Towards Inlining Untyped Code #4589
Closed
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
odersky
force-pushed
the
change-inline-untpd-2
branch
5 times, most recently
from
a791e52 to
9a69dbc
Compare
Introduce UntypedSplice and make TreeCopiers and TreeMaps more regular to take splices into account.
- extend scheme to accumulators - no need anymore to define special tpd versions of maps and accumulators - no extra `typeMap` field needed in typeMap; instead the UntypedSplice case is handled directly.
Some refactorings so that it will become easier to generate untyped trees
Use combination of Enum and Case instead.
It's useful to know something was an enum when it started, and it's necessary to pickle this when serializing untyped trees.
- Mark splices in code under -print-debug - Print owners of all definitions under -print-debug-owners
Allow UNTYPEDSPLICE roots with TYPEDSPLICE subtrees. Trees between them are untyped.
- need to maintain Type mode bit for correct desugaring - need to pickle PatDefs directly since desugaring requires too much context.
Everything should be reflected in flags and privateWithin already.
This was previously forgotten, even though unary operations such as `!` and `~` were folded later in FirstTransform. As a consequence, inlining simplifications involving constant expressions using these operations were not done. This affected in particular the trace macro, which was always expanded to an operation taking a closure argument.
Avoid creation of `op1` methods if tracing is not enabled.
Rely on call-by-name parameters instead. Fix "unused defs" logic so that referred-to cbn parameters are not eliminated.
Since all Mod values are now reflected in flags and privateWithin, there is no need anymore to test whether a Mod exists.
odersky
force-pushed
the
change-inline-untpd-2
branch
from
1375bf4 to
e8c96cb
Compare
|
test performance please |
|
performance test scheduled: 1 job(s) in queue, 1 running. |
|
I admit it would be better to split this into separate PRs. But my time-budget is already overdrawn as is, so I have chosen to push ahead and try to get the subsequent steps done instead. |
|
Performance test finished successfully: Visit http://dotty-bench.epfl.ch/4589/ to see the changes. Benchmarks is based on merging with master (228b232) |
|
Replaced by #4620 for now. Putting decision to pickle untyped trees on hold. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
We would like inline code that is only typed at the leaves. This seems to be the best way to have an inlining scheme that can generate new types. This ability to generate new types gives us some of the power of whitebox macros. It is needed for most typelevel programming idioms.
This PR is preparatory. It contains
only as carriers of positions for syntactic analysis in IDE and elsewhere.
inlineas a modifier for function parameters, use by-name instead.reducing the amount of code that's generated.
The PR can be reviewed separately. It will be followed up with another PR, introducing
transparentmethods, which allow to inline untyped trees.