[mypyc] feat: new primitive for int.bit_length

[BobTheBuidler]

This PR adds a new primitive for int.bit_length and is already ready for review.

[JukkaL]

Thanks for the PR! bit_length can be useful in low-level code, where a generic method call can really slow things down.

[BobTheBuidler]

I'm not entirely sure how to proceed here, I can't think of any reason our use of getattr would lead to segfault behavior where there was none before

[BobTheBuidler]

@JukkaL This one I could definitely use some help with though. I have no idea why 1eb2c7e might cause segfault behavior and I'm inclined to say maybe it's a non-issue and we can mark this test as 64-bit only?

[JukkaL]

I think I found an issue. Also can you measure the perf difference using a micro-benchmark?

[BobTheBuidler]

Also can you measure the perf difference using a micro-benchmark?

Will this benchmark suffice? I can deploy this commit hash and compare against 1.18 which I already have benchmarked here

[BobTheBuidler]

Also can you measure the perf difference using a micro-benchmark?

Will this benchmark suffice? I can deploy this commit hash and compare against 1.18 which I already have benchmarked here

1.18 vs this branch

There was a ~6% increase in speed for various permutations of the int_to_big_endian benchmark which includes not only a call to bit_length but a decent bit of other stuff as well

[JukkaL]

Thanks, looks good now!