Suggest using the updated jemalloc lib

[thaodt]

Hi team, I've just noticed that we're using the outdated jemallocator.
Check this PR from rust-core for more details.

Updated jemalloc lib
This updated one is from tikv

TL;DR: this upgrade could help for sized deallocation's usage correctly, particularly to rustdoc.
So somehow it can deliver a sizable performance increase. We can do some benchmarks after that.

Lets consider my suggestion. Thank you.

[escapedcat]

Thanks!

You mean this?:
https://github.com/SaitoTech/saito-rust/blob/main/Cargo.toml#L34-L35
Is jemalloc-ctl still ok?

Wanna create a PR to use the suggested alternative?

[thaodt]

Thanks!

You mean this?: https://github.com/SaitoTech/saito-rust/blob/main/Cargo.toml#L34-L35 Is jemalloc-ctl still ok?

Wanna create a PR to use the suggested alternative?

Sure, will do that. For jemalloc-ctl, its already included in tikv-jemallocator. Therefore, we just need to replace its usage.

[clayrab]

Thanks so much for pointing this out, Thao.

The reason I added jemalloc in the first place was related to jemalloc-ctl because I wanted to do some introspection on memory usage in the old codebase... we aren't currently using this because the debate about memory usage has basically been settled...

If I'm understanding the discussions linked above correctly, it seems that rust-lang is already using tikv-jemallocator, so perhaps we should just remove both jemalloc packages from the toml and delete memory-stats.rs?

That being said, if this is mostly an improvement for rustdoc, I'm not particularly concerned about that unless it would improve compile times during dev... but in theory that shouldn't be the case, right?(i.e. why would be build the docs on every compile?).

I think the proper way to do this would be as Thao suggested and do some very basic benchmarks, at least. However, I wouldn't be opposed to simply removing the packages, since that's probably the right thing to be doing anyway. If no one else wants to handle this, I'll probably just proceed this way.

[thaodt]

Thanks so much for pointing this out, Thao.

The reason I added jemalloc in the first place was related to jemalloc-ctl because I wanted to do some introspection on memory usage in the old codebase... we aren't currently using this because the debate about memory usage has basically been settled...

If I'm understanding the discussions linked above correctly, it seems that rust-lang is already using tikv-jemallocator, so perhaps we should just remove both jemalloc packages from the toml and delete memory-stats.rs?

That being said, if this is mostly an improvement for rustdoc, I'm not particularly concerned about that unless it would improve compile times during dev... but in theory that shouldn't be the case, right?(i.e. why would be build the docs on every compile?).

I think the proper way to do this would be as Thao suggested and do some very basic benchmarks, at least. However, I wouldn't be opposed to simply removing the packages, since that's probably the right thing to be doing anyway. If no one else wants to handle this, I'll probably just proceed this way.

Got your point, Clay. Upon checking, its only used on memory-stats.rs, so if that file is useless now, we can simply removing it & the related packages.