Re: [Xen-devel] [PATCH v8 38/50] x86emul: support of AVX512* population count insns

["Jan Beulich" <JBeulich@xxxxxxxx>]

>>> On 19.06.19 at 14:22, <andrew.cooper3@xxxxxxxxxx> wrote:
> On 15/03/2019 11:01, Jan Beulich wrote:
>> --- a/xen/tools/gen-cpuid.py
>> +++ b/xen/tools/gen-cpuid.py
>> @@ -269,7 +269,7 @@ def crunch_numbers(state):
>>          # AVX512 extensions acting (solely) on vectors of bytes/words are 
>> made
>>          # dependents of AVX512BW (as to requiring wider than 16-bit mask
>>          # registers), despite the SDM not formally making this connection.
>> -        AVX512BW: [AVX512_VBMI, AVX512_VBMI2],
>> +        AVX512BW: [AVX512_VBMI, AVX512_BITALG, AVX512_VBMI2],
> 
> This ordering looks suspect.  BITALG should go last, given its position
> in the feature leaf.

After re-basing this now is

@@ -268,7 +268,7 @@ def crunch_numbers(state):
         # AVX512 extensions acting (solely) on vectors of bytes/words are made
         # dependents of AVX512BW (as to requiring wider than 16-bit mask
         # registers), despite the SDM not formally making this connection.
-        AVX512BW: [AVX512_BF16, AVX512_VBMI, AVX512_VBMI2],
+        AVX512BW: [AVX512_BF16, AVX512_BITALG, AVX512_VBMI, AVX512_VBMI2],
 
         # The features:
         #   * Single Thread Indirect Branch Predictors

I don't think ordering based on (potentially unrelated) leaves should
be a criteria here. Instead, as indicated before, I think we'd be
better off using alphabetical sorting for such longer dependency
lists. I'd be happy to re-sort the AVX512F one as well.

> With this fixed, Acked-by: Andrew Cooper <andrew.cooper3@xxxxxxxxxx>

I'm not going to apply this without further clarification by you as
per the above point.

Jan



_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/mailman/listinfo/xen-devel