From: Jakub Narebski <jnareb@gmail.com>
Date: Thu, 27 Oct 2011 22:32:06 GMT
Subject: Re: [PATCH 2/2] pack-objects: optimize "recency order"
Message-ID: <m3ehxy5aoe.fsf@localhost.localdomain>
In-Reply-To: <CACBZZX7tghoHhxCygEj9DZSxvKyTvybawVA2HwHBkjBaH73Ujg@mail.gmail.com>

Ævar Arnfjörð Bjarmason <avarab@gmail.com> writes:

> On Thu, Oct 27, 2011 at 23:49, Ævar Arnfjörð Bjarmason <avarab@gmail.com> wrote:
> 
> > Actually it just seems slow in general, not just on repositories with
> > a lot of tags, on linux-2.6.git with this patch:
> 
> Here's profiling with gprof for everything with >1% of execution time
> with the patch applied:
> 
>     Each sample counts as 0.01 seconds.
>       %   cumulative   self              self     total
>      time   seconds   seconds    calls   s/call   s/call  name
>      21.07     15.99    15.99  2184059     0.00     0.00  add_descendants_to_write_order
>      20.25     31.35    15.37 1146371554     0.00     0.00  add_to_write_order
[...]
> 
> And without:
> 
>     Each sample counts as 0.01 seconds.
>       %   cumulative   self              self     total
>      time   seconds   seconds    calls   s/call   s/call  name
>      21.29      9.13     9.13 142180385     0.00     0.00  hashcmp
>      10.59     13.67     4.54 90592818     0.00     0.00  lookup_object
[...]

Errr... do or do not gprof results include time spend in libraries?
Though that might not matter for this case.

Can you repeat profiling using "perf events" or something using it
(e.g. via PAPI library like HPCToolkit)?

-- 
Jakub Narębski