You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
EVM mload and mstore instructions seem relatively hot. Their arguments are u256, but this obviously can't work -- the memory needs to be limited somehow.
Reading the geth code, it seems the relevant bits are
Ok, so the wins here are rather meaningful, about 6% of WASM gas.
I am in a bit of a doubt here though -- the correctness of the change would hinge on the fact that any memory offset larger than u32 is going to run out of gas, and, in sputnik, this is a non-local property, as gas counting and execution are very far apart from each other.
The design of oding seems to be more performant and correct here, as it simultanously executes the instruction and subtracts the gas cost, making this constraint local:
EVM mload and mstore instructions seem relatively hot. Their arguments are
u256
, but this obviously can't work -- the memory needs to be limited somehow.Reading the geth code, it seems the relevant bits are
I think what that says is that:
Switthing sputnik to use u64/usize for all offsets should speed it up somewhat significantly I think.
The text was updated successfully, but these errors were encountered: