Just tested my GPU code vs my non-GPU code.

Its a simple game of life implementation. My test is on a 80 x 40 grid running for 100,000 cycles.

The normal code took 117 seconds.
The CUDA code took 2 seconds.

Holy fuck this is terrifying.

So are you going to increase your grid until it matches the CPU time?

@bkwilliams to be honest, I have a LOT more to learn about CUDA first. I need to properly find a way of /understanding choosing threads per block.

What I'm really struggling with right now is figuring out a good way to encorporate the CUDA stuff into my preferred OOP style.

CUDA is amazing.

I remember running an n-body gravity benchmark like 16 years ago. On the GPU it ran with 100k particles and was getting very high fps. The same thing on my cpu at the time (3 core?) could do about 2k before slowing down.