Memory Access Layout for Column Reading in Row-Oriented DRAM
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for decoding product codes in DRAM require excessive row changes when reading data in a column direction, leading to inefficient memory access due to the mismatch between the storage format of two-dimensional arrays and the DRAM's row-oriented access.
Innovation Solution
A method that reads multiple data sets in a single pass from DRAM, computes column syndromes, and stores remaining data sets in SRAM for sequential processing, reducing the number of row changes and clock cycles required.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If data is read sequentially from sequential addresses in a single pass, then the reading operation is simple and fast, but column syndromes cannot be computed because data is stored in row direction while product codes require column direction access
Solution Approach 1:
The patent segments the two-dimensional array data into multiple one-dimensional arrays, where each one-dimensional array corresponds to a row in the original two-dimensional array. This segmentation allows the system to read data in row direction (matching DRAM storage) while still enabling column syndrome computation by processing segmented data collectively
Solution Approach 2:
The patent transforms the two-dimensional array access problem into a one-dimensional array access solution by storing row segments in a one-dimensional array structure. This dimensionality change allows sequential reading from sequential addresses while maintaining the ability to compute column syndromes through the segmented structure
2Reliability
If multiple row changes are performed to read data in column direction, then column syndromes can be computed, but the number of clock cycles increases significantly
Solution Approach 1:
The patent performs preliminary action by pre-segmenting the two-dimensional array into multiple one-dimensional arrays before the reading operation. This pre-segmentation allows the system to read all required row data in a single pass without requiring multiple row changes during the actual reading process, thereby reducing clock cycles
Solution Approach 2:
The patent ensures continuity of useful action by reading all required data in a single continuous pass through sequential addresses. The segmented one-dimensional array structure allows the reading operation to proceed continuously without interruption for row changes, maintaining high efficiency while enabling column syndrome computation
Data Source
AI summary
A two-dimensional array is stored in a first storage memory. A data accessing direction of the first storage memory is in a row direction. A method for reading data in the two-dimensional array in a column direction contains reading a plurality of data sets in the array from the first storage memory; performing a calculating operation on a first data set of the plurality of data sets; storing remaining data sets of the plurality of data sets into a second storage memory; and sequentially reading and applying the calculating operation on the remaining data sets stored in the second storage memory.