NVIDIA Developers Express Interest In Helping Out libc++/libstdc++ Parallel Algorithms

NVIDIA developers have expressed interest in helping the open-source GCC libstdc++ and LLVM Clang libc++ standard libraries in bringing up support for the standardized parallel algorithms.

C++17 brings parallelized versions for some of the algorithms exposed by the C++ standard library, but sadly GCC's libstdc++ and LLVM's libc++ do not yet support these parallel algorithms while the rest of their C++17 support is in great shape. Going back over a year Intel has been interested in contributing parallel support code to these C++ standard libraries that could be shared by both projects. The Intel path builds in abstractions for supporting different underlying thread/parallelism APIs.

Intel's Parallel STL code continues to be maintained and for their purposes are building off their own Threading Building Blocks (TBB) implementation.

AMD / HSA Foundation meanwhile has been working on their own libstdc++ parallel STL offloading code suited for the Heterogeneous System Architecture.

This week with the LLVM Developers' Meeting the past few days in San Jose, there's been interest from NVIDIA in helping out on these parallel algorithms implementation. The mailing list post comes from a member of NVIDIA's CUDA C++ team.

NVIDIA developers haven't pushed any code yet, but great to see more industry pressure on getting this parallel STL code into place for rounding out the C++17 functionality by these leading open-source compilers. We'll see if this support can be managed by the LLVM Clang 8.0 release in early 2019.

Those interested in the C++17 parallel STL features can learn more at software.intel.com.