On Fri, 15 Sep 2023 19:11:48 GMT, 温绍锦 <d...@openjdk.org> wrote:
> The byte[] length used by HexFormatBench is 512. In this scenario, the > performance improvement of using table lookup is significant. Is this a common use-case? I could see an argument that formatting small chunks is much more common, where users will experience relatively more cache misses from the table lookup. The gain from specializing for `StringBuilder` is impressive. My guess is that most of the gain is from enabling better loop optimizations by turning a complicated loop with interface calls into a simple counted loop over a byte[]. This could be split out into a separate PR, or perhaps we could consider more generalizable approaches to make `append` and `format` perform at the same level ------------- PR Comment: https://git.openjdk.org/jdk/pull/15768#issuecomment-1721772711
