Re: [PATCH v2 0/2] do_compare_entry: use already-computed path

From: Junio C Hamano <gitster@pobox.com>
Date: Dec 22, 2015, 18:51 UTC
Message-ID: <xmqq60zqe2pu.fsf@gitster.mtv.corp.google.com>
In-Reply-To: <1450758362.3892.7.camel@twopensource.com>

David Turner <dturner@twopensource.com> writes:

Show 18 quoted lines

> On Mon, 2015-12-21 at 15:34 -0800, Junio C Hamano wrote:
>> Great. Thanks, will queue w/o 1/2 (though I do not think it would
>> hurt).
>> 
>> On Mon, Dec 21, 2015 at 3:33 PM, David Turner <
>> dturner@twopensource.com> wrote:
>> > On Mon, 2015-12-21 at 15:27 -0800, Junio C Hamano wrote:
>> > > Thanks.  Does the number still stay at 25% improvement?
>> > 
>> > Yes.
>
> BTW, that function, via ce_in_traverse_path, gets called about 40
> million times when switching (checking out) between master and a
> branch that's a few months old (and that contains relatively small
> changes from master-as-of-then. Our repo only has approximately a
> quarter -million files.  This seems somewhat unreasonable to me,
> but I haven't really looked into what's going on.  Do you happen
> to know why this is and whether it is likely to be a bug?

That does sound excessive; smells like somebody is being overly cautious (i.e. a performance bug).

We might be seeing something similar to what e53e6b44 observed and attempted to correct. I dunno.