NA Digest Sunday, March 27, 1988 Volume 88 : Issue 13

Today's Editor: Cleve Moler

Today's Topics:


From: Hans Schneider <>
Date: Mon, 21 Mar 88 11:32:38 cst
Subject: New Editors for LAA

Linear Algebra and its Applications is pleased to announce the
appointment of three new associate editors:

Daniel Hershkowitz Roger A. Horn Leiba Rodman
Mathematics Dept Dept of Math Sciences Mathematics Dept
Technion Johns Hopkins Univ College of William
Haifa 32000 Baltimore Md 21218 and Mary
Israel USA Williamsburg Va 23185

Manuscripts (in duplicate) may be submitted to any member of the
editorial board, including the editors named above.


From: John G. Aspinall <>
Date: Tue, 22 Mar 88 10:03 EST
Subject: Re-orthonormalization

I have an NxN real orthonormal matrix representing an orientation in
N-space. Some assorted floating point operations on it (e.g. I've been
multiplying several of them together) have left it slightly

I seek suggestions and references to re-orthonormalizing the matrix;
something better than Gram-Schmidt orthonormalization (which imposes a
preferential order on the columns/rows), perhaps iterative, preferably
converging on the "nearest" (in some norm) orthonormal matrix.

John Aspinall < >


From: Frank Luk <>
Date: Wed, 23 Mar 88 12:11:03 EST
Subject: Signal Processing Workshop at Cornell

The Army Research Office, through the Cornell Mathematical Sciences
Institute, will sponsor a Workshop on Matrix-based Signal Processing
Algorithms and Architectures in Ithaca for three days ( June 27-29, 1988 ).
A preliminary list of scientists and engineers who have accepted
our invitation to speak at the Workshop is given below.

1. Sidney Burrus Rice University
2. Patricia Eberlein SUNY at Buffalo
3. Simon Haykin McMaster University
4. Mostafa Kaveh University of Minnesota
5. Marek Kowalski University of Utah and University of Warsaw
6. H.T. Kung Carnegie-Mellon University
7. Monica Lam Carnegie-Mellon University
8. John McWhirter Royal Signal and Radar Establishment
9. Charles Rader Lincoln Lab.
10. John Rice Purdue University
11. Robert Schreiber SAXPY Corp.
12. Jeffrey Speiser Naval Ocean Systems Center
13. G.W. Stewart University of Maryland
14. Donald Tufts University of Rhode Island

For further information, please contact either Adam Bojanczyk or
Franklin Luk at


From: Ian Gladwell (Manchester) <>
Date: Thu, 24 Mar 88 14:55:03 GMT
Subject: Position at Manchester



Applications are invited for a

The post is tenable from October 1st 1988 for 12 months. The salary
will be in the range UK pounds 9260 - 11680.

The post is open to candidates with good qualifications and research
experience in Numerical Analysis. Preference will be given to those
with some experience of lecturing or demonstrating.

Further particulars and application forms should be obtained from the
Registrar, University of Manchester, Manchester M13 9PL, England (Tel.
061 275 2028), to whom applications should be returned by April 22nd.

Informal enquiries may be made to Professor J. Walsh, Department of
Mathematics (Tel. 061 275 5800).


From: Arthur Wouk <wouk@BRL.ARPA>
Date: Fri, 25 Mar 88 13:31:48 EST
Subject: Defense University Research Instrumentation Program

The DoD has announced the FY88 University Research Instrumenta-
tion Program (DURIP). DURIP is a hardware support program for
universities to upgrade their equipment, including computers and
associated hardware. $25 million is to be awarded. Awards will
be made by ARO, AFOSR, DARPA and ONR. Deadline for proposals is
5:00 pm, Tuesday, May 31, 1988. A portion of the money ($3.5 mil-
lion) is specifically restricted to institutions in either of
two categories: (1) historically Black colleges and universities
and other minority institutions, and (2) academic institutions
that received less than $10 million per year in federal funds in
either FY85 or FY86.

In the past awards have averaged $135,000; few proposals over
$500,000 were funded.

For a copy of the full announcement you may call Jean Tigges,
Secretary, Math. Sciences Division, ARO at 919-549-0641, or call
or write to your favorite contact at one of the agencies listed
above. Please do not call me, since I have already given you all
that I could tell you over the phone.


From: David Hough <dgh@Sun.COM>
Date: Fri, 25 Mar 88 19:33:09 PST
Subject: ANSI C Deadline Nears

I have been advised that the public review period for the ANSI C draft dated
11 January 1988 is drawing to a close soon, so I plan to finish up my
commentary Monday and mail it to CBEMA on Tuesday 29 March.

Any of you planning your own submissions should send them in as soon as
possible, and in any event no later than April 15.
Let me know if you would like electronic copies of my submission -
it requires tbl, nroff, and -ms macros.
I encourage you to copy or adapt any parts of it that appeal to you in your
own responses. Neither reading the draft standard nor my commentary
is a 15-minute job, however.

ANSI C or something like it will probably be the common language of CS
graduates during the next decade, so it behooves numerical people
to make their objections known. Most numerical people find plenty
to object about in traditional C implementations.

Many people consider the current X3J11 draft unacceptable;
many others are in a hurry to approve this draft
or anything close as soon as possible.
If there are numerous substantial objections then there may well have
to be another major public review cycle; the preparation for that may
well permit consideration of major changes. This has positive and
negative aspects. If you have feelings one way or the other it
would be worthwhile to express them.

The address for commentary is:

ANSI X3J11 Secretariat
CBEMA, Suite 500
311 First St NW
Washington, DC 20001

The following is the abstract of my submission:

The proposed C standard suffers numerical
shortcomings - many inherited from its precursors
- in areas of interest to providers of portable
mathematical software. I comment in detail upon
the following aspects of the proposed standard:

Comment #1, Section 1.1: anticipate supplemental standards
Comment #2, Section 1.1: manifest the best way
Comment #3, Section exclude implementations explicitly
Comment #4, Section use "significand"
Comment #5, Section bound rounding errors explicitly and uniformly
Comment #6, Section round conversions between floating types
Comment #7, Section forbid implicit narrowing conversions
Comment #8, Section 3.3.3: add power or square operator for integral exponents
Comment #9, Section 3.3.4: emphasize rounding forced
Comment #10, Section 3.4: defer constant expressions with side effects
Comment #11, Section 3.5.6: encourage non-zero initialization
Comment #12, Section 4.7: SIGFPE means floating point
Comment #13, Section 4.8 variable argument lists are expensive
Comment #14, Section 4.9.6: printf/scanf duality for non-model numbers
Comment #15, Section display signed zero with printf %+
Comment #16, Section distinguish exact zero with printf %f
Comment #17, Section provide useful printf %#g
Comment #18, Section scanf requires more than minimal ungetc
Comment #19, Section strtod/atof are mathematical functions
Comment #20, Section 4.10.2: require two random number generators
Comment #21, Section <float.h> has too many names, not enough information
Comment #22, Section long double > minimal float
Comment #23, Section compiler conversion same as run-time
Comment #24, Section no implicit declarations
Comment #25, Section fix arrays
Comment #26, Section 3.7.1: standardize Fortran-77 interface
Comment #27, Section 3.8.8: predefine generalized precision macros
Comment #28, Section 4.5: specific mathematical library functions
Comment #29, Section 4.13.4: standard functions predefine generic operators
Comment #30, Section 4.5.1: make numerical exception handling uniform
Appendix #1: A Proposal for Conformant Arrays
Appendix #2: A Proposal for <float.h>
Appendix #3: Why does traditional C treat float the way it does?

The following comments are based upon Draft Proposed
American National Standard for Information Systems - Pro-
gramming Language C, document number X3J11/88-002, dated 11
January 1988, and its accompanying Rationale. The comments
are personal opinions of the author, and should neither be
construed as wholly original nor as representing the posi-
tion of any organization or other person. A number of indi-
viduals helped formulate and clarify them.


From: Gene Golub <>
Date: Sat, 26 Mar 88 15:26:18 EST
Subject: Third Parallel Circus

Third Parallel Circus

IBM Kingston, NY, May 6-7, 1988.

This is an informal gathering of researchers interested in the practice
and theory of parallel processing. The intention is to engage in
discussions about work in progress or recently completed work, and
exchange ideas for future work.

The first Parallel Circus was held in May 1987 at Yale (chaired by Prof.
M. Schultz) and the second at Cornell in November 1987 (chaired by Prof.
C. Van Loan). IBM Kingston was proposed as the site for the third Circus
because it is within comfortable driving distance of a number of universities
in the Eastern U.S. Profs. Gene Golub and Martin Schultz have consented
to be joint chairmen for the Third Parallel Circus.

A suggestion made at the previous Circus was that one session at the Circus
should have well defined focus within the broad field of parallel processing.
For the session chaired by Prof. Golub, the focus is on parallel processing
pertaining to computational techniques in linear algebra.

The atmosphere is completely casual, and the schedule of speakers is
determined by drawing names out of a hat at the beginning of the Circus.
It has been customary for the Circus to be held on Friday and Saturday
so as to cause minimal disruption to working schedules.

Please call Vijay Sonnad at 914-385-0266 if you wish to attend. If you
prefer to write, the address is: Mail Stop 428, IBM Corporation, Kingston,
NY 12401.


From: Himadri Das <uvaarpa!hudson!gibbs!>
Date: 25 Mar 88 21:16:06 GMT
Subject: SVD software in Pascal or C

I am posting this on behalf of a friend of mine who does not have access
to the net.

She is looking for Pascal routines for calculating the singular value
decomposition of a matrix. If no Pascal routines are available,
information on C routines would be appreciated.

Please send e mail to the following BITNET address : HD7Q@VIRGINIA

Or else please write to or call Karen Bofinger at:

Karen Bofinger
Department of Chemical Engineering
Thornton Hall
University of Virginia
Charlottesville, VA 22901
Phone: (804)-924-6287



End of NA Digest