Odd Z80 failure modes?

[ChildOfCv]

While writing processor tests for my diag board, I was wondering, how much of the chip needs to be tested to clear it of suspicion?  Have you encountered a case where some functions worked but others didn't?  Or was it always a working or not kind of thing?

 

I am currently testing the address lines, the NMI, INT (with or without interrupts enabled), and HALT modes.  The data lines are necessarily tested as part of the process, since bad data lines will give incorrect execution.

 

Beginning CPU test
Testing address lines
JR 0002
JR 0004
JR 0008
JR 0010
JR 0020
JR 0040
JR 0080
JR 0100
JP 0200
JP 0400
JP 0800
JP 1000
JP 2000
JP 4000
JP 8000
Testing non-maskable interrupts
LD SP, 1234
Testing HALT mode
HALT
<-halt
->int
<-halt
->nmi
NOP (CPU must ignore this instruction)
NOP (CPU must ignore this instruction)
RETN
Testing maskable interrupts
IM 1
EI
HALT
<-halt
->int
NOP (CPU must ignore this instruction)
Interrupt acknowledge
RETI
All CPU tests passed

 

[Keatah]

Speaking from a non-developer's or layman's point of view..

 

I remember my Apple II diagnostic programs. The documentation and/or discussion with the publisher mentioned they do the address and data lines, interrupts, and a good number of instructions if not all of them. There was also focus on whether the chip could do math correctly and that took time. Not something to do during startup.

 

It would also make sense to do as much as possible. Never know what corner case might arise.

 

Current Intel CPU diagnostics spend like 10-15 minutes going through tons of stuff.