home *** CD-ROM | disk | FTP | other *** search
- From: <jsh@usenix.org>
-
-
- An Update on UNIX*-Related Standards Activities
-
- May 1990
-
- USENIX Standards Watchdog Committee
-
- Jeffrey S. Haemer, Report Editor
-
- IEEE 1003.4: Real-time Extensions
-
- Rick Greer <rick@ism.isc.com> reports on the April 23-27 meeting in
- Salt Lake City, UT:
-
- 1003.4
-
- The .4 ballots went out on schedule, and most came back on schedule as
- well. We (barely) got the required 75% response, of which 43%
- approved of the draft as it stood. The small-group leaders are
- currently working to resolve the objections and will report back at
- Danvers, in July.
-
- 1003.4a
-
- Most of the work at Snowbird centered around threads (.4a). Two
- alternatives to the pthreads proposal were presented at the meeting:
- ``strands'', from John Zolnowsky of Sun, defined a minimal set of
- interfaces for multi-threaded applications; ``VP'', from Paul Borman
- of Cray, added a ``virtual processor'' layer to the pthreads
- specification, which made some multiprocessor (MP) features visible to
- applications.
-
- The primary MP hardware feature that Paul's VP proposal makes visible
- to the pthreads environment is the ability to write your own spin
- loops and expect them to work. One could, for example, have one
- thread continuously reading an in-core data base while another thread
- updates it. On an MP system, it might be more efficient to code this
- without using a mutex, although doing so on a uni-processor with a
- co-routine threads package is an absolute disaster. The new
- multiprocessor group, 1003.16, is looking into this and similar
- problems, and will probably suggest that .4a include some sort of
- system-wide attribute structure that one can check when writing
- programs that depend heavily on concurrent execution of threads.
-
- After a week's discussion (often a euphemism for argument), we settled
- into a compromise position not that far from where we started --
- pthreads. All this work without much net change was frustrating, but
-
- __________
-
- * UNIX is a registered trademark of AT&T in the U.S. and other
- countries.
-
- May 1990 Standards Update IEEE 1003.4: Real-time Extensions
-
-
- - 2 -
-
- probably unavoidable. Until fairly recently most of the committee was
- busy getting the .4 draft ready for balloting. Lacking enough time to
- have studied threads carefully, members were unwilling to accept the
- small group's conclusions before investigating some alternatives for
- themselves. Still, some progress was made. The most important was a
- more comprehensive definition of signal behavior in multi-threaded
- programs.
-
- 1003.14
-
- On the last day, a first attempt at a real-time application
- environment profile (AEP) was presented. This PAR will be an
- excellent, practical test of AEPs. Real-time applications are likely
- to vary wildly in the subsets of .4's rich features that they require.
- Some worry that the real-time AEP will force embedded systems that
- need only one or two .4 features to incorporate others just to adhere
- to the standard. The problem this poses is not just storage space
- wasted by unused code, but the expense of verifying that this extra
- code will never get in the way of the application. The group will be
- wrestling with these and similar problems in the months to come.
-
- May 1990 Standards Update IEEE 1003.4: Real-time Extensions
-
-
- Volume-Number: Volume 20, Number 5
-
-