Re: multispectral file access

Testing shows that for a 5000x5000 TIFF file, most of the time spent 
in the vips benchmark is spent reading/writing the TIFF file (both 
reading/writing are similarly slow), with hardly any time being spent 
in the crop and convolution algorithms.

Using a different format such as PNM or MIFF turns out to be quite a 
lot faster.

It is not clear why TIFF is so slow.  Perhaps GM is using non-ideal 
access patterns.  There is a bit less performance when using tiled 
TIFF since GM's internal image representation is not tiled, but the 
difference is not dramatic compared with the overall slowness.

Bob
-- 
Bob Friesenhahn
[email protected], http://www.simplesystems.org/users/bfriesen/
GraphicsMagick Maintainer,    http://www.GraphicsMagick.org/

------------------------------------------------------------------------------
Slashdot TV.  
Video for Nerds.  Stuff that matters.
http://tv.slashdot.org/