Charles Oliver Nutter

A few notes as I look through this. * Everything appears to JIT pretty much the same in both 6.6.1 and main. I'm not seeing any methods fail to JIT...

I ran a longer jfr profile and may have found something related. During a 5-minute profile, 85_000 InterruptedExceptions were thrown, which works out to 283 exceptions per second, one every...

Actually I was wrong... it seems to be from the ConditionVariable used in the main thread loop. Still a lot of exceptions and likely impacting performance.

Seems like the parser probably is copying more strings than it needs to as well. I think there's lots of optimization potential here but not sure which things have changed...

With a lot of tweaking and zero verification that my changes are valid, I have new numbers with JRuby 10.1, patched `ConditionVariable.wait` logic, and some tweaks to the parser to...

Oops, those numbers were with some other config changes I'll update when I'm back at my machine.

I had some spare time so I have run @schneems's benchmark on x86_64. These numbers are on JRuby 10.0.3.0-SNAPSHOT with no modifications. v6.6.1: ``` 12 threads and 1024 connections Thread...

Yeah this is an unfortunate side effect of using the Java process API on Windows. We don't use our usual native API because it depends on `posix_spawn` which does not...

Did not get to circle back to this but I don't think it's anything that regressed recently. I agree it should be possible to bundle such gems when using jruby-complete,...

To be honest I haven't been unable to figure out what's wrong with it. It's either something wrong in the MacOS/AArch64 FFI or something I'm missing in the FFI-based io/console...