CPU and GPU

General Discussion Other

asimes

in General Discussion • Other • 1 year ago

Please forgive my lack of knowledge on the differences, but is there a method of accessing the GPU for rendering? I have been making a number of sketches lately that involve rendering 3D points and even fairly optimized code is slow enough that it cannot be interactive. People keep mentioning how fast rendering on the GPU is in rendering tutorials I read, is that something that can be done in Processing?

EDIT: Opps, maybe this belongs in General Discussion or something, my bad. I don't see how to change it.

Replies(3)

amnon.owed

Re: CPU and GPU

1 year ago

It can be done through GLGraphics or the new OPENGL renderer implemented by andres in Processing 2.0. To store points, you can use the GLModel in GLGraphics or a (retained) PShape in the Processing 2.0 alpha's. But something like your raymarcher could be even better implemented through a (pixel/fragment) shader probably. The downside of shaders is that they are written in a different syntax based on the C programming language.

monkstone

Re: Re: CPU and GPU

1 year ago

Currently there seems to be some advantage in preferring the original GLGraphics library over the retained PShape (processing-2.0a5), I've been experimenting with PShape to produce a recursive menger. There was a memory GC issue above the number of recursions I used in this sketch. So I experimented by creating a mesh from AABB (toxiclibs) in place of my custom cube (and added the meshes together). If I rendered the mesh in OpenGL using toxis processing renderer (ToxiclibsSupport) I hit the same limit. However when I used the GLGraphics renderer ala Gerd Platl Algebraic Surface Viewer, I found the sketch would render OK with at least one more level of detail. Of course if you want to really push the limits C++ is probably the way to go.

andres

Re: CPU and GPU

1 year ago

Hello, thanks for your feedback. As you point out, the new retained mode in 2.0a5 consumes more memory than the old GLModel class in GLGraphics. The main reason is that PShape3D in 2.0a5 stores in CPU memory both the input and tessellated geometry, besides the VBOs in video memory... This redundancy allows you to do things like transparently recreating a shape if the OpenGL context gets restarted, but it also results in out-of-memory exceptions as soon as the geometry gets moderately complex.I probably need to optimize the memory utilization to make PShape3D lighter.

CPU and GPU

Replies(3)

Re: CPU and GPU

Re: Re: CPU and GPU

Re: CPU and GPU

Statistics

Tags

Actions

Related Posts