This is the mail archive of the 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] x86/Intel: accept mandated operand order for vcvt{,u}si2s{d,s}

On Thu, May 21, 2015 at 4:36 AM, Jan Beulich <> wrote:
>>>> On 21.05.15 at 12:42, <> wrote:
>> On Wed, May 20, 2015 at 11:30 PM, Jan Beulich <> wrote:
>>>>>> On 06.05.15 at 09:44, <> wrote:
>>>> Please don't just repeat yourself, but give a reason I can understand
>>>> to override the intention to conform with the Intel manual. I'm
>>>> certainly hesitant to commit changes that can't be agreed upon, but
>>>> as said before I don't feel tied to your disapproval of the changes.
>>> I guess I'll take two weeks of silence as silent withdrawal of the
>>> objection to the patches then.
>> Please don't change it. As I said before, we have discussed it at Intel
>> and we don't think the change is appropriate.
> So are you planning to change the SDM? Else I don't see what new
> aspect you are trying to tell me. I'm hesitant to commit the changes
> without your consent, but getting back silence or all the same
> vague arguments I'm afraid all I can do is give you a little more
> time (say a week; if you need more, please give a clear time line) to
> come forward with something substantial.

I checked with our SDM people and was told that

Intel Software Developer Manual only governs the output side of the binary
form of instruction byte stream matches what HW expect. Each assembly
tool product has its own implementation of transforming the input
language/dialect into the output stream.

Intel syntax supported by GNU assembler is what is implemented
in GNU assembler.  Given that there is nothing we can be compatible
with, we don't want to change it.

> (Also please recall me having stated before that the AT&T operand
> ordering has got completely screwed up over its apparent original
> intentions with all the more-than-two operand instructions that got
> added over the last so many years. This brokenness should _not_
> impact Intel syntax mode.)

I agreed with you on this.  But it was before my time and I can't
change it.


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