Explore the future of accelerator programming in C++ through Sebastian Schaetz's 46-minute conference talk from C++Now 2014. Delve into topics such as accelerator hardware models, vendor-specific considerations, and programming techniques for accelerators. Learn about concurrency, memory management strategies, and computation limitations in accelerator programming. Discover parallel primitives and fancy iterators, and examine various approaches to writing accelerator functions using C++AMP, HSL, and VexCL. Investigate kernel space exploration with Obsidian and Halide. Gain valuable insights into the evolving landscape of accelerator programming and its implications for C++ developers.
Overview
Syllabus
Intro
About Me
Accelerator Programming
Accelerator Hardware Model
Accelerators and Vendors
Programming Accelerators
Concurrency
Memory Management: Explicit (cont.)
Memory Management Convenient
Memory Management Reversed
Computation Limitations
Parallel Primitives: Fancy Iterators
Writing Accelerator Functions (C++AMP)
Writing Accelerator Functions (HSL)
Writing Accelerator Functions (VexCL)
Kernel Space Exploration (Obsidian)
Kernel Space Exploration (Halide)
Conclusion
Taught by
CppNow
