Re: [Qemu-devel] [PATCH] s390x/cpumodels: add z14 Model ZR1

[David Hildenbrand]

On 13.06.2018 11:42, Christian Borntraeger wrote:
> 
> 
> On 06/13/2018 11:00 AM, David Hildenbrand wrote:
>> On 13.06.2018 10:18, Christian Borntraeger wrote:
>>> introduce the new z14 Model ZR1 cpu model. Mostly identical to z14, only
>>> the cpu type differs (3906 vs. 3907)
>>>
>>> Signed-off-by: Christian Borntraeger <address@hidden>
>>> ---
>>>  target/s390x/cpu_models.c | 1 +
>>>  1 file changed, 1 insertion(+)
>>>
>>> diff --git a/target/s390x/cpu_models.c b/target/s390x/cpu_models.c
>>> index e10035aaa8..cfdbccf46d 100644
>>> --- a/target/s390x/cpu_models.c
>>> +++ b/target/s390x/cpu_models.c
>>> @@ -79,6 +79,7 @@ static S390CPUDef s390_cpu_defs[] = {
>>>      CPUDEF_INIT(0x2964, 13, 2, 47, 0x08000000U, "z13.2", "IBM z13 GA2"),
>>>      CPUDEF_INIT(0x2965, 13, 2, 47, 0x08000000U, "z13s", "IBM z13s GA1"),
>>>      CPUDEF_INIT(0x3906, 14, 1, 47, 0x08000000U, "z14", "IBM z14 GA1"),
>>> +    CPUDEF_INIT(0x3907, 14, 1, 47, 0x08000000U, "z14ZR1", "IBM z14 Model 
>>> ZR1 GA1"),
>>>  };
>>>  
>>>  #define QEMU_MAX_CPU_TYPE 0x2827
>>>
>>
>> This is the first time that we have two different EC variants (different
>> CPU type) with the same HW gen and GA level .
>>
>> Are they really completely equal (esp. the IBC value? that is used for
>> model detection) apart from the CPU type?
>>
>> check_unavailable_features()/arch_query_cpu_model_comparison() will
>> treat both models as equal (meaning one can run on the other).
>>
>> arch_query_cpu_model_baseline() might produce a z14ZR1-base when
>> baselining e.g. a z14 and a z14ZR1 (don't think this is a problem), as
>> s390_find_cpu_def() will always try to walk as far as possible in the
>> cpu model definition table.
>>
> Yes, we can run z14ZR1 on Z14 and vice versa.
> 
> 
> s390_find_cpu_def has this
> 
>         /* stop the search if we found the exact model */
>         if (def->type == type && def->ec_ga == ec_ga) {
>             return def;
>         }
> 
> So we should always get the best fit, no?

Yes, but one could "think" that baselining a z14 against a z14RC1 would
result in the "older" (z14) machine, while we will return the z14RC1. As
 said, I don't think this is a problem, just wanted to point it out!

-- 

Thanks,

David / dhildenb