oneAPI Libraries

oneMath covers mathematical domains such as linear algebra, transforms, random-number generation, and related operations. The project separates application-facing interfaces from underlying math libraries so supported implementations can connect to different devices and backends. Consult the project documentation for current domains, backends, compilers, and platform support.

oneTBB helps developers express logical parallelism using tasks, parallel algorithms, flow graphs, concurrent containers, and memory-allocation components. It is designed for composable, performance-oriented parallelism without requiring applications to manage threads directly. Consult the project documentation for current APIs and platform support.

oneDAL provides optimized C++ building blocks for data preparation, analysis, and machine-learning workloads, including batch and distributed computation. It supports application developers and framework authors who need reusable analytics algorithms across supported CPU and accelerator environments. Consult the project documentation for current interfaces, algorithm coverage, computational modes, and configurations.

oneCCL provides collective communication operations used to coordinate data across processes and devices in distributed-computing and machine-learning workloads. It is intended for applications and frameworks that need scalable communication patterns. Consult the project documentation for current APIs, runtimes, transports, devices, and deployment requirements.

oneDNN provides computational primitives used by deep-learning applications and frameworks. It is intended for developers who need optimized execution on supported CPUs and GPUs. Because hardware and runtime coverage evolves, consult the project documentation for current architectures, data types, primitives, graph capabilities, and support levels.

oneDPL provides parallel algorithms, execution policies, iterators, and related utilities for data-parallel programming. It helps developers apply familiar C++-style patterns to parallel work and provides supporting APIs for SYCL kernels. Consult the project documentation for current compiler requirements, API coverage, samples, and supported features.