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: Improving gcc scalability wrt _extra_ large C files


Jacques THOMAS wrote:-

> I am not sure that I understand exactly what you mean, but I am afraid 

What isn't clear?

> that --using this method-- we will still have to do a lot of  list 
> traversing. IMHO, this should still be very costly when the global scope 
> contains over 300k symbols (what I am trying to get cc1plus to compile 
> in a more decent time).

No, it won't cost anything for globals, and there is only a single list
traversal per scope exit, of length the number of identifiers declared
in that scope.  At present, we have this traversal per scope exit,
*and* per identifier lookup.  In other words, what I suggest is vastly
more efficient, uses less space, and less hash tables.

Neil.


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