Processor Error Recovery via State Snapshot and Dynamic Precision Switching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing processors lack efficient methods to manage and recover from excessive error accumulation during approximate computing, which affects accuracy and power consumption.
Innovation Solution
A processor design that includes a snapshot storage mechanism to record its state before performing approximate computations, with an error indicator to detect when errors exceed a bound, allowing the processor to restore its state and re-perform computations with full accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Use of energy by moving object
If approximate computing is used to reduce power consumption, then energy efficiency is improved, but computation accuracy deteriorates
Solution Approach 1:
The processor dynamically switches between approximate computing mode and full accuracy mode based on error accumulation. The error indicator monitors computation accuracy in real-time, and when error exceeds a threshold, the system transitions from approximate mode back to full accuracy mode, creating a dynamic adaptation mechanism that resolves the contradiction between power consumption and accuracy.
Solution Approach 2:
The system changes the computation precision parameter based on error accumulation. Initially, computations are performed with reduced precision (approximate mode) to save power. When the error indicator detects that accumulated error exceeds a threshold, the precision parameter is changed back to full accuracy, allowing the system to optimize power consumption while maintaining accuracy when necessary.
2Use of energy by moving object
If approximate computing is performed without error recovery, then power consumption is reduced, but reliability deteriorates
Solution Approach 1:
The processor takes a snapshot of its state before performing approximate computations. This preliminary action prepares the system for potential error recovery by preserving the pre-computation state, enabling reliable recovery if approximate computing produces excessive error, thus maintaining reliability while allowing power-efficient approximate mode operation.
Solution Approach 2:
The error indicator provides continuous feedback on the accumulated error during approximate computing. When the error exceeds a threshold, this feedback triggers the recovery mechanism that restores the processor state from the snapshot and re-performs computations in full accuracy mode, ensuring reliability while maintaining power efficiency benefits.
3Measurement precision
If full accuracy computing is always used, then computation accuracy is maintained, but power consumption increases
Solution Approach 1:
Instead of always performing full accuracy computing, the system applies partial action by using approximate computing for computations where full accuracy is not critical. The error indicator monitors when accuracy becomes important, and only then does the system switch to full accuracy mode, avoiding the excessive power consumption of always using full precision while maintaining accuracy when necessary.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A processor includes a storage configured to receive a snapshot of a state of the processor prior to performing a set of computations in an approximating manner. The processor also includes an indicator that indicates an amount of error accumulated while the set of computations is performed in the approximating manner. When the processor detects that the amount of error accumulated has exceeded an error bound, the processor is configured to restore the state of the processor to the snapshot from the storage.