This is the mail archive of the binutils@sourceware.org mailing list for the binutils project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCH] ix86/Intel: don't require memory operand size specifier for PTWRITE


>>> On 16.11.17 at 00:12, <hjl.tools@gmail.com> wrote:
> On Wed, Nov 15, 2017 at 5:58 AM, Jan Beulich <JBeulich@suse.com> wrote:
>>>>> On 15.11.17 at 14:24, <hjl.tools@gmail.com> wrote:
>>> On Wed, Nov 15, 2017 at 5:16 AM, Jan Beulich <JBeulich@suse.com> wrote:
>>>> Other than in 64-bit mode, in 32- and 16-bit modes operand size isn't
>>>> ambiguous.
>>
>> With this ...
>>
>>>> --- 2017-11-10/gas/testsuite/gas/i386/ptwrite.s
>>>> +++ 2017-11-10/gas/testsuite/gas/i386/ptwrite.s
>>>> @@ -9,4 +9,5 @@ _start:
>>>>
>>>>         .intel_syntax noprefix
>>>>         ptwrite ecx
>>>> +       ptwrite [ecx]
>>>
>>> Do we also need such a test in x86-64-ptwrite.s?
>>
>> ... the answer is clearly no: The operand size _is_ ambiguous
>> there. In fact I have (as part of the larger patch talked about in
>> the other thread) queued an adjustment removing the suffix-less
>> AT&T variant there.
>>
> 
> Why isn't
> 
>       ptwrite (%ecx)
> 
> ambiguous in 64-bit? Is that because the `l' suffix is implied?

Yes, it is wrongly not flagged as ambiguous; that would be fixed
as part of the "AT&T syntax operand size defaults" work.

> OK.

Thanks.

Jan


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]