Re: [Monotone-devel] Speedup chances

[Christof Petig]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Markus Schiltknecht schrieb:
> Hi,
> 
> Christof Petig wrote:
>> The problem at hand is the malloc and memset overhead introduced by
>> using a dynamic buffer.
> 
> Huh? Why do we use memset there? malloc() hardly makes for a 40%
> speedup, does it?

Botan seems to call memset on buffer free. And since we call hex_decode
some 100k times per OE roster read, this easily adds up.

>> The botan decoder might as well be used by the specialized functions
>> which give 40% speedup, but I was unable to decode the botan manual.
> 
> So you are saying it's not botan which is slow, but our use of it,
> calling malloc() and memset() within monotone, before invoking botan,
> right?

I don't know whether botan's hex_decode is slow or fast, but using
dynamic buffering here kills performance.

   Christof

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.6 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFIH0eHng+R+0ucfO0RAq0CAJ9Y9upTZ+h0lEH+FhE0klqZlBhpNgCfWtXc
dkbGPnws98E+WXEp3/fEjgo=
=vXMq
-----END PGP SIGNATURE-----