Question: cold start vs warm start compile times

[sambacha]

Basically wondering how much of a difference (if any statistically measurable amount exists that is) would there be if forge compiles from a dead cold start vs a pre-warmed state.

Pre warming would be compiling something like

pragma solidity ^0.8.0;contract dummy{}"

for x amount of times.

Is there a potential worthwhile gain to warrant exploring this?

Cheers

[gakonst]

WDYM pre-warmed state? There's caching in the compiler already, which saves some time. Calling it N times doesn't really do much. We also want to add forking caching to disk when pinned against a block, to avoid re-making network requests in forked tests.

[sambacha]

WDYM pre-warmed state? There's caching in the compiler already, which saves some time. Calling it N times doesn't really do much. We also want to add forking caching to disk when pinned against a block, to avoid re-making network requests in forked tests.

So forge will support using an external store for cache persistence, e.g. redis (if you wanted to)?

[onbjerg]

No, just disk cache 🙂

[gakonst]

Could do anything really but let's just go with disk cache, keyed by pinned block like hardhat