[PATCH v2] x86: Add {imm8}/{imm16}/{imm32} pseudo prefixes

[Jan Beulich]

On 24.03.2025 19:41, H.J. Lu wrote:
> On Mon, Mar 24, 2025 at 9:27 AM Jan Beulich <jbeulich@suse.com> wrote:
>>
>> On 24.03.2025 17:24, H.J. Lu wrote:
>>> Instruction templates with only sign-extended 8-bit immediate operand
>>> also have a second template with full-operand-size immediate operand
>>> under a different opcode.  Add {imm8}, {imm16} and {imm32} pseudo
>>> prefixes as a hint to control immediate size.  Don't match the first
>>> template if {imm16} or {imm32} are used.
>>>
>>> gas/
>>>
>>>       PR gas/32811
>>>       * config/tc-i386.c (pseudo_prefixes): Add avoid_imm8s.
>>>       (operand_size_match): Return false for templates with only sign-
>>>       extended 8-bit immediate operand if {imm16} or {imm32} are used.
>>>       (parse_insn): Handle Prefix_Imm8, Prefix_Imm16 and Prefix_Imm32.
>>>       * doc/c-i386.texi: Document {imm8}, {imm16} and {imm32}.
>>>       * testsuite/gas/i386/pseudos.s: Add tests for {imm8}, {imm16}
>>>       and {imm32}.
>>>       * testsuite/gas/i386/x86-64-pseudos.s: Likewise.
>>>       * testsuite/gas/i386/pseudos.d: Updated.
>>>       * testsuite/gas/i386/x86-64-pseudos.d: Likewise.
>>>
>>> opcodes/
>>>
>>>       PR gas/32811
>>>       * opcodes/i386-opc.h (Prefix_Imm8): New.
>>>       (Prefix_Imm16): Likewise.
>>>       (Prefix_Imm32): Likewise.
>>>       * i386-opc.tbl: Add {imm8}, {imm16} and {imm32} pseudo prefixes.
>>>       * i386-mnem.h: Regenerated.
>>>       * i386-tbl.h: Likewise.
>>>
>>> Signed-off-by: H.J. Lu <hjl.tools@gmail.com>
>>
>> What's the difference to v1? Importantly {imm16} and {imm32} are still there,
> 
> full-operand-size and avoid_imm8s.
> 
>> when imo they should be suitably folded.
>>
> 
> {imm16} and {imm32} are the same internally.  But for users,
> 
> {imm16} add $0,%bx
>  {imm8} add $0,%bx
> 
> are quite clearer than other alternatives.

And

    {imm16} add $0,%ebx

is, I'm sorry to say so, rubbish. Which however we happily accept to mean

    {imm32} add $0,%ebx

Thinking about it, unlike {disp8} - what exact purpose does {imm8} serve?
We encode with the smallest possible immediate anyway. {disp8} is useful
to request a displacement when none would be needed, yet the same situation
doesn't arise for immediates afaict. Hence the sole use looks to be to
"override" an earlier {imm16} or {imm32}. I'm not entirely certain that's
worth it.

Jan