My favorites | Sign in
Project Home
READ-ONLY: This project has been archived. For more information see this post.
Search
for
  Advanced search   Search tips   Subscriptions
Issue 19: Offload fractal rendering to the GPU, with OpenCL
1 person starred this issue and may be notified of changes. Back to list
Status:  Fixed
Owner:  ejrh00@gmail.com
Closed:  Jan 2012


 
Project Member Reported by ejrh00@gmail.com, Nov 14, 2011
I am making exploratory ventures into this.  It might not pan out.  Most GPUs should be able to do 32-bit integer and float computation, and some will do 64-bit.  I don't know what mine will do.

Starting with http://www.browndeertechnology.com/coprthr_stdcl.htm.  The windows package is for 64-bit only, whereas my MSVC is for 32-bit only, so I'm adapting and recompiling it.  Still have not got anything compiled and running yet.

I have not yet thought about how the Mandelbrot computations can be translated into OpenCL style.  I don't think the standard mfunc interface will work.  So a variation on the STANDARD drawing mode that bypasses mfuncs and instead passes larger chunks of the work to OpenCL will probably be in order.

One possibility is to make it like PARALLEL mode: distribute pixels to workers in even steps, so that all workers are likely to have similar shares of expensive pixels.

Jan 2, 2012
Project Member #1 ejrh00@gmail.com
Fixed by issue 184.

Still a lot of performance tuning to do (it's currently slower than SIMPLE,LOOP mode).  Since there isn't much variation in speed when it does deep parts, I think the main cost is the overhead of setting up the OpenCL dispatch.  So, doing more pixels at once could help.
Status: Fixed
Jan 2, 2012
Project Member #2 ejrh00@gmail.com
Bah.  Fixed by *revision* r184.

Powered by Google Project Hosting