31 Jul
2013
31 Jul
'13
7:28 p.m.
On Wed, 31 Jul 2013, Chuck Guzis wrote:
...and parity does not cause a machine to "lock
up".
True. It just triggered the code to do so.
What it does do is
trigger a non-maskable interrupt. The default action for that vector on
the 5150 is to display a message and halt--but that does not preclude
the user from supplying his own handler that returns control to the
interrupted code.
Yes.
I used to use NMIEXIT.COM
That let you load some code into memory, such as DEBUG.COM
When the NMI happened (or if you shorted A1 to B1, it would jump to
something that you could use as a monitor, and try to recover whatever was
left in RAM, and later try to repair the corruption.
I'd rather know about a memory error than discover
that my program was
producing garbage.
ABSOLUTELY!
However, I would have liked the option of a RAM dump, since at least with
text editing, a single bit bad would be a lot better than total loss of
everything since the last SAVE.
BUT, the general public would much rather have a machine with
untrustworthy results than have an interruption.
Seymour Cray is often credited with saying
"Parity is for farmers" when
the 6000-series machines lacked parity checking on memory. It's also
worth noting that his later machines not only had parity but also SECDED.
Error correction was much more needed than error detection with no
recovery option.
--
Grumpy Ol' Fred cisin at xenosoft.com