>I'd be interested in hearing about this - I'm somewhat assuming that the
>bottleneck might be the construction of the dynamic huffman tables (a piece
>of code that I couldn't be bothered to write in Slang ;-) but so far I
>couldn't see it eat up more than say 30% of the time. If you have any
>counter examples I'd be glad to see them.

That's probably the meat of it, and it's relatively constant per image,
which makes it a proportionally heavier penalty for small images.  After we
merge Bob's and Eric's code together, we'll redo the tallies and see what
pops out.  Bob's last email seems to confirm what we had seen before.

-- Duane