1Michael Wong Distinguished Engineer ● Chair of SYCL Heterogeneous Programming Language ● ISO C++ Directions Group past Chair ● Past CEO OpenMP ● ISOCPP.org Director, VP http://isocpp.org/wiki/ leading team developing HIP & CUDA backends for DPC++ Background in C++ programming models for heterogeneous systems Worked on ComputeCpp (SYCL) since its inception Contributor to the Khronos SYCL standard AI/Tensor HW Other BackendsSYCL 2020 is here! Open Standard for Single Source C++ Parallel Heterogeneous Programming SYCL 2020 is released after 3 years of intense work Significant adoption in Embedded

Sender Patterns to Wrangle Sender Patterns to Wrangle Concurrency in Embedded Devices Concurrency in Embedded Devices Michael Caisse Michael Caisse michael.caisse@intel.com michael.caisse@intel.com

Go 1.4 runtime Gopher China 2015 1. Memory Allocator 2. Garbage Collector 3. Goroutine Scheduler 1. Memory Allocator 内存分配器 base on tcmalloc. 基于成熟方案，性能优秀。随着版本升级， 针对性改进，以期与垃圾回收器更好协作。 核心：自主管理，缓存复用，无锁分配。 阈值触发，并行标记，并发清理。 定期强制回收，释放物理内存。 版本升级，垃圾回收效率总是核心问题。 gogc. 阈值检查，或强制回收。 malloc next_gc 0 gogc runtime.gc() stop start mark sweep stop start mark sweep 0 2 2 1 forcegc 2m 1 mark. 暂停用户逻辑，并行标记。 scheduler thread processor goroutine max. 系统限制，允许调整。 runtime.GOMAXPROCS 调整 P 数量，会导致 G 任务队列重新分布。 M G P scheduler max = 10000 max = 256 runtime/debug.SetMaxThreads 超出限制，会导致进程崩溃。 newproc. 创建新并发任务。

FPGAs AMD GPUs Any CPU SYCL enables Khronos to influence ISO C++ to (eventually) support heterogeneous compute SYCL, OpenCL and SPIR-V, as open industry standards, enable flexible integration and more) Any CPU Experimental SYCL enables Khronos to influence ISO C++ to (eventually) support heterogeneous compute SYCL, OpenCL and SPIR-V, as open industry standards, enable flexible integration and EXPLORERhttps://godbolt.org/z/jdhKr7e5rExpressiveness and simplicity for heterogeneous programming in modern C++ New Features Unified Shared Memory | Parallel Reductions | Subgroup Operations | Class template

Performance improvements of more than 500x overthe state-of-the-art Compilertakes in standard Unified Robot Description Format (URDF) files and generates optimized code Setup data structure to optimize

Leveraging the power of C++ for efficient machine learning on embedded devices Adrian Stanciu adrian.stanciu.pub@gmail.com CppCon, 2023 1 / 50About me ◮ I am a software engineer from Romania ◮ I have predictions ◮ Applications: ◮ Computer vision ◮ Medicine ◮ Search engines 6 / 50Embedded devices ◮ Computing devices designed to perform specific tasks within larger systems ◮ Applications: ◮ Consumer power consumption ◮ May have real-time performance constraints 7 / 50Machine learning on embedded devices ◮ Alternative to cloud-based machine learning ◮ Advantages: ◮ Real-time processing ◮ Low latency

Rust 异步 Runtime 的兼容层 施继成 @ DatenLord Introduce what’s rust async runtime # Rust async runtime Analyze the reason of runtime isolation # Async runtime binding # Compatible layer 1 Create a wheel 2 3 # Rust async runtime 1 Light-weight task • Language and compiler define tasks • How to run it? • When to run it? • How does it deal with the I/O? Rust async runtime Runtime responsibilities it’s multi-thread model Rust async runtime Available Runtimes • Tokio • Async-std • Smol • Monoio Rust async runtime # Async runtime binding 2 Which runtime to choose ? • More adopters • Rich

Designing an ultra low-overhead multithreading runtime for Nim Mamy Ratsimbazafy mamy@numforge.co Weave https://github.com/mratsim/weave Hello! I am Mamy Ratsimbazafy During the day blockchain/Ethereum multithreading: definitions and use-cases ◇ Parallel APIs ◇ Sources of overhead and runtime design ◇ Minimum viable runtime plan in a weekend 4 Understanding the design space Concurrency vs parallelism hardware threads The same distinctions can be done at a multithreaded language or multithreading runtime level. The problem 8 How to schedule M tasks on N hardware threads? Latency vs Throughput

Real-Time Unified Data Layers: A New Era for Scalable Analytics, Search, and AI v 1.1Table of Contents Introduction 1. The Interconnection of Analytics, Search, and AI 2. What is a Real-Time Unified Data Data Layer? 3. Why Do You Need a Real-Time Unified Data Layer? 4. 5.CrateDB: A Modern Real-Time Unified Data Layer1. Introduction Data teams are facing more challenges than ever. As applications generate and architecture teams must rethink traditional data infrastructures. The future lies in Real-Time Unified Data Layers—platforms that seamlessly support analytics, search, and AI workloads at scale. These

Go runtime related problems in TiDB production environment About me ● Arthur Mao(毛康力), Senior Engineer@PingCAP ● TiDB core developer (top3 contributor) ● GitBook about golang internals (@tiancaiamao) IO is ready => goroutine wake up == 4.3ms ○ Sometime even 10ms+ latency here! ○ The time spend on runtime schedule is not negligible ● When CPU is overload, which goroutine should be given priority? Analysis longer to be scheduled ● The runtime scheduling does not consider priority ● CPU dense workload could affect IO latency Conclusion Part II - Memory control ● Go Runtime ○ Allocated from OS (mmaped)