Re: COPY_DELAY could perform worse than COPY_NONE

[Sergio López]

2011/10/8 Richard Braun <rbraun@sceen.net>:
> On Sat, Oct 08, 2011 at 01:55:42AM +0200, Sergio López wrote:
>> Nowadays, the copy strategy is decided per-object, but I think the
>> desired type of access to an object should also be considered. For
>> instance, the operation generated by vm_copy should almost always use
>> COPY_NONE, while vm_map, when doing things such as providing a private
>> map, should use COPY_DELAY.
>
> I really don't see why vm_copy() should use COPY_NONE. Could you explain
> it briefly (sorry if I'm asking about something you may already have
> described, but I feel it would be better with a clear, simple,
> centralized explanation).

vm_object_copy_delayed(), which is being used when copying from an
object with the MEMORY_OBJECT_COPY_DELAY attribute, is not designed
for doing small, repetitive reads from an object (which is what we're
usually doing with vm_copy). Using it this way is slower than copying
the pages without any optimization (as with MEMORY_OBJECT_COPY_NONE).