https://github.com/erichkeane commented:
This all seems really mechanical as well, so I dont' have any concerns with
that. However, the compile-time-tracker _IS_ a bit concerning. The DRAV vs RAV
decision is very much a performance-vs-size debate, and I don't like that we're
giving up a percent+ on a few of these benchmarks (between this and the last
one) in exchange for this size.
What I would LIKE to see is some analysis of:
1- How many instantiations these have
-Thus: How much space taken, and saved by switch
2- Which are hot-path enough to affect benchmarks, and which are irrelevant
For example, the tooling, unittests, tidy/etc variants seem like an easy
decision for me, even if the answer to 1 above is "only 1 instantiation".
However, the ones in CodeGen or AST, or Debug info, etc, I'd want to see a
significant savings, since those clearly affect compile time.
https://github.com/llvm/llvm-project/pull/160290
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
