[FFmpeg-devel] [RFC] cpu specific implementation of libavutil functions

Michael Niedermayer michaelni
Sun Sep 9 16:59:40 CEST 2007


On Wed, Sep 05, 2007 at 05:10:50PM +0200, Luca Barbato wrote:
> Recently I got an altivec adler32 implementation from Sigbj?rn Skj?ret
> and the main issue is how to plant it into libavutils and find a sane
> general way.

how much faster is that adler32 code?
adler32 really should be limited by memory speed not by cpu

> Possibilities:
> - make the av_* a pointer and set it to default to the _c
> implementation, have an av_init_* function to switch it to the other.
> - duplicate completely dsputils (keep a context, init it somewhere, pass
> it to the av_* function or just keep it inside it...)

1. the different libavutil parts MUST stay independant iam very strongly
against turning libavutil into a intermingled mess where its impossible
to use a part without useing everything, so a solution with one big dsputil
like context is not ok

2. do we even have a working altivec detection routine? that what we do
in dsputil is not what i would call working and i will not accept such
hacks in libavutil

> - something better that I hadn't write/thought yet.

well ... :)

> - move dsputil in a standalone library


Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

I wish the Xiph folks would stop pretending they've got something they
do not.  Somehow I fear this will remain a wish. -- M?ns Rullg?rd
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: Digital signature
URL: <http://lists.mplayerhq.hu/pipermail/ffmpeg-devel/attachments/20070909/1fe1d880/attachment.pgp>

More information about the ffmpeg-devel mailing list