[cfe-commits] r61929 - in /cfe/trunk: Driver/RewriteObjC.cpp include/clang/AST/DeclBase.h include/clang/AST/DeclObjC.h lib/AST/DeclBase.cpp lib/AST/DeclCXX.cpp lib/AST/DeclObjC.cpp lib/CodeGen/CGObjCGNU.cpp lib/Sema/IdentifierResolver.cpp lib/Sem

[Daniel Dunbar]

I don't really understand your message Fariborz. My tester runs with a
Release-Asserts build so it should be at -O3 if you are referring to
the actual build of clang itself.

And I'm not really sure how this related to the commit, I had hoped
that there would be a speedup due to the more efficient lookup
mechanism.

 - Daniel

On Thu, Jan 8, 2009 at 5:31 PM, Fariborz Jahanian <fjahanian at apple.com> wrote:
>
> On Jan 8, 2009, at 5:20 PM, Daniel Dunbar wrote:
>
>> Hi Steve,
>>
>> My tester is pointing at this for a .98% slowdown at -fsyntax-only (of
>> parts of Sketch, essentially Cocoa.h). I was really hoping we were
>> going to see a speedup from this! Can you investigate?
>>
>> - Daniel
>
>
> Isn't construction of deeply nested inherited objects (I don't know the
> right phrase) more time consuming due
> to number of constructors called? Are these constructor calls eventually get
> inlined and thus with -O
> do not show any performance degradation (or at least this is the hope)?
>
> - fariborz
>
>
>