19 Mar
2010
19 Mar
'10
11:48 a.m.
On 18 Mar 2010 at 19:33, Tony Duell wrote:
I am talking about microcontrollers, not desktop computer CPUs...
If it's genuinely the case that the exact behavious or the device is that
hard to predict, it makes them useless for just about any application
that I would want to use them for, which generally depend on changing
external signals with precisely known timing.
And a logic analyser will tell you different
things to a breakpoint.
OK, at a breakpoint (if the debugging hardware is designed sensibly),
you can exame CPU registers, etc. But it woun't necessarily tell you
how the code got to the breakpoint. A logic analyser will/
Hardly! Maybe on vintage CPUs, but not on anything made in the last
20 years. Caching, prefetch queues, scheduling hardware and
integrated peripherals make a logic analyzer of marginal use in
debugging program flow.
Times have changed and so have methods. For example, simulation is
You say that as though it's a Good Thing. Personally, I can't think of
any change in the last 20 years that's actually made life better for me.
used a lot more than it used to be for designs, as is
on-chip
debugging support.
I've had too many bad experiences of simulators to trust them. I've yet
to meet a digital logic simulatoor that gets my test circuits right...
And it's very hard to simulate a microntroller when you're using it to
interface to some other large and complex system which you somehow have
to include in the simulator.
On-chip debugging support would be great if it was documented. It appears
not to be, at least not for PiCs.
-tony