home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!stanford.edu!rutgers!igor.rutgers.edu!geneva.rutgers.edu!hedrick
- From: hedrick@geneva.rutgers.edu (Charles Hedrick)
- Newsgroups: comp.os.linux
- Subject: Re: real time
- Message-ID: <Jan.3.14.03.48.1993.27545@geneva.rutgers.edu>
- Date: 3 Jan 93 19:03:48 GMT
- References: <1993Jan3.054159.11008@umr.edu>
- Organization: Rutgers Univ., New Brunswick, N.J.
- Lines: 26
-
- quandt@cs.umr.edu (Brian Quandt) writes:
-
- >What can anyone say about linux and real time support? Same
- >question applies to 386bsd. I tend to believe that this concept
- >would not be possible. The reason being is that this (I believe)
- >is a fundamental design change in how the kernal works. However,
- >if anyone has information to the contrary please let me know.
-
- It depends upon what level of real time you want. Linux isn't
- designed to do real-time, so if you need specific facilities (e.g. a
- scheduler that tries to guarantee latencies) you'll be in trouble.
- However some applications probably don't require that.
-
- As you probably know, depending upon your application, real time can
- involve latencies anywhere between 1 usec and 1 msec. Linux has good
- enough interrupt latency to handle fairly high speed serial I/O. I
- could do 9600 bps reliably on a 386sx under 0.12. I can do at least
- 19200 on a 486 now. This is with chips that require an interrupt for
- each character. So you should be able to get latencies from several
- hundred usec to 1 msec. I don't think you could do much better than
- this. The limit used to be the hard disk, though I think Linus has
- been working on that. If you configured a system with a RAM disk and
- ran entirely from there, maybe you could do a little better.
-
- I'm assuming that the code for which latency is critical would be
- linked into the kernel, presumably as a device driver.
-