Heterogeneous System Architecture - Why Should You Care?
summary
Heterogeneous System Architecture claims to let one source code target CPU and accelerators and deliver higher performance with lower power by exposing a platform-level integration of GPUs DSPs and FPGAs. It provides a concrete stack: CLANG/LLVM adapted compilers, an HCC open source C++14 path, runtime hooks that complement OpenCL, and migration paths for CUDA and C++ AMP. The architecture is implemented as a platform and is presented as working across multiple vendors and runtimes.
AI-written summary · about
related talks
24mMapping Sparse Applications to Accelerated Computing Systems
43mA comparison of OpenCL, CUDA, and HIP as compilation targets for a functional array(…)
01h 00mBringing Clang and C++ to GPUs: An Open-Source, CUDA-Compatible GPU C++ Compiler
31mToolchains for People in a Hurry: Latest Features of GCC/GLIBC
22mAlgorithm-Aware Hardware Optimization using E-Graph Rewriting: how should we marry(…)
47mOpen AND High Performance Computing
30mA Compiler-First Planetary Compute Engine: Automatic differentiable and performance(…)
01h 03mCLANG + C2 - Engineering/Futures/Measurements