This is the mail archive of the gcc@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: Expenses of unit-at-a-time (Re: On inlining in C++ with unit-at-a-time code(


> Jan Hubicka <hubicka@ucw.cz> writes:
> 
> [...]
> 
> | Of course it is possible to construct testcases where unit-at-a-time
> | takes much more memory than normal mode by creating large file
> | consisting of many small funcitons but I tend to believe that this is
> | rare in C and in C++ frontend already deffers most of these so the
> | difference is even smaller.
> 
> I can't parse that.
> 
> It is quite common (in the C++ community) to have large translation
> units with many small functions (usually candidates for inlining).

Yes, however this is the common case where unit-at-a-time helps.  In
majority cases I seen there is huge amount of inlining candidates and
relatively few of them are used.  unit-at-a-time releases all the
functions bodies that are not needed before assembling causing reduced
memory footprint.  Combined with improved inlining heuristics times it
is usually a win for C++. (usually means that I don't know about real
world testcase where it is not but I can construct artificial one for sure)

In gneral unit-at-a-time seems to help more to C++ compile time
performance than C compile times.  This is because C++ deferred many
functions anyway getting memory wise the worse of both worlds.

Honza
> 
> -- Gaby


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]