Depositing ETH from L1 to L2 costs too much gas

[snario]

An L1 -> L2 deposit looks as though it is costing ~240,000 gas which is significantly more than what you'd expect for a relatively simple transfer. The figure is large because the contracts do some rate limiting on the number of L1 -> L2 enQueues. It seems like that could be optimized. As a first step, one should do an audit of this code path and try to profile what operations are inflating the gas cost.

[K-Ho]

cc @ben-chain @maurelian

[K-Ho]

We need to map out a few low hanging fruit here for decreasing deposit gas cost, as over 1,000 people will be depositing ETH in a couple weeks to use Synthetix Exchange

[K-Ho]

1. Remove the messageNonce (read the length of the CTC queue instead)
2. Remove the gas burn in enqueue
3. Remove the sentMessages mapping storage. (Later, add functionality for replayMessage that allows providing an inclusion proof to the queue element)
4. Generally clean up the gateway to make fewer cross contract calls

[maurelian]

K-Ho's changes are good suggestions, I'd have to look more closely to come up with more.

[K-Ho]

cc @elenadimitrova

[ben-chain]

WRT 4., I do not think the gateway does any more than the two minimum required calls for transferring and enqueueing. 1.-3. look great though and will be big improvements :D

[elenadimitrova]

Making constructor initialised variables immutable will also help with reducing storage reads

[tynes]

A deposit into zksync takes about 60,000 gas for reference, see:
https://etherscan.io/tx/0xe16b21b5c8eb3ee61e15663502c7bec3a982e5ca2187ff98f7fe72e094b80386

[elenadimitrova]

@tynes This tx is depositing ETH rather than ERC20 and is thus much lighter. Still we can improve our numbers for sure.

[tynes]

@tynes This tx is depositing ETH rather than ERC20 and is thus much lighter. Still we can improve our numbers for sure.

Note that there is a way to deposit ETH directly, see ethereum-optimism/contracts#311