jhorstmann commented on issue #12045: URL: https://github.com/apache/arrow-rs/issues/12045#issuecomment-6662590252
That sounds a lot like re-implementing memcpy ;) With a modern `target-cpu` and `-Cllvm-args=--x86-use-fsrm-for-memcpy` you could test how an inlined `rep movsb` instruction performs vs. a call to `memcpy`. For `extend_from_dictionary` specifically, another idea could be to track the min/max lengths of strings in the dictionary. I could imagine that for many data sets, columns are actually fixed length at runtime. The extend code could then be specialized for some small lengths. -- 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]
