On 7/28/05 9:41 PM, David A. Lyons didst favor us with:
> On Jul 28, 2005, at 12:07 PM, David Duncan wrote:
>
>>> You shrink it with SetHandleSize() first.
>>>
>>> h = NewHandle( 1000 );
>>> SetHandleSize( h, 100 );
>>> SetHandleSize( h, 1000 ); <- no allocation necessary.
>>>
>>
>> There is also the potential that your original allocation was
>> rounded up. I think the current threshold is 12K - allocations
>> greater than that are page rounded, so if you asked for say 13K,
>> then your allocation would be 16K - if you never grow beyond 16K
>> the handle won't move.
>
> Have we crossed back over the line, yet, into implementation details
> that can't be counted on?
>
> I'm willing to believe there is documentation saying "handles in Mac
> OS X will never move unless you grow them, so you pretty much never
> need HLock any more" but is anything beyond that actually guaranteed?
No. I would never *rely* on a handle not being grown in the example above, I
was only explaining why I worded my comment the way I did. However, the
current implementation is optimized for performance and I don't expect that
to change. Actually releasing memory when downsizing a handle would require
a new allocation and copying the current data, and certainly there's no
reason the reallocate and copy if the currently allocation is large enough
to hold the new size.
Larry
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Carbon-dev mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
http://lists.apple.com/mailman/options/carbon-dev/email@hidden
This email sent to email@hidden
