Summary
Jason explores the practical benefits of custom memory allocation strategies in C++, particularly using PMR (Polymorphic Memory Resources). He demonstrates how custom allocators can improve performance through fewer calls to new/delete, reduced thread contention, minimized false sharing, and reduced memory diffusion. Using benchmarks, he shows significant performance improvements (~25%) and reduced runtime variance when using monotonic buffer resources and pool resources with multithreaded applications, especially as thread count increases.
Related C++ Standard Sections
This episode covers topics found in these sections of the C++ standard:
-
[mem.res]69% match -
[mem.poly.allocator.class]49% match -
[default.allocator]45% match -
[mem]40% match