An efficient multi-format low-precision floating-point multiplier It supports three 16-bit formats: Half_ALT BFloat DLFloat