Thanks for posting this, but I think this is mostly taken out of context. There are two really key bits here.
First is that this change is only in ripgrep on master. It's not in a release. It will be in the next release. If you want this in your ripgrep, then `cargo install --git https://github.com/BurntSushi/ripgrep ripgrep` should do the trick. (Add `--features pcre2` if you want that.)
Second is that this make ripgrep around twice as fast for a certain (albeit pretty common) class of regexes. For example, it likely applies to the patterns `sherlock` and `\w+\s+sherlock\s+\w+`, but not to `sherlock|watson`. The latter uses a multi-substring search which isn't SIMD-ified on aarch64 yet. (That is, loosely speaking, my next step. Then hopefully release ripgrep 14 soon there after.)
First is that this change is only in ripgrep on master. It's not in a release. It will be in the next release. If you want this in your ripgrep, then `cargo install --git https://github.com/BurntSushi/ripgrep ripgrep` should do the trick. (Add `--features pcre2` if you want that.)
Second is that this make ripgrep around twice as fast for a certain (albeit pretty common) class of regexes. For example, it likely applies to the patterns `sherlock` and `\w+\s+sherlock\s+\w+`, but not to `sherlock|watson`. The latter uses a multi-substring search which isn't SIMD-ified on aarch64 yet. (That is, loosely speaking, my next step. Then hopefully release ripgrep 14 soon there after.)