slowness of Cpp::EnvironmentFile::merge

[Milian Wolff]

Milian Wolff, 21.12.2010:
> Hey all, esp. David:
> 
> According to this:
> https://userpage.physik.fu-berlin.de/~milianw/callgrind.out.4728.bz2
> 
> the Cpp::EnvironmentFile::merge is responsible for 13.4% of the time spent
> parsing a part of boost (duchainify --force-update-recursive
> boost_1_45_0/libs/serialization/performance/xml/string256_test.cpp). Is
> this sane? I'll redo the callgrind in RelWithDebInfo soon but I hope we
> could improve that even now.
> 
> What I don't quite get are the lines like these:
> 
>   d_func_dynamic()->m_strings = (d_func()->m_strings + (file.d_func()-
> 
> >m_strings - d_func()->m_definedMacroNames)) - d_func()-
> >m_unDefinedMacroNames;
> 
> Won't that lead to duplicated entries? Essentially that is:
> 
> A = (A + (A - B)) - C
>   = 2 A - B - C
> 
> so all items in A not contained in B and C will show up twice (or the Set
> is indeed a Set and prevents that internally). Still, this should be
> faster and more clear, no?:
> 
> A -= B;
> A -= C;

this doesn't hold though:

  ReferenceCountedStringSet strings2 = (d_func()->m_strings + (file.d_func()-
>m_strings - d_func()->m_definedMacroNames)) - d_func()-
>m_unDefinedMacroNames;
  d_func_dynamic()->m_strings -= d_func()->m_definedMacroNames;
  d_func_dynamic()->m_strings -= d_func()->m_unDefinedMacroNames;
  Q_ASSERT(strings2  == d_func_dynamic()->m_strings);

I wonder why - is the set not a set after all? I.e. can it contain duplicates?

Bye
-- 
Milian Wolff
mail at milianw.de
http://milianw.de
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 198 bytes
Desc: This is a digitally signed message part.
URL: <http://mail.kde.org/pipermail/kdevelop-devel/attachments/20101221/61a2373c/attachment-0001.sig>