Sun-Spots-Request@RICE.EDU (William LeFebvre) (11/10/87)
SUN-SPOTS DIGEST Monday, 9 November 1987 Volume 5 : Issue 58
Today's Topics:
Re: linpack and Eispack for Suns
Re: Sun processors
Rumour mill: T800 transputers on a VME bus
UNIX plot(5) tool
BUG in libmath.a on Sun 3.4
getting parts for a Sun monitor
Sun hardware problems... Help!
Malformed rl in rl_free!
How to find the NFS file system on which a file is mounted (again)?
read/write label problem
Help needed in programming function keys
Fast stat(2) wanted
Need Info on SMD vs SCSI
Need information on Electronic Publishing Systems
Sun compatible tape drives?
Optical Disk Subsystems?
Fractal tool?
----------------------------------------------------------------------
Date: Mon, 26 Oct 87 17:36 EDT
From: SYSRUTH@utorphys.bitnet
Subject: Re: linpack and Eispack for Suns
Thanks to all who replied (I still don't understand why replies get to me
2 days before I see the digest with my message in it, but anyway...).
[[ Very likely because we almost always have problems getting through to
wiscvm to deliver the issues across the bitnet. We may soon do something
on our end to get around this problem. After all, wiscvm isn't going to
gateway forever. --wnl ]]
For the information of other people who may be interested :
Many people referred me to the netlib@anl-mcs.arpa service, which provides
many mathematical/statistical routines over the network. Unfortunately it
doesn't really help me in this case since I need the whole packages, and
netlib was not set up to do this. I would have to request each routine
individually and then weed out the duplications of internally-called
routines. Jack Dongarra at Argonne suggested I call IMSL in Houston to
get the complete package, and another responder suggested I talk to the
engineering group at U of Tennessee (Knoxville) who have a package which
includes Linpack and Eispack. So I will pursue those suggestions.
However in general the ANL network librarian is extremely useful, I am
sure. They have a very extensive list of routines available. For more
information send a message to netlib@anl-mcs.arpa with the text "send
index".
Thanks again.
Ruth Milner
Systems Manager
University of Toronto Physics
------------------------------
Date: Thu, 29 Oct 87 10:21:02 EST
From: ucbcad!ames.UUCP!rochester!srs!matt@ucbvax.berkeley.edu (Matt Goheen)
Subject: Re: Sun processors
I heard a talk about a month ago that said that Sun will have a 25 MIPS
machine out by June (July?) '88 and a 50 MIPS one out by Dec. '88. I
don't think they can do that w/o going multi-processor. And speaking of
MIPS, Suns equation is (from "A RISC Tutorial"):
MIPS = 1 / (I * C * (1 / S))
where:
I = instruction count for a benchmark
C = # of clock cycles per instruction
S = processor clock speed
They give figures for the SPARC (Sun-4) of:
I = 1.2 C = 1.3 S = 16.67
I assume the `I' value is based on the 68030 having an `I' of 1.0 (i.e.
the SPARC value is relative to the value they give the 68030, which is
1.0). The `C' value for the 68030 is given as 5.2. The problem, as I see
it, is that the 5.2 is for (I assume) the ENTIRE 68030 instruction set,
not the sub-set used in the benchmark. This gives a rather LARGE bias
toward the SPARC machine that has a simplified instruction set, and thus
the C value of 1.3. If both `C' values were given for the the actual
instructions used in the benchmarks, I don't think the SPARC would be
quite as fast as this equation would have you believe (relative to the
68030). BTW, the MIPS numbers (in case you don't feel like plugging them
in yourself) are:
68030: 3.21
80386: 3.44 (I = 1.1, C = 4.4)
SPARC: 10.69
Matt Goheen
S.R. Systems
{rutgers,ames,allegra}!rochester!srs!matt
------------------------------
Date: Thu, 29 Oct 87 07:27:33 EST
From: Sid Stuart <sid%linus@mitre-bedford.arpa>
Subject: Rumour mill: T800 transputers on a VME bus
Just to add to the rumour mill, I have it from good sources that a midwest
company will be offering a multiprocessor board with 16 T800 transputers
and a parallel Lisp compiler sometime in the first quarter of 88. The
board fits and runs in a Sun VME bus based system and, if it gets good
utilization of the transputers, should be about 64 VAX mips and 24 Mflops.
That is, if you believe the Inmos sales stuff.
sid stuart
------------------------------
Date: Thu, 29 Oct 87 22:32:21 -0800
From: Bob Brown <rlb@riacs.edu>
Subject: UNIX plot(5) tool
Anyone know of a good UNIX plot(5) display tool for sunwindows? We've
tried using the tek emulation things and so on and nothing was
satisfactory.
Barring something better, I wrote up something called "plotview" that can
invoke a program that generates UNIX plot(5) language and display the
output in a canvas. It has features like a customizable control panel and
macro substitution from that panel to the program and its input.
A tar file of plotview is available via anonymous FTP from host
lavalite.riacs.edu in pub/plotview.tar. A related program is in
pub/functionview.tar, but this program is much less useful, though the
document in that tar file named functionview.r describes in greater detail
the workings of the program.
The moderator may wish to copy plotview.tar into the rice archives.
[[ The moderator has not been able to connect to that host. He will put a
copy of the program in the archives as soon as he can manage to transfer
the file. --wnl ]]
If anyone has something better for this purpose, please let me know.
Bob Brown
RIACS/NASA Ames
------------------------------
Date: Fri, 30 Oct 87 10:33:48 MST
From: hi!kurt@hc.dspo.gov
Subject: BUG in libmath.a on Sun 3.4
/*
The following program tests the return value of
log(-10.1).
According to the man page, the value of the function
should be NaN and errno should be set to EDOM (33).
*/
#include <math.h>
extern int errno;
main()
{
double f;
int e;
/* f should return NaN */
f = log((double) -10.1);
/* errno should be set to 33 */
e = errno;
printf("log(-10.1) = %f\n", f);
printf("errno = %d\n", e);
}
/*****
Under Sun OS 3.4 on a Sun3/75 (68881), this program returned:
log(-10.1) = NaN /* correct */
errno = 0 /* WRONG!!! BUG!!! */
*****/
/***
Kurt D. Zeilenga
Electrical and Computer Engineering
University of New Mexico
Albuquerque, NM 87131
(505) 277-0805
zeilenga@hc.dspo.gov
***/
[[ Under 3.2, this program behaves correctly, except that "log" writes an
error message to stderr. Why is a lowly library routine *writing*
anything to stderr? That's the caller's job! --wnl ]]
------------------------------
Date: Fri, 30 Oct 87 13:07:14 PST
From: acad!robert@sun.com (Robert Wenig ext 609)
Subject: getting parts for a Sun monitor
I've achieved success in getting a replacement flyback transformer for a
Sun monitor. Dyn Service in San Jose is authorized by Phillips Electronics
to sell replacement parts. Their number is 408-432-1090. I spoke with
someone named Chris.
Robert Wenig
Autodesk
------------------------------
Date: 30 Oct 87 21:58:45 GMT
From: "Robert Condon" <rbc@amethyst.ma.arizona.edu>
Subject: Sun hardware problems... Help!
Okay, everybody out there in Sun land, here's a tough one. When I boot up
my Sun 3/260 server, I get the message:
Boot Prom Selftest
PROM Checksum Test
DVMA Reg Test
Context Reg Test
Segment Map Wr/Rd Test
Segment Map Address Test
Page Map Test
Err 2: Addr 0x10FD4000, Exp 0x2C025A97, Obs 0x3C025A97
My interpretation is that one of the Page Map RAMs is not working. I
looked on the CPU board and found 32 RAMs with a 35-nsec access time.
These, I believe, must be the RAMs I'm looking for, since a page
translation must occur as quickly as possible.
I would like to send it to Sun, but we aren't on maintainence (we can't
afford it, of course) and they want over $1000 to fix it in 30 days.
The only alternative I have is to fix it myself, but I don't know THAT
much about the hardware side of computers. I have a VERY knowledgeable
friend who said he could help me, fortunately. It would really help if we
could find out two things: 1) Do I have the correct bank of RAMs? 2) If
so, which one is it? I would think that the information in the message
could tell us about 2), which is VERY important, since there are 32 chips
on a multi-layer board and it would NOT be very fun to replace all of
them.
Can anybody help me?
Thanks in advance,
Chris Ott "Man who fall into blast furnace is
System Manager certain to feel overwrought"
Computer-Aided Engineering Lab
University of Arizona ----------> chris@spock.ame.arizona.edu
------------------------------
Date: 30 Oct 87 20:42:27 GMT
From: roy%phri@uunet.uu.net (Roy Smith)
Subject: Malformed rl in rl_free!
Just now, I noticed a message in my console window which reads "Malformed
rl in rl_free". No indication where it came from (suntools? kernel? some
random daemon?). Anybody know what it means? Couldn't be too bad; I'm
still up.
Roy Smith, {allegra,cmcl2,philabs}!phri!roy
System Administrator, Public Health Research Institute
455 First Avenue, New York, NY 10016
------------------------------
Date: 29 Oct 87 22:26:43 GMT
From: Tony E. Davis <tony%linus@mitre-bedford.arpa>
Subject: How to find the NFS file system on which a file is mounted (again)?
Question:
Given a file (or directory) how would one determine, USING SYSTEM
CALLS, the NFS file system from which it is mounted.
Previous responses:
When I originally posted this question, several people responded to the
effect: 'use "df ." and parse the result'. I hadn't thought of this
and thank those people for their responses, but what I'm really looking
for is the appropriate way to use getmntent(3) and company to solve
this problem. Using getmntent(), setmntent(), endmntent(), and
stat(2), one can search the mount entries much like one searches
password entries with getpwent(), setpwent(), etc. What I'm missing,
however, is the comparison I need to make to determine when to end the
search: How do I find file system information from the stat()
information on the current directory? and What do I compare this with
from stat() information on the mount entries to determine that the
current directory corresponds to the current mount entry? Also, are
there other problems that I need to consider when playing with mount
entries? Any information will be appreciated.
Example:
Say I want to find out which NFS file system mounts the home directory
for 'genericusername'. If genericusername's home directory is
/usr/oddusers/ourgroup/genericusername and the local machine mounts
/usr, /usr/oddusers, and /usr/oddusers/ourgroup from different places
(perhaps even different NFS servers, or different names: for example,
/usr might be mounted from /usr.private), what is the appropriate way
to determine that /usr/oddusers/ourgroup/genericusername gets mounted
with /usr/oddusers/ourgroup.
The Problem:
I'm using Sun's remote execution protocol (rex) to run processes
remotely. The rex protocol takes arguments which allow one to specify
the directory that will become the current working directory for the
remote process, but one specifies that directory via two arguments:
the NFS file system and the directory within that file system. With
the two arguments, the rex protocol can run the remote process in the
requested working directory even if that directory is not normally
mounted on the remote machine. It first ensures that the specified NFS
file system is mounted (if it isn't mounted, rex will mount it for the
duration of the remote process). Then it changes to the desired
directory within that file system.
The remote processes I'm running are started from a program on the
local machine (rather than from the command line), and I want the
remote process to have the same current working directory as the local
program. So, to give rex the appropriate arguments, I need to be able
to find the NFS file system from which the current working directory of
the local program is mounted.
Details:
Sun 3/110 running SunOS 3.2
Thanks in advance,
Tony Davis
(617) 271-2146
The MITRE Corporation MS A455 Burlington Road Bedford, MA 01730 U.S.A.
tony@linus.b.mitre.org linus!tony tony%linus@mitre-bedford.ARPA
------------------------------
Date: 1 Nov 87 19:47:44 GMT
From: umn-cs!haberman@rutgers.edu (Joe Habermann)
Subject: read/write label problem
I had a problem with a Sun not being able to boot up as a result of not it
being able to read the label on its boot disk. On boot up it was saying:
No label found, attempting to boot anyway
And would fail to boot. "Verify label" in diag was agreeing that there
was no label then asked me if I wanted to look for backups then write one.
I said ok, but it failed to write:
diag> v
verify label
read failed #5, read err - header 2, cyl=0 head=0 sector=0
NO LABEL!!!
Do you wish to search for backup labels? y
Backup label found.
id:<Fujitsu - M2351 Eagle cyl 840 alt 2 hd 20 sec 46>
Partition a: starting cyl=0, # blocks = 16560
Partition b: starting cyl=18, # blocks = 34040
Partition c: starting cyl=0, # blocks = 772800
Partition d: starting cyl=397, # blocks = 282440
Partition f: starting cyl=704, # blocks = 16560
Partition h: starting cyl=722, # blocks = 108560
Do you wish to restore the primary label? y
write failed #5, read err - header 2, cyl=0 head=0 sector=0
At this point I feared the worst. I tried running fix on (0,0,0) once but
it reported no errors and then I got the same error when trying to restore
the label again. Finally after trying the above sequence about 20 times
it allowed me to write the label and I was home free. Does anyone know
what KIND of read/write error that is? I am used to hard ECC errors and
the like. I read the diag section pretty carefully. Did I miss
something? I am used to hard ECC errors and the like. Also what is a fix
besides getting lucky? We are running Sun OS 3.4 on a 3/160 with xy450 +
FujiM2351. Much sincere thanks in advance,
Joe Habermann
rutgers!meccts!umn-cs!haberman
haberman@umn-cs.ARPA
haberman@umn-cs.cs.umn.edu
------------------------------
Date: Tue, 27 Oct 87 09:08:34 PST
From: ptsfa!pbhyf!pjc@ames.arpa
Subject: Help needed in programming function keys
/*
** Help needed in programming function keys on the sun.
** Can someone show me the errors of my ways!!!
** Why doesn't this work for function keys?
*/
#include <curses.h>
#include <sys/types.h>
#include <sys/ioctl.h>
#include <sundev/kbd.h>
#include <sundev/kbio.h>
main ()
{
char ch;
int fd;
initscr ();
cbreak ();
noecho ();
nonl ();
keypad (stdscr, TRUE);
fd = open ("/dev/kbd", 0, 0);
/* set the up arrow key */
/* this works */
set_key (fd, 69, "uparrow", STRING+UPARROW);
/* set F4 */
/* this doesn't work */
/* H E L P - why doesn't this work ? */
set_key (fd, 3, "hello", STRING+TF(4));
close (fd);
while (ch != '9')
{
ch = getch ();
printf ("\n:%d: :%c:", ch, ch);
}
endwin ();
}
set_key(fd, station, value, count)
int fd, station, count ;
char value[80] ;
{
struct kiockey key ;
key.kio_tablemask = 0 ;
key.kio_entry = count ;
key.kio_station = station ;
strcpy (key.kio_string,value) ;
ioctl (fd,(int) KIOCSETKEY,(char *) &key) ;
}
------------------------------
Date: Thu, 29 Oct 87 19:00 EDT
From: GLAUBMAN%nuhub.acs.northeastern.edu@relay.cs.net
Subject: Fast stat(2) wanted
I need a fast way to get the size of a file. It resides on the same file
system as the inquiring program, and the program knows its i-node number.
The file is not necessarily open at the time. Stat(2) is much too slow. I
have some SystemV code which reads the inode, but it won't compile under
Sun OS3.2 because of the v-node stuff. If you have any ideas, send mail
to glaubman@northeastern.edu on CSNet.
Thanx,
David Glaubman
Northeastern University
Boston, MA
(617) 437-4579
------------------------------
Date: 30 Oct 87 15:23:19 GMT
From: harvard!bunny!mdf0@seismo.css.gov (Mark Feblowitz)
Subject: Need Info on SMD vs SCSI
We are new Sun buyers looking for information on disk performance. We were
originally going to buy SCSI 141Meg drives from Sun to run on 2 or 3 3/60
nodes, but heard a rumor that the SCSI performance is several times slower
than SMD performance. We are now considering getting 1-2 diskless 3/60's
and a 3/260 with a 280Meg SMD disk (also from Sun) as a file server.
We would appreciate advice from experienced Sun users on the pros and cons
of the two configurations. We will be running Lisp and Prolog, with some
typical AI-ish tools, and will have as much memory as neceeesary to
minimize paging. Any suggestions? Thanks.
Mark Feblowitz GTE Laboratories, Inc., 40 Sylvan Rd. Waltham, MA 02254
(617) 466-2947
CSNET: feblowitz@GTE-LABS.CSNET
UUCP: feblowitz@bunny.UUCP old UUCP: harvard!bunny!mdf0
------------------------------
Date: 30 Oct 87 16:26:14 GMT
From: peterj@rosevax.rosemount.com (Peter Juhl)
Subject: Need information on Electronic Publishing Systems
Our company is finally considering buying a medium-range electronic
publishing system on the order of TEXET or Interleaf. We desperately need
advice on other systems in this price range. We need names, and if you
have used any of these systems, I would love to hear your impressions.
If you work for a company that produces any of these products, and if you
have an unfair bias, I don't care. Write me mail or post here. You may
net your company a sale, because we are ready to buy! (I suppose it would
be out of the realm of propriety to post an ad for your company to the
net, now that I think of it.) Well, send me mail, then.
[[ I don't case what you mail to Mr. Juhl, but if you send a product blurb
to the digest, please keep the biased statements to a minimum. Try to
stick to the facts. --wnl ]]
ANY information would be very much appreciated!
Peter N. Juhl
Rosemount Inc
12001 Technology Drive
Eden Prairie, Minnesota 55344
(612)828-3381
peterj@rosevax.rosemount.COM
------------------------------
Date: 29 Oct 87 07:52:24 EST (Thu)
From: drilex!dricej@harvard.harvard.edu
Subject: Sun compatible tape drives?
We're finally getting serious about getting some Suns, and I need some
information about compatible hardware.
First, does anyone know of a cartridge tape drive for a PC-AT running
Xenix which can produce tapes compatible with Suns? I've talked to
Emerald; their tapes are QIC-2; Sun's are QIC-11 or QIC-24. What are
people's experiences with cartridge interchangability?
Also, are there any recommendations on getting to a laser printer from the
Sun? I'm looking at buying a Laserwriter and getting Transcript. It
appears that cheaper packages (PPI, etc) offer less function. What are
your experiences? Are there other combinations that work well? (Elan
stuff with a LaserJet, perhaps?) Note that most of our printing will be
source listings, not troff output.
(Actual experiences can be sent to the list; manufacturer's product blurbs
are welcome also, but should be mailed to me.)
Craig Jackson
UUCP: {harvard!axiom,linus!axiom,ll-xn}!drilex!dricej
BIX: cjackson
------------------------------
Date: Thu, 29 Oct 87 08:38:33 PST
From: aidesign!mel@sun.UUCP (Mel Sibony)
Subject: Optical Disk Subsystems?
We have a Sun 3/280 and are looking for an optical disk subsystem which
perhaps uses the CDC Laser Drive disk. Does anyone have experience with a
manufacturer who is providing a very clean working solution or perhaps of
a software driver that is available for Suns? We'd be very interested to
hear from someone who has had one running for at least 2-3 months and can
provide us with reliability, ease of installation, maintenance, and
software compatibility information. We are looking primarily for an
archival type device.
Thanks in Advance
------------------------------
Date: 26 Oct 87 16:13:51 GMT
From: hao!nbires!cadnetix!sharp@rutgers.edu (Jim Sharpe)
Subject: Fractal tool?
Two years ago at the '86 Siggraph, Sun was running a demo in their booth
which I would like some futher information on. It was a fractal
generating suntool which utilized several machines to simultaneously
compute portions of the image, one of which was a very impressive Convex.
Who wrote this software? Is it in the public domain? Where can I get a copy?
Thanx in advance.
Jim Sharpe Cadnetix Corporation
5757 Central Ave. Boulder, CO 80301
UUCP: nbires!cadnetix!sharp (303) 444-8075
------------------------------
End of SUN-Spots Digest
***********************