IRM logic improvement proposal

[pakim249CAL]

IRM logic improvement proposal

[Rubilmax]

What about passing nothing to the IRM, and letting it get what it needs ?

In my PoC, the state can only be loaded all at once from outside the contract and is already loaded to memory anyway, so this way is cheaper

[MathisGD]

https://eips.ethereum.org/EIPS/eip-2330 could also help

[Rubilmax]

the state can only be loaded all at once from outside the contract

Yeah you just reminded me that I also expose extsload getters, so this sentence is wrong. I feel like it doesn't cost much considering the size of the state so it's not much of a matter to always pass it to the IRM

[Rubilmax]

Agreed, here's how I did it:

https://github.com/morpho-labs/blue/blob/63e8da9592869a8935a4dadba3b73063db19302f/src/interfaces/IRateModel.sol#L7-L10

https://github.com/morpho-labs/blue/blob/63e8da9592869a8935a4dadba3b73063db19302f/src/libraries/MarketStateLib.sol#L169C8-L172

[MathisGD]

What about passing nothing to the IRM, and letting it get what it needs ?

[makcandrov]

I agree with Patrick on that point. It is more efficient and elegant to give the IRM the variables we know it will need. Also, it limits the potential effects of the IRM, because imo it shouldn't depend on the market addresses.

[MathisGD]

I don't think that we should pass specific data to the IRM (except if we could pass everything with a market struct for example) because it would necessarily restrict the possibilities in the IRM. Blue is meant to work with any contract that returns a reasonable rate.

And I think that the IRM should return a rate and not an amount of interests, because it is a more natural split between Blue and IRMs, it let less logic to the IRM, and it seems that we want to emit the rate from Blue directly.