Aaron Lehmann wrote:
> Wow. Keep up these small-but-cool commits, especially during
> optimization week!
Yes, I noticed I took the lead in a rapid optimization sweep :-)
But I still think Thomas' vector stuff will be *very* hard to beat (two
orders of magnitude isn't something you do everyday :-) ).
> I'm hoping you'll be able to help us out becuase it seems you have the
> M$ tools, which probably includes some sort of profiler.
I use a better profiler that gives even a graphical overview of call
hierarchies and such. :-)
> The GNU profiler sucks. Perhaps you will be able to track down
> slowness in the XP code.
The "native" MSVC profiler is probably just as crappy. One question: does
gcc have an option to insert a "function_enter" (or something like that, and
a function_leave would also be helful) call to every compiled function? If
so, I could probably supply a rather decent profiling framework within a
matter of days. If gcc doesn't have such an option, perhaps someone should
suggest it to the gcc team?
/Mike - please don't cc
