From Hardware Trace to System Knowledge - Data-intensive Hardware Trace Analysis

摘要

The capture of large amounts of hardware trace data by recent hardware trace units in System-on-a-Chip (SoC) defines the new requirements for hardware trace analysis. Several hundreds of MiB need to be processed in near real-time. The requirements on analyses are further increased with long-term trace capture and increasing frequencies of SoC. Hence, hardware trace analysis has become a data-intensive computing task. The article proposes an approach based on data, functional and pipeline parallelism in combination with data stream processing. As this is a 'work in progress', we will only outline the approach and the preconditions and decisions leading to the approach.