🚀 go-pugleaf

In article <cZsNFDO00VswE1p342@andrew.cmu.edu> jaap+@andrew.cmu.edu (Jaap Akkerhuis) writes:
+ Suppose you have a (hypothetical) fast printing engine, say 100 pages a
+ minute for raw speed. Normally you have one postscript engine, and now
+ look at the speed you will get in the end. Well, since postscript is
+ interpreted, fonts might not be in the cache etc, it takes a while for
+ the pages to be processed. What is really frustrating is that you now
+ know that the printer can run on 100 pages a minute (you checked this
+ with fooling around with #copies) and you actually will see 30 pages a
+ minute or so. It is easy to imagen a setup where you have a couple of
+ processors with (quite) some disc space. Each individual job is directed
+ to a different proccessor, the bitmaps are stored to disk and the
+ printer can be fed with the bitmaps at full speed (after which the
+ bitmaps are removed) with the bitmaps of one job while the other jobs in
+ the queue are calculated. Of course, slightly more complicated setup can
+ be thought of. Note that I don't propose to take the bitmaps and store
+ them to a general purpose machine for turning them into group 3 encoding
+ or other uses. I would consider above described setup as a black box,
+ just as I consider my current printer as a (whatever colour painted)
+ black box.

Are you sure you can feed bitmaps to the marking engine that fast?  I'm
not sure, but I doubt that the bandwidth on most disks would even come
close.  Given a 300dpi frame buffer, which is roughly 1 megabyte of data,
you would have to be able to feed 100 megabytes a minute.  Maybe that's
possible.  What is the bandwidth to/from most disk drives, anybody know?

I think you just need a faster controller, or perhaps better-formed
PostScript code, to really get 100 ppm.  Maybe the disk scheme would
work, but I vaguely remember people saying that it wouldn't even work
on an 8 ppm printer, but maybe with some compression/decompression scheme
you could do it.

In any case, you raise a good point.  The bitmaps in your example are more
like a "cache", an optimization.  They still have only one purpose, to be
fed to the marking engine for which they were rasterized.

Glenn Reid