Synchronized Trace Timestamps for Multi-Core Fault Localization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In computer systems with multiple arithmetic parts, identifying faults and ascertaining the internal state is difficult due to independent data movement without a host part, making it challenging to locate failures and understand data flow.
Innovation Solution
A computer system with trace parts recording timestamp values based on event detection, synchronized across arithmetic parts, and a host part managing these timestamp values to identify failures and manage data flow.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If arithmetic parts independently move data without going through a host part, then data processing speed is improved, but fault identification becomes difficult
Solution Approach 1:
A trace part is introduced as an intermediary component between the arithmetic parts and the host part. This trace part records trace data including timestamps and arithmetic part identifiers without interfering with the high-speed independent data movement. The recorded trace data serves as evidence that can be analyzed by the host part to identify faults and monitor system state, thus resolving the contradiction between fast independent processing and fault detectability
Solution Approach 2:
The trace part performs preliminary recording of data movement information (timestamps, arithmetic part IDs, data values) as data passes through the arithmetic parts. This preliminary action captures the state of the system during operation, enabling post-event analysis of faults and performance without requiring the host part to intervene in the real-time data flow
2Stability of the object's composition
If timestamp values are synchronized across arithmetic parts with different operating frequencies, then system coordination is improved, but timestamp value accuracy may be compromised
Solution Approach 1:
The patent applies parameter transformation by multiplying timestamp values from arithmetic parts with different operating frequencies by appropriate coefficients. This parameter change converts timestamps from different frequency domains into a unified reference frame, enabling synchronization and comparison while preserving the original timing information's accuracy within each arithmetic part's context
3Difficulty of detecting and measuring
If trace data is recorded for every arithmetic part, then fault identification capability is improved, but system complexity increases
Solution Approach 1:
The trace part is designed as a universal component that can be implemented in each arithmetic part with the same basic functionality. This multi-functionality approach allows standardized trace data recording across all arithmetic parts, simplifying the overall system architecture compared to having different monitoring mechanisms for each part. The unified trace data format enables centralized analysis by the host part
Data Source
AI summary
An embodiment is a computer system which processes input data which includes a plurality of arithmetic parts; and a host part connected to the plurality of arithmetic parts and configured to control the plurality of arithmetic parts, in which the processed data is transferred between the plurality of arithmetic parts, the arithmetic part includes trace parts which record trace data using detection of a predetermined event from the input data as a trigger, the trace data has a timestamp value which is a detection time of the event based on an operating frequency of the arithmetic part, and the timestamp values of the plurality of arithmetic parts are synchronized.