Re: Making Guile slower (3.4 and 6.7%). (was Re: NIMP thing)

[Greg Harvey <Greg dot Harvey at thezone dot net>]

[snipped maciej, his mail keeps bouncing]

Mikael Djurfeldt <mdj@mdj.nada.kth.se> writes:

> Greg Harvey <Greg.Harvey@thezone.net> writes:
> 
> > > Try to find where I said that Guile got sick from NIMP food.
> > 
> > Urm... why reference the FDA, then (or is that not food & drug
> > administration over there?).
> 
> (We actually have one agency for food and a different one for drugs.)
> 
> Because they deal with risks, tests, and who has the responsibility to
> demonstrate what.
> 
> My benchmark _indicates_ that Guile _has_ got sick.  This means that
> Guile _may have_ got sick from NIMP food.
> 
> If I had made a more precise test, it may not have shown anything at
> all, in which case we would have missed that Guile has been slowed
> down due to some other of Greg's changes.  The precise test is the
> kind of test _you_ would use to show that NIMP food is safe for
> Guile.

Yeah, but I didn't think it warrented the discussion time in the
beginning, and I don't think it warrents the amount of attention it's
still getting (and it certainly doesn't warrent a few hours of
benchmarking). Yes, maybe we should profile every single change that
goes in (did the addition of goops support to the evaluator slow it
down? What about all those programs that don't use goops! The
horror!), and maybe we should worry about every instruction that gets
executed in the evaluator (while taking absurd amounts of time to
actually load the stuff being evaluated); personally, I think it'd be
a waste of time (given that there are so many other ways you could
improve the performance of guile a lot more than counting instructions
in the evaluator).

-- 
Gregh