[bug-notifications] [gv 0004368]: gv very slow

[Mantis Bug Tracker]

A NOTE has been added to this issue. 
====================================================================== 
http://www.opencsw.org/bugtrack/view.php?id=4368 
====================================================================== 
Reported By:                wcohrs
Assigned To:                james
====================================================================== 
Project:                    gv
Issue ID:                   4368
Category:                   regular use
Reproducibility:            always
Severity:                   tweak
Priority:                   normal
Status:                     assigned
====================================================================== 
Date Submitted:             2010-03-23 13:30 CET
Last Modified:              2010-03-23 15:46 CET
====================================================================== 
Summary:                    gv very slow
Description: 
gv is very slow SPARC and x86 on a Server with  Quad-Core AMD Opteron(tm)
Processor  8380 2500MHz needs gv several minutes but gs needs only few
seconds.
the same picture on a Linux Server with gv needs also only few seconds.

====================================================================== 

---------------------------------------------------------------------- 
 (0007730) james (manager) - 2010-03-23 15:46
 http://www.opencsw.org/bugtrack/view.php?id=4368#c7730 
---------------------------------------------------------------------- 
The anti-alias device used in gv is what is making gv slower.  Turn off
anti aliasing and see the difference.   Simulate the effect directly with
Ghostscript:

gs -sDEVICE=x11alpha test.ps


Try alternate aliasing:

gs -dGraphicsAlphaBits=1 -dTextAlphaBits=1 test.ps
gs -dGraphicsAlphaBits=2 -dTextAlphaBits=2 test.ps
gs -dGraphicsAlphaBits=4 -dTextAlphaBits=4 test.ps


You can set the gs parameters in gv.

The underlying problem is the Postscript itself, it just slow.  It's
plotting dots by drawing zero length lines with the linecap set to round. 
Try the dot or as procedure, gylph or similarly cached bit map, user path
or form.  Even then, it's plotting the dots over each other, I expect the
whole thing can be improved but decide if you want to spend time writing a
better prologue, time writing better output, or spend time waiting for the
Postscript interpreter.