Parallizing Antialising and other graphic concepts

Started by
3 comments, last by Kylotan 11 years, 3 months ago

Hi all ,

We have to do a end sem project on parallel computing for which we have 5 months time .

I was thinking since graphics is high processing ,we could parallelise it .

But I have no proper idea about it..The use , benifits of it .

I would like to ask a few questions :

1. Would It benifit to parallize polygon generation or dynamic lighting and stuff ? Does it have any practical use ? Maybe simulators and stuff ?

2. We are considering OpenGL for this .Does OpenGL use the processor or the graphics card to generate stuff by default ? Because we are going to use OpenMPI (parallelizing API) and we dont think it supports GPUs .

3. What are the areas we can parallize ?

4. Any suggestions on how to go about it ? Is OpenGL good or should we use something else ? As we are not sure if OpenMPI suports OPenGL

Any suggstions on how to go about it would be helpful

Advertisement

Everything that's done on a GPU is already parallel -- the GPU itself operates asynchronously from the CPU, and internally, the GPU has many cores that are all used to perform small amounts of the work that you ask it to do.

OpenGL is primarily an API used for performing work on a GPU.

Instead of using OpenGL, you could implement your own rendering API (in software) and then parallelize it using MPI. Many graphics operations are known as being "embarrassingly parallel" -- e.g. if ray-tracing an image of a sphere, then every pixel can be processed independently of any other pixel, so not much work is required to implement a parallel solution to the algorithm.

As this sounds like an MPI-based parallel computing course, I doubt that OpenGL (or anything to do with the GPU directly) will fit the bill. But as Hodgman says, a software ray-tracer (possibly working on a more complicated aspect thereof, like radiosity) would be excellent for this.

However, I would check with your professor to make sure that he is ok with an embarrassingly parallel process - my professors tended to be sticklers for requiring us to solve harder problems in parallel computing :)

Tristam MacDonald. Ex-BigTech Software Engineer. Future farmer. [https://trist.am]

Not necessarily Graphics processing, but image processing in general, i.e. image comparison, effects, i.e. extrusion, edge detection, emboss, take you pick. All of those are best suited for parallel processing. Face detection is one nice pick, and with OpenCV not so hard to do.

Pretty sure that already happens in Photoshop etc.

This topic is closed to new replies.

Advertisement