Replace solana_program::program::invoke with solana_invoke::invoke

[deanmlittle]

Creating this PR to get the conversation started. Thanks @cavemanloverboy. Running some A/B tests locally now against my own escrow program that does some basic CPI calls. Will report back shortly. Check out solana-invoke to see how it works. 🙏

[vercel]

@deanmlittle is attempting to deploy a commit to the coral-xyz Team on Vercel.

A member of the Team first needs to authorize it.

[acheroncrypto]

Thanks for creating the PR, it's an interesting one.

The README mentions:

The current CPI functions solana_program::program::invoke* perform unnecessary copies and allocations. This crate removes these inefficiencies in a manner that is 100% backwards compatible.

If this really is 100% backwards compatible, and is objectively better than solana_program's implementation, why not PR this to change it from solana-program side? Both Anchor and Native programs would get the benefits this way.

[deanmlittle]

Yes, this should definitely be merged into solana-program. At the same time, if we can significantly reduce CU usage today without waiting for an upstream merge that may or not may not even happen, would that not be better than leaving it as is?

[cavemanloverboy]

Historically, I've had trouble getting PRs into agave. Either ignored or takes far too long. Since the instruction stabilizer has a few tricky unsafe bits, I suspect it will receive pushback.

[cavemanloverboy]

Thanks for creating the PR, it's an interesting one.

The README mentions:

The current CPI functions solana_program::program::invoke* perform unnecessary copies and allocations. This crate removes these inefficiencies in a manner that is 100% backwards compatible.

If this really is 100% backwards compatible, and is objectively better than solana_program's implementation, why not PR this to change it from solana-program side? Both Anchor and Native programs would get the benefits this way.

this is why lol

anza-xyz/agave#1426