[Ffmpeg-devel] [PATCH] simple internal lzo decoder

Jindrich Makovicka makovick
Sun Jan 15 21:39:46 CET 2006


Reimar D?ffinger wrote:
> Hi,
> On Sat, Jan 14, 2006 at 04:21:44PM +0100, Reimar D?ffinger wrote:
> 
>>Also instead of the last patch I'd like to propose the attached one,
> 
> 
> A last try, the speed is really nice (though only tested on my system).
> But it needs a bit of padding for the buffers - since it is done in
> other place I guess this is okay even if I personally don't like it
> much.
> Ok, here are the values: first comes liblzo, then minilzo and then this
> implementation with the patch. And then the same but for gcc 4.0.0 (I
> know, outdated, but that's what I had).
> 
> file all zeroes, 56325B comp:
> 8.90 8.90 0.80 8.87 8.46 0.75

I am getting ~3.1s here for minilzo and ~1.7s for av-lzo on XP-M
2600+ at 2.1GHz. Either I am doing something totally wrong or minilzo just
sucks on 64-bit machines. I was using gcc-3.4, patched lzo.c and
lzo1x_decompress_safe from minilzo to be fair.

Relevant C proggy is available at

http://kmlinux.fjfi.cvut.cz/~makovick/lzobench.tar.bz2

> nwnmod, 2675408B comp:
> 7.39 6.98 6.87 7.35 7.18 6.39
>
> /dev/urandom, 10527740B comp:
> 4.28 4.13 1.72 4.28 4.24 1.73

On tarred MPlayer sources av-lzo was marginally slower than minilzo. And
a little bit faster for /dev/urandom - about 2.52s vs. 2.56s.

Regards,
-- 
Jindrich Makovicka





More information about the ffmpeg-devel mailing list