ROCm is an open software platform built for high flexibility and high performance, for accelerated computing and no programming language restrictions, enable participants in the machine learning and high-performance computing communities to use a variety of open-source computing languages, compilers, libraries, and redesigned tools to speed up code development, suitable for large-scale computing and support for multi-GPU computing.
However, ROCm has not provided effective support for Navi 1x and Navi 2x GPUs for a long time. Although workstation GPUs are supported on ROCm 5.0, it is possible that the Radeon RX 6000 series will work with ROCm 5.0 given the relevance of consumer GPUs.
Phoronix said that relevant tests will be carried out, and whether ROCm 5.0 will support the Polaris architecture GPU again remains to be confirmed. ROCm 5.0 also adds support for Red Hat Enterprise Linux 8.5 and introduces part of the kernel-mode driver code for AMDGPU/AMDKFD.
Last year AMD established a project called GPUFORT, also part of ROCm. AMD has been working hard for a long time to help developers migrate as much CUDA-specific code as possible to the interfaces supported by the Radeon open source compute stack, enabling large CUDA codebases to work outside of the NVIDIA ecosystem. In this ROCm 5.0, the related support code has also been optimized.
Since AMD did not state the ROCm 5.0 in advance, there are still many questions to be confirmed.