This series fix following issues related to OCP FP8 E4M3 conversion.
- Numeric overflow to E4M3 produces incorrect exponent values due to a
missing local variable update after calling uncanon_e4m3_overflow.
- Converting infinity to E4M3 incorrectly raises OF|NX exception flags.
Changes in version 2:
- The “fpu: Simplify OCP FP8 E4M3 NaN classification in parts_canonicalize”
patch has been removed. This change retains the flexibility to classify NaNs.
rnax
Max Chou (2):
fpu: Fix repacking issues in the uncanonical step for E4M3 overflow
fpu: Fix unexpected exception flags when converting infinity to OCP
E4M3
fpu/softfloat-parts.c.inc | 4 +++-
1 file changed, 3 insertions(+), 1 deletion(-)
--
2.52.0
