Llama.cpp b9095 Release Enhances CUDA AllReduce

How it developed

1. May 11, 2026llama.cpp ReleasesWhere it started

   Llama.cpp b9095 Release Enhances CUDA AllReduce

2. May 21, 2026llama.cpp Releases

   Llama.cpp Enhances Performance with PDL for NVIDIA GPUs

   Llama.cpp adds Programmatic Dependent Launch (PDL) support to optimize performance on NVIDIA Hopper architecture GPUs.

3. May 22, 2026llama.cpp Releases

   llama.cpp b9275 Release Optimizes Metal Kernels

   Release b9275 optimizes Metal kernels with row batching for small widths and adds 50 new reshaping test cases.