Buffer View Integration for Efficient Tensor Buffer Access
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Efficient data access presents a challenge for reconfigurable coarse-grained computing systems.
Innovation Solution
A method and system for integrating buffer views into buffer access operations in a reconfigurable computing environment by detecting buffer allocation statements, modifying tensor indexing expressions, and allocating buffers accordingly, utilizing an allocation statement detector and modifier to optimize buffer access.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If buffer views are integrated into buffer access operations in reconfigurable computing systems, then data access efficiency is improved, but device complexity increases
Solution Approach 1:
The patent merges buffer view management directly into the buffer access operations by integrating view allocation, buffer allocation, and access operations into a unified mechanism. This combination eliminates separate view management overhead and streamlines the data access path, improving efficiency while managing complexity through integration rather than addition of separate components
Solution Approach 2:
The buffer access operations are designed to handle multiple functions simultaneously - buffer allocation, view creation, and data access - through a single unified operation. This multi-functionality reduces the number of separate operations needed and improves data access efficiency without proportionally increasing device complexity
2Manufacturing precision
If tensor indexing expressions are modified to include buffer view parameters, then buffer allocation precision is improved, but instruction stream complexity increases
Solution Approach 1:
Buffer views are allocated and configured before buffer access operations execute. The view allocation statement is processed in advance to establish the buffer view parameters, which are then directly used during buffer access without requiring complex runtime computation. This preliminary action improves allocation precision while keeping the execution-phase instruction stream relatively simple
Solution Approach 2:
The modified tensor indexing expressions automatically incorporate buffer view parameters through the integrated buffer access operation mechanism. The system self-manages the integration of view parameters into indexing expressions without requiring external complex processing, improving precision while maintaining instruction stream manageability
Data Source
AI summary
A method and system for integrating buffer views into buffer access operations in reconfigurable computing environments. The method includes detecting, in a buffer allocation statement comprising a tensor indexing expression, a buffer view indicator and one or more buffer view parameters. The buffer view parameters are lowered into the tensor indexing expression, according to the buffer view indicator, to produce a modified tensor indexing expression. The buffer view indicator is removed from a buffer allocation statement to produce a modified buffer allocation statement and allocating a buffer according to the modified buffer allocation statement. The system implements the described method and further includes a non-transitory computer readable medium for executing the disclosed method.