Linear Processing Element Array Data Transmission Method
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In data-intensive scientific computations, existing methods either incur high power consumption and latency due to frequent data fetching from memory or suffer from severe storage overhead by duplicating data across multiple storage devices for local access.
Innovation Solution
A processing element array with a linear arrangement of PEs and transmission circuits, where data is shared and scattered across multiple memories, allowing only one PE to hold and deliver data at a time, reducing the need for frequent direction changes in data transmission.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If data is stored locally in multiple storage devices for fast access, then access speed is improved, but storage overhead increases severely
Solution Approach 1:
The patent segments data storage across multiple PEs in a linear array, where each PE stores a portion of the data. This allows any PE to access its local data immediately without duplicating entire datasets across all PEs, reducing storage overhead while maintaining fast access speeds through localized storage and systematic data distribution.
Solution Approach 2:
The patent introduces a temporal dimension to data access by implementing a systematic data exchange mechanism along the linear PE array. Instead of requiring all data to be simultaneously available in all PEs (spatial duplication), the system exchanges data sequentially through the linear array over time, achieving fast access without increasing storage overhead.
2Ease of operation
If data is delivered to neighbors immediately at every timing in linear topology, then data availability is improved, but computation must complete in multiple passes increasing time consumption
Solution Approach 1:
The patent implements preliminary data exchange operations before computation begins. Data is pre-positioned in appropriate PEs along the linear array through systematic exchange protocols, so that when computation starts, all necessary data is already available at the correct locations, eliminating the need for multiple passes and reducing total computation time.
3Productivity
If PEs deliver data to neighbors immediately without local storage, then bandwidth utilization is improved, but computation is blocked when data is not available
Solution Approach 1:
The patent implements dynamic data exchange protocols where PEs adaptively exchange data based on computational needs and data availability. The linear array enables flexible data movement in both directions, allowing PEs to request and receive data from neighbors dynamically without blocking computation, maintaining both high bandwidth utilization and computation continuity.
Data Source
AI summary
A processing element array includes N processing elements (PE) arranged linearly, N≥2, and an operating method of the PE array includes: performing a first data transmission procedure, where an initial value of I is 1 and the first data transmission procedure includes: operating, by an ith PE, according to a first datum stored in itself, and sending the first datum to other PEs for their operations, adding 1 to I when I<N, and performing the first data transmission procedure again, performing a second data transmission procedure when I is equal to N, which includes: operating, by the Jth PE, according to a second datum stored in itself, and sending the second datum to other PEs for their operations, reducing J by 1 when J>1 and the (J−1)th PE has the second datum, and performing the second data transmission procedure again.