AMD ROCm 10.1 Released With Many Improvements

Following the ROCm 10.0 release from late August, ROCm 10.1 is now available in continuing with the ROCm six week release cycle. ROCm 10.1 brings hipFILE improvements for AMD infinity Storage with an async fast-path backend, a batch I/O API, and multi-tier I/O statistics. ROCm 10.1 also brings NUMA-aware host memory allocations with HIP, new ROCm CLI and AMD skills, kernel replay support in beta form with the ROCprofiler, moving to the LLVM 24 compiler stack, faster rebuilds, continued WSL2 improvements, hipThreads for incremental GPU acceleration for threaded code, and a variety of other improvements. More details on today's ROCm 10.1 release can be found via the ROCm blog. There is also the documentation with more details on the individual ROCm 10.1 changes.