Coprocessor Distributed Register Grid
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing coprocessor designs face inefficiencies in operand and result storage, particularly in distributing memory across processing elements, leading to increased power consumption and area requirements due to lengthy wiring connections.
Innovation Solution
A coprocessor architecture featuring a grid of processing elements with a distributed Z memory that allows each processing element to update a portion of the result memory, reducing wiring complexity and enhancing storage efficiency by localizing memory access within the coprocessor.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If memory is distributed across processing elements, then memory access latency is reduced, but wiring complexity and area requirements increase
Solution Approach 1:
The result memory is segmented into multiple portions, with each processing element having direct access to its local portion. This segmentation allows parallel access to different memory portions by different processing elements, reducing overall memory access latency while distributing the wiring burden across multiple smaller connections rather than one large centralized connection.
2Loss of time
If memory is distributed across processing elements, then memory access latency is reduced, but area requirements increase
Solution Approach 1:
Each processing element is given local access to a specific portion of the result memory, creating local quality in the memory architecture. This allows each processing element to quickly access its designated memory portion without competing for bandwidth with other elements, reducing latency while optimizing the area distribution across the coprocessor.
3Device complexity
If centralized memory is used, then wiring complexity is reduced, but power consumption increases due to lengthy connections
Solution Approach 1:
The memory architecture transitions from a centralized single-dimension model to a distributed multi-dimension model where each processing element has access to its local memory portion. This dimensional change in memory organization reduces the average distance data must travel, lowering power consumption while maintaining manageable wiring complexity through structured distribution.
Data Source
AI summary
In an embodiment, a coprocessor includes multiple processing elements arranged in a grid of one or more rows and one or more columns. A given processing element includes an arithmetic/logic unit (ALU) circuit configured to perform an ALU operation specified by an instruction executable by the coprocessor, wherein the ALU circuit is configured to produce a result. The given processing element further comprises a first memory coupled to the execute circuit. The first memory is configured to store results generated by the given processing element. The first memory includes a portion of a result memory implemented by the coprocessor, wherein locations in the result memory are specifiable as destination operands of instructions executable by the coprocessor. The portion of the result memory implemented by the first memory is the portion of the result memory that the given processing element is capable of updating.