• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

GPU, a giga@lobal processing unit?

D

dangermaus

Junior Member
hallo people!

I am a little bit tired to explain (here is 01:19 in the morning), but I'd like to present
shortly the GPU project:

it is a Gnutella client that allows File Sharing and CPU sharing as well. Everyone agrees to provide network resources as needed and in return is able to get CPU-cycles from other clients on the network system.
The framework is extensible with plugins. These plugins are not very useful at the moment, I have to admit: right now it is possible to compute pi and solve a partial differential equation. However, a chess plugin is on the work.

I was able to build a cluster with 20 machines (in a student's computer room) and to do some computations. The program is now quite stable and can run for hours unattended (especially compared with the first versions one and a half year ago). There also is some documentation, written as semester work at ETH Zuerich.

I would be glad if some of you could try it out: I keep my laptop running day and night so that a cluster can build up. A more advanced system to allow GPUs to meet is implemented in 0.813, and is called GWebCache, but there lot of clients are required...

thank you for downloading and trying out the program at gpu.sf.net. The program is written in Delphi and released under GPL (a part of a commercial TGnutella component, bought for 100 ? from Kamil, a german programmer) and hosted by Sourceforge. I would be very happy to receive your impressions and criticism, so that we can improve the program.
 
Hi dangermaus!
Welcome on our forums.
I don't get idea of your project.. Who will be able to use my computing power, and how I will be able to use some1 other's power? Can you explain that?

Thanks in advace.
 
hi D347h,

I run now three computers in a computer room at university: if you download the program,
you can send them computational jobs. If someone would be in front of that computers,
they could also ask some computation time from you.

The allowed compuations are libraries of procedures (plugins). Plugins can be verified with
PGP (a freeware Open Source tool) to ensure they were not modified. At the moment, plugins are signed by me, so you
have to trust me... however, it would be much better if someone else could sign them,
after trying them out for a long time. To give you a little bit more confidence, notice that GPU is listed in the "Internet based DC project site" here , in the Distributed Computing section of Googleand in dmoz.org.

The GPU system is a proof of concept, and is far away to be finished. You can get an idea
if you look at the screenshots here.

A chess plugin is hidden into the /frontend/chess subdirectory of the package you can download.

bye!
 
So, it's normal p2p client + it can use your spare power...

OK, my questions now are:
1. Can I ask anyone using his soft to do computation if s/he isn't already doing something?
2. Do the user which I asked could say 'No, I won't do that' ?
3. Chess - is it ChessBrain or other Chess DC ??
4. If it's ChessBrain /I think no, but../, whose conf file will be used?

OK, it's all so far..

YGPM
 
hi D347h (is it an hexadecimal number?)

exactly, it is normal p2p that can use your spare power.

1. anyone in your so called TTL range: for Gnutella, you can reach around 2000 computers.
TTL means Time To Live and is a counter decrease each time your request goes to a new
hop (TCP/IP itself implements a TTL to kill lost packets). Normally, this counter is set to 7.
Additionally, all clients have to understand the request and have to have free resources: GPU can work on 3 requests at a time (the three slots in the graphics tab).

2. At the moment no, the idea would be that the tasks are computed with very low priority, so that the user is not bothered.

3. It is not Chessbrain (I run it on my machine as well). It is a plugin derived from an Open source component written by Tom Kerrigan and Daniel Terhell. At the moment, the plugin does not work in a distributed fashion (all clients compute the same move) and there are some problems at the end with mate positions. The main difference with Chessbrain, is that on Chessbrain all clients compute only one game, and everyone has to follow it on the
Chessbrain homepage. If the GPU plugin works, everyone will be able to play once against a supercomputer using the spare power of the other volunteers... although let's remember that GPU is a proof of concept... maybe I do not get the parallelism right...

4. no, Chessbrain is centralized. like the upcoming BOINC platform, which is nice as well.

bye!
 
D347h = Death, LoL 😀

At 2 - But is there any way to use them in DC projects? 'Cos in all there is part that you must get admin's permission...
At 3 - Idea seem to be nice. When you complete it, just let us know, I'm pretty sure some of us want to test it!

Bye!

OH, one more question: is it better than standard Gnutella cleints? Maybe I can switch from iMesh Light to it...
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top