Re: memory_object_lock_request and memory_object_data_return fnord

[Thomas Bushnell, BSG]

Neal H Walfield <neal@cs.uml.edu> writes:

> > I think we should certainly use vm_copy for whole-page copies in
> > pager_memcpy because of the badly suboptimal behavior you've described.
> 
> I have cooked up the attached implementation.  I checked everything
> but a few border cases -- I need to write a few more tests.  Perhaps,
> I will get to it this weekend.  However, I am interested in some
> feedback.

One thing I think of looking at.

It may well be that for smaller number of pages, memcpy is actually
more efficient than vm_copy because of the constant overhead in the
latter function.

So I'd like to see some instrumentation of where that cutoff actually
is.  

Until that's done, I have no problem with assuming any particular sane
value for the cutoff, but I think your function should have a tunable
parameter for it.

Thomas