Anaïs Querol

After: https://github.com/o1-labs/proof-systems/pull/2310 https://github.com/o1-labs/proof-systems/pull/2311 https://github.com/o1-labs/proof-systems/pull/2305 Subset of the failing constraints: - [ ] `SyscallReadPreimage` > thread 'main' panicked at 'called Result::unwrap() on an Err value: ConstraintNotSatisfied("Unsatisfied expression: ((Curr(x[82]) * (((Curr(x[45]) -...

Update: After https://github.com/o1-labs/proof-systems/pull/2355 https://github.com/o1-labs/proof-systems/pull/2351 https://github.com/o1-labs/proof-systems/pull/2357 ## RType - [x] ShiftLeftLogical - [x] ShiftRightLogical - [x] ShiftRightArithmetic, - [x] ShiftLeftLogicalVariable, - [x] ShiftRightLogicalVariable, - [x] ShiftRightArithmeticVariable, - [x] JumpRegister, - [x]...

There's [this PR](https://github.com/o1-labs/proof-systems/pull/2047) merging into `master` for review that integrates the changes of `experiment/break-stufff` with some ideas to fix the constraints unsatisfiability problem (basically, batch inverting before folding instead of...

Idea from Matthew after our 1:1 to avoid costly u64->Fp transformations: Perhaps we don't need the Montgomery representation until the very end of the folding, and in between we can...

I made this a draft because performance is still not as fast as desired (in my machine, Matthew's idea runs at 1.3 Mhz, but this one is about 5 times...

> Idea: not using batch_inverse from arkworks, but implementing the batch inversion ourself, ignoring zero @dannywillems actually the `batch_inverse` from arkworks already ignores zeros, so that's not a problem anymore