Re: More precise tag following

From: Jeff King <peff@peff.net>
Date: Jan 27, 2007, 13:33 UTC
Message-ID: <20070127133352.GB2417@coredump.intra.peff.net>
In-Reply-To: <7vbqkklv3h.fsf@assigned-by-dhcp.cox.net>

On Sat, Jan 27, 2007 at 12:41:54AM -0800, Junio C Hamano wrote:

Show 7 quoted lines

> > Based on some (limited) profiling with Shark it seems we spend about
> > 50% of our CPU time doing zlib decompression of objects and almost
> > another 14% parsing the tree objects to apply the path limiter.
> 
> I once tried to use zlib compression level 0 for tree objects
> and did not see much difference -- maybe I should dig it up and
> find out why.

I don't know exactly what Shawn meant, but a considerable amount of time in a blame is spent decompressing the blobs. Just for fun, some numbers:

Fully packed, warm cache, core.compression = -1: $ time git blame Makefile >/dev/null real 0m5.537s user 0m5.500s sys 0m0.032s

Fully packed, warm cache, core.compression = 0: $ time git blame Makefile >/dev/null real 0m3.001s user 0m2.984s sys 0m0.012s

That's 45% savings. The resulting pack sizes are 11932K compressed and 22308 uncompressed.

-Peff