Understanding Allocator Impact on Runtime Performance in C++
summary
Claim: using a monotonic buffer resource with a local buffer reduces execution time by cutting memory loads and resisting heap diffusion. Evidence comes from a custom benchmark framework AWPBM plus hardware counters read via libpfm and Linux perf and explored with Intel VTune. Experiments include a deterministic heap‑littering simulation, measurements of execution time, page faults and cache events, and show the PMR monotonic variants cut memory accesses and outperform the std::set baseline.
AI-written summary · about
chapters (12)
- 0:00Introduction and motivation
- 1:44Scope and limitations
- 3:33Allocator performance basics
- 4:33Mechanisms of performance gain
- 7:15Understanding allocation diffusion
- 11:06Methodology for analysis
- 13:59Tools and benchmarking
- 23:42Overcoming tool limitations
- 28:46Case study one
- 36:56Simulating heap littering
- 44:55Conclusion and takeaways
- 47:10Q&A session
related talks
01h 08mLocal (Arena) Allocators Part II
58mLocal (Arena) Memory Allocators Part 1
49mMaking Allocators Work, Part I
01h 41mLocal (arena) Memory Allocators
01h 10mBasic usage of PMRs for better performance
36mMemory Why You Need to Care
27m“From security to performance to GPU programming..."
39mMemory Efficient Java