Compile time of configure gadget

[naure]

Description

This improves compilation times.

Issue Link

No issue, quality-of-life PR.

Type of change

• [ x ] Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• This change requires a documentation update

Contents

The function configure_gadget is split into a generic and a non-generic functions.

Rationale

Generics are not free. They are the biggest contributor to compile times and to CPU code cache. configure_gadget is the largest example of this phenomenon.

I measured about 2x fewer LLVM IR with cargo llvm-lines. Even 3x in Scroll’s fork. The improvement in clock time depends a lot on exactly what you do, but it will be faster.

How Has This Been Tested?

cargo test --features test -p zkevm-circuits --release

[hero78119]

Hi @naure The changes looks neat to me!!
Just wanna further ask, is there more insightful elaboration/reference on why this craft trick can improve compiling time instead of compiler optimize itself?
Want to learn more and see whether can also optimise other parts in current codebase

[naure]

The Rust Performance Book!

[naure]

The next largest occurence of this would be create_gate. That one should be fixed naturally along with this improvement: privacy-scaling-explorations/halo2#144

[ed255]

LGTM! This is really nice!

The github action for this PR took 6m 28s to build the tests (see https://github.com/privacy-scaling-explorations/zkevm-circuits/actions/runs/4266506184/jobs/7427108345)
Previoulsy this was 7m 13s (see https://github.com/privacy-scaling-explorations/zkevm-circuits/actions/runs/4303670677/jobs/7503692807)

So it's a reduction of more than 1 minute :D