Cache AnnotatedTypes

[odersky]

This leads to an increase of cachable applied types. For typer/*.scala:

   cached applied types      uncached applied types
now   32.3K                    0.4K
was   30k                     12.4K

[smarter]

Annotations contain a tree whose hash is not stable, how does caching works for them?

[odersky]

@smarter Annotations are hashed and compared with identityHashCode and eq.

[smarter]

Right, but for caching to be useful that means we need to reuse the same tree instance in multiple Annotations, I'm curious in what situation this could happen?

[odersky]

I believe the wins are more for types containing annotated types. Previously, since annotated types were not cached any type containing them was not cached either. The numbers in the PR message show this clearly: We go from 42.4k AppliedTypes to
32.7k. Even for annotated types it's relevant. Say we do an asSeenFrom on an annotated type. Now, if the parent part produces a type that we have seen already, we can cache it.

[odersky]

test performance please

[dottybot]

performance test scheduled: 3 job(s) in queue, 1 running.

[dottybot]

Performance test finished successfully:

Visit http://dotty-bench.epfl.ch/9743/ to see the changes.

Benchmarks is based on merging with master (101e620)

[liufengyun]

LGTM