GDB-Protocol: Cycle step command?

Andrew Cagney ac131313@cygnus.com
Thu Jun 17 16:28:00 GMT 1999


"J.T. Conklin" wrote:
> 
> >>>>> "Andrew" == Andrew Cagney <ac131313@cygnus.com> writes:
> Andrew> Hello, I'd like to propose that the GDB remote protocol be
> Andrew> extended to support a cycle step command.  This command would
> Andrew> be used when either interacting with physical control devices
> Andrew> such as JTAG or (more often) cycle accurate simulators.
> Andrew>
> Andrew> Possible syntaxes include:
> Andrew>     QCycle=<NR>
> Andrew> and i<NR>
> Andrew>
> Andrew> comments suggestions additional data people thing may be
> Andrew> needed.  Unlike the ``S'' and ``C'' commands I've not included
> Andrew> an address.  Changing the address in a cycle accurate model
> Andrew> can get very complex.
> 
> This seems incomplete.
> 
> If you have a command to step through processor cycles, don't you also
> need cooresponding commands to extract processor state (pipeline, tlb,
> cache, btb, etc.).  Without such commands, I fail to see the value in
> being able to step in this manner.

I think that, while related, the two problems should be considered
separatly.

In the case of a cycle accurate simulator I'd expect the information you
mention to be displayed by interacting with the simulator directly
(possibly using a remote script command).  Most simulators of this type
already have a command line interface.

As a separate issue (independant of cycle stepping) GDB does need to
become better at displaying processor internal state (TPB registers,
....).

	enjoy,
		Andrew

PS: It can't be ``QC...''.


More information about the Gdb mailing list