Re: [Chicken-users] nursery logic

[felix winkelmann]

On 7/15/06, Brandon J. Van Every <address@hidden> wrote:
> Makefile.am has a script that configures chicken-defaults.h with a
> nursery value.  The following condition will never be reached:
>
>       if [ $$bestn -eq 0 ]; then \
>             nursery="256*1024"; \
>             echo "hm, no decent nursery value here ... using 256K"; \
>
> because
>
>         if [ $$bestn -eq 0 ]; then \
>               best=$$tm; \
>               bestn=$$n; \
>
> is always executed on the first pass,  and n is never 0.  n is a member of
>
>       for n in 16 24 32 48 64 128 256 512 1024; do \
>
> I point this out in case the desired logic is other than the actual
> logic.

I'm just a hopeless shell programmer, so don't expect too much from
the pieces of shellcode in the makefile. If you (or someone else) can
code this in a more elegant fashion, please go ahead.

>  Also I wonder whether you want less-than-or-equal or simply
> less-than:
>
>         elif [ $$tm -le $$best ]; then \
>           best=$$tm; \
>           bestn=$$n; \
>
> less-than-or-equal says "if all things are the same, then prefer a
> bigger nursery."  less-than would say "...prefer a smaller nursery."

Well, a bigger nursery would be preferable in any case (less major GCs).
This "nsampling" takes a very simple benchmark (browse) and tries
to estimate a good nursery value but its just a simple test.


cheers,
felix