What Your Allocator Is Doing to Your C++ Program?

Dynamic memory allocation is often treated as an implementation detail in C++ programs. Yet in practice, allocator behaviour strongly influences performance, scalability, memory usage, and even correctness. Seemingly unrelated symptoms -latency spikes, memory bloat, allocator contention, or hard-to-reproduce crashes -are frequently rooted in allocation patterns rather than algorithms.

Through this talk we'll examine what actually happens when C++ code allocates memory at runtime. We'll explore how common allocation strategies interact with object lifetime, threading, and cache locality, and how allocator behavior can amplify design decisions made at the C++ level.

We discuss how patterns such as short-lived allocations, shared ownership, and hidden allocation boundaries can lead to fragmentation, contention, and unpredictable performance.

Attendees will leave with a clearer mental model of allocator behavior and concrete guidance for designing C++ systems that are more predictable, scalable, and easier to reason about under real-world workloads.