hpk@vax135.att.com (Howard P Katseff) (09/25/89)
Has anybody thought about how to speed up the X window system by distributing the X server over a multiprocessor network? I recall reading about this topic in an X document. Is anybody working in this area? I would appreciate receiving any pointers to past or current work. -- Howard Katseff Room 4G-622, AT&T Bell Laboratories, Holmdel, NJ 07733 UUCP: att!vax135!hpk INTERNET: hpk@vax135.att.com PHONE: 201 949-5337
hpk@cbnewsh.ATT.COM (howard.p.katseff) (10/11/89)
I recently sent out a request for information about how to speed up the X window system by distributing the X server over a multiprocessor network. I got one useful reply: a pointer to a Masters thesis by Jeff Kelley (uunet!watmath!watcgl!jvkelley). Here is the information that I received: Abstract This thesis discusses the design and implementation of a window system for a tightly coupled multiprocessor workstation. The implementation is based on the X Window System and runs under Harmony, a multiprocessor, multitasking, real-time operating system. Bottlenecks inherent in the design of the X Window System limit the use of parallelism. To some extent, modifications to the current implementation could take better advantage of the multiprocessor architecture, but certain fundamental bottlenecks cannot be eliminated. Based upon an analysis of the X Window System and other window systems, a new window system design is proposed. This new design is intended to allow for exploitation of the maximum potential parallelism in window system functions, primarily by explicitly separating screen management functions from graphics rendering functions. End Abstract Just to make sure that there isn't any misconception, I'll point out that I don't provide specific solutions to the bottlenecks present in X (i.e. the sample server from MIT), but rather describe a generic window system architecture more appropriate for a tightly coupled multiprocessor (i.e. frame buffer(s) and rendering processors on a shared bus). Note that this is a description of an architecture, not an implementation (yet). Most of the interesting parts of the thesis were turned into a paper published in the Graphics Interface '89 proceedings. (What was left out was mostly background information on Harmony, window systems, X, as well as the description of the X implementation under Harmony.) If you can find a copy of the proceedings you may find that adequate. -- Howard Katseff Room 4G-622, AT&T Bell Laboratories, Holmdel, NJ 07733 UUCP: att!vax135!hpk INTERNET: hpk@vax135.att.com PHONE: 201 949-5337
jkrueger@uunet.UU.NET (Jon) (10/13/89)
hpk@cbnewsh.ATT.COM (howard.p.katseff) writes: >I recently sent out a request for information about how to speed up the >X window system by distributing the X server over a multiprocessor >network. >I got one useful reply: a pointer to a Masters thesis by Jeff Kelley >(uunet!watmath!watcgl!jvkelley). >Abstract >This thesis discusses the design and implementation of a window system >for a tightly coupled multiprocessor workstation...Bottlenecks >inherent in the design of the X Window System limit the use of >parallelism. You asked for a distributed X. You cite discussion of a parallel X. Which do you want? -- Jon -- Jonathan Krueger jkrueger@daitc.daitc.mil uunet!dgis!jkrueger Isn't it interesting that the first thing you do with your color bitmapped window system on a network is emulate an ASR33?