Re: GUILE GC -- Write barrier for vectors

[Miroslav Silovic]

Han-Wen Nienhuys wrote:

> I'm quite sure that free/malloc bytes reported are matched. However,
> most counts are a little underestimated. This is impossible to
> prevent, since we are using classes derived from smobbed C++
> baseclasses (You can't get correct object sizes in destructors and
> constructors).
>
>  

AFAIK, this shouldn't be the problem (it should only cause your memory 
usage to be
larger than necessary). Did you try to monitor and compare GC 
frequencies on both
machines (and log the bytes allocated etc.) ?

   Miro