Floating-point multiplication unit and floating point photonic tensor accelerator
AssigneeUniversity of Central Florida Research Foundation, Inc.
InventorsGuifang LI, Sean PANG
Life of this patent
Estimating the term…
Loading patent…
AssigneeUniversity of Central Florida Research Foundation, Inc.
InventorsGuifang LI, Sean PANG
Estimating the term…
A floating point multiplication unit is disclosed in which numerical components are encoded into properties of harmonic waveforms. A first wave modulator generates a first harmonic wave having a frequency corresponding to an exponent of a first floating point number and an amplitude corresponding to a significand of the first floating point number, thereby representing a first multiplication factor. A second wave modulator similarly generates a second harmonic wave having a frequency corresponding to an exponent of a second floating point number and an amplitude corresponding to a significand of the second floating point number, thereby representing a second multiplication factor. The two harmonic waves are combined through interference, such that the resulting waveform inherently produces a multiplication product corresponding to the first and second floating point numbers. This approach enables efficient computation by leveraging wave-based encoding and interference to perform floating point multiplication.
A floating point number multiplication unit comprising:
A photonic unit for floating point vector multiplication comprising:
Dependencies are inferred from the claim text.