[clang] small speed improvement of Sema::AddArgumentDependentLookupCandidates

[Kokan]

Summary

[clang] Improve speed of Sema::AddArgumentDependentLookupCandidates

The Sema::AddArgumentDependentLookupCandidates compile the list of
overload candidates found via ADL.

Also it makes sure not to include candidates if already found via not
ADL. It achives that distinct list via looping over candidates and
removing them from the ADL candidates(Fns).

When there is no candidate found via ADL, there is no need to loop over
the candidate list as Fns is empty so there is nothing to remove from
that.

Disclaimer: I do not have compile time measurement, but not doing a loop

should be faster.

[Kokan]

clang-format applied

[whisperity]

The context of the diff is missing. Please re-run the diff making with -U9999999999999999999.

clang/include/clang/Sema/Lookup.h

817

(void)? LLVM is a C++ project.

clang/include/clang/Sema/Lookup.h

     Decls.erase(cast<NamedDecl>(D->getCanonicalDecl()));
   }
 
+  bool empty(void) { return Decls.empty(); }
+
   using iterator =
       llvm::mapped_iterator<decltype(Decls)::iterator, select_second>;
 

clang/lib/Sema/SemaOverload.cpp

   // FIXME: Pass in the explicit template arguments?
   ArgumentDependentLookup(Name, Loc, Args, Fns);
 
+  if (Fns.empty())
+    return;
+
   // Erase all of the candidates we already knew about.
   for (OverloadCandidateSet::iterator Cand = CandidateSet.begin(),
                                    CandEnd = CandidateSet.end();