6.6.2.2.1 IEEE 754-2008 Compliance

This module is compliant with the IEEE 754-2008 Standard for Floating-Point Arithmetic for data formats, supported signaling and quiet branch predicates, exception status flags, and exception status behavior.

Note that the IEEE 754-2008 minNum(x,y) and maxNum(x,y) definitions are supported only through the largely compatible IEEE 754-2019 minimumNum(x,y) and maximumNum(x,y) operations via the FMINNUM and FMAXNUM instructions. The functional differences related to how +0 and -0 are considered:

  • IEEE 754-2008 minNum(x,y) / maxNum(x,y): Operand values +0 and -0 are regarded as equivalent. The (implementation dependent) result could therefore be either +0 or -0.
  • IEEE 754-2019 minimumNum(x,y) / maximumNum(x,y): Operand values +0 and -0 are not regarded as equivalent such that -0 compares to less than +0. The result will therefore be the correct sign of 0 based on the selected operation.