Re: [RFC 06/52] hw/cpu: Introduce hybrid CPU topology

[Zhao Liu]

On Tue, Feb 14, 2023 at 10:27:53AM +0100, Philippe Mathieu-Daudé wrote:
> Date: Tue, 14 Feb 2023 10:27:53 +0100
> From: Philippe Mathieu-Daudé <philmd@linaro.org>
> Subject: Re: [RFC 06/52] hw/cpu: Introduce hybrid CPU topology
> 
> On 14/2/23 10:30, Zhao Liu wrote:
> > On Mon, Feb 13, 2023 at 02:10:17PM +0100, Philippe Mathieu-Daudé wrote:
> > > Date: Mon, 13 Feb 2023 14:10:17 +0100
> > > From: Philippe Mathieu-Daudé <philmd@linaro.org>
> > > Subject: Re: [RFC 06/52] hw/cpu: Introduce hybrid CPU topology
> > > 
> > > On 13/2/23 10:49, Zhao Liu wrote:
> > > > From: Zhao Liu <zhao1.liu@intel.com>
> > > > 
> > > > For smp systems, the parts in one topology level are the same. But now
> > > > there are more and more systems with hybrid architectures. Different
> > > > parts of the same topology level may have differences. For example,
> > > > Intel's Alder Lake series CPU has two types of cores, so the CPU
> > > > topology is no longer symmetrical.
> > > > 
> > > > The hybrid topology is compatible with the smp topology type, that is,
> > > > different parts on the same level of the hybrid topology can set to be
> > > > the same, but the hybrid topology will introduce more complexity (need
> > > > to allocate more memory, organized with array or linked-list), so the
> > > > original smp topology support is retained while introducing the hybrid
> > > > topology, and the hybrid topology is only built when the hybrid is
> > > > explicitly required.
> > > > 
> > > > Therefore, we introduce the definition support of hybrid cpu topology
> > > > here. At the same time, in order to unify with the original smp, we
> > > > introduce a new cpu topology structure that can support smp topology
> > > > or hybrid topology. This structure will replace the CpuTopology type (in
> > > > include/hw/boards.h) used by MachineState.smp.
> > > > 
> > > > As for now, we only support two hybrid topology levels: core and
> > > > cluster.
> > > > 
> > > > Signed-off-by: Zhao Liu <zhao1.liu@intel.com>
> > > > ---
> > > >    MAINTAINERS                   |   1 +
> > > >    include/hw/cpu/cpu-topology.h | 117 
> > > > ++++++++++++++++++++++++++++++++++
> > > >    qapi/machine.json             |  12 ++++
> > > >    3 files changed, 130 insertions(+)
> > > >    create mode 100644 include/hw/cpu/cpu-topology.h
> > > > 
> > > > diff --git a/MAINTAINERS b/MAINTAINERS
> > > > index 58794885ced3..918a9418d98e 100644
> > > > --- a/MAINTAINERS
> > > > +++ b/MAINTAINERS
> > > > @@ -1742,6 +1742,7 @@ F: qapi/machine-target.json
> > > >    F: include/hw/boards.h
> > > >    F: include/hw/core/cpu.h
> > > >    F: include/hw/cpu/cluster.h
> > > > +F: include/hw/cpu/cpu-topology.h
> > > >    F: include/sysemu/numa.h
> > > >    F: tests/unit/test-smp-parse.c
> > > >    T: git https://gitlab.com/ehabkost/qemu.git machine-next
> > > 
> > > Worth splitting this section in 2, machine VS numa/clusters?
> > 
> > Do you mean splitting this header file into numa.h and cluster.h?
> 
> No, I meant the MAINTAINERS 'Machine' section.

I see. I will slit this.

> 
> > It seems that the cpu topology is not related to the content of
> > numa.h,
> 
> And you seem to agree ;)
> 
> > and it may be possible to merge some content into cluster.h,
> > but the definition in cluster.h is to use the cluster as a device
> > (if I understand correctly...) , and cpu_topology.h is just about
> > topology-related things.
> > 
> > I have not included the contents of cluster.h/cluster.c into hybrid
> > considerations, but only modified the topology defined by smp.
> > 
> > Zhao
>