[PATCH 0/3] Refine AVX10.2 mnemonics

[Jan Beulich]

On 13.01.2025 09:26, Haochen Jiang wrote:
> Hi all,
> 
> Since AVX10.2 got published, there are several discussions regarding to
> the mnemonics. After internal discussion, we will make three changes for
> AVX10.2 mnemonics.
> 
> Quick Conclusion:
> 
>   - NE would be removed for all AVX10.2 new insns
>   - VCOMSBF16 -> VCOMISBF16
>   - P for packed omitted for AI data types
> 
> The details of why doing so will be embedding at the end of this email.
> 
> All of them will get refined in AVX10.2 SPEC (ETA published this week).
> Since Binutils 2.44 window is near, I choose to send them out.
> 
> Upcoming will be the three patches to change the mnemonics according to
> the conclusion, where patch 1 on BF16 arithmetic insns, patch 2 on
> VCOMISBF16 and patch 3 on convert insns.
> 
> These three patches would probably be the last unsent patches for Intel
> Diamond Rapids in Binutils 2.44 if no exception happened (The only
> potential exception would be MOVRS APX_F EVEX.W issue). Really appreciate
> the review and discussion on AVX10.2 and AMX since August. It has been
> a really long run.
> 
> Ok for trunk?

Okay, on the assumption that the doc will be updated accordingly in due
course, and hence we're not going to end up with any back and forth.

Jan