[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [ossig] Anyone interested in doing some tech research?




On 01/09/2004 19:37 Mukhsein Johari said the following:
> 1. Overall goal is to be able to utilize multiple GPU
> cards in one box for compute intensive scientific
> calculations. The current areas of interest are
> physics, chemistry and engineering simulations type
> problems. As you can imagine, these use a lot of fp
> and  matrix type calculations.

other than working on a compiler/runtime to interpret to the GPU 
instruction set, the rest would be standard parallel processing work now 
wouldn't it ? aren't projects like this (gelertner's LINDA, MPI and 
other parallel proc projects) already instantiatedworldwide ? :)

> 2. To get there, researchers will have to create
> 'device drivers' and other kernel-side (as well as
> userland) infrastructure software.

i see the following:

1. dynamic runtime compiler/interpreter to convert from x86 instruction 
set to GPU instruction set
2. hook into kernel to provide X number of additonal virtual CPUs
3. reuse kernel's SMP scheduling and locking mechanisms.

you may also want to take a look at FreeBSD 5.x's SMPng as it has taken 
SMP into a slightly different approach wrt mutexes and locking.

> other. However, the interested individuals would be
> part of the team and you'd be surprised how much
> influence / leverage you can have in determining
> direction.

having been in academia, this is not a given. it all depends on the 
project leader and his aptitude to things. i was luck in the sense that 
usm had a dean who was not conservative.

> Commercialization:
> 1. Indeed, almost all the gov grants I currently know
> of (running/coming to end) have this end goal in mind.

end goal, but there this is rarely met or if it is met, it's met with a 
warped sense of what commercialization means.

> 2. In the U I'm attached to, researchers get the
> double-whammy of being paid to do research as well as
> have the opportunity to do their postgrad degrees.
> (yes, effectively get paid to do your postgrad without
> having to pay back a loan)

this is not abnormal. many Us do that, paying their postgrad students 
under the designation Research Assistant or RA.

> OK, hopefully I've answered the questions to your (and
> others) satisfaction. I've got 2 interested

interesting, but i think i'll pass on this. experience has taught me to 
never return to malaysian academia ever again. i'm quite happy being a 
confused capitalist socialist bastard. :)

-- 
Regards,                           /\_/\   "All dogs go to heaven."
dinesh@alphaque.com                (0 0)    http://www.alphaque.com/
+==========================----oOO--(_)--OOo----==========================+
| for a in past present future; do                                        |
|   for b in clients employers associates relatives neighbours pets; do   |
|   echo "The opinions here in no way reflect the opinions of my $a $b."  |
| done; done                                                              |
+=========================================================================+


---------------------------------------------------------
To unsubscribe: send mail to ossig-request@mncc.com.my
with "unsubscribe ossig" in the body of the message