Emulating a GPU on a CPU using Finite Field Assembly
https://open.substack.com/pub/leetarxiv/p/emulating-a-gpu-on-a-cpu-using-finite?r=2at73k&utm_campaign=post&utm_medium=web&showWelcomeOnShare=true
A Friendly yet Thorough Introduction to Finite Field Assembly programming language and the Recursive Computing Paradigm
Really interesting approach! Using Finite Field Assembly to emulate GPU operations on a
opens up some exciting possibilities.
It reminds me of how alternative instruction sets can sometimes unlock performance in unexpected ways.
The recursive computing paradigm you mention also feels relevant for workloads that demand parallelism but don’t have access to specialized hardware.
I’m curious how this method scales when the dataset grows larger? Does the overhead of emulation outweigh the benefits beyond a certain point, or is it still manageable with optimization?