NVIDIA GCC Backend Gets Ready For OpenMP Offloading
While GCC 7 feature development is officially over, one of the late patches to land for GCC 7.1 in trunk are improvements to the NVIDIA NVPTX back-end.
The big code that landed today in GCC are all of the prerequisites for supporting OpenMP offloading with the NVPTX back-end. PTX is the IR used in NVIDIA's CUDA that is then consumed by the proprietary graphics driver for converting this representation into the actual GPU machine code. The NVPTX back-end has been part of GCC for a while now as part of OpenACC offloading and now the recent focus on OpenMP offloading.
With the new code the prerequisites for being able to support OpenMP offloading for NVIDIA NVPTX are in place. The OpenMP 4.0 specification had introduced support for accelerators -- including GPUs.
Those wanting to learn more about GCC's current offloading support can read this Wiki page.
The big code that landed today in GCC are all of the prerequisites for supporting OpenMP offloading with the NVPTX back-end. PTX is the IR used in NVIDIA's CUDA that is then consumed by the proprietary graphics driver for converting this representation into the actual GPU machine code. The NVPTX back-end has been part of GCC for a while now as part of OpenACC offloading and now the recent focus on OpenMP offloading.
With the new code the prerequisites for being able to support OpenMP offloading for NVIDIA NVPTX are in place. The OpenMP 4.0 specification had introduced support for accelerators -- including GPUs.
Those wanting to learn more about GCC's current offloading support can read this Wiki page.
2 Comments