Re: [PATCH 06/10] Support Intel RAO-INT

[Jan Beulich <jbeulich@suse.com>]

On 28.10.2022 18:08, H.J. Lu wrote:
> On Fri, Oct 28, 2022 at 1:40 AM Jan Beulich <jbeulich@suse.com> wrote:
>>
>> On 28.10.2022 10:31, Jiang, Haochen wrote:
>>>> -----Original Message-----
>>>> From: Jan Beulich <jbeulich@suse.com>
>>>> Sent: Friday, October 28, 2022 4:22 PM
>>>>
>>>> On 28.10.2022 10:10, Jiang, Haochen wrote:
>>>>> BTW, should the suffix instruction dependent? It might be more operand
>>>>> related from my opinion. If that is the truth, could we just judge
>>>>> whether we should add them when dealing with memory operands?
>>>>
>>>> I'm afraid I don't really understand what you're saying/asking here.
>>>> In any event - whether a suffix is required indeed depends on insn operands.
>>>> Yet even insns with (only) GPR operands _may_ use a suffix in AT&T mode,
>>>> irrespective of it being derivable from those GPR operands. We actually apply
>>>> consistency checks between registers used and the suffix (if present).
>>>
>>> What I am saying is we could put all the suffix module out of instructions. If we
>>> found that are using those variable GPR operands, then to determine whether
>>> we should allow suffixes instead of determining at instructions.
>>
>> Once again - even with GPR operands use of suffixes is permitted (and
>> actually kind of mandated by the only AT&T spec I'm aware of). You
>> may have seen the series that I have pending to re-work some of the
>> suffix recognition, but that's certainly not going in the direction
>> you're suggesting (if I understand what you're saying; perhaps if you
>> gave an example it might become more clear).
>>
> 
> The old AT&T syntax rules, which are quite vague, don't apply to new
> instructions.

Why would they not? These rules are intended to cover the full set of
insns. I'm afraid I need to keep repeating myself: Inconsistency here
leads to non-predictable overall behavior of the assembler (and of
the disassembler as well, in suffix-always mode, just that there it
won't lead to unexpected errors).

Jan

> For new instructions, the suffix should be permitted and required only
> if it must
> be used to specify the operand size.
> 
>