Ubuntu QA:
BlogBrainstormPackage status
Log in
Ubuntu QA
The Ubuntu community has contributed 21986 ideas, 135057 comments, 2615221 votes
Idea sandbox Idea sandbox
Popular ideas Popular ideas
Ideas in development Ideas in development
Implemented ideas Implemented ideas
Idea #12333: Use GPGPU via OpenCL or Close to Metal

bug This idea is a duplicate of Idea #16362: Implement support for OpenCL API.
Written by marcocapoferri the 19 Aug 08 at 10:09. Category: System. Related project: Nothing/Others. Status: New
Rationale
Hi all,
why don't do archive (de)compression, cryptographic operations and others computational huge application with GPU instead of CPU?
Ubuntu would be really faster if we use openCL or CTM for doing this things and the CPU will be able to control the rest of the system keeping it stable and reactive for others simpler operations/applications.

Comments are always accepted and sorry for my english

[UPDATE 1]
When I wrote the idea i was thinking how my nVidia 8400M GT could help my Intel T7100. but if someone is using an older GPU (s)he can enable or disable this "power up" because the GPU don't support this technology or for other reasons (like playing 3D games)

52
votes
closed
Solution #1: Auto-generated solution of idea #12333
Written by marcocapoferri the 19 Aug 08 at 10:09.
Ubuntu Brainstorm was updated in January 2009. Since the idea #12333 was submitted before this update, its rationale and solution are not separated. Please vote accordingly, and if you have the necessary rights, please separate the rationale from the solution. Thanks!

Propose your solution

Attachments
No attachments.


Duplicates


Comments
Ssdg wrote on the 19 Aug 08 at 12:14
Hum, that might be interesting, but I don't know if there is some way to give priorities to task using the GPU.

The graphical operations shouldn't wait for the "bonus" computation to end before running.

(ie, I want to decompress something in the background and play openarena as if nothing else runs on the GPU)

neonesis wrote on the 19 Aug 08 at 12:29
Hmm, it's not bat idea, but user should be able to choose GPU or CPU. Some people(suchy as me) have GPU which hardly run GNOME, i'm not saying about Compiz or something like that. So it would be slower than decompression with CPU.

cheesehead (Brainstorm admin) wrote on the 19 Aug 08 at 12:39
Packages already in the repositories:

libbotan1.6
libbotan-dev

Botan (OpenCL): http://botan.randombit.net/
CTM: http://sourceforge.net/projects/amdctm/ (seems moribund)

gmatht wrote on the 20 Aug 08 at 02:31
This may be useful for somethings. However e.g. gzip compresses data as a single linear stream. This means that gzip does not use more than one core. The advantage of GPGPUs is that they have many cores. The disadvantage is that each individual core is slower than a CPU core for typical operations. Thus gzip is likely to be slower on GPGPUs than a regular CPU.

If upstream finds tasks that can be optimized by GPGPUs, great. I think it would be quite hard for Canonical to implement this idea in any useful way. So +0.

Magnes wrote on the 20 Aug 08 at 08:30
What about Compiz? Could you use it with OpenCL/CTM without interference?

marcocapoferri wrote on the 20 Aug 08 at 10:10
@Magnes
I can't give you a "certified" answer because I never developed with OpenCL or CTM. So this is only an hypothesis.

I think that compiz could be used even if OpenCL or CTM are running. Perhaps this could slow down a little bit during the GPGPU non graph elaborations.
But I think that if compiz will slow down from 60fps to 40fps during elaboration this wouldn't be a relevant loss


Post your comment