EIP-2537: Precompile for BLS12-381 curve operations

[emhane]

https://eips.ethereum.org/EIPS/eip-2537

op-revm

• feat(optimism): implement BLS precompiles for isthmus bluealloy/revm#2000 if eip isn't included in isthmus, we will simply have to update the name of that function at a later date to reflect the name of the hf that does include it.

op-geth

smart-contract

References

• Pectra Devnet 5 EIP List
• Proto's list of Pectra EIPs
• Pectra support on the OP Stack tracking issue: [Tracker] OP Stack: Support Pectra (Upgrade 14) optimism#12435

[meyer9]

I took a swing at this one here: bluealloy/revm#2000

[protolambda]

This needs:

• a spec PR to explicitly state this EIP is part of the upgrade
• monorepo test that the precompile is usable when the op-stack fork is active (covers at least op-geth, but op-reth testing would also be nice)
• op-program change to override the new BLS precompiles, and hook them up to the preimage oracle, like we have done with other precompiles.

Also warning: if I understood correctly from ACD calls, there are some G2mul precompiles of the BLS precompile set that were being removed, and still seem present in devnet 5, but not in final L1 Pectra scope, due to them being considered unnecessary/duplicate. We should not include these last-minute removed precompiles in L2.

[meyer9]

The removal of MUL precompiles, and pricing changes suggested in the ACD calls is implemented here in geth: ethereum/go-ethereum#30978

We'll need to pull this into op-geth once merged.

[Inphi]

We also need a performance analysis of any introduced precompiles to ensure that derivation is fault provable by cannon/op-program (and probably asterisc/kona too). Similar to the Cannon benchmarking performed by Base for the gas limit increase.

It should answer questions like: Do BLS12-381 operations require too many steps to execute under cannon? How much memory is used by the VM in the worst case execution?

[meyer9]

@Inphi I believe we can utilize the L1 precompiles for this. The reason we needed to benchmark the P256VERIFY precompile is that it isn't implemented on L1, so the fault proof program has to execute it. For this, we're implementing Isthmus as a fast follow-up to Pectra, so the BLS operations can just be passed through. Let me know if I'm missing something though!

[meyer9]

Update on this work:

• Added support to op-geth here: feat: add Isthmus BLS precompiles op-geth#472
• Added test to monorepo here: feat: add acceptance test for BLS precompiles optimism#13934

These are blocked by some fixes to get Isthmus/geth e2e tests working:

• feat: add Isthmus fork rules op-geth#471
• fix: isthmus e2e test utils and withdrawal root fix optimism#13933

Planning on working on fault proofs accelerated precompile tests soon, and testing Reth precompiles as well later this week.

[emhane]

Update on this work:

• Added support to op-geth here: feat: add Isthmus BLS precompiles op-geth#472
• Added test to monorepo here: feat: add acceptance test for BLS precompiles optimism#13934

These are blocked by some fixes to get Isthmus/geth e2e tests working:

• feat: add Isthmus fork rules op-geth#471
• fix: isthmus e2e test utils and withdrawal root fix optimism#13933

Planning on working on fault proofs accelerated precompile tests soon, and testing Reth precompiles as well later this week.

pls add links in pm repo better, this is specs repo, and when spec is merged this issue will close, but doesn't mean implementation is done

the issue doesn't exist, so please make a new one ethereum-optimism/pm#23

[meyer9]

ah makes sense! I'll make an issue there