Wednesday, September 23, 2026A blog and newsletter on Autonomous performance engineering and the search-measure-prove-keep loop and Software headroom.

Performance Optimization

CPU Cache Optimization for Hot Data Structures in C++

Layout decisions rippling down from the 64-byte cache line yield order-of-magnitude speedups.

By Dariusz Kowalenko

Cover illustration for “CPU Cache Optimization for Hot Data Structures in C++”

Profiling & Observability

Lock Contention Profiling in Multithreaded C++ Services

Standard CPU profilers miss lock contention because they can't see blocked threads.

10 min read

Benchmarking

Microbenchmark Validity Problems in Modern CPUs

Microbenchmark numbers often measure CPU state rather than actual code performance.

10 min read