Author: Gregor Overney
Date: 16:52:35 12/20/00
Go up one level in this thread
On December 20, 2000 at 18:10:38, Eugene Nalimov wrote: >I have very little experience with majority of the stuff you wrote about (I am >software developer who never wrote any network or GUI code), but even I can >address several your objections. See below. > >On December 20, 2000 at 16:49:02, Gregor Overney wrote: > >> >>>You're talking about developer stuff. I'm not. And also, I doubt that you would >>>have any problems doing anything under windows, if you had the experience. Name >>>me one thing that couldn't be done in windows, that is done in unix, that you >>>would miss. >>> >>2) A more sophisticated threading model that allows multiple threads share a >>light-weight process (LWP). Not just this simple one to one correspondence >>between threads and LWP's as implemented in NT. This requires pthreads library. > >Wrong. Use fibers. They are supported starting from NT 3.51, and I assume they >are exactly what you want. > Did you ever read one of the best books about NT programming for SW engineers called "Advanced Windows" from Jeffrey Richter? If so, you will find on page 971 of the third edition a nice chapter about fibers and that they are only a silly, little fix to allow UNIX programs to be inefficiently ported to NT. NT Fibers are a cheap excuse. They are a higher level construct without any acceptable performance. Some IBM programmers once called it NT's fix for having overlooked a crucial point. If you want to learn more about fibers and their usability you can look at comp.programming.threads and read their FAQ. Or even better check out Micro$oft's own publications. With regard to 64-bit file pointers, when was it the last time you used a 256GB file on NT? NT is for small systems (a la PC) and not for high performance systems. Wrong again? I don't think so. Gregor
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.