[llvm] [llvm-mca][FeatureRequest] Itimeline graph, note source of delay for each instruction (PR #136423)

[via llvm-commits]

sommersun wrote:

> The cycle-accurate simulators I have looked at internally have used `.` for any waiting state.
> 
> I think for OoO machines with loops, the fetch will likely get ahead of execute so "waiting on data" will be more common than "waiting on resources". It might be good to invert the two and pick a new symbol for waiting on resources, keeping the `=` for waiting on data. I'm not sure what others think of the two? I can see it being useful for them to have different symbols.

good idea!

https://github.com/llvm/llvm-project/pull/136423