mina [WIP] pickles async circuits

Feb 12 '24 17:02 mrmr1993

!ci-build-me

Feb 20 '24 15:02 mitschabaude

A few things remain to address:

[x] Failing Rosetta intg test because of the block_on_async_exn in create_trivial_snapp
[x] JS only works when this console log is added, I don't know why
[x] Commented out log_step and log_wrap
[x] How does this PR affect error feedback on constraint failures and similar errors?
EDIT - answer: even though snarky is left with a half-finished circuit state on such failures, the way we use it doesn't seem to be affected by that (but see https://github.com/o1-labs/snarky/pull/836#pullrequestreview-1903811416). the restrictions in how pickles can be used after this PR are more worrying

Feb 23 '24 16:02 mitschabaude

!ci-build-me

Feb 26 '24 09:02 mitschabaude

!ci-build-me

Feb 26 '24 10:02 mitschabaude

!ci-build-me

Feb 26 '24 12:02 mitschabaude

Note: Some of the test failures and their fixes suggest that this PR adds some brittleness to Pickles because it's no longer possible to await the results of two different compiled instances concurrently. While we take care to run circuits sequentially within one compile call, we don't ensure this across all compile calls. The error messages in these cases are also very unhelpful, e.g. seemingly random constraint failures.

EDIT: see also https://github.com/MinaProtocol/mina/pull/15084/commits/b5e4eccc6fe1a4094a6a209d98d57bce249d72f5

Feb 26 '24 12:02 mitschabaude

!ci-build-me

Feb 26 '24 13:02 mitschabaude

Tested and working in o1js here: https://github.com/o1-labs/o1js/pull/1450

Feb 27 '24 14:02 mitschabaude

Heads up @mrmr1993 - this seems to change all verification keys https://github.com/o1-labs/o1js/pull/1450/files#diff-d157735e3d120da8877dad8c85ce828fb388beae8a3080dfaf4654694ab43feb

Feb 28 '24 13:02 mitschabaude

!ci-build-me

Mar 04 '24 20:03 mitschabaude

!ci-build-me

Mar 05 '24 16:03 mitschabaude

!ci-build-me

Mar 06 '24 08:03 mitschabaude