ARROW-10682: [Rust] Improve sort kernel performance by enabling inlining of is_valid calls

[jhorstmann]

Benchmark results on an aws t3 instance (cascadelake):

$ cargo bench --bench sort_kernel
sort 2^10               time:   [168.90 us 168.98 us 169.08 us]                      
                        change: [-59.182% -59.041% -58.824%] (p = 0.00 < 0.05)
                        Performance has improved.

sort 2^12               time:   [839.88 us 840.15 us 840.46 us]                      
                        change: [-57.095% -57.040% -56.947%] (p = 0.00 < 0.05)
                        Performance has improved.

sort nulls 2^10         time:   [148.69 us 148.75 us 148.82 us]                            
                        change: [-66.823% -66.727% -66.633%] (p = 0.00 < 0.05)
                        Performance has improved.

sort nulls 2^12         time:   [721.79 us 721.95 us 722.12 us]                            
                        change: [-65.966% -65.920% -65.845%] (p = 0.00 < 0.05)
                        Performance has improved.

[jhorstmann]

This alone improved the performance a lot, but directly using ArrayData was even better. Should also improve some other kernels.

I think the data method is a real performance foot-gun and would be in favor of removing it completely. That makes it obvious to every user whether an Arc gets cloned. The cloning itself might not be a problem, but I think it sometimes prohibits other compiler optimizations.

[jorgecarleitao]

Really great finding!

[jorgecarleitao]

FWIW, I agree with you: require an explicit .clone from the user (also broadly speaking, not just for .data() in particular. We should audit the code base for similar occurrences.

[jhorstmann]

I created https://issues.apache.org/jira/browse/ARROW-10683 about removing data method.

[jhorstmann]

@jorgecarleitao I see you're currently also working on sorting, this might be of interest to you.

[github-actions]

https://issues.apache.org/jira/browse/ARROW-10682

[alamb]

👍 thank you @jhorstmann and @jorgecarleitao