Better caching of prototypes #9742
Conversation
- Make IgnoredProto a cached type - Convert IgnroedProtos to WildcardTypes in wildApprox - Use `eql` instead of `equals` when hash-consing prototypes
|
test performance please |
|
performance test scheduled: 2 job(s) in queue, 1 running. |
|
Performance test finished successfully: Visit http://dotty-bench.epfl.ch/9742/ to see the changes. Benchmarks is based on merging with master (d11e137) |
This reverts commit 83e6ad4. Also, optimize trace further to be zero overhead # Conflicts: # compiler/src/dotty/tools/dotc/reporting/trace.scala
|
test performance please |
|
performance test scheduled: 2 job(s) in queue, 1 running. |
|
Performance test finished successfully: Visit http://dotty-bench.epfl.ch/9742/ to see the changes. Benchmarks is based on merging with master (101e620) |
| (name eq that.name) && (memberProto eq that.memberProto) && (compat eq that.compat) && (privateOK == that.privateOK) | ||
| case _ => | ||
| false | ||
| } |
There was a problem hiding this comment.
Does the slight difference between equals and eql have a semantic difference or only a performance difference? In any case, for maintainability, it is worth some explanation here.
There was a problem hiding this comment.
It's only for performance.eql does a eq on all recursive calls, knowing that all elements are already hash-consed. I'll add an explanation in one of the next PRs
eqlinstead ofequalswhen hash-consing prototypes