Reconfigurable Processor Defect Repair via Virtual Addressing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Reconfigurable processors, such as coarse-grain reconfigurable arrays (CGRAs), face challenges in manufacturing yield due to defects, which render some elements unusable, and existing technologies lack effective methods to repair or bypass these defects during operation.
Innovation Solution
A spatially reconfigurable array architecture with spare elements and a parameter store that tags unusable elements, allowing for the reconfiguration of unit configuration files and interconnects to reroute data around defective units, using virtual addressing and repair control signals to maintain functionality without recompiling the entire configuration file.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If manufacturing processes are used to produce reconfigurable processors, then production volume increases, but manufacturing defects occur that render elements unusable
Solution Approach 1:
The patent applies preliminary action by tagging unusable elements during manufacturing testing before the device is deployed. The parameter store pre-identifies defective configurable units, interconnects, or switches during factory testing, so that when the device is later configured for use, the configuration tools can avoid these pre-tagged elements and redirect resources accordingly, preventing field failures.
Solution Approach 2:
The patent changes parameters by dynamically modifying the configuration parameters (placement and routing) based on the tagged unusable elements. When defects are detected, the system alters the configuration file to remap logical addresses to different physical locations, change routing paths around defective interconnects, and reassign resources, thereby adapting the device's operational parameters to work around manufacturing defects.
2Reliability
If configuration files are modified to repair defects, then device functionality is restored, but configuration complexity increases
Solution Approach 1:
The patent introduces an intermediary layer between the application and the physical hardware resources. The parameter store acts as an intermediary that translates logical resource addresses into physical locations, automatically routing around defective elements. This intermediary abstraction layer handles the complexity of defect management transparently, so application developers don't need to directly manage the complex remapping and rerouting operations.
Solution Approach 2:
The patent uses copying by creating alternative mappings in the configuration file that replicate the original logical-to-physical mapping but route through different physical paths. Instead of fundamentally redesigning the configuration system, it creates copied versions of the configuration with modified physical addresses and routing tables that bypass defective elements, maintaining the same logical interface while changing the physical implementation.
3Ease of repair
If spare elements are added to the array, then repair capability improves, but device area increases
Solution Approach 1:
The patent applies universality by making configurable units serve multiple functions: they act as both functional processing elements and as spare replacement elements. The same physical array of configurable units, interconnects, and switches is used for normal operation and for defect repair, depending on the configuration. This eliminates the need for separate spare elements, as any functional unit can be remapped to replace a defective one through the parameter store's address translation.
Solution Approach 2:
The patent changes the dimension of the solution by moving from a spatial approach (adding physical spare elements in the array) to a logical/address space approach. Instead of adding elements in the physical dimension, it creates redundancy in the logical addressing dimension through the parameter store, which can map multiple logical addresses to the same physical resource or remap addresses to alternative locations, providing repair capability without increasing physical area.
4Reliability
If routing is changed to bypass defective elements, then operational reliability improves, but communication overhead increases
Solution Approach 1:
The patent applies preliminary action by pre-calculating and storing the optimal remapped routing paths in the parameter store during configuration loading. Instead of dynamically computing new routes during operation, the configuration tools use the tagged unusable elements information to pre-determine alternative paths and encode them in the configuration file, so that during execution, data simply follows the pre-established routes without real-time overhead.
Data Source
AI summary
A device architecture includes a spatially reconfigurable array of processors, such as configurable units of a CGRA, having spare elements, and a parameter store on the device which stores parameters that tag one or more elements as unusable. Technologies are described which change the pattern of placement of configuration data, in dependence on the tagged elements. As a result, a spatially reconfigurable array having unusable elements can be repaired.