IIRC float to integer conversion is somewhat underspecified in IEEE754, and MIPS and other architectures do different things, leading to unpredictable (and bad) decoding results.
In any case we shouldn't be underflowing in the first place!!
This should fix that everywhere, and also clarifies some magic constants to be less magic.
IIRC float to integer conversion is somewhat underspecified in IEEE754, and MIPS and other architectures do different things, leading to unpredictable (and bad) decoding results.
In any case we shouldn't be underflowing in the first place!!
This should fix that everywhere, and also clarifies some magic constants to be less magic.