I have two comments regarding this CL: - You un-inlined function Size() which is not very heavily used by most benchmarks but is heavily used in GC. I applied your patch and ran oldspace. I observe 20% slowdown of marking phase which uses Size() heavily. [this BTW corresponds to 5% degradation of oldspace on golem].
So I would not recommend un-inlining Size() [at least for now]. - If you move methods' definitions from objects-inl.h to objects.cc please make sure that you remove 'inline' modifier from methods' declarations. Otherwise (some) compilers will complain that they do not see definitions of used inline functions [probably you just forgot to include objects.h into CL]. -- Vyacheslav Egorov On Fri, Jul 30, 2010 at 4:05 PM, <[email protected]> wrote: > Reviewers: Vladislav Kaznacheev, > > Message: > lib8.a compiled in Linux reduced ~1%. Performance did not degrade. > > Description: > Out of lining some methods. > > Please review this at http://codereview.chromium.org/3015042/show > > SVN Base: http://v8.googlecode.com/svn/branches/bleeding_edge/ > > Affected files: > M src/ia32/virtual-frame-ia32.h > M src/ia32/virtual-frame-ia32.cc > M src/objects-inl.h > M src/objects.cc > M src/virtual-frame-heavy-inl.h > M src/virtual-frame-heavy.cc > > > -- > v8-dev mailing list > [email protected] > http://groups.google.com/group/v8-dev > -- v8-dev mailing list [email protected] http://groups.google.com/group/v8-dev
