Subject: Re: DIAGNOSTIC in -current
To: None <firstname.lastname@example.org>
From: Christos Zoulas <email@example.com>
Date: 09/28/2006 15:44:00
In article <20060928144931.GB9794@spathi.chuq.com>,
Chuck Silvers <firstname.lastname@example.org> wrote:
>On Thu, Sep 28, 2006 at 10:45:30AM -0400, Thor Lancelot Simon wrote:
>> On Thu, Sep 28, 2006 at 07:35:01AM -0700, Chuck Silvers wrote:
>> > I was going to suggest the same thing. I'll go ahead and do both of these.
>> Please measure these. If you're going to make them always functions, you
>> may want to make them inline with shadows in libkern, because I think we
>> measured the cost of not using the macros at one point and found it higher
>> than we expected.
>what benchmark would you suggest?
>most of the correct usage of the macros (for constant-sized allocations)
>has been replaced by pools already, so the difference will be much less
>than in the past.
All the MALLOC users are now using constant-sized allocations (this is
enforced by the definition of MALLOC).