Workaround for subgroup arithmetic failing on MoltenVK with AMD GPUs#15886

0cc4m merged 1 commit intoggml-org:masterggml-org/llama.cpp:masterfrom

lksj92hs:lksj92hs/moltenvk-subgroup-workaroundlksj92hs/llama.cpp:lksj92hs/moltenvk-subgroup-workaroundCopy head branch name to clipboard