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

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

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

> Greg Harvey <Greg.Harvey@thezone.net> writes:
> 
> > There's a big difference between:
> > 
> > 1) Hey, look, all these people ate this food and got sick. Something's
> >    wrong with that food.
> > 
> > 2) All these people ate a wide variety of different foods, and became
> >    sick. xxx food was the culprit, because I don't like xxx food.
> 
> 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?). I was just extending an analogy ;)

> I think it is Greg's task to show that a particular change he made was
> safe.  My goal here is that Guile's efficiency and stability should be
> preserved.
>
> Which test is best to do then?

If it does break things (outside of silly bugs/typos) then those
things were already broken (testing NIMP on something that we're about
the treat as a non-immediate should not break things). He very likely
did change too much, too quickly, without getting someone else to look
over the code before applying it to the cvs tree, but that's not
really what we're talking about here (however, he did discuss the NIMP
change before making it). We are talking about whether the NIMP change
caused a noticable performance degradation.

-- 
Gregh