This is the mail archive of the gcc-patches@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: [PATCH][C++] Improve PR15855 (compile-time/memory hog)


On Fri, 2005-09-23 at 11:33 -0700, Mark Mitchell wrote:
> Richard Guenther wrote:
> 
> > After some sleep and coffee I re-structured the thing even a little
> > more and came up with the following patch, which looks reasonably
> > easy to follow. 
> 
> This version is OK for mainline.  After it's baked in for a week or so,
> and assuming the 4.0.2 release has gone out, please apply it to the 4.0
> branch for 4.0.3 as well.
Cool.

Now if someone wants to make another huge improvement to 15855, they
should probably look at how we're handling SFTs and how they're 
exploding in the incremental SSA updates -- there's something like
5 million mappings for SFTs for one of the incremental SSA updates.
We may have algorithmic issues with that many mappings and we may
be doing some stupid stuff in creating that many to begin with.

That's the most likely reason why 15855 has regressed so badly  in
compile-time and memory-usage. 

jeff




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