Enable YJIT

[JamieMagee]

Code improvement description

Once #9597 is complete, we should enable YJIT and monitor the performance.

References:

• https://github.com/ruby/ruby/blob/master/doc/yjit/yjit.md#command-line-options

[JamieMagee]

@landongrindheim is there enough data to decide whether or not to enable YJIT globally?

[landongrindheim]

@JamieMagee Thanks for checking in on this! Here's a graph tracking job durations (the thing we were specifically hoping to impact) 👇 I don't think there's an appreciable difference, though the max values are 🤏 improved. I suspect YJIT would make a bigger difference for longer-running processes.

---

Here's a split view of the same data, in case the overlay is hard to read.

[JamieMagee]

@landongrindheim thanks for the update. If that's kusto, could you regenerate the graphs with percentiles instead of avg. That should better show if the max values are improving.

I'd recommend using the 68-95-99.7 rule to show 1, 2, and 3 standard deviations from the mean i.e. summarize percentiles(..., 68, 95, 99.7) by yjit_enabled

[landongrindheim]

It's Datadog. And I wrote that before thinking about what it was actually reporting 🤦 The max would be the timing of an individual job, which isn't especially meaningful in this context.