PoCL is the open-source project implementing OpenCL for CPU-based execution as well as multi-device support by getting its Portable Computing Language implementation working atop NVIDIA GPUs via CUDA, AMD GPUs via HSA, and other experimental implementations through leveraging LLVM/Clang.
The PoCL 1.8-RC1 release has added support for the LLVM/Clang 13.0 toolchain, improves debugging with Valgrind and LTTND, improves operating system support for NetBSD and macOS, and improves its support for SPIR/SPIR-V intermediate representation on NVIDIA CUDA.
PoCL 1.8-RC1 downloads are available from GitHub . Learn more about this great open-source project at PortableCL.org .