhensbergen.pdf

  "Experiences porting the Plan 9 research operating system to the IBM Blue
Gene supercomputers" (Minnich and Mckie, 2009)

  "Grid Computing with Plan 9: an Alternative Solution for Grid Computing"
(Mirtchovski, 2005)

  "System Support for Many Task Computing" (van Hensbergen and Minnich, 2008)

Actually replacing MPI/Globus with Plan9 is something that I am interested in
myself.  I personally would love to find some exemplar programs for plan9 and
Inferno which I could study.  They are probably around, but I have not made
time to dig them up yet.

  EBo --

Re: [9fans] MPI

[hiro]

All right ;)
Ron, when has plan9 stopped being state of the art?

Re: [9fans] MPI

[ron minnich]

On Wed, Mar 31, 2010 at 10:51 PM, hiro <23hiro@googlemail.com> wrote:

> Ron, when has plan9 stopped being state of the art?

Remember, "the state of the are" does not mean "good".

These new flint arrowheads are state of the art! I always use them for
HPC with my MPI code!

:-)

ron

Re: [9fans] MPI

[David Leimbach]

[-- Attachment #1: Type: text/plain, Size: 453 bytes --]

On Wed, Mar 31, 2010 at 4:02 PM, ron minnich <rminnich@gmail.com> wrote:

> On Wed, Mar 31, 2010 at 10:51 PM, hiro <23hiro@googlemail.com> wrote:
>
> > Ron, when has plan9 stopped being state of the art?
>
> Remember, "the state of the are" does not mean "good".
>
> These new flint arrowheads are state of the art! I always use them for
> HPC with my MPI code!
>
> :-)
>

With some ancient Fortran sprinkled in?  :-)


>
> ron
>
>

[-- Attachment #2: Type: text/html, Size: 980 bytes --]

Re: [9fans] MPI

[EBo]

> > These new flint arrowheads are state of the art! I always use them for
> > HPC with my MPI code!
>
> With some ancient Fortran sprinkled in?  :-)

Now were were those Jacquard loom plates I had sitting around?  Are we done
with out multi-tread processing testing?  You know that was a real warped
setup... :-)

Re: [9fans] MPI

[Jorden M]

On Thu, Apr 1, 2010 at 2:19 PM, EBo <ebo@sandien.com> wrote:
>
>> > These new flint arrowheads are state of the art! I always use them for
>> > HPC with my MPI code!
>>
>> With some ancient Fortran sprinkled in?  :-)
>
> Now were were those Jacquard loom plates I had sitting around?

Under the pile of Hollerith cards.

>Are we done
> with out multi-tread processing testing?  You know that was a real warped
> setup... :-)
>
>

Re: [9fans] MPI

[Jeff Sickel]

Sometimes SOTA just has staying power.  Even now, in the days of the streaming audio and video technologies, businesses making state of the art turntables for whats that? LPs (isn't that stone-age tech?)... continue to release new products.

-jas

On Mar 31, 2010, at 6:02 PM, ron minnich wrote:

> On Wed, Mar 31, 2010 at 10:51 PM, hiro <23hiro@googlemail.com> wrote:
> 
>> Ron, when has plan9 stopped being state of the art?
> 
> Remember, "the state of the are" does not mean "good".
> 
> These new flint arrowheads are state of the art! I always use them for
> HPC with my MPI code!
> 
> :-)
> 
> ron
> 
> 

Re: [9fans] MPI

[Corey Thomasson]

But OTOH, who's still making SOTA VHS players (ignoring the VHS/DVD
combos and VHS->USB thingies)

On Thu, Apr 1, 2010 at 11:49 AM, Jeff Sickel <jas@corpus-callosum.com> wrote:
> Sometimes SOTA just has staying power.  Even now, in the days of the streaming audio and video technologies, businesses making state of the art turntables for whats that? LPs (isn't that stone-age tech?)... continue to release new products.
>
> -jas

Re: [9fans] MPI

[maht]

On 01/04/2010 17:01, Corey Thomasson wrote:
> But OTOH, who's still making SOTA VHS players (ignoring the VHS/DVD
> combos and VHS->USB thingies)
>
Search for : LG MG64 VHS VCR

Re: [9fans] MPI

[Corey Thomasson]

well damn

On Thu, Apr 1, 2010 at 12:19 PM, maht <maht-9fans@maht0x0r.net> wrote:
> On 01/04/2010 17:01, Corey Thomasson wrote:
>>
>> But OTOH, who's still making SOTA VHS players (ignoring the VHS/DVD
>> combos and VHS->USB thingies)
>>
>
> Search for : LG MG64 VHS VCR
>
>
>
>

Re: [9fans] MPI

[hugo rivera]

2010/3/31 EBo <ebo@sandien.com>:
> Other than that, you might want to download some of the models which use MPI
> and possibly play with them.  Depending on the size of the codebase this might
> scare you off a bit, but I actually find playing with the GCM WRF and RegCM3
> rather enlightening once I got past the initial frustrations...  On second
> thought I never got over my frustrations, but this brings me to a pet peve --
> if you are looking into modeling and are not a computer scientist by training,
> PLEASE, PLEASE, PLEASE at LEAST sit in a course on algorithms and data
> structures, and possibly another one in software engineering.  For all the
> theoretical physicists, ecologists, biologists, sociologists, and economists I
> have worked with, I have yet to meet one which have had more than a practicum
> in FORTRAN and/or C/C++.  If you already have a CS background, you come to
> understand why some of the code still makes me twitch...

I must admit that, in general, physicist, astrophysicists and
astronomers are very bad programmers. I've worked with all those
breeds and we really suck. Maybe the fact that I've read, and solved
almost all the exercises from "the practice of programming"
 and K&R's "the c programming language" will make you feel better.
Currently I'm going through Aho et al. "the awk programming language",
and then I'm planning to move to "The Unix Programming Environment".
Coming back to the subject, I'll start with your recommendations and
Andrew's. Probably this is just a newbie talking, but it seemed to me
a kind of problem easily solved with file servers and 9p; but again,
this was only my first impression.
Thanks to all for the feedback.

-- 
Hugo

Re: [9fans] MPI

[EBo]

> I must admit that, in general, physicist, astrophysicists and
> astronomers are very bad programmers. I've worked with all those
> breeds and we really suck.

referring to we, I take it you are one of them.

> Maybe the fact that I've read, and solved
> almost all the exercises from "the practice of programming"
>  and K&R's "the c programming language" will make you feel better.
> Currently I'm going through Aho et al. "the awk programming language",
> and then I'm planning to move to "The Unix Programming Environment".

from this I realize that you are well on your way, and I stand duly
chastised...  In other words I think I have finally met one.  Thank you sir,
may I have another!  ;-)

> Coming back to the subject, I'll start with your recommendations and
> Andrew's. Probably this is just a newbie talking, but it seemed to me
> a kind of problem easily solved with file servers and 9p; but again,
> this was only my first impression.

Some of the people here are the authorities on those subjects, and wish you
the best in your endeavors.

  Best regards,

  EBo --