This article explains Linux process virtual address spaces, details the multi‑step internals of the C++ new operator, and analyzes its memory‑fragmentation, time, and space overheads through code examples and benchmarks, then offers practical mitigation techniques such as memory pools and smart pointers.
The article reveals that using C++ new or malloc is far from cost‑free, detailing how Linux virtual address space, system‑call overhead, memory‑header metadata, fragmentation, and page‑fault latency all contribute to hidden performance penalties, and it offers concrete techniques such as stack allocation, memory pools, restrained PImpl usage, and startup pre‑allocation to mitigate these costs.
This article compares the performance of creating objects with the new operator versus Java reflection, explains why reflection incurs a large overhead, and outlines scenarios where each approach is appropriate, helping developers choose the right technique for their code.
This article compares creating Java objects with the new operator and with reflection, presents benchmark code showing a huge performance gap, explains the underlying reasons such as JIT optimization limits and Method#invoke overhead, and outlines practical scenarios for using each approach.
The article explains why allocating an array with new sample[10] yields zero‑initialized memory on Linux but uninitialized memory on Windows, clarifies the difference between default and value initialization in C++, and shows how various new expressions behave with concrete code examples.
