Semiconductor Event Profiling With Serializer-DMA Capture
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing software-based event profiling in semiconductor devices is inefficient, as it requires changing software modes, struggles to monitor hardware events, and analyzing system-on-chip operations is difficult due to asynchronous data periods.
Innovation Solution
A semiconductor device and method utilizing hardware configuration with serializers, de-serializers, timers, and DMA to collect and store event data in parallel streams at specific intervals, reducing the need for software intervention and minimizing resource usage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If software-based profiling is used, then event collection is possible, but profiling requires changing software modes and cannot be performed on-the-fly
Solution Approach 1:
The patent replaces software-based event collection with a hardware-based event collection unit that continuously monitors and captures events in real-time without requiring software mode changes. The hardware unit independently collects events and transfers them to memory, enabling on-the-fly profiling while eliminating the need for software intervention or mode switching.
2Measurement precision
If software checks software status, then software state monitoring is possible, but hardware event monitoring becomes difficult
Solution Approach 1:
The patent divides the monitoring system into separate functional units: an event collection unit that monitors multiple hardware events simultaneously, a data transfer unit that moves captured events to memory, and a processing unit that analyzes the events. This segmentation enables precise hardware event monitoring while maintaining the ability to monitor multiple different event types across various software structures.
3Quantity of substance
If multiple events are monitored with different software structures, then comprehensive event data is collected, but data analysis becomes difficult due to different periods
Solution Approach 1:
The patent implements a periodic data transfer mechanism where the data transfer unit moves captured events to memory at regular intervals. This periodic action synchronizes the collection and storage of multiple event types, ensuring that all events are captured with consistent timing intervals. The system uses a common time base or counter to timestamp events, allowing comprehensive event data collection while maintaining uniform data periods that simplify subsequent analysis.
Data Source
AI summary
A semiconductor device includes a first serializer configured to collect at least one event in a first domain to generate a first serial data stream and transmit the first serial data stream periodically at a first period, a first de-serializer configured to receive the first serial data stream to restore the first serial data stream into first parallel data streams, the first parallel data stream encoding first parallel data items, a timer configured to provide a clock signal having a second period, a direct memory access (DMA) configured to capture the first parallel data items periodically at a second period using the clock signal to generate capture data items, and a first memory configured to store the capture data items. The addresses of the first memory at which the respective capture data items are stored are arranged in an order that the respective capture data items are captured.