Implementation of a RISC-V E-Trace Encoder for Multi-Core SoCs
Abstract
Debugging timing-critical software requires observability without influencing system behavior. Tracing offers this advantage over traditional instrumentation methods. This paper presents a hardware implementation of an Efficient Trace for RISC-V (E-Trace) encoder, integrated into a multi-core System-on-Chip. Its generator-based design features configurable modules, allowing the implementation of trace encoders with different capabilities and complexities. For multi-core support, the system handles AXI-Stream packet translation with size constraints suitable for DMA operations. When implemented on an FPGA, the E-Trace infrastructure only adds a 5% (LUTs)/6% (FF) resource overhead. ASIC synthesis using SkyWater 130nm PDK shows our compressed implementation requires 13.2 mW power, $54770 \mu m^{2}$ area, and 3518 standard cells. Our encoder achieves more efficient packet compression and undergoes more thorough verification through its testing integration with the official specification test programs than a comparable open source trace encoder.