ggml : sync/merge cmake,riscv,powerpc, add common.cmake (ggml/0)

llamafile : ppc64le MMA implementation for Q4_0. (llama/12489)

SYCL: implement memset ggml backend buffer interface (llama/12580)

HIP: Add support for RDNA4 targets (llama/12372)

metal : refactor mat-vec code (llama/12569)

ggml : fix MUL_MAT_ID repack with Q8_K (llama/12544)

ggml-cpu : update KleidiAI to v1.5.0 (llama/12568)

SYCL: disable Q4_0 reorder optimization (llama/12560)

opencl: simplify kernel embedding logic in cmakefile (llama/12503)

CUDA: Fix clang warnings (llama/12540)

vulkan: fix mul_mat_vec failure in backend tests (llama/12529)

ggml : fix quantized cpy op (llama/12310)

musa: refine compute capability (llama/12493)

vulkan: Optimize mul_mat_vec p021 and nc shaders (llama/12505)

Vulkan: RTE rounding for cpy to quant (llama/12480)

vulkan: workaround for AMD Windows driver 16 bit unpack8 bug (llama/12472)

Fix build on Windows when ccache enabled (ggml/9954) (llama/9976)

sycl: cleanup oneDNN related code (llama/12097)

ggml : block interleaving support for Q4_K quantization for x86 AVX2 architecture (llama/12332)

CUDA: Improve flash decoding kernel GPU occupancy for BS=1 case (llama/12183)

vulkan: optimize iq1 coopmat2 dequant functions (llama/12427)

Fix visionOS build and add CI (llama/12415)

vulkan: Submit once enough matmul work has been recorded (llama/12406)

opencl: improve profiling (llama/12442)

musa: override warp_size of musa device to 32 (llama/12445)

SYCL: using graphs is configurable by environment variable and compile option (llama/12371)

ggml : add SVE support for q6_K_q8_K (llama/12361)

Vulkan: Default to 1GB allocations instead of 4GB to avoid fragmentation and driver issues (llama/12434)

fixed compilation warnings in ggml-sycl (llama/12424)

llama: Add support for RWKV v7 architecture (llama/12412)

cuda : enable CUDA Graph on CUDA Toolkit < 12.x (llama/12394)

ggml-vulkan: remove unused find_program(glslc) (llama/12416)

vulkan: Add N/2 and N/4 optimized paths in coopmat2 shader (llama/12312)

vulkan: subgroup size tuning (llama/12087)

vulkan: use fp32 in coopmat2 q4_k dequant function (llama/12309)

vulkan: Pad N dimension of B matrix for coopmat2 perf, to avoid bounds checking (llama/12273)

vulkan: Adjust coopmat2 tile sizes and selection heuristic (llama/12258)

cmake : enable building llama.cpp using system libggml (llama/12321)

SYCL: set extras only on GGML_TYPE_Q4_0 (llama/12366)

SYCL: Delete redundant plus sign and space (llama/12391)

SYCL : support non-contiguous tensors in binary ops (add, sub, etc) (llama/12399)

MUL_MAT optimization (llama/12382)

sycl : variable sg_size support for mmvq kernels (llama/12336)

CUDA/HIP: Fix fattn-vec-* when device warp size is not 32 (llama/12315)

vulkan: fix bug in coopmat1 mul_mat_id (llama/12316)

CUDA/HIP: refractor mmqv to unify the calculation of nwarps and rows per block between host and device code. (llama/12177)

ggml-backend : fix backend search path (llama/12330)

metal : Cache the Metal library at the device context level (llama/12265)

mat vec double buffer (llama/12188)

musa: support new arch mp_31 and update doc (llama/12296)