STL router #347
STL router #347
Conversation
integrate solmate STL, and reduce mloads
wrap erc20 for stl
contracts/TridentRouter.sol
Outdated
| unchecked { | ||
| return i + 1; | ||
| return i++; |
There was a problem hiding this comment.
This doesn't work as expected since it returns the same i value instead of incrementing it. i + 1 is fine and already optimal. I would also remove the comment since it doesn't apply for this generic function.
There was a problem hiding this comment.
I see. I suppose ++i would work here, but will revert to move this along.
revert increment flow
remove unchecked comment
| // Pay the first pool directly. | ||
| bento.transfer(params.tokenIn, msg.sender, params.path[0].pool, params.amountIn); | ||
| // Call every pool in the path. | ||
| // Pool `N` should transfer its output tokens to pool `N+1` directly. | ||
| // The last pool should transfer its output tokens to the user. | ||
| // If the user wants to unwrap `wETH`, the final destination should be this contract and | ||
| // a batch call should be made to `unwrapWETH`. | ||
| uint256 n = params.path.length; |
There was a problem hiding this comment.
Is this not an optimisation? I didn't run the numbers but afaik it is (caching-the-length-in-for-loops)
There was a problem hiding this comment.
@gasperbr I just did some testing and including cache increases gas cost
gist, https://gist.github.com/z0r0z/4187d4f39b9a2a6e89941f9f8e86682b
There was a problem hiding this comment.
let me know if this makes sense. I see explanation of how optimizer might be working to cache outside of loop here, https://gist.github.com/hrkrshnn/a1165fc31cbbf1fae9f271c73830fdda
integrate solmate STL, and reduce mloads