Re: [edk2-devel] [edk2-platforms][PATCH 1/1] AmpereAltraPkg: Update ArmPlatformLib to work with changed ARM_CORE_INFO

["Rebecca Cran" <quic_rcran@quicinc.com>]

I was under the impression that this is becoming a more standard format?

For example, the Neoverse N2 has AFF0 always 0 (it's not hyperthreaded), 
it puts the core ID in AFF1, the cluster ID in AFF2, but since it only 
has a single socket AFF3 is always 0. This differs from older cores 
where the core ID was in AFF0 and the cluster ID in AFF1.

https://developer.arm.com/documentation/102099/0000/AArch64-registers/AArch64-identification-registers/MPIDR-EL1--Multiprocessor-Affinity-Register

Am I mistaken?

-- 
Rebecca Cran

On 1/17/23 02:53, Nhi Pham wrote:
> Hi Rebecca,
> 
> That's Ampere Altra Family specific MPIDR encoding. So, we could not 
> leverage the definitions in the ArmPkg/Include/Library/ArmLib.h.
> 
> -Nhi
> 
> On 1/13/2023 9:40 PM, Rebecca Cran wrote:
>> On 1/12/23 21:21, Nhi Pham via groups.io wrote:
>>
>>> +//
>>> +// MPIDR manipulation
>>> +//
>>> +#define AC01_GET_MPIDR(SocketId, ClusterId, CoreId) \
>>> +          (((SocketId) << 32) | ((ClusterId) << 16) | ((CoreId) << 8))
>>> +#define AC01_GET_SOCKET_ID(Mpidr)  (((Mpidr) & ARM_CORE_AFF3) >> 32)
>>> +#define AC01_GET_CLUSTER_ID(Mpidr) (((Mpidr) & ARM_CORE_AFF2) >> 16)
>>> +#define AC01_GET_CORE_ID(Mpidr)    (((Mpidr) & ARM_CORE_AFF1) >> 8)
>>> +
>>
>> Ideally, this should go in ArmPkg/Include/Library/ArmLib.h, but I'm 
>> not sure how we should handle the older format where the the core was 
>> in the first 8 bits.
>>