jon-chuang

@ValarDragon 30-70 split is not really "dominated by MSM". A 10% improvement in FFT would translate to 3% overall improvement. What I don't understand is what these cosets/domains are. I...

Hi, yes, the use case is indeed to skip the subgroup check. We can use a batched subgroup check instead which is about 30-75x faster, iirc.

It would be nice to specialise the impl for vec or [T] where T: AffineCurve etc. to perform the batch subgroup check. Perhaps a change for the future.

Upon further thought, I personally dont like the idea of having checked and unchecked in the API, and am more in favour of one of the two solutions: - Remove...

Hmm yeah, I guess the solution proposed is meant to allow for minimal breakage. I guess compression and data integrity checks do form a reasonable quadrant on the type of...

> Hmm since it's useful, I propose renaming `(de)serialize_unchecked` to `(de)serialize_compressed_unchecked`. I am undecided on whether we should also rename `serialize` to `serialize_compressed` in the process. WDYT @ValarDragon @jon-chuang @kobigurk?...

> Im in favor of the approach of having 4 methods > > ``` > * serialize_compressed > * serialize_uncompressed > * serialize_compressed_unchecked > * serialize_uncompressed_unchecked > ``` > >...

> Here's some example code, that allows incorporating the batch checks as well: I like this strategy as it exposes optional fine grained control with fewer breaking changes and more...

Instead of the above approach, I propose instead to have a trait with overridable default impls ``` trait ValidData { fn validate(value: Self) -> Result { Ok(()) } fn validate_batch(values:...

Hi, @Pratyush @ValarDragon, could I get permission to push to non-protected branches?