Path: utzoo!utgpu!news-server.csri.toronto.edu!eecg.toronto.edu!drb
Newsgroups: comp.sys.sgi
From: drb@eecg.toronto.edu (David R. Blythe)
Subject: Re: Getting compor map colors.
Message-ID: <1991Apr29.020915.7092@jarvis.csri.toronto.edu>
Keywords: getmcolor
Organization: EECG, University of Toronto
References: <1991Apr22.153229.7358@cs.ruu.nl> <80085@bu.edu.bu.edu> <1991Apr25.130848.25198@cs.ruu.nl>
Date: 29 Apr 91 06:09:15 GMT
Lines: 46

In article <1991Apr25.130848.25198@cs.ruu.nl> markov@cs.ruu.nl (Mark Overmars) writes:
>In <80085@bu.edu.bu.edu> tjh@bucrf11.bu.edu (Tim Hall) writes:
>
>>In article <1991Apr22.153229.7358@cs.ruu.nl>, markov@cs.ruu.nl (Mark Overmars) writes:
>>|> I have the following problem. I want to draw colormap images inside a RGB
>>|> window. As a result I need to find out what the colors in the colormap are. To
>>|> this end I thought I should use the routine
>>|> 
>>|>   getmcolor()
>>|> 
>>|> Unfortunately, this routine only work when you are in colormap mode, not when
>>|> you are in RGB mode. (Why?) 
>>
>>	current_window = winget();   /* Remember what the current window is */
>>
>>	noport();                    /* Open a window that never appears */
>>	color_window = winopen( NULL);
>>
>>	Read the color map here.
>>
>>	winclose( color_window );
>>
>>	if ( current_window > 0 )  /* >= ????   I forget.... */
>>		winset( current_window );
>
>The problem with this solution is that it is way too slow. Even keeping the
>color window open and doing a winset on it whenever I need a colormap entry
>slows the drawing of e.g. rectangles down by a factor of three (winset seems
>to be an expensive call), at least on a GTX. On a 25/TG there was almost no
>speed penalty (????). What I do now is to keep my own colormap and reading it in
>only once. Of course, in this way I don't see it when the application changes
>the colormap but it works fast. (Strangely enough, doing this on a 25/TG was
>faster than drawing normal in colormap mode.)
>
>Mark Overmars

I saw a performance posting a couple of weeks ago that showed wildly varying
performance with colormap manipulation across the various platforms, I think
the PI was faster than the GT (then again maybe not).  I think the keeping
the window open solution should get even faster as the pressure to have
cheap (X-like) windows is answered (in 4.0?).  On the other hand, having
a window manager color map changed event also would be handy and is probably
necessary for total X'ness anyway ...
	-drb