pitrou commented on PR #50281: URL: https://github.com/apache/arrow/pull/50281#issuecomment-4874095870
> After testing the new benchmark, I discovered something unexpected: the baseline approach—`BitmapAnd` combined with `VisitSetBitRunsVoid`—actually turned out to be the fastest. I attribute this to the bitmap's minimal memory footprint (A bitmap with 1 million elements occupies only 128 KB), This is also going to vary based on memory and cache intensivity of the current workload. Micro-benchmarks generally don't hit very hard on the memory subsystem (and they are single-threaded), so temporary allocations can seem very cheap, especially if the same allocation size is repeated over and over. > Consequently, I have decided to close this PR and simply materialize the bitmap directly when needed. Well, why not have `VisitTwoBitRuns` functions that actually do the grunt job of materializing bitmap chunks and iterating on them? -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
