Sparse Look-Up Table for High Radix Division
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing division algorithms, such as SRT division, face challenges with large and time-consuming look-up tables (LUTs) that become even more prohibitive as the radix increases, leading to increased circuit size and processing time.
Innovation Solution
The implementation of a sparse look-up table (LUT) generated using a personality key, which reduces the size of the LUT by exploiting symmetries and folding it, allowing for a more efficient division process by iteratively computing current remainders and portions of the result until a predefined precision is reached.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If a look-up table is used to determine quotient digits in division operations, then the division speed is improved, but the circuit size and processing time increase significantly as the radix increases
Solution Approach 1:
The patent segments the division operation into multiple iterations, where each iteration processes a portion of the quotient digits. Instead of determining all quotient digits simultaneously through a large look-up table, the algorithm divides the problem into smaller steps, determining one or a few quotient digits per iteration. This segmentation allows the use of smaller, more manageable look-up tables while maintaining division speed through iterative processing.
Solution Approach 2:
The patent implements partial action by determining only a portion of the quotient digits in each iteration rather than all digits at once. The algorithm performs partial division operations iteratively, where each iteration computes a subset of the final quotient. This approach reduces the complexity of individual look-up table operations while achieving the complete division result through multiple passes.
2Productivity
If the radix is increased to reduce the number of iterations, then the division efficiency is improved, but the look-up table size increases greatly
Solution Approach 1:
The patent segments both the radix processing and the look-up table structure. Instead of using a single large look-up table for high radix operations, the algorithm segments the radix processing into multiple iterations with smaller look-up tables. Each iteration handles a portion of the radix processing, effectively dividing the large table requirement into multiple smaller tables that are processed sequentially.
Solution Approach 2:
The patent introduces a time dimension to the division process by using iterative computation. Instead of resolving the radix-size tradeoff in a single spatial dimension (look-up table size), the algorithm distributes the computation across multiple time steps (iterations). This dimensional transformation allows the system to use smaller look-up tables at each step while achieving high radix division efficiency through the accumulation of results across iterations.
Data Source
AI summary
According to one general aspect, an apparatus may include a key selector configured to select a personality key from at least a portion of an input value. The apparatus may include a sparse look-up table configured to determine a portion of a result based upon the personality key and a prior remainder. The apparatus may include an adder configured to compute a current remainder based upon, at least, the input value, and the portion of the result. The apparatus may be configured to iteratively compute current remainders and portions of the result until either the current remainder is zero, or a predefined level of precision is reached.


