mlprof is right (I think) (but I have improvements)

[Stephen Weeks]

> As far as -s being on by default, I could be convinced.

Whap!  Whap!

Are you convinced?

> The big question for the moment is: do we believe these results.  I did a few
> tests (the roundPage thing) and saw that it was, indeed, correct.  If you have
> any other specific thing you would like to question (like with all the flags
> on, a specific number) I can do that by hand and see if it is correct.
> I agree that it does look pretty rosy to believe it, but I am starting to
> pretty strongly.  After all, it always was some what surprising why self
> compile didn't speed up by more.  Perhaps this is the answer.

Perhaps.  I will try a run of the kit and of fxp this weekend to see
if there are similar numbers.