This is the mail archive of the
guile@cygnus.com
mailing list for the Guile project.
Re: Bad news from the module/environment front
Why isn't Aubrey interested in making all his improvments to guile
instead of SCM? One would have thought that all the development of guile
in other areas would be of benefit to him and vica-versa. Does he just
like having full control of his project or something?
Jim Blandy wrote:
>
> Yes, SCM has gotten even faster since Guile forked off. In principle,
> I'd love to see Aubrey's improvements integrated into Guile ---
> especially the delayed consing for local environments.
>
> However, either:
> 1) Aubrey often chooses speed over maintainability, or
> 2) Aubrey is much, much more clever than I am, and thus his code is
> perfectly maintainable to him, while it sometimes stumps me for
> hours.
> Possibly both. In either case, I find his code difficult to work
> with. But don't let that keep you from integrating his speedups.
>
> >SCM old guile evaluator new environment-evaluator
> >
> >14100 (10 in gc) 21520 (3430 msec in gc) 32110 (13840 msec in gc)
> >14040 (20 in gc) 21620 (3470 msec in gc) 31910 (13820 msec in gc)
> >14790 (40 in gc) 20960 (3640 msec in gc) 32070 (13940 msec in gc)
> >14030 (20 in gc) 21100 (3480 msec in gc) 33220 (13960 msec in gc)
>
> I think it's better to separate out the GC for purposes of comparison:
>
> old guile evaluator new environment-evaluator
> 18090ms + 3430ms in gc 18270ms + 13840ms in gc
> 18150ms + 3470ms in gc 18090ms + 13820ms in gc
> 17320ms + 3640ms in gc 18130ms + 13940ms in gc
> 17620ms + 3480ms in gc 19260ms + 13960ms in gc
>
> This makes it clearer that almost all the slowdown is due to increased
> GC time. Why does the new environment system allocate more than the
> old one? Is the memory image larger as well? (I hope so, or else I
> am very lost.)