Innokentii Sennovskii

Gate count jumps after finalization

3

While creating instance during IVC, if we compare the number of gates before and after finalization, we'll get the following:  So there's obviously some bug

When I tried to do ivc.accumulate in parallel, pippenger encountered a race condition.

If we change to logup for lookups, we can stop committing to values of the grand lookup polynomial that we don't need (suggestion by Zac)

Right now, if we have to add some gates at the beginning of ultra_circuit_constructor, ACIR will break, because it accesses variables directly, relying on very particular structure. This is very...

Commitment algorithms (IPA, KZG) have fixed labels for elements being added to the transcript. If 2 instances need to be used in the same transcript for whatever reason, we won't...

Batch_affine_add_chunked and batch_affine_add_internal are used in batch_mul_with_endomorphism and batch_affine_add, is it worth trying to split them out into separate functions?

Goblin Translator concatenated polynomials parallel construction optimization

1

Goblin Translator construction of concatenated polynomials right now uses at most 4 cores (1 concatenated polynomial per core). We can use more if we want to by splitting the polynomials...

Is there a better way to do combiner tests? Right now we have a templated SFINAE method that enables or disables optimisations

feat: Detect subgraphs that are completely independent from inputs or outputs

2

# Description ## Problem\* Currently a developer can create a circuit that would use unconstrained values in such a way, that a part of the circuit becomes completely disconnected from...

The cost of biggroup is extremely high with our attempt to cover all cases