Programmable Look-Up-Table Hardware for Arbitrary Bin Spacing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional look-up-table approaches for piecewise linear approximation are inflexible, as slope and intercept values are hard-coded, limiting their ability to accommodate different values at test-time and requiring fixed bin spacing, which is not suitable for arbitrarily-spaced bins.
Innovation Solution
A programmable look-up-table system using special-purpose digital-compute hardware with programmable memory for slope and intercept values, Bin ID logic with data registers for bin-threshold values, and a Fused-Multiply-Add circuit to handle arbitrarily-spaced bins and perform 16-bit floating-point arithmetic, allowing dynamic configuration of bin spacing and values.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If slope and intercept values are hard-coded into the system, then the implementation is compact, but the system lacks flexibility to accommodate different values at test-time
Solution Approach 1:
The patent transforms the static, hard-coded look-up-table into a dynamic, programmable structure. The memory device stores slope and intercept values that can be loaded and reconfigured at test-time, allowing the system to adapt to different functions and requirements while maintaining a compact implementation through efficient memory utilization.
Solution Approach 2:
The patent enables changing the parameters (slope and intercept values) of the look-up-table from fixed to variable. By using a memory device that can be programmatically loaded with different values, the system can modify its characteristics at test-time to accommodate different continuous functions and approximation requirements.
2Device complexity
If conventional look-up-table approaches are used, then the structure is simple, but the bin spacing must be fixed and cannot accommodate arbitrarily-spaced bins
Solution Approach 1:
The patent introduces dynamic bin-threshold values stored in memory that can be reconfigured to define arbitrarily-spaced bins. This allows the system to transition from fixed, uniform bin spacing to flexible, arbitrary bin spacing by simply loading different threshold values into the memory device, maintaining structural simplicity while gaining adaptability.
Solution Approach 2:
The patent creates a universal look-up-table structure that can handle both uniformly-spaced and arbitrarily-spaced bins through the same hardware architecture. By using programmable memory for storing bin thresholds and function values, the system becomes multi-functional, capable of accommodating different binning strategies without requiring separate dedicated structures.
3Ease of manufacture
If uniformly-spaced bins are used, then the implementation is straightforward, but the accuracy for modeling functions with varying gradients is reduced
Solution Approach 1:
The patent applies local quality by allowing different bin spacing strategies in different regions of the input domain. Through programmable bin-threshold values, the system can create non-uniform bin spacing that provides finer resolution in regions where the function has varying gradients and coarser spacing in regions with more uniform behavior, optimizing accuracy while maintaining implementation simplicity.
Data Source
AI summary
Special-purpose digital-compute hardware for fully-programmable look-up-tables is provided. In one aspect, a system for implementing a continuous function by piecewise linear approximation includes: at least one memory programmatically loaded with an indexed table of slope/intercept values of linear segments along a gradient of the continuous function approximating a plurality of contiguous ranges of the continuous function; at least one Bin ID logic having data registers programmatically loaded with bin-threshold values corresponding to the plurality of contiguous ranges defining a series of arbitrarily-spaced bins; and a Fused-Multiply-Add circuit configured to multiply an incoming data element by a slope value and add an intercept value from the indexed table of slope/intercept values selected based on the bin-threshold values. Comparators in the Bin ID logic can be configured to compare the incoming data-element with the bin-threshold values. A method for implementing a continuous function by piecewise linear approximation is also provided.