Re: [PATCH 3/6] x86: improve operand reversal

["Jan Beulich" <JBeulich at suse dot com>]

>>> On 02.08.18 at 14:51, <hjl.tools@gmail.com> wrote:
> On Thu, Aug 2, 2018 at 5:38 AM, Jan Beulich <JBeulich@suse.com> wrote:
>>>>> On 02.08.18 at 14:19, <hjl.tools@gmail.com> wrote:
>>> On Wed, Aug 1, 2018 at 11:49 PM, Jan Beulich <JBeulich@suse.com> wrote:
>>>> In quite a few cases the .s suffix or {load} / {store} prefixes did not
>>>> work as intended, or produced errors when they're supposed to be ignored
>>>> when it is not possible to carry out the request.
>>>>
>>>> The change here re-purposes(?) the .s suffix to no longer mean "store"
>>>> (if that's what 's' did stand for), since the forms used in the base
>>>> templates are not consistently loads (and we unlikely want to change
>>>> that). The pseudo prefixes will now fulfill what their names say, i.e.
>>>> {load} now only ever produces a load form encoding (if available) while
>>>> {store} only ever produces a store form one (again if available). This
>>>> requires minimal test suite adjustments, while the majority of the
>>>> changes there are simply additions.
>>>>
>>>
>>> I prefer not to change the behavior of the `.s' suffix, unless it is to fix
>>> the wrong encoding.  I don't see the need for the 'swap' pseudo prefix.
>>> If the programmer doesn't care load/store encoding, "swap" isn't really
>>> useful.
>>
>> There's no {swap} prefix. I've just (verbally) assigned the meaning of
>> "swap" to the .s suffix (which is the behavior it always had afaict,
>> rather than forcing a store to be used, as one could have implied from
>> it being 's' and it having had the same behavior as {store}, just that
>> the behavior was clearly wrong for {store}). This isn't a very good
>> association, but I couldn't come up with anything better that would
>> fit the 's'.
>>
>> Please take a look at the testsuite adjustments - this gives a pretty
>> good picture, and you'll notice that uses of .s continue to behave as
>> before. The adjustments (beyond the various additions) were for
>> {load} and/or {store} now behaving according to their names.
>>
> 
> There are many test changes,  Can you list a couple for before and
> after comparison?

I'm afraid I don't really understand what you're after. Looking at the
patch makes pretty obvious what the very few changes are that aren't
plain additions. Iirc there were only two mov-es with {load} (or was
it {store}) prefix where the output changed.

Jan