Currently, the code only supports one big R1CS matrix but we already have good benchmarks showing that turning the API to allow uniform circuits that repeat a lot gives drastic reduction in proving time.
The description (informal) is there. While it may seem intimdating at first, the good news is that it requires mostly changing the parameters of some functions to make it work (like the sumcheck only operates on the small circuit instead of the bigger one etc).
TODO:
Currently, the code only supports one big R1CS matrix but we already have good benchmarks showing that turning the API to allow uniform circuits that repeat a lot gives drastic reduction in proving time.
The description (informal) is there. While it may seem intimdating at first, the good news is that it requires mostly changing the parameters of some functions to make it work (like the sumcheck only operates on the small circuit instead of the bigger one etc).
TODO: