Adds Gas to Weight mapping

[crystalin]

What does it do?

It adds a better GasWeightMapper than the default one (1-1 mapping).
The new GasWeightMapper compute the gas/weight based on a given ratio.

From our benchmarking, the EVM computation can consume up to 10M Gas/second (on a 4.4ghz) with some optimizations (no trace log, using compiled WASM).

To stay on a very safe path and considering that our current targets are under the required specs, we are targeting a consumption of 4M Gas/second. The GasToWeight conversation is doing: gas * WEIGHT_PER_SECOND / GAS_PER_SECOND

Considering that our current block limit is 500ms Weight (requirement to be a parachain), it means we will have 4M Gas limit per block.
However, the current computation of block weight consider that 25% for operations (including 10% for block initialization) and 75% for the transactions (so 3M gas here)

This diagram gives an overview of the gas consumption:

This gives us a approximate effective gas per block of 3M/block (so 0.5M effective gas/s, ~50% of Ethereum mainnet)

Also, additional note for now, current extrinsics are consuming 125_000_000 weight extra (which is 1000 gas). We might want to merge that into the 21_000 already consumed by the EVM

Another addition is the fact that the maximum extrinsic weight is equal to the maximum normal class weight (75%) minux the data block initialization (10%, not sure why it is included as it doesn't belong to the normal class), minus the extrinsic base weight (125_000), (which is roughly around 64.9% of the block weight). In term of gas it means a Tx is limited to 2_599_000 gas

Is there something left for follow-up PRs?

There will be some tweaking PR to increase those numbers once our benchmark are more accurate and once we have defined our final specs.

Another PR to adapt the Ethereum Gas Limit to reflect the actual gas limit is needed. That will be possible after polkadot-evm/frontier#265 lands.

What value does it bring to the blockchain users?

It gives an accurate limit of the possibilities of the current parachain.
It prevents the network to stall when a TX saturates the EVM and is too long

[JoshOrndorff]

I think this is a good approach to start. And I appreciate the thorough description in the PR.

The Weight type is an alias for u64. Why are we using u128 for these calculations? https://crates.parity.io/frame_support/weights/type.Weight.html

Like I'm

[crystalin]

With 1.3M effective gas per block, we might be under some contract limit.
@albertov19 can you tell me what is the highest gas you ever used ?

[albertov19]

Yes its a bit low, like I mentioned in the meeting, the Oracle contract from Chainlink needs 1.8M to be deployed, and its just one example at the top of my mind.

One ERC721 (NFT token) token contract that I have (it is one with a lots of add-ons) needs 2.4M of gas. We will talk with Rarible soon (an NFT marketplace), so it is somewhat important we are able to showcase compatibility with NFTs.

With 1.3M effective gas per block, we might be under some contract limit.
@albertov19 can you tell me what is the highest gas you ever used ?

[crystalin]

Yes its a bit low, like I mentioned in the meeting, the Oracle contract from Chainlink needs 1.8M to be deployed, and its just one example at the top of my mind.

One ERC721 (NFT token) token contract that I have (it is one with a lots of add-ons) needs 2.4M of gas. We will talk with Rarible soon (an NFT marketplace), so it is somewhat important we are able to showcase compatibility with NFTs.

With 1.3M effective gas per block, we might be under some contract limit.
@albertov19 can you tell me what is the highest gas you ever used ?

Thank you. I'll raise to 2.6M effective gas / block (Equivalent to 8M gas/s on 500ms Weight)

[crystalin]

I did some changes, It might need another review @JoshOrndorff