Danny Willems
Danny Willems
Nit (not for this PR): a good documentation, IMO, is also a documentation that describes what the parameters of the function does, and in particular, explains why there are some...
Right now this will not be enough: folding needs the /sum/ of constraints, not the max(block_constraints) + common_constraints. To be optimised in the future. _Originally posted by @volhovm in https://github.com/o1-labs/proof-systems/pull/2366#discussion_r1650766720_
For a follow-up, but still important to write a proper test: we might want to add a check that some registers have been set correctly for the whole execution, after...
it doesn't work as a test, as I don't instance witness pairs to use as inputs, I made it just to show how it would be used and check there...
Accepting to move on. However, we should have an "index" or a setup phase as specified before. _Originally posted by @dannywillems in https://github.com/o1-labs/proof-systems/pull/2249#pullrequestreview-2086141663_ There might be improvements.
Agreed. I want to rework the design later on to remove `set_fixed_selectors` altogether and take selectors from some trait for the column. _Originally posted by @volhovm in https://github.com/o1-labs/proof-systems/pull/2233#discussion_r1617312033_
Yeah, in the following PR I introduce some converting methods to use multi-dimensional arrays more. Might be good to use more commonly indeed :) _Originally posted by @volhovm in https://github.com/o1-labs/proof-systems/pull/2233#discussion_r1617104007_
Hmmm, right now I was focusing more on the stack space, and I think memory considerations sound like an early optimisation. But in the future, maybe. _Originally posted by @volhovm...