“Towards Heterogeneous Programming in C++"

on YouTube ↗56m

" summary

SG14 mandates adding heterogeneous support to C++ and the practical proposal centers on SYCL: a single-source C++ model that compiles device lambdas into SPIR or vendor binaries and embeds them in one executable. It uses buffers and accessors so the runtime builds automatic data-dependency graphs and only copies when needed; kernels use parallel_for with explicit ranges. Codeplay ships ComputeCpp (community edition) as a SYCL implementation targeting OpenCL devices and SPIR, demonstrating portability across Intel/AMD hardware while noting that full performance portability still requires device-specific tuning.

AI-written summary · about

" chapters (15)

1. 0:00Introduction
2. 1:37Heterogeneous computing
3. 3:33Historical evolution
4. 7:11Language popularity
5. 8:57Challenges
6. 13:00Offloading techniques
7. 15:26Describing parallelism
8. 17:19Data management
9. 22:29Sycl standard
10. 27:04Sycl features
11. 30:36Sycl code example
12. 34:02Future of C++
13. 40:11Executors and simd
14. 43:01Standardization path
15. 45:55Conclusion and q&a

" related talks

• 01h 32mHeterogeneous Programming in C++ todayMichael S Wong · ACCU 2018
• 01h 07mHeterogeneous Modern C++ with SYCLMichael S Wong, Nevin Liber, Tom Deakin, Gordon Brown · CppCon 2021
• 41mSYCL building blocks for C++ librariesGordon Brown · Meeting C++ 2016
• 01h 01mA Modern C++ Programming Model for GPUs using Khronos SYCLGordon Brown · CppCon 2018
• 01h 01mTake Advantage of All the MIPS - SYCL & C++Michael S Wong, Nevin Liber, Ronan Keryell, Hugh Delaney, Jakub Chlanda · CppCon 2022
• 01h 33mGPU programming with modern C++Michael S Wong · ACCU 2019
• 01h 00mSYCL: Open Your Code to More Hardware ChoicesIgor Vorobtsov · C++ on Sea 2022
• 24mOpen Acceleration for RISC-V – Portability, Performance and Partners...Charles Macfarlane · embedded world 2022