ggml

ggml is a tensor library for machine learning written in plain C/C++, with the stated goal of being simple, portable and efficient with minimal setup. It is the tensor foundation behind projects such as llama.cpp. Key characteristics described in the README: - Plain C/C++ implementation without external dependencies. - Cross-platform support: x86, ARM, RISC-V, LoongArch, PowerPC, s390x and WebAssembly. - SIMD-optimized kernels for x86, ARM and RISC-V. - Broad backend support covering CPU, GPU, NPU and browser environments. - Integer quantization from 2-bit to 8-bit, plus MXFP4 and NVFP4 microscaling formats. - Zero memory allocations during runtime. Getting started is done by building from source with CMake: ```bash git clone https://github.com/ggml-org/ggml cd ggml mkdir build && cd build cmake .. cmake --build . --config Release -j 8 ``` A minimal, fully commented example covering matrix multiplication is available under examples/simple. Documentation referenced by the project includes the GGUF file format specification, an introduction to ggml published on Hugging Face, and a GGML tips and tricks wiki page. For contributions to the core ggml library, including the CMake build system, the README asks contributors to open a pull request in the llama.cpp repository instead, so that changes receive more visibility, testing and review. The project is released under the MIT license.