Re: guile/guile-core/libguile init.c

[Mikael Djurfeldt <mdj at mdj dot nada dot kth dot se>]

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

> Approx. execution time for (eval2 500000):
> 
> Old Guile               46 s
> Double cell Guile       18 s
> New GC scheme            7 s

Sorry, I think you shouldn't trust these values.  I've now discovered
strange things happening with memory on my system (Debian Woody on
Pentium).

Particularly, one cannot trust the above timings.  It seems that if
there is a certain kind of load on the system, reported execution time
gets smaller than the real value.  So, if the GC manages to load the
system in a special way, e.g. by exercising malloc, it can *fake* a
smaller reported execution time in the benchmark suite!

We have to check this.  Maybe we should measure time in some different
way.