510.81*
Of2
Vol. 8-9
1980-1981
Q.
OFF i-IME
\
Hitting Macau
Vol. Ho.
Data
£l t-2. ^nj-EO^r^O
Hitting Items not ordered became
aval labl I Ity or Importance It
doubtful
Comments:
Date:
L206
fa
Computing
Services
Office
Off-Line
University of Illinois at Urbana-Champaign
VOL. 8, NO. 3 March 1980
EDITOR: Lynn Bilger
PHONE: (217) 333-6236
164 Digital Computer Lab.
Urbana, Illinois 61801
Page
Contents
3
4
5
6
POLICY
Planned Disk System Upgrades
IBM 360 to be Replaced
SYSTEM NOTES
Reliability Report
STATISTICAL SERVICES
Multivariance VI Installed
on CYBER and IBM
SOUPAC on the CYBER
MATHEMATICAL SERVICES
Tell, Math Software
MISCELLANEOUS
Research Board Deadline for
Departmental Allocation Requests
Computer Related Discounts Available
Notice of a Founding Meeting
of an Apple Computer Users Group
Used Qume Cartridges Available
X
**
50
CSO DIRECTORY - STAFF AND SERVICES
Administrative
Director
George Badger
150
DCL
333-4103
Business Manager
Stanley Rankin
150
DCL
333-6530
Secretary
Joyce Vaughn
150
DCL
333-1637
User Services
User Accounting
Gary Bouck
162
DCL
333-7752
Systems Consulting
138
DCL
333-6133
Statistical Services Consulting
65
Comm West
333-2170
Text Processing Consulting
207
Astronomy
333-7318
Document Printing Reservation
209
Astronomy
333-8150
Asst Dir User Services
Robert Penka
173
DCL
333-4709
Mgr, Statistical Services
Larry Sautter
91
Comm West
333-2170
Mgr, UNIX Operations
Sandra Moy
177
DCL
333-4703
Documentation
Lynn Bilger
139
Astronomy
333-6236
Distribution Center
Don McCabe
164
DCL
333-6285
Keypunch Services
Darlene Hawkins
1208
W Springfield
333-6184
Hardware/Software Support
Terminal Repair Service
164
DCL
333-0969
CYBER Dial-up Numbers
110
baud
333-4000
300
baud
333-4000
1200
baud
333-4001
Asst Dir Systems and Operations
Sandra Moy
177
DCL
333-4703
Asst Dir Development
J. M. Randal
120
DCL
333-9772
Asst Dir Engineering
Cliff Carter
195
DCL
333-3723
CYBER-IBM Operations
Jack Knott
194a
DCL
333-6562
UNIX Operations
Debbie Weller
203
Astronomy
333-8150
Telecommunications
Tom Kerkering
16
DCL
333-0816
Hardware Selection Assistance
Cliff Carter
195
DCL
333-3723
Laboratory Support Project
Lee Hollaar
193
DCL
333-7904
RJE Operations
Rex Duzan
164
DCL
333-6285
RJE Sites
Agriculture
M103
Turner Hall
333-8170
Chemistry
153
Noyes Lab
333-1728
Commerce West
70
Comm West
333-4500
DCL Routing Room
129
DCL
333-6203
Electrical Engineering
146
EEB
333-4936
Florida Ave Res Hall
FAR
333-2695
Illinois St Res Hall
ISR
333-0307
Mechanical Engineering
65
MEB
333-2072
Psychology
453
Psych Bldg.
333-7815
Social Science
202
Lincoln Hall
333-0309
OFF-LINE is the newsletter of the Computing Services Office at the University of
Illinois at Urbana-Champaign. OFF-LINE is printed monthly. Articles may be re-
printed provided that the source of the article is noted. CSO operates an IBM 360
Model 75 with one million bytes of fast core and two million bytes of slow core,
under HASP and OS, and a CYBER 175 with 256K words of central memory and
512K words of ECS, under NOS, serving up to 190 simultaneously active text and
graphics terminals, and a DEC PDP-11/50 with 252K bytes of core driving a GSI
CAT-8 phototypesetter with the UNIX Operating System.
POLICY
PLANNED DISK SYSTEM UPGRADES
TEMPORARILY POSTPONED
The disk system upgrades planned for December 1979 have been temporarily postponed due
to the failure of CDC to honor their delivery commitments. This delayed upgrade affects the
two most significant service bottlenecks on the CYBER, disk space and swapping speed. We
anticipate problems, particularly with disk space (and migrated files) throughout the
remainder of the Spring semester because of this delay. Please cooperate by purging
unnecessary files and by leaving files migrated unless actually needed.
The revised delivery schedule, which we expect will be honored, calls for partial delivery in
late March and completion of delivery by June.
We regret any inconveniences caused by this delay, and will honor the price reductions which
were made in anticipation of adding this equipment.
IBM 360 TO BE REPLACED
At the January Board of Trustees meeting, approval was given for CSO to aquire a new
system, the IBM 4341, to replace the 360/75. The new system will not completely replace all
components of the 360. CSO will replace the central processing system, the memory and the
data channels, but will retain the disk system, tapes and communications equipment, and the
unit record equipment associated with the 360.
The 360, which was installed in the late 1960's, has 1 million bytes of main memory, 2
million bytes of a slow Ampex memory, 3 selector channels and 2 multiplexor channels. This
is the equipment that will be phased out.
The incoming system, the IBM 4341, will have 4 million bytes of high speed central memory
and 6 channels, one of which will be a byte multiplexor. The remaining 5 channels are
known as block multiplexors. (The block multiplexor is the successor to the selector
channel.)
The two systems are approximately equal in computing power although our tests have shown
a fairly wide range of individual performance differences depending on the characteristics of
the job involved. Despite this similarity in performance there are enormous differences based
on the changes in technology which have occurred since the design of the 360.
The outgoing equipment consists of approximately 25 large cabinets housing the processor,
memory and channels. The new system consists of three cabinets housing the entire system.
Data channels have gone from being contained in 6-foot-high boxes that are approximately
3 feet square, to being packaged 3 on a large card. The power consumption and air-
conditioning requirements have dropped by approximately 80% despite the slight increase in
the number of channels and the amount of memory. Most attractive, however, is that the
monthly maintanance cost will drop from about $7,000 to about $600 and that this difference
over a three-year period will pay for the purchase of the system.
The system is currently scheduled to arrive in August. However, we have reqested that
delivery be made earlier so we can install the system at a more convenient time during the
summer. One advantage of the new system is the compact size which will allow us to install
and test it before any steps are taken to dismantle the 360. In any event the installation, for
service purposes, will not occur until removal of the Library Circulation system scheduled for
July 1, 1980.
With the installation of the 4341 we are continuing our commitment, made at the time of the
CYBER selection, to provide IBM compatible service over the life of the CYBER. The 4341
will meet this objective quite adequately and will allow us to continue our IBM services,
including the driving of the RJE network.
Of interest to the user community is the fact that the OS/MVT operating system, with HASP,
as it is currently in operation will continue on the new machine, although additional services
are being planned. It is our anticipation that all jobs which currently run on the 360/75
should run without modification on the new system. As more definite delivery schedules are
known and as other plans are developed we will continue to inform the user community in
future issues of OFF-LINE.
SYSTEM NOTES
RELIABILITY REPORT
During January, the approximate mean time between failures for the CYBER was 24 hours
and the mean time to repair was about 23 minutes. 2550's, disk and environmental problems
were the major causes of CYBER downtime.
For the IBM 360, the approximate mean time between failures was 26 hours and the mean
time to repair was about 44 minutes. Memory failures and disk problems were the reasons
for 360 downtime.
STATISTICAL SERVICES
MULTIVARIATE VI INSTALLED
ON CYBER AND IBM
Version VI of the MULTIVARIANCE statistical program has been installed on the IBM
360/75 and the CDC CYBER 175. MULTIVARIANCE performs a generalized univariate
and multivariate analysis of variance, covariance, and regression. It is intended to replace the
MULTIVARIANCE Version V package presently available on both machines. Both Version
VI and Version V will be maintained until the end of this semester. Then Version V will be
removed from both systems.
The new manual, MULTIVARIANCE - Version VI, National Educational Resources, Chicago,
Illinois, 1978, may be purchased at the CSO Distribution Center (Room 164 DCL).
To access the program on the CYBER you should issue the following commands from the
BATCH subsystem in time-sharing or from cards:
GRAB.MULTV.
MULTV,f/'/ei,f//e2.
where:
filel Specifies the local file which contains the specifications of the problem
and data. If filel is not specified, the information is assumed to be in
the system file INPUT.
file2 Specifies the local file to which the results are written. If file2 is not
specified, the results are written to system file OUTPUT by default.
To access the program on the IBM you should request approximately 160K on the ID card
and issue the following commands from cards:
//PROCLIB DD DSN=SYS4.PROCLIB,DISP=SHR
// EXEC MULTV
Any questions or problems with this program should be directed to the CSO Statistical
Services Consultants in Room 65 Commerce West (333-2170).
SOUPAC ON THE CYBER
CYBER SOUPAC now takes data with record lengths longer than 150. However, the default
record length limit will continue to be 150. If your record length is shorter than 150, the
default limit will suffice. If you need a higher limit, you must specify the limit with an
internal file statement. This file statement should appear early in the SOUPAC program,
preferably prior to any program statements and should have this form:
#FILE(S-ur?;f no.) (mrl) (mbl)" rt" "bt"
where:
S-unit no. Is the sequential number for this file in usual SOUPAC syntax,
e.g. S16 corresponds to TAPE 16 data file
mrl Maximum record length, default = 150
mbl Maximum block length, default = mrl= 1 50
rt Record type, default =Z
bt Block type, default =C
For example (using default parameters):
#FILE(S16)(150)(150)"Z""C".
As is usually the case with SOUPAC parameters, any parameter which is to take other than
its default value should be specified, all others may be skipped or omitted. The file statement
is not needed if all parameters take default values.
Although record lengths may be quite long, limitations on sizes of formats and numbers of
variables still hold as before. Also the old external file statement is obsolete in SOUPAC.
Users should not reuse S-Units designated on #FILE statements for other purposes in the
same SOUPAC run, or for reading in subsequent SOUPAC runs unless the file statement is
given there.
File statements only apply to foreign files bringing data into SOUPAC and files written for
export to other packages, etc. Of course these files may be read or written with default values
without file statements (when they apply). SOUPAC internal scratch files never require file
statements.
The above description of #FILE can be found in the revised document, SOUPAC on the
CYBER, available in Room 65 Commerce West. See other CYBER manuals for appropriate
file statements suitable for your data.
MATHEMATICAL SERVICES
TELL, MATH SOFTWARE
Users with numerical problems or questions are invited to use the new mailbox facility to
leave messages. You may do this by first entering:
TELL.MATH SOFTWARE
(the space between MATH and SOFTWARE is required), and then entering your message
when you are prompted with a question mark.
MISCELLANEOUS
RESEARCH BOARD DEADLINE FOR
DEPARTMENTAL ALLOCATION REQUESTS
The Research Board has established an April 7, 1980 deadline for the submission of
departmental requests for research computer allocations. This deadline affects allocations for
the period July 1, 1980 through December 31, 1980.
Research Board allocations are expected to support faculty research, and thesis and
dissertation research. Departmental requests and the allocations they subsequently receive
are based on individual user requests. Those persons who will need research computer time
for this allocation period should be sure to submit their requests to their department via the
Research Board Form A. These forms and further instructions are available from the
University departments.
COMPUTER RELATED DISCOUNTS AVAILABLE
The Lier Siegler Terminal has been added to the list of CRT's available for Research and
Instructional Computing. It also has been made a component of the LSI II Project, since it
can be upgraded to a graphics terminal by the insertion of a "card" manufactured by Digital
Engineering.
Model of the CRT terminal is:
Lier Siegler ADM 3A (with upper/lower case) $750
EIA Cable $25
Model of the "card" for upgrading to a graphics terminal is:
Digital Engr. RG512 $965
A demonstration of the CRT and the graphics upgrade is available in the CSO Systems
Consulting Office (Room 138 DCL).
NOTICE OF A FOUNDING MEETING
OF AN APPLE COMPUTER USERS GROUP
Time: 8:00 - 10:00 PM
Date: Tuesday, March 18, 1980
Place: Room 28 Education Building
The initial objectives of the group will be:
. to share knowledge about Apple Computer hardware and Software
. to share experiences concerning use of the Apple Computer
• to exchange expertice in solving problems encountered with the Apple Computer
• to discuss applications and software packages of interest to campus users of Apple
Computers
. to discuss interfacing and use of various peripherals with the Apple Computer
. to deal with other concerns regarding Apple Computer systems
We invite all faculty, staff and students who are current or potential users of Apple
Computers to attend this founding meeting. An inventory of campus Apple applications and
users will be one of the first benefits of this group.
For further information, contact Jim Carter or Dave Pontius at 333-2757.
USED QUME CARTRIDGES AVAILABLE
The Center for the Study of Reading has approximately 100 used, cloth and film ribbon qume
cartridges available. They would be pleased to give these to anyone willing to recycle the
cartridges. If you are interested, please contact either Mike Nivens (333-6660) or Steve
Antos (333-7624).
OFF-LINE "s Mailing List
If you wish to be placed on our mailing list for future issues of OFF-LINE, if you wish to be
removed from the list, or if you wish to enter an address correction, please complete and
return this page. (Current subscribers are kept on the mailing list until a specific request for
removal is received, or until a mailing is returned as undeliverable.)
Check one:
□ New subscriber
□ Removal request
□ Address correction
Name:
Address:
CAMPUS or Zip Code
(If address correction, give old address and zip code below.)
Comments:
RETURN TO:
OFF-LINE
164 Digital Computer Laboratory
University of Illinois at Urbana-Champaign
Urbana, Illinois 61801
Of\)0"0
c <•><->
03 r-
-cure
>
c
m
Computing
Services
Office
Off-Line
University of Illinois at Urbana-Champaign
VOL. 8, NO. 4 April 1980
EDITOR: Lynn Bilger
PHONE: (217) 333-6236
164 Digital Computer Lab.
Urbana, Illinois 61801
Page
Contents
2
3
4
6
6
7
7
Till
THE
POLICY
Service Keypunching at CSO
SYSTEM NOTES
Reliability Report
CYBER
TELENET Address Changes
Catalog Overflow - Size
Reference Guides Reorganized
IBM
SAS Changes on the IBM
MISCELLANEOUS
CDC Documentation Updates
Notice of Meeting for
Microcomputer Users Group
HELP WANTED
Operator/Programmer Wanted
C Compiler Wanted
SPECIAL TEAR-OFF SHEET AND MEETING NOTICE
Informal DEC Users Group Meeting
CSO DIRECTORY - STAFF AND SERVICES
Administrative
Director
Business Manager
Secretary
User Services
User Accounting
Systems Consulting
Statistical Services Consulting
Text Processing Consulting
Document Printing Reservation
Asst Dir User Services
Mgr, Statistical Services
Mgr, UNIX Operations
Documentation
Distribution Center
Keypunch Services
Hardware/Software Support
Terminal Repair Service
CYBER Dial-up Numbers
Asst Dir Systems and Operations
Asst Dir Development
Asst Dir Engineering
CYBER-IBM Operations
UNIX Operations
Telecommunications
Hardware Selection Assistance
Laboratory Support Project
RJE Operations
RJE Sites
Agriculture
Chemistry
Commerce West
DCL Routing Room
Electrical Engineering
Florida Ave Res Hall
Illinois St Res Hall
Mechanical Engineering
Psychology
Social Science
George Badger
150
DCL
333-4103
Stanley Rankin
150
DCL
333-6530
Joyce Vaughn
150
DCL
333-1637
Gary Bouck
162
DCL
333-7752
138
DCL
333-6133
65
Comm West
333-2170
207
Astronomy
333-7318
209
Astronomy
333-8150
Robert Penka
173
DCL
333-4709
Larry Sautter
91
Comm West
333-2170
Sandra Moy
177
DCL
333-4703
Lynn Bilger
139
Astronomy
333-6236
Don McCabe
164
DCL
333-6285
Darlene Hawkins
1208
W Springfield
333-6184
164
DCL
333-0969
110
baud
333-4000
300
baud
333-4000
1200
baud
333-4001
Sandra Moy
177
DCL
333-4703
J. M. Randal
120
DCL
333-9772
Cliff Carter
195
DCL
333-3723
Jack Knott
194a
DCL
333-6562
Debbie Weller
203
Astronomy
333-8150
Tom Kerkering
16
DCL
333-0816
Cliff Carter
195
DCL
333-3723
Lee Hollaar
193
DCL
333-7904
Rex Duzan
164
DCL
333-6285
M103
Turner Hall
333-8170
153
Noyes Lab
333-1728
70
Comm West
333-4500
129
DCL
333-6203
146
EEB
333-4936
FAR
333-2695
ISR
333-0307
65
MEB
333-2072
453
Psych Bldg.
333-7815
202
Lincoln Hall
333-0309
OFF-LINE is the newsletter of the Computing Services Office at the University of
Illinois at Urbana-Champaign. OFF-LINE is printed monthly. Articles may be re-
printed provided that the source of the article is noted. CSO operates an IBM 360
Model 75 with one million bytes of fast core and two million bytes of slow core,
under HASP and OS, and a CYBER 175 with 256K words of central memory and
512K words of ECS, under NOS, serving up to 190 simultaneously active text and
graphics terminals, and a DEC PDP-1 1/50 with 252K bytes of core driving a GSI
CAT-8 phototypesetter with the UNIX Operating System.
POLICY
SERVICE KEYPUNCHING AT CSO
Due to the declining use of service keypunching, CSO is planning to contract all keypunch
work to a commercial keypunch service. Users will see little change in the service and will
continue to deal only with CSO when submitting or retrieving keypunch jobs. CSO itself will
provide all interfacing with the contracting agency.
Over the past 18 months, use of service keypunching has been at a level too low to justify the
cost of in-house processing. This has held true even though charges for the service have
been held at a level below the usual commercial costs. At the same time, it is impractical to
simply reduce the size of the operation since it is now at the minimum level to accommodate
large jobs with reasonable turnaround times.
The target date for conversion of the service is July 1, 1980. The exact date will be governed
by how quickly the present data entry staff can be placed in other similar positions on the
campus. It is expected that there will be a transition period during which some work will be
processed in-house, and the remainder will be processed commercially.
We emphasize that CSO is continuing to offer service keypunching as one of its services. The
method of processing submitted jobs is changing, but little else. It is expected that small
keypunch jobs (400 cards or less) will experience an additional one or two day delay because
of transit times. Larger jobs should be processed more quickly since a larger staff will be
available to keyboard the data.
CSO is planning to adjust its method and rate of charging for keypunch service to be
consistent with the contracting agency. Users who have received Research Board support for
keypunch services will not be affected by the rate change since their allocation is denominated
in the number of cards to be punched. Other users will probably incur higher data entry
costs.
SYSTEM NOTES
RELIABILITY REPORT
During February, the approximate mean time between failures for the CYBER was 74 hours
and the mean time to repair was about 54 minutes. Major problems on the CYBER were
related to disk.
For the IBM 360/75, the approximate mean time between failures was 11 hours and the
mean time to repair was about 35 minutes. Hardware failures on memory and LCS were the
major causes of IBM downtime.
CYBER
TELENET ADDRESS CHANGES
Due to TELENET expansion, the addresses for a number of TELENET hosts have been
changed. Please make a note of these changes because you will no longer be able to access
these systems using the old addresses!
HOST SYSTEM
OLD ADDRESS
NEW ADDRESS
CHANGE DATE
Carnegie-Mellon
412 CM
412 CU
412 28
412 29
03/01/80
03/01/80
U. of Delaware
302 UD
302 ZD
302 ZV
302 20
302 21
302 22
01/19/80
01/19/80
01/19/80
U. of Illinois - Urbana
217 UI
217 25
03/22/80
U. of Minnesota
612 UM
612 MN
612 123
612 124
03/22/80
03/22/80
Notre Dame
219 ND
219 65
03/22/80
Rice
713 RU
713 74
03/01/80
Wisconsin - Madison
608 UW
608 25
03/01/80
CATALOG OVERFLOW - SIZE
If you have ever tried to SAVE or REPLACE a file and received the following message:
CATALOG OVERFLOW ■ SIZE
read this article to learn why you receive such a message, and what to do about it.
If you are trying to SAVE a new permanent file, the message indicates that you have tried to
exceed the amount of permanent INDIRECT access disk space you are permitted. To verify
this, you can check your disk space limitations by entering the command
LIMITS.
and checking the displayed value for disk space. Then, by entering the command
DIRECT/PROJ/SUM.
you can determine how many PRU's of indirect files you are currently using.
If you are doing a REPLACE to update a permanent file and receive the message, the above
discussion may apply. However, there is another possibility. To understand it, we'll have to
explore the connection between files, projects, and limits.
When a file is created, the charge, project used at the last signon is stored along with the
user's data (file). This charge, project is then billed for the disk space used by the file. The
Project Manager can control the amount of disk space funded by the project by setting the
various file limits for each person in the project (e.g., the disk space limit we are currently
discussing).
The BILL or CHARGE command sets these limits for the user at signon, using the
charge, project specified. If a user signs on using one project and REPLACES a file belonging
to another project, the limits associated with the current signon are not applicable to that file.
The limits which must be enforced are those set by the owning project. Ideally, these limits
would be extracted from the same database used at signon by the BILL or CHARGE
command. This solution, however, is very difficult and costly, and an alternate solution has
been adopted.
The alternate solution is to store, at file creation, not only the owning charge, project, but also
the file limits currently in effect. Whenever a REPLACE is done on a file belonging to a
project other than the one used at signon, these stored limits are used. Note that the usual
limits (those set up at signon and displayed by the LIMITS command) are used whenever the
project stored with the file matches that used at signon.
We now come back to the error message we started with. If you receive this message after a
REPLACE command, you must first determine which limits are being enforced -- those
displayed by the LIMITS command or those stored with the file. If all of your files were
created under the charge, project you used at signon, your current project limits are being
used. If this is not the case, enter the command:
DIRECT.We/PROJ/LIMITS.
This will display both the charge,project which owns the file and the limits stored with the
file. You can update this information by entering:
CLAIM, file.
This will replace the charge, project stored with the file with the charge, project used on the last
BILL or CHARGE command, and also will replace the stored limits information with your
current limits information. The output from
DIRECT,fi/e/PROJ/LIMITS.
should then reflect the changed information. Using the CLAIM command also will cause all
disk space charges for the file to be charged to the charge, project current at the time of the
CLAIM.
REFERENCE GUIDES REORGANIZED
CSO has recently completed a reorganization of the Reference Guides. This reorganization
involved splitting the Reference Guides into several major categories, reviewing and updating
the existing Reference Guides, and writing new Reference Guides for many topics.
The categories which have been established are as follows:
Category 0 General information which applies to both the CYBER and the
IBM.
Category 1 CYBER General Information
Category 2 CYBER Language Processors
Category 3 CYBER Statistical Software
Category 4 CYBER Mathematical Software
Category 5 CYBER Plotting and Graphics Software
Category 6 CYBER Miscellaneous Software
Category 7 CYBER Utilities and Procedures
Category 1 1 IBM General Information
Category 1 2 IBM Language and Compilers
Category 13 IBM Statistical Software
Category 14 IBM Mathematical Software
Category 15 IBM Plotting and Graphics Software
Category 16 IBM Libraries
Category 17 IBM Utilities and Procedures
Most of these Reference Guides are now available at CSO North, CSO South, and the RJE
sites. However, some of the old guides are still undergoing revision, and some of the "new"
guides have yet to be written. These will be added as they are finished.
In addition to reorganizing the guides, the manner in which they are stored has been
changed. Instead of being placed on shelves (with the attendant problems of adding or
deleting guides and misplaced copies) the guides will now be maintained in folders in filing
cabinets. This should make it much easier to add or delete guides, and take up less space.
We hope the users will find this new system more useful.
Six of the Reference Guides have undergone some rather drastic revision; DIRECT, PLOT,
PRINT, PUNCHC, SENDJOB, and TYPE. Since these particular utilities have so many
options available on them, CSO has put them into a special manual entitled, CSO Local
Utilities Manual. The Reference Guides for these utilities now contain only the most
frequently used options. The manual, however, contains all of the options that are available
on these particular utilities. The manual will be available at the CSO Distribution Center,
Room 164 DCL (as soon as it is returned from the printer). It is also available on-line, and
may be obtained and printed by entering:
WRITEUP.UTILITY.
PRINT,UTILDOC/ASCII/CC.
The reorganization of the Reference Guides came about because CSO is preparing to publish
a new manual, An Index to Software on the CYBER. This index will contain a listing of all
software available on the CYBER including software that is on the system, but unsupported
by CSO. Each software package or system listed will include the level of support, a brief
description, and a list of available documentation. This manual should be available in the
near future.
IBM
SAS CHANGES ON THE IBM
The 1979 version of SAS has now been installed and tested on the IBM 360. Due to the
installation of the 1979 version, SAS76 and SAS76.6 were both removed from the system on
March 31, 1980. The new 1979 version of SAS is available by using the following JCL (you
do not need a PROCLIB card):
// EXEC SAS79
CSO Statistical Services would also like to remove the much older 1972 version of SAS from
the system. However, we understand that certain users are still making use of some of the
features from the 1972 version. We encourage any users who feel that they still need to use
this older version to contact Ron Woan (333-2172). After March 31, 1980, the 1972 version
of SAS will be available only by using the following JCL:
//PROCLIB DD DSN=SYS4.PR0CLIB.DISP=SHR
// EXEC SAS72
If there is not sufficient evidence to support keeping the 1972 version of SAS available, it will
be removed from the system at the end of the semester.
MISCELLANEOUS
CDC DOCUMENTATION UPDATES
The CSO Distribution Center (Room 164 DCL) now has updates and/or revised manuals
available from CDC for Release 5 of the operating system. In the following list, updates are
available to anyone who needs them to update their present manuals. "New" manuals will be
given only to those users who bring in their old manuals for trade (otherwise, the manuals
must be purchased).
1. NOS Vol 1 Reference Manual (60435400), Rev J
2. NOS Vol 2 Reference Manual (60445300), Rev J
3. FORTRAN Reference (60497800), Rev E
4. BASIC Reference (19983900), Rev F
5. COMPASS Reference (60492600), Rev F
6. COMPASS Instant (60492800), Rev C
7. Time-Sharing Reference (60435500), Rev G
8. LOADER Reference (60429800), Rev G
9. Interactive Debug Reference (60481400), Rev B
10. SYMPL Reference (60496400), Rev F
11. Record Manager BAM Reference (60495700), Rev E
12. Record Manager A AM Reference (60499300), Rev B
NEW
NEW
UPDATE
NEW
UPDATE
UPDATE
NEW
UPDATE
NEW
NEW
NEW
NEW
NOTICE OF MEETING FOR
MICROCOMPUTER USERS GROUP
Date: Wednesday, April 9, 1980
Time: 7:30 PM
Place: To be determined
Topic: Microcomputer Communications with the CYBER
The first meeting of the Apple Computer Users Group, held on March 1 8, was highly
successful and well attended. It was then decided that the Apple Computer Users Group
should expand and become a general Microcomputer Users Group.
Please help us get in touch with other microcomputer users in the area who do not receive
OFF-LINE by telling them of this meeting. All interested faculty, staff, students and off-
campus users are invited.
For further information about place and details of the upcoming meeting, contact Jim Carter
or Dave Pontius at 333-2757.
HELP WANTED
OPERATOR/PROGRAMMER WANTED
A local Federal agency wants to hire a 32-hour per week computer operator/ programmer.
Applicants should have FORTRAN or PL1 programming experience. Familiarity with IBM
and CYBER systems is desirable, but not necessary. If interested, please contact Ms. Helen
Larson at 398-5355.
C COMPILER WANTED
I am looking for a "C" compiler for a 6502 CPU. If you have one, or know of anyone who
does, please contact Pat Kane, 206 Astronomy, 333-1546.
SPECIAL TEAR-OFF SHEET AND MEETING NOTICE
INFORMAL DEC USERS GROUP MEETING
TIME: 4:00 - 5:30 PM
DATE: Tuesday, April 29, 1980
PLACE: 115 Digital Computer Lab
Persons with DEC computers of any size (DEC 10/20, VAX, PDP-11, PDP-8, etc.) are
invited to gather and share common hardware and software interests, problems, and ideas.
We will determine at the meeting if there is sufficient interest to form an organization which
would continue to meet on a regular basis.
Please help us to get in touch with users of DEC machines in the area by Xeroxing this page
for appropriate people (who do not receive OFF-LINE). All interested faculty, staff, students,
and off-campus users are invited. We intend this particular group to be user oriented. It is
not a policy group.
Part of the value of such a group will be knowing who has which machines and what
software. The responses to the questionnaire on the following page will be made available to
interested members.
10
The responses to the questions below will be made available to DEC users and other
interested persons. Please type or print plainly in BLACK ink and bring this form to the
meeting, or return it to:
Jerry Wray
Loomis Laboratory of Physics
University of Illinois
Urbana, Illinois 61801
1. Department/Company.
Address
2. Principal user or
Person responsible
Second contact
Phone-
Phone-
3. Location of machine(s).
4. Location of user
5. System/ Processor type(s) - list specific models.
6. Configuration (s) - Memory size, types of peripherals, special devices, etc..
7. Brief description of work (slat analysis, A/D, etc).
8. Software used? (Monitor, BASIC, FORTRAN, etc).
9. Do you have user-written software you would consider sharing? Yes No
What ?
On what medium?.
10. Should this group continue meeting1;
Yes No. If yes, how often?.
What topics should be considered at meetings?
11. Should we associate formally with DEC/DECUS?
Yes
No.
OFF-UNE's Mailing List
If you wish to be placed on our mailing list for future issues of OFF-LINE, if you wish to be
removed from the list, or if you wish to enter an address correction, please complete and
return this page. (Current subscribers are kept on the mailing list until a specific request for
removal is received, or until a mailing is returned as undeliverable.)
Check one:
□ New subscriber
□ Removal request
□ Address correction
Name:
Address:
CAMPUS or Zip Code
(If address correction, give old address and zip code below.)
Comments:
RETURN TO:
OFF-LINE
164 Digital Computer Laboratory
University of Illinois at Urbana-Champaign
Urbana, Illinois 61801
CSO DIRECTORY - STAFF AND SERVICES
Administrative
Director
Business Manager
Secretary
User Services
User Accounting
Systems Consulting
Statistical Services Consulting
Text Processing Consulting
Document Printing Reservation
Asst Dir User Services
Mgr, Statistical Services
Mgr, UNIX Operations
Documentation
Distribution Center
Keypunch Services
Hardware/Software Support
Terminal Repair Service
CYBER Dial-up Numbers
Asst Dir Systems and Operations
Asst Dir Development
Asst Dir Engineering
CYBER-IBM Operations
UNIX Operations
Telecommunications
Hardware Selection Assistance
Laboratory Support Project
RJE Operations
RJE Sites
Agriculture
Chemistry
Commerce West
DCL Routing Room
Electrical Engineering
Florida Ave Res Hall
Illinois St Res Hall
Mechanical Engineering
Psychology
Social Science
George Badger
150
DCL
333-4103
Stanley Rankin
150
DCL
333-6530
Joyce Vaughn
150
DCL
333-1637
Gary Bouck
162
DCL
333-7752
138
DCL
333-6133
65
Comm West
333-2170
207
Astronomy
333-7318
209
Astronomy
333-8150
Robert Penka
173
DCL
333-4709
Larry Sautter
91
Comm West
333-2170
Sandra Moy
177
DCL
333-4703
Lynn Bilger
139
Astronomy
333-6236
Don McCabe
164
DCL
333-6285
Darlene Hawkins
1208
W Springfield
333-6184
164
DCL
333-0969
110
baud
333-4000
300
baud
333-4000
1200
baud
333-4001
Sandra Moy
177
DCL
333-4703
J. M. Randal
120
DCL
333-9772
Cliff Carter
195
DCL
333-3723
Jack Knott
194a
DCL
333-6562
Debbie Weller
203
Astronomy
333-8150
Tom Kerkering
16
DCL
333-0816
Cliff Carter
195
DCL
333-3723
Lee Hollaar
193
DCL
333-7904
Rex Duzan
164
DCL
333-6285
M103
Turner Hall
333-8170
153
Noyes Lab
333-1728
70
Comm West
333-4500
129
DCL
333-6203
146
EEB
333-4936
FAR
333-2695
ISR
333-0307
65
MEB
333-2072
453
Psych Bldg.
333-7815
202
Lincoln Hall
333-0309
OFF-LINE is the newsletter of the Computing Services Office at the University of
Illinois at Urbana-Champaign. OFF-LINE is printed monthly. Articles may be re-
printed provided that the source of the article is noted. CSO operates an IBM 360
Model 75 with one million bytes of fast core and two million bytes of slow core,
under HASP and OS, and a CYBER 175 with 256K words of central memory and
512K words of ECS, under NOS, serving up to 190 simultaneously active text and
graphics terminals, and a DEC PDP-11/50 with 252K bytes of core driving a GSI
CAT-8 phototypesetter with the UNIX Operating System.
SYSTEM NOTES
MARCH RELIABILITY REPORT
CYBER:
16 interruptions affecting time-sharing only
6 interruptions affecting the entire system
2 of these were for more than 1 5 minutes
Availability: 98.8% of the scheduled time
360/75:
36 interruptions affecting the entire system
10 of these were for more than 15 minutes
Availability: 97.7% of the scheduled time
For the past several months, we have considered changing the format of the monthly
reliability report to make it more meaningful to the user community. The above format is
one possible way. "Interruptions affecting time-sharing only" means those interruptions from
which a user may normally recover (may continue any work being done at the time of the
interruption by doing a RECOVER,tty when the system comes back up). "Interruptions
affecting the entire system" means that a major problem has occcurred and work being done
at the time of the interruption will not be able to be recovered.
We would appreciate any comments (pro or con) and/or suggestions from you, the users,
about the new format.
CYBER
SECURITY SAFEGUARDS
Due to some recent incidents with unauthorized users on the CYBER, CSO would like to
take this opportunity to remind all users about security safeguards for files and accounting
information. It may be legal to look at files which the owner has not taken proper steps to
protect. However, it is illegal to pose as another person, through the presentation of
identification and passwords, for the purpose of making unauthorized use of their accounts.
Allocations of service supported by campus funds are for specified research and instructional
purposes. Their use for harassment of others, for violations of file privacy, or as a means of
gaining illegal access to other accounts will result in disciplinary or legal action.
To safeguard your resources from such abuse, please be sure that your files which contain
accounting information, especially passwords, are properly protected. Such files should be
restricted to private mode. If it is necessary to allow access to a few people, you can permit
the files explicitly to their user identifications.
We advise all CYBER users to change their passwords frequently. Users with null passwords
should assign passwords to their User Numbers. If someone has been using your resources
illegally, however, just changing the password may not provide adequate protection. You
should also check the permission privileges for all your files, and change the permission of
any public files to private if they should be protected.
The following commands show a number of ways to check and change the permission
privileges of files, and to change your password.
DIRECT/CT=PU.
Will list all files in your area which are public, i.e.,
can be accessed by anyone signed on to the
CYBER.
DIRECT/ PTOTAL.
GET, pfn.
PURGE,pfn.
SAVE,pfn.
PERMIT, pfn, userid= R.
PASSWOR, oldpas.newpas.
Will list all files in your area with the specific IDs
permitted to access them.
Will get a local copy of the file pfn.
Will delete the permanent copy of file pfn.
Will save the file pfn in your permanent directory
with no permission privileges assigned to it.
Permits read-only access to file pfn for the
specified userid only.
Changes the password oldpas to a new password
newpas. This changes the password immediately,
and the new password must be used the next time
you sign on. An even safer way to change your
password is to simply enter the command
PASSWOR. The computer will then prompt you
for the old and new passwords and the entries you
type in will not be echoed on the screen. If your
password is presently null, enter a carriage return
when prompted for oldpas.
Project Managers should check to ensure that no User Numbers have been illegally added to
their project. Also, if they have a null project code word, they should assign a code word to
their project immediately. To assign a code word to a project, enter:
MANAGE.
P,charge,project
CODE WORD=new:ocfe/PN
One other word of caution: it has been observed that some users have signed onto public
terminals and have not logged off when leaving for a short period of time. Perhaps the
reasoning for this is that they believe this "reserves" the terminal for them. However, it is a
dangerous practice because it allows anyone to then use that signon for whatever purpose — to
access the files, use the person's funds, etc. If you do this type of thing, you must accept the
responsibility for what happens. LOG OFF when you leave a public terminal - protect your
files and your funds!
NCAR GRAPHICS SOFTWARE
CSO has installed the NCAR Graphics Software on the CYBER. This software, obtained
from the National Center for Atmospheric Research, contains graphics utility subroutines for
drawing
. Contour plots
. 3-D surfaces with hidden lines removed
• World map projections
and other high-level applications. Plots may be produced on a Tektronix graphics terminal or
a CalComp plotter.
The NCAR Graphics Software is a library of subroutines called from a FORTRAN program.
The NCAR software achieves device independent plotting by writing a local disk file of plotter
commands, which is later plotted on a specific device requested by the user. This file of
plotter commands is called the meta-code file.
Documentation describing the subroutine usage and parameters is available for inspection in
the CSO North Consulting Office (Room 138 DCL). This same information is available on
the CYBER via a procedure, NCARDOC. The commands
GRAB.NCARDOC.
NCARDOC.
will give a brief summary and directions to obtain additional information and subroutine
writeups.
A typical use of the NCAR software involves the following sequence of commands:
GRAB,NCAR. Attach the NCAR library.
FTN,I =program,L=0. Compile the FORTRAN program with calls to NCAR
subroutines.
LGO. Execute the program. This does not generate the plot
itself, but rather, the meta-code file of plot instructions.
NCARTRN. To display the plot on a Tektronix terminal.
NCARTRN,PLOT=CALC. To display the plot on a CalComp plotter.
PLOT,TAPE99. This plot statement is necessary to actually send the file
to the plotter.
Once the meta-code file has been generated, it may be displayed at will on either Tektronix or
CalComp equipment by using the appropriate NCARTRN statement.
NOTE: The file name NC ARMC must appear on the program statement of your FORTRAN
program. This is the name of the meta-code file written by the NCAR software.
A sample program and plot are shown here:
PROGRAM NCARTST(INPUT,OUTPUT,NCARMC)
C
C THE FUNCTION
C
C Z(X,Y)=.25*(X+Y + l./(X-.l)**2+y**2 + .09)
C -l./((X + .l)**2+Y"2+.09))
C
C IS EVALUATED FOR
C
C X=-l. TO 1. IN INCREMENTS OF. 1 AND
C Y =-1.2 TO 1.2 IN INCREMENTS OF. 1.
C
C XX CONTAINS THE X-DIRECTION COORDINATE VALUES FOR Z(X,Y)
C YY CONTAINS THE Y-DIRECTION COORDINATE VALUES FOR Z(X,Y)
C Z CONTAINS THE FUNCTION VALUE
C S CONTAINS VALUES FOR THE LINE OF SIGHT FOR SRFACE.
C WORK IS A WORK ARRAY
C
REAL XX(21), YY(25), Z(21,25), S(6), WORK(1050)
C
DATA S(l), S(2), S(3), S(4), S(5), S(6)/
1 -8.0, -6.0, 3.0, 0.0, 0.0, 0.0/
C
C FILL XX AND YY COORDINATE ARRAYS AND Z FUNCTION VALUE ARRAY
C
DO 20 1 = 1,21
X= PFLOAT(I-ll)
XX(I)=X
DO 10 J = l,25
Y= PFLOAT(J-13)
YY(J)=Y
Z(I,J) = (X+Y-l-l./((X-.l)*'2-l-Y**2+.09)-
1 l./((X-hl)**2+Y**2+.09))*.25
10 CONTINUE
20 CONTINUE
CALL SRFACE (XX, YY, Z, WORK, 21, 21, 25, S, 0)
STOP
END
RECORD MANAGER BUG - ODD BLOCK SIZES
It has been noted that CYBER RECORD MANAGER, when writing records on an S-format
or L-format tape (e.g., a tape to be sent away to be read elsewhere), writes only blocks of
even length. If a block is odd in length, RECORD MANAGER adds a character to the block,
which may hamper subsequent attempts to read the tape file. This will apply whether you are
recording the tape with FORTRAN directly or with TBLOCK.
For example, if you use TBLOCK with RECSIZE = 269 (record size) and BF=11 (blocking
factor), the size of each tape block is 269*1 1 =2959, which is odd. RECORD MANAGER
will then create a block of length 2960 by adding a character to the block. If BF=10, the size
of each block is 2690, which is even. However, the last block in the file may not have a full
10 records in it; it may only have 9 records. This block then would be 269*9 = 2421 which is
odd, so, RECORD MANAGER would add a character.
For the moment, the only way around this problem is to specify an even record size, thus
ensuring that in all cases, each block is even in length.
Please note that DEBLOCK will read such an incorrect file with no errors, but if the tape file
is read on the IBM 360, an error will result. Or, if the file is read "to end of file" with
FORTRAN on the CYBER, RECORD MANAGER error 143 (insufficient data) will result.
STATISTICAL SERVICES
1979 VERSION OF BMD INSTALLED
ON THE IBM 360
The 1979 Version of BMDP is now available on the IBM 360/75. It is an improved edition of
the 1977 Version. The following JCL should be used to access it:
// EXEC BIMED,PROG=des/red program
NOTE: You no longer need to use BIMEDT to increase the core for larger problems. You
may simply specify more core on the "ID" card. If you request more than 400K, you will
need to put the region on the "EXEC" JCL card also. You now need the BIMEDT program
only when FORTRAN statements are to be used. The following JCL will invoke BIMEDT:
// EXEC BIMEDT, PROG= desired program
Any questions concerning BMDP should be directed to the CSO Statistical Services
Consultants at 65 Commerce West (333-2170).
SAS79 INTERFACE TO BMDP79
You may now call most of the BMDP programs from SAS. For a detailed description of its
usage, see the BMDP procedure as described in the SAS User's Guide -1979 Edition (available
from the CSO Distribution Center, Room 164 DCL). To utilize this interface, you must use
the following JCL:
// EXECSASBMDP
If you have any questions concerning its use, please contact the CSO Statistical Services
Consultants at 65 Commerce West (333-2170).
CORRECTION IN APRIL'S SAS ARTICLE
In the April issue of OFF-LINE, the JCL specified in the SAS article to access the 1979
Version of SAS was in error. The article said to use "EXEC SAS79". The correct JCL to use
to access the 1979 Version is as follows:
// EXEC SAS
CSO regrets any inconvenience or confusion this error may have caused.
NUMERICAL SERVICES
SLAM: SIMULATION LANGUAGE FOR
ALTERNATIVE MODELING
SLAM, a program for combined continuous, discrete and network simulation, is now
available on the CYBER. The program is similar to GASP in its mode of operation: the user
may supply a data deck for certain information, but must supply FORTRAN routines to define
states, events and special output. Currently, CSO has no formal documentation other than
the book, Introduction to Simulation and SLAM, by A. Alan B. Pritsker, available for inspection
in the CSO North Consulting Office (Room 138 DCL).
The program is accessed by the statement:
GRAB.SLAM.
This places a binary program file, SLAM, in your local file space. It is then run as follows:
SLAM.input.output.
Where input represents a file of input data and output is a file where the simulation report
should be placed. If you have auxiliary routines as well, the following method should be
used:
FTN,I= subs,...
LOAD.LGO.
SLAM.input.output.
Where subs represents a file containing your routines.
An assortment of sample SLAM problems is available via the SAMPLES procedure. This is
accessed via
GRAB.SAMPLES.
General help on SAMPLES can then be obtained by entering:
-HELP.SAMPLES.
A catalog of SLAM sample decks may be obtained by entering:
SAMPLES.SLAM.
See Stan Kerr (Room 175 DCL, 333-4715) for additional information.
MINI-DYNAMO
A version of the DYNAMO simulation language for system dynamics called Mini-DYNAMO
has been installed on the CYBER. It is documented in the Mini-DYNAMO User's Guide (has
been ordered), and in the works of Jay Forrester, the inventor of systems dynamics.
DYNAMO is accessed by the statement:
GRAB.DYNAMO.
This places a CCL procedure file called DYNAMO, and a binary program in your local file
space. Some help information may then be obtained by entering:
-HELP.DYNAMO.
To run a model, enter the statement:
DYNAMO, model, options.
Where model represents a file containing your DYNAMO model statements, and options
represents a series of letters specifying run options. The default options (if you omit options
from the above statement) cause DYNAMO to analyze and list the model, do one run of it,
then request rerun options.
Some sample models for DYNAMO are available via the SAMPLES procedure, which is
accessed with the statement:
GRAB.SAMPLES.
After accessing SAMPLES, general help information on SAMPLES may be obtained by
entering:
-HELP.SAMPLES.
A catalog of available DYNAMO samples may be obtained by entering:
SAMPLES.DYNAMO.
Other references on system dynamics are Principles of Systems by Forrester, and World
Dynamics by Forrester.
See Stan Kerr (Room 175 DCL, 333-4715) for further information or references on system
dynamics.
NEW PROCEDURES: MATH AND SAMPLES
A CCL procedure called MATH is now available which combines the facilities of the
MATHDOC, IMSLDOC, MSLDOC, NATSDOC, LINDOC, and MISCDOC procedures (all
of which will eventually be phased out) for obtaining source and/or writeups of routines in
various libraries. This procedure is accessed by entering the statement:
GRAB.MATH.
General help information may then be obtained by entering:
— HELP.MATH.
MATH is designed to supply information on a specific routine in a specific library, and is
generally used as follows:
M ATH , library, routine.
Where library represents the name of some library "known" to MATH, and routine represents
something in the library. MATH places the information (if any) in files SOURCE and/or
DOC; file SOURCE will contain FORTRAN source (if available) and file DOC will contain a
writeup (if available). SOURCE and DOC are not rewound, so several things can be
"stacked" in them by consecutive calls to MATH. If you "stack" things in SOURCE, be sure
to PACK the file before saving or compiling it.
For a given library, the statement
MATHJibrary.
places a list of available routines from the library in the file DOC. Entering just the statement
MATH.
or the statement
MATH.SUMMARY.
causes MATH to display a list of "known" libraries.
Procedure SAMPLES can be used to obtain sample decks for a number of packages on the
CYBER, demonstrating their use. It is accessed via
GRAB.SAMPLES.
You may then obtain more help information by entering
-HELP.SAMPLES.
SAMPLES is used much like MATH:
SbMPLES,package,deck.
Where package represents the name of some package "known" to SAMPLES, and deck
represents the name of some sample deck or data for that package. It will produce a file
SAMPLE with the requested sample deck in it. In some instances, an extra auxiliary file may
10
be produced; this is usually called DATA. A message is displayed if this auxiliary file has
been produced.
Entering the statement
SAMPLES.package.
lists a catalog of known sample decks for the given package in file SAMPLE. Entering the
statement
SAMPLES.SUMMARY.
produces a summary of what packages are known to SAMPLES.
Please note that you must know beforehand how to run the package you are interested in.
SAMPLES only provides files, not instructions on how to use them.
Following is a list of the libraries currently "known" for MATH:
UOILIB Locally produced and externally acquired routines
IMSL Large leased libary of math and statistical routines
MSL CDC's math routine library (no longer supported by CDC)
EISPACK Routines for eigenanalysis of both standard and generalized eigenvalue
problems
FUNPACK Special function package developed at Argonne Lab
MISC Miscellaneous unsupported routines, provided only in source with no
guarantees of performance
BSPLINE B-Spline routines from Carl de Boor's book, A practical Guide to Splines
HARWELL Very small subset of Harwell Library routines which various users have
converted for the CYBER
GRG Generalized reduced-gradient optimization program
NETWORK Network solving routines and programs
Following is a list of packages "known" to SAMPLES:
APEX CDC's large-scale linear programming package. Includes parametric
and mixed-integer programs.
MPOS Northwestern U.'s medium-scale linear/integer/quadratic programming
package
XMP Linear programming subroutines libary
11
ALTRAN
SPICE
MODEL
ACSL
FORSIM
DYNAMO
POST
ELLPDE
PDECOL
FFT9
SLAM
GASP
SPURT
GRG
BSPLINE
DBSPLIN
ITPACK
Symbolic manipulation language for rational functions
Circuit analysis package
Continuous network simulation language
Continuous simulation language
Simulation package for time-dependent PDE's
System dynamics simulation language
PDE program from Bell Labs
Routines for separable elliptic PDE's, published in Transactions on Math
Software, Sept/ 79
Collocation software for time-dependent PDE's, published in
Transactions on Math Software, Sept/ 79
A program for fast solution of Helmholtz-type PDE's, published in
Transactions on Math Software, Dec/ 79
Continuous/discrete/network simulation language
Continuous/discrete simulation language
Discrete simulation language
Generalized reduced-gradient non-linear programming
B-Spline routines from Carl de Boor's book, A Practical Guide to Splines
Double precision B-Spline routines
Iterative routines for solution of large sparse symmetric positive
definite linear systems, from U. of Texas at Austin
NEW UOILIB ROUTINES
The following routines have recently been installed in UOILIB on the CYBER:
EPISODE
A routine for solution of ordinary differential equations from the
Argonne National Laboratory.
DEPSODE
Double precision version of EPISODE.
12
LSODE A new code for solution of ordinary differential equations, by Alan
Hindmarsh of Lawrence Livermore Laboratory. This contains
algorithms for both stiff and non-stiff problems.
DLSODE Double precision version of LSODE.
STINT A routine for solution of stiff ordinary differential equations, published
in Transactions on Mathematical Software, Volume 4, Number 4,
December 1978.
GAM A routine for evaluation of incomplete gamma functions, published in
Transactions on Mathematical Software, Volume 5, Number 4,
December 1979.
DGAM Double precision version of GAM.
The UOILIB library is accessed via
GRAB.UOILIB.
For source or writeups of UOILIB routines, use the procedure MATH as follows:
GRAB.MATH.
-HELP.MATH
These statements will display information on how to use MATH to then obtain source and
writeups for the UOILIB routines.
TEST VERSION OF MINPACK
A library of minimization and optimization codes is being developed at Argonne National
Laboratory. We have been asked by the test site representative here to announce it for user
testing. Like the previous packages, EISPACK (eigenanalysis), FUNPACK (special
functions) and LINPACK (linear systems), MINPACK will be extensively tested and
validated at many user sites, on many different computer types. The final result should be a
well-integrated and reliable package of routines.
The present version contains several Levenberg-Marquardt routines for minimizing sums of
squares (as for curve-fitting); there are versions requiring derivatives to be provided by the
user, and derivative-free versions, as well as storage-economizing versions. Also, there are
routines for solution of simultaneous nonlinear systems of equations, based on a hybrid
method due to Powell; there are derivative-based and derivative-free versions of this. A
routine is also provided for checking the consistency of a Jacobian calculation, as a means of
testing one's Jacobian calculations before using one of the routines requiring derivatives.
13
A full description of the curent MINPACK routines (78 pages) can be obtained and printed
as follows:
WRITEUP.MINPACK/FUTURE.
PRINT.MINDOC/CC.
The library can be accessed and used by entering the control statements:
GRAB.MINPACK/FUTURE.
Following this, simply compile and run your program which calls the MINPACK routines.
Please direct inquiries or responses to Stan Kerr (Room 175 DCL, 333-4715) or to Mary Ann
Berg (Room 221 Altgeld, 333-2168) in the Mathematics Department.
MISCELLANEOUS
NEW REVISION PAGES FROM CDC
CSO has now received Rev. D for the COBOL Version 4 Reference Manual. Note that these
are just revision pages, not a new manual. Anyone who needs these pages may obtain them
from the CSO Distribution Center, Room 164 DCL.
15
SPECIAL TEAROFF SHEET
INDEX FOR MINI AND MICRO COMPUTER USERS
CSO is planning to publish an Index to the Use of Mini and Micro Computers by Researchers on
Campus. It will also include the use of specialized computer-related equipment, software
packages, or databases not provided by CSO.
The impetus for this document comes from the frequent questions directed to CSO in these
areas. A researcher who is facing a problem involving computers often suspects that
someone else on campus has already dealt with a similar problem. CSO is a logical place to
start trying to contact such a person.
The document will be published informally and updated periodically. Minimal effort will be
spent editing the information or data received, and inclusion of an item will not imply further
knowledge or support by CSO. We are simply trying to start useful conversations.
We would like to hear from you if
. you are using a minicomputer or microcomputer in your work.
• you have acquired experience using equipment for automated control, measurement,
data collection, or data conversion.
. you have acquired or developed software packages or databases which could be of
interest to someone else.
If you have information, please fill out the form on the reverse side of this page, and return
it to us by June 15, 1980. If you are a CYBER user, you can use the APPEND command to
send us the information through a CYBER file, COMPUSE, rather than using the form. To
do this, simply put your responses into a local file (named filex in the example below) and
then enter the following control statement:
APPEND,COMPUSE,fr/ex/UN=DOCUMNT.
16
Please fill in this form and return by June 15, 1980 to:
Editor, Documentation
164 Digital Computer Laboratory
University of Illinois at Urbana-Champaign
Urbana, Illinois 61801
It may be returned by campus mail to the above also.
Name:.
Address: Phone:.
If you are using a minicomputer or microcomputer, please supply the following information:
Type of Computer-
Peripheral hardware.
Operating system
How are you using the equipment?.
If you've developed or acquired special purpose equipment which is used in conjunction with a computer, please
tell us about it .
If you've developed or acquired software packages or databases which could be useful to someone else, please tell
us about it._ .
Other comments or information.
OFF-LINE's Mailing List
If you wish to be placed on our mailing list for future issues of OFF-LINE, if you wish to be
removed from the list, or if you wish to enter an address correction, please complete and
return this page. (Current subscribers are kept on the mailing list until a specific request for
removal is received, or until a mailing is returned as undeliverable.)
Check one:
□ New subscriber
□ Removal request
□ Address correction
Name:
Address:
CAMPUS or Zip Code
(If address correction, give old address and zip code below.)
Comments:
RETURN TO:
OFF-LINE
164 Digital Computer Laboratory
University of Illinois at Urbana-Champaign
Urbana, Illinois 61801
LU
CD
<
X <
<ojq:
•-«u.O T
C£— '<\J <
Computing
Services
Office
Off-Line
University of Illinois at Urbana-Champaign
VOL. 8, NO. 6 June 1980
EDITOR: Lynn Bilger
PHONE: (217) 333-6236
164 Digital Computer Lab.
Urbana, Illinois 61801
Page
Contents
1
2
3
3
5
5
6
8
SYSTEM NOTES
April Reliability Report
NUMERICAL SERVICES
New Version of Spice
CCL Versions of Some Procedures Available in FUTURE
Some Procedure Files to be Discontinued
FORTUOI Phase-out
TEXT PROCESSING
Diablo Reservation Service to Change
MISCELLANEOUS
Teletype Repair Discontinued
LSI- 11 Rental Systems
DEC Informal Users Group Meeting
A Cosmic Experience
CSO DIRECTORY - STAFF AND SERVICES
Administrative
Director
Business Manager
Secretary
User Services
User Accounting
Systems Consulting
Statistical Services Consulting
Text Processing Consulting
Asst Dir User Services
Mgr. Statistical Services
Mgr, UNIX Operations
Documentation
Distribution Center
Hardware/Software Support
Terminal Repair Service
CYBER Dial-up Numbers
Asst Dir Systems and Operations
Asst Dir Development
Asst Dir Engineering
CYBER-IBM Operations
UNIX Operations
Telecommunications
Hardware Selection Assistance
Laboratory Support Project
RJE Operations
RJE Sites
Agriculture
Chemistry
Commerce West
DCL Routing Room
Electrical Engineering
Florida Ave Res Hall
Illinois St Res Hall
Mechanical Engineering
Psychology
Social Science
George Badger
150
DCL
333-4103
Stanley Rankin
150
DCL
333-6530
Joyce Vaughn
150
DCL
333-1637
Gary Bouck
162
DCL
333-7752
138
DCL
333-6133
65
Comm West
333-2170
207
Astronomy
333-7318
Robert Penka
173
DCL
333-4709
Larry Sautter
■ 91
Comm West
333-2170
Larry Sautter
Lynn Bilger
Don McCabe
139
Astronomy
333-6236
164
DCL
333-6285
164
DCL
333-0969
110
baud
333-4000
300
baud
333-4000
1200
baud
333-4001
Sandra Moy
177
DCL
333-4703
J. M. Randal
120
DCL
333-9772
Cliff Carter
195
DCL
333-3723
Jack Knott
194a
DCL
333-6562
Debbie Weller
203
Astronomy
333-8150
Tom Kerkering
16
DCL
333-0816
Cliff Carter
195
DCL
333-3723
Lee Hollaar
193
DCL
333-7904
Rex Duzan
164
DCL
333-6285
M103
Turner Hall
333-8170
153
Noyes Lab
333-1728
70
Comm West
333-4500
129
DCL
333-6203
146
EEB
333-4936
FAR
333-2695
ISR
333-0307
65
MEB
333-2072
453
Psych Bldg.
333-7815
202
Lincoln Hall
333-0309
OFF-LINE is the newsletter of the Computing Services Office at the University of
Illinois at Urbana-Champaign. OFF-LINE is printed monthly. Articles may be re-
printed provided that the source of the article is noted. CSO operates an IBM 360
Model 75 with one million bytes of fast core and two million bytes of slow core,
under HASP and OS, and a CYBER 175 with 256K words of central memory and
512K. words of ECS, under NOS, serving up to 190 simultaneously active text and
graphics terminals, and a DEC PDP-1 1/50 with 252K bytes of core driving a GSI
CAT-8 phototypesetter with the UNIX Operating System.
SYSTEM NOTES
CYBER:
APRIL RELIABILITY REPORT
15 interruptions affecting time-sharing only
14 interruptions affecting the entire system
7 of these were for more than 1 5 minutes
Mean Time Between Failures = 28 hours
Mean Time to Repair = 63 minutes
Availability: 95.8% of the scheduled time
360/75:
21 interruptions affecting the entire system
8 of these were for more than 15 minutes
Mean Time Between Failures = 33 hours
Mean Time to Repair = 51 minutes
Availability: 96.7% of the scheduled time
"Interruptions affecting time-sharing only" means those interruptions from which a user may
normally recover (may continue work being done at the time of the interruption by doing a
RECOVER, tty when the system comes back up). "Interruptions affecting the entire system"
means that a major problem has occurred, and work being done at the time of the
interruption will not be able to be recovered.
CSO would appreciate any comments (pro or con) or suggestions from you, the users, about
the new format of the Reliability Report.
NUMERICAL SERVICES
NEW VERSION OF SPICE
Version 2F. 1 of the SPICE circuit analysis program has been received from the Electronics
Research Laboratory of the University of California at Berkeley. It is available via
GRAB.SPICE/FUTURE.
or
FUTURE.SPICE.
SPICE is then run as follows:
SPICE, mput.output.
Current users of SPICE should note that a "procedure" will no longer be used with the new
version.
The writeup for the new SPICE can be obtained and printed by entering:
WRITEUP.SPICE/FUTURE.
PRINT.SPCDOC/CC.
This will print 71 pages.
The new SPICE will become the default on or about July 1, 1980 and then may be accessed
by simply entering:
GRAB.SPICE.
Please take note of this if you are currently using SPICE, as the present method of using it
(CALL,SPICE...) will not be correct after that date.
CCL VERSIONS OF SOME PROCEDURES AVAILABLE IN FUTURE
CCL versions of several CSO procedure files are available via GRAB, procedure/ FUTURE.
These include USERLIB, TIDY (formerly TIDYPRC), MIMI (formerly CONVPRC),
ACSLCOM, ACSLGO, ALTCOM, ALTGO, DEBLOCK and TBLOCK. In each case, on-line
help is available by entering:
HELP
and then by entering the name of one of the above procedures when prompted with a
question mark.
The new procedures are all called in "control-card" form rather than with a dash (-) or
BEGIN. For instance, with DEBLOCK you can now enter:
DEBLOCK,DISK = xxx.TAPE=yyy
The parameters of all the new versions are the same as the old, with two exceptions: in TIDY
and MIMI, the NR parameter has been replaced with a REWIND parameter, which should be
coded as
REWIND=NO
if you wish to suppress the automatic rewinding of the files used by the procedure.
The "positional parameter" feature of CCL can be used to make some of these procedures
more convenient. For example, in translating and running an ACSL model interactively,
suppose the model definition is in file MODEL. You would enter
GRAB.ACSL/F.
ACSLCOM.MODEL
to translate it, and then enter
ACSLGO,PLOT=TEKTRON.
to run it at a Tektronix graphics terminal. Each procedure should be consulted to check
which parameters can best be given positionally.
These CCL versions will become the "official" default versions on or about September 1,
1980.
SOME PROCEDURE FILES TO BE DISCONTINUED
The procedures MATHDOC, IMSLDOC, MSLDOC, NATSDOC, LINDOC, MISCDOC,
ACSLINF, and ALTINF will no longer be available after July 31, 1980. They have been
superceded by the CCL procedures MATH and SAMPLES, which were described in the May
issue of OFF-LINE. New reference guides are also available on these procedures (RF-4.12
MATH and RF-4.13 SAMPLES).
FORTUOI PHASE-OUT
The bulk of the routines in the 360 FORTUOI library are outdated and should not be used
when substitutes from IMSL or other libraries are available. Since there may be hidden
dependencies on FORTUOI in many programs, the following plan will be used to phase out
the outdated material.
. The routines we desire to preserve will all be placed in a data set called
SYS1.FORTUOIX. On or about September 1, 1980, this data set and the current
SYS1. FORTUOI will be "switched." At this time, programs using old routines will fail
unless they explicitly access SYS1.FORTUOIX.
• Source of the outdated routines will be available via the MATHLIST procedure, using
T=FORTUOIX.
. SYS1.FORTUOIX will remain on-line until the end of the Fall 1980 semester. It will
be destroyed on or about January 1, 1981.
. Source of the deleted routines will remain available until June 1981 at least.
Disposition beyond that date has not yet been determined.
(lowing routines are to be
"removed."
BKTRNZ
DGEARZ
EIGENZ
INV3Z
ORT4Z
SCSLEZ
BROMNZ
DGELGZ
EVIITZ
LRCHZ
PLOTEZ
SLEVBZ
BROWNZ
DIFEQZ
FLPOMZ
MDETZ
PLOTZ
SMEIGZ
CEST1Z
DIFSUB
FRANCZ
MINVZ
POL1Z
SPL1Z
CFIT3Z
DOTP
FSER1Z
NOLINZ
POL2Z
TRAUBZ
CHOL1Z
DVDIFZ
GAUSZ
ORTIZ
POLVAZ
WLSQZ
CHOL3Z
ECON1Z
HOUSEZ
ORT2Z
RAMEZ
XINVZ
CHOL4Z
EIGENP
INV1Z
ORT3Z
RSSR
TEXT PROCESSING
DIABLO RESERVATION SERVICE TO CHANGE
Effective June 2, 1980, users will no longer be able to make reservations to use the Diablo
terminal located in Room 209 Astronomy. Instead, CYBER users will be required to queue
their RNF output, intended for printing on the Diablo, by executing a CYBER Control
Language (CCL) procedure. Users unfamiliar with RNF are referred to the RNF User's
Guide or the RNF Reference Manual (available in the Distribution Center, Room 164 DCL).
All RNF source files must contain the .LPT command in the first line of the file, and must be
processed by RNF into an output file. All non-RNF source files must have carriage control
and begin with a page eject. The RNF output file or the non-RNF source file can then be
submitted to the Diablo queue via the CYBER CCL procedure.
The following conventions have been established:
1. The top of the form (page) will be set to one line below the horizontal perforation
in the paper.
2. Plain white 20-lb paper, 14 7/8" wide by 11" long, will be used.
3. If the pica font is selected, the printing will be done with 10 characters and 6 lines
per inch.
4. If the elite font is selected, the printing will be done with 12 characters and 6 lines
per inch.
After you have prepared your file for submission, you should access and execute the
DIABLO procedure by issuing the following commands from time-sharing on the CYBER:
GRAB.DIABLO.
DIABLO.
The procedure then prompts you for your name, telephone number, choice of font (pica or
elite), and the name of your RNF output file. The RNF output file may be either a local or a
permanent file. If the procedure is unable to find your file, it aborts with a message to that
effect. If the procedure finds the specified file, it places it in the Diablo queue.
On-line help is available by entering the command
HELP
and then by entering DIABLO (the name of the procedure) after the question mark prompt.
Turnaround time should be no longer than 24 hours, except in cases of system downtime or
hardware malfunctions. The output of jobs received on Fridays or prior to a University
holiday will be available the next working day.
Your printouts may be picked up at 203 Astronomy. The cost will continue to be $0.15 per
page, rounded up to the nearest dollar. You may pay for the printing by charging it to a
University account number or with a personal check.
Questions regarding this service should be directed to Debbie Weller, 209 Astronomy (333-
8150).
MISCELLANEOUS
TELETYPE REPAIR DISCONTINUED
As stated in the February 1979 OFF-LINE, and again in the October 1979 OFF-LINE, the
maintenance and repair service relating specifically to "Teletype" terminals will be discontinued
June 30, 1980. This applies only to the service offered on Teletype Models 33 and 35; it does
not affect any other offering.
LSI-11 RENTAL SYSTEMS
We have now received the equipment necessary to start our rental program using the DEC
LSI-1 1. We are offering a package system with the components listed at the end of this
article. Our initial rental program is going to be fairly simple. We will rent entire systems on
a quarterly basis at $1500 per quarter.
In the event that the customer wishes to purchase the system later, it is available at $500 over
our purchase cost and 50% of the rental, which has been paid, will be credited against the
purchase price. By paying the $500, the system is included in our general LSI-1 1 support
program which was described in a recent issue of OFF-LINE (December 1979). The rental
price includes maintenance and a reasonable amount of discussion with our support
personnel. The system is fully licensed with FORTRAN, BASIC and the operating system.
KD11-HA
KEV11
CI- 1103
DLV11-J
BDV11
BA11-NE
RXV21-BA
H984
LSI- 11 RENTAL SYSTEM
LSI-11/2CPU
Arithmetic Instruction Chip for KD11
32K word MOS Memory
4 Line Serial Interface RS2326
Bootstrap/Terminator/Diagnostic
9 Slot System Box
Dual Drive Dual Density Floppy
Cabinet
Options (No charge unless noted)
TU58-BB Dual DECTAPE II
ADV11 16 Line MUX'D A/D
AAV 11 4 Line D/ A
KPV1 1 Power Sequencer/Controller with Console
Lier Siegler ADM 3A with RG512 Graphics (extra $250/quarter)
Software
RT11
FORTRAN /RT1
BASIC /RT11
Manuals
Operating System
DEC INFORMAL USERS GROUP MEETING
(The following article was submitted by J. Wray)
On April 29th people using PDP-8's, LSI-ll's, VAX's, DEC-20's and DEC-10's, from both
inside and outside the University, were noticed sitting in a room together. What was it? It
was the first area-wide informal DEC Users Group Meeting.
Some of the meeting was given over to a general discussion of possible meeting topics and
frequencies. Much of the value of such meetings is (and was) a chance to meet other users
of similar machines and find out what their problems and solutions have been. Time was left
for such discussions, and that will be planned for in the future meetings.
After the meeting the volunteer coordinating committee met and responding to some of the
suggestions set up the following for a trial run:
1. Meetings are scheduled for June 24 and July 22 in Room 115 Digital Computer
Lab. (Since there is a slight chance that we will have to change rooms if we
conflict with a scheduled class, we suggest that you arrive a little early. You may
have to walk a block or two to a different room.) The meeting will start at 4 and
the room is reserved until 5:30. All users of DEC computers on and off campus
are welcome.
2. At the June 24 meeting the Chemistry, Coordinated Science, Materials Research,
Physics and Psychology systems will be described briefly.
3. The July 22 meeting will be an RT11 Special Interest Group (SIG) meeting for
RT1 1 users in this area.
4. The questionnaires which have been returned have been xeroxed and members of
the coordinating committee have copies. CSO has distributed a similar
questionnaire for mini-micro users in general. We will provide copies of the
questionnaires we have collected to CSO for inclusion in their publication, so you
do not have to fill out both.
In addition to the above, the following topics were suggested for future meetings:
1 . RT 1 1 to RSX 1 1 M migration.
2. Software packages available and their use.
3. Presentations by non-DEC vendors (hardware and software).
The following people volunteered to serve on a coordinating committee:
Name Address Telephone
ArtGaylord 152 Noyes Lab 333-1728
Virginia Metze 244 Materials Research Lab 333-6665
Walter Schneider Psychology 333-6819
Randy Stein 255 Morrill 333-3245
Steve Wilkus 2-107 Coordinated Science Lab 333-6444
Jerry Wray 487 Loomis Lab 333-4922
All the above are at the University of Illinois. The mail address of the Psychology Building is
in Champaign, all the rest are in Urbana.
PURPOSES OF THE INFORMAL DEC USERS GROUP
Some questions and comments at the meeting made it clear that some clarification of the
distinction between the informal users group and the CSO LSI- 11 project is needed. Simply
stated, they are separate but are somewhat parallel in intent. They do share some of the
same broad goals.
If the users group had been in existence for a few years, it is likely that the LSI project would
have been suggested by the group in recognition of the specific needs of a particular subgroup
of users - the LSI users on campus. The CSO LSI- 1 1 project was formed in recognition of
that specific need.
The CSO project is currently only concerned with on-campus users of LSI-ll's. They want to
be able to help people get or stay on the air by providing quick and easy access to hardware.
Help with system software problems is desirable, but as yet it is unclear how much of it they
will be able to provide. There is, of course, a charge to the user of these services and
facilities. A major function of the project is the collection, in a single location, of information
on existing programs or hardware in use on the LSI equipment.
On the other hand, facilitating interchanges among users of all DEC machines is one of the
main reasons for setting up a users group. These interactions might be trading user software,
ideas for new hardware, ways to get around problems, fixing software bugs, etc. The users
group does not have any hardware 'warehousing' function or budget. Also, the group is open
to users of any DEC computer, not just LSI-ll's, and not just campus users. There is no
charge for participating in the users group.
This is by no means a complete list of parallels, differences, shared goals, etc, but we hope
that it is clear that the two are not competing. There are needs being met by both and areas
where both feel that the other is the obvious group to carry the ball. People who are
involved in the LSI- 11 project will probably want to be in the users group as well, and people
who are not interested in the LSI- 11 project may still find that the users group fills a need.
A COSMIC EXPERIENCE
One of our users, Paul Opryszek, was working in DCL on May 1 when the power went out.
In his own words, "the system crashing was no big thing - but the lights going out too, well,
that was truly cosmic." He has written and submitted the following poem about that
experience and we thought our readers might enjoy sharing it with us.
A Cosmic Experience
The human spirits were
Glued at video screens. The
Master nourished
only Total awareness.
Access space, never enough
Intense competition always.
For access to the
Power intense.
For the human ghosts
Flickering patterns of
Energy were more
Real than the
Three space geometry
Through which their bodies
Groped as they lived in the
System's Reality.
Lights flash brightly
Too bright, too fast.
Flashing energy pulsed
Too bright, too fast.
Terminals surged.
Buzzzzap!
First thoughts. ..a
Problem, programmable, solvable?
Darkness
Stunned, total silence
Unbelieving, uncomprehending.
Anxious, held breath.
Reality had Ended!
Reality had Crashed!
Shaft of a dim
Emergency light from
Above.
All had died.
Each contemplated
If there was
Life after Death in the
Strange new three space geometry's
reality.
Anxious questioning of a
Time after life,
Had the Creator
Wiped clean the Disk?
Was resurrection possible?
Shadows, silence, anxiety
Replace Reality of Energy.
Human ghosts
fade away
Copyright 1980, by Paul Opryszek
OFF-LINE's Mailing List
If you wish to be placed on our mailing list for future issues of OFF-LINE, if you wish to be
removed from the list, or if you wish to enter an address correction, please complete and
return this page. (Current subscribers are kept on the mailing list until a specific request for
removal is received, or until a mailing is returned as undeliverable.)
Check one:
□ New subscriber
□ Removal request
□ Address correction
Name:
Address:
CAMPUS or Zip Code
(If address correction, give old address and zip code below.)
Comments:
RETURN TO:
OFF-LINE
164 Digital Computer Laboratory
University of Illinois at Urbana-Champaign
Urbana, Illinois 61801
■> -is,
L t
0 r-
>x
■
-a
c
Computing
Services
Office
" LIBRARY Of U
JUL 2 3 mo
UHhtk.Mii »>■ in mult
Off-Line
University of Illinois at Urbana-Champaign
VOL. 8, NO. 7 July 1980
EDITOR: Lynn Bilger
PHONE: (217) 333-6236
164 Digital Computer Lab.
Urbana, Illinois 61801
Page
Contents
2
4
5
POLICY
Keypunch Service
SYSTEM NOTES
May Reliability Report
CYBER
New ZETA Plotters
Some Utiltiies Allow Continuations
More Nonlinear Programming Algorithms
FLEX and GMET
SIMULA Now Available Through GRAB
STATISTICAL SERVICES
BMDP-77 on the CYBER
HELP WANTED
Needed - An 8-Line Asynchronous ULM
CSO DIRECTORY - STAFF AND SERVICES
Administrative
Director
George Badger
150
DCL
333-4103
Business Manager
Stanley Rankin
150
DCL
333-6530
Secretary
Joyce Vaughn
150
DCL
333-1637
User Services
User Accounting
Gary Bouck
162
DCL
333-7752
Systems Consulting
166
DCL
333-6133
Statistical Services Consisting
65
Coram West
333-2170
Text Processing Consulting
207
Astronomy
333-7318
Asst Dir User Services
Robert Penka
173
DCL
333-4709
Mgr, Statistical Services
Larry Sautter
91
Comm West
333-2170
Mgr, UNIX Operations
Larry Sautter
Documentation
Lynn Bilger
Don McCabe
139
Astronomy
333-6236
Distribution Center
164
DCL
333-6285
Hardware/Software Support
Terminal Repair Service
164
DCL
333-0969
CYBER Dial-up Numbers
110
baud
333-4000
300
baud
333-4000
1200
baud
333-4001
Asst Dir Systems and Operations
Sandra Moy
177
DCL
333-4703
Asst Dir Development
J. M. Randal
120
DCL
333-9772
Asst Dir Engineering
CYBER-IBM Operations
Cliff Carter
195
DCL
333-3723
Jack Knott
194a
DCL
333-6562
UNIX Operations
Debbie Weller
203
Astronomy
333-8150
Telecommunications
Tom Kerkering
16
DCL
333-0816
Hardware Selection Assistance
Cliff Carter
195
DCL
333-3723
Laboratory Support Project
Lee Hollaar
193
DCL
333-7904
RJE Operations
Rex Duzan
164
DCL
333-6285
RJE Sites
Agriculture
M103
Turner Hall
333-8170
Chemistry
153
Noyes Lab
333-1728
Commerce West
70
Comm West
333-4500
DCL Routing Room
129
DCL
333-6203
Electrical Engineering
Florida Ave Res Hall
146
EEB
333-4936
FAR
333-2695
Illinois St Res Hall
ISR
333-0307
Mechanical Engineering
65
MEB
333-2072
Psychology
453
Psych Bldg.
333-7815
Social Science
202
Lincoln Hall
333-0309
OFF-LINE is the newsletter of the Computing Services Office at the University of
Illinois at Urbana-Champaign. OFF-LINE is printed monthly. Articles may be re-
printed provided that the source of the article is noted. CSO operates an IBM 360
Model 75 with one million bytes of fast core and two million bytes of slow core,
under HASP and OS, and a CYBER 175 with 256K words of central memory and
512K words of ECS, under NOS, serving up to 190 simultaneously active text and
graphics terminals, and a DEC PDP-11/50 with 252K bytes of core driving a GSI
CAT-8 phototypesetter with the UNIX Operating System.
POLICY
KEYPUNCH SERVICE
As announced in the April issue of OFF-LINE, CSO has discontinued internal processing of
keypunch work. This move was made for financial reasons since the amount of data entry
provided by CSO no longer justified continuance of a dedicated data entry staff.
At the time the April article was written, we planned to use an off-campus commercial data
entry service with CSO providing all interfacing with the contracting agency. Since that time,
another campus unit, Survey Research Lab (SRL), has indicated a desire to assume CSO's
data entry workload. An agreement has been worked out with SRL which we hope will
minimize the amount of information which must be transferred between CSO and SRL with
the attendant possibilities for confusion.
Users whose data entry work is funded through the Research Board will see little change.
Requests for data entry funds are to be made through existing channels, input is to be
delivered to CSO for punching at SRL, and output is to be picked up at CSO. Allocations will
continue to be made in numbers of cards to be punched and CSO will continue to provide all
accounting associated with data entry. Occasionally, a user with a complex data entry task
may be asked to talk directly to the data entry staff at SRL. Data entry work funded through
the Research Board is to be taken to the CSO Accounting Office and completed data entry
jobs are to be picked up there. This office, as well as document sales and terminal rentals, is
scheduled to move to 1208 West Springfield during July.
Users whose data entry work is being funded with hard contract money are to deal directly
with SRL. These users are to submit their work directly to the SRL data entry staff, and
claim the completed work at SRL. They will be billed directly by SRL for services delivered.
SRL's rates are $7.25/hour for normal work and $10.50/hour for priority work. The number
of cards produced per hour will vary with the data being entered. A rough estimate, based on
previous experience in CSO's data entry shop, is 100 cards per hour. Users paying for data
entry services with hard contract funds should make all arrangements with Mrs. Frances
Sykes, 310 SRL, 1005 West Nevada, Urbana (333-7328).
SRL maintains a staff of three full-time data entry operators and hires part-time help to staff a
second shift whenever the load requires it. Additional full-time personnel will be hired if
needed to maintain reasonable turnaround time.
SYSTEM NOTES
MAY RELIABILITY REPORT
CYBER 175: 5 recoverable interruptions
13 non-recoverable interruptions
7 of these were for more than 15 minutes
Mean Time Between Failures = 32 hours
Mean Time to Repair = 63 minutes
Availability: 96.5% of the scheduled time
Major cause of downtime was related to
software problems with disk.
IBM 360/75: 32 non-recoverable interruptions
15 of these were for more than 15 minutes
Mean Time Between Failures = 22 hours
Mean Time to Repair = 25 minutes
Availability: 94.9% of the scheduled time
Major cause of downtime was related to
hardware problems with CPU.
"Recoverable interruptions" means those interruptions from which a user may normally
recover (may continue work being done at the time of the interruption by doing a
RECOVER, tty when the system comes back up). "Non-recoverable interruptions" means that
a major problem has occurred, and recovery of work in progress at the time of interruption is
not possible.
CYBER
NEW ZETA PLOTTERS
Because CalComp will no longer guarantee maintenance of the CalComp 763 and CalComp
1136 plotters in use at CSO, two new ZETA plotters (Models 1453 and 3653SX) have been
installed as replacements. Software has been developed for their use on the CYBER and they
are now in limited service in parallel with the CalComp plotters.
Each plotter contains a builtin microprocessor-based controller which permits it to be
connected to the CYBER as if it were a 1200 baud terminal and provides hardware character
and vector generation. The 1453 is a small desktop-sized unit which plots on 12-inch wide
paper (only 1 1 inches of usable space). It has an increment size of .0025 inches and can plot
at rates up to 14.14 inches per second diagonally. The 3653SX is a large floor-standing unit
which plots on 36-inch wide paper (only 34 inches of usable space). It has an increment size
of .001 inches and can plot 21 inches per second diagonally. An interesting hardware feature
of the 3653SX is a sliding "throttle" which allows the operator to slow down the plotter if this
is needed to maintain plot quality with the pens, ink and paper in use. Both plotters have
four program-selectable pens and can support ballpoint, rolling writer and liquid ink pens.
Rolling writer plots, new to CSO, are darker than ballpoint pen plots and are nearly the equal
of liquid ink plots in quality. The smallest liquid ink pen supported by the ZETA plotters is
.25mm. (Note: Liquid ink pens are not currently supported, but will be in the near future.)
Libraries which generate ZETA plot codes and a new command, PLOTZ, have been installed
on the CYBER. To use the new plotters, CYBER users should use different GRAB
commands to access the new libraries:
OLD NEW
GRAB,GCSCALC. GRAB,GCSZETA.
GRAB,CALCOMP. GRAB,ZETA.
The command, PLOTZ, must be used instead of the command, PLOT, to route the files to
the plot queue. At the present time, only a single file may be used on the PLOTZ command.
CalComp users who are calling the symbol routine with an integer equivalent rather than a
character should be aware of differences in the integer equivalence tables. The ZETA integer
equivalence tables can be seen in the Systems Consulting Office (166 DCL). No other
program changes should be required.
The following options are available on the PLOTZ command:
There are two types of pens used: ROLLING and BALL. If you do not specify a
/Px= option (where x is 1, 2, 3, or 4 for the four pens, respectively), the default
action will be as follows:
/PI =ROLLING (rolling writer pen)
/P2=BALL (ball point pen)
/P3=BALL
/P4=BALL
You may change this action by specifying /PI =BALL or /P4 = ROLLING, for
example.
You may specify or change the default pen color by using the /Cx= option (where x
is 1, 2, 3, or 4 for pens 1 through 4, respectively). If you omit this option, the default
colors are:
/C1=BLACK
/C2=BLUE
/C3=GREEN
/C4 = RED
You may change the default by specifying /C2 = RED or /C4= BLACK, for example.
. The /FORMS = parameter must be one of the following:
FANFOLD 8.5 by 11 inch fanfold paper
NOTE: This is the default.
WIDE 34-inch wide roll paper
ROLL 1 1-inch wide roll paper
. The /TIME= option may be used to specify the maximum plotter time in minutes.
Default value is 5 minutes.
. The /LENGTH = option may be used to specify the maximum plot length in inches.
Default is 51 inches (which is 6 fanfold pages).
. The /COPIES = option may be used to specify the number of copies desired.
Maximum number of copies is 31; default value is 1. NOTE: Currently, only one copy
is allowed.
. The /BIN= option may be used to specify a specific bin number from 00 to 99 in
which to place the output. Default is the last two digits of your ID number.
. The /JOBNAME= option sets the jobname which will be printed on the plot burst
page. The name may not be longer than 8 characters.
Plots made on the default fanfold paper fold flat and can be transported easily without
damage. Users should be aware, however, that a pen in the "pen up" position may drag
across the "out fold" as the fold passes under the pen, producing a short unwanted line.
Because of this, best results are obtained when each plot fits on a 8.5-inch by 11-inch page.
Plots not satisfying this condition should be plotted on roll paper.
Users should also be aware that unlike the PLOT command, PLOTZ will not produce any
printer output. Plot output will be found in the usual places. Plots on fanfold paper will be
found in trays next to the roll paper plot bins with plots stored in numbered trays according
to the last digit of their bin number. Users not finding their fanfold plots are asked to check
the appropriate roll paper plot bin since we may have to generate the plot on roll paper under
some circumstances.
We are presently working on ZETA plotter support for the IBM 360/75 and expect to have
software in production by mid-July. IBM users will be urged to use this software when it is
announced. Use of the ZETA rather than the CalComp plotters by IBM users is mandated by
concerns for both reliability and turnaround time. Since the IBM 4341, which will replace the
360/75 by the start of the fall semester, will not support the on-line CalComp plotter, only
the off-line CalComp plotter will be available to process user jobs. The off-line CalComp
plotter will remain in service for a period of time sufficient to allow users to convert to the
ZETA plotters.
SOME UTILITIES ALLOW CONTINUATIONS
The local utilities, PRINT, PUNCHC, PLOT, TYPE, and SENDJOB, now allow for
continuation lines and a maximum of 20 file names. Previously, there was a limit of 10 file
names.
To indicate continuation, end the command with a comma. If you are running a card batch
or deferred batch job, or are using a procedure file, the next line should be the continuation.
If you are entering the command at a time-sharing terminal, you will be prompted with the
message:
CONTINUATION:
Then, you simply finish the command line.
You may have any number of continuation lines by simply terminating each line with a
comma. To end the continuation, place a period at the end of the command line.
MORE NONLINEAR PROGRAMMING ALGORITHMS
FLEX AND GMET
FLEX: The flexible tolerance range method, FLEX, is an algorithm for solving nonlinear
optimization problems. FLEX is available on the CYBER and can be accessed with the
statement:
GRAB.FLEX.
The use of FLEX requires that the user provide a user subroutine and a data file. The format
of the time-sharing control statements used to execute FLEX is:
FTN,REW,I = user subroutme,L=0.
P.LOAD(LGO);FLEX,dafa file.lfn.
Where Ifn is a local file to which the FLEX results are written.
GMET: GMET is an algorithm for solving geometric programming problems. GMET can be
accessed on the CYBER with the statement:
GRAB.GMET.
The format of the time-sharing control statements used to execute GMET is:
GET.dafa file.
GMET.dafa file.lfn
Where data file is the user supplied file containing information about the problem and Ifn is
the local file to which the GMET results are to be written.
A user manual for GMET and a writeup containing more information about FLEX are
available for inspection in the Systems Consulting Office, Room 166 DCL. Questions or
problems about either GMET or FLEX should be directed to Manoochehr Ghiassi (Room
193 DCL, 333-7904).
SIMULA NOW AVAILABLE THROUGH GRAB
CDC has dropped support of the SIMULA compiler with Release 5 of the NOS Operating
System. However, CSO has made the CDC SIMULA 67 compiler available through GRAB.
It may be accessed by entering the statement:
GRAB.SIMULA
On or about September 1, 1980, we will no longer allow SIMULA to be used as a system
command. Therefore, we suggest that all procedure files referring to SIMULA be converted
to include the "GRAB,SIMULA." statement. Note that this is exactly the same SIMULA
compiler as is currently on the system.
The "S" option (and all of its associated options) on the SIMULA compiler does not presently
work (and we are uncertain if it ever did work here). If time permits, we will fix this
problem some time in the future.
STATISTICAL SERVICES
BMDP-77 ON THE CYBER
The set of biomedical statistical programs, BMDP-77, has been installed on the CYBER. The
manual, BMDP Biomedical Computer Programs P-Series 1979, is available at the CSO
Distribution Center.
In addition, a document describing the differences between the CYBER BMDP-77 version
and the IBM BMDP-79 version is available at the Statistical Services Office (65 Commerce
West) or may be obtained by entering the following commands at the terminal:
WRITEUP.BMDP.
PRINT,BMDP/CC/EJ/RJE= remote site.
To run BMDP from a time-sharing terminal, assuming the problem information and data are
already in filel and the output is to be written to file2, enter the following:
GRAB.BMDP
BMDP(P=BMDPxx,l = Mei.L=f//e£i
Where BMDPxx is the name of the particular BMDP routine you wish to use, e g BMDP1D
or BMDP1M.
To run BMDP from cards or deferred batch (using SENDJOB or SUBMIT) use:
From Cards From Deferred Batch
/•CYBER
jobname.
SIGNON university id.
password.
B\LL,charge,proi jet.
PRINT/CC/EJ/RJE=remofe site
GRAB.BMDP.
BMDP(P=BMDPxx;
7/8/9
(BMDP program and data)
7/8/9
/•
/JOB
jobname.
SIGNON university id.
password.
Bl LL, charge.project
PRINT/CC/EJ/RJE=remofe site
GRAB.BMDP.
BMDP(P=BMDPxx;
/EOR
(BMDP program and data)
/EOR
Where BMDPxx is the name of the particular BMDP routine you wish to use.
HELP WANTED
NEEDED - AN 8-LINE ASYNCHRONOUS ULM
I need to switch from a 4-line asynchronous QTY (Data General Model 4060) to an 8-line
asynchronous ULM (Data General Model 4241 -A). Will have a surplus QTY to trade. If
there is a NOVA minicomputer user who needs a 4-line QTY, please contact H. B. Puckett,
333-0808.
OFF-LINE 's Mailing List
If you wish to be placed on our mailing list for future issues of OFF-LINE, if you wish to be
removed from the list, or if you wish to enter an address correction, please complete and
return this page. (Current subscribers are kept on the mailing list until a specific request for
removal is received, or until a mailing is returned as undeliverable.)
Check one: □ New subscriber
□ Removal request
□ Address correction
Name:
Address:
CAMPUS or Zip Code
(If address correction, give old address and zip code below.)
Comments:
RETURN TO: OFF-LINE
164 Digital Computer Laboratory
University of Illinois at Urbana-Champaign
Urbana, Illinois 61801
• V.
L I 1 '-
1-1
Computing
Services
Office
Off-Line
University of Illinois at Urbana-Champaign
VOL. 8, NO. 8 August 1980
EDITOR: Lynn Bilger
PHONE: (217) 333-6236
164 Digital Computer Lab.
Urbana, Illinois 61801
Page
4
5
5
6
11
12
13
13
14
14
Contents
POLICY
Rate Reduction Period
Service Unit Rates to be Reduced
Rate Changes for the IBM 4341
SYSTEM NOTES
June Reliability Report
CYBER
CSO Videotapes
Print and Punch Charges More Accurate
On-line 763 Plotter Gone
NUMERICAL SERVICES
IMSL Edition 8
ACSL Version 6A
CCL Reminder
MISCELLANEOUS
CSO Short Courses
Mailing List Update
HELP WANTED
Needed - Computer Programmer
Programmer Needed
tffc
&*#
sd
c'
tv«£
CSO DIRECTORY - STAFF AND SERVICES
Administrative
Director
George Badger
150
DCL
333-4103
Business Manager
Stanley Rankin
Joyce Vaughn
150
DCL
333-6530
Secretary
150
DCL
333-1637
User Services
User Accounting
Gary Bouck
1208
W Springfield
333-7752
Systems Consulting
166
DCL
333-6133
Statistical Services Consulting
65
Comm West
333-2170
Text Processing Consulting
207
Astronomy
333-7318
Asst Dir User Services
Robert Penka
173
DCL
333-4709
Mgr, Statistical Services
Larry Sautter
91
Comm West
333-2170
Mgr, UNIX Operations
Larry Sautter
Documentation
Lynn Bilger
139
Astronomy
W Springfield
333-6236
Distribution Center
1208
Hardware/Software Support
Terminal Repair Service
164
DCL
333-0969
CYBER Dial-up Numbers
no
baud
333-4000
300
baud
333-4000
1200
baud
333-4001
Asst Dir Systems and Operations
Sandra Moy
177
DCL
333-4703
Asst Dir Development
J. M. Randal
120
DCL
333-9772
Asst Dir Engineering
CYBER-IBM Operations
Cliff Carter
195
DCL
333-3723
Jack Knott
194a
DCL
333-6562
UNIX Operations
Debbie Weller
203
Astronomy
333-8150
Telecommunications
Tom Kerkering
16
DCL
333-0816
Hardware Selection Assistance
Cliff Carter
195
DCL
333-3723
Laboratory Support Project
Lee Hollaar
193
DCL
333-7904
RJE Operations
Rex Duzan
164
DCL
333-6285
RJE Sites
Agriculture
M103
Turner Hall
333-8170
Chemistry
153
Noyes Lab
333-1728
Commerce West
70
Comm West
333-4500
DCL Routing Room
129
DCL
333-6203
Electrical Engineering
146
EEB
333-4936
Florida Ave Res Hall
FAR
333-2695
Illinois St Res Hall
ISR
333-0307
Mechanical Engineering
65
MEB
333-2072
Psychology
453
Psych Bldg.
333-7815
Social Science
202
Lincoln Hall
333-0309
OFF-LINE is the newsletter of the Computing Services Office at the University of
Illinois at Urbana-Champaign. OFF-LINE is printed monthly. Articles may be re-
printed provided that the source of the article is noted. CSO operates an IBM 360
Model 75 with one million bytes of fast core and two million bytes of slow core,
under HASP and OS, and a CYBER 175 with 256K words of central memory and
512K words of ECS, under NOS, serving up to 190 simultaneously active text and
graphics terminals, and a DEC PDP-11/50 with 252K bytes of core driving a GSI
CAT-8 phototypesetter with the UNIX Operating System.
POLICY
RATE REDUCTION PERIOD
During the period between August 1 and September 1 inclusive, the rates charged to internal
University users for CSO's CYBER and IBM services will be reduced by 40 percent. All
services billed at the end of the job are included in this rate reduction. This includes charges
for tape and disk mounts, printing, punching, and plotting as well as charges for I/O,
memory, CPU seconds, connect time charges and, where applicable, line transmissions.
UNIX rates, and rates for disk space and keypunching will not be reduced.
The mechanism used to reduce rates during similar periods in the past has been to devalue
the dollar value of the service unit. Users whose computing was funded through the
Research Board and students who utilized class funds were not directly benefited since their
allocation of computer time was denominated in service units rather than in dollars. As a
group these users did benefit since the reduced dollar value of the bill paid by the campus on
their behalf allowed the campus to allocate more service units. Users who paid computing
charges with hard contract funds benefited directly since the bills generated at the end of the
month were for smaller dollar amounts.
A different mechanism will be used during this year's period of reduced rates. Briefly,
computer charges will be calculated internally the same as now, but a multiplier of 0.6 will be
applied to the service unit cost of each job before its cost is recorded in the billing database
and before any PS number or user number balance is decremented by the cost of the job.
Since the service unit cost of the job is being reduced, all users, regardless of the source of
their funding, will benefit directly.
SERVICE UNIT RATES TO BE REDUCED
Effective after the end of the present discounted price period (September 1, 1980), the cost
of a service unit will be reduced from $$0.68 to $0.62. This continues the planned reduction
of prices which we have been following for four years.
In addition, the prices for IBM services are being revised to reflect the new IBM 4341 system.
These pricing changes are explained in the following article.
RATE CHANGES FOR THE IBM 4341
With the installation of the IBM 4341, which is replacing the IBM 360/75, a revision in
pricing policy for IBM services will become effective September 2, 1980. These prices will be
for the continuation of old services available under OS/MVT in batch mode. Further
revisions will occur with the anticipated initiation of terminal-based services.
The two principal changes are to be in the areas of disk storage and the effect of memory on
CPU or I/O charges.
Disk charges will be reduced from 0.01 service units per track per day to 0.0017. At the rate
of $0.62 per service unit, this would mean a charge of 3.93 service units at a cost of $2.43 to
store one million bytes for one month.
Processing charges are to be computed according to the formula:
0.00036 (100T + IOH0.0014R + 0.7)
Where:
T = CPU time in seconds
IO = number of IO requests issued
R = memory Region in thousands of bytes
This wil replace the old formula:
0.00036U00T + IOM0.0045RF + 0.001 5RS + 0.5)
where RF and RS were fast and slow regions, respectively.
The effect of the change will depend on the memory required, and the relative speed of the
two machines for a specific job. The larger the memory requirement, the greater the decrease
in service units per hour. The following examples should illustrate this -- assuming that the
4341 is 85 percent as fast as the 360/75.
Region in
Service Units Charged
1000 bytes
to do a One-hour 360 Job
360/75
4341
0
64.80
106.73
128
139.44
134.05
256
214.09
161.37
512
363.40
216.02
1024
662.00
325.31
The new formula will give an approximate equivalence to CYBER batch charges, where the
memory component was reduced a year ago.
SYSTEM NOTES
JUNE RELIABILITY REPORT
CYBER 175: 6 recoverable interruptions
18 non-recoverable interruptions
1 2 of these were for more than 1 5 minutes
Mean Time Between Failures = 22 hours
Mean Time to Repair = 26 minutes
Availability: 97.9% of scheduled uptime
Major cause of downtime was related to
software problems with disk and 2550's.
IBM 360/75: 69 non-recoverable interruptions
48 of these were for more than 15 minutes
Mean Time Between Failures = 10.5 hours
Mean Time to Repair = 28 minutes
Availability: 94.2% of scheduled uptime
Major cause of downtime was related to
hardware problems with CPU.
"Recoverable interruptions" means those interruptions from which a user may normally
recover (may continue work being done at the time of the interruption by doing a
RECOVER,tty when the system comes back up). "Non-recoverable interruptions" means that
a major problem has occurred, and recovery of work in progress at the time of the
interruption is not possible.
CYBER
CSO VIDEOTAPES
CSO has recently completed the initial stages in the preparation of videotapes describing the
CYBER 175. The following three tapes are available for viewing.
CSOVT1 Introduction to Computing at CSO. This videotape is intended for the
first time computer user, and for people wishing to learn more about
CSO's facilities.
CSOVT2 This videotape consists of four sections:
. Using a Terminal - How to operate a computer terminal.
. Introduction to CYBER Time-Sharing - How to access the CYBER
using a terminal.
. File Usage - Local files and indirect access permanent files.
• Introduction to ICE Text Editing - The basics of time-sharing
editing.
CSOVT3 This videotape, Running a FORTRAN Program on the CYBER, consists
of three sections.
. Concepts involved
. PROGRAM statement
. CYBER control statements
Anyone may view these tapes by going to the Undergraduate Library in person to make a
reservation to use the videotape equipment. The library hours are 8 AM - 8 PM Monday
through Thursday, 8 AM - 5 PM Friday, and 1 PM - 5 PM Saturday and Sunday.
When you check out the videotapes, you should also pick up one of the handouts that are
available with the tapes. The handout summarizes the contents of each tape in more detail,
and facilitates note taking.
Your opinions regarding these videotapes would be greatly appreciated. You may use the last
sheet of the handout, or write directly to:
CSO VIDEOTAPES
164 DCL
CAMPUS
PRINT AND PUNCH CHARGES MORE ACCURATE
Print requests from the CYBER to any of the IBM remote job entry (RJE) sites were not
always charged correctly in the past. As a result of some changes which became effective July
23, the charges for printing to the IBM now will be as follows:
• The user number issuing the print request will be billed at the standard rate for the
current charge, project number in use.
• The PS number will be billed:
at the correct rate for the RJE site (some remotes have different rates)
only for lines actually printed - if a print request is cancelled, there will be no
charge for any unprinted portion
for unused print lines caused by specifying /EJECT (same as the IBM)
• If the print request is cancelled or sent to a remote that is charged at a non-standard
rate, the PS number will be charged correctly, but the user number will be charged as
if all lines were printed at the standard rate.
• In addition, cards sent to the IBM punch will be charged at the correct rates.
Previously, they were charged at a reduced rate.
Overall, the charging will more accurately reflect the published rates.
Until now (July 23), this situation of non-standard rates at some RJE sites required special
handling by the CSO Accounting Office. For example, the Accounting Office had to search
through the records and issue refunds for things such as charges being made for printouts
which were cancelled, etc. Due to the new changes, this will no longer be necessary.
ON-LINE 763 PLOTTER GONE
The CALCOMP 763 on-line plotter has been removed from service at CSO as planned when
the new plotters were ordered. Hardware problems caused the plotter to go down several
weeks ago. Since the cost of repair was very high, and the plotter would not be compatible
with the IBM 4341 that is currently being installed, CSO decided to remove it from service
sooner than originally anticipated.
All CYBER users are urged to use the new ZETA plotters which were recently installed. The
July issue of OFF-LINE contained an article on the ZETA plotters and how to use them.
NUMERICAL SERVICES
IMSL EDITION 8
IMSL has released Edition 8 of their subroutine library; we have received and installed both
the IBM version and the CDC CYBER version for testing.
The IBM version is available in dataset SYS9.IMSL, which can be accessed by using
LIBFILE=SYS9.IMSL in your FORTRAN JCL. For example:
// EXEC FORTLDGO,UBFILE=SYS9.IMSL
To access the CDC version, enter the statement
FUTURE.IMSL
in place of GRABJMSL. Edition 8 will become the default on September 1, 1980.
Changes in Edition 8 include four deleted routines and 41 new routines. The deleted
routines are:
GGAMS - replaced by GGAMR
GGAMT - replaced by GGAMR
GGBIR - replaced by GGBM
GGMLT - replaced by GGMTN
Exploratory Data Analysis
Four new routines implementing some of the techniques of exploratory data analysis have
been added to the Library. One routine, BEMDP, does median polish of a two-way table;
another routine, BOLTV, produces "letter value" summaries. A third new routine, USSLF,
produces stem and leaf plots, and a fourth new routine, USBOX, yields boxplots of one to
several samples on a single set of axes. The output from this latter routine allows easy
comparison of the samples and also provides clear indication of the dispersion characteristics
of each sample.
Categorized Data Analysis
Abilities for log linear model analysis have been provided in a new routine, CTLLF, that does
iterative proportional fitting. Another new routine, CTPR, computes exact probabilities for
two-way tables. This latter routine supplements the existing routine CTRBYC.
Differential Equations
A subroutine, DTPTB, which solves differential equation systems with two-point boundary
conditions has been added. This routine utilizes a multiple shooting technique, using IMSL
initial value routine DVERK to solve the differential equations each "shot." Another new
routine, DBLINT, calculates double integrals, using DCADRE to calculate each simple
integral.
Eigensystem Analysis
A code, EIGBS, to find eigenvalues and eigenvectors of band symmetric matrices is included
in Edition 8. The existing routines for real symmetric and complex Hermitian matrices have
also been extended to allow input in full storage mode.
Transforms
New routines include an inverse Laplace transform code, FLINV, and a subroutine, FFT3D,
which calculates fast Fourier transforms of two- and three-dimensional arrays.
Random Number Generation
Ten new routines have been added to Chapter G in the Library. GGUO and GGNO are
routines to generate order statistics from a uniform and a normal distribution, respectively.
Any set of order statistics from the i-th to the j-th from a given sample size may be
generated. Two other new routines are for generation of variates from a nonhomogeneous
Poisson process using an efficient thinning method. Another basic uniform generator,
GGUBT, has been added to the Library for the user who would prefer an alternate multiplier.
The shuffled generator GGUW has been modified so the user may call it from any subroutine
in the chapter if it is desired to perform shuffling prior to generation of nonuniform variates.
A routine for generation of discrete uniform deviates, GGUD, as well as two routines,
GGDA and GGDT, for generation of variates from general discrete distributions has been
provided. One of the general routines uses a table lookup method and the other uses the
alias method. In addition to the new routines added, the efficiencies of the current routines
for generation of gamma, beta and multinomial variates have been substantially improved.
Interpolation; Approximation; Smoothing
Featured additions are easy-to-use companions to the existing cubic spline interpolation and
smoothing subroutines. The easy-to-use interpolatory spline routine, ICSCCU, achieves high
accuracy without requiring user-supplied end conditions, while the easy-to-use smoothing
spline routine uses statistical considerations to determine the degree of smoothing needed.
Additional one-dimensional approximation subroutines calculate a cubic spline interpolant
with periodic end conditions (ICSPLN) and a least squares approximation using user-supplied
basis functions (IFLSQ). Two-dimensional advances include a new code, IQHSCV, which fits
a smooth surface to data given at irregularly spaced points and modifications to all the bicubic
spline routines so that they use C. de Boor's "not-a-knot" boundary conditions rather than the
less accurate "natural" boundary conditions.
Linear Algebraic Equations
Two new subroutines have been added to Chapter L for Edition 8. LLBQF computes high
accuracy solutions to linear least squares problems. LGINF, a subroutine to compute the
generalized inverse of a matrix, has also been added.
Probability Density and Distribution Functions
Two new routines in Chapter M, MDGC and MDGCI, allow evaluation of a general
continuous distribution function or its inverse, using a table of values of the density function.
A new routine has been added to Chapter N, NDKER, for nonparametric estimation of the
density function using the kernel method.
Regression Analysis
Edition 8 allows two useful alternatives to least squares estimation in regression models. The
new routine RLLAV performs a least absolute values fit of a linear model, and a second new
routine RLLMV computes a minimum maximum deviation fit.
Zeros and Extrema; Linear Programming
A more robust nonlinear equation solver ZSCNT has been added which should be used
instead of ZSYSTM for all new applications. ZSYSTM will be deleted for Edition 9. A new
linear programming routine has been added which may eventually replace ZX3LP and
ZXOLP. This subroutine, ZX4LP is expected to handle large problems with greater reliability.
User comparisons between ZX3LP and ZX4LP are invited.
A complete listing of the 41 new routines follows.
NAME PURPOSE
BDLTV Produce letter value summary
BEMDP Median polish of a two-way table
CTLLF Log-linear fit of contingency table
CTPR Compute exact probabilities for contingency tables
DBLINT Numerical integration of a function of two variables
over a rectangular region
DTPTB Solve an ordinary differential equation system with
boundary conditions at two points, using a multiple
shooting method
EIGBS Find some eigenvalues and (optionally) eigenvectors
of a real symmetric band matrix
FFT3D Compute the fast Fourier transform of a complex
valued 1, 2, or 3 dimensional array
FLINV Compute the inverse Laplace transform of a user-supplied
complex function
GGBN Binomial random deviate generator
GGDA General discrete d;stribution random deviate
generator using alias method
GGDT General discrete distribution random deviate
generator using table lookup method
GGEXT Random deviate generator for a mixture of two
exponentials
GGMTN Multinomial random deviate generator
GGNO Generate set of order statistics from normal
distribution
GGNPP Nonhomogeneous Poisson process generator with
rate function LAMDA(T)-fixed interval, fixed
number or one-at-a-time
GGPER Generate a random permutation of the integers
1 toK
GGSRS Generate a simple random sample from a finite
population
GGSTA Stable distribution random deviate generator
GGUBT Uniform (0, 1 ) pseudo-random number generator
using alternate multiplier
GGUD Discrete uniform random number generator
GGUO Generate set of order statistics from uniform
(0,1) distribution
ICSCCU Cubic spline interpolation (easy-to-use version)
ICSPLN Cubic spline interpolation with periodic end
conditions
10
ICSSCV Cubic spline smoother (easy-to-use version)
IFLSQ Least squares approximation with user-supplied
basis functions
IQHSCV Smooth surface fitting with irregularly distributed
data points
LGINF Compute the generalized inverse of a real matrix
LLBQF Solution of linear least squares problem -
high accuracy solution
MDGC General continuous probability distribution
function, given ordinates of the density
MDGCI Inverse of a general continuous probability
distribution function, given ordinates of
the density
MMPSI Logarithmic derivative of the gamma function
NDEST Evaluate probability density function at
specified points
NDKER Nonparametric probability density function
(one dimensional) estimation by the kernel
method
OCDIS Compute pairwise Euclidean distances between
the columns of a matrix
RLLAV Perform linear regression using the least
absolute values criterion
RLLMV Perform linear regression using the
minimax criterion
USBOX Print a box plot (K samples)
USSLF Print a stem and leaf display
ZSCNT Solve a system of nonlinear equations
ZX4LP Solve the linear programming problem
via the revised simplex algorithm
(alternate easy-to-use version)
11
ACSL VERSION 6A
Version 6A of the Advanced Continuous Simulation Language (ACSL) for the CYBER has
been installed in FUTURE and will become the system default on September 1, 1980. Before
then, Version 6A can be accessed by entering:
FUTURE.ACSL
Please note that, as per the JUNE 1980 issue of OFF-LINE, the procedures ACSLCOM and
ACSLGO used to run ACSL will be CCL (CYBER Control Language) procedures; "CALL"
will no longer work. ACSLCOM, for instance, will be called by entering
ACSLCOM,INPUT=ft/e,OPTION=opf/ons...
and ACSLGO will be called by entering
ACSLGO,INPUT=ft/e,OUTPUT=ouf,PLOT=p/offe/-,...
The positional parameter feature of CCL can be used to a limited extent to shorten these
calls. For example,
ACSLCOM,MODEL,OPTION=T.
can be used to translate the model definition in file MODEL, with a full translator listing
(option T). Also,
ACSLGO.RUNCON.
could be used to run a translated model, using the run control commands in file RUNCON.
Procedure SAMPLES (accessed by GRAB,SAMPLES.) can be used to obtain sample model
definitions, particularly for the examples of Appendix A of the ACSL User's Guide.
Following is a summary of the changes and enhancements for Version 6A.
. Subroutines LINES and PAGE were added to allow page control from external
FORTRAN routines.
. The debug printout has been modified to identify system variables, state variables and
algebraic variables.
• The default maximum step size MAXT has been changed to 1.0E10 from the previous
value of 1.0.
• A discrete event capability has been added, invoked by setting the algorithm to zero
(IALG=0). A full description can be obtained from Stan Kerr. This algorithm
should be used in current simulations which now have a separate DERIVATIVE
section using IALG=6, the unsynchronized euler. Algorithm 6 will be eliminated
eventually.
12
. A start has been made to introduce a linear analysis capability. At the moment it will
find a steady state (if the model is reasonably linear), evaluate the Jacobean and
evaluate the eigenvalues and eigenvectors. See Stan Kerr for a full description.
. A subroutine, LOG, has been added for selective data logging. To force high
frequency data logging at every derivative evaluation, add to the end of the
DERIVATIVE section the statement
IF (condition) CALL LOG.
which will write the OUTPUT list and save the PREPAR list every time through,
while condition is .TRUE.
. The ACSL Newsletter in which Version 6A is described also contains several
suggestions, including:
A macro by which all FORTRAN intrinsic integer functions can be
automatically declared INTEGER to ACSL, rather than the default of REAL
which ACSL assumes for all variables and functions.
A macro which can be used to facilitate solution of difference equations with
algorithm 3 or 0.
A technique for measuring average step size and integration efficiency for the
variable step algorithms.
For further details and copies of the ACSL Newsletter, see Stan Kerr, Room 175 DCL (333-
4715).
CCL REMINDER
This is to remind you that the procedures listed below, which are currently KCL procedures
invoked by CALL or "-", will become CCL procedures on September 1, 1980, as announced
in the June 1980 issue of OFF-LINE. The CCL versions are currently accessed via FUTURE.
GRAB name
USERLIB (parameter XREF changing to NX)
ACSL
CONVPRC (will be MIMI after Sept. 1 )
TIDYPRC (will be TIDY after Sept. 1 )
ALTRAN
DEBLOCK
TBLOCK
Brief help information on these is available from the CYBER HELP program.
13
MISCELLANEOUS
CSO SHORT COURSES
CSO offers a number of introductory courses every semester. Topics usually include:
. How to use the CDC CYBER 175 and the ICE text editor.
. How to use the RNF text formatter.
. How to do plotting with NCAR and GCS.
. How to use magnetic tapes.
• How to use various statistical packages, including SPSS, SOUPAC and FOSOL.
. How to use the mathematical libraries and packages.
The offerings each semester vary considerably, depending on needs and demands.
An announcement of course offerings is sent to everyone on the OFF-LINE mailing list who
lives within 50 miles of the campus. An announcement is also made via the on-line HEAR YE
program and the RJE Bulletin posted at each remote job entry site. Our biggest problem has
always been that the introductory classes fill up too quickly, and further applicants are turned
away. We are trying to solve this problem by keeping a waiting list for each course. When
we have a sufficient number of applicants on the waiting list to fill another class, we will
attempt to open another session of the class whenever possible.
If you have any questions or complaints about the current short courses, please direct them
to Scott Lathrop, 187 DCL (333-6618).
MAILING LIST UPDATE
We are currently updating the on-campus section of the OFF-LINE mailing list and would
appreciate help from our user community. Off-campus and foreign subscribers do not need
to return the form to maintain their subscription. However, if these subscribers do move or
wish to terminate their subscription, we would appreciate word from them.
If you are an on-campus subscriber (or have a mailing address in Champaign or Urbana), and
would like to continue receiving a copy of the newsletter, please take a few minutes of your
time to check the special box labelled "Continue my subscription" on the mailing list form at
the end of this issue, add your current mailing address, and return the form to 1 64 DCL.
14
Also, departmental secretaries may be of great assistance by taking any copies of OFF-LINE
that are sent to members of their department who are no longer there, marking the copies
"LEFT' or "NO LONGER HERE", and returning them to 164 DCL.
An updated mailing list keeps down expenses, saves the Mailing Center many frustrations
and returns, and I'm sure, saves departments the frustration of receiving mail month after
month for someone who has left.
Since this issue may not reach some people due to summer activities, this notice will be
repeated in the September issue. However, beginning the first of October, we will delete all
names of persons who have not returned the form. Thank you for your cooperation in
helping us maintain an updated mailing list.
HELP WANTED
NEEDED - COMPUTER PROGRAMMER
1/4 research assistantship, August 21, 1980 - at least 1 year. Skills needed: FORTRAN
programming, knowledge of behavioral science statistics, ability to run SPSS programs.
Contact Barbara Tinsley, Dept. of Psychology (333-6371).
PROGRAMMER NEEDED
We are looking for a programmer for an 11-month, 1/2 time assistantship for the 1980-81
academic year. The position involves work in computerized movement control research in
the Physical Education Department, and provides an opportunity for programming on a
PDP- 11/03 and graphics display. Requires familiarity with FORTRAN and assembly
languages, and some experience with graphics work. Interested persons please contact Karl
M. Newell, Institute for Child Behavior and Development, 51 Gerty Drive (333-6563).
OFF-LINE's Mailing List
If you wish to be placed on our mailing list for future issues of OFF-LINE, if you wish to be
removed from the list, or if you wish to enter an address correction, please complete and
return this page. (Current subscribers are kept on the mailing list until a specific request for
removal is received, or until a mailing is returned as undeliverable.)
Check one:
□ New subscriber
□ Removal request
□ Address correction
□ Continue my subscription
Name:
Address:
CAMPUS or Zip Code
(If address correction, give old address and zip code below.)
Comments:
RETURN TO:
OFF-LINE
164 Digital Computer Laboratory
University of Illinois at Urbana-Champaign
Urbana, Illinois 61801
-CIC
Computing
Services
Office
Off-Line
University of Illinois at Urbana-Champaign
VOL. 8, NO. 9 September 1980
EDITOR: Lynn Bilger
PHONE: (217) 333-6236
139 Astronomy Building
Urbana, Illinois 61801
Page
Contents
f^-
°Bfc*ffy
O^TVf
&»
'*'*
;rJi 1980
Of-
'.^(VOi
'">'%&
m
2
3
4
4
4
5
6
9
11
11
12
SYSTEM NOTES
July Reliability Report
CYBER
CSO Videotapes
CSO Short Courses for the CYBER 175
IBM
IBM PUBLIC Facility Gone
MATHEMATICAL SERVICES
HELP Information Dumped from Procs
APEX Snafu
T.O.M.S. Algorithm Tape
New Routines in UOILIB
Permanent File Subroutines
IMSL Edition 8 Manual - Errata
MISCELLANEOUS
Computer Related Discounts
Available Through Purchasing Division
Address Change
Mailing List Update
HELP WANTED
IBM Selectric I/O Writer
CSO DIRECTORY - STAFF AND SERVICES
Administrative
Director
Business Manager
Secretary
User Services
User Accounting
Systems Consulting
Statistical Services Consulting
Text Processing Consulting
Asst Dir User Services
Mgr, Statistical Services
Mgr, UNIX Operations
Documentation
Distribution Center
Hardware/Software Support
Terminal Repair Service
CYBER Dial-up Numbers
Asst Dir Systems and Operations
Asst Dir Development
Asst Dir Engineering
CYBER-IBM Operations
UNIX Operations
Telecommunications
Hardware Selection Assistance
Laboratory Support Project
RJE Operations
RJE Sites
Agriculture
Chemistry
Commerce West
DCL Routing Room
Electrical Engineering
Florida Ave Res Hall
Illinois St Res Hall
Mechanical Engineering
Psychology
Social Science
George Badger
150
DCL
333-4103
Stanley Rankin
150
DCL
333-6530
Joyce Vaughn
150
DCL
333-1637
Gary Bouck
1208
W Springfield
333-7752
166
DCL
333-6133
65
Comm West
333-2170
207
Astronomy
333-7318
Robert Penka
173
DCL
333-4709
Larry Sautter
91
Comm West
333-2170
Larry Sautter
Lynn Bilger
139
Astronomy
W Springfield
333-6236
1208
333-6760
150
DCL
333-0969
110
baud
333-4000
300
baud
333-4000
1200
baud
333-4001
Sandra Moy
177
DCL
333-4703
J. M. Randal
120
DCL
333-9772
Cliff Carter
195
DCL
333-3723
Jack Knott
194a
DCL
333-6562
Debbie Weller
203
Astronomy
333-8150
Tom Kerkering
16
DCL
333-0816
Cliff Carter
195
DCL
333-3723
Lee Hollaar
193
DCL
333-7904
Rex Duzan
162
DCL
333-6285
M103
Turner Hall
333-8170
153
Noyes Lab
333-1728
70
Comm West
333-4500
129
DCL
333-6203
146
EEB
333-4936
FAR
333-2695
ISR
333-0307
65
MEB
333-2072
453
Psych Bldg.
333-7815
202
Lincoln Hall
333-0309
OFF-LINE is the monthly newsletter of the Computing Services Office at the
University of Illinois at Urbana-Champaign. Unless otherwise indicated, permis-
sion to reprint is freely granted, provided that the author, if named, and the Com-
puting Services Office (CSO) are credited. Information in this issue is current as
of August 29, 1980.
CSO operates an IBM 4341 with four million bytes of fast core running HASP-
OS/MVT under VM, a CYBER 175 with 256K words of central memory and 512K
words of ECS, under NOS, serving up to 190 simultaneously active text and
graphics terminals, and a DEC PDP-11/50 with 252K bytes of core driving a GSI
CAT-8 phototypesetter with the UNIX Operating System.
SYSTEM NOTES
JULY RELIABILITY REPORT
CYBER 175: 17 recoverable interruptions
18 non-recoverable interruptions
18 of these were for more than 15 minutes
Mean Time Between Failures = 16 hours
Mean Time to Repair = 46 minutes
Availability: 94.3% of scheduled uptime
Major cause of downtime was related to
disk pack problems and power outage due
to storm.
IBM 360/75: 54 non-recoverable interruptions
50 of these were for more than 15 minutes
Mean Time Between Failures = 1 3 hours
Mean Time to Repair = 90 minutes
Availability = 91.3% of scheduled uptime
Major cause of downtime was related to
memory failures and power outage due to
storm.
IBM 4341: 15 non-recoverable interruptions
15 of these were for more than 15 minutes
Mean Time Between Failures = 7.5 hours
Mean Time to Repair = 83 minutes
Availability = 83.7% of scheduled uptime
Major cause of downtime was related to
installation of disk drives.
(NOTE: The IBM 4341 replaced the IBM 360
on July 27, 1980.)
"Recoverable interruptions" in time-sharing are those interruptions from which a person may
normally recover work being done at the time of the interruption by doing a RECOVER, tty
when the system comes back up. "Non-recoverable interruptions" means that a major
problem has occurred, and recovery of work in progress at the time of the interruption is not
possible.
CYBER
CSO VIDEOTAPES
This article is repeated from the August issue.
CSO has recently completed the initial stages in the preparation of videotapes describing the
CYBER 175. The following three tapes are available for viewing.
CSOVT1 Introduction to Computing at CSO. This videotape is intended for the
first time computer user, and for people wishing to learn more about
CSO's facilities.
CSOVT2 This videotape consists of four sections:
. Using a Terminal - How to operate a computer terminal.
. Introduction to CYBER Time-Sharing - How to access the CYBER
using a terminal.
. File Usage - Local files and indirect access permanent files.
• Introduction to ICE Text Editing - The basics of time-sharing
editing.
CSOVT3 This videotape, Running a FORTRAN Program on the CYBER, consists
of three sections.
. Concepts involved
. PROGRAM statement
. CYBER control statements
Anyone may view these tapes by going to the Undergraduate Library in person to make a
reservation to use the videotape equipment. The library hours are 8 AM - 8 PM Monday
through Thursday, 8 AM - 5 PM Friday, and 1 PM - 5 PM Saturday and Sunday.
When you check out the videotapes, you should also pick up one of the handouts that are
available with the tapes. The handout summarizes the contents of each tape in more detail,
and facilitates note taking.
Your opinions regarding these videotapes would be greatly appreciated. You may use the last
sheet of the handout, or write directly to:
CSO VIDEOTAPES
150 DCL
CSO SHORT COURSES FOR THE CYBER 175
CSO is once again offering short courses during the fall semester 1980 to acquaint people with
our facilities and the Control Data Corporation (CDC) CYBER 175 computer system.
To register for a course:
Come in person to Room 150 DCL, or
Phone 333-6630.
Registration is free and limited to 30 persons per class. If you find that all of the available
classes on a topic are full, please leave your name on our waiting list. We will call you if
someone drops a course or we open another session.
The following courses are being offered (check the Short Course mailout or call 333-6630 for
further details - time and place).
GENERAL COURSES
Introduction to the CYBER 175
MANAGE
CYBER Control Language (CCL)
CYBER Magnetic Tapes
Math Libraries
Differential Equations
Curve Fitting
Non-Linear Programming and NETWORK
RNF TEXT PROCESSING COURSES
Introduction to the CYBER for RNF Users
Introduction to the RNF Text Formatter
Advanced Usage of RNF
GRAPHICS COURSES
Graphics at CSO
ZETA Plotters
NCAR Plot Package
TEKTRONIX 4027 Color Terminal Usage
Introduction to GCS Plot Package
Advanced usage of GCS
STATISTICAL COURSES
FOSOL
SAS
SOUPAC
SPSS
IBM
IBM PUBLIC FACILITY GONE
The IBM "PUBLIC" facility, which allowed public access to user-contributed software, is no
longer in existence. It has not been added to in the last several years, and it was recently
noted that the supporting dataset has been idle for so long that it has disappeared completely.
Since we feel that this is indicative of the demand for the facility, we do not plan to revive it.
If there should be a strong demand for revival, we will revive it, but it may not operate in the
same way. Please contact Stan Kerr (333-4715) with comments or suggestions.
MATHEMATICAL SERVICES
HELP INFORMATION DUMPED FROM PROCS
The HELP information available from procedures MATH, SAMPLES, and DYNAMO via the
statement "-HELP, prod' has been added to the system HELP files. This information will be
removed from the procedures on October 1, 1980. After that date, enter the system
command "HELP followed by the name of the proc you wish help information for.
APEX SNAFU
We have discovered that the APEX manuals currently supplied by CDC are Revision G, for
Version 1.2 of APEX. However, we are running Version 1.1 of APEX. Therefore, be wary
of features described in the manual on pages marked with a G at the bottom, and features
marked with vertical change bars. In particular, one feature which is not available in Version
1 . 1 is the dual simplex algorithm invoked by the DUAL verb.
T.O.M.S. ALGORITHM TAPE
We now have a tape containing all algorithms published in Transactions on Mathematical
Software to the present date (the first is algorithm 493, published in Volume 1, 1975). Users
can access this tape by using the directory of files supplied in file
TOMS/UN = MATHUB.
All algorithms are supplied in their original form and may require conversion to run on the
system here; some are subroutines, some are main programs.
For example, from the directory you would find that algorithm 498, published in Volume 1,
Number 4, for calculation of Airy functions is in tape file 6 and contains 348 cards. You
could read and save this as follows:
GRAB.DEBLOCK.
LABEL,TAPE,VSN=TOMSN140,D=PE,LB=KU.F=S,CV=EB,PO=R.
SKIPF,TAPE,5.
DEBLOCK,DISK=ALG498.
SAVE.ALG498.
RETURN/TAPE.
Please note that the Systems Consulting Office (Room 166 DCL) has notebooks with all the
Collected Algorithms of the ACM (from 1 on up).
NEW ROUTINES IN UOILIB
Three new routines, NSPIV, RMFFT and CMFFT, have been added to UOILIB.
NSPIV
NSPIV is a routine for sparse Gaussian elimination, published as algorithm 533 in
Transactions on Mathematical Software, Volume 4, Number 4 (December 1978). With this
routine, a sparse matrix can be passed as a vector of non-zero values together with arrays of
pointer indices that tell where the non-zeros are positioned in the matrix. Because Gaussian
elimination may generate additional non-zero matrix elements (at each stage of the
reduction), the routine must be supplied with enough work storage to allow for this growth;
in some cases, it may be difficult to tell how much growth will occur, so the routine should be
applied only after this aspect has been analyzed.
To print a brief description of the parameters of NSPIV, enter the following:
GRAB.MATH.
MATH.UOILIB.NSPIV.
PRINT.DOC.
RMFFT and CMFFT
RMFFT and CMFFT are routines for computing the fast Fourier transform (FFT) of a large
real (RMFFT) or a complex (CMFFT) sequence or multi-dimensional array stored on a
random-access disk file. These codes were published as algorithm 545 in Transactions on
Mathematical Software, Volume 5, Number 4 (December 1979). With them, a very large
sequence or array can be broken down into small portions, stored on a CYBER FORTRAN
random-access file, and transformed.
To print descriptions of RMFFT and CMFFT, enter the following:
GRAB.MATH.
MATH.UOILIB.RMFFT.
MATH.UOILIB.CMFFT.
PRINT.DOC.
The original version of the algorithm supplied to us includes random access drivers for other
machines, including PDP-lFs. If you are interested in these, please contact Stan Kerr, Room
175 DCL (333-4715).
PERMANENT FILE SUBROUTINES
Permanent and local file subroutines for use from FORTRAN on the CYBER have been
installed in UOILIB. There is a routine, PF, for doing permanent file operations, and a
routine, LF, for local file operations (currently LF only does a RETURN).
LF is called as follows:
CALL LFf RETURN" ."/fn/V'/fn^ "Ifnrf)
Where each local file name can be given as a character constant, e.g., "FILE", or as the
FORTRAN unit number for a file specified in the PROGRAM statement, e.g., CALL
LF("RETURN',7).
PF is called as follows:
CALL PF(request.lfn,pfn,keyl.optl.key2.opt2, .keyn.optn)
Where:
request is the desired permanent file request ("APPEND", "ATTACH",
"CHANGE", "DEFINE", "GET", "PERMIT', "PURGE",
"REPLACE", or "SAVE"). The "PERMIT' request requires
special format -- see Note 3 below.
Ifn is the local file name, given in character format as for LF, or as a
FORTRAN unit number.
Pfn is the permanent file name given in character format. Only one
permanent file can be accessed per call to PF.
key and opt are pairs of keywords and options, in character format, subject to
the restrictions mentioned below.
For example:
CALL PFf,GEr,,l,"DATA":,UN,,:,USERNa,,"PW ."STRING")
CALL PFf ATTACH" ," TAPE2" ," BIG" ," M" ," N" )
CALLPFf,PURGE",5LTAPE3,"SUPERC',"NA")
CALLPF0,DEFINE\"TAPE3","ULTRAC',,,CT',"S")
The options are parellel to those provided by NOS (see the NOS Manual, Volume 1, Section
8 or the Time-Sharing User's Reference Manual) with the following additions:
"NONE" used to nullify specific keywords
"RC" returns error code in integer format
"RRC" returns error code in real format
"NA" inhibits rollout if a direct access file is busy
"SS" retains subsystem mode of saved file
"UC" retains the user control word
The following examples illustrate different ways of performing the same permanent file
functions, viz., obtain a permanent file "PERM' as a local file named "TAPE10" (assume that
TAPE10 is in your PROGRAM statement and is not equated to something else).
CALL PFfGETV'TAPEiaV'PERM")
CALL PF(3LGET,6LTAPE10,4LPERM)
CALL PFf'GETMO."PERM")
If the same name is desired for both the permanent and local files, then one of the
parameters for Ifn and pfn may be zero. The following examples all get a permanent file
named "NICKLE" as a local file of the same name.
CALL PFfGET',"NICKLE","NICKLE")
CALL PF("GET',0,"NICKLE")
CALL PFrGET',"NICKLE",0)
Please pay special heed to the following notes:
1. Make sure that local files are declared in the PROGRAM statement. If they are not,
you will be able to GET, etc., but FORTRAN cannot do I/O with the file.
2. Before calling PF, ensure that BUFFERS ARE FLUSHED! The FORTRAN
statements REWIND and ENDFILE can be called to do this (but note that either of
these adds an /EOP to a text file).
The "PERMIT' request requires a special format such that the user number (only
2one may be specified) and permission mode must include the "UN" and "m"
keywords, e.g.,
CALLPFrPERMir,"MYFILE","UN,,:,YOURS":,M","R")
ERROR PROCESSING -- The combination of the "RCV'RRC and "NA" options
select a number of responses from PF:
. If option "NA", "RC", or "RRC" is not specified, and the function fails, the
standard NOS error message is written to your dayfile and your program is
aborted. To specify NA, include
" NK', value
as a keyword/option pair in the call to PF. The value is immaterial, e.g.,
"NA",0 will do fine.
. If either "RC or "RRC is specified and the function fails, the NOS numeric
error code is placed in the return code parameter. A zero value indicates a
successful operation.
For instance, the keyword/ option pair
"RC'JRC
in your call to PF indicates you want the value of the error code, as an integer,
placed in variable IRC. The pair
"RRC'.X
in your call to PF indicates you want the value of the error code, as a real
number, placed in variable X.
. If "NA" is not specified, a call to PF for a direct access file which is currently
attached in write mode by someone else will cause your program to "roll out"
and wait until the file becomes available.
. If "NA" is specified and the function fails, your program continues, and an
error code is returned if you specified "RC or "RRC as above.
If you use a FORTRAN unit number for the local file, and that unit is equated to a
file in your PROGRAM statement, it is the name of this latter file which is used as the
local file name. If, further, this file is overridden on the "LGO" card, the name of the
overriding file is used. For example, suppose your PROGRAM statement is
PROGRAM FU(INPUT,OUTPUT,CAT,TAPEl=CAT)
and you execute using
LGO,DATA,OUT,DOG.
and, if your program contains
CALL PFCGET",l,"COW')
then, permanent file COW will become the local file DOG, and references to
FORTRAN unit 1 will use local file DOG and not local file CAT.
IMSL EDITION 8 MANUAL - ERRATA
We have discovered an error in the IMSL Version 8 Manual. Under the routine USSLF,
there is a parameter to the routine called IUNIT. The writeup in the manual claims that this
parameter is never changed by USSLF, but in fact, it can be. If you have questions about
this, please contact Stan Kerr (333-4715).
MISCELLANEOUS
COMPUTER RELATED DISCOUNTS
AVAILABLE THROUGH PURCHASING DIVISION
The following computer related discounts are available through the Urbana campus
Purchasing Division. These discounts are valid only during the Fiscal Year 1981.
CRT TERMINALS
Vendor: Kal-Tronics Corp., 3677 Woodhead Dr., Northbrook, IL 60062
LierSiegler ADM3A U/L Case $695.00
Interactive Display
Retro Graphics RG512 $860.00
(Graphics capability for ADM3A)
Vendor: Bronson & Bratton,Inc, 5161 S. Millard Ave, Chicago, IL 60632
Infoton GT101 $850.00
Visual 210 $890.00
10
Vendor: IBM, Data Proc Div, 1133 Winchester, White Plains, NY 10604
IBM Model 3101
(Good until 2/28/81)
$1046.36
PRINTING TERMINALS
Vendor: Hall-Mark Electronics, 13789 Rider Trail, Earth City, MO 63045
Texas Instrument
743
743
745
745
785
787
810
810
820
820
U Case only
$863.00
U/L Case
$935.00
U Case only,
portable
$1218.00
U/L Case, portable
$1290.00
$1899.00
$2240.00
$1380.00
with package
$1568.00
$1570.00
with package
$1737.00
Vendor: David Jamison Carlyle Co., 4200 Marine Dr., Chicago, IL 60613
DecWriter
Teletype
LA34DA
LA34AA
LAI 20
4320AAA
4320AAK
$895.00
$1084.00
$2074.00
$940.00
$1015.00
Vendor: Xerox Data Sys., 450 W. Algonquin Rd., Arlington Heights, IL
Diablo-Xerox
1640
1640
1650
1650
Receive only
Keyboard
Receive only
Keyboard
$2306.00
$2640.00
$2389.00
$2720.00
Vendor: Information Systems Inc., 9806 W. Farragut Ave., Rosemont, IL 60018
Spinwriters - NEC
5510-1
5520-1
5515-1
5525-1
5530-1
5540-1
$2361.00
$2433.00
$2693.00
$2782.00
$2362.00
$2811.00
11
COMPUTERS
Micro-mini, Apple II $1902.48
Sample Configuration
1 Apple II with: Integer and Extended Basic, 32K RAM,
Necessary cables and manuals
1 RS232 Serial Interface Card
1 Disk II drive and interface
1 12" B&W monitor
ACOUSTIC COUPLERS (300 BAUD)
Vendor: ComData Corp., 8115 N. Monticello Ave., Skokie, IL 60076
302A2-13 General Purpose $176.90
150A2-14B DecWriter, includes cable $128.00
150A2-14C TTY 43, includes cable $128.90
ADDRESS CHANGE
Due to the various moves that are continuing to be made at DCL, the mailing address for all
documentation (comments on manuals, newsletter, etc.) has been changed from 164 DCL to
150 DCL. This address change will be made in our documentation as manuals, etc., are
updated.
Also, for those users returning to campus after a summer off, the Systems Consulting Office
has been moved to Room 166 DCL, the Accounting Office and Distribution Center have both
been moved to 1208 W. Springfield.
MAILING LIST UPDATE
This article is reprinted from the August issue. I would also like to take this
opportunity to thank our users for the tremendous response I have received to this
request. - Lynn Bilger, Editor.
We are currently updating the on-campus section of the OFF-LINE mailing list and would
appreciate help from our user community. Off-campus and foreign subscribers do not need
to return the form to maintain their subscription. However, if these subscribers do move or
wish to terminate their subscription, we would appreciate word from them.
12
If you are an on-campus subscriber (or have a mailing address in Champaign or Urbana)
and would like to continue receiving a copy of the newsletter, please take a few minutes of
your time to check the special box labelled "Continue my subscription" on the mailing list
form at the end of this issue, anadd your current mailing address, and return the form to 150
DCL.
Also, departmental secretaries may be of great assistance by taking any copies of OFF-LINE
that are sent to members of their department who are no longer there, marking the copies
"LEFT' or "NO LONGER HERE", and returning them to 150 DCL.
An updated mailing list keeps down expenses, saves the Mailing Center many frustrations
and returns, and I'm sure, svaes departments the frustration of receiving mail month after
month for someone who has left.
Beginning the first of October, we will delete all names of persons who have not returned
the form. Thank you for your cooperation in helping us maintain an updated mailing list.
HELP WANTED
IBM SELECTRIC I/O WRITER
One IBM Selectric I/O Writer is available locally. If interested, contact Cliff Carter, 333-3723.
OFF-LINE's Mailing List
If you wish to be placed on our mailing list for future issues of OFF-LINE, if you wish to be
removed from the list, or if you wish to enter an address correction, please complete and
return this page. (Current subscribers are kept on the mailing list until a specific request for
removal is received, or until a mailing is returned as undeliverable.)
Check one: □ New subscriber
□ Removal request
□ Address correction
□ Continue my subscription
Name:
Address:
CAMPUS or Zip Code
(If address correction, give old address and zip code below.)
Comments:
RETURN TO: OFF-LINE
150 Digital Computer Laboratory
University of Illinois at Urbana-Champaign
Urbana, Illinois 61801
: I L L A •
Computing
Services
Office
Off-Line
University of Illinois at Urbana-Champaign
VOL. 8, NO. 10 October 1980
EDITOR: Lynn Bilger
PHONE: (217) 333-6236
139 Astronomy Building
Urbana, Illinois 61801
Page
Contents
5
6
8
9
9
10
10
12
SYSTEM NOTES
August Reliability Report
CYBER
PASCAL Bugs Fixed
IBM
New Versus Old
MATHEMATICAL SERVICES
SLAM - Updates and New Features
The PORT Library
Harwell Library on the IBM 4341
MISCELLANEOUS
Research Board Deadline
for Departmental Allocation Requests
CSO Personnel Honored
HELP WANTED
Position Available
SPECIAL EDITION ARTICLES AND INDEX
On-line Storage
Magnetic Tapes From Scratch
(THEUBRArfYOFlHfi
nov i ' mo
W^mr 0F,LUW<S
15 CSO Mailbox Facility
18 USE Command Available From Time-Sharing
19 Text Processing Services at CSO
22 Service Keypunching at CSO
22 Catalog Overflow - Size
24 Security Safeguards
25 NCAR Graphics Software
28 Diablo Reservation Service to Change
29 Keypunch Service
30 Rate Changes for the IBM 4341
31 CSO Videotapes
32 Index
CSO DIRECTORY - STAFF AND SERVICES
Administrative
Director
Business Manager
Secretary
User Services
User Accounting
Systems Consulting
Statistical Services Consulting
Text Processing Consulting
Asst Dir User Services
Mgr, Statistical Services
Mgr, UNIX Operations
Documentation
Distribution Center
Hardware/Software Support
Terminal Repair Service
CYBER Dial-up Numbers
Asst Dir Systems and Operations
Asst Dir Development
Asst Dir Engineering
CYBER-1BM Operations
UNIX Operations
Telecommunications
Hardware Selection Assistance
Laboratory Support Project
RJE Operations
Don McCabe
RJE Sites
Agriculture
Chemistry
Commerce West
DCL Routing Room
Electrical Engineering
Florida Ave Res Hall
Illinois St Res Hall
Mechanical Engineering
Psychology
CRH Snack Bar
Social Science
George Badger
150
DCL
333-4103
Stanley Rankin
150
DCL
333-6530
Joyce McCabe
150
DCL
333-1637
Gary Bouck
1208
W Springfield
333-7752
166
DCL
333-6133
65
Coram West
333-2170
207
Astronomy
333-7318
Robert Penka
173
DCL
333-4709
Larry Sautter
91
Comm West
333-2170
Larry Sautter
Lynn Bilger
139
Astronomy
W Springfield
333-6236
1208
333-6760
150
DCL
333-0969
no
baud
333-4000
300
baud
333-4000
1200
baud
333-4001
Sandra Moy
177
DCL
333-4703
J. M. Randal
120
DCL
333-9772
Cliff Carter
195
DCL
333-3723
Jack Knott
194a
DCL
333-6562
Debbie Weller
203
Astronomy
333-8150
Tom Kerkering
16
DCL
333-0816
Cliff Carter
195
DCL
333-3723
Lee Hollaar
193
DCL
333-7904
Rex Duzan
162
DCL
333-6285
1528 W Springfield
333-7720
M103
Turner Hall
333-8170
153
Noyes Lab
333-1728
70
Comm West
333-4500
129
DCL
333-6203
146
EEB
333-4936
FAR
333-2695
ISR
333-0307
65
MEB
333-1430
453
Psych Bldg.
333-7531
120
Snack Bar
333-1851
202
Lincoln Hall
333-0309
OFF-LINE is the monthly newsletter of the Computing Services Office at the
University of Illinois at Urbana-Champaign. Unless otherwise indicated, permis-
sion to reprint is freely granted, provided that the author, if named, and the Com-
puting Services Office (CSO) are credited. Information in this issue is current as
of September 25, 1980.
CSO operates an IBM 4341 with four million bytes of fast core running HASP-
OS/MVT under VM, a CYBER 175 with 256K words of central memory and 512K
words of ECS, under NOS, serving up to 190 simultaneously active text and
graphics terminals, and a DEC PDP-11/50 with 252K. bytes of core driving a GSI
CAT-8 phototypesetter with the UNIX Operating System.
SYSTEM NOTES
AUGUST RELIABILITY REPORT
CYBER 175: 8 recoverable interruptions
15 non-recoverable interruptions
1 5 of these were for more than 1 5 minutes
Mean Time Between Failures = 24 hours
Mean Time to Repair = 57 minutes
Availability: 95.9% of scheduled uptime
Major cause of downtime was related to
disk problems.
IBM 4341: 57 interruptions
30 of these were for more than 1 5 minutes
Mean Time Between Failures = 12 hours
Mean Time to Repair = 44 minutes
Availability: 85.9% of scheduled uptime
Major cause of downtime was related to
installation.
(NOTE: The IBM 4341 replaced the IBM 360
on July 27, 1980.)
"Recoverable interruptions" in time-sharing are those interruptions from which a person may
normally recover work being done at the time of the interruption by doing a RECOVER, tty
when the system comes back up. "Non-recoverable interruptions" means that a major
problem has occurred, and recovery of work in progress at the time of the interruption is not
possible.
CYBER
PASCAL BUGS FIXED
The PASCAL compiler has had a number of bugs fixed, including the following which were
brought to our attention by the users:
• Correct indexing of packed arrays with eight elements per word (i.e., 7 bit entries).
. Restrict NEW to pointer variables.
Prevent explicit assignment to FOR-DO loop control variables and nesting of
FOR-DO loops with the same control variable.
. Prevent the compiler from infinite looping when trying to compile incorrect
declarations (e.g., set of 1..N; N, a variable) or set constructions (e.g., [1..255],
[1..3.5], etc) with max>58 (Integer).
. Correct runtime Division by Zero Error message.
• Insure pointer is valid before disposing node in PSYSTEM.
To date, 40 bugs have been corrected. This version of PASCAL (S.2) is in:
GRAB.PASCAL/F.
Present binaries will not work with this new version.
IBM
NEW VERSUS OLD
We though you would enjoy seeing the results of nearly a decade and a half of progress in
building computers. The following pictures graphically demonstrate the continuing trend of
miniturization of electronic components. The "before" pictures are of the IBM 360/75 which
was installed in the late sixties. The "after" pictures are of the IBM 4341 which was installed
in late July 1980. The two machines have nearly identical computing power, but the 4341 has
more memory and channels.
BEFORE The monolithic unit (built
of many cabinets) which seems to run
the length of the room forms the central
processing unit, one megabyte of
memory, and the operator console of
the 360/75. An entire cabinet (approxi-
mately six feet long) is required to con-
tain a quarter megabyte of memory!
AFTER This is the 4341 -allot it! It
replaces the monolithic unit in the previ-
ous picture plus a number of pysically
separate boxes some of which cannot
easily be seen because of the crowded
condition of the room.
AFTER No bigger than a breadbox,
but equal of the 360/75! These circuit
boards are the heart of the 4341 and
contain 4 megabytes of memory and the
central processing unit.
BEFORE The control panel of the
360/75. It contains approximately 1300
lights and more than 100 switches and
dials.
AFTER The "control panel" of the 4341. Actually,
the engineers have access to a few other switches
inside the cabinet. The CRT console seen in the
picture of the 4341 together with an internal diag-
nostic computer serve the same function as the
switches and lights on the console of the 360/75. In
addition, the CRT displays are formatted for human
interpretation and are much easier to read than the
lights on the 360/75 console.
MATHEMATICAL SERVICES
SLAM - UPDATES AND NEW FEATURES
Updates and enhancements for the SLAM simulation package have been received and
installed. Enhancements include the following:
• It is now possible to request a summary report on selected runs. The options available
for the ISMRY field on the GEN card are:
N - No summary reports
Y - summary report after every run (default)
Y/E - summary report after every run (same as the Y option; default)
Y/F - summary report after first run only
Y/S - summary report after the first and last runs
Y/n - summary report after every nth run
. It is now possible to clear statistics based on observations selectively
between runs. Two optional fields have been added following the JJCLR
field on the INITIALIZE card, giving the card the form:
INIT,TTBEG,TTFIN,JJCLR/NCCLR/JCNET,JJVAR,JJFIL;
Statistics to be cleared are defined as follows:
Variable Options
JJCLR "Y" - clear user-collected
statistics with STAT
code less than NCCLR
Default
"N" - clear user-collected
statistics with STAT
code greater than or
equal to NCCLR
NCCLR integer equal to breaking
point for clearing of user-
collected statistics
largest statistics
code defined + 1
JCNET "Y" - clear network-collected
statistics
JJCLR
"N" - do not clear network-
collected statistics
. Two new functions have been added:
FFAWT(IFILE) will return the average waiting
time in file IFILE
NNBLK(IACT,IF1LE) will return the number of entities
associated with activity number
I ACT and blocked by the node
associated with file IFILE
If you have problems or questions about SLAM, contact Stan Kerr (Room 175 DCL, 333-
4715 or through TELL,UN=MATHLIB on the CYBER).
THE PORT LIBRARY
CSO has installed the Bell Labs PORT mathematical library on the CYBER. It can be
accessed by entering:
GRAB.PORT.
Following this you can compile and run a FORTRAN program which calls PORT routines.
There is a manual for PORT available for inspection in the Systems Consulting Office, Room
166 DCL. Manuals can be purchased for $20.00 from Bell Laboratories Computing Informa-
tion Service, 600 Mountain Avenue, Murray Hill, New Jersey (07974). A copy of Bell Labs
Computing Science Technical Report #47, which has an overview and rationale for the
library, is also available for inspection.
The purpose of the library is to provide the first "portable" library of general mathematical
routines, written in a FORTRAN which is acceptable to all present computer systems. Only
three routines in the library must be adjusted for a given computer system: R1MACH,
D1MACH and I1MACH which define real, double precision and integer machine constants,
respectively. This is in contrast to a library such as IMSL, which is maintained in slightly
differing versions for different models of computers.
PORT is divided into functional areas as follows:
1 . Approximation, Interpolation and Extrapolation
This includes routines for approximating discrete and continuous functions. For the
former, there are codes for best uniform approximation by rational functions, splines
(both cubic and the more general B-splines), and least squares approximations by
splines. For the latter, there are least squares approximations using B-splines.
2. Computer Arithmetic
This includes routines for performing complex double precision arithmetic, using a 2-
element vector of double precision numbers to represent a complex number.
3. Differential Equations
Currently, this only has a variable step algorithm for non-stiff (no sharp transient) sys-
tems using a modified midpoint rule with extrapolation.
4. Linear Algebra and Eigensystems
Currently, this only has general algorithms for solving real and complex linear sys-
tems, or obtaining least-squares solutions, and for finding eigenvectors and eigen-
values of a real matrix.
5. Mathematical Programming
This section is empty in the current PORT.
6. Optimization
Currently, this section has only one very simple routine. Others will be added later.
7. Probability and Statistics
Currently, this section has only a uniform random number generator.
8. Quadrature and Differentiation
This section presently includes several routines for adaptive quadrature, cubic spline
quadrature or differentiation, and Gaussian quadrature.
9. Roots
Currently, this includes routines for solving real or complex polynomials and for solv-
ing a set of real simultaneous nonlinear equations, with or without derivatives
required.
10. Special Functions
This includes Bessel functions I and J, of complex argument and integer order, and
several standard trigonometric functions.
1 1 . Transforms
This includes several Fast Fourier transform routines, for real and complex data.
12. Utility Routines
This includes routines for:
. centralized error handling
. dynamic storage allocation
. specification of machine-dependent quantities
. operations on one-dimensional arrays (including sorting)
. evaluation of the ceiling and floor of a number
Errors in PORT routines will be reported to Bell Labs; they will not be fixed by CSO. Ques-
tions about PORT should be directed to Stan Kerr (Room 175 DCL, 333-4715 or through
TELL,UN=MATHLIB on the CYBER).
HARWELL LIBRARY ON THE IBM 4341
Since we expect to continue IBM service for some time, we have decided to install the bulk
of the Harwell subroutine library on the IBM 4341. This library consists of several hundred
subroutines, developed at the Atomic Energy Research Establishment in Harwell, England. It
is organized by lettered chapters as follows (Note: only starred chapters are available at
present -- see Stan Kerr for others) :
D* - differential equations
E - eigenvales and eigenvectors of matrices
F* - mathematical functions (also random numbers and Fourier transforms)
G - geometrical problems
I - integer valued functions
K - sorting
L* - linear programming
M* - linear algebra (includes good sparse matrix codes)
N* - nonlinear equations
O - input/output aids
P - polynomial and rational functions
Q* - numerical integration
S - statistics
T* - interpolation and approximation
V* - optimization and nonlinear data fitting
Z - non-FORTRAN system facilities
The library is available automatically in all standard FORTRAN and PL/I procedures on the
IBM 4341.
Bugs in Harwell routines will be reported to Harwell; they will not be fixed by CSO.
A manual is available for inspection at the Systems Consulting Office, Room 166 DCL. At
present, this is the only copy. If you have questions about the Harwell subroutine library,
contact Stan Kerr (Room 175 DCL, 333-4715).
MISCELLANEOUS
RESEARCH BOARD DEADLINE
FOR DEPARTMENTAL ALLOCATION REQUESTS
The Research Board has established a November 3, 1980 deadline for submission of depart-
mental requests for research computer allocations. This deadline affects allocations for the
period December 24, 1980 through June 26, 1981.
Research Board allocations are expected to support faculty research, thesis and dissertation
research. Departmental requests and the allocations they subsequently receive are based on
individual user requests. Those persons who will need research computer time for this alloca-
tion period should be sure to submit their requests to their department via the Research
Board form A. These forms and further instructions are available from the University
departments.
CSO PERSONNEL HONORED
Several of CSO's non-academic personnel were recently honored for long service to the
University.
Tom Kerkering, who is responsible for data communications, received a 30-year award. Both
Ed Pelg and Estil Carter received 25-year awards, and Gary Bouck received a 15-year award.
We appreciate the continuing efforts of these people, and recognize the value of their accu-
mulated experience.
10
HELP WANTED
POSITION AVAILABLE
Manager of Data Processing for new state agency: Illinois Health Finance Authority. The job
is to design and maintain a data system to collect hospital cost data and to help determine
rates that hospitals can charge.
Salary: S25.OO0 - S30.0O0
Contact: Michael Koetting. Director. 312-793-5940
Location: Chicago
SPECIAL EDITION ARTICLES AND INDEX
ON-LINE STORAGE
Reprinted from OFF- USE. September 1979
The CYBER on-line storage now consists of two levels, disk and a mass storage unit. The
disk portion of the system is presently about two billion characters (three million PRU'sJ and
will be upgraded during the next 12 months to about six billion characters. Mass storage
holds about twenty billion characters and may be upgraded to forty billion.
Over the past year. CSO has developed an archive system based on disk and tape. Files
appear in the user catalog when they are at either level of this hierarchy, and movement back
and forth will become rapid and convenient as the mass storage unit replaces tape.
This archive system was developed as a prototype of the new hierarchy, and we are now test-
ing the mass storage unit in place of the magnetic tape. It is possible that other uses of the
mass store will be available later, e.g.. user-controlled transfers. The processes and behavior
of the archive system have been carefully observed since the beginning so that reasonable
policies and operating philosophies could be developed.
That portion of the storage complex which is associated with the catalog (i.e., all files known
to the system) is viewed as one resource. Actual location of specific files is chosen to maxim-
ize performance, which generally means having the smallest return traffic from mass storage
to disk and the most possible free space on the disks.
11
All of this has led to the following pricing policy for on-line cataloged storage:
. The price for on-line storage space is being cut by 80 percent on December 22, 1979.
. As of December 22, 1979, space for all files appearing in the catalog will be charged at
the same rate, independent of where they reside in the storage hierarchy.
. No charge will be made for recalling files or moving them to mass store.
. The per-file cover charge will remain as before.
. Limits will be applied to the total cataloged space. The disk space limit will be doubled
for all users and project managers.
Current CYBER disk charge practice is as follows:
. All files, regardless of type, size, or residency, are charged 0.15 SRU's per day for
overhead.
. All disk resident files are charged for space allocated at the rate of 0.03 SRU's per
PRU per day (this price may be reduced 80 percent on December 22, 1979).
• Non-disk resident (migrated) files are not charged for space allocated (this will change
on December 22, 1979).
. Charges are made once every 24 hours.
. Files which belong to identifiable projects which have run out of funds are charged for
7 days after the project runs out of funds. Thereafter, they are not charged.
. Files which belong to projects which are one (or more) of the following are not
charged:
not identifiable (i.e., AAAA)
cancelled
expired
out of funds more than 7 days
owner of file does not belong to project
For further information on CSO disk policy, see the January, 1979 issue of OFF-LINE.
12
MAGNETIC TAPES FROM SCRATCH
This article was originally published in the December 1979 issue and was adapted with
permission from Two Bits Worth, the newsletter of the University of Southwestern
Louisiana Computing Center. Alicia Towster wrote the original article on the basis of
interviews with Sam Bullard, Bob Sonnier and James Dugel.
Magnetic tapes are simply very long thin strips of mylar backing coated with a thin layer of
some material which can retain magnetic information. They are an economical and relatively
durable way to store information; for $10 you can acquire a tape capable of storing up to
about 4.5 million words of useful information on the CYBER or about 34 million bytes of
useful information on the IBM; and if you are able to keep this tape away from heat, humi-
dity, dirt, magnetic fields, and malfunctioning tape drives, it can give you reliable service for
several years. (Even under the best circumstances, tapes will eventually begin to show signs
of wear. Thus, cautious tape users will often keep more than one copy of their important
data.) Magnetic tapes have a standard width of one-half inch; several lengths are available:
400', 1200', 2400', and 3200'. At either end of a tape is a shiny aluminum patch to mark the
Beginning of Tape (BOT) and End of TAPE (EOT).
It sounds simple enough.
Why, then, when you approach a computer installation carrying a "foreign tape" (that is, one
which was not created on that particular computer), does the staff eye you with misgivings,
rather as though you were carrying a foreign virus? They are hoping that you can describe it
clearly enough that they will quickly know what treatment to prescribe.
You see, it is not nearly as simple as tape cassettes — tape them on one machine, play them
back on another. There is a considerable variety of ways that information can be written
onon a magnetic tape. And it is entirely possible to produce a tape which is totally incompati-
ble with the machine on which you desire to use it. These incompatibilities may be due to
either hardware (the kinds of tape drives which are available) or software (programs which
read and write tapes, normally supplied by the manufacturer).
First, there are differences in the ways that tape drives can physically arrange and/or access
information on a tape. Some drives will read or write nine bits (binary digits, either a zero or
one) in a row across the width of the tape; these are called "nine-track" drives, and the tapes
they produce are called, reasonably enough, nine-track tapes. There are also seven-track
drives which deal with seven-track tapes on which rows of seven bits are stored.
Information can also be arranged differently down the length of the tape: this is called tape
"density" and is measured in "bpi" which stands either for bits per inch (if you think in terms
of only one track at a time) or bytes per inch (if you think of the entire row of tracks as a
byte). Possible densities are 200 bpi, 556 bpi, 800 bpi, 1600 bpi, and 6250 bpi. A particular
drive is limited in the number of different densities that it can handle.
In addition to data bits, tapes contain additional information that enables the drives to con-
tinually be checking on whether or not they are reading your data successfully. For example,
on a nine-track 800 or 1600 bpi tape, one bit in each row is used as a "parity bit" - that is, it
is not actually part of your data; rather, its value is a function of the value of the other bits in
the same row.
13
Most typically, "odd parity" is used; that is, the ninth bit is set so that the sum of the one bits
in the row will be an odd number. On 800 bpi tape, parity information is also computed for
groups of rows and written at regular intervals along the tape. On 1600 bpi tape, the zero bits
are actually written, rather than denoted by the absence of a one bit. In addition, unique pat-
terns of bits are written at intervals for the purpose of synchronization. Normally you can
remain totally unaware of these extra bits; the tape drive hardware/ firmware generates them
and/or checks them automatically. If they fail to check out, the proper action is left to the
program which is controlling the attempted read. Often such a program will elect to retry the
read; if this does not work, you will, of course, get an error message. Such a message could
indicate a bad tape, a malfunctioning drive, or a mismatch between the way your tape actually
is (tracks, density, or parity type) and what the drive expects it to be.
Information is not written continuously along the length of the tape; it is written in "blocks"
or "physical records". These physical records may well differ in size from the logical division
of the data ("logical records") which you have placed on the tape. If logical records are short,
several may be grouped together into one physical record; long logical records may also span
several physical records. Most typically, physical records will be of uniform size, but it is also
possible for their size to vary. The space between records is called the "interrecord gap".
The tape drive itself deals only in terms of these physical records, moving them from the tape
to a buffer in the computer's memory (or vice versa, in the case of writing a tape). Because
these buffers must reside in the computer's real memory, many installations will have some
upper limit on the size of physical tape records that they can handle.
But this is only a small part of what makes tapes difficult. Consider -- how are those mean-
ingful bits of data to be interpreted? Remember that a CYBER word consists of 60 bits and
your tape consists of numerous rows of 8 bits and 8 does not go evenly into 60, whereas an
IBM word consists of 32 bits and 8 does go evenly into 32 ... or perhaps your tape was created
on yet another computer with a different word size?
Well, luckily there are standards. But unfortunately there are more of these than we might
like. First there is the theoretical standard: the American National Standards Institute's
(ANSI) specification of how to map those rows of bits on tape back and forth between words
on the computer. Then there is the de facto standard: the IBM standard, which has sheer
numbers on its side. But there is nothing to prevent any of the computer manufacturers
from developing their own internal standards, tailored to their hardware, and so they do.
This can potentially produce more efficient or appropriate use of tapes so long as you are
committed to a particular brand of computer, but will almost certainly cause problems if you
try to switch brands and take your data with you.
To complicate matters still further, these various standard tapes can have subtypes. Tapes
may be either labeled or unlabeled. Labels are, in theory, extremely helpful, since they con-
tain information about the tape you are dealing with. But there's a catch: label formats can
vary, too, and some computer installations may not have programs available to interpret tape
labels. Thus, they could actually make your tape harder to read.
Another complication arises from the different ways that information can be represented
inside various computers. The internal binary representation of machine instructions or data
will normally be specific to a particular computer; thus, it is useful to tape such binary
14
information only if the tape will be reread on exactly the same sort of computer. "Portable
tapes" (that is, tapes which are to be carried to a different sort of computer) should use one
of the standard sets of character codes to represent the information; there are two widely
used standards: EBCDIC, which is promulgated by IBM, and ASCII, the American Standard
Code for Information Exchange. In the absence of other information, you may expect most
IBM format tapes to use EBCDIC and most ANSI tapes to use ASCII.
With so many variables involved, clearly it is only sensible to write down the appropriate
information about the tape when it is created and to keep this information with the tape.
However, this does not ensure that it can be read by the computer of your choice, which sim-
ply may lack the hardware or software that you need.
This is a lot of information to keep straight, so here are some checklists.
Ways in which tapes (and installations) can vary:
• number of tracks
• density
• type of parity
• size of records
• size of block
• type of format
• labeled or unlabeled
• type of encoding
Portable tapes which both the CYBER and IBM can handle:
• nine-track
. 800 or 1600 bpi
• odd parity
• fixed record size
. fixed block size between 18 and 6000 characters (for the CYBER, it is preferable that
the blocksize not exceed 5120 characters)
• IBM and ANSI forms
15
. most types of labels
. ASCII or EBCDIC encoding
A tape of this sort could be read at many installations:
• nine-track
. 800 bpi
• odd parity
• a fixed record size
• a fixed block size no larger than 2048 characters
. IBM format
• unlabeled
. EBCDIC encoding
When you send someone a tape, be sure you also send complete information about the
characteristics of the tape. If you receive a tape from someone, be sure you also receive com-
plete information about the tape's characteristics.
Tapes may be a little difficult, but misinformation or lack of information about them can
make them much, much worse.
CSO MAILBOX FACILITY
Reprinted from OFF-LINE, January 1980
CSO has recently completed the development and testing of an electronic mailbox facility for
use on the CYBER 175. The purpose of the facility is to allow you to leave messages for
another person who uses the CYBER and/or to view messages left for you.
CAUTION: Since this is a new facility, there will be bugs or inconsistencies
for a time. Please report these along with any suggestions you may have by
using TELL,UI = 103.
There are three basic utilities connected with the mailbox facility:
• WHO to determine if a person is on the system
• TELL to leave a message for one or more people
16
. MESSAGE to view messages left for you and optionally respond to them.
You can refer to one or more people in a number of ways:
. You can use their name(s) as it has been entered under the MANAGE program.
However, this may be inadequate if someone has a common name.
• You can use their University ID number.
• You can use their user number.
• You can use their user index or their user index hash.
• You can refer to anyone in the same charge, project with which you are associated.
. You can refer to a predefined set of names that you have established in your OPTION
file.
The following features and options exist in this facility:
• A copy of the message is placed in the local file NEWMAIL automatically.
• All messages are placed in a central system file. If you send a message to many peo-
ple, only one copy of the message is kept in this central file.
. After everyone has seen the message, it is automatically removed from the central file.
• If a message remains in the central file over two weeks, the originator of the message
is warned that not everyone has seen it and that the message will be removed after
one more week (this means the message would be in the central file for a total of
three weeks). The originator would then be allowed to extend this deadline one more
week, thereby keeping the message in the central file for four weeks. After this
extended period, the originator will again be warned if not everyone has seen the mes-
sage.
• At any time, you can delete a message you have sent to others.
• After viewing a new message, you may:
REPLY to the sender.
FORWARD a copy to other people.
COPY the file to a local file or to the terminal again.
SAVE a copy of the message in a permanent file.
Go on to the next message - making the current message no longer accessible
except from the local file NEWMAIL.
17
. Get on-line assistance via the HELP command of the mailbox facility.
• If you are going on vacation, you can have any future messages that are sent to you
routed to a friend so that someone can respond to your messages while you are gone.
. You can receive a SUMMARY of all messages waiting to be seen by you, and all mes-
sages you have sent to others that have not yet been received.
. You can create a local file containing a message and use this file as input when "telling"
someone.
Full details on all options are available via the HELP commands of the mailbox facility. To
access these HELP files, enter one of the following:
MESSAGE/HELP.
TELL/HELP.
WHO/HELP
To use these three commands, in the simplest form, you do the following:
. Determine whether a person you know is on the system by using the WHO command:
WHO.THOMAS MIKE.
. Having found that this person is on the system, you can leave the person a message
by entering:
TELLTHOMAS MIKE.
You will then be prompted by:
WHAT?
Enter your message after this prompt. End the message with a bare carriage return
after the prompt of a single question mark. You then will receive an acknowledge-
ment that the person has been "told".
• You can view messages to you by entering:
MESSAGE
You will be given a summary of all messages waiting to be seen. Then you will be
asked
WHAT TO DO?
simply press the carriage return to see a message that has been sent to you.
18
When there are no more messages to see, you can exit from MESSAGE by entering
the command:
E
Whenever you login to the system, you will be told if you have any messages waiting for you.
In the future, there will be a more complete reference to the system, and all of the options
will be described more thoroughly.
USE COMMAND AVAILABLE FROM TIME-SHARING
Reprinted from OFF-LINE, January 1980
The USE command, which was previously announced for batch use {OFF-LINE, November
1978, "USE Program Now Available") is now available from time-sharing as well. The pur-
pose of USE is to guarantee that any needed file which has been migrated is reloaded before
an attempt is made to use it.
The format of the USE command is:
\JSE,filel.file2,...,filen.
or
USE, filel, file2... Wen/UN = user number.
Where filel, file2, etc. are the names of permanent files which must be on disk if subsequent
processing is to succeed. The USE command should appear before any of the named files
appear in a GET, ATTACH, or OLD command.
When used from time-sharing, USE will produce no output if all the named files are on disk.
If one or more of the named files have been migrated, USE will generate a RELOAD request
and issue the message
xxxxxxx WILL BE RELOADED
at the terminal for each file that must be reloaded. After issuing all the required RELOAD
requests, USE aborts with the message:
SOME FILES UNAVAILABLE TRY AGAIN IN 30 MINUTES.
USE behaves in a similar fashion when it is used in a batch job. In this case, however, USE
reports the files which had to be reloaded in the dayfile and waits for completion of the gen-
erated RELOAD requests instead of aborting.
Multiple USE commands are processed together. Contiguous USE commands are processed
as if there were a single USE command with an extended list of file names. For time-sharing
jobs, this behavior insures that all necessary RELOAD's are generated before aborting. For
19
batch jobs, it eliminates multiple waits for RELOAD's. A maximum of 100 file names may
be used in a contiguous group of USE commands.
TEXT PROCESSING SERVICES AT CSO
Reprinted from OFF-LINE, January 1980
CSO offers text processing services on two systems, CYBER and UNIX. Most general-
purpose text processing is done on the CYBER 1 75 at DCL, using the text editor ICE in con-
junction with the text formatter RNF. If you are not presently a CYBER user, and wish to
get started, ask for assistance from your department, or contact the CSO Accounting Office,
Room 162 DCL, 333-6769.
CYBER
Once you have acquired a CYBER signon, your next step is to learn how to create and edit
files. This is done using ICE, which is a text editor. The ICE tutorial, A Tutorial Guide to the
ICE Text Editor, contains the necessary instructions to do simple file creating and editing
tasks. For the user who wishes to acquire more proficiency, a copy of the ICE Reference
Manual will be useful. Another manual which contains useful information for a beginner on
the CYBER is the Introduction to the CYBER 175. These manuals may be obtained in the
CSO Distribution Center, Room 164 DCL.
RNF is a program which reads a file containing text, interspersed with formatting commands,
and then produces formatted output. The following are some of the more important features
of RNF:
Pagination - titles, sub-titles, page numbers
Filling of text and right-margin justification
Numbered lists
Paragraphing
Tabs
Single, double, and triple spacing
Automatic chapter, section, and sub-section numbering
Macros and variables for user-defined text structures
Output from RNF can be displayed directly on any time-sharing terminal, or it can be stored
in a file for later use. Typical output devices are CRTs, DECwriters, and the line printers at
CSO RJE sites.
20
The following is an example of some RNF output:
An RNF Example
This is some text which has been run through RNF on the
CYBER. The right margin is justified by default, but
that can be changed easily with a simple RNF command.
The text Delow is a list with justification turned off:
1 . The numbers at the beginning of each entry of
the list are generated automatically by RNF.
2. The indents are also handled automatically.
This output was done on the Diablo printer.
The user can also create
indents and tabs to suit
many documentation
needs .
For final, camera-ready copy, the best quality output can be obtained from a Diablo-type ter-
minal equipped with a carbon ribbon. CSO maintains a dial-up Diablo terminal for general
use, in Room 209 of the Astronomy building. The user must call 333-8150 to make a reser-
vation for its use. A lead time of 2 or 3 days is recommended. The cost of using the Diablo
terminal is 1 5<t per page of output. (For further information on the Diablo, see the
December issue of OFF-LINE, "Diablo Output Service")
For more information on RNF and how to use it, pick up copies of the RNF User's Guide and
the RNF Reference Manual, either in the CSO Distribution Center, Room 164 DCL, or at
CSO South, Room 70 Commerce West.
UNIX
CSO also provides a text processing service on the PDP- 11/50 UNIX operating system which
is located on the second floor of the Astronomy Building. This system features the TROFF
formatting program developed by Bell Labs, equation and table formatting programs, and a
phototypesetter for high-quality, camera-ready output. Using TROFF, its associated macro
packages, and the equation and table programs, the user can generate a wide variety of docu-
ment styles, including footnotes, automatic table of contents, multiple columns, equations,
and complex tables, in several different fonts and point sizes.
21
Following is some output from TROFF, using both the table and equation programs:
Composition of Foods
Food
Percent by Weight
Protein
Fat
Carbo-
hydrate
Apples
Halibut
Lima beans
Milk
Mushrooms
Rye bread
.4
18.4
7.5
3.3
3.5
9.0
.5
5.2
.8
4.0
.4
.6
13.0
22.0
5.0
5.0
52.7
Name
Sine
Error
Bessel
Zeta
Function
Gamma V(z)=J t:~]e~'dt
s'm(x)=i-(e'x-e-ix)
Li
erf(z)=-^(V':<//
V7T •'o
}(z)=-Ljc
J0(z)=— \ cos(zsin9) d9
of 0
£(s)=£A;-5 (Re s>\)
i=l
Access to UNIX is available on a restricted basis, which is determined in part by the need for
using its facilities. Inquiries should be directed to Debbie Weller (333-8150).
Which System to Use?
The choice of which text processing system to use depends on several factors. The primary
advantage of RNF on the CYBER is its easy accessibility for a large portion of the user com-
munity. It is easy to learn, and inexpensive (RNF runs usually cost about 10<t per page of
output, for computer time used, and line printer output costs are about 5<t per page). How-
ever, it does not have footnoting, or a reliable superscript or subscript capability. Also, its
output is limited to the standard alpha-numeric character set, and it provides no support for
devices with extended character sets or special plotting features, thus ruling out most
equation-type applications.
TROFF, on the other hand, although a great deal more powerful and versatile, is expensive.
For example, typeset copy costs from $3.00 to $4.00 per page for typical composition, and
sometimes more for complicated applications, such as matrices, multiple columns with
densely packed text, equations, and tables.
If you are already using one of the two systems, help is available from the Text Processing
Consulting Office, in Room 207 Astronomy, phone 333-7318. Consulting hours are from
9:00 AM to 12:00 noon and from 1:00 PM to 4:00 PM daily.
22
SERVICE KEYPUNCHING AT CSO
Reprinted from OFF-LINE, April 1980
Due to the declining use of service keypunching, CSO is planning to contract all keypunch
work to a commercial keypunch service. Users will see little change in the service and will
continue to deal only with CSO when submitting or retrieving keypunch jobs. CSO itself will
provide all interfacing with the contracting agency.
Over the past 18 months, use of service keypunching has been at a level too low to justify the
cost of in-house processing. This has held true even though charges for the service have
been held at a level below the usual commercial costs. At the same time, it is impractical to
simply reduce the size of the operation since it is now at the minimum level to accommodate
large jobs with reasonable turnaround times.
The target date for conversion of the service is July 1, 1980. The exact date will be governed
by how quickly the present data entry staff can be placed in other similar positions on the
campus. It is expected that there will be a transition period during which some work will be
processed in-house, and the remainder will be processed commercially.
We emphasize that CSO is continuing to offer service keypunching as one of its services. The
method of processing submitted jobs is changing, but little else. It is expected that small
keypunch jobs (400 cards or less) will experience an additional one or two day delay because
of transit times. Larger jobs should be processed more quickly since a larger staff will be
available to keyboard the data.
CSO is planning to adjust its method and rate of charging for keypunch service to be con-
sistent with the contracting agency. Users who have received Research Board support for
keypunch services will not be affected by the rate change since their allocation is denominated
in the number of cards to be punched. Other users will probably incur higher data entry
costs.
CATALOG OVERFLOW - SIZE
Reprinted from OFF-LINE, April 1980
If you have ever tried to SAVE or REPLACE a file and received the following message:
CATALOG OVERFLOW • SIZE
read this article to learn why you receive such a message, and what to do about it.
If you are trying to SAVE a new permanent file, the message indicates that you have tried to
exceed the amount of permanent INDIRECT access disk space you are permitted. To verify
this, you can check your disk space limitations by entering the command
LIMITS.
23
and checking the displayed value for disk space. Then, by entering the command
DIRECT/PROJ/SUM.
you can determine how many PRU's of indirect files you are currently using.
If you are doing a REPLACE to update a permanent file and receive the message, the above
discussion may apply. However, there is another possibility. To understand it, we'll have to
explore the connection between files, projects, and limits.
When a file is created, the charge, project used at the last signon is stored along with the
user's data (file). This charge, project is then billed for the disk space used by the file. The
Project Manager can control the amount of disk space funded by the project by setting the
various file limits for each person in the project (e.g., the disk space limit we are currently
discussing).
The BILL or CHARGE command sets these limits for the user at signon, using the
charge, project specified. If a user signs on using one project and REPLACES a file belonging
to another project, the limits associated with the current signon are not applicable to that file.
The limits which must be enforced are those set by the owning project. Ideally, these limits
would be extracted from the same database used at signon by the BILL or CHARGE com-
mand. This solution, however, is very difficult and costly, and an alternate solution has been
adopted.
The alternate solution is to store, at file creation, not only the owning charge, project, but also
the file limits currently in effect. Whenever a REPLACE is done on a file belonging to a pro-
ject other than the one used at signon, these stored limits are used. Note that the usual lim-
its (those set up at signon and displayed by the LIMITS command) are used whenever the
project stored with the file matches that used at signon.
We now come back to the error message we started with. If you receive this message after a
REPLACE command, you must first determine which limits are being enforced - those
displayed by the LIMITS command or those stored with the file. If all of your files were
created under the charge, project you used at signon, your current project limits are being
used. If this is not the case, enter the command:
DIRECT, fi/e/PROJ/LIMITS.
This will display both the charge,project which owns the file and the limits stored with the
file. You can update this information by entering:
CLAIM, file.
This will replace the charge, project stored with the file with the charge, project used on the last
BILL or CHARGE command, and also will replace the stored limits information with your
current limits information. The output from
DIRECT,/ii/e/PROJ/UMITS.
24
should then reflect the changed information. Using the CLAIM command also will cause all
disk space charges for the file to be charged to the charge, project current at the time of the
CLAIM.
SECURITY SAFEGUARDS
Reprinted from OFF-LINE, May 1980
Due to some recent incidents with unauthorized users on the CYBER, CSO would like to
take this opportunity to remind all users about security safeguards for files and accounting
information. It may be legal to look at files which the owner has not taken proper steps to
protect. However, it is illegal to pose as another person, through the presentation of
identification and passwords, for the purpose of making unauthorized use of their accounts.
Allocations of service supported by campus funds are for specified research and instructional
purposes. Their use for harassment of others, for violations of file privacy, or as a means of
gaining illegal access to other accounts will result in disciplinary or legal action.
To safeguard your resources from such abuse, please be sure that your files which contain
accounting information, especially passwords, are properly protected. Such files should be
restricted to private mode. If it is necessary to allow access to a few people, you can permit
the files explicitly to their user identifications.
We advise all CYBER users to change their passwords frequently. Users with null passwords
should assign passwords to their User Numbers. If someone has been using your resources
illegally, however, just changing the password may not provide adequate protection. You
should also check the permission privileges for all your files, and change the permission of
any public files to private if they should be protected.
The following commands show a number of ways to check and change the permission
privileges of files, and to change your password.
DIRECT/CT=PU. Will list all files in your area which are public, i.e.,
can be accessed by anyone signed on to the
CYBER.
DIRECT/ PTOTAL. Will list all files in your area with the specific IDs
permitted to access them.
GET, pfn. Will get a local copy of the file pfn.
PURGE,/?/>j. Will delete the permanent copy of file pfn.
SAVE,pfn. Will save the file pfn in your permanent directory
with no permission privileges assigned to it.
25
PERMIT, pfn,userid=R. Permits read-only access to file pfn for the
specified userid only.
PASSWOR, oldpas.newpas. Changes the password oldpas to a new password
newpas. This changes the password immediately,
and the new password must be used the next time
you sign on. An even safer way to change your
password is to simply enter the command
PASSWOR. The computer will then prompt you
for the old and new passwords and the entries you
type in will not be echoed on the screen. If your
password is presently null, enter a carriage return
when prompted for oldpas.
Project Managers should check to ensure that no User Numbers have been illegally added to
their project. Also, if they have a null project code word, they should assign a code word to
their project immediately. To assign a code word to a project, enter:
MANAGE.
P,charge,project
CODE WORD=newcode/PN
E
One other word of caution: it has been observed that some users have signed onto public ter-
minals and have not logged off when leaving for a short period of time. Perhaps the reason-
ing for this is that they believe this "reserves" the terminal for them. However, it is a
dangerous practice because it allows anyone to then use that signon for whatever purpose - to
access the files, use the person's funds, etc. If you do this type of thing, you must accept the
responsibility for what happens. LOG OFF when you leave a public terminal - protect your
files and your funds!
NCAR GRAPHICS SOFTWARE
Reprinted from OFF-LINE, May 1980
CSO has installed the NCAR Graphics Software on the CYBER. This software, obtained
from the National Center for Atmospheric Research, contains graphics utility subroutines for
drawing
. Contour plots
. 3-D surfaces with hidden lines removed
• World map projections
and other high-level applications. Plots may be produced on a Tektronix graphics terminal or
a CalComp plotter.
26
The NCAR Graphics Software is a library of subroutines called from a FORTRAN program.
The NCAR software achieves device independent plotting by writing a local disk file of plotter
commands, which is later plotted on a specific device requested by the user. This file of
plotter commands is called the meta-code file.
Documentation describing the subroutine usage and parameters is available for inspection in
the CSO North Consulting Office (Room 138 DCL). This same information is available on
the CYBER via a procedure, NCARDOC. The commands
GRAB.NCARDOC.
NCARDOC.
will give a brief summary and directions to obtain additional information and subroutine
writeups.
A typical use of the NCAR software involves the following sequence of commands:
GRAB,NCAR. Attach the NCAR library.
FTN,I=/vogram,L=0. Compile the FORTRAN program with calls to NCAR
subroutines.
LGO. Execute the program. This does not generate the plot
itself, but rather, the meta-code file of plot instructions.
NCARTRN. To display the plot on a Tektronix terminal.
NCARTRN,PLOT=CALC. To display the plot on a CalComp plotter.
PLOT,TAPE99. This plot statement is necessary to actually send the file
to the plotter.
Once the meta-code file has been generated, it may be displayed at will on either Tektronix or
CalComp equipment by using the appropriate NCARTRN statement.
NOTE: The file name NCARMC must appear on the program statement of your FORTRAN
program. This is the name of the meta-code file written by the NCAR software.
A sample program and plot are shown here:
PROGRAM NCARTST(INPUT,OUTPUT,NCARMC)
C
C THE FUNCTION
C
C Z(X,Y)=.25*(X+Y + l./(X-.l)**2+y**2-l-.09)
C -l./((X + .l)*'2+Y**2+.09))
C
C IS EVALUATED FOR
C
C X=-l. TO 1. IN INCREMENTS OF .1 AND
C Y =-1.2 TO 1.2 IN INCREMENTS OF. 1.
27
C XX CONTAINS THE X-DIRECTION COORDINATE VALUES FOR Z(X,Y)
C YY CONTAINS THE Y-DIRECTION COORDINATE VALUES FOR Z(X,Y)
C Z CONTAINS THE FUNCTION VALUE
C S CONTAINS VALUES FOR THE LINE OF SIGHT FOR SRFACE.
C WORK IS A WORK ARRAY
C
REAL XX(21), YY(25), Z(21,25), S(6), WORKU050)
C
DATA S(l), S(2), S(3), S(4), S(5), S(6)/
1 -8.0, -6.0, 3.0, 0.0, 0.0, 0.0/
C
C FILL XX AND YY COORDINATE ARRAYS AND Z FUNCTION VALUE ARRAY
C
DO 20 1 = 1,21
X= .l*FLOAT(I-ll)
XX(I)=X
DO 10J = 1,25
Y= .l*FLOAT(J-13)
YY(J)=Y
Z(I,J) = (X+Y + l./((X-.l)"2+Y**2 + .09)-
1 l./((X + .l)**2+Y**2 + .09))*.25
10 CONTINUE
20 CONTINUE
CALL SRFACE (XX, YY, Z, WORK, 21, 21, 25, S, 0)
STOP
END
28
DIABLO RESERVATION SERVICE TO CHANGE
Reprinted from OFF-LINE, June 1980
Effective June 2, 1980, users will no longer be able to make reservations to use the Diablo
terminal located in Room 209 Astronomy. Instead, CYBER users will be required to queue
their RNF output, intended for printing on the Diablo, by executing a CYBER Control
Language (CCL) procedure. Users unfamiliar with RNF are referred to the RNF User's
Guide or the RNF Reference Manual (available in the Distribution Center, Room 164 DCL).
All RNF source files must contain the .LPT command in the first line of the file, and must be
processed by RNF into an output file. All non-RNF source files must have carriage control
and begin with a page eject. The RNF output file or the non-RNF source file can then be
submitted to the Diablo queue via the CYBER CCL procedure.
The following conventions have been established:
1. The top of the form (page) will be set to one line below the horizontal perforation
in the paper.
2. Plain white 20-lb paper, 14 7/8" wide by 11" long, will be used.
3. If the pica font is selected, the printing will be done with 10 characters and 6 lines
per inch.
4. If the elite font is selected, the printing will be done with 12 characters and 6 lines
per inch.
After you have prepared your file for submission, you should access and execute the DIA-
BLO procedure by issuing the following commands from time-sharing on the CYBER:
GRAB.DIABLO.
DIABLO.
The procedure then prompts you for your name, telephone number, choice of font (pica or
elite), and the name of your RNF output file. The RNF output file may be either a local or a
permanent file. If the procedure is unable to find your file, it aborts with a message to that
effect. If the procedure finds the specified file, it places it in the Diablo queue.
On-line help is available by entering the command
HELP
and then by entering DIABLO (the name of the procedure) after the question mark prompt.
Turnaround time should be no longer than 24 hours, except in cases of system downtime or
hardware malfunctions. The output of jobs received on Fridays or prior to a University holi-
day will be available the next working day.
29
Your printouts may be picked up at 203 Astronomy. The cost will continue to be $0.15 per
page, rounded up to the nearest dollar. You may pay for the printing by charging it to a
University account number or with a personal check.
Questions regarding this service should be directed to Debbie Weller, 209 Astronomy (333-
8150).
KEYPUNCH SERVICE
Reprinted from OFF-LINE, July 1980
As announced in the April issue of OFF-LINE, CSO has discontinued internal processing of
keypunch work. This move was made for financial reasons since the amount of data entry
provided by CSO no longer justified continuance of a dedicated data entry staff.
At the time the April article was written, we planned to use an off-campus commercial data
entry service with CSO providing all interfacing with the contracting agency. Since that time,
another campus unit, Survey Research Lab (SRL), has indicated a desire to assume CSO's
data entry workload. An agreement has been worked out with SRL which we hope will
minimize the amount of information which must be transferred between CSO and SRL with
the attendant possibilities for confusion.
Users whose data entry work is funded through the Research Board will see little change.
Requests for data entry funds are to be made through existing channels, input is to be
delivered to CSO for punching at SRL, and output is to be picked up at CSO. Allocations will
continue to be made in numbers of cards to be punched and CSO will continue to provide all
accounting associated with data entry. Occasionally, a user with a complex data entry task
may be asked to talk directly to the data entry staff at SRL. Data entry work funded through
the Research Board is to be taken to the CSO Accounting Office and completed data entry
jobs are to be picked up there. This office, as well as document sales and terminal rentals, is
scheduled to move to 1208 West Springfield during July.
Users whose data entry work is being funded with hard contract money are to deal directly
with SRL. These users are to submit their work directly to the SRL data entry staff, and
claim the completed work at SRL. They will be billed directly by SRL for services delivered.
SRL's rates are $7.25/hour for normal work and $10.50/hour for priority work. The number
of cards produced per hour will vary with the data being entered. A rough estimate, based on
previous experience in CSO's data entry shop, is 100 cards per hour. Users paying for data
entry services with hard contract funds should make all arrangements with Mrs. Frances
Sykes, 310 SRL, 1005 West Nevada, Urbana (333-7328).
SRL maintains a staff of three full-time data entry operators and hires part-time help to staff a
second shift whenever the load requires it. Additional full-time personnel will be hired if
needed to maintain reasonable turnaround time.
30
RATE CHANGES FOR THE IBM 4341
Reprinted from OFF-LINE, August 1980
With the installation of the IBM 4341, which is replacing the IBM 360/75, a revision in pric-
ing policy for IBM services will become effective September 2, 1980. These prices will be for
the continuation of old services available under OS/MVT in batch mode. Further revisions
will occur with the anticipated initiation of terminal-based services.
The two principal changes are to be in the areas of disk storage and the effect of memory on
CPU or I/O charges.
Disk charges will be reduced from 0.01 service units per track per day to 0.0017. At the rate
of $0.62 per service unit, this would mean a charge of 3.93 service units at a cost of $2.43 to
store one million bytes for one month.
Processing charges are to be computed according to the formula:
0.00036(100T + IOM0.0014R + 0.7)
Where:
T = CPU time in seconds
IO = number of IO requests issued
R = memory Region in thousands of bytes
This wil replace the old formula:
0.00036U00T -I- IOH0.0045RF + 0.001 5RS + 0.5)
where RF and RS were fast and slow regions, respectively.
The effect of the change will depend on the memory required, and the relative speed of the
two machines for a specific job. The larger the memory requirement, the greater the decrease
in service units per hour. The following examples should illustrate this - assuming that the
4341 is 85 percent as fast as the 360/75.
Region in
Service Units Charged
1000 bytes
to do a One-hour 360 Job
360/75
4341
0
64.80
106.73
128
139.44
134.05
256
214.09
161.37
512
363.40
216.02
1024
662.00
325.31
31
The new formula will give an approximate equivalence to CYBER batch charges, where the
memory component was reduced a year ago.
CSO VIDEOTAPES
Reprinted from OFF-LINE, August 1980
CSO has recently completed the initial stages in the preparation of videotapes describing the
CYBER 175. The following three tapes are available for viewing.
CSOVT1 Introduction to Computing at CSO. This videotape is intended for the
first time computer user, and for people wishing to learn more about
CSO's facilities.
CSOVT2 This videotape consists of four sections:
• Using a Terminal - How to operate a computer terminal.
• Introduction to CYBER Time-Sharing - How to access the CYBER
using a terminal.
• File Usage - Local files and indirect access permanent files.
. Introduction to ICE Text Editing - The basics of time-sharing edit-
ing.
CSOVT3 This videotape, Running a FORTRAN Program on the CYBER, consists
of three sections.
• Concepts involved
. PROGRAM statement
. CYBER control statements
Anyone may view these tapes by going to the Undergraduate Library in person to make a
reservation to use the videotape equipment. The library hours are 8 AM - 8 PM Monday
through Thursday, 8 AM - 5 PM Friday, and 1 PM - 5 PM Saturday and Sunday.
When you check out the videotapes, you should also pick up one of the handouts that are
available with the tapes. The handout summarizes the contents of each tape in more detail,
and facilitates note taking.
32
Your opinions regarding these videotapes would be greatly appreciated. You may use the last
sheet of the handout, or write directly to:
CSO VIDEOTAPES
164 DCL
CAMPUS
INDEX
An index of the September 1979 through September 1980 issues off OFF-LINE is provided
on the following three pages. If you would like a copy of any particular issue after looking
through this index, call Lynn Bilger (Editor), 333-6236.
POLICY
CYBER Rates Vol. 7, No. 9, September 1979
On-line Storage Vol. 7, No. 9, September 1979
Protect Your Computer Funds Vol. 7, No. 9, September 1979
Committee on Instructional Computer Use Vol. 7, No. 10, October 1979
CYBER Rates Vol. 7, No. 11, November 1979
Rate Reduction Period Vol. 7, No. 11, November 1979
Christmas Schedule Vol. 7, No. 12, December 1979
LSI 11 Project Vol. 7, No. 12, December 1979
Caveat Emptor Vol. 8, No. 1, January 1980
Planned Disk System Upgrades Vol. 8, No. 3, March 1980
IBM 360 to be Replaced Vol. 8, No. 3, March 1980
Service Keypunching at CSO Vol. 8, No. 4, April 1980
Keypunch Service Vol. 8, No. 7, July 1980
Rate Reduction Period Vol. 8, No. 8, August 1980
Service Unit Rates to be Reduced Vol. 8, No. 8, August 1980
Rate Changes for the IBM 4341 Vol. 8, No. 8, August 1980
SYSTEM NOTES
Reliability Report Vol. 7, No. 9, September 1979
Reliability Report Vol. 7, No. 10, October 1979
Reliability Report Vol. 7, No. 11, November 1979
Reliability Report Vol. 7, No. 12, December 1979
Reliability Report Vol. 8, No. 1, January 1980
Reliability Report Vol. 8, No. 2, February 1980
Reliability Report Vol. 8, No. 3, March 1980
Reliability Report Vol. 8, No. 4, April 1980
March Reliability Report Vol. 8, No. 5, May 1980
April Reliability Report Vol. 8, No. 6, June 1980
May Reliability Report Vol. 8, No. 7, July 1980
June Reliability Report Vol. 8, No. 8, August 1980
July Reliability Report Vol. 8, No. 9, September 1980
CYBER
Permanent File Catalog Entries Vol. 7, No. 9, September 1979
Changes in ICE Vol. 7, No. 9, September 1979
EDUNET Network Vol. 7, No. 9, September 1979
1200 Baud Dial-up Service Update Vol. 7, No. 9, September 1979
CYBER Manuals - Revisions Vol. 7, No. 9, September 1979
Network Algorithms Available Vol. 7, No. 10, October 1979
Interested in IFPS? Vol. 7, No. 10, October 1979
CSO Purchases Tape Cleaner Vol. 7, No. 12, December 1979
New Version of GCS AVAILABLE Vol. 7, No. 12, December 1979
Magnetic Tapes From Scratch Vol. 7, No. 12, December 1979
CSO Mailbox Facility Vol. 8, No. 1, January 1980
EDUNET News: Computer Conferencing System Vol. 8, No. 1, January 1980
SIGMA - A New Conversational Graphical Language Vol. 8, No. 2, February 1980
TELENET Address Changes Vol. 8, No. 4, April 1980
Catalog Overflow - Size Vol. 8, No. 4, April 1980
Reference Guides Reorganized Vol. 8, No. 4, April 1980
Security Safeguards Vol. 8, No. 5, May 1980
NCAR Graphics Software Vol. 8, No. 5, May 1980
RECORD MANAGER Bug - Odd Block Sizes Vol. 8, No. 5, May 1980
New ZETA Plotters Vol. 8, No. 7, July 1980
Some Utilities Allow Continuations Vol. 8, No. 7, July 1980
More Nonlinear Programming Algorithms, FLEX and GMET Vol. 8, No. 7, July 1980
SIMULA Now Available Through GRAB Vol. 8, No. 7, July 1980
CSO Videotapes Vol. 8, No. 8, August 1980
PRINT and PUNCH Charges More Accurate Vol. 8, No. 8, August 1980
On-line 763 Plotter Gone Vol. 8, No. 8, August 1980
CSO Videotapes Vol. 8, No. 9, September 1980
CSO Short Courses for the CYBER 175 Vol. 8, No. 9, September 1980
IBM
SAS Changes on the IBM Vol. 8, No. 4, April 1980
IBM PUBLIC Facility Gone Vol. 8, No. 9, September 1980
STATISTICAL SERVICES
Statistical Manuals on Reserve Vol. 7, No. 9, September 1979
CYBER SPSS Release 8 Available Vol. 7, No. 12, December 1979
SOUPAC on the CYBER: New Release Vol. 7, No. 12, December 1979
LISREL Version IV Installed on the CYBER Vol. 8, No. 1, January 1980
SPSS on the CYBER Vol. 8, No. 2, February 1980
Multivariance VI Installed on CYBER and IBM Vol. 8, No. 3, March 1980
SOUPAC on the CYBER Vol. 8, No. 3, March 1980
1979 Version of BMD Installed on the IBM 360 Vol. 8, No. 5, May 1980
SAS79 Interface to BMDP79 Vol. 8, No. 5, May 1980
Correction of April's SAS Article Vol. 8, No. 5, May 1980
BMDP-77 on the CYBER Vol. 8, No. 7, July 1980
NUMERICAL SERVICES
ALPS Removed : Vol. 7, No. 9, September 1979
Full Release of LINPACK Vol. 7, No. 9, September 1979
IMSL Notes Vol. 7, No. 10, October 1979
MSL Catalog Vol. 7, No. 10, October 1979
FORMAC Vol. 7, No. 11, November 1979
Software for Partial Differential Equations Vol. 7, No. 11, November 1979
ACSL on the CYBER Vol. 7, No. 11, November 1979
XMP on the CYBER Vol. 7, No. 12, December 1979
TELL,MATH Software Vol. 8, No. 3, March 1980
SLAM: Simulation Language for Alternative Modeling Vol. 8, No. 5, May 1980
Mini-DYNAMO Vol. 8, No. 5, May 1980
New Procedures: MATH and SAMPLES Vol. 8, No. 5, May 1980
New UOILIB Routines Vol. 8, No. 5, May 1980
Test Version of MINPACK Vol. 8, No. 5, May 1980
New Version of SPICE Vol. 8, No. 6, June 1980
CCL Versions of Some Procedures Available in FUTURE Vol. 8, No. 6, June 1980
Some Procedure Files to be Discontinued Vol. 8, No. 6, June 1980
FORTUOI Phase-out Vol. 8, No. 6, June 1980
IMSL Edition 8 Vol. 8, No. 8, August 1980
ACSL Version 6A Vol. 8, No. 8, August 1980
CCL Reminder Vol. 8, No. 8, August 1980
HELP Information Dumped From Procs Vol. 8, No. 9, September 1980
APEX Snafu Vol. 8, No. 9, September 1980
T.O.M.S. Algorithm Tape Vol. 8, No. 9, September 1980
New Routines in UOILIB Vol. 8, No. 9, September 1980
Permanent File Subroutines Vol. 8, No. 9, September 1980
IMSL Edition 8 Manual - Errata Vol. 8, No. 9, September 1980
TEXT PROCESSING
Diablo Output Service Vol. 7, No. 12, December 1979
Text Processing Consulting Vol. 7, No. 12, December 1979
Text Processing Services at CSO Vol. 8, No. 1, January 1980
Diablo Reservation Service to Change Vol. 8, No. 6, June 1980
MISCELLANEOUS
Terminal Tables for Sale Vol. 7, No. 9, September 1979
DECwriter II Maintenance and Repair Vol. 7, No. 9, September 1979
Give Us Your Name Vol. 7, No. 9, September 1979
Research Board Deadline for Departmental Allocation Requests Vol. 7, No. 10, October 1979
CSO Personnel Honored Vol. 7, No. 10, October 1979
No August Issue of OFF-LINE. Vol. 7, No. 10, October 1979
Teletype Repairs Vol. 7, No. 10, October 1979
1200 Baud Modems - Update Vol. 7, No. 10, October 1979
Computer Related Discounts Available Through Purchasing Division Vol. 7, No. 10, October 1979
Correction Vol. 7, No. 11, November 1979
SLIP Package Needed Vol. 7, No. 11, November 1979
DECwriter II Discontinued Vol. 7, No. 12, December 1979
New Terminal Site Opened Vol. 8, No. 1, January 1980
Ordering CSO Documentation Vol. 8, No. 1, January 1980
Give Us Your Name Vol. 8, No.l, January 1980
IBM 082 Sorter Vol. 8, No. 2, February 1980
Computer Related Discounts Through Purchasing Division - Update Vol. 8, No. 2, February 1980
Research Board Deadline for Departmental Allocation Requests Vol. 8, No. 3, March 1980
Computer Related Discounts Available Vol. 8, No. 3, March 1980
Notice of a Founding Meeting of an Apple Computer Users Group Vol. 8, No. 3, March 1980
Used Qume Cartridges Available Vol. 8, No. 3, March 1980
New Revision Pages From CDC Vol. 8, No. 5, May 1980
Teletype Repair Discontinued Vol. 8, No. 6, June 1980
LSI-11 Rental Systems Vol. 8, No. 6, June 1980
DEC Informal Users Group Meeting Vol. 8, No. 6, June 1980
A Cosmic Experience Vol. 8, No. 6, June 1980
CSO Short Courses Vol. 8, No. 8, August 1980
Mailing List Update Vol. 8, No. 8, August 1980
Computer Related Discounts Available Through Purchasing Division Vol. 8, No. 9, September 1980
Address Change Vol. 8, No. 9, September 1980
Mailing List Update Vol. 8, No. 9, September 1980
OFF-LINE's Mailing List
If you wish to be placed on our mailing list for future issues of OFF-LINE, if you wish to be
removed from the list, or if you wish to enter an address correction, please complete and
return this page. (Current subscribers are kept on the mailing list until a specific request for
removal is received, or until a mailing is returned as undeliverable.)
Check one:
□ New subscriber
□ Removal request
□ Address correction
Name:
Address:
CAMPUS or Zip Code
(If address correction, give old address and zip code below.)
Comments:
RETURN TO:
OFF-LINE
1 50 Digital Computer Laboratory
University of Illinois at Urbana-Champaign
Urbana, Illinois 61801
orvcro
T]J> — I*/-!
l/ll- >— '
i-.ro|—
r
j> x
-< J.<_
D
m
Computing
Services
Office
Off-Line
University of Illinois at Urbana-Champaign
VOL. 8, NO. 11 November 1980
EDITOR: Lynn Bilger
PHONE: (217) 333-6236
139 Astronomy Building
Urbana, Illinois 61801
Page
Contents
1
5
6
10
11
POLICY
Extension of Computing Service
CSO Consulting Services
Vacation Schedules
SYSTEM NOTES
September Reliability Report
IBM
IBM SSP to Go
MATHEMATICAL SERVICES
Q-GERT: Simulation Language for Modelers
MISCELLANEOUS
New CSO Documentation Available
DEC Users Group Meetings
CSO DIRECTORY - STAFF AND SERVICES
Administrative
Director
Business Manager
Secretary
User Services
User Accounting
Distribution Center
Systems Consulting
Statistical Services Consulting
Text Processing Consulting
Asst Dir User Services
Mgr, Statistical Services
Mgr, UNIX Operations
Documentation
Hardware/Software Support
Terminal Repair Service
CYBER Dial-up Numbers
Asst Dir Systems and Operations
Asst Dir Development
Asst Dir Engineering
CYBER-IBM Operations
UNIX Operations
Telecommunications
Hardware Selection Assistance
Laboratory Support Project
RJE Operations North
RJE Operations South
RJE Sites
Agriculture
Chemistry
Commerce West
DCL Routing Room
Electrical Engineering
Florida Ave Res Hall
Illinois St Res Hall
Mechanical Engineering
Psychology
CRH Snack Bar
Social Science
George Badger
150
DCL
333-4103
Stanley Rankin
Joyce McCabe
150
DCL
333-6530
150
DCL
333-1637
Gary Bouck
1208
W Springfield
333-7752
1208
W Springfield
333-6760
166
DCL
333-6133
65
Comm West
333-2170
207
Astronomy
333-7318
Robert Penka
173
DCL
333-4709
Larry Sautter
91
Comm West
333-2170
Larry Sautter
Lynn Bilger
139
Astronomy
333-6236
150
DCL
333-0969
110
baud
333-4000
300
baud
333-4000
1200
baud
333-4001
Sandra Moy
177
DCL
333-4703
J. M. Randal
120
DCL
333-9772
Cliff Carter
195
DCL
333-3723
Jack Knott
194a
DCL
333-6562
Debbie Weller
203
Astronomy
333-8150
Tom Kerkering
16
DCL
333-0816
Cliff Carter
195
DCL
333-3723
Mike Gardner
193
DCL
333-7904
Rex Duzan
162
DCL
333-6285
Don McCabe
1208
W Springfield
333-2171
333-7752
M103
Turner Hall
333-8170
153
Noyes Lab
333-1728
70
Comm West
333-4500
129
DCL
333-6203
146
EEB
333-4936
FAR
333-2695
ISR
333-0307
65
MEB
333-1430
453
Psych Bldg.
333-7531
120
Snack Bar
333-1851
202
Lincoln Hall
333-0309
OFF-LINE is the monthly newsletter of the Computing Services Office at the
University of Illinois at Urbana-Champaign. Unless otherwise indicated, permis-
sion to reprint is freely granted, provided that the author, if named, and the Com-
puting Services Office (CSO) are credited. Information in this issue is current as
of October 24, 1980.
CSO operates an IBM 4341 with four million bytes of fast core running HASP-
OS/MVT under VM, a CYBER 175 with 256K words of central memory and 512K
words of ECS, under NOS, serving up to 190 simultaneously active text and
graphics terminals, and a DEC PDP-11/50 with 252K bytes of core driving a GSI
CAT-8 phototypesetter with the UNIX Operating System.
POLICY
EXTENSION OF COMPUTING SERVICE
The service plans for CSO have recently been revised to include a major
expansion of CYBER facilities. A request for approval of the acquisition of a
CYBER 174 has been submitted to the Board of Trustees, and if approved at
their November 20, 1980 meeting, the new CYBER 174 will be installed during
the semester break.
Presented below is a statement to the Computer Policy Committee which
served as the basis of their discussion and approval of this plan.
More details regarding the use planned for the system will be in the next issues
of OFF-LINE.
In the fall of 1976 CSO installed the CDC CYBER 175. The selection of this machine was
the result of a very extensive competition between vendors, including benchmarking and a
competitive financial bid. In concluding the final details of the acquisition it was necessary to
make a balance between the financial life of the machine and the capacity of the machine to
meet the demands of the campus. Since unlimited funds were not available, a financial plan
in terms of the life of the machine was a necessity and had to be followed up by making the
allocation system support that financial life in a reasonable fashion.
The balance that was achieved was a combination of a lease and purchase agreement with
purchase payments extending over seven years. At the end of that seven-year period, the
majority of the system would be owned with no outstanding debt.
The demand for computing service, through the allocations systems of Instructional and
Research committees and through the contract support, has continued to grow at least as
rapidly as the computing systems would allow. In addition to those needs which have been
well met by the current equipment, there are at least two categories of need which still have
not been met in a satisfactory fashion. The availability of low cost, large scale computation
has been partially met through the bulk system. However, there are increasing numbers of
faculty and graduate students whose problems have reached substantial size where the
availability of computing is a major determinant in the speed with which they progress.
Second, the introduction to computing for students in lower level courses is carried out in
what is now an obsolete fashion - through the use of punch cards and printed output rather
than interacting with computers through a terminal.
CSO is now proposing a modification of the seven-year plan consisting of the addition of
another machine and the extension of the plan to a total of nine years. To treat the time
period first, our initial plan called for the equipment to be paid off and the financial life of the
machine to terminate in the middle of 1983. At that point the majority of the funds currently
going into the CYBER system would become available for a successor system. The ability to
continue operating the CYBER while utilizing the successor system would, of course,
consume part of the funds during the early years, but this would be a relatively low share
because the mortgage payments would have been completed. The new plan would leave us in
essentially the same position in 1985 with ownership of all present computers, the 4341 IBM
system, the CYBER 175 and the proposed system the CYBER 174. The equipment being
proposed is discussed in more detail later.
Over the past year and a half CSO has been looking at alternative means of supporting
elementary teaching of computing and the general increase of computer literacy on campus.
There are, through the Computer Science service courses, approximately 2,000 students a
semester involved in introductory level computing. In addition, there are many courses given
by individual departments for their students introducing the use of the computer to relatively
large numbers. Perhaps another one to two thousand students are involved in this. The
majority of these courses have used punched cards as their input, paper as their output, and
have done their processing using the EXPRESS system on the 360.
There have been two major complaints voiced with increasing frequency concerning this
manner of teaching. The first is simply that the approach to the computer through cards is
obsolete and does not represent a modern intoduction to computing capability which the
student should be exposed to. The second is, perhaps, the more serious and comes from
those who teach advanced courses which use the computer. The complaint is that what was
learned in the 100 level courses does not provide an introduction to computing that is useful
in the advanced courses, since the students in advanced courses are almost completely on
terminal-based systems. The difference between using an editing system with files and using
punched cards must be bridged by taking up class time in each course where the computer is
used. Since this is frequently not well taught by the faculty in the advanced courses it
represents a substantial consulting and support load as well as a burden on the instructors.
We have looked at a number of alternatives for meeting both of these complaints. Most of
these alternatives address one but not both. For example the use of small computers such as
a PDP-11 to edit programs for submission to a host has been widely criticized here because
the editing system and the access to the computer is not similar to that which would be used
in advanced courses. This is despite the fact that this approach has been extremely successful
at many other places and that it is a relatively cost-effective approach.
The use of the present equipment is an attractive alternative except that the addition of
several thousand students would seriously interfere with present commitments to research
programs and advanced instruction. Although the machine is very fast at computation,
adding many simultaneous users would be a serious problem. The strong desire which has
been expressed has been for continuity from introduction through advanced courses and this
has been a major determinant in our seeking out a solution which has been presented here.
The equipment being proposed is another member of the CYBER 170 family, the model 174.
It is exactly the same vintage as the 175 and is capable of running the identical software. The
significant elements of the configuration we are considering, the processor and memory, are
compatible with the present system, both in hardware characteristics and in operational
potential. The 174 consists of two central processors, each approximately 25% to 30% of the
speed of the 175. Thus for straight computation the equipment is approximately half of the
175. The memory is 196,000 words which allows the largest jobs being run on the 175 to also
be run on the 174. Thus no artificial distinctions between users, based on problem
characteristics, would be necessary. The operating system and the file system including the
disk hardware would also be compatible with that presently in use.
It is likely that rather than attempt to absolutely synchronize the two machines' software, one
machine might receive software upgrades earlier than the other to allow testing before all
services were committed to new releases of software.
Given the systems we have looked at and a number of ways in which they could be operated,
two are receiving final discussion. The method of operation has to be looked at in terms of
how the machines are interconnected and how the clienteles are divided between them. The
two reasonable ways of running the machines are: each machine operating in a standalone
fashion, or both machines fully integrated with all peripherals, such as disks, being available
to both machines. For reasons of practicality we would probably start with the systems
running as completely separate and gradually integrate them until they were fully shared. This
integration might take as long as a year and a half to be completed, since we don't want to
disturb the quality of the service on the old machine in order to make these changes.
The division of clientele which makes most sense to us is to devote one machine largely or
exclusively to support of instruction while devoting the other to graduate student and faculty
research. As an operational definition instructional computing could be considered that which
is supported with instructional allocations while research computing is that which is supported
by the Research Board, and Grants and Contracts. Obviously when the students are not on
campus the machine would be useful for other purposes and we would make arrangements so
that other work could be carried out on it.
Given this kind of split and some surveys we have done of other installations running a
CYBER 174, we feel that the new system could support approximately 100 terminals for
introductory level work plus 100 terminals for the advanced student work. This would
represent a substantial increase over the present load of advanced students and would solve
the problem of the introductory level students. In addition to the use of this machine for
introductory courses those students who are going to remain on the IBM service throughout
their academic career would be given their introduction using terminals on the IBM 4341.
Some upgrading of the IBM facility will probably occur to support this, and is within the five-
year financial plan.
At the present time, when the academic program is in full swing, the distribution of terminal
time is about half research and half instruction. This is the highest fraction of instructional
computing that goes on so it should be recognized that a substantial load will be moved off
the 175 during that time. Our experience has been that the 175 research load does not go
down substantially when school is not in session so a more uniform appearance of the service
and capacity throughout the year would be presented to the faculty.
Given the proposed division of facilities between Research and Instruction, the highly
seasonal work load presented by students leaves large gaps in demand. These can readily be
converted into a combination of expanded batch services and bulk computing. In terms of
computing, we should easily be able to obtain a doubling of the present allocation in terms of
real compute power.
In addition to the increased capacity presented by working countercyclical^ with the student
load, the removal of students from the 175 presents substantial room for growth by the
research community. It probably does not represent an opportunity to get further bulk
service on the 175 simply because the time of day loading has already been taken into account
in researching the present level.
In the longer term, if the machines are closely linked, a reasonable way of operation in times
of relatively light demand is to present all of the time-sharing services through the CYBER
while retaining the CYBER 175 as a batch machine. The ability to go back and forth between
the various modes of operation, utilizing identical software and with files held in common
between the two machines, gives us the opportunity to exploit the capacity of the systems
effectively.
Other Alternatives
It is always difficult to speculate on the timing of the availability of new machines with
substantially improved cost-effectiveness. It is widely believed, however, that within the next
year, IBM will announce their next generation of equipment which has gone under the name
H-series. If this date is accurate then we would expect an additional one to two years before
the earliest possible installation of that newer equipment and the associated peripherals
necessary to take advantage of it. This would of course, fit very nicely with the 1983 time if
we were interested in having one of the earliest deliveries which we could achieve.
Once IBM has announced their new generation of equipment, Control Data will, of course,
respond with theirs which could be expected to have the relative performance and price that
defines CDC's place in the current market. Again, we would be talking about a 1983 to 1984
installation of a machine the size of the 175 or larger.
Based on these dates our extension of the financial plan to 1985 means that we would not get
the earliest possible delivery of new equipment but rather we would get the equipment after it
had been in the field and through the intial testing. It is quite likely that any large successor
machine would come from either the IBM compatible or CDC families and that we would
overlap whichever family we had not chosen from.
The question of large machines vs multiple small machines continues to heat up but the
answers are far from becoming clear. The extension of an additional one or two years on our
present equipment gives more time for those debates to present better alternatives or clearer
guidance. CSO is actively looking at the networking requirements for use of small machines,
and the potential use of small systems in courses.
The issues of continuity, both for the researcher over a period of years and for the student
through a sequence of courses, is viewed with considerable concern on the campus because of
the very heavy investment in learning and program development. The additional continuity
offered by this expansion has been one of its best received features. This plan, of course,
represents a very strong commitment to continuity for at least another five years and that is a
comforting thought to many people.
Summary
What is proposed above is an extension of a general pattern of services utiltizing IBM and
CDC hardware and software systems, presenting the majority of our services through
terminals with batch processing of work prepared at terminals accounting for much of the
larger scale work. The selection of an additional CYBER machine represents an extension of
at least two years in the serviceability of the CDC equipment and an expansion in the amount
of terminal hours which can be presented. It represents a substantial upgrade in the amount
of non-prime BATCH service which can be presented to the research community.
The plan is financially sound in that it provides for the University to own all of its equipment
in 1985 with no outstanding debt and to have the full cash flow associated with the old
systems available for selection of either a large scale replacement or a collection of smaller
computers to meet the next generation needs. The only cost which is traded off against this
immediate expansion of service is a deferral for two years of the options of making wholesale
replacements based on the seven-year life of the CYBER 175.
After discussion with various committees and members of the faculty and after studying the
needs for student computing which have been presented over the last year, we feel that this is
a good plan and should be undertaken as quickly as possible. We request the support of the
Computer Policy Committee in setting this direction.
CSO CONSULTING SERVICES
CSO provides consulting help to general users, users of statistical packages, and users of text
processing facilities. Since these services are both vital and scarce, we want to tell you what
consulting help is available and how to use it efficiently.
The Systems Consultants, located in 166 DCL (333-6133), can help you with general system
questions. They can help you select the proper documentation, identify or analyze system or
program malfunctions, recommend appropriate programs and software packages, and help you
interpret diagnostic messages. The Systems Consultants will not provide individualized
instruction on the use of the computer, write or design programs for users, or provide
lengthy diagnoses of program logic errors. In the latter case you will be given suggestions as
to how to diagnose the error yourself.
The Statistical Consultants, located at 65 Commerce West (333-2170), will answer general
systems questions, but specialize in helping users with the statistical packages available on
CSO's computers. They will help you select the appropriate statistical package and supporting
documentation, provide advice on how to implement a model and prepare data for input, and
help you interpret output. Consulting assistance usually does not include help with choosing
a model, selecting the appropriate statistical technique, writing or running programs, or
interpreting the statistical meaning of results.
The Text Processing Consultant, located at 207 Astronomy (333-7318), will help you use the
text processing services available through CSO. He will explain the services available to you,
answer questions about their use, and help you solve text processing problems you encounter.
Bring your problem to a consultant only after first trying to solve it yourself using the
documentation known to you. Because there are heavy demands for consulting help, you
should do your part to minimize the amount of time the consultant must spend on your
problem. This includes, first and foremost, bringing a complete and current set of supporting
output: source listings, program output demonstrating the error, load maps, dumps, etc. All
output must be current. A source listing which has been manually updated for recent
program changes, for instance is worthless. Explain your problem briefly and concisely and
then allow the consultant to approach the problem in his or her own way. There is little to be
gained by the consultant repeating your own diagnosis.
Realize that consultants are human and not extensions of the system which has frustrated
you. Common courtesy will result in the best service. Don't expect instant solutions to
complex problems or be put off if you are referred to another consultant more knowledgeable
in your problem area. Graciously accept a request to make an additional run to produce
additional diagnostic output. The consultant is only asking you to do what he/she would do if
the problem were the consultant's own.
The Systems Consulting and Statistical Consulting offices are open 9AM - 5PM Monday
through Friday. Text Processing Consulting is available 9AM - Noon, 1PM - 4PM, Monday
through Friday. Since there is only one text processing consultant you should call first to
make sure that he is free.
VACATION SCHEDULES
The CYBER 175, IBM 4341 and UNIX will be available throughout the scheduled
Thanksgiving vacation, finals week and Christmas vacation. However, during these periods,
offices and RJE sites will operate according to the following schedules. Any additional
changes to the schedules will be posted on the doors of the RJE sites, and published in
HEARYE.
THANKSGIVING VACATION
CSO Departmental and Consulting Offices
Thursday Nov 27 - Sunday Nov 30 CLOSED
Monday Dec 1 Resume regular hours
ISR, FAR, SB (Snack Bar)
Wednesday Nov 26 - Sunday Nov 30 CLOSED
Monday Dec 1 Resume regular hours
CSO South - COM (70 Comm West)
Wednesday Nov 26 Close at 4:00 PM
Thursday Nov 27 - Saturday Nov 29 CLOSED
Sunday Nov 30 Resume regular hours
Agriculture (M-103 Turner Hall), Chemistry (153 Noyes Lab), Electrical Engineering
(146 EE), Mechanical Engineering (65 ME), Psychology (453 Psych),
Social Sciences (202 Lincoln Hall)
Wednesday Nov 26 Close at 5:00 PM
Thursday Nov 27 - Sunday Nov 30 CLOSED
Monday Dec 1 Resume at 8:00 AM
CSO North - LOCAL (129 DCL)
Thursday Nov 27 CLOSED
Friday Nov 28 - Saturday Nov 29 8:00 AM - 5:00 PM
Sunday Nov 30 8:00 AM - 2:00 AM
Monday Dec 1 Resume regular hours
FINALS WEEK
ISR, FAR, SB (Snack Bar)
Saturday Dec 13 - Friday Dec 19 Open 12:00 noon to 4:00 PM
CSO South - COM (70 Coram West)
Saturday Dec 13 - Friday Dec 19 Open 8:00 AM to 8:00 PM
CHRISTMAS VACATION
CSO Departmental and Consulting Offices
Wednesday Dec 24 Close at 12 noon
Thursday Dec 25 - Sunday Dec 28 CLOSED
Monday Dec 29 - Tuesday Dec 30 Resume regular hours
Wednesday Dec 3 1 - Sunday Jan 4 CLOSED
Monday Jan 5 Resume regular hours
ISR, FAR, SB (Snack Bar)
Friday Dec 19 Close at 4:00 PM
Saturday Dec 20 - Sunday Jan 18 CLOSED
Monday January 19 Resume regular hours
CSO South - COM (70 Comm West)
Tuesday Dec 23 Close at 5:00 PM
Wednesday Dec 24 - Wednesday Jan 7 CLOSED
Thursday Jan 8 Resume regular hours
Agriculture (M-103 Turner Hall)
Friday Dec 19 Close at 5:00 PM
Saturday Dec 20 - Sunday Jan 4 CLOSED
Monday Jan 5 Resume regular hours
Chemistry (153 Noyes Lab), Electrical Engineering (146 EE), Mechanical
Engineering (65 ME), Psychology (453 Psych), Social Sciences (202Lincoln Hall)
Wednesday Dec 24 Close at 12 noon
Thursday Dec 25 - Sunday Jan 4 CLOSED
Monday Jan 5 Resume regular hours
CSO North - LOCAL (129 DCL)
Wednesday Dec 24 Close at 4:00 PM
Thursday Dec 25 CLOSED
Friday Dec 26 - Wednesday Dec 31 8:00 AM - 5:00 PM
Thursday Jan 1 CLOSED
Friday Jan 2 - Sunday Jan 4 8:00 AM - 5:00 PM
Monday Jan 5 Resume regular hours
SYSTEM NOTES
SEPTEMBER RELIABILITY REPORT
CYBER 175: 18 recoverable interruptions
13 non-recoverable interruptions
8 of these were for more than 15 minutes
Mean Time Between Failures = 30 hours
Mean Time to Repair = 42 minutes
Availability: 97.4% of scheduled uptime
The major downtime of nine hours occurred
on September 4 for handling disk problems.
We have been running since then with no
further hardware trouble on the new disk
drives. It is hoped that problems related
to the installation of new equipment are
finally over.
IBM 4341: 13 interruptions
8 of these were for more than 15 minutes
Mean Time Between Failures = 51.9 hours
Mean Time to Repair = 40 minutes
Availability: 96.2% of scheduled uptime
Major cause of downtime was related to
diagnostic error and software problems.
"Recoverable interruptions" in time-sharing are those interruptions from which a person may
normally recover work being done at the time of the interruption by doing a RECOVER,/^
when the system comes back up. "Non-recoverable interruptions" means that a major
problem has occurred, and recovery of work in progress at the time of the interruption is not
possible.
IBM
IBM SSP TO GO
Because it is extremely outdated, the IBM SSP library will not be automatically available after
the 1980 Fall Semester - specifically, it will not be available after January 5, 1981. After that
date, it will only be available for those who still may have a need to use it by specifying:
UBFILE='SYS1.F0RTSSP'
For example:
// EXEC FORTLDGO,LIBFILE=,SYSl.FORTSSP•
If you are using SSP, we ask that you convert to an equivalent IMSL or HARWELL routine.
See Stan Kerr in Room 175 DCL (333-4715), or use TELL,UN=MATHLIB if you have
questions. It is likely that SSP will eventually be removed entirely.
MATHEMATICAL SERVICES
Q-GERT: SIMULATION LANGUAGE FOR MODELERS
Q-GERT is a network approach to modeling procedural systems. The Q-GERT analysis
program has been designed and built to simulate Q-GERT networks. Currently, the only
documentation available is the book, Network Modeling and Analysis Using Q-GERT, by A.
Alan B. Pritsker. This book is available for inspection at the Systems Consulting Office,
Room 166 DCL.
QGERT is accessed by entering the statement:
GRAB.QGERT.
This places a binary file, QGERT, in your local file space.
10
This file can then be run in one of two ways:
1. If your network is described entirely by a QGERT data deck, enter
QGERT, input.output.
Where input represents a file containing the data deck, and output represents the file
that will receive the simulation report. To PRINT this output file, you must use the
/CC option on the PRINT command. Neither the input nor the output file is rewound
by QGERT.
2. If you must include auxiliary FORTRAN routines, enter:
FTN,l=sute,...
LOAD.LGO.
QGERT.input.output.
Where subs is the file containing the routines. "..." represents other parameters you
may want to use with FTN such as REW, L=0, etc.
A number of sample QGERT problems, taken from the text of the book mentioned above,
are available from the procedure SAMPLES. For information on SAMPLES, see Reference
Guide RF-4.13, or obtain the on-line information by entering the command HELP and typing
SAMPLES in response to the question mark prompt.
Questions about QGERT should be addressed to Stan Kerr (333-4715) or Manoochehr
Ghiassi (333-3938).
MISCELLANEOUS
NEW CSO DOCUMENTATION AVAILABLE
A CSO manual called An Index to Software on the CDC CYBER 175 is now available free
from the CSO Distribution Center, 1208 West Springfield. The index contains brief
descriptions, support levels (maintenance and consulting), and a list of available
documentation for each software package supported by CSO on the CYBER. An index of
unsupported software is in progress.
A document called the List of Mini- and Micro-Computers in the Local Area is also available to
interested persons at the Distribution Center. This document contains the listing of those
persons (and their respective computers) who returned the DEC and CSO survey sheets
published in OFF-LINE. An on-line copy of the list is available by entering:
WRITEUP.SURVEY.
11
This will place the on-line document CMPLIST in your local file space to be viewed or
printed.
Hopefully, by the time this issue of OFF-LINE is on the stands, the new ZETA manuals will
have arrived. CSO is currently writing several manuals to help users convert from CalComp
to ZETA. Announcements will appear in HEARYE when these manuals are ready.
DEC USERS GROUP MEETINGS
The DEC Users Group meetings scheduled for November and December are:
November 1 1 Cliff Carter will speak on what CSO has
learned about various terminals on the market.
December 9 Walter Schneider will speak on trouble
shooting on small machines.
The meetings will be held at 4 PM in the auditorium of the Coordinated Science Building,
located at the corner of Springfield and Goodwin in Urbana.
OFF-LINE'S Mailing List
If you wish to be placed on our mailing list for future issues of OFF-LINE, if you wish to be
removed from the list, or if you wish to enter an address correction, please complete and
return this page. (Current subscribers are kept on the mailing list until a specific request for
removal is received, or until a mailing is returned as undeliverable.)
Check one: □ New subscriber
D Removal request
□ Address correction
Name:
Address:
CAMPUS or Zip Code
(If address correction, give old address and zip code below.)
Comments:
RETURN TO. OFF-LINE
150 Digital Computer Laboratory
University of Illinois at Urbana-Champaign
Urbana, Illinois 61801
Computing
Services
Office
Off-Line
University of Illinois at Urbana-Champaign
VOL. 8, NO. 12 December 1980
EDITOR: Lynn Bilger
PHONE: (217) 333-6236
139 Astronomy Building
Urbana, Illinois 61801
Page
Contents
1
2
3
4
5
8
8
11
POLICY
CYBER Files - Daily Backup Procedure
Time-Sharing Weekend Rates Reduced
SYSTEM NOTES
October Reliability Report
CYBER
New Diablo Services
Please Convert to the ZETA Plotter
FORTRAN Version 5
STATISTICAL SERVICES
New FOSOL Procedures
New SOUPAC Procedure
MATHEMATICAL SERVICES
MINPACK on the CYBER
DOCUMENTATION
New Section Added to OFF-LINE
OFTHf
DEC 16 1980
VSV^eSiTY OF ILLINOIS
CSO DIRECTORY - STAFF AND SERVICES
Administrative
Director
Business Manager
Secretary
George Badger
Stanley Rankin
Joyce McCabe
150 DCL
150 DCL
150 DCL
333-4103
333-6530
333-1637
User Services
User Accounting
Distribution Center
Systems Consulting
Statistical Services Consulting
Text Processing Consulting
Asst Dir User Services
Mgr, Statistical Services
Mgr, UNIX Operations
Documentation
Gary Bouck
Robert Penka
Larry Sautter
Larry Sautter
Lynn Bilger
1208 W Springfield
1208 W Springfield
166 DCL
65 Comm West
207 Astronomy
173 DCL
91 Comm West
139 Astronomy
333-7752
333-6760
333-6133
333-2170
333-7318
333-4709
333-2170
333-6236
Hardware/Software Support
Terminal Repair Service
CYBER Dial-up Numbers
Asst Dir Systems and Operations
Asst Dir Development
Asst Dir Engineering
CYBER-IBM Operations
UNIX Operations
Telecommunications
Hardware Selection Assistance
Laboratory Support Project
RJE Operations North
RJE Operations South
RJE Sites
Agriculture
Chemistry
Commerce West
CRH Snack Bar
DCL Routing Room
Electrical Engineering
Florida Ave Res Hall
Illinois St Res Hall
Mechanical Engineering
Psychology
Social Science
150 DCL
110 baud
300 baud
1200 baud
Sandra Moy
J. M. Randal
Cliff Carter
Jack Knott
Debbie Weller
Tom Kerkering
Cliff Carter
Mike Gardner
Rex Duzan
Don McCabe
177
DCL
120
DCL
195
DCL
194a
DCL
203
Astronomy
164
DCL
195
DCL
164
DCL
162
DCL
1208
W Springfield
-1103
Turner Hall
153
Noyes Lab
70
Comm West
120
Snack Bar
129
DCL
146
EEB
FAR
ISR
65
MEB
453
Psych Bldg.
202
Lincoln Hall
333-0969
333-4000
333-4000
333-4001
333-4703
333-9772
333-3723
333-6562
333-8150
333-0816
333-3723
333-7904
333-6285
333-2171
333-7752
333-8170
333-1728
333-4500
333-1851
333-6203
333-4936
333-2695
333-0307
333-1430
333-7531
333-0309
OFF-LINE is the monthly newsletter of the Computing Services Office at the
University of Illinois at Urbana-Champaign. Unless otherwise indicated, permis-
sion to reprint is freely granted, provided that the author, if named, and the Com-
puting Services Office (CSO) are credited. Information in this issue is current as
of November 24, 1980.
CSO operates an IBM 4341 with four million bytes of fast core running HASP-
OS/MVT under VM, a CYBER 175 with 256K words of central memory and 512K
words of ECS, under NOS, serving up to 190 simultaneously active text and
graphics terminals, and a DEC PDP-11/50 with 252K bytes of core driving a GSI
CAT-8 phototypesetter with the UNIX Operating System.
POLICY
CYBER FILES - DAILY BACKUP PROCEDURE
The procedure for providing daily backup of CYBER files has been changed, primarily to
improve the time it takes to restore a destroyed disk pack. Keep in mind while reading about
the new procedure that all backups are done after midnight.
The previous method was to backup daily on tape all files modified within the last 24-hour
period. This was done every day of the week except on Sunday when a full backup of all files
was done. Each of these daily tapes were kept for that week and then reused the next week.
The new method still provides for the full backup of all files on Sunday. However, the daily
backup method has changed. Each day of the week, all files that are on disk and have been
modified since Sunday (the last full backup) are dumped to tape. This implies that a full disk
pack restore only requires the processing of two sets of tapes; a full backup tape and the most
recent incremental tape of files modified since Sunday that are still on disk. Under the old
scheme, the worst case required up to seven sets of tapes.
There is a significant difference in how daily backup tape sets are used. Under the old
scheme, seven tape sets (one for each day) were required; under the new scheme, three tape
sets are required and are used as follows:
Daily
Tape Set Used on
1 Monday and Thursday
2 Tuesday and Friday
3 Wednesday and Saturday
The significance of this is that should you create a file on Monday, it will be backed up on the
Tuesday (early morning) daily tape. If you then accidentally purge the file on Tuesday, it will
not appear on the Wednesday or Thursday incremental tapes. If you then wait until Friday to
come to the Consultants to restore your purged tape, it will be too late because the Tuesday
tape will have already been written over by the early morning Friday tape.
NOTE: This means that you no longer have a full week in which to get
a file restored; you now have only two days!
This approach improves our recovery time in the event of a disk problem, and reduces the
CSO tape storage needs. For example, a worst-case (Saturday) disk failure results in a reload
of four 6250 tapes in contrast to the 50-60 1600 BPI tapes previously required; thus, reducing
the chances of tape problems and considerably simplifying the restore procedure, as well as
requiring less time to spin through all necessary tapes.
TIME-SHARING WEEKEND RATES REDUCED
Starting on Friday, November 21, CSO began charging reduced rates for weekend time-
sharing use on the CYBER. The period of reduced rates begins each Friday afternoon at 4:00
PM and ends the following Monday morning at approximately 6:00 AM when the system
goes down for scheduled engineering.
Time-sharing jobs which both begin and end during the low-cost period are billed at 60% of
the full rate. Session costs reported at logoff do not reflect the reduced rate; the rate
reduction is applied when billing information is processed at the end of the day.
IBM jobs and CYBER batch jobs are billed at the regular rate. Time-sharing jobs which logon
before the period of reduced rates begins and logoff during the period of reduced rates are
also billed at the regular rate.
SYSTEM NOTES
OCTOBER RELIABILITY REPORT
CYBER 175: 21 recoverable interruptions
12 non-recoverable interruptions
22 of these were for more than 15 minutes
Mean Time Between Failures = 16 hours
Mean Time to Repair = 31 minutes
Availability: 95.3% of scheduled uptime
Major cause of downtime was related to
CPU failures.
IBM 4341: 16 interruptions
1 1 of these were for more than 15 minutes
Mean Time Between Failures = 44 hours
Mean Time to Repair = 33 minutes
Availability: 95.7% of scheduled uptime
Major cause of downtime was related to
disk and other hardware problems.
"Recoverable interruptions" in time-sharing are those interruptions from which a person may
normally recover work being done at the time of the interruption by doing a RECOVER, tty
when the system comes back up. "Non-recoverable interruptions" means that a major
problem has occurred, and recovery of work in progress at the time of the interruption is not
possible.
CYBER
NEW DIABLO SERVICES
Effective December 1, 1980, a new Diablo service will be offered on the CYBER. The new
service will be similar to the queued service now being offered with a few changes.
Users of the Diablo service will still run a procedure on the CYBER. However, the new
procedure will prompt you for your surname and whether or not you wish to charge the
printing to a UOI account number. Costs for printing can only be charged to "real money"
UOI accounts; class accounts or research board accounts are not allowed.
If you choose to pay by UOI account number, you will be prompted for the account number
and title at the time the procedure is running. If you choose to pay by check, you will not be
prompted for an account number or title. You will also be prompted for a choice of the pica
or elite font and you will be asked for the name of your local file containing your RNF
output. NOTE: The procedure will not perform a get if it cannot find a local file by the name
given.
Finally, at the conclusion of the procedure, you will be told the number of PRUs (one PRU
equals 640 upper case characters, 320 lower case characters) and the dollar cost associated
with its printing. The new charge is $0.05 per PRU and is comparable to the old charge of
$0. 1 5 per page. All charges continue to be rounded up to the nearest dollar.
You will be sent a message on the CYBER once your file has been printed on the Diablo.
You may pick up your printout at DCL. If you chose to pay by check, you will make out a
check at that time to the University of Illinois for the amount indicated on the billing sheet
you receive with your printout. If you chose to pay by UOI account number, your account
will automatically be billed for the cost indicated on the billing sheet.
It should be pointed out that once something has been submitted to the queue it will be
printed and you will be charged! Please take every precaution to print your RNF output on
an upper and lower case line printer or DECwriter prior to submission to the Diablo queue.
Refunds will not be given for printouts submitted by mistake.
If you have any questions concerning this service, please call the operator at 333-8150.
PLEASE CONVERT TO THE ZETA PLOTTER
The Calcomp plotter has become increasingly unreliable and hardware support is
unpredictable. The Calcomp company has dropped support of the plotter as part of a
reorganization of their product line. As a result, availability of repair parts is not guaranteed
and Calcomp engineers are under no obligation to respond to requests for repairs.
CSO has installed three replacement plotters manufactured by the Zeta company. These
plotters have been in service on the CYBER for several months, and IBM support has
recently been completed. We are asking all users of the Calcomp plotter to convert as quickly
as possible to the ZETA plotters. CSO plans to remove the Calcomp plotter from service at
the end of this semester (December 21, 1980).
CYBER Users
If you generate your plots on the CYBER, you can convert to the ZETA plotters by simply
using different control statements as follows:
Present Control Card Replacement Control Card
GRAB.CALCOMP. GRAB.ZETA.
GRAB.GCSCALC. GRAB.GCSZETA.
GRAB.GCSCALC/F. GRAB.GCSZETA/F.
PLOT, filename. PLOTZ.filename.
New parameters on the PLOTZ command eliminate the "special handling" statement; other
parameters have a different format from those on the PLOT command. Parameters accepted
by PLOTZ are documented in Reference Guide RF-7.31.
IBM Users
If you have been using the IBM to generate your plot, conversion may be as simple as adding
two subroutine calls to your program and changing your JCL, or it may require more
extensive code changes. Approximately half of the subroutines available in the IBM Calcomp
library are available in identical form in the new ZETA library; a fourth are available in
slightly altered form which will require minor source changes; the remaining fourth have no
ZETA equivalents. A conversion document, IBM Calcomp to ZETA Conversion, is available at
the Distribution Center, 1208 W. Springfield (or from the Routing Room personnel or the
Systems Consultants).
If you cannot convert your programs to the ZETA by the planned date (December 21) for the
removal of the Calcomp plotter, please contact Bob Penka, 173 DCL (333-4709).
FORTRAN VERSION 5
Earlier this year, CDC introduced a new FORTRAN compiler, FORTRAN Version 5. At
that time CDC indicated that it would support both the new compiler and the current
compiler, FORTRAN Extended Version 4, for approximately two years and then drop
support for Version 4. Although CSO is not yet running a version of the operating system
under which this product is officially supported by CDC, we have adapted FORTRAN
Version 5 to our current operating system in order to offer our users a chance to acquaint
themselves with the new compiler and avail themselves of its added capabilities. The new
compiler is accessed via the FTN5 control statement as in,
FTN5.I = source,l= listing.
FTN5 was designed to conform to the new ANSI FORTRAN Standard (X3.9-1978).
Although most of the new features in this standard have been implemented in some existing
FORTRANs, many of them will be new to CYBER users. Among the new features are the
following:
• Many of the syntactic rules have been liberalized to allow expressions where only
more restricted forms such as simple variables or constants were allowed before. For
example, constant expressions (i.e., expressions not involving variables) may be used
in array dimension specifications and expressions may be used to control DO loops.
• The PARAMETER statement allows one to define symbolic constants which can then
be used in most places that explicit constants can be used. For example,
PARAMETER (NEQS=10,NRHS=2,TWOPI=2*3.14159)
REAL A(NEQS,NEQS),B(NEQS,NRHS)
INTEGER IWRK((NEQS*(NEQS+l))/2)
Y=SIN(TWOPI*X)
DO loops may now be controlled by REAL variables and expressions as well as
INTEGERS. If the increment expression has a negative value, DO loops will run
backwards. If the terminal condition of a DO loop is satisfied by the initial value of
the control variable, the body of the loop is not executed. (Such "zero-trip" loops are
in contrast to the "one-trip" loops produced by the current compiler, in which the loop
body is always executed at least once. FTN5 offers an option to compile "one-trip"
loops if they are required for program compatibility.)
Array subscripts are no longer required to start at 1. For example,
REAL ARRAY(0:99)
. Most of the intrinsic functions are now generic, i.e., they may be applied to arguments
of more than one type. For example, SQRT can now be applied to either REAL or
DOUBLE PRECISION arguments, rather than requiring SQRT for the former and
DSQRT for the latter, as in the current compiler.
. A new CHARACTER data type allows characters to be stored without worrying about
the specific characteristics of the machine on which the program is running. For
example,
CHARACTER*80 CARD
New operations and intrinsic functions facilitate the manipulation of this new data
type.
. A new block-IF construct facilitates easier selection among blocks of code. For
example,
IF (ANSWER.EQ.'YES') THEN
IOPT=l
PRINT *, 'AFFIRMATIVE'
ELSE IF (ANSWER.EQ.'NO) THEN
IOPT=2
PRINT *, 'NEGATORY'
ELSE
PRINT \ 'I DIDN'T UNDERSTAND'
GO TO 1000
END IF
Now OPEN and CLOSE statements give the programmer greater control over the files
manipulated by a FORTRAN program and, indidentally, free the programmer from
the requirement of listing all files to be used by a program on the PROGRAM
statement. An INQUIRE statement is now available to determine the characteristics
of the files available to a FORTRAN program.
FORTRAN now supports a random access file capability similar to that offered by
DEFINE FILE on the IBM machine. (The old READMS/WRITMS routines are also
still available.)
FORMATs have been extended in several minor, but useful ways.
READ and WRITE may be used in conjunction with CHARACTER variables to
provide a capability similar to that offered by DECODE and ENCODE. (DECODE
and ENCODE are still available, but non-standard.)
. CHARACTER expressions may now be used as FORMATS. In particular, statements
such as
WRITE(IUNIT,,(1X,3U0)')I,J,K
are now allowed.
. END= and ERR= specifiers, similar to those already familiar to IBM FORTRAN
users, may be used in place of the EOF function to detect end of file. Alternatively,
the specifier IOSTAT= can be used to return the input/output status in a variable.
. The ENTRY statement now has its own parameter list, as in IBM FORTRAN.
. Alternate RETURNS are now supported in a manner similar to IBM FORTRAN.
. A new declaration, SAVE, can be used to indicate which variables in a subprogram are
to retain their values between executions. This new declaration helps assure that this
happens, even when the subprogram is part of a segment overlayed program.
More complete information on the language supported by FTN5 and the control statements
used to run it may be found in the FORTRAN Version 5 Reference Manual, which may be
found in the CSO manual racks or the Consulting Office, or which may be purchased at the
Distribution Center, 1208 W. Springfield.
Users should be aware that we are running the first official release of FTN5 and that it has
known deficiencies. A list of known problems is being maintained in the file
FTN5BUG/UN=CONSULT. This file may be examined via the command
TYPE,FTN5BUG/UN=CONSULT/ ASCII.
As we receive new versions of FTN5, we will investigate the possibility of adapting them to
our current operating system if they appear to have significantly fewer deficiencies.
Although FTN5 has been designed to offer the same capabilities present in FTN, in some
cases the syntax used to express these capabilities has changed. CDC has provided a
translation program, F45, to assist in the conversion of FTN source to FTN5 source. Access
to F45 is obtained through the GRAB command, as in,
GRAB.F45.
F45 may then be invoked by the command,
F45.I = source, L= listing,? ' = newsource.
More complete information on the conversions performed by F45 and its control statement
options may be found in the FORTRAN Extended Version 4 to FORTRAN Version 5 Conversion
Aid Program Version 1 Reference Manual, which may be found in the CSO manual racks or in
the Consulting Office. This manual has been ordered and should be available for purchase at
the Distribution Center within the next few weeks.
STATISTICAL SERVICES
NEW FOSOL PROCEDURES
Two new FOSOL procedures written in CYBER Control Language (CCL) have been installed
on the CYBER. They are intended to replace the KCL procedures previously available with
the GRAB,FOS and GRAB,MAN commands. The new CCL procedures have somewhat
different features than their predecessors. GRAB,FOS and GRAB,MAN will be removed
from the system on January 9, 1981!
To access FOSOL using the new procedure, use the following commands:
GRAB.FOSOL
F0S0L(l=fi/ei.L=fi/e2)
Where filel is the name of your local file containing the FOSOL problem information
(INPUT is the default) and file2 is the name of the local file to which your output will be
written (OUTPUT is the default).
The FOSOL Interactive Manual is accessed by entering
GRAB.FOSMAN.
F0SMAN(L=fi/e2)
Where file2 is the name of the local file to which output from the EXCERPT mode in the
FOSOL Interactive Manual is to be placed (OUTPUT is the default).
On-line help files are available for both of these procedures and may be accesssed by entering
the command, HELP, and then entering
FOSOL
or
FOSMAN.
after the question mark (?) prompt. If you have any questions concerning FOSOL, please
contact the CSO Statistical Services at 65 Commerce West (333-2170).
NEW SOUPAC PROCEDURE
A new SOUPAC procedure written in CYBER Control Language (CCL) has been installed on
the CYBER. It is intended to replace the KCL procedure previously available with the
GRAB,SOUP command. It is almost identical in operation. GRAB,SOUP will be removed
from the system on January 9, 1981!
To access SOUPAC using the new procedure, use the following commands:
GRAB.SOUPAC.
S0UPAC(l=fr/ei,L=fi/e2)
Where filel is the name of your local file containing the SOUPAC problem information
(INPUT is the default) and file2 the name of the local file to which your output will be
written (OUTPUT is the default).
An on-line help file is available and can be accessed by entering the command, HELP, and
then entering
SOUPAC.
after the question mark (?) prompt. If you have any questions concerning SOUPAC, please
contact the CSO Statistical Services at 65 Commerce West (333-2170).
MATHEMATICAL SERVICES
MINPACK ON THE CYBER
MINPACK, a library of nonlinear least-squares routines and nonlinear equation solvers
developed at Argonne National Laboratory, is now available on the CYBER via GRAB. A
test version of MINPACK was announced in the May issue of OFF-LINE
The library is accessed by entering the control statement,
GRAB.MINPACK.
Following this, you only need to compile and run a FORTRAN program which calls one or
more MINPACK routines.
Source and writeups of MINPACK routines are available via the MATH procedure (see
Reference Guide RF-4.12). A complete set of the writeups (93 pages) can be obtained and
printed by entering
WRITEUP.MINPACK.
PRINT.MINDOC/CC/EJ/RJE=LCCAL
The library consists of five basic routines, each paired with an "easy-to-use" version. There
are three least-squares routines and two equation solvers. One of the least-squares routines
and one of the equation solvers are derivative-free; you need not code the Jacobian matrix of
partial derivatives.
10
All of the least-squares routines use a modification of the Levenberg-Marquardt method to
solve the following problem:
Minimize the sum of the squares of m nonlinear functions in n variables.
The equation solvers handle the following problem:
Find a zero of a system of n nonlinear functions in n variables.
by using a modified form of the Powell hybrid method. For those routines which require you
to supply FORTRAN code for evaluating both the function and its Jacobian, a routine
CHKDER is supplied to help you check the consistency of your code before using the other
MINPACK routines.
Following is a list of the routine pairs (names ending in 1 (one) are the "easy-to-use"
versions) :
LMDER/LMDER1
LMSTR/LMSTR1
LMDIF/LMDIF1
HYBRJ/HYBRJ1
HYBRD/HYBRD1
A least-squares routine. You must supply a routine which
evaluates both the functions and the Jacobian (the matrix of
partial derivatives of the functions with respect to the unknown
variables). The memory required is on the order of m*n real
numbers, which may be uncomfortably large in a curve-fitting
problem where m would be the number of observations.
A least-squares routine; a "space-saving" version. You must
supply a routine which evaluates both the function and a
specified row of the Jacobian. The memory required is on the
order of n2 + 2m real numbers; since n is usually small, this is
"reasonable".
A least-squares routine; the derivative-free version. You must
supply a routine to evaluate the function; the Jacobian is
estimated by MINPACK using forward differences. The
memory required, however, is on the order of m*n real
numbers.
An equation solver. You must supply a routine to evaluate the
functions and to evaluate the Jacobian. The memory required is
on the order of 1.5*«2 real numbers.
An equation solver. You must supply a routine to evaluate the
functions; the Jacobian is estimated by MINPACK using forward
differences. The memory is on the order of 1.5*«2 real numbers.
Questions about or problems with MINPACK should be directed to Mary Ann Berg, 121
Altgeld Hall (333-2168) who is the local Argonne contact for MINPACK, or to Stan Kerr,
175 DCL (333-4715 or TELL,UN=MATHLIB).
11
DOCUMENTATION
NEW SECTION ADDED TO OFF-LINE
We have added this new section, entitled DOCUMENTATION, to OFF-LINE to keep the
user community better informed about new or revised documentation.
Basically, it will be divided into four general categories: New Documentation, Revised
Documentation, On-line Documentation, and Reference Guides. Listings of new
documentation will include the price and an abstract. Listings of revised documentation will
specify whether the revision is major or minor, and will contain brief descriptions of the
revisions. The on-line section will list new manuals which have been added to the WRITEUP
facility and new HELP files. The listings for Reference Guides will follow the same format as
for the revised documentation.
We hope our readers will find this new section useful. Comments or suggestions should be
directed to Lynn Bilger, 139 Astronomy (333-6236).
New Documentation
Compatible Plotting Subroutines - FORTRAN Reference Manual $6.00
This manual, published by ZETA, describes vendor-supplied software for the
ZETA plotters. A locally-produced addendum accompanies this manual.
IBM Calcomp to ZETA Conversion Manual Free
This manual is directed toward IBM users of CSO's Calcomp plotter and
documents changes which must be made to convert to the ZETA plotters.
(The ZETA plotters are replacements for the soon to be discontinued Calcomp
plotter.)
FOR TRAN Version 5 Reference Manual $ 1 2.40
This CDC publication describes the FORTRAN language as defined by the
1977 ANSI Standard and the use of the CDC FTN5 compiler which implements
it.
On-line Documentation
On-line documentation has been added for MINPACK, a library of nonlinear least-squares
routines and nonlinear equation solvers from Argonne Labs. It may be accessed and printed
by entering
WRITEUP.MINPACK.
PRINT,MINDOC/CC/EJ/RJE=LOCAL
12
Several new help files have been added to the HELP facility. The new files are as follows:
FOSOL
FOSMAN
MESSAGE
SOUP AC
TELL
WHO
Reference Guides
With the addition of the following two Reference Guides, our reorganization of the Reference
Guides has been completed:
RF-0.6 Disk Policy
RF-2.5 LISP
OFF-LINE's Mailing List
If you wish to be placed on our mailing list for future issues of OFF-LINE, if you wish to be
removed from the list, or if you wish to enter an address correction, please complete and
return this page. (Current subscribers are kept on the mailing list until a specific request for
removal is received, or until a mailing is returned as undeliverable.)
Check one: □ New subscriber
□ Removal request
□ Address correction
Name:
Address:
CAMPUS or Zip Code
(If address correction, give old address and zip code below.)
Comments:
RETURN TO: OFF-LINE
1 50 Digital Computer Laboratory
University of Illinois at Urbana-Champaign
Urbana, Illinois 61801
o.— <->-o
"o r~oo
r- z.
■— -<
><
<
m
Computing
Services
Office
Off-Line
University of Illinois at Urbana-Champaign
VOL. 9, NO. 2 Jan-Feb 1981
EDITOR: Lynn Bilger
PHONE: (217) 333-6236
139 Astronomy Building
Urbana, Illinois 61801
Page
Contents
POLICY
5
6
7
9
10
12
13
15
16
16
17
17
18
19
21
#\£
Special Weekend Rates
Revised Rates for Plotting
New Method for Computing Disk Charges
CSO Short Courses - Spring 1981
SYSTEM NOTES
November and December Reliability Reports
CYBER
CYBER CANCEL Command Installed
CYBER 174 Installed
Gandalf Switch Installed
More Equipment Changes
New 6250 BPI Tape Drives
CSO Videotapes Available
Changes to Microcomputer Software in GRAB
MATHEMATICAL SERVICES
IMSL Newsletter
DOCUMENTATION
January-February Issues Combined
Revised CSO Documentation
MISCELLANEOUS
Graphics Presentation
Central Illinois Users Group Meetings (DEC)
HELP WANTED
Research Associate and/or Co-author
Wanted for Computer Work
SPECIAL TEAR-OFF SHEETS
Math Short Course Survey
Request for Information--Tektronix Graphic Terminals
UB^
Of
-rvtf
F£BQ*
•
CSO DIRECTORY - STAFF AND SERVICES
Administrative
Director
Business Manager
Secretary
User Accounting
Distribution Center
Systems Consulting
Statistical Services Consulting
Text Processing Consulting
Terminal Repair Service
CYBER Dial-up Numbers
Asst Dir User Services
Asst Dir Systems and Operations
Asst Dir Statistical Services
and UNIX System
Asst Dir Development
Asst Dir Engineering and
Hardware Selection
Documentation
CYBER-IBM Operations
UNIX Operations
Telecommunications
Laboratory Support Project
RJE Operations North
RJE Operations South
RJE Sites
Agriculture
Chemistry
Commerce West
CRH Snack Bar
DCL Routing Room
Electrical Engineering
Florida Ave Res Hall
Illinois St Res Hall
Mechanical Engineering
Psychology
Social Science
George Badger
150
DCL
333-4103
Stanley Rankin
150
DCL
333-6530
Joyce McCabe
150
DCL
333-1637
fare Support
1208
W Springfield
333-7752
1208
W Springfield
333-6760
166
DCL
333-6133
65
Comm West
333-2170
207
Astronomy
333-7318
150
DCL
333-0969
CYBER 175
110-300
baud
333-4000
CYBER 175
1200
baud
333-4001
CYBER 174
110-300
baud
333-4004
Robert Penka
173
DCL
333-4709
Sandra Moy
177
DCL
333-4703
Larry Sautter
91
Comm West
333-2170
J. M. Randal
120
DCL
333-9772
Cliff Carter
195
DCL
333-3723
Lynn Bilger
139
Astronomy
333-6236
Jack Knott
194a
DCL
333-6562
Debbie Weller
171
DCL
333-8150
Tom Kerkering
164
DCL
333-0816
Mike Gardner
164
DCL
333-7904
Rex Duzan
162
DCL
333-6285
Don McCabe
1208
W Springfield
333-2171
333-7752
M103
Turner Hall
333-8170
153
Noyes Lab
333-1728
70
Comm West
333-4500
120
Snack Bar
333-1851
129
DCL
333-6203
146
EEB
333-4936
FAR
333-2695
ISR
333-0307
65
MEB
333-1430
453
Psych Bldg.
333-7531
202
Lincoln Hall
333-0309
OFF-LINE is the monthly newsletter of the Computing Services Office at the University of
Illinois at Urbana-Champaign. Unless otherwise indicated, permission to reprint is freely
granted, provided that the author, if named, and the Computing Services Office (CSO) are
credited. Information in this issue is current as of January 20, 1981.
CSO operates a CDC CYBER 175 with 256K words of central memory and a CDC CYBER
174 with 131 K words of central memory. The 175 and 174 run under the NOS Operating
System and share 512K words of ECS. The 175 serves over 200 simultaneously active text
and graphics terminals and the 174 serves over 100 simultaneously active terminals. CSO
also operates an IBM 4341 with 4 million bytes of fast core running HASP-OS/MVT under
VM. In addition, CSO operates a DEC PDP-1 1/50 with 252K bytes of core driving a GSI
CAT-8 phototypesetter and a DEC VAX 11/780 with 2 million bytes of core, both running
under the UNIX Operating System.
POLICY
SPECIAL WEEKEND RATES
As of January 23, 1981, low-cost weekend rates were extended to include both batch and
time-sharing on the CYBER 174 and 175 systems. The period of reduced rates is to begin
each Friday at 4 PM and end each Monday at approximately 6 AM when the system is taken
down for engineering. These lower rates are restricted to accounts of internal University
users; they do not apply to bulk jobs.
To take advantage of these lower weekend rates you must both logon and logoff between the
specified hours. For example, if you logon at 4:15 PM and logoff at 5:30 PM on Friday, you
will be charged at the reduced rate. However, if you logon at 3:30 PM and logoff at 5:30 PM
Friday, you will be charged at the usual rate since you logged on before the 4 PM start of the
rate-reduction period.
During the rate-reduction period, batch and time-sharing jobs are charged at 60% of the usual
rate. The rate reduction applies to all charges except those for printing and punching. The
costs reported to you at the end of a batch job or a time-sharing session do not reflect the
reduced rates. The rate reduction is applied at the end of the day when billing information is
entered into the billing database, and the balances at the various accounting levels are
decremented by the cost of the job.
Any jobs submitted to be run during this period must finish execution by the 6 AM Monday
deadline. If these jobs are completed after the deadline, they will be charged at the regular
rates. Users who do not wish their jobs to run at the regular rates can use the CANCEL
command early Monday morning to delete (cancel) any job which has not yet completed
execution. See the article, CYBER CANCEL COMMAND INSTALLED, under the CYBER
section of this issue for further details.
Due to other pressing demands on manpower, no additional software support is being
provided for this service. If practical experience dictates that further software support is
required, the low-cost batch service will be withdrawn until such support can be provided.
A similar low-cost service is being planned for the IBM 4341. An announcement detailing its
use will be made when this service becomes available.
REVISED RATES FOR PLOTTING
The published rate for plotting has for some time been set at 20 Service Units per hour. Due
to overhead costs for collecting some plot billing data, charges were never issued for plots
generated on the CYBER or plotted on the off-line plotter. Now that the Calcomp plotter is
no longer in use and the ZETA plotters have been integrated into the CYBER system, usage
data is being automatically gathered for all plot jobs. Starting February 16, we will charge for
all plots on the basis of this data.
Rates charged for plotting have been reviewed and an attempt has been made to tie charges
directly to costs. We will, for instance, produce plots on the default plotter quite cheaply
since this plotter is run with a minimum of operator attention. No refunds will be given for
defects such as pen skips. On the other hand, plots produced on the special plotter will be
charged at a higher rate. This plotter is given close operator attention and plots produced on
it are of guaranteed quality. Similarly, a higher charge will be made for wide paper than for
narrow paper. A special charge will be made for use of liquid ink pens since they require
both operator preparation and cleanup.
The new plotting rates are based on SRUs (100 SRUs = 1 Service Unit) and are as follows:
Default Plotter (minutes) 750 SRUs per hour
Special Plotter (minutes) 1200 SRUs per hour
Wide Plotter (minutes) 2400 SRUs per hour
Narrow Paper (inches) 2 SRUs per inch
Wide Paper (inches) 3 SRUs per inch
Special Paper (inches) 7 SRUs per inch
Liquid Ink Setups 250 SRUs per pen
NEW METHOD FOR COMPUTING DISK CHARGES
We are now computing disk charges on a weekly rather than a daily basis. Billing programs
are run between 6 PM and midnight each Saturday night. Any file in existence when the
billing program is run is charged for a full week's residency. Conversely, no charge is issued
for any file or dataset which is purged before the billing program is run. This change was
made to reduce the excessive number of billing records resulting from the present method of
computing disk charges daily.
CSO SHORT COURSES - SPRING 1981
CSO is offering the following short courses for the CYBER 174 and 175 during the first half
of the 1981 spring semester. Additional courses (to be announced at a later date) will be
offered during the second half of the semester.
Registration is free, but limited to 30 people for some classes. To register for a course either
• come in person to Room 150 DCL, or
• phone 333-6630.
If all of the available classes on the topic you are interested in are full, leave your name on
our waiting list. If anyone cancels, or another section of the class is set up, we will contact
you immediately.
The short courses being offered are as follows:
INTRODUCTION TO THE CYBER (3 classes - 3 sessions each)
February 2,4,6 11-12 AM in Room 115 DCL
February 9, 11, 13 11-12 AM in Room 115 DCL
March 9,11,13 3-4 PM in Room 239 DCL
CYBER MAGNETIC TAPES (1 class - 3 sessions)
February 16, 18, 20 3-4 PM in Room 239 DCL
GRAPHICS AT CSO (1 class - 1 session)
February 5 12-1 PM in Room 115 DCL
CALCOMP TO ZETA CONVERSION (1 class - 1 session)
February 12 2-3 PM in Room 239 DCL
NCAR PLOT PACKAGE (1 class - 1 session)
February 19 2-3 PM in Room 239 DCL
GCS PLOT PACKAGE - INTRODUCTION (1 class - 6 sessions)
February 23, 25, 27 and 3-4 PM in Room 239 DCL
March 2, 4, 6
GCS PLOT PACKAGE - ADVANCED (1 class - 4 sessions)
February 23, 25, 27, 30 1 1-12 AM in Room 239 DCL
MATH LIBRARIES (1 class - 3 sessions)
March 9, 11, 13 11-12 AM in Room 239 DCL
INTRODUCTION TO THE CYBER FOR RNF USERS (2 classes - 3 sessions each)
February 16, 18, 20 11-12 AM in Room 239 DCL
March 16, 18, 20 3-4 PM in Room 239 DCL
INTRODUCTION TO RNF (2 classes - 3 sessions each)
February 23, 25, 27 1 1-12 AM in Room 239 DCL
March 23, 25, 27 3-4 PM in Room 239 DCL
NON-LINEAR PROGRAMMING AND NETWORK (1 class - 2 sessions)
February 24, 26
MANOVA (1 class - 4 sessions)
February 24, 26, and
March 3, 5
SAS (1 class - 3 sessions)
February 16, 18, 23
SPSS (1 class - 2 sessions)
February 9, 1 1
2-3 PM in Room 239 DCL
7-9 PM in Room 140 Com West
7-9 PM in Room 241 Com West
6:30-9:30 PM in Room 241 Com West
SYSTEM NOTES
NOVEMBER AND DECEMBER RELIABILITY REPORTS
CYBER 175
November
21 recoverable interruptions
24 non-recoverable interruptions
28 of these were for more than 1 5 minutes
Mean Time Between Failures = 12 hours
Mean Time to Repair = 33 minutes
Availability: 94.6% of scheduled uptime
Major cause of downtime was related to
CPU errors and many hardware problems.
The major problem was resolved by
December 7, 1980.
December
30 recoverable interruptions
52 non-recoverable interruptions
53 of these were for more than 15 minutes
Mean Time Between Failures = 6 hours
Mean Time to Repair = 46 minutes
Availability: 83.9% of scheduled uptime
Major cause of downtime was related to
CPU failures, power failures, generator
problems, and installation of the 174.
IBM 4341:
18 interruptions
11 of these were for more than 15 minutes
Mean Time Between Failures = 39.8 hours
Mean Time To Repair = 18.8 minutes
Availability: 98.8% of scheduled uptime
Major cause of downtime was related to
software and hardware problems.
19 interruptions
9 of these were for more than 1 5 minutes
Mean Time Between Failures = 38 hours
Mean Time To Repair = 44 minutes
Availability: 98.7% of scheduled uptime
Major cause of downtime was related to
power failure.
CYBER
CYBER CANCEL COMMAND INSTALLED
We have installed a new command, CANCEL, which allows a user to delete (cancel) his/her
job from the input, rollout, fetch, print, plot, or punch queue on the CYBER. The format of
the command is
CANCEUxxx.
where xxx is the FNT ordinal of the job. (The FNT ordinal is the number displayed to the
left of the jobname in the response to a QUERY command.)
You can cancel only your own jobs. This means that when canceling a job you must be
logged in under the same user number which you were using when the job was queued.
(Technically, the user index hashes of the canceling request and the job to be canceled must
match.) Explicitly, when canceling a batch job you must do the following:
• If you submitted the batch job through a card reader, you must signon under the
same user number that you used in the batch job.
. If you used the SUBMIT or SENDJOB command to submit the job from another
CYBER batch or time-sharing job, you must signon under the same user number that
you were using when you issued the SUBMIT or SENDJOB command.
CYBER 174 INSTALLED
CSO is now operating a CYBER 174 in parallel with the CYBER 175. The two systems are
running with identical operating systems and share all permanent files.
Although the two machines are running with identical operating systems, there are several
differences:
. Due to the expense of obtaining second licenses, some software packages are being
offered only on the CYBER 175. The packages involved are COBOL, GPSS, APEX,
SORT/MERGE, Interactive Debug, and ALGOL-68.
. The CYBER 174 has no direct communication link with the IBM 4341. As a result,
the following facilities are not available to jobs running on the CYBER 174:
SENDJOB and SUBMIT cannot be used to submit jobs to the IBM 4341.
The HASP (&D) command is not available.
CYBER 174 batch jobs cannot be submitted from a card reader. All the card
readers are attached to the IBM 4341 which in turn is attached to the CYBER
175. Any CYBER job read into one of the card readers will be sent to the
CYBER 175; however, all CYBER accounts are authorized to run batch jobs
on the 175.
Output cannot be sent from the IBM 4341 to the FETCH queue on the
CYBER 174.
• The CYBER 174 and CYBER 175 do not share queues. As a result, each machine has
its own FETCH queue. It is also impossible for a job on one machine to enquire
about the status of a job on the other. This has a practical bearing on print, plot, and
punch jobs generated on the CYBER 174. These jobs initially are placed into the
appropriate queues on the CYBER 174. While there, a CYBER 174 job can enquire
on their status. However, these jobs are eventually transferred to the CYBER 175
where they are put into the corresponding print, plot, or punch queue. Once a job is
passed to the CYBER 175 its status can no longer be determined by a CYBER 174 job.
. CYBER 174 jobs are limited to 35K (104,207 octal) words of memory between 8 AM
and 8 PM and 75K (222,370 octal) words of memory after 8 PM.
• There are no user accessible tape drives on the CYBER 174.
The CYBER 174 is being dedicated to instructional use and the CYBER 175 to faculty and
graduate research. Accordingly, all CYBER interactive computing by projects funded through
class allocations is being restricted to the CYBER 174; all other interactive use is being
restricted to the CYBER 175. There are no similar restrictions on batch jobs. If the hardware
or software restrictions noted above make it impossible for a class to use the CYBER 1 74,
that class will be permitted to use the CYBER 175. Instructors of such classes should contact
the CSO Accounting Office.
To facilitate class preparation, project managers of class accounts will automatically be
permitted to use both the CYBER 174 and CYBER 175. Similarly, teaching assistants and
graders can be granted permission to use the CYBER 175 at the request of the project
manager.
Approximately equal numbers of public terminals are available for use on the CYBER 1 74
and the CYBER 175. An attempt has been made to place the terminals near the populations
using them, with CYBER 1 74 terminals concentrated at sites most used by students. For a
complete guide to the location of public terminals, see Reference Guide RF-0.3 Job Entry
Sites. Twelve dial-up ports (phone number 333-4004) are also available on the CYBER 174.
An attempt has been made to equalize costs between the two machines. This is accomplished
by applying a multiplier to the actual CYBER 1 74 CPU time before it is recorded. The
multiplier is chosen so that, for the average job, the CPU time recorded will equal the CPU
time that would have been recorded had the job run on the CYBER 175. Since the recording
of CPU time is done at the lowest level of the system, all code dealing with limits
enforcement, usage reporting, or charging uses this adjusted value of CPU time.
Users will notice that jobs take about four times as long to execute on the CYBER 174 as
they do on the CYBER 175. This is longer than might be expected considering that the
CYBER 174 has approximately half the throughput capacity of the CYBER 175. This
discrepancy is explained by the fact that the CYBER 1 74 is a dual processor but a job can
make use of only one of the processors at a time.
GANDALF SWITCH INSTALLED
Over the break between semesters, we installed a Private Automatic Computer eXchange
(PACX) built by the Gandalf Company. This device serves as an electronic switch between
terminals and computers and is often descriptively referred to simply as "the Gandalf Switch"
or "the switch."
A total of 58 public terminals are connected to one side of the PACX, and the CYBER 174
and CYBER 175 are connected to the other. A user stepping up to one of the terminals
marked as being "on the Gandalf switch" can use that terminal on either the CYBER 174 or
CYBER 175.
When using one of the switchable terminals, a modified signon procedure, documented on
the terminal, must be used. We thought you might like a better idea of what's happening, so
here's a fuller explanation of how to logon at one of these terminals:
. First, depress the BREAK key once and hold it down for about a second. When the
PACX sees the BREAK signal, it puts your terminal into a queue of terminals waiting
for service and responds by sending a sequence of null characters. Unfortunately,
most terminals discard any null characters sent to them so you wo not receive any
visible acknowledgement that your BREAK character has been received. (The IBM
terminals provided at several RJE sites are an exception. These terminals will print
the null character as a backward question mark.) Some DOs and DONTs follow.
DO hold the BREAK key down for about a second. The PACX ignores a very
short BREAK signal, considering it to be electronic noise. While some
terminals will always send a BREAK signal of reasonable duration, others
transmit only while the key is depressed. A good typist using one of these
latter terminals can generate a BREAK signal so short that the PACX considers
it noise.
DON'T hit the BREAK key more than once. Once a terminal has been placed
in the service queue, another BREAK character will be treated as a request to
remove it from the queue.
Wait two seconds and hit the RETURN key. The two-second wait gives the PACX
time to assign a server to your terminal. (A server is a component of the PACX
which can communicate with you and complete the connection with your chosen
computer.) The RETURN key signals the server to proceed. The terminal responds
with "ENTER CLASS", the PACX's way of asking which computer you want to use. If
your terminal does not respond, there are a couple of possible reasons:
A server has not yet been assigned to your terminal. There are a limited
number of them available and you may have to wait for one to free up. Wait a
couple of seconds and hit RETURN again. Try this at least three times before
giving up on this diagnosis.
The PACX might not have seen your initial request for service. (This should
not happen if you hold the BREAK key down for a full second.) If this has
happened, the remedy is to go back to step one and depress the BREAK again.
This is always a last ditch diagnosis, given the PACX's treatment of a BREAK
for a terminal that is already in the service queue.
Enter a three-digit code to identify the computer you wish to use, followed by a
carriage return. The correct code is easy to remember; it's 174 for the CYBER 174
and 175 for the CYBER 175. If all goes well, your terminal responds with "CLASS
xxx START" where xxx is the code you entered. You might, however get one of the
following messages:
CLASS xxx UNASSIGNED
The three digit code xxx is unknown. You will be prompted for another.
CLASS xxx UNAVAILABLE
The class xxx is temporarily not available. The chosen computer might, for
instance, be out of service for repairs. You will be prompted for another code.
CLASS xxx RESTRICTED
The code you have given is not acceptable from the terminal you are using.
You might, for instance, have accidentally given a code that is recognized but
is used only for diagnostic purposes. You will be prompted for another code.
CLASS xxx BUSY
QUEUE SIZE yyyy DO YOU WISH TO QUEUE?
This pair of messages indicates that all paths to the computer you have chosen
are busy and that the number of terminals waiting for a path to free up to this
computer is yyyy. If you want to be put at the end of this waiting list, enter Y,
but be aware that once you enter Y, you are placed in the queue and cannot do
anything to get out of the queue until a path to the computer becomes
available. While you are waiting you will receive periodic reports of your queue
position. When a path becomes available, you will be prompted with a START
message.
If you do not want to wait but want to try to connect to another computer
instead, enter N. You will be prompted for another three digit code. If you
want to abandon your attempt to use the computer, type Q. You will be
disconnected from the PACX.
BYE
You waited too long (more than 20 seconds) to enter the requested three digit
code. Your terminal has been disconnected from the switch and you will have
to start over by depressing the BREAK key.
Enter RETURN to initiate the usual CYBER logon sequence. From this point on, the
PACX simply transmits every signal generated by your terminal directly to the
computer. There is no way for your terminal to communicate with the PACX. The
connection will be broken when the computer you are using sends a disconnect signal
to the PACX, something that is normally done at logoff.
MORE EQUIPMENT CHANGES
Since the last issue of OFFLINE, the CYBER card reader, printer, and Calcomp plotter have
been removed from service. The card reader and printer were removed for financial reasons.
The money previously applied to their support has been applied to the new CYBER 174.
Other installed equipment has sufficient capacity to absorb their workload. The Calcomp
plotter was removed for reasons of reliability and has been replaced by three plotters
manufactured by the Zeta Company.
All jobs which would normally have printed on the CYBER printer are now automatically
being routed to the IBM system for printing at DCL. While waiting in the IBM print queue
such jobs are given an IBM jobname of the form XXXXYYY where XXXX is the first four
characters of the CYBER jobname and YYY is the bin number used to file the output. Jobs
will be printed with an IBM burst page. The IBM printers will split any line longer than 133
characters (including carraige control) into two lines.
10
Users accustomed to using the CYBER card reader will have to make the following changes
in their decks before reading them in on any card reader attached to the IBM system:
. A card containing the characters /"CYBER in columns 1-7 and blanks everywhere
else must be placed at the beginning of the deck.
. The 6/7/8/9 multipunch card used to mark the end of the deck must be replaced with
a card containing the two characters /* in columns 1-2 and blanks everywhere else.
. Any IBM Job cards contained in the deck must be removed.
Three facilities provided by the CYBER card reader which have been lost are:
• The ability to read an end of partition card (6/7/9 multipunch)
• The ability to read CYBER binary decks
. The ability to read and automatically translate decks punched in 026 format. In this
case other provisions can be made. See a consultant for details.
Steps required to convert from use of the Calcomp to the Zeta plotters were documented in
the December issue of OFFLINE. Consult it for details or see a consultant for conversion
advice.
In the way of future changes, we are setting up procedures for using a Honeywell "page
printer" being installed at the campus AISS computing center. This device prints on 8 1/2" by
1 1" sheets of paper at the rate of 8000 lines per minute, using an electrostatic printing
process. We plan to use this device for printing very large output files. Details will be
announced when the service becomes available.
NEW 6250 BPI TAPE DRIVES
In September, CSO acquired new tape drives for the CYBER 175, including 9-track drives
capable of reading and writing at densities of both 1600 bpi and 6250 bpi. The transition to
the new drives is now complete, and the tape drive configuration on the CYBER 175 is now
. Three 9-track 1600 bpi / 6250 bpi drives
. One 9-track 800 bpi / 1600 bpi drive
• One 7-track 800 bpi / 556 bpi drive (which can also read at 200 bpi)
This configuration introduces 6250 capability and reduces 9-track 800 bpi capability from four
drives to one drive.
11
The 9-track tapes use odd parity. The 7-track tapes use odd parity for binary data and even
parity for coded data.
Successful 6250 bpi usage requires that the type of tape used for writing at this density be the
precise type to which the tape drives have been tuned. Our local drives have been tuned to
Scotch 701 3200 foot tapes rated at 6250 cpi (cpi=bpi), and only that specific tape should be
used here for writing at 6250 bpi. (The critical aspect seems to be the writing of the tape.
Almost any tape written at 6250 bpi appears reliable for reading.) Scotch 701 3200 foot tapes
may be purchased from the CSO Distribution Center, 1208 W. Springfield. (Scotch 701 in
shorter standard lengths is thicker and will not match the tuning of the drives.)
Tape written at 6250 bpi can hold about three times as much information in a given space as
tape written at 1600 bpi. Thus, 6250 bpi is well suited for recording massive quantities of
data. But remember, our IBM 4341 does not have 6250 drives; therefore, this high density
cannot be used to transfer data between the two computers, nor can the IBM analysis and
data recovery programs be used on any problematic 6250 bpi tape. Also, when preparing tape
for transferring information to another site, 1600 bpi or perhaps 800 bpi is preferable to 6250
bpi because not all sites have 6250 capability.
The usage of 6250 bpi and the new tape configuration require few changes, simply more
attention to the density parameter in the LABEL statement and changes in the RESOURC
statement. The details are summarized below.
The following discussion refers to 9-track tapes only. The usage of 7-track tapes is
unchanged.
The default for CYBER 175 tape usage will remain 9-track 1600 bpi. To specify 9-track tape,
include NT in the LABEL statement or omit the track designation and get 9-track by default.
To read or write at the various 9-track densities, include the following in the LABEL
statement:
. For 6250 bpi, always specify D=6250 or D=GE
. For 800 bpi, always specify D = 800 OR D=HD
. For 1600 bpi, omit the D= specification and get 1600 by default, or specify D = 1600
or D=PE (Including the density specification rather than using the default is good
documentation.)
Some examples of 9-track LABEL statements are given below. For more information on the
LABEL statement see Chapter 10 of the CDC NOS Version 1 Reference Manual.
LABEL(TAPE,VSN=ABCDEFD444,D=GE,P0=W.SI=ABCDEF,QN = 1,W)
LABEL(TAPE,VSN=MYTAPE-F555,D=PE,PO=W,SI=MYTAPE.QN=9999)
LABEL(TAPE,NT.VSN = NEWDATTEMP,D=800,PO=R.LB=KU.F=S,CV=EB)
12
The RESOURC statement is required when more than one tape will be used concurrently in a
job. In the RESOURC statement, you must now specify the 9-track tape drives by the
alphabetic characters for density:
. GE for 6250 bpi
. PE for 1600 bpi
. HD for 800 bpi
To specify two 9-track 1600 bpi drives, use
RESOURC(PE=2)
To specify one 9-track 800 bpi drive and one 9-track 1600 bpi drive, use
RESOURC(HD=l,PE=l)
To specify one 7-track drive and one 9-track 6250 bpi drive, use
RESOURC(MT=l,GE=l)
Users of the EXAMINE and ARCHIVE programs should be aware that the report of the
amount of tape used that is returned by these programs will be too large for 6250 bpi tapes.
Also, the tape density must not be specified in the EXAMINE statement for 6250 bpi tapes,
and EXAMINE does not give a correct report of density for 6250 bpi tapes. In all other
respects, the programs work satisfactorily with these tapes.
CSO VIDEOTAPES AVAILABLE
CSO has produced a series of eight videoptapes which introduce the novice to the CYBER
System. A viewing guide containing the major pictorials used in the series is available and
can be used to facilitate note taking.
The title and a brief synopsis of each of the videotapes is given below. Running time is 10-15
minutes for each videotape.
Introduction to Computing at CSO
A brief look at the steps required to solve a problem using a computer and some
of the hardware used.
Using a Terminal
A description of the physical operation of a terminal and some of the keys that
have a special meaning to the CYBER.
13
Introduction to CYBER Time-Sharing
A tutorial on the logging on and off of the CYBER.
File Usage - Local File and Indirect Access to Permanent Files
An introduction to CYBER files and the commands used to manipulate them.
Introduction to ICE Text Editing
A tutorial on entering and modifying files with ICE.
Running a FORTRAN Program - Concepts
A discussion of the comcepts of compilation loading and execution.
Running a FORTRAN Program - The PROGRAM Statement
A discussion of the PROGRAM statement and its relation to files accessed by the
program.
Running a FORTRAN Program - Control Statement
A discussion of the control statements used to compile, load, and execute a
FORTRAN program.
Anyone can view these videotapes by going to the Undergraduate Library in person to make a
reservation for use of the videotape equipment. Ask for a copy of the viewing guide when
you checkout the videotape for viewing.
Copies (Betamax format) of these videotapes are available for loan from CSO to any
instructor wishing to use them in class. They were effectively used in this environment
several times last semester with the instructor stopping the playback equipment whenever
he/she wished to elaborate further or questions arose from the class.
To borrow a videotape for classroom used and obtain copies of the viewing guide for
classroom distribution, call Scott Lathrop (333-6618). If you do not already have access to
the required videotape equipment, Betamax viewing equipment can be borrowed from the
Office of Instructional Resources (333-3690).
CHANGES TO MICROCOMPUTER SOFTWARE IN GRAB
On January 29, 1981, changes were made to four of the microcomputer support products
available on the CYBER via GRAB. The four affected products are: INT8080, Z80, M6800,
and M68000.
14
From a user's point of view, the most extensive changes were made to INT8080. Entering
the command
GRAB.INT8080.
still places the three CCL procedure files I80ASM, I80SIM, and I80PLM in your local area.
The I80PLM procedure file has not changed; however, I80ASM and I80SIM have been
changed as follows:
. I80ASM now runs only the Intel 8080 cross-assembler (previously, it also ran the
interpreter). 180 ASM is called by entering:
\80ASW=source,L=listing,B=lntel binary,J=MUMS binary
. I80SIM runs the Intel 8080 simulator. It is called by entering:
I80SIM,B= Intel binary input,\ = command mput,L= output.
In the Z80 procedures, some of the parameter names have been changed:
. Z80ASM is now called by entering:
Z80ASM.I = source,L=listing,B= binary
. Z80LINK is now called by entering:
Z80LI NK,B = filel.P=file2,N = file3,\ = file4,L0= file5.
Where:
filel is the file containing the object output from the last Z80 assembly.
file2 is the file containing a library of segmented text.
file3 is the file containing the object output of the link.
file4 is the file containing the directives to be executed.
file5 is the file receiving the diagnostics and map.
The main change to the M6800 product is that the default assembly language output file for
M68HMPL has been changed to COMPILE. In addition, there has been a major bug fix in
this version.
15
In the M68000 product, M68KSIM has become a binary file rather than a procedure file. An
"REW" parameter has been added to allow the rewinding of all files before the simulation. If
not specified, no files are rewound.
Additional details are available in the system HELP files and in the new microcomputer
Reference Guides: RF-20.1 INT8080, RF-20.2 Z80, RF-20.3 M6800, and RF-20.4 M68000.
MATHEMATICAL SERVICES
IMSL NEWSLETTER
Issue 26 of IMSL's Numerical Computations Newsletter has been received and can be
examined in the Consulting Office at DCL, in the IMSL General Information Manual. The
newsletter contains articles on:
. Producing probability plots from data.
. A version of the IMSL Library for the PRIME 300/400/500 and 50 series.
. New packages available through the IMSL distribution service: MINPACK-1 and
ITPACK-2A.
. User services provided by IMSL.
If you wish to be on the mailing list for the newsletter, write to IMSL at
Sixth Floor-NBC Building
7500 Bellaire Blvd.
Houston, TX. 77036
and indicate that you belong to a subscribing organization, or tell Stan Kerr (175 DCL, 333-
4715, or TELL,UN=MATHLIB).
16
DOCUMENTATION
JANUARY-FEBRUARY ISSUES COMBINED
I am sure all of our users are aware of the many moves and changes being made within CSO.
During the semester break, the PDP 11/50 and the phototypesetter used to produce OFF-
LINE were moved from the Astronomy Building to DCL. Since this movement of
equipment necessitated a prolonged period of downtime, it was decided that we would
combine the two issues rather than produce an issue using a different format on another
machine.
REVISED CSO DOCUMENTATION
Revised Manuals
Introduction to the CYBER 175 Free
An introductory manual for persons using the CYBER 175 for the first time. This is a
completely new version of the manual; it has been reorganized, reformatted, and
rewritten.
Fortran Conversion Guide Free
A manual for FORTRAN users who wish to convert from IBM system FORTRAN to
CDC CYBER 175 FORTRAN Extended. Revisions have been made on pages 17-19
(CONVPRC procedure changed to MIMI) and pages 27-29 (TIDYPRC procedure
changed to TIDY).
RNF Tutorial Free
A working document currently used for the RNF short course. This working
document temporarily replaces the RNF Users Guide. A new system version of RNF
is currently being worked on and will be announced along with updated documentation
at a later date.
Archive Reference Manual Free
A manual for users who wish to keep copies of permanent files on tape. Revisions
include: RESOURC statement on pages 4, 5, 19, and 20; AD parameter on page 11;
and DV parameter on page 12.
Reference Guides
RF-1.4 Creating CYBER File From Card Deck - has been revised to reflect removal
of the CYBER card reader.
17
RF-1.2 CYBER Rates -- has been revised to reflect the new plotting charges.
RF-5.2 GCS Library -- examples of programs and a plot have been added. Also,
commands for accessing version 3-D have been corrected.
RF-11.2 4341 Rate/Charge Estimator - has been revised to reflect the new plotting
charges and the new rate for terminal repair.
Four new microcomputer Reference Guides have been added:
RF-20.1 INT8080
RF-20.2 Z80
RF-20.3 M6800
RF-20.4 M68000
MISCELLANEOUS
GRAPHICS PRESENTATION
Representatives from Tektronix will be giving a presentation from 2:00 to 3:45 PM on
Tuesday, February 10, 1981 in the Coordinated Sciences Lab (CSL) auditorium. Slightly less
than half this period will be spent in a discussion of graphics in general with emphasis on
Tektronix equipment. A Tektronix 4052 Graphic Computing System and a new hardcopy
unit will be shown. This will be followed by a presentation of Tektronix' Interactive Graphics
Library (IGL) software. This is a powerful language which can be individually tailored to the
needs of an application.
CENTRAL ILLINOIS USERS GROUP MEETINGS (DEC)
The following meetings of the Central Illinois Local Users Group (Mainly planned for users
or potential users of DEC machines) have been scheduled. Unless otherwise noted, the
meetings will be at 4 PM in the Auditorium at the Coordinated Science Laboratory (CSL).
February 10 Graphics terminals. Al Tuchman from CSO will share some of
his impressions of the wide variety of graphics terminals on the
market today. (See announcement of Tektronics demonstration
elsewhere in this issue.)
March 10 Panel discussion on "word processing". What you should look
for and expect in a word processing system, whether it is based
on a PDP-1 1, PDP-8 or some other system. (Tentative)
18
April 14 Technical description of VAX by DEC representatives. (Time
and place to be announced.)
HELP WANTED
RESEARCH ASSOCIATE AND/OR CO-AUTHOR
WANTED FOR COMPUTER WORK
Any graduate student or other person who would like to earn over $5.00 an hour doing some
interesting computer work with Professor Stuart Nagel of the Political Science Department
should phone him at 359-8541. Total hours per week are about 10 hours at the convenience
of the person doing the work.
The work mainly involves running research data through SPSS or SOUPAC statistical
programs and through some mathematical optimizing routines. The subject matter of the
data generaly deals with a causal or evaluative analysis of alternative public policies.
Previous experience with statistical computer work or related work is expected. No
programming will be involved, but the person should have a knowledge of how to write
control cards for a statistical software package and possibly experience with mathematical
software.
This work was formerly performed partly by a graduate student named Marian Neef before
she received her Ph.D. In doing the work, she became the co-author of numerous articles,
book chapters, and books. This could be an excellent publishing opportunity for another
graduate student.
If you are interested or if you desire further information, phone 359-8541.
19
SPECIAL TEAR-OFF SHEETS
MATH SHORT COURSE SURVEY
In the Spring Semester, as part of our Short Course series, there will be a Math Libraries
course giving a general survey of mathematical/numerical software available. We would like
to offer at least two or three additional courses on specific numerical or software topics.
Please help us decide what these should be by circling your preferences below and returning
this sheet to Stan Kerr, 175 DCL.
Curve Fitting Differential Equations
APEX Nonlinear Least Squares
ACSL Nonlinear Programming
Simulation General Linear Programming
EISPACK Eigenvalue Problems
Splines Partial Differential Equations
MPOS Solving Nonlinear Equations
GASP SLAM
LINPACK
Symbolic Manipulation -- doing formal algebra by computer
Large Linear Problems -- sparse matrices
Name
Address/Phone
RETURN TO:
Stan Kerr
175 Digital Computer Lab
Campus
21
REQUEST FOR INFORMATION--TEKTRONIX GRAPHIC TERMINALS
Quite often staff members specify Tektronix Graphic Terminals as hardware items in
proposals or in requests for purchase orders. Because of this some effort is being made to
obtain an agreement with Tektronix for a quantity reduction in cost.
One item of data is needed from you in order to open the negotiations. Tektronix needs to
know an approximate number of terminals planned for acquisition during the next 18
months. If any apply to you, please enter a number in the appropriate blanks below and
return to C.E. Carter, 195 DCL.
Tektronix Computer
Graphic Terminal Purchase Propose Plan
4006-1
4010-1
4012
4014-1/4015-1
4016-1
4027
4024
4025
4051
4052
4054
4662
4663
4631
4632
Explanation
The purchase column means that funds are available and requests are going to be made.
The propose column means that funds are not available but graphic terminals have been
included as budget items in proposals.
The plan column means neither of the above but a Tektronix terminal will be included when
you do propose or purchase.
Name
Address/Phone
RETURN TO:
Cliff Carter
195 Digital Computer Lab
Campus
OFF-LINE's Mailing List
If you wish to be placed on our mailing list for future issues of OFF-LINE, if you wish to be
removed from the list, or if you wish to enter an address correction, please complete and
return this page. (Current subscribers are kept on the mailing list until a specific request for
removal is received, or until a mailing is returned as undeliverable.)
Check one:
□ New subscriber
□ Removal request
□ Address correction
Name:
Address:
CAMPUS or Zip Code
(If address correction, give old address and zip code below.)
Comments:
RETURN TO:
OFF-LINE
150 Digital Computer Laboratory
University of Illinois at Urbana-Champaign
Urbana, Illinois 61801
PRISCILLA YJ
CDLLECTIQNJ DEV DEPT
111 *IAIN LI:3RA^Y
CAMPUS
t/
Computing
Services
Office
Off-Line
University of Illinois at Urbana-Champaign
VOL. 9, NO. 3 March 1981
EDITOR: Lynn Bilger
PHONE: (217) 333-6236
139 Astronomy Building
1011 W. Springfield
Urbana, Illinois 61801
\e
Contents
SYSTEM NOTES
1
January Reliability Reports
CYBER
1
2
A Procedure File for the BILL Command
New Easy Graphing Software
THE LIBRARY OF THE
MATHEMATICAL SERVICES
APR 01 1:
3
3
4
5
Transactions on Mathematical Software
Math Notes
New Math Software
XMP Updates
UNrVERSITY OF ILL
10
TEXT PROCESSING
Diablo Services
MISCELLANEOUS
From Plug-Board to Pascal
SALES - EXCHANGES - HELP WANTED
Tektronix Graphics Display
CSO DIRECTORY - STAFF AND SERVICES
Administrative
Director
Business Manager
Secretary
User Accounting
Distribution Center
Systems Consulting
Statistical Services Consulting
Text Processing Consulting
Terminal Repair Service
CYBER Dial-up Numbers
Asst Dir User Services
Asst Dir Systems and Operations
Asst Dir Statistical Services
and UNIX System
Asst Dir Development
Asst Dir Engineering and
Hardware Selection
Documentation
CYBER-IBM Operations
UNIX Operations
Telecommunications
Laboratory Support Project
RJE Operations North
RJE Operations South
RJE Sites
Agriculture
Chemistry
Commerce West
CRH Snack Bar
DCL Routing Room
Electrical Engineering
Florida Ave Res Hall
Illinois St Res Hall
Mechanical Engineering
Psychology
Social Science
George Badger
150
DCL
333-4103
Stanley Rankin
150
DCL
333-6530
Joyce McCabe
150
DCL
333-1637
rare Support
1208
W Springfield
333-7752
1208
W Springfield
333-6760
166
DCL
333-6133
65
Comm West
333-2170
207
Astronomy
333-7318
150
DCL
333-0969
CYBER 175
110-300
baud
333-4000
CYBER 175
1200
baud
333-4001
CYBER 174
110-300
baud
333-4004
Robert Penka
173
DCL
333-4709
Sandra Moy
177
DCL
333-4703
Larry Sautter
91
Comm West
333-2170
J. M. Randal
120
DCL
333-9772
Cliff Carter
195
DCL
333-3723
Lynn Bilger
139
Astronomy
333-6236
Jack Knott
194a
DCL
333-6562
Debbie Weller
171
DCL
333-8150
Tom Kerkering
164
DCL
333-0816
Mike Gardner
164
DCL
333-7904
Rex Duzan
162
DCL
333-6285
Don McCabe
1208
W Springfield
333-2171
333-7752
M103
Turner Hall
333-8170
153
Noyes Lab
333-1728
70
Comm West
333-4500
120
Snack Bar
333-1851
129
DCL
333-6203
146
EEB
333-4936
FAR
333-2695
1SR
333-0307
65
MEB
333-1430
453
Psych Bldg.
333-7531
202
Lincoln Hall
333-0309
OFF-LINE is the monthly newsletter of the Computing Services Office at the University of
Illinois at Urbana-Champaign. Unless otherwise indicated, permission to reprint is freely
granted, provided that the author, if named, and the Computing Services Office (CSO) are
credited. Information in this issue is current as of February 24, 1981.
CSO operates a CDC CYBER 175 with 256K words of central memory and a CDC CYBER
174 with 131 K words of central memory. The 175 and 174 run under the NOS Operating
System and share 512K words of ECS. The 175 serves over 200 simultaneously active text
and graphics terminals and the 174 serves over 100 simultaneously active terminals. CSO
also operates an IBM 4341 with 4 million bytes of memory running HASP-OS/MVT under
VM. In addition, CSO operates a DEC PDP-11/50 with 252K bytes of core driving a GSI
CAT-8 phototypesetter and a DEC VAX 11/780 with 2 million bytes of core, both running
under the UNIX Operating System.
SYSTEM NOTES
JANUARY RELIABILITY REPORTS
CYBER 175 CYBER 174
28 recoverable interruptions 9 recoverable interruptions
27 non-recoverable interruptions 2 non-recoverable interruptions
40 of these were for more than 15 minutes 4 of these were for more than 15 minutes
Mean Time Between Failures = 10 hours Mean Time Between Failures = 30 hours
Mean Time to Repair = 20 minutes Mean Time to Repair = 16 minutes
Availability: 96.6% of scheduled uptime Availability: 99.2% of scheduled uptime
Major cause of downtime was related to Major cause of downtime was related to
hardware and disk problems. hardware and disk problems.
IBM 4341:
10 interruptions
iu interruptions
4 of these were for more than 15 minutes
Mean Time Between Failures = 74 hours
Mean Time To Repair = 86.9 minutes
Availability: 99% of scheduled uptime
Major cause of downtime was hardware problems.
CYBER
A PROCEDURE FILE FOR THE BILL COMMAND
From time to time, CYBER users request that additional information be provided at logon.
Since users vary widely in what they'd like to see, we suggest that you create a procedure to be
invoked at logon to generate the desired information.
As a simple example, suppose that you have created a permanent indirect access file named
PROCFIL which contains the following record. (In the simplest case, this would mean that file
PROCFIL contains only the following lines.)
PROC.START
$BILL,chg,prj.
$IF(.NOT.FILE(OPTION,LO))GET,OPTION/NA.
$FETCH.
$HEARYE/F/S = 3.
$REVERT
Where chg,prj is your charge and project number respectively. You can then invoke this
procedure file at logon instead of issuing the usual BILL or CHARGE command. When
invoked, the procedure will supply your BILL command, get your option file if you have one
and it's not already a local file, list the names of any files you might have in the FETCH queue,
and list the titles of any HEARYE bulletins generated during the last three days. To invoke
this procedure, use the command
BEGIN.START.
If you are using time-sharing this command can be abbreviated to
-START.
If the proc above is the first (or only) record in file PROCFIL and if PROCFIL is either
rewound or is not a local file, the word START need not be specified on the above commands.
In this case, logon to a time-sharing terminal at its briefest would look like:
SIGNON: 123456789
PASSWORD:
RECOVER/CHARGE: -
CCL provides more elaborate facilities you might want to investigate. You can, for instance,
have more than one proc in PROCFIL and can pass parameters to procs at the time they are
invoked. There are CCL statements which can make use of passed parameters to govern the
sequence of commands to be executed. CCL is documented in the NOS Reference Manual,
Volume I, chapter 4.
NEW EASY GRAPHING SOFTWARE
In early March, CSO will install Tektronix' PLOT- 10 Easy Graphing plotting program. No
programming knowledge is necessary for a user to produce plots, bar charts, or pie charts using
Easy Graphing. Although this program is intended to be used on the Tektronix graphics 4010
family of terminals, we have modified Easy Graphing to also produce plots on the ZETA
plotters. The April issue of OFFLINE will contain a more detailed description of Easy
Graphing, its capabilities, and examples of its use. A user's pocket reference guide for Easy
Graphing may be purchased for $1.00 at the Distribution Center, 1208 W. Springfield. The
complete users manual is expected to be in stock very soon. We plan to offer a short course
for Easy Graphing later this semester.
MATHEMATICAL SERVICES
TRANSACTIONS ON MATHEMATICAL SOFTWARE
Two articles appeared in the December 1980 issue of Transactions on Mathematical Software
which may be of special interest to our users:
• A Survey of Software for Partial Differential Equations by Marek Machura and
Roland A. Sweet. This article describes a number of known PDE packages.
• PERUSE: An Interactive System for Mathematical Programs by William Kurator and
Richard O'Neill. This article describes a system for interactively examining the data
and solutions of large linear programs.
We now have on tape all TOMS algorithms through December 1980. For information about
this tape, enter the following:
GET,TOMS/UN=MATHUB.
PRINT.TOMS.
MATH NOTES
A series of informal writeups on various math software topics, prepared by Stan Kerr, is
available on the CYBER. These writeups are accessed through a procedure file called MNOTES
in user number MATHLIB, and identified by number (currently 0-10). MNOTES is accessed
by the command:
GET,MNOTES/UN=MATHUB.
Then, to print note number 6, for example, you would enter:
MNOTES,6,OUT=/fn. *
PRINT./fh/AS/CC.
Where Ifn is any valid file name. If you omit the parameter OUT=/rn, the default file OUT is
used for Ifn, so you can enter:
MNOTES.6.
PRINTOUT/ AS/CC.
To print all current notes, enter the following:
MNOTES.ALL
PRINT.OUT/AS/CC.
If you enter just the command
MNOTES.
you will receive on your terminal a directory of currently available notes.
As of this writing, the current notes are:
0 A description of MNOTES.
1 A brief outline of math software available on the CYBER.
2 A description of procedure CAT/UN = MATHLIB, which gives an annotated
directory of MATHLIB.
3 (unused at present)
4 Summary of math libraries with descriptions.
5 Description of IMSL products and services.
6 Description of TOMS ( Transactions on Mathematical Software).
1 Summary of simulation languages and programs.
8 A list of possible sources of software, if you are searching for something.
9 Routines for ordinary differential equations.
10 A nice writeup on LSODE, the latest version of the Gear algorithm for ordinary
differential equations.
We stress that these notes are informal and will certainly change with time, as new information
is added. Your comments and suggestions are welcome. Please direct any questions to Stan
Kerr, 175 DCL (333-4715), or TELL,UN=MATHLIB.
NEW MATH SOFTWARE
CSO has recently acquired the following software. Although this sofware is on the system and
can be accessed and used, it is not currently available via the GRAB command. If you wish to
use any of the new software, or have questions, please contact Stan Kerr, 175 DCL (333-4715).
• Algorithms from the December 1980 issue of TOMS (Transactions on Mathematical
Software) :
563 A program for linearly constrained discrete l\ problems.
564 A test problem generator for discrete linear l\ approximation problems. (As
received, this one is incomplete, but a correction is forthcoming.)
• NL2S0L, an adaptive nonlinear least squares algorithm by J. E. Dennis, D. M. Gay and
R. E. Welch of the University of Wisconsin at Madison, distributed by IMSL.
• Algorithms for integer goal programming, using branch-and-bound and implicit
enumeration, by Sang M. Lee of the University of Nebraska.
ARSTEC, an adaptive random search ("Monte Carlo") technique for nonlinear
optimization problems, from Argonne National Laboratory.
ADBASE, a multi-objective function linear programming package by Ralph Steuer of
the University of Kentucky. This was purchased by Dave Whitford of Finance and
donated for public use.
XMP UPDATES
Updates for the linear programming library XMP have been installed and are available by
entering:
GRAB.XMP/F.
If no problems arise, this version will become the default XMP by May 1, 1981. The updates
include:
. Faster versions of XCAND and XDCHQ. The calling sequence for these has changed.
• Generalized upper bounding (GUB) routines.
• A parametric routine XPARAB.
See Stan Kerr, 175 DCL (333-4715) or TELL,UN = MATHLIB for documentation on the
updates.
TEXT PROCESSING
DIABLO SERVICES
We have been getting quite a few questions about the Diablo Service and felt that a repeat of
the information published earlier in OFF-LINE may be of help to our users.
All RNF source files must contain the .LPT command in the first line of the file, and must be
processed through RNF into an output file. Users unfamiliar with RNF are referred to the RNF
User's Guide ox the RNF Reference Manual (available in the Distribution Center, 1208 W.
Springfield). All non-RNF source files must have carriage control and begin with a page eject.
The RNF output file or the non-RNF source file can then be submitted to the Diablo queue via
a CYBER CCL procedure called DIABLO.
After you have prepared your file for submission, you should access and execute the DIABLO
procedure by issuing the following commands from time-sharing on the CYBER:
GRAB.DIABLO.
DIABLO.
The procedure prompts you for your surname and whether or not you wish to charge the
printing to a UOI account number. Costs for printing can only be charged to "real money" UOI
accounts; class accounts or research board accounts are not allowed.
If you choose to pay by UOI account number, you will be prompted for the account number
and title at the time the procedure is running. If you choose to pay by check, you will not be
prompted for an account number or title. You will also be prompted for a choice of the pica or
elite font and you will be asked for the name of your local file containing your RNF output.
NOTE: The procedure will not perform a GET if it cannot find a local file by the name given.
Finally, at the conclusion of the procedure, you will be told the number of PRUs (one PRU
equals 640 upper case characters, 320 lower case characters) and the dollar cost associated with
its printing. The charge is $0.05 per PRU. All charges continue to be rounded up to the
nearest dollar.
You will be sent a message on the CYBER once your file has been printed on the Diablo. You
may pick up your printout at room 123 DCL. If you chose to pay by check, you should make
out a check at that time to the University of Illinois for the amount indicated on the billing
sheet you receive with your printout. If you chose to pay by UOI account number, your
account will automatically be billed for the cost indicated on the billing sheet.
It should be pointed out that once something has been submitted to the queue it will be printed
and you will be charged! Please take every precaution to print your RNF output on an upper
and lower case line printer or DEC writer prior to submission to the Diablo queue. Refunds will
not be given for printouts submitted by mistake.
The following conventions have been established for files printed on the Diablo:
• The top of the form (page) will be set to one line below the horizontal perforation in
the paper.
• Plain white 20-lb. paper, 14 7/8" wide by 11" long, will be used.
. If the pica font is selected, the printing will be done with 10 characters and 6 lines per
inch.
• If the elite font is selected, the printing will be done with 12 characters and 6 lines per
inch.
On-line help is available by entering the command
HELP
and then by entering DIABLO (the name of the procedure) after the question mark prompt.
If you have any questions concerning this service, please call the operator at 333-8150.
MISCELLANEOUS
FROM PLUG-BOARD TO PASCAL
This article is a reprint from INTERFACE (volume 7, no. 10, May 1980), the
newsletter of Arizona State University Computer Services. The author, Billy G. Wood,
is an Associate Professor in the College of Engineering, Department of Technology at
Arizona State University.
From plug-board to PASCAL and from sticks and stones to silicon chips, we can trace the development
of the present day microcomputer. Man has built and used mechanical aids to computing since long
before the present number system was fully developed. Probably one of the first such aids used was a
quantity of sticks or stones which acted as a tally and represented an early herdsman's wealth. Today,
that same wealth is represented by tiny, invisible, magnetized spots on a roll of magnetic tape stored in
a bank vault. Most of us are familiar with the computational aids which were popular in the earlier part
of our century; specifically, the mechanical calculator and the slide rule. The mechanical calculator and
adding machine had their beginning in the toothed wheel devices used as tallying or adding machines in
the ninteenth century. The slide rule which was the badge of the science or engineering student on
campus in the late forties began with Napier, ' who also invented tables of logarithms. Most of us have
used those to aid in computing numbers to a high degree of precision. Here we see examples of two
methods of mechanical computing, one discrete, or digital, and the other linear, or analog. The adding
machine or mechanical calculator represents the relatively simple counting or tallying method of
computing, and the slide rule represents analog computing.
World War II furnished much of the early impetus to improve mechanical computing of all types.
Accurate tables of trigonometric and logarithmic values were needed for better maps and navigation.
More complete ballistic tables were needed to do accurate gun aiming, and as weapons became more
and more sophisticated, including automatic weapon directing by radar and other complex detectors,
electronic methods of computing had to be improved to match the capabilities of the rest of the
weaponry. Analog computing reached a high level of perfection shortly after the war. Aerospace
companies such as Boeing and Goodyear Aerospace built, used, and sold vacuum tube analog
computers whose accuracy was limited only by the accuracy of the components built to program them.
This limitation proved to be fatal to the analog computer, since it meant that a physical limit would be
reached beyond which more and more money and effort would be required to make the computer only
slightly more accurate. In the case of even the early digital computers, however, designers could easily
see that no limit to the precision of the result would be reached, provided more and more steps were
accomplished in the program.
Two devices were developed in 1946 which were to culminate in the present microcomputer. One of
these was the transistor, and the second was ENIAC. : The transistor received some publicity from Bell
Laboratories, but the first digital computer, ENIAC, could not even generate support from venture
capital to finance its own successor, until the Bureau of Census decided it needed a computer.
ENIAC contained 18,000 vacuum tubes and had to operate with a 1 in 1014 probability of malfunction
to keep it going for 12 error-free hours.1 The first transistor was a crude, delicate, and low bandwidth
device. However, even at its beginning, it was already much more efficient than a vacuum tube
because it did not require any filament power. In addition, and ultimately just as important, it was
much smaller than a vacuum tube.
Vacuum tubes were first used as the active devices for the first generation of computers - from 1946 to
1956 and beyond. Machines of that time like the IBM 407 were unit record machines usually capable
of one or two simple operations like reading a deck of time cards, multiplying them individually times a
deck of personnel cards, and thus producing a set of paychecks. The programmer was a mysterious
individual schooled in the black art of how to plug a huge bundle of wires into a plug-board and store a
program which would do that week's payroll.
The IBM 650* was many present-day computer manager's introduction into digital computing. It had a
drum memory as the basic storage device for both program and data (just as John Von Neumann had
suggested),4 and thus had the important capability of modifying and testing its own program as it was
executing it.
With the introduction of the core memory and a read-modify-write cycle time reduced to less than a
microsecond, the digital computer began to take on the architecture which we recognize today.
Operating systems were invented to aid in program development, and high-level languages like
FORTRAN were written to simplify the tedious coding and debugging of long programs. The IBM
709* and the SAGE air defense computer were examples of machines of this, the most advanced of
tube-type computers.
By 1959, the alloy junction transistor was well perfected and being produced in sufficient quantities to
be used in one of the first mass-produced second-generation computers, the IBM 1401*. This machine
usually cost about $250,000. It used transistors as the active devices and other discrete parts with
which to build individual flipflops or gates on single-sided print circuit cards. The cards were then
assembled like building blocks into the complete machine. These features made possible an important
facet of the manufacturing of this machine; that is, automatic production methods. The printed circuit
boards used were all prduced by an automatic assembly line and, more importantly, the back-plane
intermodular wiring between printed circuit boards was done by automatic wire-wrapping machines not
subject to human error. All of the interconnections were stored in a computer and the wiring diagrams
which accompanied the 1401* in the field were then printed on an IBM 1403* line printer which was in
turn controlled by a 1401 computer. This must certainly rank as a machine lifting itself by its
bootstraps!
By 1964, the term "integrated" circuits was established, and several devices of these types were available
in the marketplace. They were expensive and consisted of just 20 to 25 active devices (transistors) on
each silicon chip less than 1/8" square, but they were the seed which would grow in the next decade to
the present microcomputer industry.
In the 1960's another type of computing system was introduced, called the minicomputer. This type of
computer is usually defined as costing less than an arbitrary number of dollars; $50,000 was one such
limit. One of the first to be manufactured in large numbers was the Digital Equipment Corporation
PDP-8.** Because minicomputers cost much less than any previous large-scale system, schools and
government labs were able to justify the purchase of a computer much more easily, and more and more
students and teachers were able to see the value of computers in business and schools.
Users groups began to trade universal programs which any laboratory could use in research or teaching.
Although the systems were relatively low cost, they had sufficient computing power so that extensive
operating systems were developed including time sharing one machine between several simultaneous
users; something which had previously been done only when using large systems.
It is readily apparent that as time went on more and more computing capability was being bought for
less and less money. For example, the IBM 1401* costing $250,000 had less arithmetic capability and
memory than the simple DEC PDP-8,** just discussed at less than $50,000.
The economic pressure which was causing this came largely from the semiconductor industry. If a plot
against time,-"' was made of the number of active transistors constructed on each silicon chip, it was
startingly evident that semiconductor density was increasing at a geometric rate. Since a given process
used to make these devices took the same number of manufacturing steps regardless of the ultimate
complexity of the chip, the consumer was getting more and more transistors for the same money.
Thus, the cost of computing capability was steadily decreasing. In addition, the semiconductor
manufacturers were faced with a kind of unprecedented problem for any industry. What does one
build, given the ability to put literally thousands of decision-making devices in one piece of equipment?
The finished device would be less than 1/4 inch on a side and would cost less than $5.00 to $10.00
when mass produced, and must appeal to hundreds of thousands of people in order to support a
production run. The natural evolution of computing capability provided the answer; first with the
scientific calculator, and finally in the microcomputer.
The Motorola 6800 and INTEL 8080 were first produced about a year apart in 1973 and 1974. Each
contained about six thousand transistors within their integrated circuits. On the front cover of Popular
Electronics magazine in January 1975, another era of computing was ushered on the world's stage.
Using the INTEL 8080, MITS, Incorporated of Albuquerque, New Mexico, made the first personal
computer, the Altair 8800.*** Shortly after, they followed it with the Altair 680,*** using the Motorola
6800. It was now possible for anyone to buy a computer for less than $1000, containing more
arithmetic capability and memory than the IBM 1401* of less than 10 years previous.
The pace of increasing circuit density has yet to slacken. Today (1980), Motorola is beginning to build
the 68000 with over 70,000 active devices in one integrated circuit, and designers are projecting chips
with 1,000,000 devices in the immediate future.
It is now routine to build an entire computing system using only one circuit. Memory, input/output,
and arithmetic capability, all are simultaneously manufactured ready to plug in, turn on, and operate.
In the memory space of a microcomputer, a program is usually permanently recorded within a type of
memory designated as "read only memory" or ROM where it cannot be erased or changed. This
program will execute as one of the high-level languages, such as BASIC . t A complete computer
system has recently been manufactured which executes the relatively new high-level language,
PASCAL. Thus, we have now come full circle from the plug-board with its semipermanent memory or
"ROM" executing each step of an early program to the new semiconductor ROM containing an entire
language in permanent memory ready to operate as soon as power is supplied.
At a local department store before Christmas,. 1978, for less than $50, a machine could have been
purchased which could have given this address without it having to be spoken by a human being.
Texas Instruments is presently selling a teaching aid or game called "Speak and Spell" t t which
synthesizes human speech without it being prerecorded, or spoken. Again, surely an unprecedented
example of a machine lifting itself by its own bootstraps. It would have been an address spoken by a
computer, about the history of how it came to exist. Since writing this article, another earth-changing
development has come upon the scene. In November of last year (1979), a company offered a board
for less than $2000 which can be taught to recognize from 40 to 100 spoken words. Now the machine
can listen, understand, and then use the voice mentioned above to reply. Before long, will we be able
to tell the difference between the machine and people?
FOOTNOTES
* Registered trademarks, IBM Corporation, Armonk, New York.
** Registered trademarks, Digital Equipment Corporation, Maynard, Massachusetts.
*** Registered trademark, MITS Incorporated, Albuquerque, New Mexico.
t Beginners All-Purpose Symbolic Instruction Code, Dartmouth College, Dartmouth,
Pennsylvania.
t t Registered trademark, Texas Instruments, Dallas, Texas.
10
REFERENCES
1. A History of Mathematics, J. F. Scott, Taylor and Francis, London. 1975.
2. Implementing the ENIAC, John W. Mauchly. CHEMTECH, December 1975. pp. 722-727.
3. Ibid.
4. Von Neumann, J., Goldstine, H., On the Principles of Large Scale Computing Machines, Vol. 5.,
A.H. Tanb, Editor. Macmillan, New York, New York. 1963.
5. Microprocessor Interfacing Techniques, Austin Lesea and Rodney Azks, Sybex. 1977. p. 17.
SALES - EXCHANGES - HELP WANTED
TEKTRONIX GRAPHICS DISPLAY
I have a Tektronix Graphics Display Terminal (4010-1) and Hard Copy Unit (4610) that I would like to
exchange for other terminal equipment. Anyone interested in obtaining these at a significant discount
over the manufacturer's price, please contact Jim Karr, 102a Vivarium, 333-1633 before March 12 (or
after March 31).
OFF-LINE's Mailing List
If you wish to be placed on our mailing list for future issues of OFF-LINE, if you wish to be
removed from the list, or if you wish to enter an address correction, please complete and
return this page. (Current subscribers are kept on the mailing list until a specific request for
removal is received, or until a mailing is returned as undeliverable.)
Check one: □ New subscriber
□ Removal request
□ Address correction
Name:
Address:
CAMPUS or Zip Code
(If address correction, give old address and zip code below.)
Comments:
RETURN TO: OFF-LINE
150 Digital Computer Laboratory
University of Illinois at Urbana-Champaign
1304 West Springfield Avenue
Urbana, Illinois 61801
LU
r\>-
-Y
-><
aia:
>- 1-1
• -t < IS)
us: -5
— <<
Computing
Services
Office
Off-Line
University of Illinois at Urbana-Champaign
VOL. 9, NO. 4 April 1981
EDITOR: Lynn Bilger
PHONE: (217) 333-6236
139 Astronomy Building
1011 W. Springfield
Urbana, Illinois 61801
Page
1
3
3
5
6
7
8
16
17
17
17
18
Contents
POLICY
THE UBRARY 7.7 T
APR 17 19S1
UNIVERSITY GF II L
Research Board Announcement:
New Computer Allocation Mode (SARA)
Research Board Deadline
for Departmental Allocation Requests
CSO Short Courses
SYSTEM NOTES
February Reliability Reports
Page Printing System (PPS) Now Available
CYBER
Update on Low-cost Weekend Rates
Easy Graphing
IBM
Reduction in 2314 Disk Drive Service
Updates for IBM 3101 Operator Manual
DOCUMENTATION
Revised CSO Documentation
SALES - EXCHANGES - HELP WANTED
Half-time Research Assistant
Research Programmer - Temporary
CSO DIRECTORY - STAFF AND SERVICES
Administrative
Director
Business Manager
Secretary
User Accounting
Distribution Center
Systems Consulting
Statistical Services Consulting
Text Processing Consulting
Terminal Repair Service
CYBER Dial-up Numbers
Asst Dir User Services
Asst Dir Systems and Operations
Asst Dir Statistical Services
and UNIX System
Asst Dir Development
Asst Dir Engineering and
Hardware Selection
Documentation
CYBER-1BM Operations
UNIX Operations
Telecommunications
Laboratory Support Project
RJE Operations North
RJE Operations South
RJE Sites
Agriculture
Chemistry
Commerce West
CRH Snack Bar
DCL Routing Room
Electrical Engineering
Florida Ave Res Hall
Illinois St Res Hall
Mechanical Engineering
Psychology
Social Science
George Badger
150
DCL
333-4103
Stanley Rankin
150
DCL
333-6530
Joyce McCabe
150
DCL
333-1637
vare Support
1208
W Springfield
333-7752
1208
W Springfield
333-6760
166
DCL
333-6133
65
Comm West
333-2170
207
Astronomy
333-7318
150
DCL
333-0969
CYBER 175
110-300
baud
333-4000
CYBER 175
1200
baud
333-4001
CYBER 174
110-300
baud
333-4004
Robert Penka
173
DCL
333-4709
Sandra Moy
177
DCL
333-4703
Larry Sautter
91
Comm West
333-2170
J. M. Randal
120
DCL
333-9772
Cliff Carter
195
DCL
333-3723
Lynn Bilger
139
Astronomy
333-6236
Jack Knott
194a
DCL
333-6562
Debbie Weller
171
DCL
333-8150
Tom Kerkering
164
DCL
333-0816
Mike Gardner
164
DCL
333-7904
Rex Duzan
162
DCL
333-6285
Don McCabe
1208
W Springfield
333-2171
333-7752
M103
Turner Hall
333-8170
153
Noyes Lab
333-1728
70
Comm West
333-4500
120
Snack Bar
333-1851
129
DCL
333-6203
146
EEB
333-4936
FAR
333-2695
ISR
333-0307
65
MEB
333-1430
453
Psych Bldg.
333-7531
202
Lincoln Hall
333-0309
OFF-LINE is the monthly newsletter of the Computing Services Office at the University of
Illinois at Urbana-Champaign. Unless otherwise indicated, permission to reprint is freely
granted, provided that ihe author, if named, and the Computing Services Office (CSO) are
credited.
CSO operates a CDC CYBER 175 with 256K words of central memory and a CDC CYBER
174 with 13 IK words of central memory. The 175 and 174 run under the NOS Operating
System and share 512K words of ECS. The 175 serves over 200 simultaneously active text
and graphics terminals and the 174 serves over 100 simultaneously active terminals. CSO
also operates an IBM 4341 with 4 million bytes of memory running HASP-OS/MVT under
VM. In addition, CSO operates a DEC PDP-11/50 with 252K bytes of core driving a GSI
CAT-8 phototypesetter and a DEC VAX 11/780 with 2 million bytes of core, both running
under the UNIX Operating System.
POLICY
RESEARCH BOARD ANNOUNCEMENT:
NEW COMPUTER ALLOCATION MODE (SARA)
The following article explains a new option to be
used by the Research Board for computer alloca-
tions. A new form, Form X, will be used in sub-
mitting requests for these funds. This form will
be included in the package that the departments
will receive from the Research Board. The second
article contains the deadline dates for submission
of all requests.
General Description
With this round of allocations, an experiment will be initiated
to reduce the paperwork involved in making occasional use of
campus computing facilities. For those faculty and dissertation
students whose use is less than about 1000 service units per
year, an automatic and recurring allocation will be offered as an
alternative to the present semiannual allocation request pro-
cedure.
The effect of this change will be to exempt those users electing
this option from the standard allocation review process. Based
on current usage patterns, a significant reduction in administra-
tive work should result if qualifying users accept this alterna-
tive. It should be emphasized that the main purpose of this
change is to simplify the campus allocation procedures. J_t does
not increase the total resources provided by CSO. The total com-
puter allocations must continue to follow the managed rate of
growth established in the last three years.
The success and continuation of this experiment will depend on
user reaction. The new allocation alternative assumes that small
users will continue to use the computer according to the present
pattern. Under these circumstances, the new alternative should
prove much more convenient for smaller users as well as for their
departmental units. However, CSO remains responsible for pro-
tecting the general quality of available computer services and
for maintaining the financial integrity of the system. If either
of these requirements is jeopardized by this experiment, it will
be necessary to make compensating adjustments.
How It Works
For concise identification, the new allocation mode will be
referred to as SARA (Small Automatic Recurring Allocation).
Under the SARA alternative, any faculty member may simply sign up
(see next section); all processing from that point on will occur
automatically. An initial allocation of 100 service units will
be given and each week 30 service units will be added. Accumula-
tion will be limited to 200 service units, but will carry forward
continuously without expiration as long as the account is used.
Users whose patterns of computer use are incompatible with this
allocation mode should make a specific request for computer time
through their departments. The weekly allocation schedule
described here will not be modified for individuals.
How to Apply
To start this new system off, the department may list any faculty
members who wish to elect this alternative on Form X, but exclude
them from the departmental request on Forms A and B. No justifi-
cation form (Form A) is necessary for those choosing this option.
No further action is needed in future periods unless the user
wishes to convert to the standard allocation system and be
included in the departmental request.
Rules for Users
The SARA alternative is intended as a substitute for, rather than
a supplement to, the departmental allocation process. Thus,
there are several conditions on its use.
9 For those who choose to use SARA, no other campus allocation
will be given (i.e., a user can't be in both the departmen-
tal pool and the SARA pool).
9 SARA allocations are for individuals and are not transfer-
able to other users.
» Use of SARA allocations will not be charged against depart-
mental allocations.
» SARA allocations will become available on June 28, 1981.
» User status under SARA will be inactivated after six months
of no use.
Graduate students, at the time when they are working on their
thesis or dissertation research, will also be eligible for SARA
allocation, although they cannot be on both this and the "free
student" account (PSB2*1^). (CS0 will remove any user from PSB1^
who is given an allocation of this type.) To include a
dissertation or thesis student user, please identify both the
dissertation/thesis committee chairman and the student on Form X.
After this initial group of faculty and students is activated
(June 28, 1981), other qualifying users may request this alloca-
tion mode by simply filling in a request form and presenting it
to CSO.
One potential problem raised by this easier access to campus sup-
port is that it might divert available grant and contract funds
of small users from paying for CSO services. This income is
absolutely essential to CSO. Therefore, SARA usage will be
closely monitored to determine whether levels of support from
departments as a whole decline. Additional steps to encourage
users to use external funds in partial support of CSO service
charges will be announced soon.
RESEARCH BOARD DEADLINE
FOR DEPARTMENTAL ALLOCATION REQUESTS
The Research Board has established an April 15, 1981 deadline for
the submission of all departmental requests for research computer
allocations and for submission of requests for SARA allocations.
This deadline affects allocations for the period June 28, 1981
through December 22, 1981.
Research Board allocations are expected to support faculty
research, thesis and dissertation research. Departmental
requests and the allocations they subsequently receive are based
on individual user requests. Those persons who will need
research computer time for this allocation period should be sure
to submit their requests to their department via the Research
Board forms. These forms and further instructions are available
through the University departments.
CSO SHORT COURSES
CSO is offering the following short courses during the remainder
of the 1981 spring semester to acquaint people with our facili-
ties and the Control Data Corporation (CDC) CYBER 17*» and CYBER
175 computer systems. To register for a course, either come in
person to Room 150 DCL or call 333-6630.
Registration is free and limited to 30 people in some classes.
If all available classes on a topic are full, leave your name on
our waiting list. We will call you if an opening occurs. The
courses being offered are as follows:
1. INTRODUCTION TO THE CYBER COMPUTER SYSTEM
This course is intended for the first-time user of the CYBER
computer system. One class (3 sessions) is being offered:
April 6, 8, 10 11-12 AM, Room 239 DCL
2. NCAR PLOT PACKAGE
This class is an overview of the facilities and general use
of the NCAR plotting package. NCAR allows 3-D plotting, con-
tour plots and world map projections. Familiarity with FOR-
TRAN and the CYBER is assumed. One class (1 session) is
being offered:
April 9 2-3 PM, Room 239 DCL
3. EASY GRAPHING
This is a high-level interactive plotting program for X-Y
plots, bar charts and pie charts. Its English-like commands
require no programming experience to generate plots. Fami-
liarity with the CYBER is assumed. One class (2 sessions) is
being offered:
April 21, 23 2-3 PM, Room 239 DCL
4. INTRODUCTION TO RNF
This is a beginning-level introduction to using the RNF text
formatter, used for preparation of letters, manuals and other
documents. Topics will include tabbing, underlining, mar-
gins, spacing, paragraphing, and justification. Familiarity
with the CYBER is assumed. One class (3 sessions) is being
offered:
April 6, 8, 10 3-4 PM, Room 239 DCL
5. INTERMEDIATE RNF
This class assumes basic knowledge from the RNF introduction.
Topics will include macros, variables, arrays and applica-
tions. One class (3 sessions) is being offered:
April 13, 15, 17 3-4 PM, Room 239 DCL
6. CURVE FITTING
This class will cover the mathematical software available to
do curve fitting. One class (3 sessions) is being offered:
April 6, 8, 10 10-11 AM, Room 239 DCL
SOLVING LARGE LINEAR PROBLEMS
This class will cover the mathematical software used to
tackle such large problems. One class (3 sessions) will be
offered:
April 27, 29, May 1
3-4 PM, Room 239 DCL
8. SAS
This is an introductory course on the SAS statistical pack-
age. Course topics will include:
Preparation of data for input
The SAS 'DATA' step
»
Manipulating the input data: modifying variables, creat-
ing new variables, deleting observations or variables
• Sorting data
• Obtaining basic statistics, frequencies, plots
• Analysis of variance, correlation, regression
One class (3 2-hour sessions) is being offered:
April 6, 8, 13 7-9 PM, Room 241 Comm West
SYSTEM NOTES
FEBRUARY RELIABILITY REPORTS
CYBER 175
16 recoverable interruptions
8 non-recoverable interruptions
9 of these were for more than
15 minutes
MTBF = 21 hours
MTR = 21 minutes
Availability: 97. 2% of sched-
uled uptime
Major cause of downtime related
to ECS and disk problems.
CYBER 174
11 recoverable interruptions
3 non-recoverable interruptions
9 of these were for more than
15 minutes
MTBF = 36 hours
MTR = 36 minutes
Availability: 98.2$ of sched-
uled uptime
Major cause of downtime related
to ECS and disk problems.
IBM 43^1
10 interruptions
5 of these were for more than 15 minutes
MTBF = 63 hours
MTR =58.9 minutes
Availability: 98.2$ of scheduled uptime
Major cause of downtime was related to
software problems.
MTBF = Mean Time Between Failures
MTR = Mean Time to Repair
PAGE PRINTING SYSTEM (PPS) NOW AVAILABLE
CSO has made arrangements to print users' output on the Honeywell
Page Printing System (PPS) installed at the campus office of
Administrative Information Systems and Services (AISS). This
machine uses an electrostatic printing process to produce output
on 8.5 by 11.5 inch sheets of paper at a rate of 90 pages per
minute. Lines are oriented along the 11.5 axis. The top of each
sheet is punched to fit a standard three-ring binder. The char-
acter set used is identical to that on the IBM printers at DCL.
Print quality is very good.
IBM users can request that their output be printed on the PPS by
including the ID card
/»ID F0RMS=PPS
in their job. CYBERS users can direct a file to the PPS by using
the /F0RMS=PPS/RJE=L0CAL options on the PRINT command. For exam-
ple:
PRINT/F0RMS=PPS/RJE=L0CAL, local file name.
The usual paper-saving measures are applied to PPS output. Users
wishing to have page ejects inserted in their output rather than
the normal substitute of three blank lines must specifically
request this processing. To request this, IBM users should
include the ID card
/•ID EJECT=YES
in their job. CYBER users should use the /EJECT option on the
PRINT command.
Users requesting PPS printing should be aware of several limita-
tions:
» The PPS can print at most 64 lines/page. Pages containing
more than 64 lines will overflow onto a second page.
» Only the following carriage control characters will be
honored:
single space (blank)
page eject (1)
double space (0)
triple space (-)
other carriage control characters will cause unexpected
results. In particular, note that overprinting is not pos-
sible.
Jobs to be printed on the PPS are dumped to tapes which are car-
ried to AISS for processing according to the following schedule
(Monday through Friday only):
Jobs dumped to tape Output available for pickup
at the DCL Routing Room
8:00 AM 2:00 PM same day
9:30 PM 9:00 AM following day
The charge for using the PPS is identical to the charge for using
the line printer.
CYBER
UPDATE ON LOW-COST WEEKEND RATES
Based on the utilization between January 20, 1981 and February
28, 1981, the effect of low-cost computing on the weekends was a
little better than expected. About 19 percent of the work was
billed at the low-cost rates. The largest benefit went to users
of Research Board allocations. This benefit more than offset the
decreases in allocations made in anticipation of the effect.
EASY GRAPHING
Easy Graphing is here! It is running on the CYBER 174 and CYBER
175. The English-like commands used in the Easy Graphing package
require no programming knowledge and allow the user to produce
plots, bar charts and pie charts with little effort. Some
experience with the CYBER system is helpful, however. Experi-
enced users and programmers will also find Easy Graphing a useful
tool for producing plots because of its easy-to-use commands and
its interactive nature.
The program, as purchased from Tektronix, supports the Tektronix
4006, 4010 and 4014 graphics terminals. Locally we have modified
Easy Graphing so it can produce a hard copy of the plot on our
ZETA plotters. Figure 1, on the following page, shows some sam-
ple plots which were produced using the Easy Graphing package.
These plots illustrate the typical format of Easy Graphing 's out-
put.
This article describes the production of a simple plot. For
details describing the use of Easy Graphing, the Plot- 10 Easy
Graphing Users Manual is available for purchase at the CSO Dis-
tribution Center, 1208 W. Springfield. A copy of this manual is
available for inspection in the Systems Consulting Office, Room
166 DCL. The manual is tutorial in style, and is strongly recom-
mended for those users who have little experience with graphing.
A summary of the Easy Graphing commands can be found in the Easy
Graphing User's Reference Guide, also available for purchase at
the Distribution Center. This pocket-sized manual may be suffi-
cient documentation for those users more experienced with com-
puter graphics software. The guide is quite handy to have during
a terminal session.
FIGURE 1
THOUSANDS
40
PETROLEUM USE VS. NATURAL GAS ANO COBL
1966-1975
RNNURL REPORT
Net sales
Orders receive
E arn i ngs
Init ioI Sales
U.S. soles
1966 1967 1966 1969 1970 1971 1972 1973 197* 197S
TERR
1974
TERR
THOUSANDS
2S
GENERAL MERCHANDISING. INC.
Fiscal 1977
S .
"Los Angeles
*""'*0etpoi t
^"Average Soles
1 Net 1 ncosie
JAN FEB MAR APR MAT JUN JUL AUG SEP OCT NOV DEC
TEAR
CTCLOID. PROLATE
/
\
v
/
\
\
V
/
\
/
A
A
/
(
\
A
\
\
V
\
I
u
1
HON THE TAX DOLLAR IS SPENT
l is.nl 1977 Est .sots
0 2.
X - T -
S S 7. S 10 12.5 15 17.5 20 22.5 25
2«SIN(T1 T - 1 - 2sC0SIT! T RANGES FROM 0 TO 25
9enef ■ t Parents.
40-39. 97.
Other Opsrat.ons
11-10. 9*
Nst I ntsrsst
6-7. 9X
Nat >onal Osf snss
26-25. 91
10
Let us follow Professor Terri Scalare, an ichthyologist, as she
plots some of her experimental data. Professor Scalare has been
breeding and raising fish, Tropheus moorii and Tropheus duboisi,
from Lake Tanganyika. Every few days the fry (newly hatched
fish) are measured and their size recorded. The data collected
is shown in Table 1 .
TABLE 1
Growth of Tropheus fry*
T.
moorii
day
size (cm)
1
1.5
4
1.8
7
2.0
11
2.3
17
2.7
23
3.0
28
3.1
34
3.2
* Data
from Scheuerm;
pheus",
in Buntbarsche
T.
duboisi
day
size (cm)
7
1.5
11
1.8
17
2.1
23
2.4
28
2.6
34
2.8
American Cichlid Association, No. 71, April 1979.
Professor Scalare now logs onto the CYBER at a graphics terminal,
and issues the following control statements:
GRAB,EZGRAPH.
EZGRAPH.
Easy Graphing begins, and prompts for input with an asterisk (*).
Now Terri enters her first set of data into Easy Graphing vari-
ables. She types the command ENTER, the variable name, and the
values:
ENTER DAY1 1 4 7 1 1 17 23 28 34
ENTER MOORII 1.5 1.8 2.0 2.3 2.7 3.0 3-1 3-2
Then she types the GRAPH command, choosing the X and Y variable
names :
GRAPH DAY1 MOORII
She is rewarded with a plot on her screen that looks like Figure
2.
11
FIGURE 2
3.2
i 1 1 1 i 1 1 1 1 1 1 r
2.5 5 7.5 10 12.5 15 17.5 20 22.5 25 27.5 30 32.5
Excited by the simplicity of use, Terri decides to refine the
format of the graph. For example, she prefers that the range of
the X axis extend from 0 to 35 days and that of the Y axis from 0
to 4 cm. She also wants the plot labelled with a title, subtitle
(Easy Graphing calls this a date) and axis titles on each axis,
so she enters:
XRANGE 0 35
YRANGE 0 4
TITLE 'Growth of Tropheus fry'
DATE 'First 35 Days'
XTITLE 'Days '
YTITLE 'Size in cm'
GRAPH
Terri now sees Figure 3«
12
S
i
z
e
o
m
1 _
0.5 _
FIGURE 3
Growth of Tropheus Fry
F i rst 35 Days
10
15
20
25
30
35
Days
Note that the command GRAPH always causes a new graph to be gen-
erated. This form, given without variable names, redisplays the
previous graph, but with all the newly chosen options included.
For example, to now change the X axis title, Terri enters:
XTITLE 'Days Since Release'
The next time a GRAPH command is entered, this X axis title will
be used. Terri would like to add the second set of data to this
existing graph, so she types:
ENTER DAY2 7 11
ENTER DUBOISI 1.5
ADD DAY2 DUBOISI
GRAPH
17 23
1.8 2
28
,1
34
2.4
2.6 2.8
She now sees Figure 4,
13
z
e
c
m
3.5
3 _
2.5 _
2 .
1.5 _
1 .
0.5 _
FIGURE 4
Growth of Tropheus Fry
F irst 35 Days
Days Smoe Release
Easy Graphing automatically chooses a different line type (dash)
to distinguish between curves (Figure M), but our experimenter
wants a legend to identify the curves. Curve 1 is for T. moorii,
and curve 2 for T. duboisi, so the commands are:
LEGEND 1 "T. moorii'
LEGEND 2 'T. duboisi'
GRAPH
This produces Figure 5.
11*
FIGURE 5
Growth of Tropheus Fry
F i rst 35 Clays
2
e
l
n
c
m
3.5 _
3 _
2.5 _
2 .
1.5 .
1 _
0.5 .
T.
T.
noon i i
dubo i s i
Days Since Release
Terri wants a small symbol to mark each data point on each curve.
Checking her Reference Guide she notes that a diamond is symbol
number 3 and a circle is symbol number 5. As a final touch, the
legend is repositioned to make it more prominent (see the manual
for an explanation of the legend positioning). To produce this
final plot, Terri enters:
SYMBOL 1 3
SYMBOL 2 5
LEGEND 25 80
GRAPH
The final plot then looks like Figure 6.
15
S
z
i
n
3.5
3
2.5 _
2 _
1.5 -
1 .
0.5
FIGURE 6
Growth of Tropheue Fry
F irst 35 Doye
Days Sinoe Release
The Easy Graphing session is terminated with the command:
BYE
This exits Easy Graphing and returns to the CYBER monitor level.
Note that it does not log off the terminal.
Terri Scalare's sample session demonstrates the basic use and
interactive nature of Easy Graphing. Other features include:
» Reading data from files
9 Bar and pie graphs (shown in Figure 1)
• Ability to change an incorrectly entered value
» Arithmetic expression to define a variable as a function of
other variables
16
• Abbreviated form of data entry
» Listing the values of a variable
• Listing a summary of the last graph displayed
» Saving a graph definition in a local file
» Running "procedure" files of Easy Graphing commands
• Log X and/or Y axes
» Axis labelling in months
9 Axis labelling using user specified labels
» ZETA plot output
• Multiple commands can be entered on a single line
» Commands can be broken across many input lines
These features are described in the Plot- 10 Easy Graphing Users
Manual .
IBM
REDUCTION IN 2314 DISK DRIVE SERVICE
The 2314 drives attached to the IBM 4341 have become obsolete
and, thus, are becoming more difficult and expensive to maintain.
CSO plans to remove them from service before they become an
embarrassment.
Known users of 2314 packs will be contacted to make arrangements
for converting their data to other media. If you have any 2314
packs and have not yet been contacted, or want to ensure that
your data is being converted, please write or call:
Mike Randal
120 DCL
333-9772
After July 1, 1981 only those users who have demonstrated a good
reason for continued use of the 2314 drives will be permitted to
use them. More definite plans will be made about removal of the
equipment after a review of our users' needs.
17
UPDATES FOR IBM 3101 OPERATOR MANUAL
If you have purchased an IBM 3101 Display Terminal, updates for
your Operator Reference Manual are available at the Systems Con-
sulting Office, Room 166 DCL.
DOCUMENTATION
REVISED CSO DOCUMENTATION
Reference Guides
RF-0.1 REFERENCE GUIDE LIST — updated; new guides.
RF-0.2 DOCUMENTATION LIST — updated; revisions and new
manuals.
RF-0.3 Job Entry Sites (RJE) — revised to reflect locations
of terminals dedicated to the CYBER 17^ and 175, and the
switchable terminals.
SALES - EXCHANGES - HELP WANTED
HALF-TIME RESEARCH ASSISTANT
CSO has an opening for a half-time Research Assistant to work as
a system consultant. The primary job function is to aid users
experiencing difficulty in using CSO's services. The position
offers a good opportunity to learn a variety of software packages
and languages.
The applicant should be a knowledgeable user of the CYBER system
and must have a working knowledge of FORTRAN. Familiarity with
CSO's IBM services or expertise in areas such as the use of
mathematical libraries or graphics software is desirable.
This position will be available in early summer. If you would
like to discuss it further, contact Robert Penka, Assistant
Director for User Services, 173 DCL (333-^709).
18
RESEARCH PROGRAMMER - TEMPORARY
A consortium of departments is starting an image processing and
graphics facility.
A programmer is needed for one year to do initial software
acquisition and installation. In addition to image and graphics
equipment, this project involves a small dedicated computer
(LSI-11/23) and will probably acquire a VAX.
Emphasis is on physical science applications and experience with
digitized data is valuable.
The position is available immediately and must be filled by May
15, 1981. Salary up to $22,000. Anyone interested should con-
tact George Badger, 150 DCL (333-4104) .
OFF-LINE's Mailing List
If you wish to be placed on our mailing list for future issues of OFF-LINE, if you wish to be
removed from the list, or if you wish to enter an address correction, please complete and
return this page. (Current subscribers are kept on the mailing list until a specific request for
removal is received, or until a mailing is returned as undeliverable.)
Check one: □ New subscriber
□ Removal request
□ Address correction
Name:
Address:
CAMPUS or Zip Code
(If address correction, give old address and zip code below.)
Comments:
RETURN TO: OFF-LINE
150 Digital Computer Laboratory
University of Illinois at Urbana-Champaign
1304 West Springfield Avenue
Urbana, Illinois 61801
i i^-c i a
>» ^-L^-»-
ii— r-^
u 1 ^
c^mo
i/i*»r li— >
i-i -HI-
; i-.r
UK
1-2
"-H -<
DJOC
Aim
j><
/U
<L
m
TJ
C ~T
Computing
Services
Office
Off-Line
Universit^n^linoi^^Urban^Chgmgaigi^
VOL. 9. NO. 5 May 1981
EDITOR: Lynn Bilger
PHONE: (217) 333-6236
139 Astronomy Building
1011 W. Springfield
Urbana, Illinois 61801
Page
1
2
4
7
10
10
10
11
11
11
12
12
12
Contents
POLICY
CYBER Shared-File System
Dial-up Usage Control
Dedicated Lines For Private Facilities
New Accounting Forms
SYSTEM NOTES
March Reliability Reports
CYBER
PPS Output
Self-Service Printer
IBM
New Release of SAS on IBM
STATISTICAL SERVICES
EFAP Statistical Package Removed
DOCUMENTATION
OFF-LINE Mailing List
Ordering CSO Documentation
SALES - EXCHANGES - HELP WANTED
PDP-8 Components, Accessories, and Documentation
DECsystem-10 for Sale
Tapes for Sale
Half-Time Research Assistant
Programmer Wanted
UttlVtKSlTYOFlU
CSO DIRECTORY - STAFF AND SERVICES
Administrative
Director
Business Manager
Secretary
User Accounting
Distribution Center
Systems Consulting
Statistical Services Consulting
Text Processing Consulting
Terminal Repair Service
CYBER Dial-up Numbers
Asst Dir User Services
Asst Dir Systems and Operations
Asst Dir Statistical Services
and UNIX System
Asst Dir Development
Asst Dir Engineering and
Hardware Selection
Documentation
CYBER-IBM Operations
UNIX Operations
Telecommunications
Laboratory Support Project
RJE Operations North
RJE Operations South
RJE Sites
Agriculture
Chemistry
Commerce West
CRH Snack Bar
DCL Routing Room
Electrical Engineering
Florida Ave Res Hall
Illinois St Res Hall
Mechanical Engineering
Psychology
Social Science
George Badger
150
DCL
333-4103
Stanley Rankin
150
DCL
333-6530
Joyce McCabe
150
DCL
333-1637
rare Support
1208
W Springfield
333-7752
1208
W Springfield
333-6760
166
DCL
333-6133
65
Comm West
333-2170
207
Astronomy
333-7318
150
DCL
333-0969
CYBER 175
110-300
baud
333-4000
CYBER 175
1200
baud
333-4001
CYBER 174
110-300
baud
333-4004
Robert Penka
173
DCL
333-4709
Sandra Moy
177
DCL
333-4703
Larry Sautter
91
Comm West
333-2170
J. M. Randal
120
DCL
333-9772
Cliff Carter
195
DCL
333-3723
Lynn Bilger
139
Astronomy
333-6236
Jack Knott
194a
DCL
333-6562
Debbie Weller
171
DCL
333-8150
Tom Kerkering
164
DCL
333-0816
Mike Gardner
164
DCL
333-7904
Rex Duzan
162
DCL
333-6285
Don McCabe
1208
W Springfield
333-2171
333-7752
M103
Turner Hall
333-8170
153
Noyes Lab
333-1728
70
Comm West
333-4500
120
Snack Bar
333-1851
129
DCL
333-6203
146
EEB
333-4936
FAR
333-2695
ISR
333-0307
65
MEB
333-1430
453
Psych Bldg.
333-7531
202
Lincoln Hall
333-0309
OFF-LINE is the monthly newsletter of the Computing Services Office at the University of
Illinois at Urbana-Champaign. Unless otherwise indicated, permission to reprint is freely
granted, provided that the author, if named, and the Computing Services Office (CSO) are
credited. Information in this issue is current as of April 27, 1981.
CSO operates a CDC CYBER 175 with 256K words of central memory and a CDC CYBER
174 with 131K words of central memory. The 175 and 174 run under the NOS Operating
System and share 512K words of ECS. The 175 serves over 200 simultaneously active text
and graphics terminals and the 174 serves over 100 simultaneously active terminals. CSO
also operates an IBM 4341 with 4 million bytes of memory running HASP-OS/MVT under
VM. In addition, CSO operates a DEC PDP-11/50 with 252K bytes of core driving a GSI
CAT-8 phototypesetter and a DEC VAX 11/780 with 2 million bytes of core, both running
under the UNIX Operating System.
POLICY
CYBER SHARED-FILE SYSTEM
At the time of the installation of the CYBER 174 a decision had to be made about the rela-
tionship between the 174 and the 175. In particular, the issue of a shared-file system was the
most critical item from the standpoint of ease of use.
Now that we have used a shared-file system for several months it's worth reviewing what has
been gained and what has been lost. In a shared-file system, such as we have been running, a
single file system is common to both machines so that for most purposes, a user can be com-
pletely indifferent as to which machine he is logged into.
The advantages of a shared-file system fall into several categories. First, the fact that all files
are common to both systems makes it very easy for instructors to deal with files they wish to
share with their students, and to avoid duplication or transfer of files between systems. A
second benefit is that when only one of the systems is in operation, either because of
scheduled activities or unscheduled difficulties, all users can be allowed on a single system
without the necessity of moving files around. This benefit is enhanced since many terminals
are now on a switch which allows access to either system. A third benefit is that software is
common and there is no problem with coordinating the updating of information on two sys-
tems; changes appear on both systems simultaneously as do new files and a variety of other
pieces of information. The final benefit is that the accounting system which is already
extremely complex has no further complexities introduced by running two systems. Current
balances in accounts for instance, appear the same on both systems as do passwords, a list of
eligible users, and all of the many details which could get out of synchronization on two
different systems.
The above advantages are offset by a few problems which are introduced by closely coupling
the systems rather than having them run more or less autonomously. The most significant of
these is that some classes of reliability problems automatically involve both systems. Gen-
erally, the most time consuming reliability problems are those related to the file system, and
the majority of the extended periods of downtime have been related to preserving the
integrity of the file system. This has meant that both CYBER systems were out of service
simultaneously. It is worth noting, however, that only one of these problems was caused by
the running of a shared-file system.
A second area with some problems caused by the close coupling is performance. The way in
which the file systems are intertwined from a hardware point of view does introduce conten-
tion in scheduling of disk and table updates and provides a more difficult environment in
which to optimize system performance. During the initial stages of installation it is clear that
two systems operating independently could support more users than two systems operating in
a closely coupled fashion. Over a period of time this disadvantage will be reduced but there
are inherent difficulties that cannot be completely removed. In general, it is true that a
shared-file system is more complex for operators, engineers, and programmers than two simi-
lar but separate systems. On the other hand it is far simpler for users.
Up to the present time, no serious efforts have been made to balance the load between the
two systems and the emphasis has been on isolating the instructional from the research loads.
Since these do not always peak at identical times, and since vacation periods have a different
impact on the two communities, we will have to look at some questions of load balancing in
the future. The closely coupled systems present a far better environment in which to do this
then an environment which depends on the user making guesses as to which machine is most
attractive at a given time.
It is our assessment that the advantages of close coupling, particularly as they can be
exploited in the future, and as they present simplicity to the user, out-weigh the disadvan-
tages.
DIAL-UP USAGE CONTROL
Earlier this semester, we encountered a substantial amount of difficulty with the peak loading
on the dial-up facilities on both the CYBER 174 and CYBER 175. A number of steps were
taken to make short-term corrections to the situation as well as looking at the problem for the
longer run. A brief review of the steps which have been taken and an assessment of the
situation for the future is presented here.
To reduce the demand for dial-up services, a number of facilities which made use of dial-ups
were converted to hardwired connections through the recently installed switch. Included in
this conversion were a number of terminals at Mechanical Engineering, several of the dormi-
tory sites and a few other sites. This removed the demand from the sites that could almost
be guaranteed to be using dial-up service during peak periods.
As a temporary measure, the "hello" facility was removed simply to keep lines circulating as
rapidly as possible. This facility, however, was restored recently both because the change was
not deemed to be needed currently and because sufficient ways around it have been found so
that it presents an inequitable change.
A number of the heavy users of dial-ups were contacted. They cooperated by releasing lines
whenever possible and, in general, trying to be good users. In addition, some of the CSO
staff use during peak times was reduced.
Finally, at the end of March there was a 15% increase in the number of lines installed. As
the result of these combined efforts, the situation seems to be in a tolerable balance at the
present time. While there are still occasional periods when all lines are busy, it is unusual to
need to make very many attempts before getting a line and it is only for relatively short
periods that this high traffic is in effect.
For the longer run, we are very concerned about the problem. We have recently learned that
during the first nine months of this fiscal year approximately 250 terminals and 150 acoustical
couplers were acquired on campus and we would expect that the majority of these are related
to the use of CSO facilities. In addition to this, while most of these new terminals are prob-
ably aimed at the 300 baud facility, there is a recognition that most of the terminals are capa-
ble at running at higher speed and that people will, at the first opportunity, acquire 1200 baud
dial-up modems. We would hope that the increased demand for 1200 baud can be met
largely by converting existing 300 baud lines. If this is not the case, the problem will be quite
serious because the expense of adding many additional lines is very substantial. We are
caught in the situation where anyone with a few hundred dollars to invest in a terminal
expects several thousand dollars worth of equipment at the other end to be ready and waiting
for their call. Since we are largely an allocation based rather than a hard money economy this
increased traffic does not bring with it any of the money required to increase the facilities.
There are two other approaches which will be necessary in order to address the problem
sufficiently. In the following article, the offering of dedicated lines for private facilities is
presented. In this case the dedicated lines are available on a hard money only basis and gen-
erate the capital required to provide the service. It does this in a way that is financially attrac-
tive to those with hard money. We recognize that this does present a discrimination between
users of the allocation system and those with hard money, but we feel it is not substantially
different than their ability to get private terminals and other more desirable access to comput-
ing.
The graph presented below represents the time-of-day pattern of utilization on our systems
on a busy working day. As you can see, there are extreme swings in the amount of loading
both in total users and in those accessing the system through dial-up terminals. At the present
time, the number of lines available at 300 baud are 66 lines for the CYBER 175 and 16 lines
for the CYBER 174. The system-wide limit of dial-ups plus hardwired is approximately 250
on the CYBER 175 and 112 on the CYBER 174, and these limits generally are not being
approached. Since the peak traffic is far in excess of the average traffic, there is a great deal of
room for smoothing out usage patterns without going beyond reasonable working hours for
faculty and students, to the extent that the number of dial-ups is a rationing mechanism at
smoothing this demand out. We must find ways of introducing fairness into the system so
that the available lines are not dominated unreasonably by a few users. If we are able to
reduce the costs of increasing the number of ports, this too will contribute to allowing growth
within the dial-up user group.
Cyber 175 Terminal fictivity
10:30PM. Rpnl 20 to 10:00PM. April 21
N
u
n
b
e
r
T
e
r
a
i
n
a
1
s
i
n
U
s
e
160
140 .
120
100 .
80
60
40
20 J
fi 1 1 Term i nals
Oialup Terminals
12
The treatment of a problem like this, both in terms of policies for allocating the resource and
policies for determining how much of our budget is devoted to handling the peak load, is
under continuous discussion. We appreciate any comments from the users who are affected
by these policies on how these problems can best be resolved.
DEDICATED LINES FOR PRIVATE FACILITIES
CSO is now offering a cost saving alternative to departments utilizing large amounts of con-
nect time paid for with contract money or departmental funds. For a fee of $1800 per port
you can purchase a dedicated connection to either a specific computer or the switch which
allows access to any of our systems on which you have an account. For an additional fee of
$1200 per port (minimum of 4) CSO will provide the communications link between your
building and DCL with all associated communications equipment. In either case, there will be
a reduction of charges for connect time to a constant $0.50 per hour. It is possible that a
minimum charge per connect hour will be instituted in the future. If so, the minimum
charge on these terminals will reflect the reduction in connect time. All initial installation and
maintenance charges until June 30, 1984 will be paid by CSO. All equipment will be selected
by CSO, but will belong to the department paying the initial fee. A trade-in program will be
maintained so far as is practical to allow for departmental growth.
The $0.50 per hour fee will not be increased before July 1, 1984 but will be subject to
increase then to no more than 70 percent of the connect time charge to other users.
An analysis of the total cost for this approach versus the cost of normal connect time is
presented here along with a diagram (page 6) illustrating possible paths of connection to the
computers.
Dedicated Port Only
You provide:
Terminal
Line to CSO
Modems (as needed) at each end
Maintenance on above
We provide:
Computer ports
Switching (if requested)
Reduction in connect charge to $0.50 per hour
Up to 1200 baud
No busy signal
Dedicated Port With Communications
(minimum of 4 terminals) in a single location
You provide:
Terminal and its maintenance
We provide:
Lines, multiplexors, Modems
Computer ports
Switching (if requested)
Reduction in connect charge to $0.50 per hour
Up to 1200 baud
No busy signal
(A separate pool of computer ports will be maintained for
switched terminals, with sufficient numbers to avoid busy sig-
nals (or their equivalent).)
Connect Hours
Connect Hours
Charges
Charges with
Per Month
July 1, 1981 to
without
dedicated port
Per Port
June 30, 1984
dedicated port*
incl. $3,000 fee
50
1800
2250
3900
100
3600
4500
4800
150
5400
6750
5700
200
7200
9000
6600
250
9000
11250
7500
300
10800
13500
8400
350
12600
15750
9300
* Assumes average rate of $1.25 per connect hour. We do not expect weekday connect rates
to go below this level.
w
H
H
X
P
S o o w 33
S O P w X
w
a
8"
£ O P W X
S O P W X
w
H
pa
H
3S P X
W S M H U PC
o
CM
Pi
pq r^
o
m
CN
S o P w S3
pq en
O
CO
o .h
oo
- ~^~ P
9
S P X
w
w
5
p
33 O P W s
X O P W 32
N
S o p w X
p
H
B
w
H
NEW ACCOUNTING FORMS
Due to the new accounting structure of the University, all PS and account forms now in use
for CSO accounts will be obsolete on July 1,1981. All departments should discard these
forms by this date and start using the new CSO forms. The new forms may be picked up at
the CSO Accounting Office, 1208 W. Springfield.
SYSTEM NOTES
MARCH RELIABILITY REPORTS
CYBER 175 CYBER 174
6 recoverable interruptions 1 recoverable interruption
21 non-recoverable interruptions 13 non-recoverable interruptions
13 of these were for more than 15 minutes 8 of these were for more than 15 minutes
MTBF = 25 hours MTBF = 48 hours
MTR = 18 minutes MTR = 14 minutes
Availability: 98.4% of scheduled uptime Availability: 99.0% of scheduled uptime
Major cause of downtime was related to Major cause of downtime was related to
mainframe errors and disk problems. channel and disk problems.
IBM 4341:
26 interruptions
8 of these were for more than 15 minutes
MTBF = 28 hours
MTR = 11 minutes
Availability: 98.1% of scheduled uptime
Major cause of downtime was related to
disk problems.
MTBF = Mean Time Between Failures
MTR = Mean Time to Repair
CYBER
PPS OUTPUT
As noted in last month's OFF-LINE, CSO has made arrangements to print users' output on
the Honeywell Page Printing System (PPS) installed at the campus office of the Administra-
tive Informations Systems and Services (AISS). Since that article was written the internal
handling of PPS output has been streamlined and early problems centered around tape com-
patibility between CSO and the PPS printer appear to be resolved. We now feel free to
expand use of the PPS printer. We would like to offload the present printing equipment to
reduce print backlogs and extend the useful life of the printing equipment owned by CSO.
Since the PPS printer is so fast (approximately 90 pages per minute) it is ideally suited to
printing large jobs. Starting June 1 we plan to default the printing of all large jobs to the PPS
printer. Since some users may need to continue using the line printer, an override of this
default will be possible. As of this writing the final details surrounding this change have not
been settled. Complete details will be announced in a HEARYE and an RJE bulletin. The
purpose of this article is to give you advance warning of this upcoming change and to urge
you to try the PPS before June 1 . For your convenience the following information is
repeated from last month's OFF-LINE.
The PPS uses an electrostatic printing process to produce output on 8.5 by 11.5 inch sheets of
paper. Lines are oriented along the 11.5 axis. The top of each sheet is punched to fit a stan-
dard three-ring binder. The character set used is identical to that on the IBM printers at
DCL. Print quality is very good.
IBM users can request that their output be printed on the PPS by including the ID card
/*ID FORMS = PPS
in their job. CYBERS users can direct a file to the PPS by using the /FORMS =PPS options
on the PRINT command. For example:
PRINT/FORMS =PPS,local file name.
The usual paper-saving measures are applied to PPS output. Users wishing to have page
ejects inserted in their output rather than the normal substitute of three blank lines must
specifically request this processing. To request this, IBM users should include the ID card
/*ID EJECT = YES
in their job. CYBER users should use the /EJECT option on the PRINT command.
Users requesting PPS printing should be aware of several limitations:
The PPS can print at most 64 lines/page. Pages containing more than 64 lines will overflow
onto a second page.
Only the following carriage control characters will be honored:
single space (blank)
page eject (1)
double space (0)
triple space (-)
Other carriage control characters will cause unexpected results. In particular, note that over-
printing is not possible. Jobs to be printed on the PPS are dumped to tapes which are carried
to AISS for processing according to the following schedule (Monday through Friday only):
Jobs dumped Output available for pickup
to tape at the DCL Routing Room
8:00 AM 2:00 PM same day
9:30 PM 9:00 AM following day
The charge for using the PPS is identical to the charge for using the line printer.
SELF SERVICE-PRINTER
A new self-service printer has been installed in room 131 DCL. This printer is intended for
use by individuals wanting to get a small listing printed quickly and willing to fetch their own
output off the printer. The printer is not tended by an operator. The system will print jobs
on this printer only if print lines number 1000 or fewer. Larger jobs directed toward this
printer are simply held by the system.
The printer is attached to the IBM system through an RJE station and is identical to other
printers in use on the RJE network. It prints both upper and lower case and is rated at 600
lines per minute. Output generated on the IBM system may be directed to this printer by
including the following ID cards:
/*ID PRINT = SELFSV
CYBER output may be directed to this printer by including the switch /RJE = SELFSV of the
PRINT command, e.g.,
PRINT/RJE=SELFSV,local file name.
10
IBM
NEW RELEASE OF SAS ON IBM
SAS 79.5, a statistical package, has been installed on the IBM 4341 and replaces the 79.3 ver-
sion. The new release is the optimizing version and mainly fixes "bugs" found in the last
release. One difference that users may notice is that the minimum core requirement has
increased from 140K to 170K.
Questions concerning use of the SAS package should be directed to the CSO Statistical Con-
sultants in Room 65 Commerce West (333-2170).
STATISTICAL SERVICES
EFAP STATISTICAL PACKAGE REMOVED
Due to the lack of use over the past few years, the EFAP (Exploratory Factor Analysis Pack-
age) will be removed from the CYBER systems on May 14, 1981. If its removal from the
system will cause you any problem, please contact the Statistical Consultants in Room 65
Commerce West (333-2170) immediately.
DOCUMENTATION
OFF-LINE MAILING LIST
Once again, it is time to update our mailing list.
If you are graduating or leaving campus permanently and wish to be deleted from the mailing
list or have OFF-LINE sent to your new address, please fill out the form on the back of this
issue and return to us.
If you plan to be gone for the summer but return in the fall you have two options. We can
flag your name on the list with a HOLD which means no issues will be sent until you return
in the fall, or we can change your address to your summer address. Please let us know which
you desire.
11
If you will be on campus, but your address will change due to dorm closings, please send us
your new address.
ORDERING CSO DOCUMENTATION
The CSO Distribution Center tries to maintain well-stocked shelves of all documentation for
the user community. In the past, these supplies often were drastically depleted when a faculty
member decided to distribute copies of a certain manual or guide to his entire class. Since
reprinting of a document requires at least three to four weeks, we can no longer permit more
than five copies of a document to be taken by any individual unless we have received prior
notification.
We will be happy to make arrangements to provide a ready supply of our documents for
classes. If you plan to use one of our documents for a class, please notify Lynn Bilger (Room
139 Astronomy Building, 333-6236) at least one month in advance and indicate the name of
the document and the number of copies you will need.
SALES - EXCHANGES - HELP WANTED
PDP-8 COMPONENTS, ACCESSORIES, AND DOCUMENTATION
Wanted: Surplus PDP-8 components and accessories. Documentation is also sorely needed.
Please contact Mike Berger, 252 ERL, 333-7452.
DECSYSTEM-10 FOR SALE
I have a DECsystem-10 for sale which will be available in September 1981. A partial listing
of the system includes:
1 KI-10 processor
4 MF10G memory modules
4 RP04 disk drives
1 LP02 line printer
2 TU10 tape drives
1 DC 10 32 line communications control subsystem
The system is under DEC maintenance contract.
12
For further information and system description contact:
Harold Ravlin
Coordinated Science Laboratory
1101 West Springfield Avenue
Urbana, Illinois 61801
(217) 333-0267
TAPES FOR SALE
CSO has several boxes of tapes for sale at $50.00 per box of 10 tapes ($5.00 a tape). These
are tapes which have been used once for CSO dumps. Tapes are all 1600 bpi and 2400-foot
reels. This special will be run only until these boxes of tapes are gone. No returns on pur-
chased tapes. Tapes can be purchased at the CSO Accounting Office, 1208 W. Springfield.
Contact Jack Knott at 333-6562 for further details.
HALF-TIME RESEARCH ASSISTANT
Half-time research assistant needed for statistical programming using SPSS and SAS. Also,
management of several large social science datasets stored on tapes. Ability to keep track of
lots of data and to document carefully.
Half-time for the Summer 1981, with the possibility of continuing during the academic year.
Work is for an economist studying the impact of education and federal policies on the status
of women (compared to men) in the labor market.
Contact Prof. A Beller, 333-7257 or 333-2412 (leave message).
PROGRAMMER WANTED
Must have knowledge of PASCAL, FORTRAN, and SPSS. Work to be on an hourly basis
during the summer with a 1/4 research assistantship for the fall and spring semesters. If
interested, please contact Barbara Tinsley at 333-6371 between the hours of 12 noon to 4PM.
OFF-LINE's Mailing List
If you wish to be placed on our mailing list for future issues of OFF-LINE, if you wish to be
removed from the list, or if you wish to enter an address correction, please complete and
return this page. (Current subscribers are kept on the mailing list until a specific request for
removal is received, or until a mailing is returned as undeliverable.)
Check one:
□ New subscriber
□ Removal request
□ Address correction
Name:
Address:
CAMPUS or Zip Code
(If address correction, give old address and zip code below.)
Comments:
RETURN TO:
OFF-LINE
150 Digital Computer Laboratory
University of Illinois at Urbana-Champaign
1304 West Springfield Avenue
Urbana, Illinois 61801
C >— C X"
J> i— USJ
^£.>—r~ *->
T3 (-0O
(_ 2mc 1
i/l>d>-l
t-i— II-
Z-n-
U>
r-^:
<-> -*
- ^L_
xrr
J><
.<<-
<u
m
T3
Computing
Services
Office
Off-Line
Universit^nilinoi^i^Urbana^hamgmgn^
VOL. 9, NO. 6 June 1981
EDITOR: Lynn Bilger
PHONE: (217) 333-6236
139 Astronomy Building
1011 W. Springfield
Urbana, Illinois 61801
Page
Contents
POLICY
Modifications to the Accounting Practices
SYSTEM NOTES
April Reliability Reports
CYBER
Changes to the P Command
MATHEMATICAL SERVICES
ITPACK
IMSL - New Newsletter
MISCELLANEOUS
Current Weather Data Available
Through Computer Access
SALES - EXCHANGES - HELP WANTED
Error in DECSYSTEM-10 Ad
Grad Assistant
Grad Research Assistant
SPSS Programmer - Part Time
'UN .-
Of,,
TXf
">'0/c
CSO DIRECTORY - STAFF AND SERVICES
Administrative
Director
Business Manager
Secretary
User Accounting
Distribution Center
Systems Consulting
Statistical Services Consulting
Text Processing Consulting
Terminal Repair Service
CYBER Dial-up Numbers
Asst Dir User Services
Asst Dir Systems and Operations
Asst Dir Statistical Services
and UNIX System
Asst Dir Development
Asst Dir Engineering and
Hardware Selection
Documentation
CYBER-IBM Operations
UNIX Operations
Telecommunications
Laboratory Support Project
RJE Operations North
RJE Operations South
RJE Sites
Agriculture
Chemistry
Commerce West
CRH Snack Bar
DCL Routing Room
Electrical Engineering
Florida Ave Res Hall
Illinois St Res Hall
Mechanical Engineering
Psychology
Social Science
George Badger
150
DCL
333-4103
Stanley Rankin
150
DCL
333-6530
Joyce McCabe
150
DCL
333-1637
rare Support
1208
W Springfield
333-7752
1208
W Springfield
333-6760
166
DCL
333-6133
65
Comm West
333-2170
207
Astronomy
333-7318
150
DCL
333-0969
CYBER 175
110-300
baud
333-4000
CYBER 175
1200
baud
333-4001
CYBER 174
110-300
baud
333-4004
Robert Penka
173
DCL
333-4709
Sandra Moy
177
DCL
333-4703
Larry Sautter
91
Comm West
333-2170
J. M. Randal
120
DCL
333-9772
Cliff Carter
195
DCL
333-3723
Lynn Bilger
139
Astronomy
333-6236
Jack Knott
194a
DCL
333-6562
Debbie Weller
171
DCL
333-8150
Tom Kerkering
164
DCL
333-0816
Mike Gardner
164
DCL
333-7904
Rex Duzan
162
DCL
333-6285
Don McCabe
1208
W Springfield
333-2171
333-7752
M103
Turner Hall
333-8170
153
Noyes Lab
333-1728
70
Comm West
333-4500
120
Snack Bar
333-1851
129
DCL
333-6203
146
EEB
333-4936
FAR
333-2695
ISR
333-0307
65
MEB
333-1430
453
Psych Bldg.
333-7531
202
Lincoln Hall
333-0309
OFF-LINE is the monthly newsletter of the Computing Services Office at the University of
Illinois at Urbana-Champaign. Unless otherwise indicated, permission to reprint is freely
granted, provided that the author, if named, and the Computing Services Office (CSO) are
credited. Information in this issue is current as of May 27, 1981.
CSO operates a CDC CYBER 175 with 256K words of central memory and a CDC CYBER
174 with 131K words of central memory. The 175 and 174 run under the NOS Operating
System and share 512K words of ECS. The 175 serves over 200 simultaneously active text
and graphics terminals and the 174 serves over 100 simultaneously active terminals. CSO
also operates an IBM 4341 with 4 million bytes of memory running HASP-OS/MVT under
VM. In addition, CSO operates a DEC PDP-11/50 with 252K bytes of core driving a GSI
CAT-8 phototypesetter and a DEC VAX 11/780 with 2 million bytes of core, both running
under the UNIX Operating System.
POLICY
MODIFICATIONS TO THE ACCOUNTING PRACTICES
Much has been said recently about the revisions to overhead rates related to Federal Govern-
ment contracts. While it is not entirely clear what will be happening, the following is our
impression of how it relates to computing.
On July 1 1981, the University will be converting from an overhead rate based on salary and
wages, approximately 68%, to a lower rate based on "modified total direct cost". The rate on
this will be somewhere in the low 40s but it will apply to things beyond salary and wages.
Another part of the change which has occurred is a requirement imposed by the Federal
Government, that specialized research facilities which have a noticeable impact on the deriva-
tion of the overhead rate must be removed and have separate rates established. For this rea-
son although Computing Services would be one of the direct costs, it will not be billed with
the overhead of approximately 40%, but an overhead figure will be included in the rate. Our
indications are that this will be approximately 8% applied to the computer charges as presently
defined. Thus, unlike normal overhead, it will not be calculated automatically and identified
as a separate charge, but will appear within the rate with no separate identification. It will not
show up as a separate line in monthly statements from the business office.
While the addition of a factor to our rates at first appears painful, the realization that it could
have been a 40% addition rather than 8% makes it somewhat more palatable.
Given the level of confusion regarding these modifications to the accounting practices,
specified by Bureau of the Budget Circular A-21, it is quite possible that this is not the entire
story and we will keep you informed as we get more information.
CYBER 175
SYSTEM NOTES
APRIL RELIABILITY REPORTS
CYBER 174
20 recoverable interruptions
15 non-recoverable interruptions
1 8 of these were for more than 1 5 minutes
MTBF = 20 hours
MTR = 32 minutes
Availability: 96.4% of scheduled uptime
Major cause of downtime was related to
power, console, telex, and disk
problems.
12 recoverable interruption
8 non-recoverable interruptions
10 of these were for more than 15 minutes
MTBF = 28 hours
MTR = 52 minutes
Availability: 96.1% of scheduled uptime
Major cause of downtime was related to
power, telex, humidity and disk problems.
IBM 4341:
15 interruptions
7 of these were for more than 15 minutes
MTBF = 46 hours
MTR = 20 minutes
Availability: 96.0% of scheduled uptime
Major cause of downtime was related to
software problems.
MTBF = Mean Time Between Failures
MTR = Mean Time to Repair
CYBER
CHANGES TO THE P COMAMND
On May 26, 1981, the P command was converted from KCL to CCL. The P command allows
terminal users to enter multiple control statements on one line. This is done by typing
followed by control statements which are separated by semicolons. The program P reads the
remainder of the command line, puts the control statements it finds into a local CCL pro-
cedure file called PDOT with the separating semicolons changed to end-of-lines, and causes
this file to be executed.
If there is no period, right parenthesis, or colon in the command, P places a period at the end
of the command line in the PDOT file. Blanks, semicolons, and commas before a statement
are skipped. Empty or blank statements are ignored. Two semicolons in a row in a statement
are converted into a single semicolon in the file PDOT rather than to a new line. This is how
a user can have embedded semicolons in a command line. If a statement begins with a minus
sign, it is assumed to be a procedure call and the minus is replaced by $BEGIN, in PDOT.
Once the entire comment field (the portion of the line after P.) has been written to PDOT in
this manner, the command
REVERT. PDOT
is placed at the end of the procedure file and the control statement
$BEGIN,,PDOT.
is issued to execute the control statements in PDOT.
Until the user executes another P command, he can execute the same sequence of commands
as often as required by "name calling" the procedure file PDOT. As shown in the examples
below, "name calling" simply involves using the name by itself as the call to the procedure.
If desired, the name of the procedure file can be changed from PDOT to any other valid file
name. This is accomplished by invoking P with the desired names as the first control card
argument.
For example, the control statement
P.GET,PROG,DATA1;R;FTN,I = PROG,L=0;LGO,DATA1,OFL;TYPE,OFL
writes the following statements to the file PDOT
PROC,PDOT.
GET,PR0G,DATA1.
R.
FTN,I = PROG,L=0.
LGO,DATAl,OFL.
TYPE,OFL.
SREVERT. PDOT
The procedure file PDOT is then executed.
This procedure file can subsequently be reexecuted by issuing the following command:
PDOT.
For a second example, the control statement
P(Q) LDSET,USE = $RTN;;;;$,LIB = LIB1/LIB2; LOAD,LGO; NOGO,BIN
writes the following statements to the file Q:
PROCQ.
LDSET,USE = $RTN;;$,LIB = LIB1/LIB2.
LOAD,LGO.
NOGO,BIN.
SREVERT. Q
The procedure file Q is then executed. This procedure file can be re-executed by issuing the
command:
Q
For a third example, the control statement
P,G.MODIFY,P=SYSOPL,LO=E,Z./*CREATE SS/*EDIT SS;-,COMPILE
writes the following statements to the file G:
PROC,G.
MODIFY,P = SYSOPL,LO = E,Z./*CREATE SS/*EDIT SS
$BEGIN„COMPILE.
SREVERT. G
The procedure file G is then executed.
The P command also allows for continuations both when it is invoked from a procedure file
or batch job and when it is entered directly as a command. A line ending in a comma indi-
cates that the following line is to be taken as a continuation. If P is invoked from a procedure
file or batch job, the next line in the procedure file or command file is taken as the continua-
tion. If the P command is invoked directly from time-sharing (by far the most common case),
CONTINUATION: is written at the terminal. The continuation line is entered in response.
Note that since leading commas on a line are deleted a ";," on the end of a P command line
will cause the current command to be terminated and the input on the continuation line to be
treated as a new command.
An example of continuations from time-sharing is as follows:
P,G.R.MODIFY,P = ASSOPL,LO=B,I = ASSMODS,C = ASSCOMP;,
CONTINUATION: COMPASS,l=ASSCOMP,L=LIST,B=BINASS,S = NOSTEXT
This would produce and execute the following file:
.PROC,G.
R.MODIFY,P = ASSOPL,LO=E,I=ASSMODS,C = ASSCOMP.
COMPASS,I = ASSCOMP,L = LIST,B = BINASS,S=NOSTEXT.
SREVERT. G
MATHEMATICAL SERVICES
ITPACK
ITPACK is a collection of routines on the CYBER for the iterative solution of large, sparse
symmetric and positive definite systems of linear equations. It was developed at the Center
for Numerical Analysis at the University of Texas at Austin, initially for systems arising from
partial differential equations and later for general systems. We have a User's Guide which
may be examined in the Systems Consulting Office, 166 DCL.
ITPACK is set up as a library and is accessed by entering:
GRABJTPACK.
Following this, you can compile and run a FORTRAN program which calls ITPACK routines.
There are seven major routines in ITPACK:
JCG Jacobi conjugate gradient method
JSI Jacobi method with semi-iteration (or Chebychev acceleration)
SOR Successive overrelaxation
SSORCG Symmetric SOR with conjugate gradient acceleration
SSORSI Symmetric SOR with semi-iteration
RSCG Reduced system method with conjugate gradient acceleration
RSSI Reduced system method with semi-iteration
The matrix must be stored in a special sparse form. This requires three
vectors: a vector of the non-zero coefficients of the upper triangle of
the matrix, with all coefficients for a given row being together;
an integer vector giving the column indices for each coefficient in the
first vector; and an integer vector giving the indices in the first
vector where each row begins. ITPACK includes routines to help you
set up the matrix in this form.
Routine MA31A in the Harwell library also covers this class of problems.
You may wish to look at this also.
The MATH procedure can be used to obtain a catalog of ITPACK routines,
or the source of a particular routine.
IMSL - NEW NEWSLETTER
We have received a new issue of IMSL User News (formerly their
Numerical Computations Newsletter). It contains an article on
"Confidence Intervals for Predicted Response in Regression Analysis"
explaining the use of IMSL routines for this problem, and includes the
announcement of a package called STPLAN for interactive study planning.
STPLAN has been developed at the University of Texas System Cancer
Center in Houston and will be distributed by IMSL at a price of $100.
According to the announcement, "STPLAN is an interactive computer
program which performs sample size and power related calculations
necessary to plan research studies requiring statistical analysis."
If you are interested in obtaining STPLAN, you can contact IMSL directly at
IMSL
Sixth Floor, NBC Building
7500 Bellaire Blvd.
Houston, Texas 77036
ph. (713) 772-1927
or see Stan Kerr (175 DCL, 333-4715).
A copy of the newsletter is on view in the Systems Consulting Office in 166 DCL, in the gray
IMSL General Information Manual. Some extra copies are available from Stan Kerr. If you
want to be added to the mailing list for the newsletter, tell Stan Kerr, or write to IMSL and
tell them you are a member of a subscribing organization (since we lease their library).
MISCELLANEOUS
CURRENT WEATHER DATA AVAILABLE
THROUGH COMPUTER ACCESS
CSO was recently informed of this weather information program and thought some of our
users might be interested in it. The information in this article was furnished to us by Steven
J. Troester from the Illinois Institute of Natural Resources, Section of Economic Entomology.
Current weather data for approximately 25 Illinois locations is being obtained weekly through
the National Weather Service and a computer file is being maintained for public access on the
University Of Illinois CYBER 175 computer. The file is updated every Tuesday noon, and
consists of current temperatures to date, 10-day forecasts, and deviated 30-year normals for
the remainder of the year. In addition to maximum and minimum temperatures, 4" soil tem-
peratures, precipitation, dew, and relative humidity are available for several locations. A
similar file consisting of 1980 weather observations is also available.
A full description of the files and instructions for accessing them are contained in the text-
edited file obtained by issuing the following CYBER 175 command:
GET,TEXT1/UN = 3NHSNHS
and then printed using the following command:
PRINTJEXT1/CC/EJ/ ASCII.
Computer programs have been developed which use these weather files as follows: (1) to cal-
culate and sum 'heat units' for estimating crop and insect development and (2) to project
development of the black cutworm and predict when cutworm damage will occur (Black
Cutworm Event Simulator). A description of these programs is contained in the text-edited
file 'TEXT2' and can be accessed and printed similarly.
For further information, contact Steven Troester at 333-2359.
SALES - EXCHANGES - HELP WANTED
ERROR IN DECSYSTEM-10 AD
In the DECSYSTEM-10 FOR SALE ad that appeared in the May issue, we erroneously listed
Mr. Ravlin as being the person interested in selling the system. Actually, the Coordinated Sci-
ence Laboratory is the department interested in selling the system, and Mr. Ravlin is the per-
son at CSL to contact if you are interested. His phone number is 333-0267.
GRAD ASSISTANT
The Department of Physical Education is looking for a half-time Graduate Assistant to work
20 hours/week over 11 months. Experience with PDP 11/03 (LSI 11), Fortran and Assem-
bler languages, also Graphics. Background in A-D conversion and data manipulation. Some
electronics helpful.
Salary - $5,500. plus proportional increase for next year.
Submit credentials to:
Terry Ward
21 2D Freer
Campus
GRAD RESEARCH ASSISTANT
A one-third time graduate research assistantship is available for the period September 1981 -
May 1982 for work in digital-analog conversion of audio signals with CSO's UNIX/PDP11/50
system. It is desirable that the applicant have experience with and interest in logic circuits
(particularly, interface circuits), I/O programming, and a higher level structured programming
language. The application area is musical sound synthesis and analysis, the pay is approxi-
mately $3400 for nine months including tuition and fee waiver. For information, contact
Prof. James Beauchamp (333-1089 or 344-3307) or Mr. Pat Kane (333-7886).
SPSS PROGRAMMER - PART TIME
A graduate student is looking for someone with knowledge of statistical programming using
SPSS on CYBER IMMEDIATELY. Work will take approximately 20 hours and the salary is
negotiable. Call Patty at 1-489-2921 anytime.
OFF-LINE's, Mailing List
If you wish to be placed on our mailing list for future issues of OFF-LINE, if you wish to be
removed from the list, or if you wish to enter an address correction, please complete and
return this page. (Current subscribers are kept on the mailing list until a specific request for
removal is received, or until a mailing is returned as undeliverable.)
Check one: □ New subscriber
□ Removal request
□ Address correction
Name:
Address:
CAMPUS or Zip Code
(If address correction, give old address and zip code below.)
Comments:
RETURN TO: OFF-LINE
150 Digital Computer Laboratory
University of Illinois at Urbana-Champaign
1304 West Springfield Avenue
Urbana, Illinois 61801
I If— r—
(_ >
OOLOJJ
i_jc-
Computing
Services
Office
Off-Line
^niversitj^onilinois^^rb^ajChamgaig^
VOL. 9, NO. 7 July 1981
EDITOR: Lynn Bilger
PHONE: (217) 333-6236
139 Astronomy Building
1011 W. Springfield
Urbana, Illinois 61801
THE LIBl
Page
Contents
POLICY
1
1
1
CS Courses Now Using Terminals
Further CSO Moves
SYSTEM NOTES
2
May Reliability Reports
CYBER
3
4
CSO Videotapes Available
FORTRAN Calling PASCAL Procedu
11
16
16
16
Added to UOILIB
MATHEMATICAL SERVICES
Software for Sparse Matrices
MISCELLANEOUS
CILLILUG Meeting
Demonstration of Tektronix Equipment
SALES - EXCHANGES - HELP WANTED
Half-Time Programming Position
CSO DIRECTORY - STAFF AND SERVICES
Administrative
Director
Business Manager
Secretary
User Accounting
Distribution Center
Systems Consulting
Statistical Services Consulting
Text Processing Consulting
Terminal Repair Service
CYBER Dial-up Numbers
Asst Dir User Services
Asst Dir Systems and Operations
Asst Dir Statistical Services
and UNIX System
Asst Dir Development
Asst Dir Engineering and
Hardware Selection
Documentation
CYBER-IBM Operations
UNIX Operations
Telecommunications
Laboratory Support Project
RJE Operations North
RJE Operations South
RJE Sites
Agriculture
Chemistry
Commerce West
CRH Snack Bar
DCL Routing Room
Electrical Engineering
Florida Ave Res Hall
Illinois St Res Hall
Mechanical Engineering
Psychology
Social Science
George Badger
150
DCL
333-4103
Stanley Rankin
150
DCL
333-6530
Joyce McCabe
150
DCL
333-1637
rare Support
1208
W Springfield
333-7752
1208
W Springfield
333-6760
166
DCL
333-6133
65
Comm West
333-2170
207
Astronomy
333-7318
150
DCL
333-0969
CYBER 175
110-300
baud
333-4000
CYBER 175
1200
baud
333-4001
CYBER 174
110-300
baud
333-4004
Robert Penka
173
DCL
333-4709
Sandra Moy
177
DCL
333-4703
Larry Sautter
91
Comm West
333-2170
J. M. Randal
120
DCL
333-9772
Cliff Carter
195
DCL
333-3723
Lynn Bilger
139
Astronomy
333-6236
Jack Knott
194a
DCL
333-6562
Debbie Weller
171
DCL
333-8150
Tom Kerkering
164
DCL
333-0816
Mike Gardner
164
DCL
333-7904
Rex Duzan
162
DCL
333-6285
Don McCabe
1208
W Springfield
333-2171
333-7752
M103
Turner Hall
333-8170
153
Noyes Lab
333-1728
70
Comm West
333-4500
120
Snack Bar
333-1851
129
DCL
333-6203
146
EEB
333-4936
FAR
333-2695
ISR
333-0307
65
MEB
333-1430
453
Psych Bldg.
333-7531
202
Lincoln Hall
333-0309
OFF-LINE is the monthly newsletter of the Computing Services Office at the University of
Illinois at Urbana-Champaign. Unless otherwise indicated, permission to reprint is freely
granted, provided that the author, if named, and the Computing Services Office (CSO) are
credited. Information in this issue is current as of June 26, 1981.
CSO operates a CDC CYBER 175 with 256K words of central memory and a CDC CYBER
174 with 13 IK words of central memory. The 175 and 174 run under the NOS Operating
System and share 512K words of ECS. The 175 serves over 200 simultaneously active text
and graphics terminals and the 174 serves over 100 simultaneously active terminals. CSO
also operates an IBM 4341 with 4 million bytes of memory running HASP-OS/MVT under
VM. In addition, CSO operates a DEC PDP-11/50 with 252K bytes of core driving a GSI
CAT-8 phototypesetter and a DEC VAX 11/780 with 2 million bytes of core, both running
under the UNIX Operating System.
POLICY
CS COURSES NOW USING TERMINALS
At the close of the spring semester, the campus reached an important milestone in the use of
computers. For the last time, the introductory courses in Computer Science were taught
using punch cards. Beginning this summer, all of those courses will have been converted to
the use of terminals, some on the IBM and some on the CYBER systems. This represents a
major investment in terminals and other support because of the very large number of stu-
dents involved.
With this changeover we will be drastically reducing the number of keypunches at our public
sites. Since a large percentage, well above half, of all the cards read this past year were
involved with the elementary courses, the availability of keypunches for those who still need
them should be adequate. Initially we will be leaving a few keypunches at each of our RJE
sites although this may be adjusted as the traffic indicates.
In order to keep the problem due dates in these elementary courses from seriously impacting
facilities, the terminals related to elementary computer science will be separate and identified.
These terminals will not be usable by other customers but on the other hand, the general col-
lection of public terminals will not be usable by these students. We expect utilization on
these terminals to be extremely high and also to be controlled through a reservation system
to insure that all students get fair access. Terminals to support these courses will be located
in Mechanical Engineering, Commerce, and the Snack Bar, with a few terminals at other loca-
tions.
FURTHER CSO MOVES
Over the past few years CSO has been rearranging the locations of a number of its activities
and attempting to consolidate its professional staff in DCL. As you are aware, we share the
building with the department of Computer Science and the pressure for space in the building
has constantly increased as the programs of the two departments have expanded. By the
opening of school in the fall, a number of further changes will have occurred. Most notable
among these will be the movement of the consulting office to 1208 W. Springfield where it
will be located with manual and supplies distribution and user accounting. We think that
when the repairs to this building are complete that most of the user services for which you
have had to visit DCL will be in a single location with a good operating environment.
Another change which will affect a few of you is that the terminal repair service will be
located in Room 209 of the Astronomy building. This should affect only those who are
bringing in terminal equipment to be repaired or who are picking it up.
We will be continuing the trend for user terminals and other services to be provided at sites
other than DCL. Because of the requirement for physical proximity of our high speed
printers the bulk printing capacity will remain in DCL at this time, although it will be moving
into room 16 in the basement. Also the public terminal room will be placed adjacent to it in
the basement and the number of terminals at this location will be somewhat reduced. We
anticipate that the personal computers which have been available in the DCL terminal room
will be located somewhere else by fall but will continue to be available. The hours of service
will continue as they are at the present time.
The precise timing of these changes is not known because of the need for some remodeling at
both Astronomy and 1208 W. Springfield and then after areas in DCL have been vacated
there will be some repair work here. We hope to have it all completed within the next three
to four months and to have those services which affect the users not be disrupted.
SYSTEM NOTES
MAY RELIABILITY REPORTS
CYBER 175 CYBER 174
20 recoverable interruptions 1 1 recoverable interruption
6 non-recoverable interruptions 5 non-recoverable interruptions
9 of these were for more than 15 minutes 10 of these were for more than 15 minutes
MTBF = 25 hours MTBF = 42 hours
MTR = 10 minutes MTR = 13 minutes
Availability: 98.4% of scheduled uptime Availability: 98.9% of scheduled uptime
Major cause of downtime was related to Major cause of downtime was related to
telex, mass storage and disc problems. telex, humidity and disc problems.
IBM 4341:
24 interruptions
12 of these were for more than 15 minutes
MTBF = 30 hours
MTR = 25 minutes
Availability: 97.9% of scheduled uptime
Major cause of downtime was related to
disk problems.
MTBF = Mean Time Between Failures
MTR = Mean Time to Repair
CYBER
CSO VIDEOTAPES AVAILABLE
CSO has produced a series of eight videotapes which introduce the novice to the CYBER Sys-
tem. A viewing guide containing the major pictorials used in the series is available and can be
used to facilitate note taking.
The title and a brief synopsis of each of the videotapes is given below. Running time is 10-15
minutes for each videotape.
Introduction to Computing at CSO
A brief look at the steps required to solve a problem using a computer and some
of the hardware used.
Using a Terminal
A description of the physical operation of a terminal and some of the keys that
have a special meaning to the CYBER.
Introduction to CYBER Time-Sharing
A tutorial on the logging on and off of the CYBER.
File Usage - Local Files and Indirect Access to Permanent Files
An introduction to CYBER files and the commands used to manipulate them.
Introduction to ICE Text Editing
A tutorial on entering and modifying files with ICE.
Running a FORTRAN Program - Concepts
A discussion of the concepts of compilation, loading and execution.
Running a FORTRAN Program - The PROGRAM Statement
A discussion of the PROGRAM statement and its relation to files accessed by the
program.
Running a FORTRAN Program - Control Statement
A discussion of the control statements used to compile, load, and execute a FOR-
TRAN program.
Anyone can view these videotapes by going to the Undergraduate Library in person to make a
reservation for use of the videotape equipment. Ask for a copy of the viewing guide when
you checkout the videotape for viewing.
Copies (Betamax format) of these videotapes are available for loan from CSO to any instruc-
tor wishing to use them in class. They were effectively used in this environment several
times in the past with the instructor stopping the playback equipment whenever he/she
wished to elaborate further or questions arose from the class.
To borrow a videotape for classroom use and obtain copies of the viewing guide for classroom
distribution, call Scott Lathrop (333-6618). If you do not already have access to the required
videotape equipment, Betamax viewing equipment can be borrowed from the Office of
Instructional Resources (333-3690).
FORTRAN CALLING PASCAL PROCEDURE
ADDED TO UOILIB
PASCLL, a routine which allows a FORTRAN routine or program to call a PASCAL pro-
cedure or function has been added to UOILIB. This routine allows those who wish to use the
programming capabilities of PASCAL (including the recursive and the dynamic structure
capabilities) to also use the I/O capabilities of FORTRAN. This can be accomplished by hav-
ing a FORTRAN main program call a PASCAL routine via PASCLL which in turn does all of
its I/O through external FORTRAN routines which it calls. This, in effect, allows a PASCAL
user to use the GCS routines. PASCLL is called as follows:
CALL PASCLL(proc,arr,arrsize,argl,arg2,...,argn)
where:
proc PASCAL routine to be called.
arr Storage area set aside for PASCAL'S stack and heap.
arrsize Integer size of arr.
argl,...,argn
Parameters of call.
proc must be a PASCAL procedure or function (i.e., not a main program) and must be com-
piled with the E+ option. The E+ option can be obtained by using either a /E+ on the end
of the PASCAL control card or with (*$E+ *) in the PASCAL program before the
occurrence of the routine. The name of the PASCAL procedure (signified proc above) must
be in an "EXTERNAL" statement in the FORTRAN calling routine.
arr needs to be dimensioned to be at least arrsize words long.
ansae must be at least 85 + number of words used by the PASCAL program's stack and
heap. The number of words used in the stack is the sum of:
. the space used by the declared variables for that routine and any routines it calls,
. one word for every argument of an external FORTRAN routine called (+1),
. three words for every additional level of procedure call nesting (i.e., run time nesting
rather than compile time nesting).
The parameters in the procedure declaration need to be counted as local variables -- VAR.
FUNCTION, and PROCEDURE parameters being one word each and call by value whatever
their storage size is. Remember to take into account recursive calls when calculating stack
size.
The size of the heap can be calculated as the sum of the number of words in an object
created with a NEW statement (+2) for each object created.
Any parameters you may wish to pass to or from the PASCAL routine must have been
declared VAR in the PASCAL routine called. The PASCAL routine can be compiled with
any Xn option.
PASCLL may be used to call functions also. The function must return a one word value. In
order to use PASCLL to call a PASCAL function, call PASCLL as a FORTRAN function and
include the PASCAL function name as the subroutine argument.
If you are desiring to do FORTRAN I/O (this includes GCS calls), the last main program
loaded must be a FORTRAN main program. If you are desiring to do PASCAL I/O, it must
be a PASCAL main program. So, you can not do both FORTRAN and PASCAL I/O from
the same program. All of this (i.e., a FORTRAN main program loaded last to allow FOR-
TRAN I/O) could be accomplished with the following sequence of control statements:
GRAB,UOILIB.
PASCAL, pascal source,listing/E + .
FTN,I=fortran source, 1 = listing.
LGO.
There is a distinct disadvantage to calling PASCAL as a subroutine from FORTRAN. That is
that there is no error post-mortem dump facilities available to issue diagnostics in case some
problem occurs during the execution of the PASCAL routine. Modular programming and
testing of the PASCAL routines can eliminate many of the problems along this line, though.
It should also be clearly noted that in future releases of either FORTRAN or PASCAL, it
may no longer be possible to interface PASCAL and FORTRAN through a routine such as
PASCLL. Currently PASCLL interfaces FTN 4.8 and FTN 5.1 to PASCAL 3.2.0. So, we
discourage the development of long term programs with PASCLL.
The following is an example of a FORTRAN main program, set of PASCAL subroutines and
control statements to generate a Sierpinski curve using GCS.
The FORTRAN main program:
PROGRAM FORSIER(INPUT,OUTPUT,TAPE5=INPUT,TAPE6=OUTPUT)
EXTERNAL SIERP
INTEGER N,STACK(400),STCKSIZ
WRITE(6,900)
900 FORMATC ENTER RECURSION LEVEL:K")
READ (5,*)N
STCKSIZ = 400
CALL USTART
CALL P ASCLL (SIERP,ST ACK,STCKSIZ,N)
CALL UFLUSH
CALL UPAUSE
CALL UEND
RETURN
END
The PASCAL routines in their main program:
PROGRAM MAIN(INPUT,OUTPUT);
(*$E + *)
PROCEDURE MOVE(VAR X,Y : REAL); FORTRAN;
PROCEDURE UDRAW(VAR X,Y : REAL); FORTRAN;
PROCEDURE SIERP (VAR N : INTEGER);
(* PLOT A SIERPINSKI CURVE OF ORDER N *)
CONST HO = 100.0;
VAR I,J : INTEGER;
HI,H,X,Y,X0,Y0 : REAL;
PROCEDURE A (I : INTEGER); FORWARD;
PROCEDURE B(I : INTEGER); FORWARD;
PROCEDURE C(I : INTEGER); FORWARD;
PROCEDURE D(I : INTEGER); FORWARD;
PROCEDURE A;
BEGIN IF I > 0 THEN
BEGIN A(I-l); X := X + H; Y : = Y-H; UDRAW(X,Y);
B(I-l); X := X + 2*H; UDRAW(X,Y);
D(I-l); X := X+H; Y := Y + H; UDRAW(X,Y);
A(I-l);
END;
END;
PROCEDURE B;
BEGIN IF I > 0 THEN
BEGIN B(I-l); X := X-H; Y := Y-H; UDRAW(X,Y);
C(I-l); Y := Y-2*H; UDRAW(X,Y);
A(I-l); X := X + H; Y := Y-H; UDRAW(X,Y);
B(I-l);
END;
END;
PROCEDURE C;
BEGIN IF I > 0 THEN
BEGIN C(I-l); X := X-H; Y := Y+H; UDRAW(X,Y);
D(I-l); X := X-2*H; UDRAW(X,Y);
B(I-l); X := X-H; Y := Y-H; UDRAW(X,Y);
C(I-l);
END;
END;
PROCEDURE D;
BEGIN IF I > 0 THEN
BEGIN D(I-l)
A(I-l); Y
C(I-1);X
D(I-l);
END;
END;
X := X + H; Y := Y+H; UDRAW(X,Y);
= Y + 2*H; UDRAW(X,Y);
= X-H; Y := Y+H; UDRAW(X,Y);
BEGIN (* SIERP •)
I:=N;
HI := 4.0; FOR J := 1 TO N DO HI := HP2.0;
H := HO/HI;
X0 := 2*H; Y0 := HO-H;
X := X0; Y := Y0;
UMOVE(X,Y);
A(I); X := X+H; Y := Y-H; UDRAW(X,Y);
B(I); X := X-H; Y := Y-H; UDRAW(X,Y);
C(I); X := X-H; Y := Y+H; UDRAW(X,Y);
D(I); X := X+H; Y := Y + H; UDRAW(X,Y);
END;
BEGIN
END.
The CYBER control statements to compile, load, and execute this:
REWIND,FORSIER,SIERPIN,LGO.
GRAB,UOILIB.
GRAB,GCSTEKT.
PASCAL,SIERPIN,LIST.
FTN,I=FORSIER,L=LIST.
LGO.
It is assumed that the FORTRAN source is in FORSIER and the PASCAL source is in SIER-
PIN.
The control cards to obtain an assembly listing of PASCLL are:
GRAB,MATH.
MATH,UOILIB,PASCLL.
REWIND,SOURCE.
COMPASS,I=SOURCE,L=LIST.
PRINT,LIST/CC/EJ.
PASCLL also allows for easy and clean access of low memory by means of global variables.
If you are interested in this, see Daniel Pommert at 179 DCL.
FORTRAN FUTURES
Part 1 • A Brief History Lesson
The first FORTRAN compiler was delivered nearly 25 years ago. It was predicted that "FOR-
TRAN should virtually eliminate coding and debugging," since there would now be an obvi-
ous correspondence between mathematical formulas and the numerical programs that used
them. As we all know, this didn't happen. Instead, higher level languages such as FOR-
TRAN have expanded our horizons, allowing us to program more complex problems and
commit more sophisticated errors.
As FORTRAN replaced machine language in numerical computations, it became possible for
the first time to write programs which could be run on a variety of machines without repro-
gramming. The value of this portability was limited by the fact that each manufacturer pro-
vided a FORTRAN that was slightly different, thus making it difficult for a programmer to
insure that his FORTRAN program would be accepted and interpreted the same way on each
machine. As time went by, these portability problems became increasingly severe. To com-
bat them, a standard for FORTRAN was adopted in 1966, specifying the form and interpreta-
tion of those features of FORTRAN which a programmer could expect to be available on all
FORTRAN processors. Most of the FORTRAN compilers available today were designed to
meet this standard, although most also provide significant extensions to the language beyond
the 1966 standard.
As more time passed, it became clear that the 1966 standard had serious deficiencies. It was
written so tersely that even the people that wrote it could not agree on exactly what parts of it
meant. It contained many restrictions which were based on what could be easily interpreted
when FORTRAN was originally developed and which were no longer necessary. As a result,
most FORTRAN processors offered many more features than were guaranteed to be portable
by the standard, and the standard was losing effectiveness as a portability tool. A revision
undertaken to correct these problems proved to be time-consuming and was not completed
until 1977, with formal adoption of the standard taking place in 1978. Most manufacturers
have revised or rewritten their FORTRAN compilers to bring them into compliance with
FORTRAN 77 (e.g., FTN5 and M77 support the FORTRAN 77 standard), but since this is
also a time-consuming process, such compilers are only beginning to be generally available
and most programmers have yet to feel the impact or benefits of the new standard.
In completing FORTRAN 77, there were some areas of possible revision to FORTRAN
which were not considered because it was felt that to consider them thoroughly would have
resulted in significant further delays in completing the revision. Once FORTRAN 77 was
adopted, X3J3 (the FORTRAN standards committee) immediately began work on a further
revision to be adopted in the mid to late 1980's, including consideration of these previously
ignored areas. The changes between this revision and FORTRAN 77 are expected to be at
least as extensive as those between FORTRAN 77 and the original 1966 standard, so if you
expect to be programming in FORTRAN in the 1990's, the work being done now may have a
significant affect on you.
This article is the first of a series intended to help make you aware of the actions of X3J3 and
the directions their work is taking. If you are elated or appalled by material appearing in
them, or simply interested in learning more details, please feel free to talk with Kurt Hir-
chert, a member of the CSO consulting staff who has been participating in the work of X3J3
for more than two years. He will be happy to answer your questions and assist you in making
your opinions heard by X3J3.
Part 2 - The Shape of Things to Come
Early in its work on the 198X revision of the FORTRAN standard, X3J3 was faced with a
dilemma. On the one hand, FORTRAN could not be allowed to grow much larger without
seriously affecting the likelihood that it could be fully implemented on small computers. On
the other hand, there was great pressure to extend FORTRAN in some very extensive ways,
such as including array expressions and other more advanced array processing facilities. If
X3J3 were to ignore these pressures, it seemed likely that, at least on some larger computers,
these facilities would be implemented anyway in nonstandard and nonportable ways. X3J3's
response to this dilemma was to adopt the "core + modules" model for the next revision.
Under this model, the standard would be partitioned into core language features. Thus,
implementors wishing to provide advanced facilities would allow programs using them to be
portable, at least among processors willing to support such facilities, but implementors on
smaller machines or machines aimed at other markets would not be burdened with the
requirement of supporting these facilities. The key features of this "core + modules" model
are described in the following paragraphs.
The Core -- It is hoped that the core language will be slightly smaller than FORTRAN 77, but
with comparable or even greater power for expressing general programming applications.
This will be accomplished by eliminating some of the redundant capabilities in FORTRAN 77
and by replacing some of its more peculiar features with more regular features which provide
the same functions, but are easier to describe and implement. In the long run, this regulari-
zation of FORTRAN should make it an easier language to learn and use.
Language Extension Modules — Features of an experimental nature or of interest to only a
limited portion of the FORTRAN user community would be described in language extension
modules. The availability of language extension modules isolates the decision to standardize
the use of a particular facility in FORTRAN from arguments that it is to expensive or special-
ized for general implementation, thus extending the portability benefits of standardization to
subcommunities of the users of FORTRAN.
10
Obsolete Features Module (s) -- One of the major problems in producing a revision of a
language standard is that such revisions are rarely 100% upwards compatible with their prede-
cessors. Thus, users of the language must be concerned with questions of how and whether
programs conforming to the old standard can be made to run properly under processors sup-
porting the revised standard. Most vendors provide some combination of conversion pro-
grams and support for the old facilities to simplify such transitions, but the portability of such
transitional programs is often poor. In future revisions of the FORTRAN standard, the
obsolete features module (s) will describe features that when added to those of the core
language will be sufficient to properly execute programs written in conformance to the previ-
ous version's core (which for this revision will be taken to be all of FORTRAN 77). The
availability of processors supporting the obsolete features module (s) should eliminate the
need for "instantaneous" transitions from one standard to the next and should allow transition
periods comparable to the revision cycle of the language (typically 5 to 7 years or more).
Application Facilities Modules -- Standards specifying the interface between FORTRAN and
externally defined applications facilities are becoming increasingly common. For example, the
graphics standards committee is in the process of developing a standard for the names and
arguments of subprograms providing standard graphics services to FORTRAN 77 programs.
The concept of application facilities modules provides a formal recognition of the relation
between such standards and the FORTRAN standard. X3J3 will also provide language
features and administrative control to help resolve situations, such as the use of identical pro-
cedure names for different purposes in different application facilities modules, that would oth-
erwise interfere with the usage of multiple application facilities modules in the same program.
Part 3 - Recent X3J3 Actions
The 78th meeting of X3J3 took place May 11-15, 1981 in Toronto Ontario. Actions at that
meeting included the following:
. Further work was done on a facility to specify the representation of real and complex
numbers in terms of minimum precision requirements rather than in machine depen-
dent terms like single and double precision.
. The criteria to be used for determining which language features are part of the core
language were revised.
• The interaction between array processing facilities and data structuring facilities was
discussed. Special attention was given to the interpretation of a construct which could
be described as representing an array of arrays.
• Further extensions to the array processing facilities were discussed, especially in the
areas of array valued procedure arguments and array valued functions.
• Additional work was done on an internal procedures facility intended to replace and
extend the statement function.
11
. Additional work was done on an enhanced procedure interface facility that would allow
such things as arguments specified by keyword, compile-time checking of the number
and type of arguments to a procedure, and the provision of default values for omitted
arguments.
• Exploratory work was done on a name-directed input/output facility which would pro-
vide functionality similar to NAMELISTs (a non-standard facility), but in a more reg-
ular form.
. There was also discussion of providing a standard input/output form that would offer
the useful functionality provided by the non-standard BUFFER IN/ BUFFER OUT
facility.
. A facility was adopted for using compile-time variables and compile-time control con-
structs to effect conditional compilation and variant subprogram source.
• There was discussion of the issues to be resolved in order to provide a means of speci-
fying all attributes of a FORTRAN entity in a single declaration, as opposed to the
current FORTRAN specification statements, which specify only a single attribute, but
may include several entities.
• As an economy measure, X3J3 has reduced its meeting schedule from five to four
meetings per year, starting in 1982. As a result of this change and several other fac-
tors, the projected schedule for revision of the standard has been updated. It is now
expected that the revision will be in its final form in mid 1983, with final adoption of
the revision taking place in late 1986.
The next meeting of X3J3 will take place August 10-14, 1981 in Los Alamos, New Mexico.
MATHEMATICAL SERVICES
SOFTWARE FOR SPARSE MATRICES
We now have a number of routines on the CYBER for solving large sparse systems of linear
equations Ax=b, for both real and complex A. (A is "sparse" if most of its elements are
zero, typically up to 95% or more.)
For general nonsymmetric A, the routines we have are NSPIV (in UOILIB), a package called
Y12M consisting of several routines (in UOILIB), the Yale sparse matrix package (in
UOILIB), and several routines in the Harwell library (GRAB,HARWELL/S) :
MA28A/MA28B/MA28C for general real matrices, ME28A/ME28B/ME28C for general com-
plex matrices, and an as-yet-unconverted routine MA32AD for applying the frontal method
using disk storage.
12
For symmetric positive definite matrices A, there is MA31A in Harwell, a library of iterative
routines called ITPACK (GRABJTPACK), and several routines in the Yale package in
UOILIB.
Codes that deal with sparse matrices usually use a special data structure, to avoid having to
store a large number of zero elements. For the routines that we have there are four storage
schemes (except for MA32AD, which uses a scheme specially designed for the frontal
method):
la. Store all the nonzero elements of A in one long vector and use two integer
arrays to contain the corresponding row and column indices.
This scheme is used by MA28A/B/C, ME28A/B/C, Y12M, and several other
Harwell routines, and is used for convenient input rather than internal opera-
tion : each routine converts the input to a scheme more like 2a or 2b below for
internal use.
lb. For a symmetric A, use scheme la, but store only the upper triangle of A.
This is used by MA31A.
2a. Store all the nonzero elements of A in a big vector VA, ordered so that the
elements of row i precede those of i + 1; then use two integer vectors I A and
J A so that JA(I) is the column number of the element in VA(I) and IA(K)
gives the position in VA where the elements of row K begin (or equals IA(K-
1) if the row is all zeros).
This scheme is used by NSPIV and by the Yale routines for nonsymmetric
matrices.
2b. For a symmetric matrix A, use scheme 2a but store only the upper triangle of
A.
This scheme is used by ITPACK and by the Yale routines for symmetric posi-
tive definite matrices.
An important fact of life about these data structures is that -- except for the iterative methods
which use them -- they "grow" during the solution process. That is, the arrays you use must
be dimensioned larger than what is necessary to store the initial matrix itself, to allow room
for "fill-in" during elimination. Unfortunately, unless you have analyzed the matrix before-
hand, it is not possible to say just how large the arrays should be, since you don't know how
much fill-in there will be. You can use a similar matrix which you have run before, or do a
"benchmark" on a matrix of smaller size with a similar pattern of elements, to gauge the
amount of storage that may be necessary. If neither of these is possible, proceed with care; it
is possible to have a matrix which, though very sparse, results in huge amounts of fill-in and
requires more memory to solve than you can get.
13
Following are brief descriptions of each piece of software mentioned above, and how to get
documentation on them.
NSPIV was published in Transactions on Mathematical Software (TOMS) volume 4 number 3
(December 1978); the same issue contains a paper by Andrew Sherman explaining the tech-
nique it uses. It is oriented toward solving one system Ax=b and preserves no information
which can be used to solve other systems with the same or a similar matrix. A brief descrip-
tion of NSPIV parameters can be obtained on-line as follows:
GRAB,MATH.
MATH,UOILIB,NSPIV.
PRINT,DOC.
To actually use NSPIV, you must access the UOILIB library, as follows:
GRAB,UOILIB.
Following this, you would compile and run a FORTRAN program which calls NSPIV.
Y12M (in UOILIB) is a package of routines developed at the Regional Computing Center at
the University of Copenhagen in Denmark. The core of the package consists of three rou-
tines : Y12MBE, Y12MCE, and Y12MDE. (These are the names of the single precision ver-
sions; those of the double precision version are Y12MBF, Y12MCF, and Y12MDF. The
documentation of Y12M often refers to routines generically - Y12MB, for instance, refers to
Y12MBE or Y12MBF, whichever is appropriate.) Y12MB re-orders your matrix information
(entered according to scheme la above) to a more convenient form for internal operations;
Y12MC performs the actual reduction, using a special "partial pivoting" strategy designed to
minimize fill-in without sacrificing accuracy (which is not easy!), and saving information as it
goes; and Y12MD uses the saved information from Y12MC to solve for a given right-hand-
side vector. This division of functions makes it easy to reduce a matrix once and then solve
for a whole series of different right-hand-sides. Y12M uses a new idea as part of its reduction
algorithm -- the "drop tolerance". This is a number which you can select, such that new
matrix elements which are generated during elimination, and which are smaller than the
drop-tolerance, are ignored (dropped, not stored, thrown away!). This results in an inaccu-
rate reduction, but the accuracy can usually be regained during solution of Ax = b by doing
"iterative refinement" : the residual vector b-Ax is used as a new right-hand-side to get a
solution which is an incremental improvement to the previous one; an updated solution is
formed, and the process repeated until the solution stabilizes. A special driver routine,
Y12MFE (sorry, no double precision version for this one) is provided which inputs your
problem information, including the drop tolerance, calls Y12MBE and Y2MCE to reduce the
matrix, then calls Y12MDE repetitively for the iterative refinement process. If you don't
wish to use the drop-tolerance, there is a driver Y12MAE (double precision version
Y12MAF) which inputs your problem information, makes default assumptions about all inter-
nal tolerances to be used, and calls Y12MBE, Y12MCE and Y12MDE to solve the problem.
Y12MAE is very short, since all it does is call the other routines; the authors of Y12M
intend that you should modify your own source copy of Y12MAE (or Y12MAF) to suit your
exact needs.
14
To obtain the source or an on-line writeup for any of the Y12M routines, enter the following
CYBER commands:
GRAB,MATH.
GRAB,UOILIB,Y12Mxx.
where xx represents which one you want (AE, AF, BE, BF, etc.). You will be provided with
a local file SOURCE containing complete source code for that routine, and a file called DOC
containing just the writeup portion of the source code. Y12MAE, Y12MCE, Y12MDE, and
Y12MFE contain extensive writeups, derived from the user manual for them written at
Copenhagen. A copy of this manual, is on view at the DCL Consulting Office or can be bor-
rowed from Stan Kerr (175 DCL, 333-4715). A research report, Direct Methods for Sparse
Matrices, by Ole Osterby (who brought the package here) and Zahari Zlatev, can also be bor-
rowed.
The Harwell Library contains one principal package - MA28 - for solving sparse real sys-
tems (ME28 is a complex version of MA28). This package consists of three routines,
MA28A, MA28B, MA28C which divide the work somewhat the same way Y12M does.
MA28A analyzes and factors a given matrix, MA28B factors another matrix of the same or a
similar pattern of nonzero elements as was last handled by MA28A, and MA28C solves
Ax = b for a given b, using information saved by MA28A or MA28B. The technique used is
to try re-ordering the rows and columns of the matrix so as to obtain a block lower triangular
matrix, then use gaussian elimination on the diagonal blocks thus formed. Besides the MA28
and ME28 routines themselves, there is a scaling routine MC19A which can (and should) be
used when the matrix elements differ widely in magnitude.
For a symmetric positive definite matrix A, Harwell routine MA31A can be used. It uses a
method of "pre-conditioned conjugate gradients" rather than a direct elimination method.
Writeups on the Harwell routines can be found at the Consulting Office in 166 DCL. The
library is accessed by the command
GRAB,HARWELL/S.
The /S means HARWELL is a public, shared entity but is not presently a fully supported pro-
duct. Not all Harwell routines are available on the CYBER, because of the necessity of
conversion from their original IBM FORTRAN. Moreover, the library has been compiled
with the new FTN5 compiler (mainly because conversion of many of the routines to FTN5 is
much easier than to the current FTN), so you may need to compile your program with FTN5
or take special measures to make sure everything hangs together correctly (see the consul-
tants!). Source of the available Harwell routines can be had by entering
GRAB,MATH.
MATH,HARWELL,xxx.
where xxx is the name of a Harwell routine. The source is put in a local file called SOURCE.
Some routines - MA28A, ME28A, MC19A, and MA31A in particular - have had extensive
internal writeups added to their source decks (most Harwell routines have no writeups inside
them, so you usually have to consult the printed manual at the consulting office for instruc-
tions on a particular routine).
15
Two papers relating to the Harwell routines are:
"Some Design Features of a Sparse Matrix Code"
by I.S. Duff and J.K. Reid
Transactions on Mathematical Software, March 1979
(This describes the design of MA28.)
"Solving Sparse Symmetric Set of Linear Equations by Pre-Conditioned
Conjugate Gradients"
by N. Munksgaard
Transactions on Mathematical Software, June 1980
(This discusses the MA31A algorithm.)
The Yale sparse matrix package consists of a set of routines for symmetric and nonsym-
metric matrices, with 5 principal driver routines which are used to access them. The routines
have suffered some name changes to make them easily distinguishable from other UOILIB
routines; the drivers are named YALEC, YALEN, YALEO, YALES, and YALET. YALEO
and YALES are the drivers for symmetric problems: YALEO attempts to re-order the matrix
rows and columns to ensure reasonably low fill-in when YALES does the actual elimination.
YALEC, YALEN, and YALET are the drivers for nonsymmetric problems. YALEN and
YALEC are two versions of the elimination algorithm, the former using "uncompressed
pointer storage" and the latter using "compressed pointer storage"; YALEN is faster than
YALEC but uses more memory. YALET is a version which can be used to efficiently solve
for additional right-hand-sides, whereas YALEN and YALEC save such information.
Writeups on the Yale routines can be obtained as follows:
GRAB,MATH.
MATH,UOILIB,YALEO.
MATH,UOILIB,YALES.
MATH,UOILIB,YALEN.
MATH,UOILIB,YALEC.
MATH,UOILIB,YALET.
This will produce one local file called DOC containing writeups on all 5 routines; this file can
then be PRINTed. If you only need one writeup, just do one call to MATH for that routine.
Two reports from Yale describing the routines can be viewed at the DCL Consulting Office,
or borrowed from Stan Kerr; in reading these reports you must take account of the name
changes of the routines.
We are interested in knowing of your particular requirements in solving large sparse systems,
and of your experience with any of the routines mentioned here. If you have any remarks or
suggestions, please send them to Stan Kerr (175 DCL, 333-4715, or TELL,UN = MATHLIB
on the CYBER).
16
MISCELLANEOUS
CILLILUG MEETING
The next meeting of the campus DEC Local Users Group will be July 14 at 3 PM in room
158 of Loomis Laboratory of Physics. There will be at least one floppy/ Winchester disk sys-
tem available for a demonstration. Walt Schneider will give a brief description of his experi-
ences with some of the available systems at 3:30. (Also, see information about demonstra-
tion of Tektronix equipment in the same building the same day.)
DEMONSTRATION OF TEKTRONIX EQUIPMENT
Jim Stallings will demonstrate the 4114 and 4112 models of the Tektronix 4100 series on July
14 in Room 420, Loomis Laboratory of Physics. Demonstrations are scheduled during the
hours: 10:30 AM - noon and 1:30 - 5:00 PM.
SALES - EXCHANGES - HELP WANTED
HALF-TIME PROGRAMMING POSITION
The High Energy Physics Group, Department of Physics is looking for an advanced undergra-
duate or graduate student to work approximately 20 hours per week. Previous programming
experience is required and experience with FORTRAN and Assembler languages is desirable.
(Students who have completed the 200 level sequence in Computer Science may be able to
substitute courses for the experience requirement.)
Computers used by the group include a DEC KI10 running the TOPS- 10 monitor and DEC
LSI-ll/2's and LSI-ll/23's. Members of the programming staff maintain monitors, system
programs and libraries for these computers, do consulting with the users and are also
involved with some microcoded special hardware for the data taking activities of the group.
Applicants should be able to commit to at least 18 months in the position.
Submit resume (it may be an informal one) to:
Jerald Wray
487 Loomis Lab
333-4922
OFF-LINE's Mailing List
If you wish to be placed on our mailing list for future issues of OFF-LINE, if you wish to be
removed from the list, or if you wish to enter an address correction, please complete and
return this page. (Current subscribers are kept on the mailing list until a specific request for
removal is received, or until a mailing is returned as undeliverable.)
Check one: □ New subscriber
D Removal request
□ Address correction
Name:
Address:
CAMPUS or Zip Code
(If address correction, give old address and zip code below.)
Comments:
RETURN TO: OFF-LINE
150 Digital Computer Laboratory
University of Illinois at Urbana-Champaign
1304 West Springfield Avenue
Urbana, Illinois 61801
Oi— I i"0
>■!— LJ/*-1
jSh-|— l-l
\J f~ (/•.
Cismo
<^. J>C 1 ■—
M^r
^- — i-
□ j>
r-H
— ■ -<
uoc
pom
>«t
*>
<c
m
-o
Hr
Computing
Services
Office
Off-Line
Univer^t^nilinoi^^Urbana^hamgajgi^
VOL. 9, NO. 8 August 1981
EDITOR: Lynn Bilger
PHONE: (217) 333-6236
139 Astronomy Building
1011 W. Springfield
Urbana, Illinois 61801
Page
1
2
3
4
5
6
7
8
9
12
13
16
Contents
POLICY
Changes in IBM Disk Policy
Conversion to New University Accounting
Updates on Page Printing System
New Charge Rate For Terminal Repairs
SYSTEM NOTES
June Reliability Reports
CYBER
FETCH Command Updated
QUERY Command Updated
New Version of ICE
FEATURE ARTICLES
Ed Pelg Retires
Special Plotter Output Service
General Plotting Service
DCL Routing Room
FORTRAN Futures
SALES - EXCHANGES - HELP WANTED
Graduate Research Assistant
Hi
»il
SEpo9 mi
CSO DIRECTORY - STAFF AND SERVICES
Administrative
Director
Business Manager
Secretary
User Accounting
Distribution Center
Systems Consulting
Statistical Services Consulting
Text Processing Consulting
Terminal Repair Service
CYBER Dial-up Numbers
Asst Dir User Services
Asst Dir Systems and Operations
Asst Dir Statistical Services
and UNIX System
Asst Dir Development
Asst Dir Engineering and
Hardware Selection
Documentation
CYBER-IBM Operations
UNIX Operations
Telecommunications
Laboratory Support Project
RJE Operations North
RJE Operations South
RJE Sites
Agriculture
Chemistry
Commerce West
CRH Snack Bar
DCL Routing Room
Electrical Engineering
Florida Ave Res Hall
Illinois St Res Hall
Mechanical Engineering
Psychology
Social Science
George Badger
150
DCL
333-4103
Stanley Rankin
150
DCL
333-6530
Joyce McCabe
150
DCL
333-1637
rare Support
1208
W Springfield
333-7752
1208
W Springfield
333-6760
166
DCL
333-6133
65
Comm West
333-2170
207
Astronomy
333-7318
150
DCL
333-0969
CYBER 175
110-300
baud
333-4000
CYBER 175
1200
baud
333-4001
CYBER 174
110-300
baud
333-4004
Robert Penka
173
DCL
333-4709
Sandra Moy
177
DCL
333-4703
Larry Sautter
91
Comm West
333-2170
J. M. Randal
120
DCL
333-9772
Cliff Carter
195
DCL
333-3723
Lynn Bilger
139
Astronomy
333-6236
Jack Knott
194a
DCL
333-6562
Debbie Weller
171
DCL
333-8150
Tom Kerkering
164
DCL
333-0816
Mike Gardner
164
DCL
333-7904
Rex Duzan
162
DCL
333-6285
Don McCabe
1208
W Springfield
333-2171
333-7752
M103
Turner Hall
333-8170
153
Noyes Lab
333-1728
70
Comm West
333-4500
120
Snack Bar
333-1851
129
DCL
333-6203
146
EEB
333-4936
FAR
333-2695
ISR
333-0307
65
MEB
333-1430
453
Psych Bldg.
333-7531
202
Lincoln Hall
333-0309
OFF-LINE is the monthly newsletter of the Computing Services Office at the University of
Illinois at Urbana-Champaign. Unless otherwise indicated, permission to reprint is freely
granted, provided that the author, if named, and the Computing Services Office (CSO) are
credited.
CSO operates a CDC CYBER 175 with 256K words of central memory and a CDC CYBER
174 with 131K words of central memory. The 175 and 174 run under the NOS Operating
System and share 512K. words of ECS. The 175 serves over 200 simultaneously active text
and graphics terminals and the 174 serves over 100 simultaneously active terminals. CSO
also operates an IBM 4341 with 4 million bytes of memory running HASP-OS/MVT under
VM. In addition, CSO operates a DEC PDP-11/50 with 252K bytes of core driving a GSI
CAT-8 phototypesetter and a DEC VAX 11/780 with 2 million bytes of core, both running
under the UNIX Operating System.
POLICY
CHANGES IN IBM DISK POLICY
In the past, datasets residing on CSO's public disk packs which satisfied any of several criteria
were dumped to tape and purged from disk. To have a purged file restored, a user had to
contact the Systems Consultants. With the installation of a disk management package on the
IBM-4341, a dataset purged from disk after August 1, 1981 may be restored to disk with the
RESTORE proc, eliminating the need to contact the Systems Consultants. The RESTORE
proc may be used in the following manner:
II jobname JOB
/*ID (necessary ID card parameters)
II EXEC RESTORE
//SYSIN DD *
RESTORE DSNAME=name[,NEWNAME=newname]
where
name is the name of the dataset to be restored
newname is the dataset name to be assigned to the restored dataset. If the
NEWNAME parameter is omitted, the data set will be restored with its
original name.
More details on the RESTORE proc are available in Reference Guide RF-17.8, available at
any RJE site.
For the present, the dataset purging policy will be unchanged. Once each week datasets which
satisfy any of the following criteria will be archived to tape and purged from disk. Such
datasets:
have not been accessed for more than 30 days,
belong to cancelled PS numbers,
belong to PS numbers which have been inactive for more than 30 days,
are not cataloged (PUBLIC only), or
have names which do not conform to the standard naming conventions at CSO.
With the termination of 2314 setup disk service, there will be a significant increase in the
number of datasets competing for residence on the public disks. To offset this load, more
public disk space will be made available. The above policy will be changed as necessary to
maintain sufficient space on the public packs.
CONVERSION TO NEW UNIVERSITY ACCOUNTING
Our conversion to the new eleven-digit university account number has caused some confu-
sion for users who must specify an account number when dealing with CSO's Accounting and
Distribution Office. If you pay for computer use through an invoice number you are not
affected by this change. If you deal with account numbers as a computer user there are a
number of facts you should know:
. All new account numbers contain eleven digits, a seven-digit root followed by a four-
digit "object code". Roughly, the first seven digits identify a pool of money and the
object code identifies the type of income or expenditure. Use of the object code is
mandatory.
. Each account number of the kind in use before July 1 has been assigned a seven-digit
root under the new accounting system. The University Accounting Division sent each
departmental office a list of account numbers used by the department and the
corresponding new seven-digit roots to be used under the new system. The depart-
mental business office must choose the object codes to be appended to these seven-
digit numbers to form the new account numbers. The Office of Business Affairs sent
each departmental business office a list of recognized object codes and their
definitions.
. Account numbers used to fund computer time at CSO must have one of the following
object codes:
Account numbers representing a research board allocation of computing service
units must have an object code of 5571.
Account numbers representing a campus allocation of computing service units
for class use must have an object code of 5573.
Any other account number must have an object code of 556x where x is
chosen by the department and may be any digit.
. If an account number is to be used to pay for manuals or terminal rental, the depart-
ment must decide how the expense is to be classified and choose the appropriate
object code.
We automatically converted all account numbers in our database to the new format using a
conversion table supplied by AISS on June 25. We used 5560 as the object code for any
account number which was not a research board or class account. If your department wants
to use a different final digit for the object code, you must fill out the necessary account and
PS forms at the CSO Accounting Office to make the change.
If you have any questions relating to the account number conversion, please contact your
department's business office. Our correspondence table between old and new account
numbers is no longer current. Furthermore, with the exception of research board and class
accounts, the departmental office must choose one or more digits of the object code.
UPDATES ON PAGE PRINTING SYSTEM
We are now automatically routing some large print jobs to the Page Printing System (PPS).
Only jobs which run on the local CPUs and which would normally print on standard forms
are affected. Jobs which run at UIC, jobs requesting special forms, and jobs which are routed
to an RJE other than DCL are not affected.
For IBM jobs "large" is defined as more than 30,000 lines of output. For CYBER jobs it is
defined as 4000 or more PRUs of disk space. Both definitions approximate 500 pages of out-
put.
Users can override this default and force a large job to print on the usual line printer by using
the FORMS =NOPPS parameter. From IBM jobs the parameter appears on the ID card. For
example:
/*ID FORMS = NOPPS
For CYBER jobs the parameter appears on the PRINT command. For example:
PRINT/FORMS =NOPPS.
Users generating large printouts should be aware of the following limitations of the PPS and
specify FORMS =NOPPS if they are a problem:
. The PPS can print at most 64 lines/page. Pages containing more than 64 lines will
overflow to a second page.
. The only carriage control characters honored by the PPS are single space (blank), page
eject (1), double space (0), and triple space (-). Any other carriage control character
will cause unexpected results. Note that overprinting is not supported.
We are now dumping PPS output to tape for subsequent printing three times each day. The
schedule is as follows (Monday through Friday only):
Jobs Dumped to Tape Output available for pickup at DCL
8:00 AM 2:00 PM same day
12:00 Noon 5:00 PM same day
9:30 PM 9:30 AM next day
Depending on the backlog of PPS work and the amount of output to be printed for CSO, it
may be impossible for AISS to produce output in time for the 5 PM pickup noted above. In
this event the output will be available for pickup the following morning at 9:30.
NEW CHARGE RATE FOR TERMINAL REPAIRS
Effective July 1, 1981, CSO will charge $24.00 per hour for terminal repairs with a minimurr
of 1 hour. This will not include parts.
CYBER 175
SYSTEM NOTES
JUNE RELIABILITY REPORTS
CYBER 174
1 1 recoverable interruptions
18 non-recoverable interruptions
13 of these were for more than 15 minutes
MTBF = 22.4 hours
MTR = 8.2 minutes
Availability: 98.8% of scheduled uptime
Major cause of downtime was related to
TELEX, disk, and software problems.
IBM 4341:
4 recoverable interruption
12 non-recoverable interruptions
10 of these were for more than 15 minutes
MTBF = 41.3 hours
MTR = 18.6 minutes
Availability: 94.2% of scheduled uptime
Major cause of downtime was related to
TELEX, upgrade of core and disk problems.
20 interruptions
7 of these were for more than 1 5 minutes
MTBF = 33.6 hours
MTR = 14.6 minutes
Availability: 93.5% of scheduled uptime
Major cause of downtime was related to
software and disk problems.
MTBF = Mean Time Between Failures
MTR = Mean Time to Repair
CYBER
FETCH COMMAND UPDATED
The FETCH command now allows you to fetch multiple files with one command and to
specify the local file name to be assigned to a fetched file. Command format is:
FETCH, filename 1, filename2, ... , filenamen.
where filenamel, filename2, etc. have any of the following forms:
local file name to be assigned = fetch file name
local file name to be assigned = FNT ordinal of file in the FETCH queue
fetch file name
FNT ordinal of file in the FETCH queue
For example, suppose that you have four jobs awaiting fetch and that a QUERY command
results in:
# JOBNAME QUEUE
123 ABCDFFJ FETCH OUTPUTX
234 ABCDAAB FETCH MODEL
345 ABCDEEK FETCH LISTING
456 ABCDEFI FETCH RESULTS
Note that the first column gives the FNT ordinal of each file while the last column gives the
fetch file name. Then the command
FETCH,OUT=OUTPUTX.
would retrieve the first file in the list above and give it the local file name OUT.
FETCH,MODEL2 = 234.
would retrieve the second file in the list above giving the resulting local file the name
MODEL2. Note that file to be fetched has been identified by its FNT ordinal.
FETCH, LISTING.
would retrieve the third file in the list above, giving the resulting local file the same name
(LISTING) as the file had in the FETCH queue.
FETCH,456.
would retrieve the fourth file in the list above. Note that the FNT ordinal has once again
been used to identify the file to be fetched. In this case the resulting local file is given the
file's jobname ABCDEFI, not its fetch file name. (Note: FETCH will consider any number
used as a fetch file identifier to be an FNT ordinal. This means that you cannot use a
number as a fetch file name.)
Finally, all of this could have been accomplished with the single command
FETCH,OUT=OUTPUTX,MODEL2=234,LISTING,456.
If a FETCH command produces a local file with a name which is already in use, the existing
local file is first returned. Similarly, if a single FETCH command produces several local files,
all with the same name, only the last file fetched will exist at completion of the command.
QUERY COMMAND UPDATED
The QUERY command has been changed to optionally report approximate queue positions
and to optionally eliminate any empty queue from the report of queue lengths.
When QUERY is used with the switch /P, the output generated will include the approximate
queue position for any job reported which is in the TIELINE or INPUT queue. For example:
/QUERY/P.
# JOBNAME QUEUE
125 ABCDCFG TAPE MOUNT
146 ABCD345 PPS
34 ABCDABC # 2 TIELINE PRT
731 ABCD212 # 89 TIELINE PUN
12 ABCDFGH # 2 INPUT
In this example, the queue position is reported in the optional third column. Here the job
named ABCDABC is the second job in the TIELINE PRINT queue, i.e., one job will be pro-
cessed before it. Similarly, job ABCD212 is the 89th job in the TIELINE PUNCH queue,
and job ABCDFGH is the second job in the INPUT queue.
When used with the switch /F QUERY will generate a report of queue lengths but will omit
any empty queue. For example:
QUERY/F.
QUEUE TOTALS LENGTH
TIELINE 57
FETCH 103
The absence of the TAPE MOUNT, INPUT, PLOT, UNIX, PPS, and IBM INPUT queues
from this report indicate that these queues are empty.
Finally, any report which shows a job in a TIELINE PRINT or PUNCH queue will also report
the number of IBM jobs awaiting transmission to the IBM system for execution if there are
any. This number is of interest since TIELINE will transmit all such jobs before processing
any job in the TIELINE PRINT or PUNCH queue.
NEW VERSION OF ICE
On Monday, August 10, 1981 CSO will install Version 2.7 of ICE. ALthough most of the
changes in Version 2.7 involve internal performance enhancements, the following three
changes will affect our users:
• Entering the control statement
ICE, filename
when filename is either non-existent or not a local file, will result in the message
CREATE: filename
being displayed immediately after the ICE 2.7.0 banner. This is being done as a warn-
ing to users who have forgotten to do a GET on a file they are trying to ICE.
. Files which are not needed by ICE will not be created. Currently, if a user either does
not have an OPTION file or tries to ICE a non-existent file and immediately exits,
both the OPTION file and the source file will be created, but empty. In the new ver-
sion, these files will not be created.
. A group-oriented command (e.g., PG) will not cause dropping out of the loop if the
group reaches the bottom of the window.
WARNING: ICEWORK files will not be compatible between the current ICE (Version 2.6)
and Version 2.7. Upon trying to use an old ICEWORK file, ICE 2.7 will abort with the mes-
sage:
OBSOLETE WRKFIL VERSION
The current version will continue to be available for those who need it and can be accessed
by entering:
GRAB,ICE26/PAST.
FEATURE ARTICLES
ED PELG RETIRES
Ed Pelg, a longtime employee of CSO will be retiring at the end of August,
used CSO's terminal repair service, you probably recognize him.
If you've ever
Ed began work with the University in the Fall of 1954. Computing was still in its infancy.
The Digital Computer Laboratory, a separate entity under the graduate college, had just com-
pleted ILLIAC I and declared it ready for full time (24 hour) service.
Ed's first job was in the construction shop
operated by the Digital Computer Labora-
tory. The shop was located on the ground
floor of CERL (formerly ERL), a space
that is now filled with Plato terminals. The
construction shop was responsible for
maintaining ILLIAC I and for building
peripherals for it. ILLIAC I was at the
frontier of computing hardware and its
designers envisioned peripherals which
were not commercially available. Con-
struction and maintenance used vacuum
tube technology. ILLIAC I continued in
service until 1961.
In the early sixties Ed moved to the new
Digital Computer Laboratory building to
work with transistors and their applications
in ILLIAC III. For nearly ten years he
produced printed circuits and custom
assemblies for the new Department of
Computer Science. If you used PLORTS,
the time-sharing system on the IBM
360/75, you were utilizing channel inter-
facing equipment Ed helped build.
In the early seventies Ed joined the newly formed Computing Services Office as an engineer
for construction and repair. Here he continued, in part, to build electronic equipment,
including a hardware monitor locally referred to as the "heart lung machine".
The heart lung machine was the key element used to tune the IBM operating system for its
final measure of performance during the early to mid seventies. If you were computing then,
you benefited from his work.
During the major part of the last ten years Ed's reliability served CSO well. He was responsi-
ble for testing communications equipment used by time-sharing users for problems on an
ongoing basis. He spent the bulk of his time repairing terminals which belong to CSO, to
people on campus, or to people at about a dozen other schools in Illinois. His helpfulness
made him a goodwill ambassador for CSO in this capacity.
Ed's plans include travel and pursuit of his hobbies of fishing and golf. We all thank him for
his service to CSO and the University and wish him well.
SPECIAL PLOTTER OUTPUT SERVICE
Increasing numbers of both CYBER and IBM users are taking advantage of CSO's graphics
capabilities. Since many of these users intend to use the graphics output in reports, theses, or
other printed material, it is essential that the output be of high quality.
To meet the needs of these users, CSO offers a Special Plotter Output Service. This article
provides information about the service, briefly describes some of the available options, and
suggests several ways to get high-quality output. If you require quick turnaround or only
need "proof type plots, you should use the general plotting facilities provided at the CSO
Routing Room rather than the special plotting service.
The Special Plotter Output Service is located in Room 123 DCL. Special plots are run on a
1453b ZETA plotter or a 3653sx ZETA plotter. The 1453b plotter is a "tabletop" device,
designed to use 11-inch wide paper. The 3653sx plotter is much larger and is contained in a
free-standing cabinet. It is designed to use 34-inch wide paper. Paper selection is the only
difference between the two special plotters; all other options are the same.
Users may request special plotting by using the appropriate parameters (options) on the
CYBER PLOTZ command or the appropriate IBM procedure. CYBER users should obtain a
copy of Reference Guide RF-7.31 PLOTZ for command format and a full list of options.
IBM users should obtain a copy of Reference Guide RF-15.1. Both of these Reference
Guides are available at the RJE sites. It is also recommended that persons planning to use
the plotting facilities should obtain other plotting documentation from the CSO Accounting
and Distribution Center, 1208 W. Springfield.
The options presented below are available only through the Special Plotter Output Service. If
any of these special options are selected, even though you may have included the
PLOT=NORMAL option, your job will automatically be queued for the special plotters. It is
possible, however, to select only non-special or default options and still request special plot-
ting by using the PLOTTER = SPECIAL option.
10
Special Paper Options
There are three types of paper available for the small special plotter: FANFOLD, ROLL, and
ACETATE. Only one type of paper, WIDE, is available for the large plotter.
FANFOLD paper is the default. It is white, unlined, 11-inch wide paper, with perforations
dividing the paper every 8 1/2 inches (at the fold). If you are using fanfold, you must
remember to begin each plot on a new page to avoid crossing the perforations. Pens crossing
the perforations may skip, tear the paper, or even jam the plotter. It is also available through
the general plotting service with standard pen options.
ROLL paper is the same as fanfold except it has no perforations; it is a continuous roll of
paper. It is also available through the general plotting service with standard pen options.
ACETATE is a roll of 11 -inch wide, clear, thin plastic. If you select acetate, you must use
nylon pens because ink pens smear and ball-point pens will not draw on the plastic.
The WIDE paper used on the large plotter is the same as the roll paper on the small plotter
except it is 34 inches wide.
See the FORMS option in Reference Guide RF-7.31 or RF-15.1 for further details.
Special Pen Options
All of the ZETA plotters have four holders for pens, referred to in CSO documentation as
Pen 1 through Pen 4. The user has the option of using the defaults or of selecting a different
pen style and/or color for each pen. The user also has the option of using only one pen, any
combination of pens, or all four pens.
There are four pen styles available: ROLLING, BALL, INK, and NYLON.
ROLLING refers to a rolling writer pen which produces a thin, dark line. This is the default
for Pen 1.
BALL refers to a ball-point pen which produces a thin, light line. This is the default for Pen
2, Pen 3 and Pen 4.
INK refers to a cartridge pen which holds liquid ink. This type of pen produces a dark line
and is often the optimal selection if the plot is to be photographically reproduced. The thick-
ness of the line is determined by the size of the cartridge tip (specified by the user). The
default line size is approximately 0.35mm thick, but the user may select cartridge tip sizes to
create lines ranging from 0.30mm to 1.00mm in thickness.
NYLON refers to a pen which has a hard, nylon tip and produces a medium-bold line. Use
of the nylon pen is mandatory if you select acetate for plotting.
11
All of the above pen styles are available in four colors: BLACK, BLUE, GREEN and RED.
The shade of each color varies slightly according to the pen style. Also, the intensity of a
color is affected by the density of the line.
See the P, S, and C options in Reference Guide RF-7.31 or RF-15.1 for further details.
Length and Time Options
The LENGTH option is used to specify the maximum length of your plot in inches. The
TIME option is used to specify the maximum amount of time, in minutes, needed to run
your plot. Please note that incorrect specification of time or length will abort your plot, even
if it has not completed. For this reason, we recommend proofing your plots on the Routing
Room plotter before sending them to the special plotters.
If you are using ink pens to produce your plots, you should increase your time specification
by at least 50 percent. This is necessary because the operator must reduce the plotter's speed
for ink plots.
We have found that plot jobs which need more than 90 minutes, or those which are longer
than 300 inches, often require a number of operator restarts before an acceptable final output
is produced. For example, a complicated plot might need the default length, and might take
180 minutes on the plotter; or, a large number of plots might be combined into one job such
that the job requires over 300 inches but can be plotted in a reasonable amount of time. In
either case, any number of unavoidable problems could ruin a portion of the plot. The opera-
tor must restart the job from its beginning each time a problem occurs. We often reproduce
a lot of output that is mostly acceptable, but has to be done over because of a small portion
that is not acceptable. Jobs such as these greatly reduce special plot turnaround time.
However, we are hesitant to enforce maximum time and length restrictions. Instead, we ask
that you consider these problems and avoid submitting such jobs whenever possible.
Choosing Appropriate Combinations
Because the density of plotted lines can range from very fine to very bold, you have to con-
sider a number of factors when choosing your pen style. For instance:
• Do you need to photographically reproduce the plot?
• What is the overall size of the plot, and does that impose a particular range of line
densities?
. How complex is your plot? If the distinction of white space between the plotted lines
is important for clarity, are your lines fine enough to achieve that effect?
12
If your main requirement is defined by the Graduate College thesis specifications, remember
that those specifications were devised to maintain document consistency and clarity of repro-
duction. In general, masters theses are photocopied; doctoral theses are photocopied and
microfilmed. Most plots produced with ink pens, rolling writer pens, or nylon-tip pens can be
photocopied without a significant loss of quality.
To maintain optimal quality for plots being microfilmed, the original plot should be drawn
with sharply-defined, dark lines. This is often best achieved by using an ink pen with the
appropriate tip size and color combination most suitable for your plot. Also remember that
any color other than black is less likely to produce high-quality results when the plot is photo-
copied or microfilmed.
If you have chosen to use ink pens in more than one position, keep in mind that liquid inks
dries out rapidly when not in use. In general, if an ink pen has not been used for a while, it
should be primed before use. If it is not primed, the next time it is selected there will prob-
ably be either a pen skip or an ink smudge at the beginning of the line. Substitute nylon pens
or prime the ink pen before using it by moving outside the plot margins and drawing a short
line.
If you have questions about using CSO graphics programs and utilities, contact the Systems
Consultants, 166 DCL (333-6133). If you have problems, questions or suggestions about
CSO's Special Plot Output Service, contact Debbie Weller, 171 DCL (333-8150).
GENERAL PLOTTING SERVICE
DCL ROUTING ROOM
Currently, the General Plotting Service in the DCL Routing Room handles only those plots
using the default or standard options and fanfold paper. Effective August 17, 1981, they will
also handle plots using default options, but roll paper.
In other words, after August 17, if you use default options and either fanfold or roll paper,
your plot will automatically be sent to the plotter in the Routing Room. If you specify roll
paper and still want your job to be processed on the special plotter, you must specify this by
using the option:
PLOT = SPECIAL
13
FORTRAN FUTURES
This article is part of a series on the work of the FOR TRAN Standards
Committee X3J3 in producing the next revision of the FOR TRAN standard.
The reader is reminded that the features described in this article are not a
part of any FORTRAN compiler currently available, but rather are proposed
requirements for FORTRAN processors produced in the late 1980's through
the mid 1990's. Although every effort has been made to accurately describe
the current position ofX3J3 on these matters, the development of a revision to
the FOR TRAN standard is an evolutionary process, and these proposals may
be subject to refinement, revision, or even retraction. Comments on these
proposals may be given to Kurt Hirchert of the CSO Systems Consulting
staff, who is a member ofX3J3.
In order to write programs in a language, one of the first things one must learn is its source
form, that is, the rules for expressing its statements. X3J3 has passed a number of proposals
modifying the source form of FORTRAN.
One commonly heard complaint about FORTRAN is that six characters are not enough to
produce meaningful variable names. X3J3's answer to this complaint has been to increase the
limit on symbolic names to thirty-one characters. In order to keep such names readable, the
underscore or break character (_) has been made a permissible character in symbolic names.
The current FORTRAN standard, known as FORTRAN 77, added a number of features such
as CHARACTER variables and constants and the option to use them in place of the FOR-
MAT statement label in an input/output statement. Since formats often contain quoted
strings, these quotes must be represented by two quotes. If such quoted strings are, in turn,
to contain quotes, these quotes must be represented by sets of four quotes. For example,
WRITE(6,'(" C = "",,,A,",","),)C
As you can see, the quotes quickly become so numerous that they are difficult to read. To
help control these difficulties, X3J3 has voted to make the double quote character an alterna-
tive for delimiting character constants. The previous example could then be written
WRITE(6,"C C=",,A,",,)")C
This reduces the proliferation of quotes to the same level as would occur in a normal
unquoted FORMAT statement.
The features of the FORTRAN 77 standard are available to CSO users
through the FNT5 and M77 compilers.
Because of the extensive nature of some of the proposals for adding to FORTRAN, X3J3 has
voted to add to the set of characters which can be used to write FORTRAN programs. The
special characters common to EBCDIC, ASCII, and all of the international variations on
ASCII were added. These are exclamation point (!), double quote ("), percent (%), amper-
sand (&), semicolon (;), less than (<), greater than (>), question mark (?), and underscore
(_). Other proposals have specified uses for some, but not all, of these characters.
14
As terminals supplant keypunches as the typical source of computer input, the handling of
lower case letters becomes an increasingly important issue. Although no requirement has
been made that a FORTRAN processor support lower case letters, X3J3 has approved rules
for handling lower case letters in processors where they are available. Briefly, these rules call
for lower case letters to differ from their upper case counterparts in character constants and
other similar contexts, but to be equivalent in all other syntactic uses. For example, xyz
denotes the same variable as XYZ, and a FORMAT statement could also be a format state-
ment, a Format statement, or even a fOrMaT statement.
The most extensive source form proposals X3J3 has passed concern the elimination of restric-
tions on the columns in which the various FORTRAN source elements may appear. All cases
where the interpretation of a syntactic element was dependent on the column in which it
appeared have been eliminated and source may now extend beyond 72 columns. Some of the
associated changes are relatively straightforward. For example, a statement label is now sim-
ply a contiguous string of digits at the beginning of a statement. On the other hand, other
changes are more drastic. For example, a C in column 1 can no longer denote a comment,
since statements can begin with C and may begin in column 1. In the proposed revised
source form, the exclamation point (!) begins a comment. In fact, any exclamation point not
in a character string or similar context turns the remainder of the input record into a com-
ment, thus allowing comments on the same lines as statements, but only on the right hand
end. As with FORTRAN 77, blank lines and lines containing nothing but comments are
ignored. Since column 6 can no longer be used to indicate continuations, a different con-
tinuation convention has been adopted. The continuation mark must appear on the end
(except for comments and/or blanks) of the line to be continued, rather than the continua-
tion line. The continuation mark is an ampersand (&). In the special case of continuing a
character constant or similar item, there can be no comments following the continuation
mark, since these would be indistinguishable from characters intended to be part of the char-
acter constant, and there must be a confirming continuation mark as the first nonblank char-
acter of the continuation line. Neither the continuation mark nor the confirming continuation
mark are part of the character constant value. Finally, because there are cases where readabil-
ity would be enhanced rather than reduced by the appearance of more than one statement on
an input line, the character semicolon (;) has been designated to separate statements in such
cases.
Example:
! A PROGRAM TO CONVERT CARTESIAN COORDINATES TO POLAR
PROGRAM MAIN
REAL X,Y ! THE INPUT CARTESIAN COORDINATES
REAL RJHETA ! THE OUTPUT POLAR COORDINATES
10 PRINT V ENTER DESIRED CARTESIAN COORDINATES'
READ(*,*,END = 99) X,Y
! CHECK FOR INDETERMINANT CASE
IF(X.EQ.O.AND.Y.EQ.O) THEN; PRINT *, ' POLAR COORD&
&INATE VALUES ARE INDETERMINANT'
ELSE
R=SQRT(X**2+Y**2);THETA = ATAN2(Y,X)
15
PRINT *, 'POLAR COORDINATE VALUES ARE ', &
R, THETA
END IF
GO TO 10
99 END
Ironically, one of the most controversial of the changes made by X3J3 will have little effect
on most programmers. Blanks, which are currently ignored in interpreting FORTRAN state-
ments, are to become a significant part of FORTRAN syntax. Much of the controversy con-
cerning this change results from a misunderstanding of what is implied. Most of the places
where blanks will be required are places where most programmers would have written a blank
anyway. Similarly, most of the places where blanks will be prohibited are places where pro-
grammers would not normally write blanks. Programmers will still be free to use multiple
blanks to space their programs as they wish. What is being eliminated is the ability to use
blanks in unusual ways that are more often confusing than helpful. For example, do you
recognize these FORTRAN statements?
REALM X,Y,Z
READY, SET, GO
GOT ONE
If not, perhaps they will be more familiar when spaced conventionally.
REAL MX,Y,Z
READ Y, SET,GO
GO TO NE
On the other hand, X3J3 is not in the habit of restricting or eliminating features simply
because they can be misused. What benefits are to be gained by making blanks significant?
This change will certainly make FORTRAN more like other languages, thus making it easier
for the users of other languages to learn FORTRAN and vice versa. Significant blanks will
also make error detection easier. Consider the following:
DO 10 1 = 1.10
Most FORTRAN programmers would recognize that this is an erroneously written DO state-
ment (most likely the result of accidentally replacing a comma by a period). FORTRAN com-
pilers, on the other hand, would take this to be a valid assignment statement that assigns the
value 1.10 to the variable DO10I. This is just one example of the way the current
insignificance of blanks makes it impossible for a compiler to use visual spacing clues to
recognize errors. A more serious concern in terms of the current revision effort is the prob-
lem that if blanks are not used to delimit the various elements of a FORTRAN statement,
then other punctuation characters must be used. This tends to result in an overpunctuated
language that is less convenient to use.
16
For example, consider the PARAMETER statement added to FORTRAN 77:
PARAMETER (NDIM = 100)
In early proposals for this statement and most early implementations, the parentheses were
not required.
PARAMETER NDIM = 100
Unfortunately, with insignificant blanks and an extension to longer variable names (such as is
proposed for this revision), this statement would be indistinguishable from the assignment
statement
PARAMETERNDIM = 100
X3J3 was thus forced to add the parentheses to avoid ambiguity. Making blanks significant
will help avoid such overpunctuation in the coming revision.
SALES - EXCHANGES - HELP WANTED
GRADUATE RESEARCH ASSISTANT
One-quarter time graduate research assistant needed beginning August 21, 1981 for statistical
programming and data management. Knowledge of SPSS and SAS desirable. Ability to keep
track of and document carefully a number of large social science datasets stored on tape.
Work is for an economist studying the impact of federal policy on sex differentials in the
labor market.
Contact Prof. A. Beller after August 17 at 333-7257, or leave a message at 333-2412.
OFF-LINE's Mailing List
If you wish to be placed on our mailing list for future issues of OFF-LINE, if you wish to be
removed from the list, or if you wish to enter an address correction, please complete and
return this page. (Current subscribers are kept on the mailing list until a specific request for
removal is received, or until a mailing is returned as undeliverable.)
Check one:
□ New subscriber
□ Removal request
□ Address correction
Name:
Address:
CAMPUS or Zip Code
(If address correction, give old address and zip code below.)
Comments:
RETURN TO:
OFF-LINE
150 Digital Computer Laboratory
University of Illinois at Urbana-Champaign
1304 West Springfield Avenue
Urbana, Illinois 61801
1 ^»— 1 _■
. >- .
ir-t-^-
U 1 I
C3mn
i/it>t i >—
Mnr
2^r
wJ>
(~2
< ■<
TJm
J>«C
/O
<C
rn
"D
Computing
Services
Office
Off-Line
^niver^tj^nUinois^a^rbanajQianjgai^
VOL. 9. NO. 9 September 1981
EDITOR: Lynn Bilger
PHONE: (217) 333-6236
139 Astronomy Building
1011 W. Springfield
Urbana, Illinois 61801
Page
1
4
13
21
21
22
22
23
24
25
27
27
Contents
POLICY
CSO Short Courses - Fall Semester
CSO Videotapes Available
SYSTEM NOTES
July Reliability Reports
CYBER
IDA Statistical Package Installed on CYBERs
IBM
OSIRIS IV on the IBM 4341
PDISK1 Available for User Datasets
FEATURE ARTICLES
FORTRAN Futures
MISCELLANEOUS
Computer-Related Discounts
INTEL to Hold Open House
National Educational Computing Conference
SALES - EXCHANGES - HELP WANTED
Programmer Wanted
Information Wanted
SPECIAL EDITION ARTICLES AND INDEX
CYBER Files - Daily Backup Procedure
Special Weekend Rates
New 6250 bpi Tape Drives
Easy Graphing
Updates on Page Printing System
CSO DIRECTORY - STAFF AND SERVICES
Administrative
Director
Business Manager
Secretary
User Accounting
Distribution Office
Systems Consulting
Statistical Services Consulting
Text Processing Consulting
Terminal Repair Service
CYBER Dial-up Numbers
Asst Dir User Services
Asst Dir Systems and Operations
Asst Dir Statistical Services
and UNIX System
Asst Dir Development
Asst Dir Engineering and
Hardware Selection
Documentation
CYBER-IBM Operations
UNIX Operations
Telecommunications
Laboratory Support Project
RJE Operations North
RJE Operations South
RJE Sites
Agriculture
Chemistry
Commerce West
CRH Snack Bar
DCL Routing Room
Electrical Engineering
Florida Ave Res Hall
Illinois St Res Hall
Mechanical Engineering
Psychology
Social Science
George Badger
150
DCL
333-4103
Stanley Rankin
150
DCL
333-6530
Joyce McCabe
150
DCL
333-1637
rare Support
1208
W Springfield
333-7752
1208
W Springfield
333-6760
1208
W Springfield
333-6133
65
Comra West
333-2170
207
Astronomy
333-7318
150
DCL •
333-0969
CYBER 175
110-300
baud
333-4000
CYBER 175
1200
baud
333-4001
CYBER 174
110-300
baud
333-4004
Robert Penka
173
DCL
333-4709
Sandra Moy
177
DCL
333-4703
Larry Sautter
91
Comm. West
333-2170
J. M. Randal
120
DCL
333-9772
Cliff Carter
195
DCL
333-3723
Lynn Bilger
139
Astronomy
333-6236
Jack Knott
194a
DCL
333-6562
Debbie Weller
123
DCL
333-8150
Tom Kerkering
164
DCL
333-0816
Mike Gardner
164
DCL
333-7904
Rex Duzan
162
DCL
333-6285
Don McCabe
1208
W Springfield
333-2171
333-7752
M103
Turner Hall
333-8170
153
Noyes Lab
333-1728
70
Comm West
333-4500
120
Snack Bar
333-1851
129
DCL
333-6203
146
EEB
333-4936
FAR
333-2695
ISR
333-0307
65
MEB
333-1430
453
Psych Bldg.
333-7531
202
Lincoln Hall
333-0309
OFF-LINE is the monthly newsletter of the Computing Services Office at the University of
Illinois at Urbana-Champaign. Unless otherwise indicated, permission to reprint is freely
granted, provided that the author, if named, and the Computing Services Office (CSO) are
credited. Information in this issue is current as of August 31, 1981.
CSO operates a CDC CYBER 175 with 256K words of central memory and a CDC CYBER
174 with 196K words of central memory. The 175 and 174 run under the NOS Operating
System and share 512K words of ECS. The 175 serves over 200 simultaneously active text
and graphics terminals and the 174 serves over 100 simultaneously active terminals. CSO
also operates an IBM 4341 with 4 million bytes of memory running HASP-OS/MVT under
VM. In addition, CSO operates a DEC PDP-11/50 with 252K bytes of core driving a GSI
CAT-8 phototypesetter and a DEC VAX 11/780 with 2 million bytes of core, both running
under the UNIX Operating System.
CSC
£**, Dote rtanSSb^ or before the
ne"- mutilation _.j
CSO is offering the folic
with our facilities and thfirp ft 4*
computer systems. ** J^ ><
To register for a course: Jg?
ER
1981 to acquaint people
R 175 and CYBER 174
please come in r.
phone 333-6630
Registration is free and
able classes on a topic
someone drops a class
PLEASE NOT
UNTIL THE V
TAUGHT!
We are enforcing this
had. Just this past sf
were registered. Rei
LINE.
The classes offered diw ' Ll6'-oio96
INTRODUCTION TO THE CYBER COMPUTER SYSTEM
u find that all of the avail-
ing list. We will call you if
I CLASS
>S IS TO BE
nber of no-shows we have
> in which over 20 people
^d in each issue of OFF-
This course is intended for the first-time CYBER computer user. The emphasis is on time-
sharing usage with discussions on the ICE text editor and card batch usage. Three classes will
be offered:
Sept. 14, 16, 18, 21
Oct. 6, 8, 13,15
Nov. 2, 4, 6, 9
12 noon - 1 PM
12 noon - 1 PM
12 noon - 1 PM
115 DCL
115 DCL
115 DCL
INTRODUCTION TO GRAPHICS AT CSO
This class will be an overview of the various hardware devices and software packages available
at CSO. One class is being offered:
Sept. 17
2 PM - 3 PM
201 Astronomy
CSO DIRECTORY - STAFF AND SERVICES
Administrative
Director
Business Manager
Secretary
User Accounting
Distribution Office
Systems Consulting
Statistical Services Consulting
Text Processing Consulting
Terminal Repair Service
CYBER Dial-up Numbers
Asst Dir User Services
Asst Dir Systems and Operations
Asst Dir Statistical Services
and UNIX System
Asst Dir Development
Asst Dir Engineering and
Hardware Selection
Documentation
CYBER-IBM Operations
UNIX Operations
Telecommunications
Laboratory Support Project
RJE Operations North
RJE Operations South
RJE Sites
Agriculture
Chemistry
Commerce West
CRH Snack Bar
DCL Routing Room
Electrical Engineering
Florida Ave Res Hall
Illinois St Res HM1
Mechanical Engineering
Psychology
Social Science
George Badger
150 DCL
333-4103
Stanley Rankin
150 DCL
333-6530
Joyce McCabe
150 DCL
333-1637
rare Support
1208 W Springfield
333-7752
1208 W Springfield
333-6760
1208 W Springfield
333-6133
65 Comm West
333-2170
207 Astronomy
333-7318
150 DCL ■'•'
333-0969
CYBER 175
110-300 baud
333-4000
CYBER 175
1200 baud
333-4001
CYBER 174
110-300 baud
333-4004
Robert Penka
173 DCL
333-4709
Sandra Moy
177 DCL
333-4703
Larry Sautter
91 Comm West
333-2170
J. M. Randal
120 DCL
333-9772
Cliff Carter
195 DCL
333-3723
Lynn Bilger
139 Astronomy
333-6236
Jack Knott
194a DCL
333-6562
Debbie Weller
123 DCL
333-8150
Tom Kerkering
164 DCL
333-0816
Mike Gardner
164 DCL
333-7904
Rex Duzan
162 DCL
333-6285
Don McCabe
1208 W Springfield
333-2171
333-7752
M103 Turner Hall
333-8170
153 Noyes Lab
333-1728
70 Comm West
333-4500
120 Snack Bar
333-1851
129 DCL
333-6203
146 EEB
333-4936
FAR
333-2695
ISR
333-0307
65 MEB
333-1430
453 Psych Bldg.
333-7531
202 Lincoln Hall
333-0309
OFF-LINE is the monthly newsletter of the Computing Services Office at the University of
Illinois at Urbana-Champaign. Unless otherwise indicated, permission to reprint is freely
granted, provided that the author, if named, and the Computing Services Office (CSO) are
credited. Information in this issue is current as of August 31, 1981.
CSO operates a CDC CYBER 175 with 256K words of central memory and a CDC CYBER
174 with 196K words of central memory. The 175 and 174 run under the NOS Operating
System and share 512K words of ECS. The 175 serves over 200 simultaneously active text
and graphics terminals and the 174 serves over 100 simultaneously active terminals. CSO
also operates an IBM 4341 with 4 million bytes of memory running HASP-OS/MVT under
VM. In addition, CSO operates a DEC PDP-11/50 with 252K bytes of core driving a GSI
CAT-8 phototypesetter and a DEC VAX 11/780 with 2 million bytes of core, both running
under the UNIX Operating System.
\
POLICY
CSO SHORT COURSES - FALL SEMESTER
CSO is offering the following short courses during the fall semester 1981 to acquaint people
with our facilities and th£ Control Data Corporation's (CDC) CYBER 175 and CYBER 174
computer systems.
To register for a course:
please come in person to Room 150 DCL, or
phone 333-6630.
Registration is free and limited to 30 people in some classes. If you find that all of the avail-
able classes on a topic are full, please leave your name on our waiting list. We will call you if
someone drops a class.
PLEASE NOTE THA T YOU MAY NOT REGISTER FOR A CLASS
UNTIL THE WEEK PRECEDING THE WEEK THE CLASS IS TO BE
TAUGHT!
We are enforcing this new policy in an attempt to reduce the number of no-shows we have
had. Just this past semester, only 8 people showed up for a class in which over 20 people
were registered. Reminders of upcoming classes will be published in each issue of OFF-
LINE.
The classes offered are as follows:
INTRODUCTION TO THE CYBER COMPUTER SYSTEM
This course is intended for the first-time CYBER computer user. The emphasis is on time-
sharing usage with discussions on the ICE text editor and card batch usage. Three classes will
be offered:
Sept. 14, 16, 18, 21 12 noon - 1 PM 115 DCL
Oct. 6, 8, 13,15 12 noon - 1 PM 115 DCL
Nov. 2, 4, 6, 9 12 noon - 1 PM 115 DCL
INTRODUCTION TO GRAPHICS AT CSO
This class will be an overview of the various hardware devices and software packages available
at CSO. One class is being offered:
Sept. 17 2 PM - 3 PM 201 Astronomy
NCAR PLOT PACKAGE
This class will be an overview of the facilities and general usage of the NCAR plotting pack-
age. This package allows for 3-D plotting, contour plots and world map projections. Familiar-
ity with FORTRAN and the CYBER is assumed. One class is offered:
Sept. 24 2 PM - 3 PM 201 Astronomy
EASY GRAPHING
This is a discussion of a high-level interactive plotting program for X-Y plots, bar charts and
pie charts. Its English-like commands require no programming experience to generate plots.
Familiarity with the CYBER is assumed. Two classes are offered:
Sept. 28, 30, Oct. 2 12 noon - 1 PM 115 DCL
Nov. 16, 18, 20 12 noon - 1 PM 115 DCL
GRAPHICS COMPATIBILITY SYSTEM (GCS)
This class on GCS will cover plots on various graphics devices. Familiarity with FORTRAN
and the CYBER is assumed. One class is offered:
Oct. 5, 7, 9, 12, 14, 16 12 noon - 1 PM 115 DCL
INTRODUCTION TO RNF
This is a beginning-level discussion of the RNF text formatter used in the preparation of
letters, manuals, or other documents. Topics will include tabbing, margins, spacing, para-
graphing, and justification. Familiarity with the CYBER is assumed. Three classes are
offered:
Sept. 28, 30, Oct. 2 2 PM - 3 PM 201 Astronomy
Oct. 19, 21, 23 2 PM - 3 PM 201 Astronomy
Nov. 16, 18, 20 2 PM - 3 PM 201 Astronomy
INTERMEDIATE RNF
This class is a continuation of the Introduction to RNF. Topics will include macros, variables,
arrays and applications. Two classes are offered:
Oct. 5, 7, 9 2 PM - 3 PM 201 Astronomy
Oct. 26, 28, 30 2 PM - 3 PM 201 Astronomy
CYBER MAGNETIC TAPES
This class will cover the use of magnetic tapes on the CYBER. Familiarity with the use of the
CYBER is assumed. One class is offered:
Oct. 26, 28, 30 12 noon - 1 PM 115 DCL
SIMULATION PACKAGES
This class will be an overview of the various simulation packages that are available. One class
is offered:
Sept. 21, 23, 25 2 PM - 3 PM 201 Astronomy
ALGEBRAIC MANIPULATION PACKAGES
This class will be an overview of the various packages that can perform algebraic manipula-
tion. One class is offered:
Nov. 9, 11,13 2 PM -3 PM 201 Astronomy
CURVE FITTING ROUTINES
This class will cover the mathematical software available to do curve fitting. One class is
offered:
Oct. 12, 14, 16 2 PM - 3 PM 201 Astronomy
SAS
This is an introductory course on the SAS statistical package. Course topics will include:
• Preparation of data for input
. The SAS 'DATA' step
• Manipulating the input data: modifying variables, creating new variables, deleting
observations or variables
• Sorting data
• Obtaining basic statistics, frequencies, plots
The course, consisting of three 2-hour lectures, will be offered once:
Nov. 2, 4, 9 7 PM -9 PM 115 DCL
SOUPAC
A lecture/laboratory will be presented on the SOUPAC statistical package on the CYBER.
Handouts and examples will be provided illustrating basic statistics, regression and analysis of
variance. Enough general CYBER terminology will be given to facilitate using SOUPAC. In
the lab session, the use of keypunches and terminals will be demonstrated, and exercises will
be provided on editing, running and modifying SOUPAC programs. Instructors will act as
consultants for solving problems during the lab session. The course, consisting of two 2-hour
lectures and one 3-hour lab will be offered once:
Sept. 29, Oct 1 7 PM - 9 PM 243 Comm West
Oct. 3 (lab) 9 AM - 12 noon 243 Comm West
SPSS
This is an introductory course on the SPSS statistical package on the CYBER. The course
contents will include:
Preparation of data for input
Basic components of the SPSS language
Running SPSS programs from cards or at a terminal
Using SPSS to transform the data
Obtaining basic statistics and crosstabulation tables
The course, consisting of two 3-hour lectures, will be offered twice:
Sept. 22, 24 6:30 PM - 9:30 PM 115 DCL
Oct. 5, 7 6:30 PM - 9:30 PM 115 DCL
CSO VIDEOTAPES AVAILABLE
CSO has produced a series of eight videotapes which introduce the novice to the CYBER Sys-
tems. A viewing guide containing the major pictorials used in the series is available and can
be used to facilitate note taking.
The title and a brief synopsis of each of the videotapes is given below. Running time is 10-15
minutes for each videotape.
Introduction to Computing at CSO
A brief look at the steps required to solve a problem using a computer and some of
the hardware used.
Using a Terminal
A description of the physical operation of a terminal and some of the keys that have a
special meaning to the CYBER.
Introduction to CYBER Time-Sharing
A tutorial for logging on and off the CYBER.
File Usage - Local File and Indirect Access to Permanent Files
An introduction to CYBER files and the commands used to manipulate them.
Introduction to ICE Text Editing
A tutorial on entering and modifying files with ICE.
Running a FORTRAN Program - Concepts
The concepts of compilation, loading and execution.
Running a FORTRAN Program - The PROGRAM Statement
The PROGRAM statement and its relationship to files accessed by the program.
Running a FORTRAN Program - Control Statement
The control statements used to compile, load and execute a FORTRAN program.
Anyone can view these videotapes by going to the Undergraduate Library in person to make a
reservation for use of the videotape equipment. Ask for a copy of the viewing guide when
you check out the videotape for viewing.
SYSTEM NOTES
JULY RELIABILITY REPORTS
CYBER 175 CYBER 174
12 recoverable interruptions 12 recoverable interruptions
13 non-recoverable interruptions 22 non-recoverable interruptions
15 of these were for more than 15 minutes 27 of these were for more than 15 minutes
MTBF = 27 hours MTBF = 20.1 hours
MTR = 15.0 minutes MTR = 29 minutes
Availability: 98.5% of scheduled uptime Availability: 97.2% of scheduled uptime
Major cause of downtime was related to Major cause of downtime was related to
TELEX, disk, and power problems. TELEX, disk, and power problems.
IBM 4341:
8 interruptions
2 of these were for more than 15 minutes
MTBF = 33.6 hours
MTR = 14.6 minutes
Availability: 98.2% of scheduled uptime
Major cause of downtime was related to
power and disk problems.
MTBF = Mean Time Between Failures
MTR = Mean Time to Repair
CYBER
IDA STATISTICAL PACKAGE INSTALLED ON CYBERS
The IDA (Interactive Data Analysis) conversational statistical package marketed by SPSS Inc.,
has been installed on the CDC CYBER 175 and 174. To quote from the IDA Users' s Manual,
Robert F. Ling, McGraw-Hill, 1980:
"IDA is designed with both statistical capabilities and user convenience in mind.
Its main emphasis is on statistical tools associated with regression analysis of
linear models and related model-building techniques, as well as the high level
of man-machine interaction that is typically required to carry out such tech-
niques properly. Thus IDA is designed to emphasize what has come to be
called model identification and diagnostic checking ...
Besides its interactive feature in statistical data analysis, the other major feature
of the system is its user-convenience capabilities. Elaborate error detection and
recovery codes, internal help files and bi-level prompts, and other special
features of IDA are all capabilities specifically designed for the convenience of
the user."
To access the IDA package use the following sequence of time-sharing commands:
GRABJDA/FUTURE.
IDA.
The manual may be purchased for $13.00 at the Accounting and Distribution Office, 1208 W.
Springfield. Questions concerning the IDA package should be directed to the CSO Statistical
Consultants in Room 65 Commerce West (333-2170).
IBM
OSIRIS IV ON THE IBM 4341
OSIRIS IV, a statistical and data management package, has been installed on the IBM 4341.
This social science package program has extensive data management capabilities, including the
capacity to handle heirarchical files.
OSIRIS IV is invoked by the following EXEC statement:
// EXEC OSIRIS4
OSIRIS IV uses at least 150K of memory.
Questions concerning the use of the OSIRIS IV package should be directed to the Social Sci-
ences Quantitative Laboratory in Room 208 Lincoln Hall (333-6750). Documentation is also
available in Room 208.
PDISK1 AVAILABLE FOR USER DATASETS
The PUBLIC disk pack on the IBM system has been consistently short of free space. To
alleviate this situation, on August 14, 1981 another disk, PDISK1, was devoted to user
datasets. To place a dataset on this disk, use the JCL parameter
VOL=SER = PDISKl
instead of the parameter
VOL=SER=PUBLIC
FEATURE ARTICLES
FORTRAN FUTURES
This article is part of a series on the work of the FORTRAN Standards Committee X3J3 in producing the
next revision of the FORTRAN standard. The reader is reminded that the features described in this article
are not a part of any FOR TRAN compiler currently available, but rather are proposed requirements for
FORTRAN processors produced in the late I980's through the mid 19%'s. Although every effort has been
made to accurately describe the current position of X3J3 on these matters, the development of a revision to
the FOR TRAN standard is an evolutionary process, and these proposals may be subject to refinement, revi-
sion, or even retraction. Comments on these proposals may be given to Kurt Hirchert of the CSO Systems
Consulting staff, who is a member of X3J3.
During the public comment and review period for the 1978 FORTRAN standard, one sugges-
tion frequently given X3J3 was that "structured programming" control structures be added to
FORTRAN. In preparing its response to these comments, X3J3 found that although the ideal
of structured programming was widely supported, there was much disagreement about how to
put this concept into practice. Reaching a consensus on how to add "structured programming"
to FORTRAN would have delayed the completion and adoption of the new FORTRAN stan-
dard, so X3J3 felt it must delay consideration of "structured programming" to the next revi-
sion. One aspect of "structured programming", however, was found to enjoy the necessary
consensus, so the block-IF was adopted as a part of FORTRAN 77.
A brief description of the block-IF may be in order for those of you who have not worked
with a FORTRAN 77 based compiler such as FTN5 or M77. In its simplest form, the block-
IF allows one to replace code such as
IF(I.NE.J) GO TO 1318
1318 CONTINUE
by code such as the following, which contains no GO TOs or statement labels:
IF(I.EQ.J) THEN
END IF
An ELSE statement can be inserted preceding a block of statements to be executed when the
expression given is false. One or more ELSE IF statements can be inserted to allow a series
of expressions to be evaluated until a true one is found (with the ELSE block then being exe-
cuted only if all the expressions are false). More complicated control structures can be built
by nesting block-IFs inside other block-IFs.
Thus the following set of block-IFs
IF (M.LT.N) THEN
IF (X.GT.O) THEN
ELSE
END IF
ELSE IF (M.GT.N) THEN
IF (Y.NE.Z) THEN
END IF
ELSE
IF (Y.NE.X) THEN
ELSE IF (Z.EQ.O) THEN
END IF
END IF
might replace this code written with GO TOs and statement labels
IF (M.GE.N) GO TO 3716
IF (X.LE.O) GO TO 3715
GO TO 3720
3715 CONTINUE
GO TO 3720
3716 IF (M.LE.N) GO TO 3718
IF (Y.EQ.Z) GOTO 3717
10
3717 CONTINUE
GO TO 3720
3718 CONTINUE
IF
3719
(Y.EQ.X) GO TO 3719
IF (Z.NE.O) GO TO 3720
3720 CONTINUE
X3J3 has developed two other "structured programming" constructs to be used with the
block-IF in the next revision of the FORTRAN standard. The first of these is the block-DO.
In its simplest form
DO
REPEAT
the block-DO establishes an "infinite" loop which must be terminated by tests contained
within the loop. A new statement, EXIT, has been provided for this purpose. The use of
EXIT within a loop transfers control to the first statement following the end of the loop.
Although the EXIT statement has been provided for the purpose of terminating loops, the
programmer is free to use other means, such as GO TOs, for this purpose. This separation of
the basic looping mechanism from the means by which it is terminated avoids the controversy
about whether the termination condition of a loop should be tested at the top or bottom of a
loop by giving the programmer the freedom to explicitly place the termination test wherever
it is desired, at the top, at the bottom, or even somewhere in the middle.
An extended form of the block-DO
DO (variable = expression,expression [,expression] )
REPEAT
provides the same function as the current DO loop, but in a form consistent with the rest of
the "structured" looping facility. (There are some indications that the array processing facili-
ties being added to FORTRAN will make most of the uses of such loops unnecessary. These
array processing facilities will be the subject of a future article in this series.)
11
A different extend form
DO (expression TIMES)
REPEAT
will be available when one needs to iterate a known number of times, but access to the
counter is not needed. The EXIT statement can also before the indicated number of itera-
tions have been executed.
The second new "structured programming" construct planned for the next revision of the
FORTRAN standard is a structured alternative to the use of the computed-GO TO statement.
Its use can probably be best shown with an example.
SELECT CASE(CHAR)
CASEOO'^YAVF')
CASEC ',',')
CASE(V')
CASE DEFAULT
END SELECT
When the SELECT CASE statement is executed, the expression in parentheses is evaluated
(in this case, the value of the character variable CHAR is determined), and the block of
statements following the "matching" CASE statement is executed. When the block of state-
ments has been completed, execution continues following the END SELECT statement. In
this example, the first block of statements will be executed if CHAR contains a hexadecimal
digit, the second it it contains a blank or comma, the third if it contains a slash, and the
fourth if it contains any other character. As the example illustrates, more than one possible
value of the expression may be associated with a block of statements and lists of consecutive
values may be abbreviated with a kind of range notation. Any given value may be specified
in at most one CASE statement. The CASE DEFAULT statement is equivalent to a CASE
statement specifying all possible values of the selection expression that have not been
specified in one of the other CASE statements. As this example illustrates, it is expected that
CASE DEFAULT block will typically be placed last, but this is not required. It is not
required to have a CASE DEFAULT block. If it is omitted, the possibility may exist that the
value of the selection expression "matches" none of the CASE statements. Such a failure to
"match" is an error and may result in unpredictable action by the processor.
12
There is at least a remote possibility that an additional "structured programming" control
structure (related to error handling) may be added to the coming revision at a later date.
Recent actions by X3J3
The 79th meeting of X3J3 was held August 10-14, 1981 in Los Alamos, New Mexico.
Actions taken at that meeting include the following:
• The basic outlines of a "name-directed" input/output facility were adopted. This facil-
ity should ofTer funtionality similar to what is available through NAMELISTs, but in a
manner which is more consistent with the existing standard FORTRAN input/output
facilities.
. A facility to allow a subprogram to deal with the omission of some of its arguments
was adopted. The effects of such an omission could be as simple as the provision of a
default value or could extend to more far-reaching modifications in the actions of the
subprogram.
. INVERSE and DETERMINANT were deleted from the list of new array processing
intrinsics to be provided.
• Action was taken to correct, clarify, or extend previously passed facilities in several
areas, including internal procedures, data structures, and numeric precision control.
• There was extensive discussion on the issue of how to handle the interaction between
previously passed facilities for data structuring and array subsection referencing.
• Mechanisms for certain classes of array valued functions were discussed.
• Mechanisms for handling errors, exceptions, and other "events" were discussed,
including user-defined "events".
• A proposal to determine which language facilities would be part of the "core" language
was extensively discussed.
. Suggested responses to several requests for official interpretations of the current
(1978) standard were discussed in anticipation of official action to be taken at the 80th
meeting.
13
MISCELLANEOUS
COMPUTER-RELATED DISCOUNTS
IBM
The IBM CRT terminal has changed in price from $1046 to $1116.52. For those doing
work (such as text editing) that requires heavy keyboard usage, this terminal is recom-
mended.
IBM Model 3101-10 $1116.52
EIA Cable 30.00
The ordering address is:
Illinois Educational Consortium
1306 South Sixth Street
Springfield, IL 62703
Maintenance is provided through warranty ship-back to Chicago depot-repair. The war-
ranty can be renewed for $70/year.
HAZELTINE
The reason that Hazeltine has so many models listed is due to the number of years they
have been in the business and the rather wide customer base they have served.
For Research and Instructional customers the low-end model 1500 or 1410 is quite good
enough for use with CYBER. Although, the GT 100A, ADM 3A, MIME 314 will do just
as well.
Presently the only one maintained by CSO personnel is the H-1500.
Hazeltine H-1410 $615.00
H-1420 725.00
H-1500 765.00
H-1510 895.00
H-1520 1145.00
Modular I - Editing 1500.00
EX 80 - Model 20 1075.00
EX 80 -Model 30 1345.00
H-1421 650.00
H-1552 950.00
EIA Cable 30.00
14
The ordering address is:
Hazeltine Corporation
10 E. 53rd Street 22nd Floor
New York City, NY 10022
GENERAL TERMINAL
General Terminal is the name of the company previously known as Infoton, but none of
the models available now are like the old ones. The Vistar series has changed to the GT
and Visual series. Even though CSO repairs the old Infotons it does not repair any of the
new models.
General Terminal GT 101 $810.00
GT 100 A 710.00
Visual 210 850.00
Visual 200 785.00
EIA Cable 30.00
The ordering address is:
Bronson and Bratton Inc.
5161 South Millard Avenue
Chicago, IL 60632
Phone: 312-735-6200 Robert Raatz
LEAR SIEGLER
Lear Siegler makes a series of ADM terminals similar to the Hazeltine, GT, and Mime.
The reason for its being included is to provide an inexpensive form of graphics. The
ADM 3 A with Retrographic 512 gives a TEK 4010 compatible graphics terminal for less
than $1600 (the 4010 costs $5000).
Lear Siegler ADM 3A U/L Case $695.00
ADM 3A + U/L Case (replaced by ADM 5) 705.00
ADM 5 U/L Case 725.00
ADM 31 U/L Case 825.00
ADM 42 U/L Case 1565.00
Retrographic Card (512) for ADM 3 A or 5 840.00
EIA Cable 30.00
The ordering address is:
Dytec/South Inc.
11657 AdieRoad
Maryland Heights, MO 63043
Phone: 314-569-2990 Robert Finnegan
15
MICROTERM
Microterm is located in the St. Louis area and makes a line of terminals that mimic other
manufacturers' terminals. The Mime 314 can operate as a Hazeltine 1500, an ADM 3 A,
or an ACT IV. The warranty is for two years and ship-back is to the factory.
Microterm MIME 314 $622.00
EIA Cable 30.00
The ordering address is:
Technical Representative
502 Earth City Plaza
Suite 201
Earth City, MO 63045
Phone: 314-291-0001
TEXAS INSTRUMENTS
The Texas Instrument's Silent 700 series and the Dot Matrix 800 series are available.
The 700 series is the thermal printer type which includes the lightweight portable. The
800 series models are meant to be competitive with the Digital Equipment Corporation's
DECwriter.
Texas Instrument Silent 743 U $890.00
970.00
1280.00
1350.00
1430.00
1508.00
1636.00
895.00
1016.00
The ordering address is:
The David Jamison Carlyle Corporation
704 North Wells Street
Chicago, IL 60610
Phone: 312-975-1500
Silent
743 U
743 U/L
745 U
745 U/L
Dot Matrix
810 RO
820 RO
820 KSP
840 RO
840 KSR
16
DIGITAL EQUIPMENT CORPORATION
The Digital Equipment Corporation's DECwriter is available in three models. The
LS34DA and AA differ only in the way the paper is handled. The LAI 20 is a 1200 baud
version of the original LA36.
Digital Equipment DECwriter LA 34 DA $857.00
Corporation LA 34 AA 857.00
LA 120 2015.00
VT-100 1310.00
The ordering address is:
The David Jamison Carlyle Corporation
704 North Wells Street
Chicago, IL 60610
Phone: 312-975-1500
TELETYPE CORPORATION
The Teletype Corporation builds a Decwriter-type device called a Teletype 43. Some like
the appearance of the resulting characters and also the smaller size. Otherwise, the 43 is
similar to the LA34.
Teletype Corporation 4320 AAA (TTL) $955.00
4320 AAK (EIA) 1025.00
4320 AAB 1235.00
The ordering address is:
Bronson and Bratton Inc.
5161 South Millard Avenue
Chicago, IL 60632
Phone: 312-735-6200 Robert Raatz
XEROX CORPORATION
The Diablo-Xerox Terminal is a daisy-wheel printer with a typewriter-like quality. The
application print speed is usually 30 characters/ second.
Xerox Corporation Diablo-Xerox 1740 RO $2337.00
1740 KSR 2508.00
1750 RO 2588.75
1750 KSR 2603.00
630 RO 2085.25
17
The ordering address is:
Xerox Corporation - Printing Systems Division
450 West Algonquin Road
Arlington Heights, IL 60005
Phone: 312-981-7400 Alex Henderson
NIPPON ELECTRIC CORPORATION
The Nippon Electric Corporation has marketed a replacement for the Diablo daisy-wheel
document printer. The Spinwriter has proven to be more reliable, but no more expen-
sive.
Nippon Electric
Corporation
Spinwriter
5510-1 or 7710-1
$2256.00
5520-1 or 7720-1
2559.00
5515-1 or 7715-1
2284.00
5525-1 or 7725-1
2627.00
5530-1 or 7730-1
2256.00
5540-1
2655.00
The ordering address is:
Inland Associates Inc.
13100 Manchester Road
St. Louis, MO 63131
Phone: 314-821-3742 Bob Omer
COMDATA
Comdata has couplers and modems. The T212A is a replacement for the Rixon that we
have used before. The major advantage of doing business with Comdata is their no-cost
ship-back-to-repair policy.
Comdata
Coupler
Modem
302A2-13
302A2-33
T212A
370E2-12
370E2-42
$176.90
226.90
851.90
278.90
328.90
The ordering address is:
Comdata Inc.
7900 North Nagle
Morton Grove, IL 60053
Phone: 312-470-9600 Philip Towle
18
MULTI-TECH SYSTEMS
Multi-Tech Systems supplies a 212A compatible modem for dial-up at 1200 baud only.
For those who do not want the dual 21 2 A modem from Comdata, the MT 212D-S is
available at $560.00.
Multi-Tech
Modem Mt212D-S
(single unit with CA 211 Cable)
Modem MT212D-R
(rack mount with CA 211 Cable)
$560.00
470.00
The ordering address is:
Multi-Tech Systems Inc.
82 Second Avenue S.E.
New Brighton, MN 55112
Phone: 612-631-3550 Tom Heimerman
RADIO SHACK
Radio Shack has offered about the same discount on their TRS80 equipment as the Byte
Shop does on their Apple equipment. It is recommended that demonstrations be given
by these vendors before you purchase. The demo should include your application.
Radio Shack
TRS-80 mod I peripheral
TRS-80 mod II
TRS-80 mod III
TRS-80 Color Computers
TRS-80 Pocket Computers
TRS-80 Color Video Receivers
TRS-80 Videotex Terminals
TRS-80 Software
-22%
-15%
-22%
-22%
-22%
-10%
none
-22%
The ordering address is:
Radio Shack
National Bid Department
1600 One Tandy Center
Fort Worth, TX 76102
The Country Fair Radio Shack can be used for equipment demonstrations since equip-
ment ordered will come to that store.
19
DIGITAL EQUIPMENT CORPORATION
The Digital Equipment Corporation has extended a bulk pricing for microcomputer pro-
ducts based upon the LSI 11/2 and 11/23 systems and software. A sample hardware sys-
tem is shown here:
KDF11-HF
KEF11-AA
MXV11-AA
MXV11-A2
128K/23 with Mem. Management
Floating Point
Multi-function 2 SLU/BOOT/8KRAM
Boot ROM CHIPS
$2880.00
269.00
448.00
38.00
Choice of System Boxes:
BA11-ME
BA11-NE
1024.00
1280.00
Choice of Disc:
5 meg
10 meg
(or alternate disc from DSD)
4320.00
4968.00
Discount is approximately 36%. The ordering address is:
Digital Equipment Corporation
2400 North Main Street
East Peoria, IL 61611
Phone: 309-694-4235 Jim Divit
APPLE COMPUTER SYSTEMS
The Apple Computer Systems are available locally through the Byte Shop. Two sample
systems are shown here:
Basic System
Apple 11+ 48K Computer
Disk II with Controller
12" Black and White Video Monitor
IDS 445G Paper Tiger Printer with Interface
TOTAL BASE BID A $2561 .07
If the 12" Video Monitor is not desired,
subtract from Base Bid 117.25
If the IDS 445G Paper Tiger Printer with
interface is not desired, subtract 755.50
20
Apple ///System Packages
Apple ///'s are sold as package systems. Apple /// option A includes:
Apple /// 128K bytes of RAM
Information Analyst Package
12" Black and White Video Monitor
Silent-type Printer with Apple /// interface
TOTAL BASE BID E $4103.40
Apple II Systems range in price from $2000 to $5000.
Apple /// Systems range in price from $4000 to $7000.
The ordering address is:
Byte Shop
P.O. Box 1678
1602 South Neil Street
Champaign, IL 61820
Phone: 217-352-2323 Dwayne Garrett
DATA SYSTEMS DESIGN
Data Systems Design has offered floppy disc and combination disc (Winchester/ floppy) as
alternative choices to those offered by DEC.
DISC (Data Systems Design) for LSI 11/2 and 11/23 Systems
Alternatives to DEC DISC
DSD 880 D/8-L11-A $5175.00
DSD480L11-2-A 3371.00
DSD 470 3221.00
DSD440L11-2-A 2921.00
DSD 430-A 2546.00
The ordering address is:
Data Systems Design Inc.
2560 Mission College Blvd.
Suite 108
Santa Clara, CA 95051
Phone: 408-727-3163 Tom Walker
21
INTEL TO HOLD OPEN HOUSE
Intel Corporation will be hosting an open house on September 23 at 6:00 P.M. in the Levis
Center. This is by invitation only for faculty, staff and graduate students.
An overview of how Intel views the future of the industry will be given by engineers
representing all divisions within the company.
To receive an invitation, please contact:
John M. Fox
Intel Corporation
Suite 815
2550 Golf Road
Rolling Meadows, IL 60008
Phone: 312-981-7200
NATIONAL EDUCATIONAL COMPUTING CONFERENCE
The National Educational Computing Conference (NECC) has issued a call for participation
for the annual conference to be held in Kansas City, Missouri June 28-30, 1982. Papers to be
presented at this conference should be submitted by January 15, 1982.
NECC provides a broad forum for discussion among individuals from all institutions with
interests at all levels in educational computing. Based on previous conferences, approximately
1000 people are expected to attend. Papers are solicited which describe actual experiences
with classroom computer use or the consequences of such use on the educational process in
general. Papers reporting negative results are also encouraged, especially when the results
could have a profound effect in the way educational computing should be viewed. Guidelines
for submission of papers are available from:
Gerald L. Engel
Computer Science Department
Christopher Newport College
50 Shoe Lane
Newport News, VA 23606
22
SALES - EXCHANGES - HELP WANTED
PROGRAMMER WANTED
Should have considerable experience in handling large data sets on magnetic tape. Work will
include writing, sorting, and merging programs and documentation of tape files. Would need
to be available for approximately 200 hours of work during Fall and Spring semesters; the
majority of work to be done in Fall semester. Send credentials to Dr. K. Parkay, 293 Bevier
Hall, 905 S. Goodwin Ave., Urbana, IL 61801; or telephone 333-0606, or leave message at
333-2412.
INFORMATION WANTED
I am interested in knowing if anyone is using C.A.L. programs on the CYBER or the IBM,
particularly as educational aids in the biological sciences.
I am mainly interested in programs in the following areas:
• ecology
• evolution
• population dynamics
• genetics (population and quantitative)
I would also like to know if anyone is using materials from the CONDUIT system. If you
have information about the above, please contact me:
Keith Garbutt
134 Morrill Hall
Phone: 333-8507 (if no answer, leave a message at 333-6177)
23
SPECIAL EDITION ARTICLES AND INDEX
CYBER FILES - DAILY BACKUP PROCEDURE
Reprinted from OFF-LINE, December 1980
The procedure for providing daily backup of CYBER files has been changed, primarily to
improve the time it takes to restore a destroyed disk pack. Keep in mind while reading about
the new procedure that all backups are done after midnight.
The previous method was to backup daily on tape all files modified within the last 24-hour
period. This was done every day of the week except on Sunday when a full backup of all files
was done. Each of these daily tapes were kept for that week and then reused the next week.
The new method still provides for the full backup of all files on Sunday. However, the daily
backup method has changed. Each day of the week, all files that are on disk and have been
modified since Sunday (the last full backup) are dumped to tape. This implies that a full disk
pack restore only requires the processing of two sets of tapes; a full backup tape and the most
recent incremental tape of files modified since Sunday that are still on disk. Under the old
scheme, the worst case required up to seven sets of tapes.
There is a significant difference in how daily backup tape sets are used. Under the old
scheme, seven tape sets (one for each day) were required; under the new scheme, three tape
sets are required and are used as follows:
Daily
Tape Set Used on
1 Monday and Thursday
2 Tuesday and Friday
3 Wednesday and Saturday
The significance of this is that should you create a file on Monday, it will be backed up on the
Tuesday (early morning) daily tape. If you then accidentally purge the file on Tuesday, it will
not appear on the Wednesday or Thursday incremental tapes. If you then wait until Friday to
come to the Consultants to restore your purged tape, it will be too late because the Tuesday
tape will have already been written over by the early morning Friday tape.
NOTE: This means that you no longer have a full week in which to get
a file restored: you now have only two days!
This approach improves our recovery time in the event of a disk problem, and reduces the
CSO tape storage needs. For example, a worst-case (Saturday) disk failure results in a reload
of four 6250 tapes in contrast to the 50-60 1600 BPI tapes previously required; thus, reducing
the chances of tape problems and considerably simplifying the restore procedure, as well as
requiring less time to spin through all necessary tapes.
24
SPECIAL WEEKEND RATES
Reprinted from OFF-LINE, Jan-Feb, 1981
As of January 23, 1981, low-cost weekend rates were extended to include both batch and
time-sharing on the CYBER 174 and 175 systems. The period of reduced rates is to begin
each Friday at 4 PM and end each Monday at approximately 6 AM when the system is taken
down for engineering. These lower rates are restricted to accounts on internal University
users; they do not apply to bulk jobs.
To take advantage of these lower weekend rates you must both logon and logoff between the
specified hours. For example, if you logon at 4:15 PM and logoff at 5:30 PM on Friday, you
will be charged at the reduced rate. However, if you logon at 3:30 PM and logoff at 5:30 PM
Friday, you will be charged at the usual rate since you logged on before the 4 PM start of the
rate-reduction period.
During the rate-reduction period, batch and time-sharing jobs are charged at 60% of the usual
rate. The rate reduction applies to all charges except those for printing and punching. The
costs reported to you at the end of a batch job or a time-sharing session do not reflect the
reduced rates. The rate reduction is applied at the end of the day when billing information is
entered into the billing database, and the balances at the various accounting levels are decre-
mented by the cost of the job.
Any jobs submitted to be run during this period must finish execution by the 6 AM Monday
deadline. If these jobs are completed after the deadline, they will be charged at the regular
rates. Users who do not wish their jobs to run at the regular rates can use the CANCEL
command early Monday morning to delete (cancel) any job which has not yet completed exe-
cution.
Due to other pressing demands on manpower, no additional software support is being pro-
vided for this service. If practical experience dictates that further software support is
required, the low-cost batch service will be withdrawn until such support can be provided.
NOTE: A similar low-cost service was extended to IBM users in May. Jobs read in by HASP
between 4 PM Friday and 8 AM the following Monday are billed at 60% of the usual rate,
independent of the time of execution. The charges reported on the burst page will reflect the
reduced rates.
25
NEW 6250 BPI TAPE DRIVES
Reprinted from OFF-LINE, Jan-Feb 1981
In September, CSO acquired new tape drives for the CYBER 175, including 9-track drives
capable of reading and writing at densities of both 1600 bpi and 6250 bpi. The transition to
the new drives is now complete, and the tape drive configuration on the CYBER 175 is now:
. Three 9-track 1600 bpi / 6250 bpi drives
. One 9-track 800 bpi / 1600 bpi drive
. One 7-track 800 bpi / 556 bpi drive (which can also read at 200 bpi)
This configuration introduces 6250 capability and reduces 9-track 800 bpi capability from four
drives to one drive.
The 9-track tapes use odd parity. The 7-track tapes use odd parity for binary data and even
parity for coded data.
Successful 6250 bpi usage requires that the type of tape used for writing at this density be the
precise type to which the tape drives have been tuned. Our local drives have been tuned to
Scotch 701 3200 foot tapes rated at 6250 cpi (cpi=bpi), and only that specific tape should be
used here for writing at 6250 bpi. (The critical aspect seems to be the writing of the tape.
Almost any tape written at 6250 bpi appears reliable for reading.) Scotch 701 3200 foot tapes
may be purchased from the Accounting and Distribution Office, 1208 W. Springfield. (Scotch
701 in shorter standard lengths is thicker and will not match the tuning of the drives.)
Tape written at 6250 bpi can hold about three times as much information in a given space as
tape written at 1600 bpi. Thus, 6250 bpi is well suited for recording massive quantities of
data. But remember, our IBM 4341 does not have 6250 drives; therefore, this high density
cannot be used to transfer data between the two computers, nor can the IBM analysis and
data recovery programs be used on any problematic 6250 bpi tape. Also, when preparing tape
for transferring information to another site, 1600 bpi or perhaps 800 bpi is preferable to 6250
bpi because not all sites have 6250 capability.
The usage of 6250 bpi and the new tape configuration require few changes, simply more
attention to the density parameter in the LABEL statement and changes in the RESOURC
statement. The details are summarized below.
The following discussion refers to 9-track tapes only. The usage of 7-track tapes is
unchanged.
The default for CYBER 175 tape usage will remain 9-track 1600 bpi. To specify 9-track tape,
include NT in the LABEL statement or omit the track designation and get 9-track by default.
26
To read or write at various 9-track densities, include the following in LABEL statement:
. For 6250 bpi, always specify D=6250 or D=GE
. For 800 bpi, always specify D = 800 or D=HD
. For 1600 bpi, omit the D= specification and get 1600 by default, or specify D = 1600
or D=PE (including the density specification rather than using the default is good
documentation).
Some examples of 9-track LABEL statements are given below. For more information on the
LABEL statement see Chapter 10 of the CDC NOS Version 1 Reference Manual.
LABEL (TAPE, VSN = ABCDEF-D444,D =GE,PO = W,SI = ABCDEF,QN = 1 ,W)
LABEL (TAPE, VSN=MYTAPE-F555,D=PE,PO=W,SI=MYTAPE,QN=9999)
LABEL (TAPE,NT, VSN =NEWD AT-TEMP,D = 800,PO = R,LB = KU,F =S,CV = EB)
The RESOURC statement is required when more than one tape will be used concurrently in a
job. In the RESOURC statement, you must specify the 9-track drives by the alphabetic char-
acters for density:
. GE for 6250 bpi
. PE for 1600 bpi
. HD for 800 bpi
To specify two 9-track 1600 bpi drives, use:
RESOURC (PE = 2)
To specify one 9-track 800 bpi drive and one 9-track 1600 bpi drive, use:
RESOURC(HD = l,PE = l)
To specify one 7-track drive and one 9-track 6250 bpi drive, use:
RESOURC(MT = l,GE = l)
Users of the EXAMINE and ARCHIVE programs should be aware that the report of the
amount of tape used that is returned by these programs will be too large for 6250 bpi tapes.
Also, the tape density must not be specified in the EXAMINE statement for 6250 bpi tapes,
and EXAMINE does not give a correct report of density for 6250 bpi tapes. In all other
respects, the programs work satisfactorily with these tapes.
27
EASY GRAPHING
Reprinted from OFF-LINE, April 1981
Easy Graphing is here! It is running on the CYBER 174 and CYBER 175. The English-like
commands used in the Easy Graphing package require no programming knowledge and allow
the user to produce plots, bar charts and pie charts with little effort. Some experience with
the CYBER system is helpful, however. Experienced users and programmers will also find
Easy Graphing a useful tool for producing plots because of its easy-to-use commands and its
interactive nature.
The program, as purchased from Tektronix, supports the Tektronix 4006, 4010 and 4014
graphics terminals. Locally we have modified Easy Graphing so it can produce hard copy of
the plot on our ZETA plotters.
For details describing the use of Easy Graphing, the PLOT-10 Easy Graphing Users Manual is
available for purchase at the Accounting and Distribution Office, 1208 W. Springfield. A copy
of this manual is available for inspection in the Systems Consulting Office at 1208 W.
Springfield. The manual is tutorial in style, and is strongly recommended for those users who
have little experience with graphing. A summary of the Easy Graphing commands can be
found in the Easy Graphing User's Reference Guide, also available for purchase at the Distribu-
tion Office. This pocket-sized manual may be sufficient documentation for those users more
experiences with computer graphics software. This guide is quite handy to have during a ter-
minal session.
UPDATES ON PAGE PRINTING SYSTEM
Reprinted from OFF-LINE, August 1981
We are now automatically routing some large print jobs to the Page Printing System (PPS).
Only jobs which run on the local CPUs and which would normally print on standard forms
are affected. Jobs which run at UIC, jobs requesting special forms, and jobs which are routed
to an RJE other than DCL are not affected.
For IBM jobs "large" is defined as more than 30,000 lines of output. For CYBER jobs it is
defined as 4000 or more PRUs of disk space. Both definitions approximate 500 pages of out-
put.
Users can override this default and force a large job to print on the usual line printer by using
the FORMS = NOPPS parameter. For IBM jobs, the parameter appears on the ID card. For
example:
/*ID FORMS = NOPPS
For CYBER jobs, the parameter appears on the PRINT command. For example:
PRINT/FORMS = NOPPS.
28
Users generating large printouts should be aware of the following limitations of the PPS and
specify FORMS =NOPPS if they are a problem:
. The PPS can print at most 64 lines/page. Pages containing more than 64 lines will
overflow to a second page.
• The only carriage control characters honored by the PPS are single space (blank) , page
eject (1), double space (0), and triple space (— ). Any other carriage control character
will cause unexpected results. Note that overprinting is not supported.
We are now dumping PPS output to tape for subsequent printing three times each day.
The schedule is as follows (Monday through Friday only):
Jobs Dumped to Tape Output Available for pickup at DCL
8:00 AM 2:00 PM same day
12:00 Noon 5:00 PM same day
9:30 PM 9:30 AM next day
Depending on the backlog of PPS work and the amount of output to be printed for
CSO, it may be impossible for AISS to produce output in time for the 5 PM pickup
noted above. In this event the output will be available for pickup the following morn-
ing at 9:30.
POLICY
Extension of Computing Services Vol. 8, No. 11, November 1980
CSO Consulting Services Vol. 8, No. 11, November 1980
Vacation Schedules Vol 8. No. 11, November 1980
CYBER Files - Daily Backup Procedure Vol. 8, No. 12, December 1980
Time-Sharing Weekend Rates Reduced Vol. 8, No, 12, December 1980
Special Weekend Rates Vol. 9, No. 2, Jan-Feb 1981
Revised Rates for Plotting Vol. 9, No. 2, Jan-Feb 1981
New Method for Computing Disk Charges Vol. 9, No. 2, Jan-Feb 1981
CSO Short Courses - Spring 191981 Vol. 9, No. 2, Jan-Feb 81
Research Board Announcement: New Computer Allocation Mode (SARA) Vol. 9, No. 4, April 1981
Reaserach Board Deadline for Departmental Allocation Requests Vol. 9, No. 4, April 1981
CSO Short Courses Vol. 9, No. 4, April 1981
CYBER Shared-File system Vol. 9, No. 5, May 1981
Dial-up Usage Control Vol. 9, No. 5, May 1981
Dedicated Lines For Private Facilities Vol. 9, No. 5, May 1981
New Accounting Forms Vol. 9, No. 5, May 1981
Modifications to the Accounting Practices Vol. 9, No. 6, June 1981
CS Courses Now Using Terminals Vol. 9, No. 7, July 1981
Further CSO Moves Vol. 9, No. 7, July 1981
Changes in IBM Disk Policy Vol. 9, No. 8, August 1981
Conversion to New University Accounting Vol. 9, No. 8, August 1981
Updates on Page Printing System Vol. 9, No. 8, August 1981
New Charge Rate For Terminal Repairs Vol. 9, No. 8, August 1981
SYSTEM NOTES
August Reliability Report Vol. 8, No. 10, October 1980
September Reliability Report Vol. 8, No. 11, November 1980
October Reliability Report Vol. 8, No. 12, December 1980
November and December Reliability Report Vol. 9, No. 2, Jan-Feb 1981
January Reliability Report Vol. 9, No. 3, March 1981
February Reliability Report Vol. 9, No. 4, April 1981
Page Printing System (PPS) Now Available Vol. 9, No. 5, April 1981
March Reliability Report Vol. 9, No. 5, May 1981
April Reliability Report Vol. 9, No. 6, June 1981
May Reliability Report Vol. 9, No. 7, July 1981
June Reliability Report Vol. 9, No. 8, August 1981
CYBER
PASCAL Bugs Fixed Vol. 8, No. 10, October 1980
New Diablo Services Vol. 8, No. 12, December 1980
Please Convert to the ZETA Plotter Vol. 8, No. 12, December 1980
FORTRAN Version 5 Vol. 8, No. 12, December 1980
CYBER CANCEL Command Installed Vol. 9, No. 2, Jan-Feb 1981
CYBER 174 Installed Vol. 9, No. 2, Jan-Feb 1981
Gandalf Switch Installed Vol. 9, No. 2, Jan-Feb 1981
More Equipment Changes Vol. 9, No. 2, Jan-Feb 1981
New 6250 BPI Tape Drives Vol. 9, No. 2, Jan-Feb 1981
CSO Videotapes Available Vol. 9, No. 2, Jan-Feb 1981
Changes to Minicomputer Software in GRAB Vol. 9, No. 2, Jan-Feb 1981
A Procedure File for the BILL Command Vol. 9, No. 3, March 1981
New Easy Graphing Software Vol. 9, No. 3, March 1981
Update on Low-Cost Weekend Rates Vol. 9, No. 4, April 1981
Easy Graphing Vol. 9, No. 4, April 1981
PPS Output Vol. 9, No. 5, May 1981
Self-Service Printer Vol. 9, No. 5, May 1981
Changes to the P Command Vol. 9, No. 6, June 1981
CSO Videotapes Available Vol. 9, No. 7, July 1981
FORTRAN Calling PASCAL Procedure Added to UOILIB Vol. 9, No. 7, July 71
FETCH Command Updated Vol. 9, No. 8, August 1981
QUERY Command Updated Vol. 9, No. 8, August 1981
New Version of ICE Vol. 9, No. 8, August 1981
IBM
New Versus Old Vol. 8, No. 10, October 1980
IBM SSP to Go Vol. 8, No. 11, November 1980
Reduction in 2314 Disk Drive Service Vol. 9, No. 4, April 1981
Updates for IBM 3101 Operator Manual Vol. 9, No. 4, April 1981
New Release of SAS on IBM Vol. 9, No. 5, May 1981
STATISTICAL SERVICES
New FOSOL Procedures Vol. 8, No. 12, December 1981
New SOUPAC Procedure Vol. 8, No. 12, December 1981
EFAP Statistical Package Removed Vol. 9, No. 5, May 1981
MATHEMATICAL SERVICES
SLAM - Updates and New Features Vol. 8, No. 10, October 1980
The PORT Library Vol. 8, No. 10, October 1980
Harwell Library on the IBM 4341 Vol. 8, No. 10, October 1980
Q-GERT: Simulation Language for Modelers Vol. 8, No. 11, November 1980
MINPACK on the CYBER Vol. 8, No. 12, December 1980
IMSL Newsletter Vol. 9, No. 2, Jan-Feb 1981
Transactions on Mathematical Software Vol. 9, No. 3, March 1981
Math Notes Vol. 9, No. 3, March 1981
New Math Software Vol. 9, No. 3, March 1981
XMP Updates Vol. 9, No. 3, March 1981
ITPACK Vol. 9, No. 6, June 1981
IMSL - New Newsletter Vol. 9, No. 6, June 1981
Software for Sparse Matrices Vol. 9, No. 7, July 1981
DOCUMENTATION
New Section Added to OFF-LINE Vol. 8, No. 12, December 1981
January-February Issues Combined Vol. 9, No. 2, Jan-Feb 1981
Revised CSO Documentation Vol. 9, No. 2, Jan-Feb 1981
Revised CSO Documentation Vol. 9, No. 4, April 1981
OFF-LINE Mailing List Vol. 9, No. 5, May 1981
Ordering CSO Documentation Vol. 9, No. 5, May 1981
MISCELLANEOUS
Research Board Deadline for Departmental Allocation Requests Vol. 8, No. 10, October 1980
CSO Personnel Honored Vol. 8, No. 10, October 1981
NewCSO Documentation Available Vol. 8, No. 11, October 1981
DEC Users Group Meetings Vol. 8, No. 11, November 1981
Graphics Presentation Vol. 9, No. 2, Jan-Feb 1981
Central Illinois Users Group Meetings (DEC) Vol. 9, No. 2, Jan-Feb 1981
From Plug-Board to Pascal Vol. 9, No. 3, March 1981
Current Weather Data Available Through Computer Access Vol. 9, No. 6, June 1981
CILLILUG Meeting Vol. 9, No. 7, July 1981
Demonstration of Tektronix Equipment Vol. 9, No. 7, July 1981
SPECIAL EDITION ARTICLES
1979-1980 Index Vol. 8, No. 10, October 1980
On-Line Storage Vol. 8, No. 10, October 1980
Magnetic Tapes From Scratch Vol. 8, No. 10, October 1980
FEATURE ARTICLES
Special Plotter Output Service Vol. 9, No. 8, August 1981
New Procedures and Special Requests Vol. 9, No. 8, August 1981
FORTRAN Futures Vol. 9, No. 8, August 1981
OFF-LINE 's Mailing List
If you wish to be placed on our mailing list for future issues of OFF-LINE, if you wish
to be removed from the list, or if you wish to enter an address correction, please com-
plete and return this page. (Current subscribers are kept on the mailing list until a
specific request for removal is received, or until a mailing is returned as undeliver-
able.)
Check one:
□ New subscriber
D Removal request
□ Address correction
Name:
Address:
CAMPUS or Zip Code
(If address correction, give old address and zip code below.)
Comments:
RETURN TO:
OFF-LINE
150 Digital Computer Laboratory
University of Illinois at Urbana-Champaign
1304 West Springfield Avenue
Urbana, Illinois 61801
vOJ>
c
>o
r-m
r-O
om
xi
1/1
no
"0
c
1/1
Computing
Services
/iiy j j_._
Office
Ujj-Line
University of Illinois at Urbana-Champaign
VOL. 9. NO. 10 October 1981
EDITOR: Lynn Bilger
PHONE: (217) 333-6236
139 Astronomy Building
1011 W. Springfield
Urbana, Illinois 61801
Page
1
Contents
POLICY
CSO Short Course Reminder
SYSTEM NOTES
THE -
41981
2
August Reliability Reports
CYBER
3
3
TELENET Telephone Number Changed
New Version of NCAR Graphics Software
STATISTICAL SERVICES
IBM SPSS Release 9 Available
FEATURE ARTICLES
FORTRAN Futures
MISCELLANEOUS
1982 ADCIS Conference
SALES - EXCHANGES - HELP WANTED
Used Hazeltine Keyboard Wanted
CSO DIRECTORY - STAFF AND SERVICES
Administrative
Director
Business Manager
Secretary
User Accounting
Distribution Office
Systems Consulting
Statistical Services Consulting
Text Processing Consulting
Terminal Repair Service
CYBER Dial-up Numbers
Asst Dir User Services
Asst Dir Systems and Operations
Asst Dir Development
Asst Dir Engineering and
Hardware Selection
Manager Statistical Services
Documentation
CYBER-IBM Operations
UNIX Operations
Telecommunications
Laboratory Support Project
RJE Operations North
RJE Operations South
RJE Sites
Agriculture
Chemistry
Commerce West
CRH Snack Bar
DCL Routing Room
Electrical Engineering
Florida Ave Res Hall
Illinois St Res Hall
Mechanical Engineering
Psychology
Social Science
George Badger
150
DCL
333-4103
Stanley Rankin
150
DCL
333-6530
Joyce McCabe
150
DCL
333-1637
rare Support
1208
W Springfield
333-7752
1208
W Springfield
333-6760
1208
W Springfield
333-6133
65
Comm West
333-2170
207
Astronomy
333-7318
150
DCL
333-0969
CYBER 175
110-300
baud
333-4000
CYBER 175
1200
baud
333-4001
CYBER 174
110-300
baud
333-4004
Robert Penka
173
DCL
333-4709
Sandra Moy
177
DCL
333-4703
J. M. Randal
120
DCL
333-9772
Cliff Carter
195
DCL
333-3723
Beth Richardson
91
Comm West
333-2170
Lynn Bilger
139
Astronomy
333-6236
Jack Knott
194a
DCL
333-6562
Debbie Weller
123
DCL
333-8150
Tom Kerkering
164
DCL
333-0816
Mike Gardner
164
DCL
333-7904
Rex Duzan
162
DCL
333-6285
Don McCabe
1208
W Springfield
333-2171
333-7752
M103
Turner Hall
333-8170
153
Noyes Lab
333-1728
70
Comm West
333-4500
120
Snack Bar
333-1851
129
DCL
333-6203
146
EEB
333-4936
FAR
333-2695
ISR
333-0307
65
MEB
333-1430
453
Psych Bldg.
333-7531
202
Lincoln Hall
333-0309
OFF-LINE is the monthly newsletter of the Computing Services Office at the University of
Illinois at Urbana-Champaign. Unless otherwise indicated, permission to reprint is freely
granted, provided that the author, if named, and the Computing Services Office (CSO) are
credited. Information in this issue is current as of September 21, 1981.
CSO operates a CDC CYBER 175 with 256K words of central memory and a CDC CYBER
174 with 196K words of central memory. The 175 and 174 run under the NOS Operating
System and share 512K words of ECS. The 175 serves over 200 simultaneously active text
and graphics terminals and the 174 serves over 100 simultaneously active terminals. CSO
also operates an IBM 4341 with 4 million bytes of memory running HASP-OS/MVT under
VM. In addition, CSO operates a DEC PDP-11/50 with 252K bytes of core driving a GSI
CAT-8 phototypesetter and a DEC VAX 11/780 with 2 million bytes of core, both running
under the UNIX Operating System.
POLICY
CSO SHORT COURSE REMINDER
This is a reminder about the short courses that will be offered this month by CSO. Registra-
tions are accepted one week prior to the start of the course. You may register by coming to
150 DCL, or by calling 333-6630. (See last month's issue for full descriptions of the
courses.)
Introduction to the CYBER Computer Systems
Oct 6, 8, 13, 15
Graphics Compatibility System (GCS)
Oct 5, 7, 9, 12, 14, 16
Introduction to RNF
Oct 19, 21, 23
Intermediate RNF
Oct 26, 28, 30
CYBER Magnetic Tapes
Oct 26, 28, 30
Curve Fitting Routines
Oct 12, 14, 16
SPSS
Oct 5, 7
12 noon - 1 PM
12 noon - 1 PM
2 PM - 3 PM
2 PM - 3 PM
12 noon - 1 PM
2 PM - 3 PM
6:30 PM-9:30 PM
115 DCL
115 DCL
201 Astronomy
201 Astronomy
115 DCL
201 Astronomy
115 DCL
CYBER 175
SYSTEM NOTES
AUGUST RELIABILITY REPORTS
CYBER 174
8 recoverable interruptions
17 non-recoverable interruptions
7 of these were for more than 15 minutes
MTBF = 27.2 hours
MTR = 15 minutes
Availability: 99% of scheduled uptime
Major cause of downtime was related to
TELEX and power problems.
1 recoverable interruption
7 non-recoverable interruptions
5 of these were for more than 15 minutes
MTBF = 85.2 hours
MTR = 35 minutes
Availability: 99.2% of scheduled uptime
Major cause of downtime was related to
main frame and disk controller problems.
IBM 4341:
9 interruptions
7 of these were for more than 15 minutes
MTBF = 77.9 hours
MTR = 60 minutes
Availability: 94.2% of scheduled uptime
Major cause of downtime was related to
power, disk, software, and air condi-
tioning problems.
MTBF = Mean Time Between Failures
MTR = Mean Time to Repair
CYBER
TELENET TELEPHONE NUMBER CHANGED
TELENET recently removed the 384-0011 telephone number for local access to TELENET.
The new telephone number is 384-6428 which is an auto-baud number for both 300 and 1200
baud access to TELENET.
TELENET has upgraded three 300-baud connections to the CYBER to 1200 baud. This
increase in baud rate only means that you will now be able to access the CYBER via
TELENET at either 300-baud baud or 1200 baud. The 300 service is automatically offered
because of the buffering capabilities of TELENET.
NEW VERSION OF NCAR GRAPHICS SOFTWARE
On October 19, a new version of the NCAR Graphics Software will be installed. This version
and all associated documentation are currently available by using either of the following
GRAB,NCAR/FUTURE.
FUTURE. NCAR.
instead of GRAB, NCAR. After October 19, the current NCAR software will be available
only via PAST, NCAR., for a short period before being removed from the system.
The new version of the NCAR library has the following new features:
. New subroutines CONRAN, CONRAQ and CONRAS can contour randomly spaced
data.
. Other new subroutines, PWRITX and SCROLL, generate high-quality text.
• The ZETA plotter metacode translator supports four pens.
. A translator is provided for the Tektronix 4027 color graphics terminal.
The new library is incompatible with the old library in the following ways:
• The subroutine STRML2 has been replaced with STRMLN.
• The metacode file format has changed. Thus, old metacode files cannot be translated
under the new translator.
. The translator parameter file is now in FORTRAN NAMELIST format instead of list
directed format.
A document describing the use of the NCAR library, as well as the translators with their
parameters, can be obtained by entering:
WR1TEUP,NCAR.
The NCARDOC procedure used to obtain on-line documentation on the NCAR library has
also been changed. To use the NCARDOC procedure, enter:
GRAB,NCARDOC.
NCARDOC, routine.
(NOTE: You must use FUTURE,NCARDOC until after October 19.) This will place the
source for the named routine into the local file SOURCE, and a writeup into the file DOC. If
the routine name is omitted from the NCARDOC command, a catalog of NCAR routines is
written instead.
Sample programs illustrating NCAR use are available through the procedure SAMPLES by
entering:
GRAB,SAMPLES.
SAMPLES,NCAR, routine.
(NOTE: You must use FUTURE,SAMPLES until after October 19.) This will give you a
sample program for the named routine. Omit the routine name to receive a catalog of avail-
able samples.
Copies of the complete NCAR manual are now available for purchase at the CSO Distribution
Office, 1208 W. Springfield.
STATISTICAL SERVICES
IBM SPSS RELEASE 9 AVAILABLE
SPSS Release 9 for the IBM 4341 computer is now available. It features three new pro-
cedures for statistical analysis:
. BOX- JENKINS for identification, estimation, and forecasting of univariate time series.
. MANOVA for a general linear models procedure.
. NEW REGRESSION for multiple regression analysis with extensive facilities for resi-
dual analysis.
The documentation for these procedures is contained in the manual SPSS Update 7-9: New
Procedures and Facilities for Releases 7-9, Hull and Nie, McGraw Hill. This manual is available
for purchase at the CSO Distribution Office, 1208 W. Springfield. In addition, the manual is
on reserve in the Undergraduate Library and available for viewing in the user manual rack in
the Statistical Consulting Office, 65 Commerce West.
The IBM SPSS Release 9 software which CSO has obtained is a 1000 variable version of
SPSS. Assuming that you use the default PARM of 80K for workspace and transpace, you
will need to specify a REGION of about 280K when running SPSS programs. At the present
you may access Release 9 by including the following statements in your program:
/*ID REGION = 280K
// EXEC SPSS9
By adjusting the PARM, you can lower the memory requirements as in:
/*ID REGION = 220K
// EXEC SPSS9,PARM = 20K
After November 1, 1981 you may access the various SPSS systems as follows (assuming the
default PARM for workspace and transpace) :
SPSS Release 9
/*ID REGION = 280K
// EXEC SPSS
SPSS Release 8, 500 variable limit
/*ID REGION = 220K
//PROCLIB DD DSN=SYS4. PROCLIB, DISP = SHR
// EXEC SPSSH8
SPSS Release 8, 1000 variable limit
/*ID REGION = 260K
//PROCLIB DD DSN=SYS4. PROCLIB, DISP = SHR
// EXEC SPSSM8
All versions of SPSS prior to Release 8 will be removed from the system on November 1,
1981.
The proc for Release 9 has been changed slightly from previous versions to make it more
compatible with the procs used at other IBM SPSS installations. You may obtain a listing of
the Release 9 proc, a listing of the known errors in Release 9, and a listing of the errors
corrected in Release 9 at the Statistical Consulting Office, 65 Commerce West (open
Monday-Friday 10AM-4PM). Any questions or problems with using SPSS Release 9 should
be brought to the attention of Beth Richardson, Box 86 Commerce West, 333-2172.
FEATURE ARTICLES
FORTRAN FUTURES
This article is part of a series on the work of the FORTR AS Standards Committee X3J3 in producing the
next revision of the FOR TR 4 V standard. The reader is reminded that the features described in this article
are not a pan of am FOR TRAX compiler currently available, but rather are proposed requirements for
FOR TRAN processors produced in the late I980's through the mid I990's. Although every effort has been
made to accurately describe the current position of X3J3 on these matters, the development of a revision to
the FOR TRAX standard is an evolutionary process, and these proposals may be subject to refinement, revi-
sion, or even retraction. Comments on these proposals may be given to Kurt Hirchert of the CSO Systems
Consulting staff, who is a member of X3J3.
One of the areas in which FORTRAN has traditionally been considered weak is the structur-
ing of data. If related pieces of information were of the same type, they could often be com-
bined into an array, but if they were of different types, the options were more limited. In
some cases, an array could still be used by clever application of the EQUIVALENCE state-
ment. Unfortunately, most of these approaches were not in strict conformance with the
FORTRAN standard and would not always transport correctly to new machines. In many
cases the only portable approach was to use related names for the variables containing related
information and with variable names of only six characters, that wasn't easy. X3J3 has
adopted features for inclusion in the next FORTRAN standard which will make this task
much easier.
The first step in using this new facility will be to declare what kinds of information are
related. For example, if one were buffering character input/output, one might use a CHAR-
ACTER variable to hold the characters and two INTEGER variables to indicate where data
should be added and removed from the buffer. The declaration for this might look some-
thing like
FORM BUFFER
INTEGER IN. OUT
CHARACTER* (BUFSIZE) CHARS
END FORM BUFFER
where it is assumed that BUFSIZE is a symbolic constant defined by an earlier PARAMETER
statement. This declaration doesn't actually create a buffer; it just says what one looks like.
Separate declaration can then be used to create as many variables or arrays of buffers as one
wishes.
BUFFER: TTY_BUFFER.TAPE_BUFFER(0:99)
BUFFER: PRINT_BUFFER
The individual fields of a structured variable can be accessed by a combination of the variable
name and the field name.
TTY_BUFFER.CHARS(TTY_BUFFER.IN:TTY_BUFFER.IN) = 'A'
TTY BUFFER. IN = TTY BUFFER. IN + 1
Limited operations are also allowed on structures as a whole. Equivalent structures can be
assigned and compared for equality.
TAPE_BUFFER(6) = TTY_BUFFER
IF (PRINT_BUFFER EQ. TTY_BUFFER) THEN
PRINT_BUFFER.IN = 0
PRINT_BUFFER.OUT = 0
END IF
(Two structures are considered equal if and only if each of the corresponding component
fields is equal.) Similarly, a structure can be written and read as a whole using unformatted
input/output. For formatted input/output, the appearance of a structure name is equivalent
to the appearance of its components in the order they were declared. Thus,
PRINT *, TTY_BUFFER
would be equivalent to
PRINT *, TTY_BUFFER.IN,TTY_BUFFER.OUT,TTY_BUFFER.CHARS
Structures can be passed to subprograms and a function subprogram can return a structured
value. A special case of this is the use of the FORM name as a function for constructing
structure values.
TAPE_BUFFER(J) = BUFFER (0,0,CHARACTER_VARI ABLE)
Although all of the preceding examples involved only a single level of structuring, there is no
reason why one of the components of a structure can't itself be structured.
FORM FILE
INTEGER UNIT
BUFFER: FILE_BUFFER
END FORM FILE
FILE: FILE_INFO,FILE2
FILE_INFO.FILE_BUFFER.IN = FILE_INFO.FILE_BUFFER.IN + 1
FILE2 = FILE(10,BUFFER(0,0,CHARACTER_VARIABLE))
The work on developing this facility in the FORTRAN standard is not complete. Future
developments in this area may include the following:
. The syntax used in this facility will likely be reviewed. The use of a period as the
punctuation between structure and component names is quite natural, but unfor-
tunately it can lead to ambiguities. For example, is A.EQ.B a comparison between
structures A and B, or the component named B of the component named EQ of the
structure A? The syntax used to declare structures is also still subject to much debate.
The current rules consider two structures equivalent if they have the same kinds of
components declared in the same order, but there is some sentiment that operations
should be allowed between structures only if they are declared using the same FORM.
As can be seen in the examples, the length of structure references could become quite
long, so a facility for abbreviating such references may be introduced.
Consideration is being given to "variant" structures, i.e.. structures whose later com-
ponents depend in some way on the value of an earlier component.
The ability to define or redefine operations on structures may prove useful. In our
buffer example, one may have wished an equality test to consider only those charac-
ters between the in and out pointers and one may have wished to define additional
operations such as insertion into the buffer.
The interaction between these data structuring facilities and the new array processing
facilities (to be described in a later article) are being actively considered.
MISCELLANEOUS
1982 ADCIS CONFERENCE
The Annual ADCIS (Association for the Development of Computer-based Instructional Sys-
tems) Conference is a forum for both formal and informal sharing of computer-based instruc-
tion ideas, experiences, research findings, courseware, procedures, policies, standards, and
educational strategies. Academic, commercial, government, and private groups are invited to
attend and participate by discussing ways in which the field of computer-based instruction can
be properly advanced. Demonstrations of hardware, software, and courseware are also soli-
cited.
Papers on all topics relevant to computer-based instruction are welcome, but topics on this
year's themes. Computer Literacy and Intelligent CAI, will be given top priority.
The conference will be held June 7-11, 1982 at the Hyatt Regency Hotel. 655 Burrard Street,
Vancouver, B.C.. Canada. A poster about the conference has been posted at DCL. Further
information may be obtained by calling 800-228-9000 or writing to either of the following:
Prof. Emile Attala ADCIS International Headquarters
ADCIS Conf. Prog. Chairman Computer Center
Dept. Computer Science Western Washington University
California Polytechnic State U. Bellingham. W'A 98225
San Luis Obispo. CA 93407
SALES - EXCHANGES - HELP WANTED
USED HAZELTINE KEYBOARD WANTED
We are looking for a used Hazeltine keyboard, Model #1552. If you have one, or know of
someone who does, please contact:
Al Clemens
Printing Services
134 University Press
Phone: 333-9200
OFF-LINE's Mailing List
If you wish to be placed on our mailing list for future issues of OFF-LINE, if you wish to be
removed from the list, or if you wish to enter an address correction, please complete and
return this page. (Current subscribers are kept on the mailing list until a specific request for
removal is received, or until a mailing is returned as undeliverable.)
Check one: □ New subscriber
□ Removal request
□ Address correction
Name:
Address:
CAMPUS or Zip Code
(If address correction, give old address and zip code below.)
Comments:
RETURN TO: OFF-LINE
150 Digital Computer Laboratory
University of Illinois at Urbana-Champaign
1304 West Springfield Avenue
Urbana, Illinois 61801
- : I L L A fj
LU I . ARY
CA^lPuS
PT
Computing
Services
Office
Off-Line
University of Illinois at Urbana-Champaign
VOL. 9. NO. 11 November 1981
EDITOR: Lynn Bilger
PHONE: (217) 333-6236
139 Astronomy Building
1011 W. Springfield
Urbana, Illinois 61801
POLICY
LiBRAJBTCElBE
1
1
2
The Wheels of Justice
Changes at Chicago Circle
CSO Short Course Reminder
SYSTEM NOTES
JTY OF ILLINOIS
- -»* in**
3
September Reliability Reports
CYBER
4
ZETAVU: A Utility for Previewing
MATHEMATICAL SERVICES
*
4
5
7
9
9
10
12
Transactions on Mathematical Software - TOMS
IMSL Edition 8.1
FORSIM: A Partial Differential
Equations Program
ITPACK Version 2A Installed
REDUCE: A Symbolic Algebra Language
New Version of Slam
Partial Differential Equations
Software from NCAR
13
16
FEATURE ARTICLES
FORTRAN Futures
DOCUMENTATION
New and Revised Documentation
SALES - EXCHANGES - HELP WANTED
18
18
Software Package Wanted
Computer Programmer
CSO DIRECTORY - STAFF AND SERVICES
Administrative
Director
Business Manager
Secretary
User Accounting
Distribution Office
Systems Consulting
Statistical Services Consulting
Text Processing Consulting
Terminal Repair Service
CYBER Dial-up Numbers
Asst Dir User Services
Asst Dir Systems and Operations
Asst Dir Development
Asst Dir Engineering and
Hardware Selection
Manager Statistical Services
Documentation
CYBER-IBM Operations
UNIX Operations
Telecommunications
Laboratory Support Project
RJE Operations North
RJE Operations South
RJE Sites
Agriculture
Chemistry
Commerce West
CRH Snack Bar
DCL Routing Room
Electrical Engineering
Florida Ave Res Hall
Illinois St Res Hall
Mechanical Engineering
Psychology
Social Science
George Badger
150
DCL
333-4103
Stanley Rankin
150
DCL
333-6530
Joyce McCabe
150
DCL
333-1637
rare Support
1208
W Springfield
333-7752
1208
W Springfield
333-6760
1208
W Springfield
333-6133
65
Comm West
333-2170
207
Astronomy
333-7318
150
DCL
333-0969
CYBER 175
110-300
baud
333-4000
CYBER 175
1200
baud
333-4001
CYBER 174
110-300
baud
333-4004
Robert Penka
173
DCL
333-4709
Sandra Moy
177
DCL
333-4703
J. M. Randal
120
DCL
333-9772
Cliff Carter
195
DCL
333-3723
Beth Richardson
91
Comm West
333-2170
Lynn Bilger
139
Astronomy
333-6236
Jack Knott
194a
DCL
333-6562
Debbie Weller
123
DCL
333-8150
Tom Kerkering
164
DCL
333-0816
Mike Gardner
164
DCL
333-7904
Rex Duzan
162
DCL
333-6285
Don McCabe
1208
W Springfield
333-2171
333-7752
M103
Turner Hall
333-8170
153
Noyes Lab
333-1728
70
Comm West
333-4500
120
Snack Bar
333-1851
129
DCL
333-6203
146
EEB
333-4936
FAR
333-2695
ISR
333-0307
65
MEB
333-1430
453
Psych Bldg.
333-7531
202
Lincoln Hall
333-0309
OFF-LINE is the monthly newsletter of the Computing Services Office at the University of
Illinois at Urbana-Champaign. Unless otherwise indicated, permission to reprint is freely
granted, provided that the author, if named, and the Computing Services Office (CSO) are
credited. Information in this issue is current as of October 21, 1981.
CSO operates a CDC CYBER 175 with 262K words of central memory and a CDC CYBER
174 with 196K words of central memory. The 175 and 174 run under the NOS Operating
System and share 512K words of ECS. The 175 serves over 200 simultaneously active text
and graphics terminals and the 174 serves over 100 simultaneously active terminals. CSO
also operates an IBM 4341 with 4 million bytes of memory running HASP-OS/'MVT under
VM. In addition, CSO operates a DEC PDP-11/50 with 252K bytes of core driving a GSI
CAT-8 phototypesetter and a DEC VAX 11/780 with 2 million bytes of core, both running
under the UNIX Operating System.
POLICY
THE WHEELS OF JUSTICE
We are including this article which documents the results of an investigation of theft of com-
puter time to underscore the seriousness of the offense. Computer time costs money and
charges mount up rapidly. So do the penalties.
Last November, a project manager noticed that his project's funds were being depleted at an
unusually rapid rate. He approached us with the problem. We examined our system logs and
determined that someone was indeed making unauthorized use of his funds.
The problem was brought to the attention of Campus Security and an investigative officer was
assigned. CSO worked with this officer and succeeded in identifying two suspects. Both were
eventually apprehended.
It was subsequently determined that each student had used approximately $1500.00 of unau-
thorized computer time.
As the theft of computer time was considered to be a crime against the state of Illinois, the
campus police turned the matter over to the office of the State Attorney. The State
Attorney's office has placed the two students in the Adult Diversion Program as an alterna-
tive to criminal prosecution. Under this program, each student is required to make full resti-
tution of the misused funds and to perform from 50 to 150 hours of public service work.
The public service work is restitution to the state for the time spent in pursuing the case.
The number of hours worked will be equal to the estimated number of hours spent by state
personnel on the case. Failure to fulfill the requirements of the Adult Diversion Program
within a specified amount of time will result in full criminal prosecution of the case. The
offense is considered to be a felony theft punishable by a prison term and up to a $10,000
fine. Finally, the records and transcripts for each student will not be released by the univer-
sity until all public service work has been completed and all stolen funds have been repaid.
The case also was turned over to the Student Disciplinary Committee. The committee put
each student on probation for one year.
CHANGES AT CHICAGO CIRCLE
The computer center at Chicago Circle has informed us that they will install a new operating
system in late December. The new software will not support the HASP to HASP communica-
tions protocol which is used for the current link to UIC. As a result, CSO users will not be
able to submit jobs for execution at UIC after December 22, 1981.
The current UIC system contains a considerable amount of code which was developed by sys-
tem programmers at UIC to support the link with our system. They have indicated that they
would rather not repeat the effort in the new system to be installed in December. Since use
of the link has decreased to the vanishing point and since we can now run jobs requiring large
amounts of core on our system, we have agreed.
There is a "generation gap" between our system and the one in use at UIC which makes com-
munications difficult. We are investigating other ways of handling our RJE network, using
newer software. It is possible that such software will again permit communications with UIC
sometime in the future.
CSO SHORT COURSE REMINDER
This is a reminder about the short courses that will be offered this month by CSO. Registra-
tions are accepted one week prior to the start of the course. You may register by coming to
150 DCL, or by calling 333-6630. (See the September issue for full descriptions of the
courses.)
Introduction to the CYBER Computer Systems
Nov 2, 4, 6, 9
EASY GRAPHING
Nov 16, 18, 20
Introduction to RNF
Nov 16, 18, 20
Algebraic Manipulation Packages
Nov 9, 11, 13
SAS
Nov 2, 4, 9
12 noon - 1 PM
12 noon - 1 PM
2 PM - 3 PM
2 PM - 3 PM
7 PM - 9 PM
115 DCL
115 DCL
201 Astronomy
201 Astronomy
115 DCL
CYBER 175
SYSTEM NOTES
SEPTEMBER RELIABILITY REPORTS
CYBER 174
13 recoverable interruptions
15 non-recoverable interruptions
16 of these were for more than 15 minutes
MTBF = 36 hours
MTR = 22 minutes
Availability: 90% of scheduled uptime
Major cause of downtime was related to
communications hardware and power
outages.
IBM 4341:
14 recoverable interruption
9 non-recoverable interruptions
12 of these were for more than 15 minutes
MTBF = 28 hours
MTR = 53 minutes
Availability: 93% of scheduled uptime
Major cause of downtime was related to
main frame, communications hardware,
and power outages.
21 interruptions
10 of these were for more than 15 minutes
MTBF = 32.8 hours
MTR = 16 minutes
Availability: 95.2% of scheduled uptime
Major cause of downtime was related to
power, CPU, software, and console
problems.
MTBF = Mean Time Between Failures
MTR = Mean Time to Repair
CYBER
ZETAVU: A UTILITY FOR PREVIEWING
ZETA PLOT FILES
All ZETA plot files produced by ZETA software, GCS, NCAR, Easy Graphing or other
software may be displayed on Tektronix Graphics terminals. ZETAVU is a new program
which reads these ZETA plot files (often named TAPE99), interprets the instructions con-
tained therein, and displays the graphics on a Tektronix 4006, 4010, 4014, or 4027 (color)
graphics terminal. ZETAVU is run on the CYBER with the commands:
GRAB, ZETAVU.
ZETAVU.
A 16-page user manual for ZETAVU may be obtained and printed by entering:
WRITEUP, ZETAVU.
PRINT, ZVUDOC/CC/ASCII/EJ.
MATHEMATICAL SERVICES
TRANSACTIONS ON MATHEMATICAL SOFTWARE - TOMS
We now have on tape all algorithms published in Transactions on Mathematical Software up to
September 1981 (from algorithm 493 published in the first issue of March 1975 up to algo-
rithm 577). The directory of algorithms in file TOMS/UN =MATHLIB has been updated; to
print it, enter the following:
GET,TOMS/UN = MATHUB.
PRINT, TOMS.
The "Math Note" series described in the March issue of OFF-LINE includes a note summariz-
ing all algorithms published in TOMS to date. To print this note (it is note number 6), enter
the following:
GET,MNOTES/UN = MATHLIB.
MNOTES.6.
PRINT, OUT/AS/CC/EJ.
(To print instructions on the use of MNOTES, substitute a 0 (zero) for the 6 above.)
The algorithms published in the June 1981 and September 1981 issues are as follows:
In the June issue:
#569: COLSYS: Collocation Software for Boundary-Value ODEs
#570: LOPSI: A Simultaneous Iteration Algorithm for Real Matrices
#571: Statistics for Von Mises' and Fisher's Distributions of
Directions: I[(X)/I0(X),Il5(X)/I05(X),
and Their Inverses
#572: Solution of the Helmholtz Equation for the Dirichlet
Problem on General Bounded Three-Dimensional Regions
In the September issue:
#573
#574
#575
#576
#577
NL2SOL: An Adaptive Nonlinear Least Squares Algorithm
Shape-Preserving Osculatory Quadratic Splines
Permutations for a Zero-Free Diagonal
A FORTRAN Program for Solving Ax = b
Algorithms for Incomplete Elliptic Integrals
If you are looking for mathematical software, Transactions on Mathematical Software is one of
the journals to check, along with Numerische Mathematik, BIT, the Computer Journal, and vari-
ous Publications of ACM (the Association for Computing Machinery) and SIAM (the Society
for Industrial and Applied Mathematics).
IMSL EDITION 8.1
Edition 8.1 of the IMSL subroutine library has replaced the current edition on the CYBER.
Most changes to the library are minor internal corrections which are usually invisible to those
using the routines; no routines have been added or deleted, and the instructions for using the
routines have not changed, except for some clarifications to writeups. The routines for which
a visible change in documentation has occurred or which were corrected for significant inter-
nal errors are as follows:
. BESRB - corrected computation for STAT (9)
. DCSEVU - eliminated dummy dimensioning with a variable that could be zero
. EQRH3F — corrected condition which caused underflow
• GGAMR - modified code so as to ensure correct branch is taken for all values of A
and reinitialized local variables for use in subsequent calls to routine.
. GGETR - modified so as to ensure correct branch is taken for all values of P and Q
. GGCAY -- modified source code to improve efficiency and protect against division by
zero
. GGDA — code changed for greater efficiency. Input vector P is no longer destroyed in
output
. GGNSM — modified code to increase precision in accumulation of RVEC
. GGUW -- zeroed out WK vector in a data statement in order to assure its integrity
from call to call
. LGING -- corrected to handle rows of all zeros
. LLSQF -- made element switching mandatory for KBASIS = 1
. MDFD -- argument checking procedure changed to guard against user input errors
. MDNRIS - revised routine to accommodate arguments less than machine epsilon
(smallest number X such that 1+X .NE. 1)
. MDTD -- corrected code for nonintegral DF
. MDTPS — code modified to avoid underflow and to improve efficiency
. MERFCI -- revised routine to accommodate arguments less than the machine epsilon
. MERRC - corrected results for very small arguments
. NHEXT -- modified code for computation of P(l) and P(2)
. OFCOMM — revised convergence technique to ensure convergence in a finite number
of iterations
. RLEAP - modified code to prevent division by zero when the problem is too ill-
conditioned; extended the definition of I JOB (2) to allow for early termination and
modified code accordingly
• RLFOTH — modified code to avoid division by zero when perfect fit is obtained and
added warning error when perfect fit is obtained with a model of lower degree than
expected when RSQ = 100
. RLFOTW — same as for RLFOTH; also corrected code so that constant response vari-
able will not result in a terminal error
. RLMUL - code modified so that coefficient of variation ANOVA(13) is not computed
if the mean response is zero
. RLONE - code changed to avoid division by zero when dependent variable is con-
stant. Changed documentation of IER = 129
. RSMITZ - modified code and added terminal error condition to avoid possible
overflow
. RSMSSE -- changed code to avoid overflow
. USBOX -- changed to handle constant input correctly
. USBOX 1 -- replaced DASH with PLUS for two adjacent hinges
. USPLT -- corrected plot width when IOPT=0
. USSLF -- eliminated possible resetting of variable IUNIT
. ZXMIN - corrected handling of case where N = l and IOPT=0, and improved accu-
racy of initial hessian when IOPT = 2 or 3
. ZX4LR - corrected inadequate handling of artificial vector
The following is an exhaustive list of routines which suffered some actual code alteration,
whether major or minor:
BESRB
CTRBYC
DCSEVU
EQRH3F
GGAMR
GGETR
GGCAY
GGDA
GGNSM
GGSTA
GGUBT
IQHSCV
LGING
LLSQF
MDFD
MDNRIS
MDTD
MDTPS
ERF
MERFCI
ERFC
GAMMA
ALGAMA
MMBSIO
MMBSI1
MMBSJO
MMBSJ1
MMBSKO
MMBSK1
MMBSYN
MMPSI
MSMRAT
NDKER
NHEXT
NRWRST
OFCOMM
RLEAP
RLFOTH
RLFOTW
RLMUL
RLONE
RSMITZ
RSMSSE
UGETIO
UHELP1
USBOX
USBOX 1
USPLT
USSLF
VUAFB
VUAFQ
VUAFS
VUASB
VUASQ
ZXMIN
ZX4LR
Edition 8.1 updates for the IBM have been received, but are not yet installed.
FORSIM: A PARTIAL DIFFERENTIAL
EQUATIONS PROGRAM
FORSIM, a program for the solution of systems of ordinary or partial differential equations,
has been installed on the CYBERs. The program, from Atomic Energy of Canada at Chalk
River, Ontario, has actually been here in test versions for a few years. It is designed to allow
the user to easily solve a system of ordinary or (parabolic) partial differential equations
without the usual worries of using a library routine with many parameters. The integration
algorithms offered by the package are among the best currently known general methods: the
Runge-Kutta-Fehlberg method, and the Gear algorithm (which is designed for "stiff" sys-
tems). Systems with very low or sparse coupling can be handled efficiently by sparse matrix
techniques in the Gear algorithm.
FORSIM can be used to solve parabolic partial differential equations, like the following:
dU_=f(ndU_ B^U v
Bt JKU,dx'dx2'
where "t" usually represents time and "x" is a spatial variable. Up to 3 space variables can be
handled, with increasing levels of difficulty. Such a system is handled by the method of lines;
spatial derivatives are replaced by finite difference approximations, and the system is reduced
to a time-varying system of coupled ordinary differential equations. Boundary conditions of
two types can be handled:
or
BU_R
where U represents one of the functions at the boundary; Bl, B2, and B3 may depend on U
but not on its derivatives.
The package is accessed by entering the command:
GRAB, FORSIM.
This makes several files available in your local file space, one of which, FORSIM, is a pro-
cedure file. To use the package, you must write (usually) just one routine called UPDATE,
in which the problem is completely defined, then link it to the rest of the package and run it
with a data file. If UPDATE is in file PROBLEM, then the sequence of commands might
look like this:
FTN,REW,I = PROBLEM,L=0,ER,T.
FORSIM, DATA, OUT.
PRINT, OUT/CC.
The package is explained in a manual, The FORSIM VI Simulation Package for the Automated
Solution of Arbitrarily Defined Partial and/or Ordinary Differential Equation Systems from Chalk
River. This manual will be available for purchase at the CSO Distribution Office (1208 W.
Springfield) as soon as we complete a list of errata for it; it can also be ordered directly from
Chalk River for $9.00. The errata will also be put on-line under
WRITEUP, FORSIM.
PRINT, FORDOC/AS/CC/EJ.
for those persons who already may have a copy of the manual.
If you have any questions about FORSIM, please see Stan Kerr (175 DCL, 333-4715, or
TELL,UN=MATHLIB on the CYBER).
ITPACK VERSION 2A INSTALLED
ITPACK, the library of iterative routines from the Center for Numerical Analysis at the
University of Texas at Austin for solving large sparse symmetric positive definite systems of
linear equations, is now at version 2A. It can be accessed by the command:
GRAB, ITPACK.
There is an on-line writeup for ITPACK which can be accessed and printed as follows:
WRITEUP, ITPACK.
PRINT, ITDOC/AS/CC/EJ.
Other software for solving sparse linear systems was discussed in the July 1981 issue of OFF-
LINE.
REDUCE: A SYMBOLIC ALGEBRA LANGUAGE
REDUCE, a language for symbolic algebraic manipulation, has been installed on the CYBER.
The language has been under development by its author, Anthony Hearn, since the late
1960s. Recently a version was created that will run on our system and this is the version we
have installed. It has the following features:
Expansion and ordering of polynomials and rational functions
Symbolic differentiation
Substitutions and pattern matching in a wide variety of forms
Calculation of the greatest common divisor of two polynomials
Automatic and user-controlled simplification of expressions
Calculations with symbolic matrices
A complete language for symbolic calculations in which the REDUCE program itself is
written
Calculations of interest to high energy physicists, including spin 1/2 and spin 1 algebra
Tensor operations
The language is accessed by the following command:
GRAB, REDUCE.
10
Following this, you can enter
REDUCE, version.
where version may be A, B, or C and defaults to A. The versions differ in the amount of
memory they allow (177000 octal for A and C, 250000 octal for B), and whether they include
the high energy physics functions (A and B do, C does not). Other versions with different
characteristics can be created if the need arises. REDUCE is intended to be run interactively,
but it can be run in a batch mode; the manual contains instructions on how to do this.
The REDUCE manual is available on-line and can be obtained and printed as follows:
WRITEUP, REDUCE.
PRINT, REDDOC/AS/CC/EJ.
This prints 89 pages.
If you have questions about REDUCE or other available symbolic manipulation programs,
please contact Stan Kerr (175 DCL, 333-4715 or TELL,UN=MATHLIB).
NEW VERSION OF SLAM
A new version of the SLAM simulation package has been received and installed in FUTURE
on the CYBER. It can be accessed by either of the following commands:
GRAB.SLAM/F. or FUTURE, SLAM.
The memory requirements for the new SLAM have gone up by about 6000 (decimal) words,
to 123000 (octal).
Called SLAM II by the vendor, Pritsker and Associates, the updated package includes the fol-
lowing enhancements:
1. Event Calendar
Information regarding the event calendar (summary statistics, contents) are printed
along with the information about other files. This information appears as that of the
last file in the report, in other words, the next file after the largest file number defined
by the user.
2. MONTR Statement, TRACE option
A negative value included in the list of up to 5 attributes specified in the
MONTR,TRACE statement will result in the printing of the corresponding XX(.) vari-
able instead of an attribute. For example, the statement
MONTR, TRACE, 0.0, 50.0, 2, -1 ;
causes an event trace with ATRIB(2) and XX (1) to be printed from time 0.0 to time
50.0.
11
3. Plotting Capabilities
Plot information generated by SLAM is no longer restricted to storage in either core or
on peripheral devices, but not both. Now, if more than one plot is generated, one
may be stored in core and others on peripheral devices. The limitation that only one plot
may be stored in core is still in effect.
Also, with respect to user-defined plot numbers, it is required that the user begin his
number at 1 and continue sequentially (although it is not required that his RECORCD
statements defining user-defined plot numbers be input in this order), to assure
proper processing of plot information.
4. New Functions
Function FFAWT(IFILE) has been added to return the average waiting time in file
IFILE. This information also appears in the file statistics section of the summary
report.
Function NNBLK(IACT,IFILE) has been added to return the number of entities
currently in activity IACT and blocked by the node associated with file IFILE.
5. Multiple Run Capabilities
Additional input fields have been defined to allow suppression of the summary report
on selected runs and selective clearing of COLCT statistics between runs, as follows:
• The options now available for the ISMRY field on the GEN card are:
N - No summary reports
Y or Y/E - Summary report after every run (default)
Y/F - Summary report after first run only
Y/S - Summary report after first and last runs
Y/n - Summary report after every nth run
. Two optional fields have been added following the JJCLR field on the INI-
TIALIZE card, giving the card the form:
INIT,TTBFG,TTFIN,JJCLR/NCCLR/JCNET,JJVAR,JJFIL;
Statistics to be cleared are defined as follows:
Variable Options Default
JJCLR "Y" - Clear user-collected "Y"
statistics with STAT
code less than NCCLR
"N" - Clear user-collected
statistics with STAT
code greater than or
equal to NCCLR
12
Variable
Options
NCCLR
Integer equal to
breaking point for
clearing of user-
collected statistics
JCNET
ti-vnt
- Clear network-
collected statistics
"N"
- Do not clear
network-collected
statistics
Default
largest
statistics
code defined
+ 1
JJCLR
6. Binary Search Algorithm
SLAM has been upgraded to include a binary search indexed list algorithm to
expedite file entry insertion. To invoke the binary search procedure, only data
input changes are required. An additional field has been added to the LIMITS
statement which specifies the total number of tags to be used by all files. The
format of the LIMITS statement is now
LIMITS, MFIL,MATR,MNTRY,NNTAG;
where NNTAG specifies the total number of tags. NNTAG must be positive if
the binary search algorithm is to be used by any file. When NNTAG is posi-
tive, 3*NNTAG words in NSET/QSET storage are reserved for the NNTAG
tags.
These and other enhancements are described in complete detail in a supplement to the SLAM
text, Introduction to Simulation and SLAM. A copy of this supplement may be inspected at the
Systems Consulting Office, 1208 W. Springfield. The supplement can be ordered directly
from Pritsker and Associates for $15.00 (plus $3.00 handling per order). There is also a six-
panel reference card for SLAM II, printed on heavy stock, which sells for $1.00 (we are ord-
ering a supply of these).
If you have any questions about SLAM, please contact Stan Kerr (175 DCL, 333-4715, or
TELL,UN=MATHLIB) or Dan Pommert (179 DCL, 333-3740).
PARTIAL DIFFERENTIAL EQUATIONS
SOFTWARE FROM NCAR
We have installed on the CYBER a library of routines for partial differential equations
obtained from the National Center for Atmospheric Research (NCAR) in Colorado. The
bulk of the package, which is set up as a subroutine library, consists of FISHPAK, an
integrated collection of routines for solving various classes of elliptic partial differential equa-
tions. For this reason, the library also has been named FISHPAK.
13
This library can be accessed by the command:
GRAB, FISHPAK.
This makes the library available in your local file space and adds it to your global library set;
you then need only run a FORTRAN main program which calls the FISHPAK routines.
Two extra pieces of software are included in FISHPAK. They are routine FINPDF, which
can be used to find the coefficients for a finite difference formula to approximate a given
mixed partial derivative on a rectangular grid; and RKFPDE, a package of routines (accessed
through the main routine RKFPDE) for solving certain systems of parabolic partial
differential equations coupled to a single elliptic equation.
The MATH procedure can be used to obtain documentation or source for FISHPAK routines
as follows:
GRAB, MATH.
MATH, FISH PAK, routine.
where routine is the name of one of the routines in the library. This will provide you with
two local files: SOURCE contains the complete FORTRAN source of the routine requested,
and DOC contains just the writeup. A catalog of short descriptions of all the routines can be
printed as follows:
GRAB, MATH.
MATH, FISHPAK.
PRINT, DOC.
As given, this will print at DCL. A paper by Swartztrauber and Sweet in the September 1979
issue of Transactions on Mathematical Software described a slightly older version of FISHPAK.
FEATURE ARTICLES
FORTRAN FUTURES
This article is part of a series on the work of the FORTRAN Standards Committee X3J3 in producing, the
next revision of the FORTRAN standard. The reader is reminded that the features described in this article
are not a part of any FORTRAN compiler currently available, but rather are proposed requirements for
FORTRAN processors produced in the late IfStls through the mid IVWs. Although every effort has been
made to accurately describe the current position ofX3J3 on these matters, the development of a revision to
the FORTRAN standard is an evolutionary process, and these proposals may be subject to refinement, revi-
sion, or even retraction. Comments on these proposals may be given to Kurt llirchert of the CSO Systems
Consulting stqfj. who is a member of X3J3.
Pipelined and parallel computers are becoming increasingly important, especially in the perfor-
mance of large scientific or engineering calculations. Efficient use of such processors requires
the recognition of the parallelism present in such problems. This recognition is simplified if
14
the languages used to express a computation are capable of making such parallelism explicit
rather than requiring analysis to determine the implicit parallelism. For this reason and oth-
ers, X3J3 has spent a large part of its time and other resources in the development of array
processing facilities for FORTRAN.
Fundamental to these new facilities is the idea that an array can be thought of not only as a
collection of "variables" which can be individually set, but also as "variables" which can be set
to a collection of values. The simplest use of this concept is an array assignment. If A and B
are each 10x10 arrays, the FORTRAN 77 statements
DO 100 1 = 1,10
DO 100 J = l,10
100 A(I,J) = B(1,J)
could be replaced by the single statement
A = B
The scalar operations and intrinsics may be similarly applied in parallel. Thus the statements
DO 200 1 = 1,10
DO 200 J=l,10
A(I,J) = SQRT(B(I,J))*A(I,J)
200 B(I,J) = B(I,J)+A(I,J)
might be replaced by
A = SQRT(B)*A
B = B+A
(This does imply that the new value of the A array will be entirely computed before the new
value of the B array, rather than interleaving the computation as with the DO loops, but in
this case the results are the same.)
Ordinary scalar variables and values can also be freely used in such parallel computations:
B = 0
A = A/A(l,l)
The latter example also illustrates another important principle. The entire right-hand side is
evaluated before anything is stored into the left-hand side so all of A is divided by the old
value of A (1,1) before A(l,l) is changed to its new value of 1.
User-defined scalar functions will also be applicable in parallel if they are declared to be
ELEMENTAL.
Such parallel computation is allowed as long as the arrays involved have the same shape, i.e.,
the same number of elements in each dimension.
15
If we had the following declarations
REALR(10),S(0:9),T(-5:4),U(11),V(10,1),W(1,10),X(2,5)
Arrays R, S, and T could be used together for parallel computations. Arrays U, V, W, and X
could not be used in parallel computation with R, S, and T, or with each other. U is also 1-
dimensional, but has a different number of elements. V, W, and X are each 2-dimensional
with different shapes (10x1, 1x10, and 2x5 respectively).
Since an entire array is often more than is wanted, several new ways of specifying subscripts
have been added. One can specify the value of selected subscripts of an array while leaving
others free by specifying the free subscripts with an asterisk. Thus, V(*,l) could be used in
the previous example to indicate the 1 -dimensional array that is the first (and only) column
of V. Similarly, W(l,*) would be the first (and only) row of W. Both V(*,l) and W(l,*)
could be used with R, S, and T in parallel computations. If all subscripts are specified with an
asterisk, the effect is the same as if the subscripts were not specified at all. Thus, R(*)
denotes the same array as R. If the asterisk is preceded by a minus sign, the elements of the
array are referenced in the reverse direction along that dimension. Thus, R( — *) denotes the
same elements as R, but in the reverse order. Ranges of index values can also be specified.
Thus, U(l:10) denotes a 1 -dimensional array of 10 elements that would be compatible with
R, S, and T. A skip factor optionally may be specified. Thus, T(l:10:2) would denote a 1-
dimensional array consisting of the five odd numbered elements of T. Negative skip factors
can also be used. S(10:l: — 1) denotes the same array as S(— *).
Another available approach will be to make your arrays the right size in the first place. In
FORTRAN 77, a specification such as
REAL Z(N)
is allowed only if Z is a dummy argument of a subroutine or function with its actual argument
properly dimensioned in the calling program. In the forthcoming revision of FORTRAN,
such specifications will be allowed more generally, with arrays of the proper size being created
dynamically when a subroutine or function is called and discarded when the subroutine or
function terminates. This kind of flexibility should not only make it easier to create arrays
that are the right size for the parallel computation to be done in a problem, but should also
make it easier to write programs which adapt to the size of a problem as specified by the data.
Another benefit of such adjustable array specifications will be the elimination of the need to
pass "work" arrays to library routines such as those in IMSL, since such routines will be able
to generate the necessary arrays internally with adjustable array declarations.
Another extension in subprograms will be the ability to specify that the shape of an array
dummy argument is unknown (although the number of dimensions must still be known).
The shape of the array will then be determined by the shape of the array or array expression
that is the actual argument. Intrinsic functions will be available to determine the extent and
upper and lower bounds of each dimension. (These intrinsics could, in turn, be used in an
adjustable array declaration like those described in the preceding paragraph.) Once again, this
will simplify the use of library routines, since it will no longer be necessary to pass a parame-
ter indicating the size of an array separate from the array itself.
16
Although the unconditional parallel computation, extended subscripting, and extended array
specification facilities described in the article do much to extend the array processing features
of FORTRAN, they represent only "the tip of the iceberg." More array processing facilities
will be described in next month's FORTRAN FUTURES.
DOCUMENTATION
NEW AND REVISED DOCUMENTATION
Manuals
Easy Graphing User's Manual $10.00
A tutorial-style manual, strongly recommended for those users who have little experi-
ence with graphing. (See the April 1981 issue for the writeup of Easy Graphing.)
Easy Graphing Reference Guide $1.00
A summary of the Easy Graphing commands.
GCS Primer $7.50
A basic manual; introduction to using GCS (Graphics Compatibility System).
GCS Programmer's Reference Manual $7.50
Alphabetical listing of all GCS subroutines and options.
NCAR Package $14.00
The package combines three NCAR manuals: An Introduction to the SCD Graphics
System; The SCD Graphics Utilities; and The System Plot Package.
IDA $13.00
IDA (Interactive Data Analysis) is a conversational statistical package marketed by
SPSS. (See the September 1981 issue for writeup of IDA.)
SPSS/ONLINE Manual $2.00
SPSS Update Release 7-9 $9.95
SPSS Manual/SPSS Update $18.95
These manuals reflect the updates to SPSS; descriptions of new routines, etc.
17
Introduction to the CYBER Systems Free
This manual replaces the Introduction to the CYBER 175; revised to include usage of
the CYBER 174.
Reference Guides
Many of the Reference Guides were recently updated to reflect the move of the Systems
Consulting Office from DCL to 1208 W. Springfield. Since this was the only change to these
guides, we are not listing them here.
RF-0.1 Reference Guide List -- Has been updated and revised to show the latest revi-
sion date for each guide.
RF-0.2 Documentation List -- Revised to reflect new manuals and prices.
RF-0.3 Job Entry Sites (RJE) — Revised to reflect latest information on equipment at
each site and new hours.
RF-4.19 GRG -- New reference guide. GRG is a program for non-linear optimization
using the generalized reduced gradient method.
On-Line Documentation
On-line documentation has been added for the following products:
HARTMAN - an on-line listing of the computer-related equipment available at
discount through CSO and the University (see September 1981 issue). This listing
may be obtained and printed by entering:
WRITEUP, HARTMAN.
PRINT, COMPDIS.
NCAR - a graphing package (see May 1980 issue for description of software). The
locally-written writeup for this package may be obtained and printed by entering:
WRITEUP, NCAR.
PRINT, NCARWRT/AS/CC/EJ.
ZETAVU - a utility for previewing ZETA plot files (see article on page 4 of this issue).
The writeup may be obtained and printed by entering:
WRITEUP, ZETAVU.
PRINT, ZVUDOC/AS/CC/EJ.
18
FORSIM - a program for the solution of systems of ordinary or partial differential
equations (see article on page 7 of this issue). The writeup contains errata for the
FORSIM manual and may be obtained and printed by entering:
WRITEUP, FORSIM.
PRINT, FORDOC/AS/CC/EJ.
ITPACK - a library of iterative routines (see article on page 9 of this issue). The
writeup may be obtained and printed by entering:
WRITEUP, ITPACK.
PRINT, ITDOC/AS/CC/EJ.
REDUCE - a language for symbolic algebraic manipulation (see article on page 9 of
this issue). The writeup may be obtained and printed by entering:
WRITEUP, REDUCE.
PRINT, REDDOC/AS/CC/EJ.
SALES - EXCHANGES - HELP WANTED
SOFTWARE PACKAGE WANTED
Wanted: Software package to handle lists of references on CYBER. Program must enable
user to sort references by author (s), combinations of authors, year, journal, single keywords
and combinations of keywords. If you have knowledge of such a program please call:
G. Kling
201 A Ornamental Horticulture
333-3363
COMPUTER PROGRAMMER
Available immediately. Ten hours per week until the end of the spring semester. $4.50 per
hour. Requires familiarity with the IBM and the CYBER at DCL. Experience in FORTRAN.
Contact:
Nancy Hansen
116 Natural Resources Building
615 East Peabody, Champaign
Phone: 344-1481, Ex. 203
OFF-LINE's Mailing List
If you wish to be placed on our mailing list for future issues of OFF-LINE, if you wish
to be removed from the list, or if you wish to enter an address correction, please com-
plete and return this page. (Current subscribers are kept on the mailing list until a
specific request for removal is received, or until a mailing is returned as undeliver-
able.)
Check one:
□ New subscriber
□ Removal request
□ Address correction
Name:
Address:
CAMPUS or Zip Code
(If address correction, give old address and zip code below.)
Comments:
RETURN TO:
OFF-LINE
150 Digital Computer Laboratory
University of Illinois at Urbana-Champaign
1304 West Springfield Avenue
Urbana, Illinois 61801
CAMPUS
Computing
Services
r%ff w :-.*
Office
Ujj-Line
University of Illinois at Urbana-Champaign
VOL. 9, NO. 12 December 1981
EDITOR: Lynn Bilger
PHONE: (217) 333-6236
139 Astronomy Building
1011 W. Springfield
Urbana, Illinois 61801
Page
Contents
POLICY
THE LIBRARY BfETHB
1
1
2
Finals Week RJE Schedule
Christmas Schedule
Interruptions to Service
SYSTEM NOTES
UNIVERSITY OF II i noma
3
October Reliability Reports
CYBER
4
5
Graphics Software Trial Period
Terminal Signon Sequences
FEATURE ARTICLES
8
FORTRAN Futures
SALES - EXCHANGES - HELP WANTED
12
Half-time Research Assistant
CSO DIRECTORY - STAFF AND SERVICES
Administrative
Director
Business Manager
Secretary
User Accounting
Distribution Office
Systems Consulting
Statistical Services Consulting
Text Processing Consulting
Terminal Repair Service
CYBER Dial-up Numbers
Asst Dir User Services
Asst Dir Systems and Operations
Asst Dir Development
Asst Dir Engineering and
Hardware Selection
Manager Statistical Services
Documentation
CYBER-IBM Operations
UNIX Operations
Telecommunications
Laboratory Support Project
RJE Operations North
RJE Operations South
RJE Sites
Agriculture
Chemistry
Commerce West
CRH Snack Bar
DCL Routing Room
Electrical Engineering
Florida Ave Res Hall
Illinois St Res Hall
Mechanical Engineering
Psychology
Social Science
George Badger
150
DCL
333-4103
Stanley Rankin
150
DCL
333-6530
Joyce McCabe
150
DCL
333-1637
rare Support
1208
W Springfield
333-7752
1208
W Springfield
333-6760
1208
W Springfield
333-6133
65
Comm West
333-2170
207
Astronomy
333-7318
150
DCL
333-0969
CYBER 175
110-300
baud
333-4000
CYBER 175
1200
baud
333-4001
CYBER 174
110-300
baud
333-4004
Robert Penka
173
DCL
333-4709
Sandra Moy
177
DCL
333-4703
J. M. Randal
120
DCL
333-9772
Cliff Carter
195
DCL
333-3723
Beth Richardson
91
Comm West
333-2170
Lynn Bilger
139
Astronomy
333-6236
Jack Knott
194a
DCL
333-6562
Debbie Weller
123
DCL
333-8150
Tom Kerkering
164
DCL
333-0816
Mike Gardner
164
DCL
333-7904
Rex Duzan
162
DCL
333-6285
Don McCabe
1208
W Springfield
333-2171
333-7752
M103
Turner Hall
333-8170
153
Noyes Lab
333-1728
70
Comm West
333-4500
120
Snack Bar
333-1851
129
DCL
333-6203
146
EEB
333-4936
FAR
333-2695
ISR
333-0307
65
MEB
333-1430
453
Psych Bldg.
333-7531
202
Lincoln Hall
333-0309
OFF-LINE is the monthly newsletter of the Computing Services Office at the University of
Illinois at Urbana-Champaign. Unless otherwise indicated, permission to reprint is freely
granted, provided that the author, if named, and the Computing Services Office (CSO) are
credited. Information in this issue is current as of November 21, 1981.
CSO operates a CDC CYBER 175 with 262K words of central memory and a CDC CYBER
174 with 196K words of central memory. The 175 and 174 run under the NOS Operating
System and share 512K words of ECS. The 175 serves over 200 simultaneously active text
and graphics terminals and the 174 serves over 100 simultaneously active terminals. CSO
also operates an IBM 4341 with 4 million bytes of memory running HASP-OS/MVT under
VM. In addition, CSO operates a DEC PDP-11/50 with 252K bytes of core driving a GSI
CAT-8 phototypesetter and a DEC VAX 11/780 with 2 million bytes of core, both running
under the UNIX Operating System.
POLICY
FINALS WEEK RJE SCHEDULE
The schedule for the RJE sites during finals week, December 14-22 is:
ISR, FAR, CRH (Snack Bar)
Dec 14 - Dec 22 8 AM - 4 PM
COMM WEST
Dec 14 - Dec 22 8 AM - 6 PM
All other sites will maintain regular hours.
CHRISTMAS SCHEDULE
The schedule for the CSO staff offices and the Consulting Offices has not
yet been finalized. Please watch HEARYE for the announcement of these hours.
The schedule for the RJE sites will be as follows:
CSO NORTH (DCL - Routing and Terminal Rooms)
Close at 4 PM
Closed
8 AM - 5 PM
Closed
8 AM - 5 PM
Resume regular hours
Dec 24
Dec 25
Dec 26 - Dec 31
Jan 1
Jan 2 - Jan 3
Jan 4
COMM WEST
Dec 23
Dec 24 - Jan 3
Jan 4 - Jan 15
Jan 18
Close at 5 PM
Closed
8 AM - 6 PM (closed weekends)
Resume regular schedule
ISR, FAR, CRH (Snack Bar)
Dec 22 Close at 4 PM
Dec 23 - Jan 17 Closed
Jan 18 Resume regular schedule
LH, ME
Dec 24
Dec 25 - Jan 3
Jan 4 - Jan 17
Jan 18
Close at 12 noon
Closed
8 AM - 6 PM
Resume regular schedule
CHEM, EE, PSYCH, AGRI
Dec 24 Close at 12 noon
Dec 25 - Jan 3 Closed
Jan 4 Resume regular schedule
INTERRUPTIONS TO SERVICE
There will be several interruptions to our services in the latter part
of December.
During the week of December 27, air conditioning to DCL will be
turned off, necessitating the shutdown of all of our computers.
Equipment will be installed which will minimize the energy used to
cool the computer rooms in DCL during the winter months. Whenever
possible, it will utilize cold outside air to provide cooling rather
than the usual steam powered equipment. If all goes well, installation
will be completed in a single day and service will be restored the
following day. At the time of this writing, we haven't learned the
exact date on which the work will take place. We will publicize it
in a HEARYE bulletin when we are notified of the exact date.
A lesser interruption, but one with an ongoing effect, will concern only
those people who make use of the RJE facilities at DCL. Sometime between
December 15 and January 1 the user facilities, the line printers, plotter,
and card punches now located in rooms 127, 129, and 131 DCL will be moved
to room 16 in the basement of DCL. There will be a slight reduction
in the number of terminals available. The date of the move, which is
expected to take a day, will be announced in a HEARYE bulletin.
On the day of the move, no services will be available at DCL; scheduled
service will continue without interruption at all other RJE sites. This
will complete the series of moves of user facilities first announced
in OFF-LINE last July.
The relocation of the RJE facilities at DCL will allow CSO to consolidate
its professional staff, something we have wanted to do for several years.
The continued deployment across campus, first of RJE equipment and
subsequently of large numbers of time-sharing terminals, has eroded the
status of DCL as the focal point of computing on campus, making this
move possible.
Room 16 is located in the northwest corner of the basement of DCL.
It can be reached conveniently by the stairs adjacent to the West
entrance to the building.
SYSTEM NOTES
OCTOBER RELIABILITY REPORTS
CYBER 175
CYBER 174
9 recoverable interruptions
14 non-recoverable interruptions
( 3 of these were for more than 15 minutes)
MTBF = 29.3 hours
MTR = 7 minutes
Availability: 99.2% of scheduled uptime
Major cause of downtime was related to
communications hardware and air
conditioning problems.
6 recoverable interruption
12 non-recoverable interruptions
( 7 of these were for more than 15 minutes)
MTBF = 38.2 hours
MTR = 30 minutes
Availability: 99.4% of scheduled uptime
Major cause of downtime was related to
main frame, communications hardware,
and air conditioning problems.
IBM 4341
19 interruptions
( 9 of these were for more than 15 minutes)
MTBF = 39.8 hours
MTR = 17 minutes
Availability: 99.6% of scheduled uptime
Major cause of downtime was related to
CPU, software, console, and air cond-
tioning problems.
MTBF = Mean Time Between Failures
MTR = Mean Time to Repair
CYBER
GRAPHICS SOFTWARE TRIAL PERIOD
If you are looking for graphics software for a minicomputer or mainframe, this article may be
of interest to you.
As individual departments and research groups acquire their own computing facilities, they
frequently discover they need some graphics output capabilities for data representation,
engineering drawing, or various types of analyses. Most users have few problems in the
selection of the graphics hardware, but find that the acquisition of software presents many
problems.
The traditional approach taken by many of these minicomputer users has been one of the fol-
lowing:
. Developing their own software which is generally incompatible with other software,
untransportable, usually poorly documented, and lacking in long-term support.
• Obtaining free or inexpensive software from the hardware vendor or an outside source
such as another university. The same problems as above usually result.
We have been asked by a number of minicomputer users to supply or recommend graphics
software. After discussing the problem, we feel that the following requirements must be met:
• The software should be reasonably priced for minicomputer users as well as main-
frame users.
. The software should be transportable among computers.
• The software should have moderate memory requirements.
• The software should recognize current ANSI standardization efforts.
• The software should support readily available graphics hardware.
We feel that meeting these goals would encourage campus-wide sharing of applications
software and easy transporting of application programs between minicomputers and the
CYBERs. It would also provide a large user base for support and development of applica-
tions.
With these ideas in mind, we have negotiated an agreement with Tektronix to examine their
Interactive Graphics Library (IGL). We have a trial copy of IGL running on the CYBER
until the end of February 1982. During this trial period (Nov 1981 - Feb 1982), we
encourage users to examine the documentation and try the software. In particular, any users
who are interested in running copies of IGL on other computer systems should evaluate the
package.
At the end of the trial period, if there is sufficient campus-wide interest, we will purchase
IGL. If there is not enough interest in the package, we will return our copy of IGL to Tek-
tronix. Due to the possibility of this latter option, we do not encourage development of any
large-scale, long-term IGL applications at this time.
We are currently planning meetings to present the IGL package, its features and abilities, and
the pricing structure. If you are interested in attending such a meeting, or in licensing infor-
mation, contact Allan Tuchman, 181 DCL (333-2048).
IGL documentation is available for inspection at the Systems Consulting Office, 1208 W.
Springfield. To access the IGL library of FORTRAN subroutines, enter:
GRAB, IGL
Following this, you simply compile and execute a FORTRAN program which calls the desired
IGL subroutines.
Direct your comments about IGL to Allan Tuchman directly or through
TELL, GRAPHIC.
TERMINAL SIGNON SEQUENCES
It has been brought to our attention that users are sometimes confused by the different sig-
non sequences required at various terminals. We hope this article will lessen the confusion.
If you have used a CYBER terminal at any of the RJE sites, you have probably observed the
colored strip attached to the keyboard. The strip is about the size of three postage stamps
and is green, blue, red, or black in color. Both the color of the strip and the words printed
on it identify the machine to which the terminal is connected.
Terminals marked with a red strip are attached directly to the CYBER 174. Such terminals
can be used ONLY on the CYBER 174. Similarly, terminals marked with a blue strip are
connected directly to the CYBER 175 and terminals marked with a black strip are connected
directly to the IBM 4341. These terminals can be used ONLY on the machines to which they
are connected. On any of these types of terminals, you initiate the signon sequence by
depressing the RETURN (CR) key.
You can probably see the drawback of this feature: terminals cannot be shared between
machines. If all terminals were connected in this fashion, the effect would be to increase the
number of terminals required.
Enter the Gandalf "switch". This device is an electronic switch with terminals connected to
one side and computers to the other. It contains a small processor which communicates with
a terminal user and, on the basis of this interaction, connects the terminal to the computer of
choice. Note the involvement of the terminal user in completing the connection.
The majority of the terminals at RJE sites are marked with a green strip which indicates that
the terminal is connected to the switch. From the above paragraph, it is clear that you must
use a more elaborate signon sequence when using one of these terminals. Before signon can
even begin, you must specify to the switch which computer you intend to use.
As is usual when communicating with machines, there is a prescribed format which must be
followed when communicating with the switch. The signon sequence is documented at each
switchable terminal, but in very brief form. A fuller description of how to logon at one of
these terminals is presented below:
. First, depress the BREAK key once and hold it down for about one second. When
the switch detects the BREAK signal, it puts your terminal into a queue of terminals
waiting for service and responds by sending a sequence of null characters. Unfor-
tunately, most terminals discard any null characters sent to them so you do not
receive any visible acknowledgement that your BREAK character has been received.
(The IBM terminals provided at several RJE sites are an exception. These terminals
will print the null character as a backward question mark.) Some DOs and DON'Ts
follow.
DO hold the BREAK key down for about one second. The switch ignores a
very short BREAK signal, considering it to be electronic noise. While some
terminals will always send a BREAK signal of reasonable duration, others
transmit only while the key is depressed. A good typist using one of these
latter terminals can generate a BREAK signal so short that the switch considers
it noise.
DON'T hit the BREAK key more than once. Once a terminal has been placed
in the service queue, another BREAK character will be treated as a request to
remove the terminal from the queue.
. Wait two seconds and hit the RETURN (CR) key. The two-second wait gives the
switch time to assign a server to your terminal. (A server is a component of the
switch which can communicate with you and complete the connection with your
chosen computer.) The RETURN (CR) key signals the server to proceed. The termi-
nal responds with "enter class" which is the switch's way of asking which computer you
want to use. If your terminal does not respond, there are a couple of possible reasons:
A server has not yet been assigned to your terminal. There are a limited
number of them available and you may have to wait for one to free up. Wait a
couple of seconds and depress the RETURN key again. Try this at least three
times before giving up on this diagnosis.
The switch might not have detected your original request for service. (This
should not happen if you held the BREAK key down for a full second.) If this
has happened, the remedy is to go back to step one and depress the BREAK
key again. This should always be a last-ditch diagnosis, given the switch's
treatment of a BREAK signal for a terminal that is already in the service
queue.
Enter a three-digit code to identify the computer you wish to use, followed by a car-
riage return. The correct code is easy to remember: it is 174 for the CYBER 174 and
175 for the CYBER 175. (Some Computer Science students, using a special pool of
terminals, can specify other numbers. We will not discuss these special numbers
here.) If all goes well, your terminal will respond with "CLASS xxx START', where
xxx is the code you entered. You might, however, get one of the following messages:
CLASS xxx UNASSIGNED
The three-digit code xxx is unknown. You will be prompted for another.
CLASS xxx UNAVAILABLE
The class xxx is temporarily not available. The chosen computer might,
for instance, be out of service. You will be prompted for another code.
CLASS xxx RESTRICTED
The code you have given is not acceptable from the terminal you are
using. You might, for instance, have accidentally given a code that is
recognized but used only for diagnostic purposes. You will be prompted
for another code.
CLASS xxx BUSY
QUEUE SIZE yyyy DO YOU WISH TO QUEUE?
This pair of messages indicates that all paths to the computer you have
chosen are busy and that the number of terminals waiting for a path to
free up to this computer is yyyy. If you want to be put at the end of this
waiting list, enter Y. While you are waiting, you will receive periodic
reports of your queue position. (A BREAK character entered while you
are waiting in the queue will disconnect your terminal from the switch,
offering an escape route if you tire of waiting.) When a path becomes
available, you will be prompted with a START message.
If you do not want to wait, but want to try to connect to another com-
puter, enter N. You will be prompted for another three-digit code. If you
want to abandon your attempt to use the computer, type Q. You will be
disconnected from the switch.
BYE
You waited too long (more than 20 seconds) to enter the requested three-
digit code. Your terminal has been disconnected from the switch and you
will have to start over by depressing the BREAK key.
Enter RETURN (CR) to initiate the usual CYBER signon sequence. From this point,
the switch simply transmits every signal generated by your terminal directly to the
computer. There is no way for your terminal to communicate further with the switch.
The connection will be broken when the computer you are using sends a disconnect
signal to the switch, something that is normally done when you type BYE to logoff.
FEATURE ARTICLES
FORTRAN FUTURES
This article is part of a series on the work of the FORTRAN Standards Committee X3J3 in producing the
next revision of the FORTRAN standard. The reader is reminded that the features described in this article
are not a part of any FORTRAN compiler currently available, but rather are proposed requirements for
FORTRAN processors produced in the late 1980's through the mid IWQ's. Although every effort has been
made to accurately describe the current position of X3J3 on these matters, the development of a revision to
the FORTRAN standard is an evolutionary process, and these proposals may be subject to refinement, revi-
sion, or even retraction. Comments on these proposals may be given to Kurt Hirchert of the CSO Systems
Consulting staff, who is a member of X3J3.
Last month's article described the basics of a FORTRAN array processing facility, including
parallel application of scalar operators and extensions in declaring and referencing arrays. The
description of the FORTRAN array processing facility continues in this month's article.
Although parallel application of scalar operations provides the means to express many kinds
of parallel operations on arrays, there is another common form of array parallelism, opera-
tions along the length of one or more dimensions within an array. Depending on the nature
of results computed, such operations are called recurrences or reductions. Beyond the use of
DO loops, X3J3 has not yet adopted a mechanism for the general expression of recurrences
and reductions, but it has adopted intrinsic functions to express the most common reduc-
tions, including sum, product, maximum and minimum value, logical sum, logical product,
and the count of true values in a logical array.
All of the parallel operations described so far are unconditional, but in many problems, they
need to be applied to selected elements, so conditional forms of these facilities will also be
available. Parallel application of scalar operations is made conditional using a WHERE state-
ment similar to the logical IF in ordinary FORTRAN:
WHERE (logical array expression) array assignment statement
For example
WHERE(A.NE.O)C = B/A
There is also a block-WHERE similar to the block-IF:
WHERE {logical array expression)
array assignment statements
OTHERWISE
array assignment statements
END WHERE
For example
WHERE(A.NE.O)
C = B/A
OTHERWISE
C=l
END WHERE
In either form, evaluation of the right hand side and actual assignment is performed only for
those positions corresponding to the true elements in the logical array (the false elements for
the OTHERWISE block). Obviously, the logical array and the arrays in the assignment state-
ments must have the same shape. This kind of limited evaluation also applies to the
operands of several of the array intrinsic functions, including a function to merge elements
from two array expressions according to a logical mask and extended forms of the reduction
functions which accept a logical mask indicating which elements are to be included in the
computation.
Another important array operation is the selection and reorganization of elements in an exist-
ing array. Array subscripting is a special case of this, but there are applications which go
beyond what one can represent with the subscripting extensions. For example, one may wish
to treat the diagonal of a square array as a 1 -dimensional array or treat a 1 -dimensional array
as though it were a 2-dimensional array with the odd numbered elements in one column (or
row) and the even numbered elements in the other. A general facility, called the IDENTIFY
statement, can be used to define a "virtual" array which is actually composed of elements in
another array referenced by subscripts which are a linear combination of the subscripts of the
"virtual" array. Our examples could be written:
REAL SQUARE(10,10),VECTOR(0:49)
IDENTIFY<10>DIAGONAL(I) = SQUARE(I,I)
IDENTIFY<0:24,2>TWO_COLUMNS(M,N) = VECTOR(2*M + N-1)
IDENTIFY<2,0:24>TWO_ROWS(M,N)=TWO_COLUMNS(N,M)
There are also intrinsic functions to perform some of the more common reorganizations that
cannot be represented by subscripting. These include a matrix transpose function and a func-
tion to create an array of higher dimensionality whose new subscripts have no effect on the
value of element returned (e.g., "spreading" a vector into a matrix whose columns (or rows)
are all identical to the original vector).
There are also facilities for reorganizing in nonlinear ways. 1 -dimensional arrays can be used
as subscripts to perform operations like permutations. The PACK statement can be used to
transfer the elements of an array to a 1 -dimensional array in canonical order and the
UNPACK statement can be used to perform the reverse transfer. Both PACK and UNPACK
can be used under control of a WHERE statement to transfer only selected elements.
10
The remaining array intrinsics fall into several categories:
. Intrinsics for vector dot product and matrix multiply will be provided, although these
operations can also be performed using combinations of more basic operations. Intrin-
sics for matrix inverse, determinant, and the solution of linear systems were proposed,
but are not currently part of the array processing facility because of concerns
expressed by the numerical community that such intrinsics would be used when they
are computationally inappropriate.
. Special purpose selection and reorganization intrinsics include functions to perform cir-
cular and "end-off' shifts, a function to concatenate an array to itself a specified
number of times along a specified dimension, a function which reduces the dimen-
sionality of its argument through selection, and a function which produces a diagonal
matrix with specified values on the diagonal (such as an identity matrix).
. Array generating functions include a function to generate a vector containing a
sequence of integer values and a function to generate a vector containing alternating
sequences of true and false values.
. Intrinsics will be available to isolate the first and last true values in a logical array or
along a specified dimension of such an array.
. Intrinsics will be available to return the dimensionality, shape, bounds, and size of an
array.
Several areas in array processing are still under consideration:
. As noted earlier, no general mechanism for expressing recurrences or reductions has
been proposed.
• The interaction between array processing and data structuring proposals is being inves-
tigated.
. Consideration has been given to integration or regularization of the array processing
facilities and facilities for manipulating character and bit strings. This includes con-
catenation of arrays.
. A means for creating array valued constants is being proposed.
. Work is progressing on a means of providing user-defined array valued functions.
11
Recent Actions by X3J3
X3J3 met November 2-6, 1981 in Yorktown Heights, New York. Actions taken at that meet-
ing include the following:
. The use of REAL control variables in the new block-DO was eliminated. The use of
REAL control variables in the FORTRAN 77 DO loop remains.
. Several formal interpretations were made on the FORTRAN 77 standard. Preliminary
work was done on additional interpretations.
. A proposal was adopted to merge the mechanism for interprocedure data sharing with
the mechanism used for data access in internal procedures.
. Several corrections and clarifications on the array processing facilities were adopted.
. The rules for adjustable array bounds expressions were liberalized in preparation for
user-defined functions returning array values.
. X3J3 expressed support for the proposed change in X3H2's formal program of work to
include the definition of language-independent database manipulation semantics.
. Rules were adopted for argument association of real and complex arguments where
precision requirements are explicitly specified.
. The basics for user-defined generic functions (i.e., functions accepting arguments of
various types) were adopted.
. A variant on the E format item was adopted. It would restrict the exponent value to a
multiple of three, in keeping with notational practice in many engineering disciplines.
. The basics for a macro facility were adopted. In its current form, this facility provides
only text inclusion with no argument substitution.
. A proposal defining the contents of the core language was discussed extensively.
• The problems in resolving the differences between the FORTRAN 77 argument asso-
ciation model and the model proposed for the next standard were explored.
. The meaning of the SAVE and DATA statements in recursive routines was discussed.
Proposals based on the results of these discussions will be presented at a forthcoming
meeting.
• In addition to the development and maintenance of standards for the FORTRAN
language, X3J3's formal program of work calls for the development of a standard for
database manipulation from FORTRAN. The task group charged with primary
responsibility for this work has disbanded. The future of this work was discussed.
12
A presentation was made on event/error handling in an asynchronous multi-tasking
environment.
SALES - EXCHANGES - HELP WANTED
HALF-TIME RESEARCH ASSISTANT
CSO has an opening for a half-time Research Assistant to work as a system consultant. The
primary job function is to aid users experiencing difficulty in using CSO's services. The posi-
tion offers a good opportunity to learn a variety of software packages and languages.
The applicant should be a knowledgeable user of the CYBER system and must have a work-
ing knowledge of FORTRAN. Familiarity with CSO's IBM services or expertise in areas such
as the use of mathematical libraries or graphics software is desirable.
This position will be available January 1, 1982. If you would like to discuss it further, contact
Robert Penka, Assistant Director for User Services, 173 DCL (333-4709).
OFF-LINE's Mailing List
If you wish to be placed on our mailing list for future issues of OFF-LINE, if you wish
to be removed from the list, or if you wish to enter an address correction, please com-
plete and return this page. (Current subscribers are kept on the mailing list until a
specific request for removal is received, or until a mailing is returned as undeliver-
able.)
Check one:
□ New subscriber
D Removal request
D Address correction
Name:
Address:
CAMPUS or Zip Code
(If address correction, give old address and zip code below.)
Comments:
RETURN TO:
OFF-LINE
150 Digital Computer Laboratory
University of Illinois at Urbana-Champaign
1304 West Springfield Avenue
Urbana, Illinois 61801
OkPX
SHI?-1
f>> o— '
►-S-wr
34f-
o>
H2
H ■<
roc
jorti
^<
'
rr
£HS©®[p[s)[
.®
25071 BLACK
25072 LIGHT BLUE
25073 DARK BLUE
25074 LIGHT GRAY
25075 LIGHT GREEN
25076 DARK GREEN
25077 TANGERINE
25078 RED
25079 EXECUTIVE RED
25070 YELLOW
GENUINE PRESSBOARD
ACCO INTERNATIONAL INC
CHICAGO, ILLINOIS 6061&
UNIVERSITY OF ILLINOIS-URBANA
Q 510 84 0F2 C001 v 8 9(1980 1981
Ofl line
3 0112 089505975