simh simulation speed

From: John A. Dundas III <dundas_at_caltech.edu>
Date: Mon Sep 27 18:34:10 2004

At 3:00 PM -0700 9/27/04, Tom Jennings wrote:
>On Mon, 2004-09-27 at 14:27, John A. Dundas III wrote:
>
>> However this does cause problems for some I/O devices. In some
>> cases, extra code has gone into working through situations in which
>> an OS makes implicit assumptions about how quickly an I/O operation
>> will complete. Sometimes SIMH executes these too quickly and much
>> fun ensues.
>
>Good point!
>
>The LGP-30 doesn't even have the hardware concept of "IO ready" for the
>console flexowriter, relying entirely on drum software-delay routines to
>avoid overrunning it, so I'll have to go look at how simh handles that.

It's complicated but not fatally so. At one time I coded up a model
for Bob of how this would work for a variety of disks on the PDP-11.
The model took into account seek time (from current location),
rotational latency, and data transfer latency. It was tested on the
RL and RP/RM disks and seemed to work quite well. That is,
operations took as long (!) as they would have with the real drives.
This was about the time Bob was working on an RSX problem where disk
I/Os completed too quickly under certain circumstances. The code is
not incorporated into the release version of SIMH, but Bob did
generate a paper from some of the explorations.

Bottom line: don't be put off by SIMH if you think that might be the
way to go. It can be done.

John
Received on Mon Sep 27 2004 - 18:34:10 BST