Blinkenlights (Re: New Book: Programming Embedded Systems in C and C++)
Jens Kilian
> #Old-time hackers sometimes get nostalgic for blinkenlights because
> #they were so much more fun to look at than a blank panel. Sadly, very
> #few computers still have them (the three LEDs on a PC keyboard
> #certainly don't count). The obvious reasons (cost of wiring, cost of
> #front-panel cutouts, almost nobody needs or wants to interpret
> #machine-register states on the fly anymore) are only part of the
> #story. Another part of it is that radio-frequency leakage from the
> #lamp wiring was beginning to be a problem as far back as transistor
> #machines. But the most fundamental fact is that there are very few
> #signals slow enough to blink an LED these days! With slow CPUs, you
> #could watch the bus register or instruction counter tick, but at
> #33/66/150MHz it's all a blur.
The BeBox (no longer manufactured by Be, Inc.) and the new "Silverline" PCs
from Fujitsu have blinkenlights; under BeOS, they are used to show current
CPU usage.
Bye,
Jens "mine are pegged at 100% cracking RC5" Kilian.
--
mailto:j...@acm.org phone:+49-7031-14-7698 (HP TELNET 778-7698)
http://www.bawue.de/~jjk/ fax:+49-7031-14-7351
PGP: 06 04 1C 35 7B DC 1F 26 As the air to a bird, or the sea to a fish,
0x555DA8B5 BB A2 F0 66 77 75 E1 08 so is contempt to the contemptible. [Blake]
Gene Gajewski
news:F95o9...@midway.uchicago.edu...
<snip>
> The BeBox (no longer manufactured by Be, Inc.) and the new
> "Silverline" PCs from Fujitsu have blinkenlights; under BeOS, they
> are used to show current CPU usage.
They don't make out to commercial products, but blinkenlights are alive and
well in product development & R&D... we love 'em....
Mike Albaugh
: Jens Kilian <Jens_...@bbn.hp.com> wrote in message
: news:F95o9...@midway.uchicago.edu...
:
: > The BeBox (no longer manufactured by Be, Inc.) and the new
: > "Silverline" PCs from Fujitsu have blinkenlights; under BeOS, they
: > are used to show current CPU usage.
:
: They don't make out to commercial products, but blinkenlights are alive and
: well in product development & R&D... we love 'em....
Amen. And inside the box, you will often find them, for production-
line trouble-shooting. I once added a resistor, cap, and voltmeter to
(essentially) the "servicing an IRQ" line of a processor, so I could
dynamically observe the IRQ load. This gave me the idea (never pursued)
that a "home computer" that never got built should have an "off the
shelf" dual VU meter displaying cache-misses and page-faults :-)
Mike
| alb...@agames.com, speaking only for myself
Victor Eijkhout
> #Old-time hackers sometimes get nostalgic for blinkenlights because
> #they were so much more fun to look at than a blank panel. Sadly, very
> #few computers still have them
Didn't (doesn't?) the Intel Paragon have blinkenlights to indicate
which of the links between processors are active?
The CM5 definitely had oodles of blinkenlights!
--
Victor Eijkhout
"Presidential hopeful Pat Buchanan's assertion ... that he personally
is not descended from monkeys explains a lot to those of us who _are_
from this planet." -- Gerald L. Epstein in Science, Aug. 16, 1996
gr...@apple2.com
alb...@agames.com (Mike Albaugh) wrote:
>Gene Gajewski <ggaj...@asctmd.com> wrote:
>>Jens Kilian <Jens_...@bbn.hp.com> wrote:
>>> The BeBox (no longer manufactured by Be, Inc.) and the new
>>> "Silverline" PCs from Fujitsu have blinkenlights; under BeOS, they
>>> are used to show current CPU usage.
>> They don't make out to commercial products, but blinkenlights are alive and
>> well in product development & R&D... we love 'em....
> Amen. And inside the box, you will often find them, for production-
> line trouble-shooting. I once added a resistor, cap, and voltmeter to
> (essentially) the "servicing an IRQ" line of a processor, so I could
> dynamically observe the IRQ load. This gave me the idea (never pursued)
> that a "home computer" that never got built should have an "off the
> shelf" dual VU meter displaying cache-misses and page-faults :-)
Yeah, about the only place you still find them in consumer products are
external modem status lights, and even then they move them into the case
and just maybe simulate them with some software on screen.
Nowadays you're hard-pressed to find a single drive indicator light in a
current production system. (With Apple's G3s, you can see some of the
glow of the laser of a DVD-ROM drive through the blue panels.)
--
Nine quadrillion, nine hundred ninety-nine trillion, nine hundred ninety-nine
billion, nine hundred ninety-nine million, nine hundred ninety-nine thousand,
eight hundred ninety-seven bottles of beer. You take one down, pass it around,
nine quadrillion, nine hundred ninety-nine trillion, nine hundred ninety-nine
H C Pumphrey
Victor Eijkhout <eijk...@nala.cs.utk.edu> writes:
|> The CM5 definitely had oodles of blinkenlights!
Now, is this the thing that looks like a large black cube covered with
red lights that light up in strange moving patterns to either
(a) show patterns of usage among the many processors
or
(b) mean nothing but look *DEEPLY* cool while doing so?
Or am I thinking about somenting else?
Hugh
--
==========================================================================
Hugh C. Pumphrey | Telephone 0131-650-6026
Department of Meteorology | FAX 0131-662-4269
The University of Edinburgh | Replace 0131 with +44-131 if outside U.K.
EDINBURGH EH9 3JZ, Scotland | Email h...@met.ed.ac.uk
OBDisclaimer: The views expressed herein are mine, not those of UofE.
==========================================================================
Barry Margolin
H C Pumphrey <h...@scotsman.ed.ac.uk> wrote:
>In article <F96p3...@midway.uchicago.edu>,
>Victor Eijkhout <eijk...@nala.cs.utk.edu> writes:
>
>|> The CM5 definitely had oodles of blinkenlights!
>
>Now, is this the thing that looks like a large black cube covered with
>red lights that light up in strange moving patterns to either
The black cubes were the CM-1 and CM-2. The CM-5 looks like a bunch of the
2001 slabs in a stairstep pattern, with blinkenlights along the thin edges.
The CM-1/CM-2 blinkenlights were definitely better.
--
Barry Margolin, bar...@bbnplanet.com
GTE Internetworking, Powered by BBN, Burlington, MA
*** DON'T SEND TECHNICAL QUESTIONS DIRECTLY TO ME, post them to newsgroups.
Please DON'T copy followups to me -- I'll assume it wasn't posted to the group.
Ric Werme
>s...@s.net (Schol-R-LEA) writes:
>> #Old-time hackers sometimes get nostalgic for blinkenlights because
>> #they were so much more fun to look at than a blank panel. Sadly, very
>> #few computers still have them
The main reason I get nostalgic for the lights is because they were
useful. Fun yes, but, very useful if you knew what they meant.
A lot of field service people didn't.
>Didn't (doesn't?) the Intel Paragon have blinkenlights to indicate
>which of the links between processors are active?
>The CM5 definitely had oodles of blinkenlights!
When I was at Alliant, most of the OS crew had PDP-10 experience with
the KA and KI models (lotsa lights). We'd often go through a hardware
design meeting and at the end tell the hardware folks what lights they
missed. They wouldn't give us too many, but we had enough to tell
which processors and I/O channels were busy. When we did a Cluster
with HIPPI links I wrote a test program that would pass a buffer from
node to node. We could watch the buffer getting passed from node to if
we made it big enough.
IIRC, the CM-5 lights were simply an output peripheral, they didn't
indicate what the machine was doing. Earlier models had light coupled
to the real action.
Some other poster posited that signals are too fast these days for meaningful
lights. I disagree. We regularly identified these types of problems
on the -10s from the lights, and could today:
Halt (frozen lights)
Monitor loop (user mode off, many lights stuck on or off, only
a few in between)
User loop (as above, but user mode on. I told a lot of users
their program was in an infinite loop.)
PI-7 glow (a scheduler hang where more and more processes are
waiting for something to complete, making the schedule look deeper
for something to run. The scheduler ran in Program Interrupt level
7.)
These days, Windows hangs, you have no idea what's really happening.
--
Ric Werme | http://people.ne.mediaone.net/werme
we...@mediaone.net | http://www.cyberportal.net/werme
Ken
> Some other poster posited that signals are too fast these days for meaningful
> lights. I disagree. We regularly identified these types of problems
> on the -10s from the lights, and could today:
When I worked around -10's, we sometimes had a high-baudrate terminal
line open because someone had unplugged the terminal, and the
unterminated wire would act like an antenna. The -10 would freeze from
the interrupt activity. This could be quickly diagnosed by observing
that light zero, representing the null job, was solidly lit. Our
long-term solution was to connect some pull-down resistors across the
receive lines on the UART boards.
There's an amusing story about a magic switch in a PDP-10 in the Jargon
File. The current link is "http://www.jargon.org/jargon_toc.html#SEC53",
or go to the top and search for "magic".
--
Ken
mailto:sh...@well.com
http://www.well.com/user/shiva/
Carl R. Friend
>
> Some other poster posited that signals are too fast these days for
> meaningful lights. I disagree. We regularly identified these types
> of problems on the -10s from the lights, and could today:
>
This could be one of two things: 1) a real HALT (RUN light off) or
2) a wedge (RUN light on, indicators frozen). On KAs and KIs, you could
tell what timestate got trashed (unless you got _very_ unlucky and the
state didn't transfer correctly).
> Monitor loop (user mode off, many lights stuck on or off, only
> a few in between)
The brightness of the indicator is a gague of the duty cycle of the
line which the indicator is attached to. In reality, that lamp is
blinking faster than you can see it (and in the case of incandescent
lamps faster than the filiament can cool off). So, yes, even though
it's blinking "too fast to be useful" you still get useful information
from it. I agree with the thesis that lights would be useful today.
They're just too expensive (and RF unfriendly, possibly).
> User loop (as above, but user mode on. I told a lot of users
> their program was in an infinite loop.)
This trick only works until the scheduler preempts the looping job.
In the -10s, though, the currently executing job was displayed in the
data lights on the console (right halfword). On dual CPU systems, the
job running on the slave showed up in the left halfword.
> PI-7 glow (a scheduler hang where more and more processes are
> waiting for something to complete, making the schedule look
> deeper for something to run. The scheduler ran in Program
> Interrupt level 7.)
Wasn't this usually the precursor to a DED (Dead Scheduler)
stopcode?
> These days, Windows hangs, you have no idea what's really happening.
Even when it seems to be running, you can't be too sure.
--
+------------------------------------------------+---------------------+
| Carl Richard Friend (UNIX Sysadmin) | West Boylston |
| Minicomputer Collector / Enthusiast | Massachusetts, USA |
| mailto:carl....@stoneweb.com | |
| http://www.ultranet.com/~crfriend/museum | ICBM: N42:22 W71:47 |
+------------------------------------------------+---------------------+
Ric Werme
>Ric Werme wrote:
>> We regularly identified these types
>> of problems on the -10s from the lights, and could today:
>>
>> Halt (frozen lights)
> This could be one of two things: 1) a real HALT (RUN light off) or
>2) a wedge (RUN light on, indicators frozen).
Memory is getting fuzzy, but I think on the KAs there was a group of six
lights on the left side of console:
1 4
2 5
3 6
Lights 1-3 were usually on, 5 was user mode, I think 4 was "Program stop"
and lit when halt executed. In the 4 series monitor, when it detected
something was wrong, it just halted. In the 5 series, it used the STOPCD
macro, range bells, etc. Fortunately that came along just before the
engineering lab hooked up a siren to the Halt light. That siren would have
woken a dead operator!
>> Monitor loop (user mode off, many lights stuck on or off, only
>> a few in between)
> The brightness of the indicator is a gague of the duty cycle of the
>line which the indicator is attached to. In reality, that lamp is
>blinking faster than you can see it (and in the case of incandescent
>lamps faster than the filiament can cool off).
Incandescent lamps also seem to have a more "human friendly" apearrance.
One running at a 50% duty cycle "looks" right. A LED at 50% is only 3 db
down from full brightness and is looks almost like one at full brightness.
>> User loop (as above, but user mode on. I told a lot of users
>> their program was in an infinite loop.)
> This trick only works until the scheduler preempts the looping job.
>In the -10s, though, the currently executing job was displayed in the
>data lights on the console (right halfword). On dual CPU systems, the
>job running on the slave showed up in the left halfword.
That only showed up if you had the console address switches set to the
right location. IIRC, that location was maintained for the lights.
Another couple locations were used by the scheduler to tell the swapper
who to swap in or out. You could set them manually and the system would
do it.
>> PI-7 glow (a scheduler hang where more and more processes are
>> waiting for something to complete, making the schedule look
>> deeper for something to run. The scheduler ran in Program
>> Interrupt level 7.)
> Wasn't this usually the precursor to a DED (Dead Scheduler)
>stopcode?
Hmm. don't recall.
>> These days, Windows hangs, you have no idea what's really happening.
> Even when it seems to be running, you can't be too sure.
And when it does die, you can't dump it for post-mortem analysis.
--
Ric Werme | http://people.ne.mediaone.net/werme
we...@nospam.mediaone.net | http://www.cyberportal.net/werme
^^^^^^^ delete
Phil Gustafson
we...@nospam.mediaone.net (Ric Werme) writes:
>The main reason I get nostalgic for the lights is because they were
>useful. Fun yes, but, very useful if you knew what they meant.
>A lot of field service people didn't.
>
>When I was at Alliant, most of the OS crew had PDP-10 experience with
>the KA and KI models (lotsa lights). We'd often go through a hardware
>design meeting and at the end tell the hardware folks what lights they
>missed...
On the KA-10, this problem was simplified by a design rule: one light
per flip-flop. The [BRS]-series discrete-transistor flip-flops each
had an extra output, decoupled by a resistor, for the lamp driver, and
the output had to be used. One could get a very good idea of the state
of a stopped system by studying the lights.
This rule went out with the IC modules. It's interesting to speculate
on the implications of such a rule on, say, an UltraSPARC. What would
the insertion force be?
Phil Gustafson
--
))
(( Phil Gustafson Urban Legends FAQ: http://www.urbanlegends.com
C|~~| Java FAQ: http://www.afu.com
`--' <ph...@panix.com>
Ric Werme
>Memory is getting fuzzy, but I think on the KAs there was a group of six
>lights on the left side of console:
> 1 4
> 2 5
> 3 6
>Lights 1-3 were usually on, 5 was user mode, I think 4 was "Program stop"
>and lit when halt executed.
Ah - I didn't lend Megan my original phone book manual, otherwise known
as the "pdp10 reference handbook" (no caps).
The lights are
1) Run
2) PI on (i.e. interrupts enabled)
3) Power (utterly useless)
4) Program stop
5) User mode
6) Memory stop
The last was triggered by referencing the location set in the address switches
when the address stop switch was set.
To the right of that are three rows of seven lamps:
PI in progress (where PI7 glow showed up)
PI request
PI active (which interrupts are enabled)
Then theres a 4th set to the right of the PI in progress lights for IOB PI
request.
otto714 otto714
> On the KA-10, this problem was simplified by a design rule: one light
> per flip-flop. The [BRS]-series discrete-transistor flip-flops each
> had an extra output, decoupled by a resistor, for the lamp driver, and
> the output had to be used. One could get a very good idea of the state
> of a stopped system by studying the lights.
Greetings, I am impressed by 10 folk. Now shall we move on to pdp 8 land, or
pdp 1140 land (the 1135 had LEDs I think). Seriously yes with the lites you
could monitor system activity by noticing how "things" looked with feet up
and coffee in hand.
otto