Greg Kurz <gr...@kaod.org> writes:
> On Fri, 02 Sep 2016 15:04:47 +0530
> Nikunj A Dadhania <nik...@linux.vnet.ibm.com> wrote:
>
>> Greg Kurz <gr...@kaod.org> writes:
>>
>> > On Fri, 2 Sep 2016 12:02:54 +0530
>> > Nikunj A Dadhania <nik...@linux.vnet.ibm.com> wrote:
>> >
>> >> Signed-off-by: Nikunj A Dadhania <nik...@linux.vnet.ibm.com>
>> >> ---
>> >
>> > Shouldn't this patch be the last one, when all other issues have been
>> > addressed ?
>> >
>> >> target-ppc/kvm.c | 2 +-
>> >> target-ppc/kvm_ppc.h | 2 +-
>> >> 2 files changed, 2 insertions(+), 2 deletions(-)
>> >>
>> >> diff --git a/target-ppc/kvm.c b/target-ppc/kvm.c
>> >> index dcb68b9..20eb450 100644
>> >> --- a/target-ppc/kvm.c
>> >> +++ b/target-ppc/kvm.c
>> >> @@ -2090,7 +2090,7 @@ void kvmppc_set_mpic_proxy(PowerPCCPU *cpu, int
>> >> mpic_proxy)
>> >>
>> >> int kvmppc_smt_threads(void)
>> >> {
>> >> - return cap_ppc_smt ? cap_ppc_smt : 1;
>> >> + return cap_ppc_smt ? cap_ppc_smt : 8;
>> >
>> > If KVM is there but does not support SMT processor modes, it looks
>> > wrong to return anything but 1. This check needs kvm_enabled().
>>
>> This also gets called when emulating PPC on PPC.
>>
>
> Yes and the current value of 1 is the default for non HV KVM and TCG. If
> you want to change the default for MTTCG, then you need separate paths.
>
>> > Also, why 8 ? This depends on the CPU model.
>>
>> Not sure if I need to emulate according to the host cpu model. I had
>> selected 8, as that was the highest number of threads possible for POWER.
>>
>
> If running in full emulation and cpu type is POWER7, this shouldn't be
> higher than 4.
>
> In the end, something like:
>
> int kvmppc_smt_threads(void)
> {
> if (kvm_enabled()) {
> return cap_ppc_smt ? cap_ppc_smt : 1;
> } else {
> return 8_or_4_or_2_or_1_depending_on_the_cpu_model;
> }
> }
Sure, will need something like this.
Regards,
Nikunj
