Accelerator Nonlinear Function Mapping for Circuit and Power Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for implementing nonlinear functions in accelerators do not optimally minimize circuitry and power consumption while meeting the required calculation accuracy.
Innovation Solution
A computer-readable recording medium stores a program that processes information to map a nonlinear function to an accelerator, ensuring calculation accuracy, and iteratively selects different mapping methods to satisfy implementation constraints such as circuitry and power consumption.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a nonlinear function is implemented in a hardware accelerator, then calculation accuracy is improved, but the amount of circuitry and power consumption increase
Solution Approach 1:
The patent applies parameter changes by adjusting the number of iterations in the Newton-Raphson method and selecting different polynomial approximation degrees to optimize the balance between calculation accuracy and circuit complexity. By varying these parameters, the system achieves required accuracy while minimizing hardware resources.
Solution Approach 2:
The patent implements partial action by using iterative approximation methods (Newton-Raphson) that can terminate early when sufficient accuracy is achieved, rather than always performing the maximum number of iterations. This allows the circuit to use fewer resources when lower precision is acceptable.
2Measurement precision
If a nonlinear function is implemented in a hardware accelerator, then calculation accuracy is improved, but power consumption increases
Solution Approach 1:
The patent optimizes power consumption by dynamically adjusting operational parameters such as the number of Newton-Raphson iterations and polynomial approximation order. This allows the system to consume less power when high accuracy is not required, while still meeting accuracy constraints when needed.
Solution Approach 2:
The iterative approximation approach enables partial execution of computational steps based on achieved accuracy, allowing the circuit to reduce power consumption by performing fewer iterations when the required precision has already been reached.
3Speed
If mapping is performed using a predetermined mapping method, then mapping speed is improved, but the ability to satisfy implementation constraints deteriorates
Solution Approach 1:
The patent implements a dynamic mapping approach where the system first applies a predetermined mapping method for speed, then evaluates whether implementation constraints are satisfied. If constraints are not met, the system adaptively adjusts the mapping parameters and re-evaluates, creating a flexible two-stage process that balances speed and constraint satisfaction.
Solution Approach 2:
The patent employs feedback by evaluating the results of the predetermined mapping against implementation constraints and using this information to determine whether additional mapping adjustments are needed. This feedback loop ensures that the final mapping satisfies all constraints while maintaining efficiency.
Data Source
AI summary
A non-transitory computer-readable recording medium stores a program for causing a computer to execute a process for processing information which includes: obtaining a first nonlinear function to be mapped, calculation accuracy, and an implementation constraint, which are required to implement a nonlinear function in an accelerator; mapping the first nonlinear function to the accelerator to satisfy the calculation accuracy using a predetermined mapping method; determining whether or not a result of the mapping to the accelerator satisfies the implementation constraint; and when the implementation constraint is not satisfied, repeating the mapping and the determining using a mapping method different from the predetermined mapping method.