[PATCH v4 0/9] Support Intel APX EGPR

[Jan Beulich]

On 20.12.2023 09:50, Cui, Lili wrote:
>> On 19.12.2023 13:12, Cui, Lili wrote:
>>> *** BLURB HERE ***
>>> Future optimizations to be made.
>>> 1. The current implementation of vexvvvvv needs to be optimized.
>>> 2. The handling of double VEX/EVEX templates in check_register() needs to
>> be optimized.
>>
>> I hope this is just stale here, and the dependency on templates was now
>> removed again from check_register().
> 
> In fact, I didn't remove it in V4, I didn't find a better place to deal with it. I don't know if you agree with this implementation below.

I'm afraid I don't, both because it still isn't clear to me what's wrong
with my alternative proposal, and also for the formal reason of ...

> /* For dual VEX/EVEX templates, evex encoding is required when the input has
>    egpr.*/
> static INLINE void
> vex_with_Egpr_requires_evex_encoding (const insn_template *t)
> {
>   for (unsigned int op = 0; op < i.operands; op++)
>     {
>       if (i.types[op].bitfield.class != Reg)
>         continue;
> 
>       if (i.op[op].regs->reg_flags & RegRex2)
>         i.vec_encoding = vex_encoding_evex;

... it not being okay to override i.vec_encoding like this, when it
may already have been set to another value.

Jan

>     }
> 
>   if ((i.index_reg && (i.index_reg->reg_flags & RegRex2))
>       || (i.base_reg && (i.base_reg->reg_flags & RegRex2)))
>     i.vec_encoding = vex_encoding_evex;
> }
> 
> static INLINE void
> install_template (const insn_template *t)
> {
>   unsigned int l;
> 
>   i.tm = *t;
> 
>   /* Dual VEX/EVEX templates need stripping one of the possible variants.  */
>   if (t->opcode_modifier.vex && t->opcode_modifier.evex)
>     {
>       vex_with_Egpr_requires_evex_encoding (t);
> 
> 
> Regards,
> Lili.