MPX: can we remove it?

[Jan Beulich]

On 19.08.2020 10:26, Martin Liška wrote:
> On 8/19/20 10:12 AM, Jan Beulich wrote:
>> On 19.08.2020 10:03, Martin Liška wrote:
>>> On 8/18/20 11:33 PM, H.J. Lu wrote:
>>>> On Tue, Aug 18, 2020 at 1:33 PM Martin Liška <mliska@suse.cz> wrote:
>>>>>
>>>>> Hello.
>>>>>
>>>>> MPX was already removed from GCC and Linux Kernel.
>>>>> Can we do the same in binutils (and later in gdb)?
>>>>> If so, I can prepare a patch for it.
>>>>>
>>>>
>>>> We should keep assembler and disassembler parts.
>>>
>>> Why is it still needed?
>>
>> Because people may be using it for varying purposes.
> 
> Hmm, do you have an example or two?

1) Decisions taken by gcc or Linux may not be follow by other compilers
or OSes.

2) Personally I use gas for building test cases for my own disassembler
library.

>> I don't think we
>> should ever remove any insn that was properly documented and was part
>> of some production hardware. SSE5 support, so take an example, didn't
>> meet this criteria afaik, and hence was reasonable to remove again.
> 
> People can still use a legacy toolchain that supports that.

Not necessarily: In order to use newer insns newer tool chains may be
needed. If removal of features happened even just slightly more
frequently, this model of building code would quickly grow beyond
what's reasonable to maintain.

> In case of MPX, it was removed from GCC, Linux kernel, LLVM and from Intel
> silicons.

It's not going to be in new silicon, but I doubt it was "removed" in
the physical sense from existing chips. If at all there a ucode update
available disabling it, but that not "remove" (not the least because
people aren't required to install it).

Jan