Re: [PATCH] testsuite, X86, Darwin: Fix bf16 ABI tests for Mach-O/macOS ABI.

[Uros Bizjak <ubizjak@gmail.com>]

On Sun, Dec 4, 2022 at 9:30 PM Iain Sandoe <iain@sandoe.co.uk> wrote:
>
>
>
> > On 4 Dec 2022, at 20:20, Uros Bizjak via Gcc-patches <gcc-patches@gcc.gnu.org> wrote:
> >
> > On Sun, Dec 4, 2022 at 12:51 PM Iain Sandoe <iains.gcc@gmail.com> wrote:
> >>
> >> This is almost a completely Darwin-local patch, but there is one (repeated)
> >> place where a general change is needed - which is in making xmm_regs and
> >> x87_regs extern in the three copies of args.h (this is consistent with the
> >> other saved vars).  These fails represent most of the current testsuite noise
> >> on x86 Darwin.
> >>
> >> tested on x86-64 Darwin and Linux.
> >>
> >> OK for master?
> >> Iain
> >>
> >> -- >8 --
> >>
> >> These tests have failed since introduction since they assume that the
> >> assembler output is ELF and that the ABI targeted supports the addressing.
> >>
> >> For Darwin, Mach-O and ABI we need to make several changes:
> >> 1. Use the __USER_LABEL__PREFIX__
> >> 2. Remove the use of ELF-specific constructs (.size, .type etc.)
> >> 3. We cannot make direct access to common variables in the ABI, so that we
> >>   must move these to BSS.
> >>
> >> Since that set is quite significant, I elected to make a separate source
> >> section for Darwin.  This is introduced by #elif defined(__APPLE__) because
> >> __MACH__ is also used by HURD.
> >>
> >> There are potentially other X86 targets (e.g. XCOFF) that could have yet
> >> more changes, so I added a catchall section that #errors if the object format
> >> is neither ELF or Mach-O.
> >>
> >> Signed-off-by: Iain Sandoe <iain@sandoe.co.uk>
> >>
> >> gcc/testsuite/ChangeLog:
> >>
> >>        * gcc.target/x86_64/abi/bf16/args.h:
> >>        * gcc.target/x86_64/abi/bf16/asm-support.S:
> >>        * gcc.target/x86_64/abi/bf16/m256bf16/args.h:
> >>        * gcc.target/x86_64/abi/bf16/m256bf16/asm-support.S:
> >>        * gcc.target/x86_64/abi/bf16/m512bf16/args.h:
> >>        * gcc.target/x86_64/abi/bf16/m512bf16/asm-support.S:
> >
> > Missing descriptions in ChangeLog entry.
>
> oops, here:
>
> gcc/testsuite/ChangeLog:
>
>         * gcc.target/x86_64/abi/bf16/args.h: Make xmm_regs, x87_regs extern.
>         * gcc.target/x86_64/abi/bf16/m256bf16/args.h: Likewise.
>         * gcc.target/x86_64/abi/bf16/m512bf16/args.h: Likewise.
>         * gcc.target/x86_64/abi/bf16/asm-support.S: Add Mach-O variant.
>         * gcc.target/x86_64/abi/bf16/m256bf16/asm-support.S: Likewise.
>         * gcc.target/x86_64/abi/bf16/m512bf16/asm-support.S: Likewise.

Please note that in other directories asm-support-darwin.s is
introduced and included via .exp file. Is there a reason a different
approach is introduced here?

Uros.