ggml-org has released llama.cpp b11550, with a build-system change for s390x, IBM’s 64-bit Linux architecture. The release disables the z17 target when the compiler in use does not support it.
The release page also lists prebuilt downloads for macOS, iOS and several Ubuntu configurations, including CPU, Vulkan and CUDA 12.8 versions. It does not describe other changes or give performance results.
Why it matters
For people building llama.cpp on s390x, the change may avoid selecting a z17 target their compiler cannot handle; the release notes do not specify which compilers are affected.
