This is an archive of the discontinued LLVM Phabricator instance.

[clangd] Get rid of AST matchers in SymbolCollector. NFC
ClosedPublic

Authored by ioeric on Sep 14 2018, 5:54 AM.

Details

Summary

The AST matcher pops up in performance profile. Although not the most expensive thing we could optimize, this still brings indexAST latency from ~5.3s to ~4.4s for a large TU with big preamble.

Diff Detail

Repository
rL LLVM

Event Timeline

ioeric created this revision.Sep 14 2018, 5:54 AM
This revision is now accepted and ready to land.Sep 14 2018, 6:19 AM
ioeric edited the summary of this revision. (Show Details)Sep 14 2018, 11:51 PM

why?

I wanted to get some numbers and update the patch summary, but somehow forgot. Sorry about that and thanks for asking!

The AST matcher pops up in performance profile. Although not the most expensive thing we could optimize, this still brings indexAST latency from ~5.3s to ~4.4s for a large TU with big preamble.

why?

I wanted to get some numbers and update the patch summary, but somehow forgot. Sorry about that and thanks for asking!

The AST matcher pops up in performance profile. Although not the most expensive thing we could optimize, this still brings indexAST latency from ~5.3s to ~4.4s for a large TU with big preamble.

Thanks! That's what I was looking for :) Sounds good to me.

This revision was automatically updated to reflect the committed changes.