fetch_add vs. compare_exchange

Benchmark C++ standard atomic operations fetch_add, fetch_or, and compare_exchange.

Two types of benchmarks:

• refcount — aggressive concurrent usage of shared_ptr, implemented with:
  • std::atomic_compare_exchange_weak()
  • std::atomic_compare_exchange_strong()
  • std::atomic_fetch_sub()
  • baseline: std::shared_ptr
• spinlock — aggressive concurrent usage of a spinlock, implemented with:
  • std::atomic_compare_exchange_weak()
  • std::atomic_compare_exchange_weak() + pause
  • std::atomic_compare_exchange_strong()
  • std::atomic_compare_exchange_strong() + pause
  • std::atomic_fetch_add()
  • std::atomic_fetch_or() + pause
  • baseline: std::mutex

Build:

make

Run benchmark (produces docs/result.csv):

make bench

Set up the Python venv for plotting (one-time):

./venv_install.sh

Show charts on screen:

make show

Render charts to PNG files:

make png

See docs/article.md for the full write-up and analysis.