Godkända
Real-Time Lossless Compression of SoC Trace Data
Jing Zhang (2011)
Start
2015-01-19
Presentation
2015-09-28
Plats:
Avslutat:
2015-12-17
Examensrapport:
Sammanfattning
Today’s SoCs has a high level of complexity and with a multi-core architecture the debugging of a software or hardware problem is a daunting task. In order to further the ability of the engineer to proceed with this task in an efficient way, debugging features is available in the SoC. One of these features is the ability to trace on-chip what is happening in real-time and then off-load this information for background post processing. The problem is that tracing data from today’s SoCs tend to be very large. For example, collecting just one byte of information per executed instruction generates approximately gigabytes of data per second on a high-end multi-core SoC. Thus, the tracing process must rely on very large on-chip buffers to store the traces of large program segments and offline storing all the huge amount of interested traced data can be both a real challenge. This thesis project aims to investigate the state-of-the-art compression algorithms and coding schemes, implemented as ANSI C code to evaluated for complexity, operator requirements, memory footprint and run-time, and also to implement chosen compression algorithms on one of Ericsson’s internal developed cores, investigate and compare real compression performance. As a final step, it is needed to implement one of the chosen compression solution in RTL and shown to be able to perform in real time.
Handledare: Lars Johan Fritz (Ericsson) och Erik Larsson (EIT) och Liang Liu (EIT)
Examinator: Peter Nilsson (EIT)