AMD continues with their aggressive efforts to enhance their GPU software compute ecosystem with ROCm. The fire under them has been lit and they have been taking their software efforts more expeditiously in recent times to better compete with NVIDIA’s CUDA ecosystem and ensuring their Instinct hardware is properly primed to compete. The release dance has begun for ROCm 7.1.
It was just one and a half months ago that ROCm 7.0 officially released in mid-September. Since then was ROCm 7.0.1 and ROCm 7.0.2 as the current stable version. Plus AMD also introduced the oddly-numbered ROCm 7.9 as their new technology preview stream for representing what will become their next-generation ROCm next year, presumably ROCm 8.0. Now ROCm 7.1 looks like it will be out before the end of October.
Overnight Git release tags for ROCm 7.1 began appearing across AMD’s ROCm GitHub repositories. AOMP rocm-7.1 was tagged for their AOMP LLVM/Clang downstream for targeting GPU offloading. There was also the HIP rocm-7.1.0 tag too for that important piece of the ROCm stack.
Similarly, the ROCm-libraries 7.1.0 was tagged and the various other repositories making up ROCm.
The ROCm documentation for ROCm 7.1 has yet to be published and the release binaries haven’t yet been uploaded, but it’s likely a matter of time now. As soon as today or in the coming days otherwise it looks like ROCm 7.1 will officially be released. It’s great seeing AMD’s ROCm software finally moving at an extremely brisk pace. More ROCm 7.1 feature details and likely benchmarks to come following that official announcement.
