Greg Tatum

C++ code already has locations that come in as raw memory addresses. We could map those to individual assembly instructions. Ideally it would be nice to map these back to...

Oh, and I could also see JIT invalidations being an issue on a single hot function.

1. `intervalIndex`: Looking at [one particular case](https://searchfox.org/mozilla-central/rev/ce9ff94ffed34dc17ec0bfa406156d489eaa8ee1/tools/profiler/public/GeckoProfiler.h#736-762) we could use thread local storage to store a static incrementing index that is incremented on the constructor for `AutoProfilerTracing`. Then we could...

per mstange in Slack: > Kannan wants to add more JS markers and is intimidated by the fact that he needs to update the frontend

I spent a bit of time simplifying the marker schema. I think some of the more complicated bits should be handled on the front-end only, in order to have a...

Seems like this defines the problem and a solution, but it's not broken out with more detailed work.

We could also use this infrastructure to create meta information about what markers are appropriate to show for what types of profiling presets: https://bugzilla.mozilla.org/show_bug.cgi?id=1589681 This way, for example, if we...

`updateProfileSelection` is really hot when using mouse events, it dispatches an action really quickly, and causes unneeded component and store churn.

> However, the risk isn't 0%, in case we are subtly mutating things and are not realizing it. Literally I just modified a test and it failed because the snapshot...

Looking through samples of the data I frequently see mismatched punctuation on the source and target languages, including inconsistent quote usage. I wonder if this is causing some of these...