Real-Time Lossless Compression of SoC Trace Data

Nowadays, with the increasing complexity of System-on-Chip (SoC), traditional debugging approaches are not enough in multi-core architecture systems. Hardware tracing becomes necessary for performance analysis in these systems. The problem is that the size of collected trace data through hardware-based tracing techniques is usually extremely large due to the increasing complexity of System-on-Chips. Hence on-chip trace compression performed in hardware is needed to reduce the amount of transferred or stored data. In this dissertation, the feasibility of different types of lossless data compression algorithms in hardware implementation are investigated and examined. A lossless data compression algorithm LZ77 is selected, analyzed, and optimized to Nexus traces data. In order to meet the hardware cost and compression performances requirements for the real-time compression, an optimized LZ77 compression algorithm is proposed based on the characteristics of Nexus trace data. This thesis presents a hardware implementation of LZ77 encoder described in Very High Speed Integrated Circuit Hardware Description Language (VHDL). Test results demonstrate that the compression speed can achieve16 bits/clock cycle and the average compression ratio is 1.35 for the minimal hardware cost case, which is a suitable trade-off between the hardware cost and the compression performances effectively