Custom Memory Allocator

Custom Memory Allocator explores how allocator strategy affects throughput, overhead, and scaling under realistic workloads. It is a low-level systems project built to compare tradeoffs rather than chase a single synthetic benchmark.

What it includes

• Four allocation strategies: best-fit, fixed-size, pool, and segregated
• Thread-safe access paths and AVX2-aware alignment work
• Reproducible benchmarks against system malloc, jemalloc, and tcmalloc
• CSV outputs and automated plots for comparing throughput, scaling, and overhead ratio

Why it matters

Allocator behavior becomes visible only when workloads are measured carefully. This project turns allocator design into something inspectable and repeatable, especially for inference-style and RL-adjacent workloads where memory patterns matter.

Code