Jonathan Wang

Nice write up! I believe the optimization Han mentioned is https://github.com/scroll-tech/halo2/pull/28#issuecomment-1407968119 We have also observed it greatly reduces memory usage.

Here is an example: the diff of the PR shows what changed, and there is a RAM example test. https://github.com/axiom-crypto/halo2-lib/pull/206

Is there an issue tracking the fact that if I have multiple workspaces/windows open, when I update Zed and restart it will only re-open one of the workspaces? I often...

It would be great to reopen the other issue to track workspace serialization if that's still being worked on!

I have a PR for this here: https://github.com/halo2-ce/halo2/pull/13 You can cherry-pick the last 2 commits and test it out.

I'm curious - how do you feel about just using `#[derive(Serialize, Deserialize)]`? That would avoid all of this. I have yet to benchmark but I would guess the `bincode` serialization/deserialization...

In that case since N is around 2^20 anyways probably there's no immediate need. It'd only be potentially interesting if we have some gates that all access the same set...

What's the status of this PR? It doesn't seem like there's any blockers, and a faster MSM would certainly be very nice. @Brechtpd

Ah thanks for pointing out that other issue. That is very interesting. I can't guess one way or the other which would end up being faster for the halo2 particular...

`Column` implements Copy, so you probably don't want to add a string or even a `&str` to the struct itself. But you could have the `ConstraintSystem` keep a separate vector...