Tachyon: Python 3.15's sampling profiler is faster...
summary
Tachyon ships in Python 3.15 and samples at the microsecond sampling level by reading memory from live interpreters with process_vm_readv/writev. It reduces syscalls via frame-chunk copying, caches code objects with generation checks to avoid stale pointers, and writes a last-frame slot so many samples require zero copies—enabling effective rates beyond the syscall limit (authors claim 1M+ samples/sec in practice for Python stacks). Reporters include flamegraphs, heatmaps, opcode and async-aware views; Kronon handles native frames (~200k/s).
AI-written summary · about
chapters (6)
- 0:00Introduction to Tachyon
- 1:41How sampling profilers work
- 7:29Tachyon optimization tricks
- 15:26Using Tachyon features
- 25:31Limitations and future tools
- 27:36Q&A session
related talks
32mProfiling at the speed of light
37mTuning Python applications can dramatically increase performance
33mA low latency deepdive of Python with Cython
01h 03mA tour of CPython's runtime
21mFrom PyPerf to py-spy - Everything You Need to Know About Python Profilers
11mPync: Function-Level Incremental Execution for Python Scripts
01h 12mFurious & Fast Python 7: Writing Fast Python Code
30mHow to Tune your Python Analysis Pipeline: A Profiler Guide