> > Something like that, yes.  But why does the memory allocator get slower
> > and slower?  Isn't this some kind of a bug?
>
>That depends.  It could be that Applixware itself is doing lots of mallocs
>and frees, maybe using an mmap-based allocator.  This can fragment the
>VM map for that process.

Hmmmm .... doesn't the X server use a mmap()-based allocator?  That
may be part of the problem (I seem to remember that from a while ago).

Now, is it possible to example the VM map for a process?  That would
help determine if this is the root cause.

>UVM doesn't currently collapse adjacent VM map entries .... partially
>because doing that is also expensive ... Although, a clever implementation
>could take hints about where to begin and end the collapse... chalk one
>up on the TODO list...

Maybe collapsing VM map entries could be done if you got past a certain
fragmentation point .... again, something for the TODO list :-)

--Ken