New lint [inefficient_pow]

[Centri3]

changelog: New lint [inefficient_pow]

[rustbot]

r? @Alexendoo

(rustbot has picked a reviewer for you, use r? to override)

[Centri3]

I'm not entirely sure how useful this lint is, considering the cost of using .pow() anyway isn't too large (see godbolt, there are a ton of branches though) but maybe it's desired somewhere ^^

[Centri3]

Huh, I've just looked at the assembly for 1 << 200 (while ensuring rustc doesn't optimize it out) and it seems it uses shl edi,cl, with and ecx,31, which is, uh, weird. I guess it ands it with the size of the integer. That's probably why it breaks in a really weird way when it overflows 🤷‍♀️ At least without it, it wouldn't roll back over and instead be 1100 -> 1000.

Unfortunately shl doesn't set the overflow flag afaik so it can't really be a rustc issue. An issue might be desirable just in case somebody smarter than me knows a way to detect overflow on this (stackoverflow shows nothing particularly clever), that way rustc can optimize out .pow() and it'll still be correct even if it overflows, maybe you can shl edi,1 repeatedly but that's way slower and probably slower than just .pow().

[bors]

☔ The latest upstream changes (presumably #11012) made this pull request unmergeable. Please resolve the merge conflicts.

[bors]

☔ The latest upstream changes (presumably #11115) made this pull request unmergeable. Please resolve the merge conflicts.

[Centri3]

I don't particularly like the suggested code, and think this should be an optimization pass in rustc instead. I'll work on that instead