Revert k256 removal

[bkchr]

Let's keep using k256 in wasm to not fuck up people compiling their stuff and apparently it also started to include __stack_chk_fail on my machine as "host function".

[gilescope]

Good call.

[pgherveou]

just fyi, this seems to have quite bad perf implication, this is a result of benchmarking ran today on pallet-contracts

[gilescope]

It's not a performance regression if it was never released. This lib also likely has less unsafety. Personally I think we should put in the effort to improve the performance of the pure rust libs.

[bkchr]

just fyi, this seems to have quite bad perf implication, this is a result of benchmarking ran today on pallet-contracts

When was the last time this benchmark was run?

[pgherveou]

It's not a performance regression if it was never released.

Yes indeed, just sharing some data from the last benchmark, @bkchr the last one was from this commit
b55dbcb 4 days ago

[bkchr]

This was also run on the same machine? I didn't followed, but I knew that we switched or wanted to switch now to the benchmarks on VM.

[pgherveou]

not too sure they were both ran with the b*t bench $ pallet dev pallet_contracts command

[bkchr]

The weight files contains the configuration of the machine. You can compare this.

[pgherveou]

Yes same machine

[bkchr]

Okay, yeah then the pure rust implementation is really that bad. 🙈

But yeah, we know about them as otherwise we would use k256 everywhere: #8089 (comment)

[ggwpez]

I compared it again on the same machine and it used just 35% of the time before this MR was merged. But okay, if it causes build error then we cannot use it now.