| T.R | Title | User | Personal
Name | Date | Lines |
|---|
| 63.1 | Halt is Good Enough | DSSDEV::SAUTER | John Sauter | Fri Jul 11 1986 13:36 | 18 |
| I think reporting the error back to the processor would be overkill.
The Amiga is a home computer; it is not intended to be used to drive
a commercial teleprocessing application, or a military command-and-
control application. Under home computer conditions I think it
would be sufficient for the memory to hang (thereby hanging the
processor) and turn on an error LED that you could see by disassembling
the system. To clear the error condition, power cycle the machine.
This gives you the minimum (and therefore lowest cost) error reporting
hardware, and doesn't require that you trust software that will
only be executed once in a blue moon (when there is an error).
If you want more reliability, and not just error checking, use ECC.
A soft ECC error returns correct data to the processor and lights
a yellow LED. A hard ECC error hangs the memory and lights a red
LED. Occasionally check the yellow LED to see if your memory needs
replacing.
John Sauter
|
| 63.2 | I know - best way to halt though? | APOLLO::BERKSON | | Fri Jul 11 1986 14:25 | 12 |
|
The LEDs seem like a good cheap way to do it. I didn't intend to
make a mil spec fail operational memory. I am just looking for the
easiest way to take the next step up from no notification at all
of a memory error. I thought that there might have been some software
that the 68000 already had which it could branch to if the memory
signalled an error rather than proceeding along as if nothing had
happened. Do most home computers not even have parity or just the
Amiga? I would rather have the system hang than have it make an
error and not know about it.
Mitch
|
| 63.3 | Why halt? | ALIBUT::SANTIAGO | Ed Santiago | Fri Jul 11 1986 14:36 | 17 |
| <FLAME ON>
Why hang the machine? One of the (many, many) things I despise
about the I*M PC is that any memory parity error will instantly
kill the machine. This may be fine if you're an ultraprofessional
that needs 100% reliability, but if you're like most people, and
say maybe have spent a good hour working on something without
saving a copy (not wise, but everybody does it occassionally),
and a small glitch appears on the bus making the memory controller
think it's got bad RAM, BOOM you have just become one very angry
person. I wouldn't mind having a requester appear to warn me or
something, if I could continue from it (not like the ordinary
variety which says "Finish ALL activity", pretty useless since
you can't do anything because the machine goes guru as soon
as you click CONTINUE). It would be nice to maybe just dump
everything to disk before powering down.
<FLAME OFF>
|
| 63.4 | | BAGELS::BRANNON | Dave Brannon | Fri Jul 11 1986 20:38 | 11 |
| Most home computers i've heard of have no memory integrity checks.
Maybe just a diagnostic program to test memory if you thinks something
is wrong it.
Halting the machine is a very, very nasty way to notify the user
that there is a problem with his memory. Reporting the error to
the user, lighting the LED, and continuing to work as normally as
possible is a much better way. A good compromise would be to
have it come up with a fatal error on the next powerup selftest.
-Dave
|
| 63.5 | Degrees of Nastyness | DSSDEV::SAUTER | John Sauter | Sat Jul 12 1986 10:36 | 4 |
| Halting the machine is nasty, but corrupting the user's data without
telling him is even nastier. If you want to be able to continue
after a memory error, invest in ECC.
John Sauter
|
| 63.6 | Fuel for the fire | ELWOOD::PETERS | | Sat Jul 12 1986 12:02 | 15 |
|
FYI, a quick survey of home computers memory protection.
TRS80 no protection ---------
Apple II no protection ---------
Rainbow parity Halt system, Error message
IBM PC parity Halt system, Error message
I have worked many hours on a Rainbow and have never had a memory
error. Therefore I feel the the protection is nice and halting the
system with some kind of indicator ( Led or message ) is good enough.
Steve Peters
|
| 63.7 | Is It a Real Problem? | ERLANG::FEHSKENS | | Mon Jul 14 1986 20:20 | 12 |
| There was a somewhat hysterical letter in a recent issue of some
personal computing magazine lambasting the Amiga for not having
ECC/parity memory. The letter gave me the impression that the Amiga was
unusual in this regard. How about Macs? ATs/XTs? I use my Amiga
a lot and haven't noticed any corrupted data. I'd expect a crash
from a bogus address reference or some such error to be as likely
as undetected corrupted data. I think this whole issue is a tempest
in a teapot. Anybody got any real data on expected error rates
for the Amiga's memory?
len.
|
| 63.8 | I'd like to know... | ALPHA::KOPP | Lowell Kopp | Tue Jul 15 1986 19:04 | 13 |
| I'd feel a lot more confortable it there were parity on the memory..
There may not be a problem with the base level system or minimal
expansion (to 512k, 1 external drive) but what happens when you
add more peripherals or memory (well designed or not), or otherwise
load the system? A lot of people discovered the I*M PC's power supply
limitations when they loaded up the expansion slots and discovered
that they were suddenly getting parity errors on memory. I'd like
to know when my system is operating under marginal conditions, and
at least get a chance to do something about it, rather than wonder
was it software or hardware that made the system crash this time..
-- Lowell
|