[PATCH] D20401: [Lexer] Don't merge macro args from different macro files

Thu May 20 10:19:30 PDT 2021

vsk added a comment.

In D20401#2770059 <https://reviews.llvm.org/D20401#2770059>, @nickdesaulniers wrote:

> I know this was sped up slightly in 3339c568c43e4644f02289e5edfc78c860f19c9f, but this change makes `updateConsecutiveMacroArgTokens` the hottest function in clang in a bottom up profile of an entire build of the Linux kernel.  It thrashes the one entry LastFileIDLookup cache, and we end up looking up the same FileID again and again and again for each token when we expand nested function like macros.
>
> Is there anything we can do to speed this up?  Is there any way to record which FileID corresponds to a given Token so that we don't have to keep rematerializing that?  Is it possible to find whether two SourceLocations correspond to the same FileID without having to figure out which FileID in particular each belongs to?

Perhaps you could try:

- using SourceManager::isInFileID(NextLoc, getFileID(CurLoc), ...) (to halve the number of necessary getFileID lookups), or
- using a 2-element cache in getFileID?

>> I discussed this bug with Argyrios off-list, who lgtm'd on the condition that it doesn't introduce a performance regression.
>
> Well, I'd say it's a performance regression, though perhaps reported 5 years too late.

If the performance issue manifests on Linux kernel sources from 5 years ago, then sure, I'd agree :).

Repository:
  rL LLVM

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D20401/new/

https://reviews.llvm.org/D20401