Postpone HirIdification until HirId functions are optimized

[ljedrz]

Perf took a pretty big hit today and I'm suspecting that one of the HirIdification methods might be the cause, because they aren't optimized yet (some of them convert HirId to NodeId and call the old NodeId function for now). If this is indeed the cause, we will want to optimize these functions before using them.

Can I get a try+perf run?

[rust-highfive]

r? @estebank

(rust_highfive has picked a reviewer for you, use r? to override)

[ljedrz]

Cc @Zoxc

[Xanewok]

@bors try

[bors]

⌛ Trying commit 5fa4351 with merge fdfb126...

[bors]

☀️ Test successful - checks-travis
State: approved= try=True

[estebank]

@rust-timer build fdfb126

[rust-timer]

Success: Queued fdfb126 with parent 57d7cfc, comparison URL.

[rust-timer]

Finished benchmarking try commit fdfb126

[estebank]

r=me with a vengeance if you feel it's ready to do so, that's a hell of a lot of green. Can you take the lead on investigating the perf hit?

That being said, in the diff I see a couple of places that are only hot if there has been an error. Those cases could probably be kept without reverting (thinking of https://github.com/rust-lang/rust/pull/58376/files#diff-9d22f4d35c9bd98748f3e1a7cda1b3e7 and https://github.com/rust-lang/rust/pull/58376/files#diff-a4796471067f8f3d7549cb74eca68a4e).

[Zoxc]

@bors r-

This is likely due to #58085 (@Mark-Simulacrum just turned this off on perf, which would explain the perf result here)

[ljedrz]

@Zoxc yep, the perf including these commits is green again. Whew ^^.