On Wed Dec 23 14:21 , Klaus Rudolph  sent:

>[email protected] schrieb:
>
>> An "assert" that depends on the values of registers in more than one cpu.
>Nice feature! Looks like watchpoints, but with multiple cpu's.
>
>BTW: Why gdb is not supporting watchpoints for avr? Is there anybody out
>there who can enable this feature in gdb. To add the feature to the
>simulator is not a problem.
>
>> An assert that takes zero time to calculate.
>Breakpoints actually need no "simulated time" and if watchpoints will be

When I'm trying to debug, I usually end up either
with breakpoints that break way too often or with
breakpoints that require auxiliary variables.
The latter is fine if I'm debugging a mathematical algorithm.
Not so fine if I'm debugging something that depends on cycle-correct timing.
YMV.

>supported, they also need no time. Connecting multiple watchpoints is a
>feature which could not directly controlled by gdb I think.
>
>Do you want to write some scripts for creating this asserts?
>
>Like:
>
>assert ((dev1->R[1]== 0x14) && ( dev2->PC > 0x240 && dev2->PC 
>"Condition 1 reached");

That is the kind of thing I had in mind.
I'd always assumed that to do it I would have to write plug-ins
that would be executed by the simulator between cycles.
I picked simulavrxx because I deemed it the AVR simulator I
was most likely to be able to modify to my taste.

>Maybe it is possible to use multiple cpus like multi threaded code in
>one gdb session... a bit tricky but not impossible.
>
>But at first I think watchpoints should be enabled in gdb. If available
>it could be added to the gdbserver interface and the breakpoints.

---
Michael Hennebry
[email protected]
"War is only a hobby."
---- Msg sent via CableONE.net MyMail - http://www.cableone.net

_______________________________________________
Simulavr-devel mailing list
[email protected]
http://lists.nongnu.org/mailman/listinfo/simulavr-devel

Reply via email to