A survey about tile compression algorithm - results

Dmitry Kazakov dimula73 at gmail.com
Sat Jun 26 21:27:27 CEST 2010


> Ooh! Numbers! But math-challenged as I am -- what conclusions do you
> suggest?
>

1) For most purposes, including swapper, we need compression only (undo is
not really a common usecase). So LZF with linearization is the best choice.

2) Linearization of colors makes compression ratio and compression speed
better. It injures decompression speed a bit (8%), but as defined in 1), it
is not a common usecase for us.

3) The tile fits into the cache of my cpu quite well, so memcpy operation on
a single tile is very cheap! It is 1/90 of compression time and 1/40 of
decompression time.

4) The best compression we can achieve on full-color images (like a tile
from hakonepa.png) is 75% (compressed/uncompressed*100%)

5) A tile, filled with color will be squeezed to 5% of original.

-- 
Dmitry Kazakov
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.kde.org/pipermail/kimageshop/attachments/20100626/4ddc5797/attachment.htm 


More information about the kimageshop mailing list