Self-Awareness in C++ Code

on YouTube ↗42m

" summary

Built a production memory profiler that instruments allocations inside the Bloomberg Terminal by intercepting heap calls via IAT patching, calling kernel heap APIs and using CaptureStackBackTrace to produce unique call‑stack hashes, per‑pointer maps and a tree histogram UI. Uses thread_local guards to avoid reentrancy, tags allocations with pool/context info and stacks counting allocators (PMR style) to separate pool‑held vs in‑use memory. Finds concrete bugs (example: a logger pool holding huge strings that turns a 500 MB component into 800 MB), runs inside the app with low overhead and produces actionable hotspots.

AI-written summary · about

" chapters (32)

1. 0:00Intro
2. 3:45Tech At Bloomberg
3. 5:36The Bloomberg Terminal
4. 7:48We care about performance
5. 9:21Memory Leaks? In 2018??
6. 10:33Memory Usage: A Case Study
7. 11:51Usual Strategies
8. 13:33Building a memory profiler
9. 14:20Where is malloc?
10. 15:02Strategy #1: Intercepting malloc using the allocator shim
11. 17:36Strategy #2: Intercepting malloc using Import Address Table redirection
12. 21:11Capturing the traceback: helper func
13. 21:57Step 3: Build a callstack / size histogram
14. 22:52Data structures
15. 24:36Gotcha #1: recursive malloc
16. 24:56Gotcha #1 fix: recursion protection
17. 25:12RAll recursion protection
18. 25:48Gotcha #2 fix: Thread-local storage
19. 26:12Instrumentation walkthrough: 2
20. 27:10Instrumentation walkthrough: example
21. 27:37Tree representation
22. 27:57Building the tree
23. 28:12Hierarchical analysis
24. 29:03Graphical analysis
25. 29:42Example 2: Simple leak
26. 30:28Remember our Case Study
27. 30:55Case Study: normal run (no errors)
28. 31:09Case Study: with some errors
29. 31:43Case Study: The logging?
30. 34:46Combining callstacks and instance data
31. 35:19Allocation context
32. 38:44Mixed-mode systems

" related talks

• 27m“From security to performance to GPU programming..."Sergey Zubkov · CppCon 2017
• 39mMemory Efficient JavaKirk Pepperdine · Devnexus 2020
• 51mUnderstanding Allocator Impact on Runtime Performance in C++Parsa Amini · CppCon 2022
• 04m“Profiling Rust Memory Usage with jemalloc”Brennan Vincent · RustConf 2025
• 53mPlugging Memory Leaks With The Chrome Dev ToolsBen Dilts · UtahJS Conf 2017
• 21mMemory Efficient JavaKirk Pepperdine · GOTO Chicago 2020
• 58mC++ Performance and OptimisationHubert Matthews · NDC Oslo 2017
• 58mLocal (Arena) Memory Allocators Part 1John Lakos · Meeting C++ 2017