ggml : add Flash Attention (llama/5021)

Vulkan k-quant mmq and ggml-backend offload functionality (llama/6155)

sync : ggml (#2001)

llama : add pipeline parallelism support (llama/6017)

ggml : remove old quantization functions (llama/5942)

Vulkan Improvements (llama/5835)

ggml : introduce ggml_status (ggml/750)

ggml-vulkan: fix VULKAN_CHECK_RESULTS flag, which was previously broken (llama/5813)

make portability_enumeration_ext apple only (llama/5757)

code : normalize enum names (llama/5697)

Introduce backend GUIDs (ggml/743)

Refactor validation and enumeration platform checks into functions to clean up ggml_vk_instance_init()

Add check for VK_KHR_portability_enumeration for MoltenVK support

Add preprocessor checks for Apple devices.

Resolve ErrorIncompatibleDriver with Vulkan on MacOS.

cmake : fix VULKAN and ROCm builds (llama/5525)

vulkan: Find optimal memory type but with fallback (llama/5381)

vulkan: only use M-sized matmul on Apple GPUs (llama/5412)

src : relocate new backend sources