map[Code, Performance] myMCU{?}
summary
map[Code, Performance] myMCU{?} argues that specifics beat slogans: measure cycles on ARMv7M using the DWT cycle counter, capture with J-Link + OpenOCD, and run code on real boards via be bench.com. Concrete findings: optimized vs nonoptimized builds change a tiny function by ~4×, software floating‑point emulation is ≈6× slower and input‑dependent, CRC via a lookup table is ~8× faster than direct, caches and flash latency make high‑frequency runs behave very differently, and on‑chip hardware CRC wins and returns correct results after careful tuning.
AI-written summary · about
chapters (11)
- 0:00Introduction
- 3:49We dont know
- 5:07Consequences
- 6:20Observations
- 7:26Architecture
- 9:32Basic Optimization
- 11:14Pipeline
- 12:55Lowlevel assembly
- 14:59CSE calculation
- 16:20Demo
- 24:04Conclusion
related talks
01h 05mPerformance Optimization in Software Development - Being Friendly to Your Hardware
05mLightning Talk: Micro C++ Benchmarks & Murphy's Law
01h 06mPerformance engineering: being friendly to your hardware
59mCounting Nanoseconds Microbenchmarking C++ Code
01h 04mDesign for Performance
42mPractical Optimizations
58mOptimization Tips - Mo' Hustle Mo' Problems
51mDesigning for Performance