Polar Decoder Partial Sum Scheduling for Lower Latency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current polar code decoding methods require significant time and hardware resources due to the need for extensive partial sum updates and traversals in Data Dependency Graphs, which increases decoding latency and power consumption.
Innovation Solution
The method involves computing and updating partial sums during the decoding of subsequent bits in a received word, traversing nodes in a reverse order relative to their indices in the Data Dependency Graph, and combining lower-order partial sums in a reverse order to reduce latency and resource usage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If partial sum updates are performed extensively during polar code decoding, then decoding accuracy is improved, but decoding latency and power consumption increase
Solution Approach 1:
The patent performs preliminary partial sum computations during the decoding of previous bits, so that when a subsequent bit is being decoded, the required partial sums are already computed and available. This eliminates the need for extensive real-time partial sum updates during the critical decoding path, thereby reducing decoding latency while maintaining accuracy
Solution Approach 2:
The patent dynamically adjusts the computation timing of partial sums, performing them in advance during idle periods in the decoding pipeline rather than uniformly during every decoding step. This dynamic scheduling optimizes the balance between computation completeness and decoding speed
2Ease of operation
If nodes in Data Dependency Graph are traversed in conventional order, then decoding follows standard procedure, but hardware resource usage and power consumption increase
Solution Approach 1:
The patent inverts the conventional traversal order of nodes in the Data Dependency Graph by processing them in reverse order of their indices. This reversal allows partial sum computations to be performed more efficiently during the decoding of subsequent bits, reducing the need for separate update operations and thereby lowering power consumption
3Ease of manufacture
If lower-order partial sums are combined in conventional order, then computation follows standard algorithm, but decoding latency increases
Solution Approach 1:
The patent combines lower-order partial sums in reverse order of their indices rather than in the conventional sequential order. This reversed combination order allows the computation to be performed more efficiently during the decoding of subsequent bits, reducing the critical path delay and overall decoding latency
4Reliability
If extensive partial sum updates are performed, then decoding completeness is ensured, but temporary storage requirements increase
Solution Approach 1:
The patent performs partial sum computations in advance during the decoding of previous bits, so that the results are already available when needed for subsequent bit decoding. This preliminary computation eliminates the need to store extensive intermediate partial sum values in temporary memory, reducing storage requirements while ensuring decoding completeness
Data Source
AI summary
Bits in a received word that is based on a codeword of a polar code are decoded to generate decoded bits. A lower-order partial sum is updated based on the decoded bits, and a higher-order partial sum based on the lower-order partial sum is computed. The higher-order partial sum computation is a live computation performed during decoding of a subsequent bit in the received word in some embodiments. In decoding the subsequent bit, nodes in a Data Dependency Graph (DDG) of the polar code may be traversed in a reverse order relative to node indices of at least some of the nodes in the DDG. A reverse order may also be applied to partial sum computations, to combine multiple lower-order partial sums that are based on previously decoded bits according to a reverse order relative to an order in which at least some of the previously decoded bits were decoded.