Path: utzoo!utgpu!jarvis.csri.toronto.edu!cs.utexas.edu!swrinde!zaphod.mps.ohio-state.edu!uakari.primate.wisc.edu!aplcen!uunet!mcsun!ukc!pyrltd!advsys!brian
From: brian@advsys.UUCP (Brian Rippon)
Newsgroups: comp.lang.postscript
Subject: Re: 9600 bps bottleneck on Laserwriter Plus
Message-ID: <564@purdey.advsys.UUCP>
Date: 17 Jan 90 14:56:01 GMT
References: <10213@microsoft.UUCP> <15469@well.UUCP>
Reply-To: brian@advent.co.uk (Brian Rippon)
Organization: Advent Systems Ltd, Wokingham, England
Lines: 29

In article <15469@well.UUCP> Jef Poskanzer <jef@well.sf.ca.us> writes:
>That was here.  As I recall, I posted my version of the PostScript
>decompression routine ...			 and asked whether any
>better PostScript hackers could come up with a more optimal routine
>that might be a win.  I got about five different versions posted or
>mailed, and none were faster than uncompressed.

I wasn't reading news then, so forgive me if this has been covered.
Was the FrameMaker technique discussed? The PostScript is Copyright
Frame Technology, Inc, so you'll have to find your own Frame output
to see what I mean (how about it Frame?). Essentially they run-length
encode short runs to a byte, allow unencoded strings prefixed by a
byte, and repeated bytes as a byte followed by the pattern. The decode
is performed by a lookup from the byte to an array of procedures.

Trying the technique on a typical(?) Sun screen (several text windows,
mailtool, dbxtool, some grey background), it generates 37k of PS, compared
with 260k using the normal method. Considering that the compress'd version
of the latter is 22k, that's not bad. In Group 3, it's 70k!

Running a QMS PS-800+ from a Sparc at 38400, I get execution times
of 55 seconds compared with 67 seconds for uncompressed. As Jef says,
not worth it on AppleTalk or Centronics, but I don't have a choice;
anyway Centronics from a PC is no great shakes.

I don't know how this compares with Jef's code - I haven't found
a local copy of PBMPLUS yet :(.

Brian.