Antoine Pitrou
Antoine Pitrou
How long does it take to generate those files on the fly from the benchmarks? In general `parquet-testing` is for interoperability testing between different Parquet implementations, not for benchmarking of...
Could you explain: 1. is "sorted" mode functionally different from "non-sorted" mode? 2. what explains the perf difference between the two?
So, to my question "are the two modes functionally different", is the answer "no"?
Is it worth keeping the two modes if the only difference is performance? Usually, it's better if users don't have to make low-level performance decisions themselves.
> We will make decisions automatically internal and use a mode with better performance. I'm not sure I understand. Do you mean you'll remove the option in a later PR?...
Note that we're not married to setuptools_scm. If we find out that something else would work better for us, then we can switch to it. Found this comparison using a...
Duplicate of #43057
@abandy Could you please rebase from git main, if not recently done, so that we get cleaner CI results?
> Proof that this PR reduces binary size: Nice! Can you also post numbers obtained with the `size` utility for completeness?
I've run the Take micro-benchmarks locally with this (AMD Zen 2, gcc 12.3.0). The changes a bit all over the place and show that compilers are generally capricious and difficult...