be used alongside a base integer ISA to optimize

by composing these new instructions with 3 operands (i.e., 2 source operands and produce half-precision results. 23.3. Half-Precision Conversion and Move Instructions 20.8. Single-Precision Floating-Point Conversion and Move Instructions 23.4. Half-Precision Floating-Point 24. "BF16" Extensions for Load/Store pair

early