ggml: fix compilation error s390x (llama/12848)

cpu: fix cpu backend's supports-op for GET_ROWS_BACK. fixes a fatal when running test-backend-ops with only the CPU backend (ggml/1190)

CANN: Support more ops (llama/12841)

Fixes #12823 (llama/12830)

ggml-cpu-impl.h: do not redefine bool on POWER9 (llama/12856)

ggml-impl.h: fix build on POWER9 (llama/12855)

CANN: Support Opt CONV_TRANSPOSE_1D and ELU (llama/12786)

vulkan: In coopmat2 mmq, load q4_k/q5_k scales through shared memory (llama/12833)

vulkan: Use fp16 for the flash attention P*V multiplication (llama/12783)

cuda : add f32 to bf16 copy op (llama/12806)

llama : fix FA when KV cache is not used (i.e. embeddings) (llama/12825)

ggml: don't include arm_neon.h when using CUDA 12 with ARM Neon (ggml/1187)

ggml : add bilinear upscale support (ggml/1185)

ggml : add more generic custom op, remove deprecated custom ops (ggml/1183)

Revert "sycl:remove redundant memcopy in function ggml_backend_sycl_buffer_set_tensor" (llama/12812)

opencl: better identify Adreno GPU (llama/12760)

cuda : fix HIP and MUSA BF16 (llama/0)

sycl: remove redundant memcopy in function ggml_backend_sycl_buffer_set_tensor (llama/12734)

CANN: fix typo in ggml-cann (llama/12733)

CANN: Refactor to reduce duplicate code (llama/12731)

musa: fix compilation warnings in mp_22/31 (llama/12780)

vulkan: fix NaN issue in flash attention shader (llama/12776)

vulkan: Use unclamped loads for flash attention mask (llama/12720)

Vulkan: Tune Vulkan mmq int dot shader for performance (llama/12767)

sycl: allow ggml-sycl configuration and compilation using Visual Studio project/solution (llama/12625)

cmake: fix ggml-shaders-gen compiler paths containing spaces (llama/12747)

vulkan: Hybrid waitForFences/getFenceStatus to reduce fence latency (llama/12630)

vulkan: set cmake minimum and project name in vulkan-shaders (llama/12744)

CUDA: Prefer vector flash decoding kernel for Gemma models (llama/12738)

vulkan: Fix missing cmake logic for dot product extension (llama/12721)

fix MUSA compiler warning (llama/12704)

CANN: Support operator SIN COS ARGMAX (llama/12709)

Simplify and improve CUDA graphs through use of indirect copy pointers (llama/9017)

CANN: Fix failed test cases (llama/12708)

opencl: use `max_alloc_size` in backend ctx instead of querying again (llama/12705)

vulkan: Implement split_k for coopmat2 flash attention. (llama/12627)

cmake: remove caching from vulkan coopmat checks (llama/12719)

vulkan: Implement grouped query attention in the coopmat2 FA shader (llama/12559)

Vulkan: Fix mmq int dot float cache size (llama/12722)

llama : add option to override model tensor buffers (llama/11397)

ggml : simplify Arm fp16 CPU logic (ggml/1177)

CUDA: don't convert BF16 weights to FP32 (ggml/1174)

coreml : set convert_to="mlprogram" in convert

ci : disable freeBSD job in build.yml (#3064)

examples : add HEAPU8 to exported runtime methods (#3062)

ruby : make Ruby bindings installed with build options (#3056)

whisper : add no_context parameter to whisper_params (#3045)

examples : add FFmpeg v7.0 support to ffmpeg-transcode.cpp (#3038)

ruby: use CMake in build process (#3043)

docs : update README.md to note newer nvidia gpus (#3031)