11 Oct
2022
11 Oct
'22
3:54 p.m.
On 10/11/22 09:11, Paul Koning wrote:
Z errors may well be acceptable. The thing to do is
examine the code to see what it wants to happen, then check the model differences appendix
B of the PDP11 architecture handbook to see if your target machine does what is wanted.
The primary source of Z errors I remember is the use of the same register in an
autoinc/autodec mode and as a direct source operand; in some machines, both operand
addresses are determined first and then the source register is fetched; in others, the
source is handled, then the destination. See page B-2 of the architecture handbook.
Of course another option would be to replace the offending instruction by 2-instruction
sequences that achieve the intended result without Z errors.
Yes, it's the autincrement/decrement when the source/destination
operands are the same; that's covered in the earlier docs. They removed
most of that description in the later manuals, which was annoying.
I am under a looming deadline to get this running, and it's already
quite a busy few days here with hurricane recovery on top of regular
work, so I really don't have time to hack on it. What I'm trying to
build is a supported configuration according to the APL-11
documentation. It just doesn't seem to work, so I just figured I'd ask
here if anyone had done it.
Late last night I started to wonder if the Z errors were actually
supposed to have been warnings, as it did generate object files. So I
moved forward to the next step in the build, and it generated a task
file...which doesn't run. It exits immediately, no messages of any
kind. Frustrating.
-Dave
--
Dave McGuire, AK4HZ
New Kensington, PA