[SME] Add __arm_agnostic("sme_za_state") keyword attribute

[sdesmalen-arm]

The __arm_agnostic keyword attribute enables the user to specify that a function is agnostic to a specified piece of architectural state. That means that the function must preserve this state when it exists, or otherwise ignores its contents.

The reason for not naming this something like __arm_za_compatible was so that we might want use the attribute keyword for other architectural state in the future.

---

name: Pull request
about: Technical issues, document format problems, bugs in scripts or feature proposal.

---

Thank you for submitting a pull request!

If this PR is about a bugfix:

Please use the bugfix label and make sure to go through the checklist below.

If this PR is about a proposal:

We are looking forward to evaluate your proposal, and if possible to
make it part of the Arm C Language Extension (ACLE) specifications.

We would like to encourage you reading through the contribution
guidelines, in particular the section on submitting
a proposal.

Please use the proposal label.

As for any pull request, please make sure to go through the below
checklist.

Checklist: (mark with X those which apply)

• If an issue reporting the bug exists, I have mentioned it in the
  PR (do not bother creating the issue if all you want to do is
  fixing the bug yourself).
• I have added/updated the SPDX-FileCopyrightText lines on top
  of any file I have edited. Format is SPDX-FileCopyrightText: Copyright {year} {entity or name} <{contact informations}>
  (Please update existing copyright lines if applicable. You can
  specify year ranges with hyphen , as in 2017-2019, and use
  commas to separate gaps, as in 2018-2020, 2022).
• I have updated the Copyright section of the sources of the
  specification I have edited (this will show up in the text
  rendered in the PDF and other output format supported). The
  format is the same described in the previous item.
• I have run the CI scripts (if applicable, as they might be
  tricky to set up on non-*nix machines). The sequence can be
  found in the contribution
  guidelines. Don't
  worry if you cannot run these scripts on your machine, your
  patch will be automatically checked in the Actions of the pull
  request.
• I have added an item that describes the changes I have
  introduced in this PR in the section Changes for next
  release of the section Change Control/Document history
  of the document. Create Changes for next release if it does
  not exist. Notice that changes that are not modifying the
  content and rendering of the specifications (both HTML and PDF)
  do not need to be listed.
• When modifying content and/or its rendering, I have checked the
  correctness of the result in the PDF output (please refer to the
  instructions on how to build the PDFs
  locally).
• The variable draftversion is set to true in the YAML header
  of the sources of the specifications I have modified.
• Please DO NOT add my GitHub profile to the list of contributors
  in the README page of the project.

[sdesmalen-arm]

@rsandifo-arm

[rsandifo-arm]

The patch should also update the “Inline assembly” section, since ZA can now be “on” on entry to an inline asm in cases that the current text doesn't cover.

[rsandifo-arm]

Nit: would be good to keep the entries sorted by name.

[rsandifo-arm]

__arm_new is only meaningful for definitions, not declarations. I'm not sure it should be a semantic requirement, since:

__arm_new("za") void foo(void) …stuff… {
    …
}

isn't semantically different from:

__arm_new("za") void foo1(void) {
    …
}
void foo(void) …stuff… {
    foo1();
}

(It's reasonable for an implementation to say that it doesn't support __arm_new for now though. This would be the equivalent of a GCC “sorry” diagnostic, but it'd be an implementation choice rather than something that the ACLE should document.)

I think we can then say:

It is not valid … to [share](#shares-state) PSTATE.ZA state with its caller.

[rsandifo-arm]

Since the AAPCS64 patch (rightly) creates a three-way choice between private, shared & agnostic, we should update the description under “Mapping to the Procedure Call Standard” section. I think the text here should then say something like:

This attribute affects the ABI of object code functions, as described under … It is the compiler's responsibility to ensure that such object code functions honor the ABI requirements.

and then delete the following two bullet points. AFAICT, the attribute has no effect on the behaviour of the abstract machine outside of inline asms.

[rsandifo-arm]

The previous comment about updating the inline asm documentation still stands.

[sdesmalen-arm]

The previous comment about updating the inline asm documentation still stands.

Thanks, I missed that comment previously. I've now added a comment that I believe is sufficient to cover the agnostic-ZA case, but let me know if there's still anything missing!

[rsandifo-arm]

I still think we should drop these two paragraphs. IMO the purpose of this section is to describe how ACLE features map to the ABI, and it's the ABI's job to describe the actual binary interface.

I was hoping the new “The use of this attribute does not imply that SME is available.” statement would replace the second paragraph, since IMO it's more precise. E.g. it's valid to do:

  void f() __arm_agnostic("sme_state") __arm_streaming {}

and in that case, the implementation can assume that PSTATE.ZA is available.

In other words, the point is that __arm_agnostic does not by itself guarantee that SME is available, but the compiler might still know that SME is available from other sources, e.g. from +sme or from other attributes. __arm_agnostic doesn't override information from other sources.

[rsandifo-arm]

Looks good, but this should be reflected in the table at the end of the section as well.

I think we also need to decide what to do about:

*   S is not an input to the asm and S is not an output from the asm.
    However, the contents of S after executing the asm's instructions might
    be different from the contents of S before executing the instructions.
    This is indicated by adding S's string to the asm's clobber list.

    If an asm takes this option for state that is controlled by PSTATE.ZA,
    the asm itself is responsible for handling the [[AAPCS64]](#AAPCS64)
    lazy save scheme.

The current rule implicity assumes that ZA can only be dormant or off. If ZA is active, then is the asm required to preserve ZA despite the clobber? Does the compiler do this? (In which case, does it also handle the lazy-save scheme, meaning that the asm doesn't need to?) Or is this option not valid in agnostic-ZA functions?

[sdesmalen-arm]

My thinking on this was that for __arm_agnostic("sme_za_state") functions, if the asm clobbers any of the state enabled by PSTATE.ZA, the user is responsible for letting the function preserve this state using the provided routines such that on return from the function the ABI is honoured.

[rsandifo-arm]

LGTM.