Re: [RFC 09/52] hw/machine: Introduce core type for hybrid topology

[Zhao Liu]

On Mon, Feb 13, 2023 at 02:13:40PM +0100, Philippe Mathieu-Daudé wrote:
> Date: Mon, 13 Feb 2023 14:13:40 +0100
> From: Philippe Mathieu-Daudé <philmd@linaro.org>
> Subject: Re: [RFC 09/52] hw/machine: Introduce core type for hybrid topology
> 
> On 13/2/23 10:49, Zhao Liu wrote:
> > From: Zhao Liu <zhao1.liu@intel.com>
> > 
> > Under the hybrid cpu topology, some CPUs care about the core type.
> > For example, Intel's Alder Lake series CPU contains two types of cores:
> > Intel Core and Intel Atom. The type information of these two types is
> > exposed in 1A leaf of CPUID.
> > 
> > Core type should also be part of the hybrid topology, and
> > MachineState.cpu_type cannot provide different type information for
> > different cpus in the same machine, so add a type field for the core
> > level in the hybrid cpu topology.
> > 
> > Additionally, add a helper to get core type information from
> > MachineState. Though core_type is only used in hybrid case, don't
> > use assert since it may be used to initialize some generic fields.
> > 
> > Signed-off-by: Zhao Liu <zhao1.liu@intel.com>
> > ---
> >   hw/core/machine-topo.c        | 12 ++++++++++++
> >   include/hw/boards.h           |  3 +++
> >   include/hw/cpu/cpu-topology.h |  2 ++
> >   3 files changed, 17 insertions(+)
> > 
> > diff --git a/hw/core/machine-topo.c b/hw/core/machine-topo.c
> > index b20160479629..e0ec07b53d42 100644
> > --- a/hw/core/machine-topo.c
> > +++ b/hw/core/machine-topo.c
> > @@ -51,6 +51,18 @@ unsigned int machine_topo_get_smp_threads(const 
> > MachineState *ms)
> >       return ms->topo.smp.threads;
> >   }
> > +unsigned int machine_topo_get_hybrid_core_type(const MachineState *ms,
> > +                                               unsigned int cluster_id,
> > +                                               unsigned int core_id)
> > +{
> > +    if (!machine_topo_is_smp(ms)) {
> > +        return ms->topo.hybrid.cluster_list[cluster_id]
> > +                       .core_list[core_id].core_type;
> > +    } else {
> > +        return 0;
> 
> Is '0' an invalid type?

My intention is to use 0 as the default type under smp (it can be
regarded as invalid under hybrid, this should depend on the check
of the architecture itself), I'll use a macro instead of hardcode.

> 
> > +    }
> > +}
>