Introduction to RISC-V in gem5
summary
Hands-on use of gem5 Standard Library to build a RISC-V board inside a Docker prebuilt environment, run GMEM benchmark with RVV, and collect stats, traces and config.ini outputs. Experiments vary RVV VLEN 128→256 and memory models (DDR4 vs HBM2 mentioned) and show doubling VLEN yields almost 2× runtime speed and ~50% instruction reduction for GEMM here. Workflow uses parse scripts, a Konata pipeline trace viewer, and region-of-interest fast‑forwarding to keep multi-hour simulations feasible.
AI-written summary · about
chapters (12)
- 0:00Introduction to RISC-V in gem5
- 2:15What is gem5?
- 8:20Why use gem5?
- 14:24gem5 software architecture
- 17:12System instantiation with gem5
- 21:48Running simulations in gem5
- 26:06Analyzing simulation results
- 28:35Hands-on demo setup
- 34:21Case study: Vector width exploration
- 46:18Reviewing simulation results
- 50:40Pipeline visualization with Konata
- 55:21Wrap-up and Q&A
