Adding a visualizer for SpiderMonkey MIR/LIR graphs

[eqrion]

The SpiderMonkey team has been prototyping a new web-based MIR/LIR graph visualizer. It takes a JSON format generated in SpiderMonkey and renders an interactive graph. It's a react component and we could publish it as an NPM package or vendor it here, whatever is easier.

We'd be interested in integrating it in the profiler when using samply/jitdump so that we can visualize the CFG of control flow graphs. There is already support for IONPERF=ir which gives us a linearized version of the graph. But it's a lot of data that's hard to scroll through and understand. We also have ideas about visualizing optimization passes, filters, and other things that wouldn't work in a flat text file.

We could emit the JSON separately and manually visualize them somewhere besides the profiler frontend, but then we lose out on sample count information (which is really useful for knowing what to look at).

┆Issue is synchronized with this Jira Task

[eqrion]

I'm familiar with SpiderMonkey's support for JitDump [1] and adapted it to work for WebAssembly. So I can hook it up so that we write the JSON into temp files, and then either emit a JitDumpDebugRecord referencing it, or some new kind of non-standard record.

I am less familiar with how samply exposes the JitDump information, and how that gets into the frontend.

[1] https://searchfox.org/mozilla-central/source/js/src/jit/Jitdump.h

[eqrion]

cc' @mstange from our conversation yesterday.