Path: utzoo!utgpu!jarvis.csri.toronto.edu!cs.utexas.edu!tut.cis.ohio-state.edu!ucbvax!bloom-beacon!EXPO.LCS.MIT.EDU!keith From: keith@EXPO.LCS.MIT.EDU (Keith Packard) Newsgroups: comp.windows.x Subject: Re: X11R4 server on DS3100; pro/con Message-ID: <9001041944.AA03635@xenon.lcs.mit.edu> Date: 4 Jan 90 19:44:25 GMT References: <1103@ursa-major.SPDCC.COM> Sender: daemon@athena.mit.edu (Mr Background) Organization: The Internet Lines: 32 In color, there are a very few operations for which the DEC product server is substantially faster than the MIT server; the two I can remember offhand are transparent stipples (including polytext) and zero-width lines. In most instances the MIT server equals or surpasses the performance of the current DEC product. Furthermore, the MIT server will use much less memory than the DEC product. This is certainly *not* because the engineering effort at DEC was inferior to the effort done at MIT, in fact DEC helped substantially in the release by donating many of the improvements which are in R4. The very short schedule that produced the DS3100 was rather tough on server optimization efforts, we at MIT benifited from the work that was done by them and were able to focus additional effort on some critical graphics operations (besides; DEC didn't have any competition to spur development like I did...) In monochrome, the choice is much easier; DEC spent little time optimizing monochrome before releasing the PMAX, for R4 I've spent a bit of time speeding up some common monochrome operations (text in particular) which make the MIT server a clear performance winner. The MIT server also has support for several new extensions which many MIT clients can use (SHAPE in particular). I don't expect to see those in a DEC product for several more generations. If you're not just out for the fastest server, but also consider things like support, testing, documentation, etc., you'll want to look more closely at the DEC products. They provide a good level of performance (the best that I know of) for a production piece of code. Keith Packard MIT X Consortium