ggml : fix weak alias win32 (#0)

android : fix builds (#0)

sync : ggml

files : remove old sources (part 2)

sync : ggml

files : remove old sources

talk-llama : sync llama.cpp

sync : ggml

metal : use less stack memory in FA kernel (llama/14088)

ggml-cpu : split arch-specific implementations (llama/13892)

cuda : fix device sync on buffer clear (llama/14033)

CANN: Simplify the environment variable setting(#13104)

sycl: Add reorder to Q6_K mmvq implementation (llama/13885)

cuda : fix buffer type check with integrated GPUs (llama/14069)

SYCL: Implement few same quantized type copy kernels (llama/13739)

vulkan: Enable VK_KHR_cooperative_matrix extension for Intel Xe2 GPUs (llama/14001)

llama : allow using mmap without PrefetchVirtualMemory, apply GGML_WIN_VER to llama.cpp sources (llama/14013)

vulkan: automatically deduce size of push constants (llama/13936)

ggml-vulkan: adds support for op CONV_TRANSPOSE_1D (llama/13813)

releases : use dl backend for linux release, remove arm64 linux release (llama/13996)

CUDA: fix FTZ in FA for Gemma 3 (llama/13991)

vulkan: fix warnings in perf logger querypool code (llama/13937)

opencl: add `backend_synchronize` (llama/13939)

OpenCL: Add concat, tsembd, upscale, tanh, pad and repeat (llama/13840)

metal : use F32 accumulators in FA kernels (llama/13975)

cmake : Handle mixed-case 'Power' strings in POWER CPU detection (llama/13966)

sycl: quantize and reorder the input to q8_1 when reorder is enabled (llama/13826)

gguf: fix failure on version == 0 (llama/13956)

ggml: check if non-native endian model is being loaded (llama/13943)

Add in-build ggml::ggml ALIAS library (ggml/1260)

ruby : output format (#3237)

ci : build and publish main-intel image (#3231)

docker : add main-intel dockerfile (#3229)

ruby : Add parallel transcription support (#3222)

ci : add mirror for ports.ubuntu.com (ARM packages) (#3221)

bindings.java : apply whisperParams in fullTranscribeWithTime instead of ignoring them (#3201)

musa: correct MUSA SDK rc4.0.1 download URL (#3217)

ci : use mirrors.kernel.org for Ubuntu packages (#3220)

node : add language detection support (#3190)

talk-llama : sync llama.cpp

sync : ggml

threading: support for GGML_SCHED_PRIO_LOW, update thread info on Windows to avoid throttling (llama/12995)

CUDA: add a prop in ggml_cuda_device_infor for distinguish iGPU or dGPU in cuda (#13856) (llama/13895)

CUDA: fix typo in FlashAttention code (llama/13926)

sched : avoid changing cur_copy when a graph is already allocated (llama/13922)

cuda : prevent using split buffers with 3d/4d matrices (llama/13919)

SYCL: Add mrope kernel (llama/13755)

cmake: Guard GGML_CPU_ALL_VARIANTS by architecture (llama/13890)

arm64: optimize q4_k_q8_k kernel with i8mm (llama/13886)

cmake: Factor out CPU architecture detection (llama/13883)