Event Notification Messages for Neural Network Bottleneck Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Complex computer systems, such as neural networks, face difficulties in identifying low-performing portions due to their complexity, making it challenging to optimize performance effectively.
Innovation Solution
The generation of event notification messages that automatically track the set and clear times of events, allowing for synchronization and analysis of data and instruction transportation, without affecting system performance, to identify bottlenecks and improve resource allocation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If complex circuits with multiple processing units are used to improve performance, then processing capability is improved, but difficulty of detecting and measuring performance bottlenecks worsens
Solution Approach 1:
The patent introduces event notification messages as an intermediary mechanism between the complex parallel processing circuits and the performance analysis system. These messages automatically track event indicators (set/clear times) and provide synchronized timing information, enabling external tools to measure performance bottlenecks without directly interfering with the complex internal operations of multiple processing units
Solution Approach 2:
The system implements feedback by generating event notification messages that provide timing information about when events occur in the parallel processing system. This feedback loop allows the system to automatically monitor its own performance, track execution timelines, and identify bottlenecks in specific layers, nodes, or operations without adding manual measurement complexity
2Measurement precision
If event tracking is implemented to identify bottlenecks, then measurement precision is improved, but device complexity worsens
Solution Approach 1:
The system employs self-service by automatically generating event notification messages that track their own internal event indicators without requiring external intervention. The processing units themselves generate the timing information about their operations, eliminating the need for complex external monitoring hardware and reducing overall system complexity while maintaining high measurement precision
3Loss of information
If monitoring mechanisms are added to track performance, then information completeness is improved, but loss of time worsens
Solution Approach 1:
The system applies preliminary action by pre-configuring event notification mechanisms that automatically generate timing information as events occur during normal operation. Event indicators are set and cleared at predetermined points in the processing workflow, and notification messages are generated in advance to capture timing data without requiring post-processing analysis or additional measurement time
Data Source
AI summary
An integrated circuit is configurable to generate a notification message when an indicator of an event used to synchronize the execution of different functional blocks of the integrated circuit changes status. The indicator of the event is cleared when an operation is triggered and is set when the operation completes. The notification message includes a timestamp indicating the time when the indicator of the event changes status. The notification message is used to determine the execution timeline of a set of instructions executed by integrated circuit and to identify bottlenecks in the set of instructions or the integrated circuit.


