sammccall added a comment.
Trying this on a few files, this seems like it increases preamble sizes up to
1% or so:
| Before | After
---------------------+-----------+----------
AST.cpp | 42249648 | 42419732
XRefsTests.cpp | 56525116 | 56763768
SelectionDAGISel.cpp | 37546764 | 37668564
internal_s.cc | 59582216 | 60024876
internal_m.cc | 192084984 | 193850560
internal_l.cc | 365811816 | 368841388
I can't see any reason to think that RAM/CPU usage would be out of proportion
to this.
A 1% regression here isn't trivial but seems worthwhile for significant
functional improvements.
I suppose we can add:
- diagnostics (already with this patch)
- inlay hints
- signature help
WDYT about keeping this behind a flag until these are working?
For diagnostics alone, I'm not sure this is a good tradeoff (otherwise why
would we restrict it to variadics? FWIW allowing all function templates to be
parsed is +4% to preamble size for internal_l.cc.
I'm less concerned about the effects on the main file, partly because preambles
are a bigger performance cliff, and partly because we're probably only
regressing in cases where the user really is seeing benefits too.
Repository:
rG LLVM Github Monorepo
CHANGES SINCE LAST ACTION
https://reviews.llvm.org/D124688/new/
https://reviews.llvm.org/D124688
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
