Kokkos Tea time with Jolly Chen “C++ Reflection for Data Layout Abstraction” скачать в хорошем качестве

Kokkos Tea time with Jolly Chen “C++ Reflection for Data Layout Abstraction”

Jolly Chen give a talk entitled “C++ Reflection for Data Layout Abstraction” Jolly Chen is a second year PhD student in Computer Science at CERN, affiliated with the University of Twente in The Netherlands. The topic of her PhD is “Optimizing Memory Access patterns through Data Layout Transformation”. Broadly speaking, her research interests are Performance Engineering, GPU Programming, and Performance Modelling. C++ Reflection for Data Layout Abstraction Good performance requires good memory access patterns. A common optimization dilemma is choosing between array-of-structures (AoS) and struct-of-arrays (SoA). However, in C++, changing the data layout between AoS to SoA requires modifying both the data structure declarations and the access syntax. This work is repetitive, time consuming, and can lead to less intuitive code. For example, with AoS, we can have an array of particles and access the third particle’s momentum with the syntax particles[2].x. In contrast, SoA requires the syntax particles.x[2]. To address this dilemma, we want to decouple memory organization from logical data access while preserving the standard C++ access syntax. In this talk, Jolly present different approaches to abstracting SoA as AoS at compile-time, using C++26 reflection and some Code Injection with Token Sequences as proposed by P3294, which she compare with manually defining SoAs. In addition, she share her experience with using C++ reflection.