x86: IAMCU, L1OM, and K1OM

x86: IAMCU, L1OM, and K1OM

[Jan Beulich]

H.J.,

may I ask what the status / purpose of these is? They were added years
ago, but actually useful functionality has never appeared. Since I've
noticed a couple of issues with the involved gas code, I'm now
wondering: Should I try to make fixes, or can the logic rather be
purged altogether? Even if the overhead isn't high, the presence of
support for these not really usable sub-architectures is causing some
extra cycles to be wasted on each and every gas invocation.

Thanks for any insight into possible plans here,
Jan

Re: x86: IAMCU, L1OM, and K1OM

[H.J. Lu]

On Mon, Mar 7, 2022 at 3:04 AM Jan Beulich <jbeulich@suse.com> wrote:
>
> H.J.,
>
> may I ask what the status / purpose of these is? They were added years
> ago, but actually useful functionality has never appeared. Since I've
> noticed a couple of issues with the involved gas code, I'm now
> wondering: Should I try to make fixes, or can the logic rather be
> purged altogether? Even if the overhead isn't high, the presence of
> support for these not really usable sub-architectures is causing some
> extra cycles to be wasted on each and every gas invocation.
>
> Thanks for any insight into possible plans here,
> Jan
>

We can deprecate L1OM and K1OM.  But we should
keep IAMCU.

Thanks.

-- 
H.J.

Re: x86: IAMCU, L1OM, and K1OM

[Jan Beulich]

On 07.03.2022 15:09, H.J. Lu wrote:
> On Mon, Mar 7, 2022 at 3:04 AM Jan Beulich <jbeulich@suse.com> wrote:
>> may I ask what the status / purpose of these is? They were added years
>> ago, but actually useful functionality has never appeared. Since I've
>> noticed a couple of issues with the involved gas code, I'm now
>> wondering: Should I try to make fixes, or can the logic rather be
>> purged altogether? Even if the overhead isn't high, the presence of
>> support for these not really usable sub-architectures is causing some
>> extra cycles to be wasted on each and every gas invocation.
> 
> We can deprecate L1OM and K1OM.

And "deprecate" == "delete" or something less heavy?

>  But we should keep IAMCU.

IOW there's stuff to come to actually fill this?

Jan

Re: x86: IAMCU, L1OM, and K1OM

[H.J. Lu]

On Mon, Mar 7, 2022 at 6:14 AM Jan Beulich <jbeulich@suse.com> wrote:
>
> On 07.03.2022 15:09, H.J. Lu wrote:
> > On Mon, Mar 7, 2022 at 3:04 AM Jan Beulich <jbeulich@suse.com> wrote:
> >> may I ask what the status / purpose of these is? They were added years
> >> ago, but actually useful functionality has never appeared. Since I've
> >> noticed a couple of issues with the involved gas code, I'm now
> >> wondering: Should I try to make fixes, or can the logic rather be
> >> purged altogether? Even if the overhead isn't high, the presence of
> >> support for these not really usable sub-architectures is causing some
> >> extra cycles to be wasted on each and every gas invocation.
> >
> > We can deprecate L1OM and K1OM.
>
> And "deprecate" == "delete" or something less heavy?

We can delete them and make an announcement.

> >  But we should keep IAMCU.
>
> IOW there's stuff to come to actually fill this?
>

What do you mean?  IAMCU is used.

-- 
H.J.

Re: x86: IAMCU, L1OM, and K1OM

[Jan Beulich]

On 07.03.2022 15:19, H.J. Lu wrote:
> On Mon, Mar 7, 2022 at 6:14 AM Jan Beulich <jbeulich@suse.com> wrote:
>> On 07.03.2022 15:09, H.J. Lu wrote:
>>>  But we should keep IAMCU.
>>
>> IOW there's stuff to come to actually fill this?
> 
> What do you mean?  IAMCU is used.

Are you suggesting there is a product which must have appeared long
after the original Pentium, but which only supports i586 insns:

  { "CPU_IAMCU_FLAGS",
    "Cpu186|Cpu286|Cpu386|Cpu486|Cpu586" },

I guess I'm puzzled ...

Jan

Re: x86: IAMCU, L1OM, and K1OM

[H.J. Lu]

On Mon, Mar 7, 2022 at 6:24 AM Jan Beulich <jbeulich@suse.com> wrote:
>
> On 07.03.2022 15:19, H.J. Lu wrote:
> > On Mon, Mar 7, 2022 at 6:14 AM Jan Beulich <jbeulich@suse.com> wrote:
> >> On 07.03.2022 15:09, H.J. Lu wrote:
> >>>  But we should keep IAMCU.
> >>
> >> IOW there's stuff to come to actually fill this?
> >
> > What do you mean?  IAMCU is used.
>
> Are you suggesting there is a product which must have appeared long
> after the original Pentium, but which only supports i586 insns:

Yes:

https://www.intel.com/content/www/us/en/embedded/products/quark/overview.html

>   { "CPU_IAMCU_FLAGS",
>     "Cpu186|Cpu286|Cpu386|Cpu486|Cpu586" },
>
> I guess I'm puzzled ...
>
> Jan
>


-- 
H.J.

Re: x86: IAMCU, L1OM, and K1OM

[Jan Beulich]

On 07.03.2022 15:48, H.J. Lu wrote:
> On Mon, Mar 7, 2022 at 6:24 AM Jan Beulich <jbeulich@suse.com> wrote:
>>
>> On 07.03.2022 15:19, H.J. Lu wrote:
>>> On Mon, Mar 7, 2022 at 6:14 AM Jan Beulich <jbeulich@suse.com> wrote:
>>>> On 07.03.2022 15:09, H.J. Lu wrote:
>>>>>  But we should keep IAMCU.
>>>>
>>>> IOW there's stuff to come to actually fill this?
>>>
>>> What do you mean?  IAMCU is used.
>>
>> Are you suggesting there is a product which must have appeared long
>> after the original Pentium, but which only supports i586 insns:
> 
> Yes:
> 
> https://www.intel.com/content/www/us/en/embedded/products/quark/overview.html

Oh, I see, thanks for the info. I would never have made a link between iamcu
an Quark.

Jan