Back to results
US 2026/0203017 A1Published application
Local-Bias Adaptive Drive for Timing Stabilization in Pipelined Mixed-Precision Processing Elements
OwnerKia Silverbrookinventor
Life of this patent
Estimating the term…
Loading patent…
Loading patent…
OwnerKia Silverbrookinventor
Estimating the term…
A processing element includes a mixed precision fused multiply add datapath configured to receive an FP4 weight operand and an FP8 activation operand in a two stage pipeline. A first stage generates a mixed precision product and a second stage accumulates the product. Control circuitry adjusts local device bias within the processing element based on detected switching activity to maintain propagation delay within a predetermined range. A computing device may include an array of such processing elements, each operating with independent bias adaptation. A method of operation includes generating a mixed precision product, detecting switching activity, adjusting local device bias, and accumulating the product while maintaining controlled timing.
A processing element comprising a mixed precision fused multiply add datapath configured to receive an FP4 weight operand and an FP8 activation operand, a first pipeline stage configured to generate a mixed precision product, and control circuitry configured to adjust local device bias within the processing element in response to measured switching activity such that propagation delay through one or more datapath subcircuits is maintained within a predetermined delay range.
A computing device comprising an array of processing elements, each processing element comprising a mixed precision fused multiply add datapath having a first pipeline stage configured to generate a mixed precision product from an FP4 weight operand and an FP8 activation operand, and control circuitry configured to adjust local device bias in response to switching activity within the processing element.
A method of performing mixed precision fused multiply add computation comprising generating a mixed precision product from an FP4 weight operand and an FP8 activation operand in a first pipeline stage, detecting switching activity within a processing element during operation, adjusting local device bias based on the detected switching activity, and accumulating the mixed precision product in a second pipeline stage while maintaining propagation delay within a predetermined range.
Dependencies are inferred from the claim text.