scott@applix.UUCP (Scott Evernden) (12/09/88)
What exactly do folks expect in terms of graphics performance from X Windows? I've seen little mention of this topic since I've started reading this newsgroup, and I'm curious to hear what others have to say. I've been working with R2 (haven't gotten R3 yet) for about 6 weeks now on a B&W Sun3, and I must say I'm pretty disappointed with performance. I understand the Sun server is completely unoptimized, but I feel that even a 10 fold speed improvement won't cut it for the types of graphical interaction I am attempting. As a random example, I wanted to be able to fill a polygon with a graduated shade. In order to accomplish this, I needed to ask for several differently filled polygons (in fact, they were rects) to be drawn. The result was so slow that I seriously doubt this sort of thing can be done effectively under any X Windows server. Even simple rubber-banding of lines, rects, etc. is, I feel, unacceptably slow. I would expect the update to look as tho the line/rect is _locked_ to the mouse cursor- like on a Macintosh or my Amiga. (I believe that the X server should be able to do these and other sorts of "common" interaction techniques locally, without any intervention by the client.) Does anyone really believe that xfig, under the current Sun server, is actually even close to usable?? How about if it was 10 times faster? What is the likelihood of a useful "paint" (i.e., pixel pumping) program being developed under X Windows. How about animation? I mean, is it believed that the client/server model is appropriate to meet the demands of this type of application? What types of applications is X Window _unsuitable_ for? I've played a little bit with a "tuned" server on a different platform, and I still think the performance is marginal. Xfig is no more usable. I've look at a beta Visual XDS. I'd be unhappy with it on my desk, yet it appears some people find it's speed adequate. Certainly there must be platforms which run exceptionally well- can someone talk about them?? -scott
burzio@mmlai.UUCP (Tony Burzio) (12/09/88)
In article <878@applix.UUCP>, scott@applix.UUCP (Scott Evernden) writes: > I've been working with R2 (haven't gotten R3 yet) for about 6 weeks now on > a B&W Sun3, and I must say I'm pretty disappointed with performance. I > understand the Sun server is completely unoptimized, but I feel that even > a 10 fold speed improvement won't cut it for the types of graphical > interaction I am attempting. According to several sources I found, the SUN X server is not only unoptimized, but de-optimized (HP and NBS). > As a random example, I wanted to be able to fill a polygon with a graduated > shade. In order to accomplish this, I needed to ask for several differently > filled polygons (in fact, they were rects) to be drawn. The result was so > slow that I seriously doubt this sort of thing can be done effectively under > any X Windows server. This depends. I have an HP350 with the TurboSRX graphics accelerator. This is one fast X Windows device! Even considering that I have a beta-release of the HP version of X11, we are talking refreshes at video sync rates. The SUN X11 is pretty worthless in comparison. In addition, the SRX can be set up with X running in the foreground planes, with the background graphics planes reserved for Starbase (HP graphics library) calls. This feature (supported as an X windows subroutine library) will give you animation speed in 3D... **************************************************************************** Tony Burzio * I want my X10 xterm title bars back... Martin Marietta Labs * ****************************************************************************
dshr@SUN.COM (David Rosenthal) (12/12/88)
Tony Burzio writes: > According to several sources I found, the SUN X server is not only > unoptimized, but de-optimized (HP and NBS). > All together now, say after me: "I believe everything the salesman says to me because he has my best interests at heart" As one of the people responsible for the Sun port, and for its admittedly less-than-adequate performance I can assure you that I am not part of a conspiracy. The reasons bear repeating for the xpert listeners who weren't around in the early days when we were struggling to turn the hype into reality. At the start of the alpha-test phase of X11, Adam de Boor of UCB and I did the first ports of X11 to something other than the monochrome uVAX. We put our efforts together, and gave the code to MIT. Sun also got the server running on color hardware for the first time and donated the code to MIT. We wrote the book ("Godzilla's Guide") on porting X11, and spent considerable time helping HP use our code to start their X11 porting effort. Both the monochrome and color Sun drivers were deliberately designed to be extremely portable rather than to provide good performance. The reason for this was that during the alpha- and beta-test phases of X11, very few ports were available and thus the system was not getting adequate testing. We were trying to make it as easy as possible for others to port to their hardware, to make X11 something people could actually use. Since then, we (like others in the industry) have been building a an X11 product (in our case, X11/NeWS). We have not had effort to divert from this to speeding up the MIT sample server. Doing so is not hard - Purdue has done some work and anyone with the ability to profile the server will find obvious areas for further work. David.
burzio@mmlai.UUCP (Tony Burzio) (12/12/88)
In article <8812112304.AA07903@devnull.sun.com>, dshr@SUN.COM (David Rosenthal) writes: > As one of the people responsible for the Sun port, and for its admittedly > less-than-adequate performance I can assure you that I am not part of a > conspiracy. The reasons bear repeating for the xpert listeners who weren't > around in the early days when we were struggling to turn the hype into > reality. Hmmm... Perhaps. Do you expect X11/News to be fast? Is SUN through supporting proprietary windowing systems? Stay tuned for the race results... ****************************************************************************** Tony Burzio * I believe everyone once, except DEC, who I don't Martin Marietta Labs * believe at all... ******************************************************************************
dshr@SUN.COM (David Rosenthal) (12/12/88)
The answer to your question is, as usual, RTFM. Sun does not support the MIT sample server, as made clear in the source code, manual pages, and README. That includes (but is not limited to) not answering questions, supplying fixes, or developing drivers for obsolete hardware. You had no reason to expect X11R2 to run on a bwone - there is a list of the configurations on which it will run in the Xsun manual page, and README. Developing an X11 driver for the bwone would involve a complete new port. This is explicitly stated in the file server/ddx/sun/README. If anyone still has one, this applies also to the CG1. David.