The heuristic to subtract 1 million gas to estimate max counts is OK

[PaulRBerg]

Context

I said here that I will follow up on the cross-chain gas usage issue.

My Input

It turns out that the heuristic introduced by @smol-ninja in #908, i.e. to subtract 1 million gas from the block gas limit, accounts for the whole gamut of cross-chain differences in Minimum Transaction Gas. So we can keep it like that.

I did some comprehensive research today, and it appears that:

1. No chain has a Minimum Transaction Go greater than 1 million
2. Execution costs are mostly equivalent to Ethereum Mainnet.

The only exception to point no. 2 is zkSync, which provides gas refunds. However, this is not a problem because it means that we will overestimate, not underestimate. In other words, our estimated max counts will be more than enough to cover the actual execution cost on zkSync.

Gas Fee Models

Chain	Type	Min Tx Gas	Model	Explanation
Arbitrum	Optimistic Rollup	35-70k	Multifunctional gas limit	Bake L1 data fee into L2 gas limit
BNB Chain	Sidechain	21k	EVM	Same as vanilla EVM
Gnosis	Sidechain	21k	EVM	Same as vanilla EVM
OP Stack	Optimistic Rollup	21k	Two-dimensional fees	Explicit L2 execution fee and implicit L1 DA fee
Polygon	Sidechain	21k	EVM	Same as vanilla EVM
Scroll	ZK Rollup	21k	Two-dimensional fees	Explicit L2 execution fee and implicit L1 DA fee
zkSync	ZK Rollup	120k-190k	Custom: AA + state diffs	Gas costs fluctuate a lot due to Account Abstraction and state diffs

References

• feat: gas costs mds1/evm-diff#31
• https://ethereum.stackexchange.com/q/163222/24693

---

cc @sablier-labs/solidity

[smol-ninja]

Thank you for the research on this subject. This is a very good summary tbh. 👏👏👏

[PaulRBerg]

Cheers.