[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-devel] [PATCH v8 31/50] x86emul: support remaining misc AVX512{F, BW} insns



>>> On 18.06.19 at 18:42, <andrew.cooper3@xxxxxxxxxx> wrote:
> On 15/03/2019 10:58, Jan Beulich wrote:
>> This completes support of AVX512BW in the insn emulator, and leaves just
>> the scatter/gather ones open in the AVX512F set.
>>
>> Signed-off-by: Jan Beulich <jbeulich@xxxxxxxx>
>> ---
>> v5: New.
>> ---
>> TBD: The *blendm* inline functions don't reliably produce the intended
>>      insns, as the respective moves are about as good a fit for the
>>      compiler when looking for a match for the intended operation. We'd
>>      need to switch to inline assembly if we wanted to guarantee the
>>      testing of those insns. Thoughts?
> 
> Do you mean __builtin_ia32_blendm_* here?

Yes, or their wrappers in the *intrin.h headers (which we don't use for
various reasons).

>  I guess it depends on the
> writemask as to whether mov instructions would be more efficient.

Right, but we don't use write masks here. Their use in general is imo
rather cumbersome in C level code anyway, and it wouldn't fit the
general structure of simd_test() very well if we wanted to make use
of them.

> I think we can probably rely on fuzzing to cover the difference, seeing
> as there is nothing interesting or nonstandard about these
> instructions.  That certainly seems preferable to a custom bode.
> 
> Acked-by: Andrew Cooper <andrew.cooper3@xxxxxxxxxx>

Thanks.

Jan



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

 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.