Re: [PATCH v2 0/6] tests/tcg/aarch64: Fix inline assemblies for clang

["Alex Bennée" <alex.bennee@linaro.org>]

Akihiko Odaki <akihiko.odaki@daynix.com> writes:

> Unlike GCC, clang checks if the operands in assembly matches with the
> type in C. It also does not support "x" constraint for AArch64 and
> complains about them.

I guess there are more needed:

  ninja: no work to do.
  /home/alex/lsrc/qemu.git/builds/all.clang/pyvenv/bin/meson introspect --targets --tests --benchmarks | /home/alex/lsrc/qemu.git/builds/all.clang/pyvenv/bin/python3 -B scripts/mtest2make.py > Makefile.mtest
    BUILD   aarch64-linux-user guest-tests
  tests/tcg/aarch64-linux-user: -march=armv8.1-a+sve detected
  tests/tcg/aarch64-linux-user: -march=armv8.1-a+sve2 detected
  tests/tcg/aarch64-linux-user: -march=armv8.2-a detected
  tests/tcg/aarch64-linux-user: -march=armv8.3-a detected
  tests/tcg/aarch64-linux-user: -march=armv8.5-a detected
  tests/tcg/aarch64-linux-user: -mbranch-protection=standard detected
  tests/tcg/aarch64-linux-user: -march=armv8.5-a+memtag detected
  tests/tcg/aarch64-linux-user: -Wa,-march=armv9-a+sme detected
  tests/tcg/aarch64-linux-user: -march=armv9-a+sme-i16i64 not detected
  <inline asm>:11:2: error: instruction requires: sve or sme
          ptrue p0.s, vl4
          ^
  <inline asm>:12:2: error: instruction requires: sve or sme
          fmov z0.s, #1.0
          ^
  <inline asm>:20:2: error: instruction requires: sve or sme
          st1w {z0.s}, p0, [x0]
          ^
  <inline asm>:22:2: error: instruction requires: sve or sme
          st1w {z1.s}, p0, [x0]
          ^
  <inline asm>:24:2: error: instruction requires: sve or sme
          st1w {z2.s}, p0, [x0]
          ^
  <inline asm>:26:2: error: instruction requires: sve or sme
          st1w {z3.s}, p0, [x0]
          ^
  6 errors generated.
  make[1]: *** [Makefile:116: sme-outprod1] Error 1
  make: *** [/home/alex/lsrc/qemu.git/tests/Makefile.include:50: build-tcg-tests-aarch64-linux-user] Error 2

>
> Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com>
> ---
> Changes in v2:
> - Removed spurious a compiler flag change for normal SME tests.
> - Fixed sme-i16i64 detection.
> - Link to v1: https://lore.kernel.org/r/20240626-tcg-v1-0-0bad656307d8@daynix.com
>
> ---
> Akihiko Odaki (6):
>       tests/tcg/arm: Fix fcvt result messages
>       tests/tcg/aarch64: Fix test architecture specification
>       tests/tcg/aarch64: Explicitly specify register width
>       tests/tcg/aarch64: Fix irg operand type
>       tests/tcg/aarch64: Do not use x constraint
>       tests/tcg/arm: Manually bit-cast half-precision numbers
>
>  tests/tcg/aarch64/bti-1.c         |   6 +-
>  tests/tcg/aarch64/bti-3.c         |   6 +-
>  tests/tcg/aarch64/mte-1.c         |   2 +-
>  tests/tcg/aarch64/sme-smopa-2.c   |   2 +-
>  tests/tcg/arm/fcvt.c              |  20 +-
>  tests/tcg/aarch64/Makefile.target |  11 +-
>  tests/tcg/aarch64/fcvt.ref        | 604 +++++++++++++++++++-------------------
>  7 files changed, 330 insertions(+), 321 deletions(-)
> ---
> base-commit: 046a64b9801343e2e89eef10c7a48eec8d8c0d4f
> change-id: 20240624-tcg-bf8116e80afa
>
> Best regards,

-- 
Alex Bennée
Virtualisation Tech Lead @ Linaro