Re: [PATCH 09/10] Support Intel AMX-FP16

[Jan Beulich <jbeulich@suse.com>]

On 18.10.2022 11:23, Jan Beulich wrote:
> On 18.10.2022 11:01, Cui, Lili wrote:
>>>> +  { "CPU_AMX_FP16_FLAGS",
>>>> +    "CpuAMX_FP16" },
>>>>    { "CPU_IAMCU_FLAGS",
>>>>      "Cpu186|Cpu286|Cpu386|Cpu486|Cpu586|CpuIAMCU" },
>>>>    { "CPU_ADX_FLAGS",
>>>
>>> Can you please insert next to the other similar AMX entries? Seeing the flaw
>>> here, I'll be making a patch to address the lack of CPU_AMX_TILE_FLAGS in
>>> the similar pre-existing entries. When you move the insertion, it'll be easier
>>> to keep things in sync.
>> Done.
>>  
>>>> -    "CpuAMX_TILE|CpuAMX_INT8|CpuAMX_BF16" },
>>>> +    "CpuAMX_TILE|CpuAMX_INT8|CpuAMX_BF16|CpuAMX_FP16" },
>>>>    { "CPU_ANY_MSRLIST_FLAGS",
>>>>      "CpuMSRLIST" },
>>>> +  { "CPU_ANY_AMX_FP16_FLAGS",
>>>> +    "CpuAMX_FP16" },
>>>>  };
>>>
>>> Same here then.
>> Done.
> 
> I guess my comment here was a little misleading (I'm sorry for that), in
> that in addition I was expecting you to consider the comment regarding
> the need for the *_ANY_* constants that I did give for several of the
> patches in this series. I think the question applies here as well: Are
> there dependent features known to appear? If not, there's no need for the
> extra constant.

I have to apologize and take back the _ANY_-related comment here (it
applies in the other places I've explicitly given it for this series).
It was only yesterday (in the course of automating the dependency
recording as much as possible) that I realized that there's a two way
requirement for re-using the non-ANY form in the initializer of the
3rd field of cpu_flags[] entries: There must not be dependents _and_
there must also not be dependencies. Otherwise disabling a feature
(via command line option or directive) would also disable the
feature's prereq ones. With my work to (mostly) automate this, I
guess there's no strong need to patch this separately - there are
many more issues that will need taking care of once the base logic is
in place.

Jan