[patch 3/8] powerpc: cpu_idle implementation for systemsim

Arnd Bergmann arndb at de.ibm.com
Sun Jan 15 23:44:11 EST 2006


On Saturday 14 January 2006 03:01, Olof Johansson wrote:
> > +config SYSTEMSIM_IDLE
> > +     bool "   Optimized idle loop for systemsim"
> > +     depends on PPC_SYSTEMSIM
> > +     help
> > +       Selecting this option will enable an more optimized idle loop
> > +       for running on the IBM Full System Simulator that
> > +       significantly reduces the load on the host system when
> > +       simulating an idle system.
> 
> Do you really need this configured separately from the rest of
> systemsim?

Maybe not. The idea here is that if you want cycle-accurate simulation
of the system, then you might want to use the regular idle loop
instead of the faster one, but you might still rely on one of the
device drivers.

> > +/*
> >   * Called into from start_kernel, after lock_kernel has been called.
> >   * Initializes bootmem, which is unsed to manage page allocation until
> >   * mem_init is called.
> > @@ -600,6 +617,8 @@ void __init setup_arch(char **cmdline_p)
> >  
> >       ppc_md.setup_arch();
> >  
> > +     setup_systemsim_idle();
> > +
> 
> Idle loops are generally probed and setup in the platform-specific inits
> (setup_arch). Can't you do this from there too on the affected
> platforms, it's only two after all. You can move the
> setup_systemsim_idle() function to idle_systemsim.c to share it.

Good idea, the setup_systemsim_idle() function should definitely
be moved to the other file.

I don't know really how many of the platforms in the current kernel
also have a systemsim port (there are only two publically available
ones, but a lot more for other systems that still have a closed user
base).

It's certainly cleaner to move this into the platforms, but it might
be helpful if any new systemsim target 'just works' with this.

	Arnd <><




More information about the Linuxppc64-dev mailing list