[0/2] target/s390x: Fix the floating-point multiply-and-add NaN rules

Message

Ilya Leoshkevich Oct. 22, 2024, 11:59 p.m. UTC

Hi,

Peter reported on IRC that FMA NaN handling was likely broken on s390x.
Patch 1 of this series fixes the issue, patch 2 adds a test. For the
sake of readability, the test sacrifices the 90-character line length
rule.

Best regards,
Ilya

Ilya Leoshkevich (2):
  target/s390x: Fix the floating-point multiply-and-add NaN rules
  tests/tcg/s390x: Add the floating-point multiply-and-add test

 fpu/softfloat-specialize.c.inc    |  19 +++
 target/s390x/tcg/fpu_helper.c     |   8 +-
 target/s390x/tcg/vec_fpu_helper.c |  12 +-
 tests/tcg/s390x/Makefile.target   |   5 +-
 tests/tcg/s390x/float.h           | 104 +++++++++++++
 tests/tcg/s390x/fma.c             | 233 ++++++++++++++++++++++++++++++
 tests/tcg/s390x/vfminmax.c        | 223 ++++++++++------------------
 7 files changed, 449 insertions(+), 155 deletions(-)
 create mode 100644 tests/tcg/s390x/float.h
 create mode 100644 tests/tcg/s390x/fma.c