Tutorial: Performance Tools
summary
A Sparta C++ performance model called Olympia simulator runs retired‑instruction traces produced with the STF library (Drystone trace included), lets users tweak microarchitecture via YAML parameters and architectures, and emits detailed reports, CSV time series and pipeouts for the Argos pipeline visualization suite. Live demo shows ~500–750k instructions/sec on a laptop, and adding a second ALU drops ALU utilization to near zero and raises Dhrystone DMIPS/ MHz from about 2.3 toward a measurable uplift.
AI-written summary · about
chapters (12)
- 0:00Introduction to tutorial
- 1:45Model overview
- 3:11Generating instruction traces
- 5:40Running Olympia simulator
- 8:32Parameter tweaking and architectures
- 16:47Report generation basics
- 21:23Fine-tuning reports
- 27:20Time series analysis
- 32:04Pipeline visualization
- 35:08Bottleneck analysis with dry Stone
- 39:19Future of Olympia model
- 42:17Q&A session
related talks
56mWrite Fast Code Like a Native
01h 04mUnlocking Performance Through Reverse Engineering
01h 00mFastware
17mUnleashing the Power of RISC-V E-Trace with a Highly Efficient Software Decoder
14mSave what must be saved: Secure context switching with Sailor
56mPath Tracing Three Ways: A Study of C++ Style
01h 01mGoing Nowhere Faster
22mDemo: From Blueprint to Reality: Navigating SoC Tradeoffs, IP, and Ecosystem