[PATCH v2] x86: Remove the restriction for size of the mask register in AVX10

[Jan Beulich]

On 18.12.2023 04:26, Haochen Jiang wrote:
> This is the v2 patch to remove the restiction for the size of the mask register
> in AVX10.
> 
> Changes in v2: Remove attribute Vsz in opcode_modifier since it is no longer
> used. But I suppose we still need a similar enum in gas to check encoding is
> right.

Indeed. Change is almost okay in this shape:

> --- a/gas/config/tc-i386.c
> +++ b/gas/config/tc-i386.c
> @@ -823,6 +823,11 @@ static unsigned int sse2avx;
>  static unsigned int use_unaligned_vector_move;
>  
>  /* Maximum permitted vector size. */
> +static enum {
> +    VSZ128 = 0,
> +    VSZ256,
> +    VSZ512,
> +} Vsz;

Looks like this static variable is unused?

>  #define VSZ_DEFAULT VSZ512
>  static unsigned int vector_size = VSZ_DEFAULT;

And this variable would then better be of the new enum type? Albeit ...

> @@ -6967,12 +6972,10 @@ VEX_check_encoding (const insn_template *t)
>  
>    /* Vector size restrictions.  */
>    if ((vector_size < VSZ512
> -       && (t->opcode_modifier.evex == EVEX512
> -	   || t->opcode_modifier.vsz >= VSZ512))
> +       && t->opcode_modifier.evex == EVEX512)
>        || (vector_size < VSZ256
>  	  && (t->opcode_modifier.evex == EVEX256
> -	      || t->opcode_modifier.vex == VEX256
> -	      || t->opcode_modifier.vsz >= VSZ256)))
> +	      || t->opcode_modifier.vex == VEX256)))
>      {

... the use of < here makes me wonder whether an enum is really the
best thing to use. Maybe better to stick to #define-s?

Jan