Speaker: Shivam Kunwar

Audience level: [ Intermediate ]

In this session, we will explore the crucial aspects of multithreaded performance optimization - false sharing and hardware destructive interference. False sharing can significantly impact performance by causing unnecessary cache invalidations, while hardware destructive interference provides a solution by guiding optimal data structure alignment for improved cache efficiency. Join us to understand the causes and consequences of false sharing and delve into the concept of hardware destructive interference. Through real-world examples and performance benchmarks, you will learn practical strategies to mitigate false sharing and leverage hardware destructive interference to unlock maximum performance in your multithreaded applications.

Key Takeaways:

- Identify and mitigate false sharing issues that degrade multithreaded performance

- Understand the concept of hardware destructive interference and its impact on cache efficiency

- Learn techniques to align data structures based on hardware destructive interference for optimal performance in multithreaded applications

- Gain insights from real-world examples and performance benchmarks to apply the concepts effectively

This session is essential for software developers, system architects, and researchers seeking to maximize the performance of their multithreaded applications and overcome cache-related performance bottlenecks.