[Bioclusters] some mpi questions
nick
bioclusters@bioinformatics.org
Tue, 2 Nov 2004 16:29:27 -0800
<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<title></title>
</head>
<body>
<div class=3D"showOnlyOnPrint">
<table noborder=3D"" cellpadding=3D"0">
<tbody>
<tr>
<td nowrap=3D"nowrap" align=3D"right" valign=3D"top"><font face=3D"A=
rial,sans-serif" size=3D"2"><b>Subject:</b></font></td>
<td valign=3D"top"><font face=3D"Arial,sans-serif" size=3D"2"><b>some
questions about MPI</b></font></td>
</tr>
<tr>
<td nowrap=3D"nowrap" align=3D"right" valign=3D"top"><font face=3D"A=
rial,sans-serif" size=3D"2"><b>Date:</b></font></td>
<td valign=3D"top"><font face=3D"Arial,sans-serif" size=3D"2">11/2/200=
4,
4:21 PM</font></td>
</tr>
<tr>
<td nowrap=3D"nowrap" align=3D"right" valign=3D"top"><font face=3D"A=
rial,sans-serif" size=3D"2"><b>From:</b></font></td>
<td><font face=3D"Arial,sans-serif" size=3D"2"><a href=3D"mailto:nic=
kfzx@aol.com">nick <nickfzx@aol.com></a></font></td>
</tr>
<tr>
<td nowrap=3D"nowrap" align=3D"right" valign=3D"top"><font face=3D"A=
rial,sans-serif" size=3D"2"><b>To:</b></font></td>
<td><font face=3D"Arial,sans-serif" size=3D"2"><a href=3D"mailto:bio=
clusters@bioinformatics.org">bioclusters@bioinformatics.org</a></font></td>
</tr>
<tr>
<td nowrap=3D"nowrap" align=3D"right" valign=3D"top"><font face=3D"A=
rial,sans-serif" size=3D"2"><b>Organization:</b></font></td>
<td valign=3D"top"><font face=3D"Arial,sans-serif" size=3D"2">-</font>=
</td>
</tr>
</tbody>
</table>
</div>
<font face=3D"Courier New" size=3D"2">Dear MPI programmers,
<br>
<br>
We are a startup working on structure from motion (3D reconstruction
<br>
from video data).
<br>
<br>
We are investing in a small cluster of workstations (8 Athlon 64
<br>
machines) and are hoping to use MPI to paralyze our application.
<br>
<br>
We are still in the early stages of thinking about how we can
<br>
effectively paralyse. We will probably be doing the same task on each
<br>
node but each node working with different data. We are using C++ with
a
<br>
lot of OpenGL in there.
<br>
<br>
Right now the only MPI I have done is a program that divides a number
<br>
into chunks, distributes the pieces, does something with them on each
<br>
node and sends them back to the root. Very simple, but soon its going
<br>
to get complicated :-) I have also used Paraview to do some
distributed
<br>
computation and rendering with MPI in the past but this didn=E2=80=99t invol=
ve
<br>
programming.
<br>
<br>
Is there a standard way for dealing with pointers to memory addresses
<br>
when paralyzing. Do we need to implement a global 'ID' scheme to get
<br>
round the problems associated with paralyzing pointers?
<br>
<br>
Does anyone use RDMA (remote direct memory access) to simulate one big
<br>
memory space (only in MPI 2 I think).
<br>
<br>
What implementation is best for scientific applications using Linux and
<br>
Windows, I am under the impression that MPICH is the most widely used
<br>
but is there a compelling reasons to consider alternatives?
<br>
<br>
Are people using MPI 2 and if so, what new features made people switch?
<br>
<br>
And finally a more general question about paralyzing
<br>
applications....what are the most commonly used parallel program
models,
<br>
especially in the field of protein folding as I think the way people
<br>
paralyze those processes may map closest to what we are hoping to
achieve?
<br>
<br>
Has anyone any advice on how best to approach paralyzing a scientific
<br>
program? Tips/techniques, etc.
<br>
<br>
Any help would be greatly appreciated.
<br>
<br>
Regards
<br>
<br>
Nick Franklin
<br>
<br>
------------------------------
<br>
computer vision startup
<br>
<br>
TEL: 650 363 1540
<br>
<br>
EMAIL: <a class=3D"moz-txt-link-abbreviated" href=3D"mailto:nickfranklin@gma=
il.com">nickfranklin@gmail.com</a>
<br>
------------------------------
<br>
<br>
</font>
</body>
</html>