Unknown reasons for the failure in verifying the Groth16 proof about ivls HOT 11 CLOSED

Are the tests in the Groth16 repo passing?

from ivls.

Yes, tests in PCD and Groth16 are both passing.

But IVLS even before the commit before your commit is failing.

from ivls.

So, I need to dig deeper. One of my guesses is whether the problem only pops up when the FFT instance is sufficiently large or very close to something.

from ivls.

Do you know what the size of the constraint system is here?

from ivls.

Let me quickly find a number...

from ivls.

It appears that the number is about 218246 for MainCircuit and 33322 for HelpCircuit. These seem to be some common sizes and are likely not the reasons for failing.

Groth16 verifier constraint is where it fails. In other words, when I removed the Groth16 verifier from both MainCircuit and HelpCircuit, PCD no longer complains that the constraint system is not satisfactory, but the final proof is not verified outside either.

The next step of my testing is likely to see if both raw proofs can be verified outside SNARK---if it cannot, then likely there is an issue with Groth16. If it can, then there is likely a problem with the Groth16 verifier gadget.

from ivls.

@weikengchen maybe this is the issue: arkworks-rs/algebra#243 ?

from ivls.

More info:

• Groth16 prover/verifier passes for 300k constraints.
• Groth16 verifier circuit passes for 300k constraints.

It seems that arkworks-rs/algebra#243 may be very related. So I will likely wait on that one...

from ivls.

@weikengchen Do these tests pass now?

from ivls.

The tests do indeed pass, so that should resolve this issue!

from ivls.

It passes!

from ivls.