Is Writing Performant Code Too Expensive?

Explore the intricacies of writing performant code in this 48-minute conference talk from Devoxx Poland 2019. Delve into hardware counters, NUMA nodes, and vector registers to understand the possibilities of current hardware and the true cost of scalability. Learn why optimizing code is crucial despite rising developer salaries and time-to-market pressures. Discover how proper algorithms and hardware knowledge can lead to 10-100x speed improvements, bringing significant value to products and clients. Gain insights on memory bandwidth for AMD EPYC and Intel Xeon systems, and understand the impact of branch predictors on performance. Through practical examples and benchmarks, develop a better intuition for identifying optimization opportunities and balancing software development with hardware capabilities.

Introduction
About me
Fads and Fat
Salaries
The Sprint
Bigger server
Titan X Developer
The Reality
The Conference
Database Problems
Efficient Query Plans
MySQL Benchmark
Scalability
Loops
Cost
Opportunity
Refresh
Memory banks
QPI
Topology
Experiment
Libraries
Benchmark
Micro benchmarks
Intel architecture
Reach throughput
Quiz
Branch predictor
Leslie Lamport
Binary operations
Mask operations
Speed of memory
Dome
Summary