Experimental evaluation of techniques for capturing and compressing hardware traces in multicores
Date of Award
Master of Science in Engineering (MSE)
Electrical and Computer Engineering
Rhonda Kay Gaede
B. Earl Wells
Parallel processing (Electronic computers), High performance processors., Embedded computer systems., Debugging in computer science.
Modern embedded systems are indispensable in all aspects of modern life. The increasing complexity of hardware and software stacks and tightening time-to-market deadlines make software development and testing the most critical aspects of system development. To help developers find software bugs faster, modern embedded systems increasingly rely on on-chip resources for debugging and tracing. Unfortunately, cap-turing and streaming all hardware events of interest for program debugging is cost-prohibitive in multicores where tens of processor cores work concurrently at very high speeds. This thesis focuses on capturing control-flow and data traces in multicores. It introduces two new techniques: mcfTRaptor for capturing control-flow traces and mlvCFiat for capturing load data value traces. The effectiveness of the commercial state-of-the-art and the proposed techniques are experimentally evaluated by measur-ing the number of bits needed to be streamed off the chip for both functional and timed traces. The results show that the proposed techniques are very effective, while requiring modest hardware support.
Tewar, Amrish K., "Experimental evaluation of techniques for capturing and compressing hardware traces in multicores" (2015). Theses. 133.