Porting appleseed to CUDA


#1

Hello Franz,
My name is Michael Blackshore and I’m interested of coordinating\manage a CUDA port of the appleseed render.
The dev I have checked with (whom I have not worked with before)
Provided this:
Personal characteristics:
* Skilled in C++ & C / ASM & CUDA.
* OpenCL composability implementation experience.
* Member of Moscow CUDA Center of Excellence (http://hpc.msu.ru/?q=node/116)
* Achieved Computer Science Master degree in Unheterogenous Systems.
* Have in-depth understanding of Math analysis / Data structuring / Linux / Low-level optimization and related challenges of performance speedup.
* Intel Threading Building Block parallel algorithms contributor.
* Author of various commercially successful concurrency work-stealing algorithms (all additional information can be provided).
* IBM Bluegean/P hyper-thore cluster experience.

That should be enough as I understand since C++ and ASM is the needed bit for cuda?
Do you have any additional input?
Do you know any studios that would be interested in a such port?

Kindest regards
Michael Blackshore
[email protected]


#2

Hello Michael,

Thanks for getting in touch!

Porting appleseed to CUDA is an incredibly interesting project, but a challenging one.

Maybe you and the dev would like to join the dev team on Slack to discuss the project?


#3

I would be very interested in this too, if the idea was merely to accelerate rendering via CUDA but making use of system memory. If textures and geometry were expected to fit into VRAM I think you’d quickly lose people. Too many (generally useless) GPU-only renderers out there these days. I haven’t heard of many short films being rendered with Octane, for instance. My $0.02.

Ive been exploring Gaffer, and itd be neat to be able to render an animation with appleseed using just a couple of consumer grade GPUs.