
oneAPI Implementations
Codeplay’s DPC++ Implementation Supports SYCL and other Languages for NVIDIA GPUs
Read More
oneAPI Elements

Tim Harris, Principal Architect
Azure AI, Microsoft

How oneAPI helped Microsoft Azure with its open standards-based initiatives
“The industry needs a programming model where developers can take advantage of an array of innovative hardware architectures. The goal of oneAPI is to provide increased choice of hardware vendors, processor architectures, and faster support of next-generation accelerators. Microsoft has been using oneAPI elements across the Intel hardware offerings as part of its initiatives and supports the open standards-based specification. We are excited to support our customers with choice and accelerate the growth of AI and machine learning.”
Languages

DPC++
Data parallel programming for CPU's and accelerators based on SYCL

hipSYCL++
Library-based implementation of SYCL

Numba
Compiler for data parallel programming in Python

DPCTL
Python bindings for SYCL classes

Julia
Compiler for data parallel programming in Julia
Deep Learning

TensorFlow
Deep learning framework

PyTorch
Deep learning framework

ONNX Runtime
Deep learning framework

Apache MXNet
Deep learning framework

PaddlePaddle
Deep learning framework
oneCCL
Communication primitives for distributed deep learning
oneDNN
Computation primitives for distributed deep learning
Data Science

Modin
Accelerated Pandas

Scikit-learn-intelex
Accelerated Scikit-learn
oneDAL
Accelerated machine learning algorithms for C++, Python, and Java
Video, Ray Tracing, and Rendering
Libraries

dpNP
NumPy-like API accelerated with SYCL
Level Zero
Low-level runtime for oneAPI

MPICH
High-performance implementation of MPI

oneDPL
Library for implementing data parallel algorithms with DPC++
oneMKL
Math kernel library
oneTBB
Threading building blocks
oneVPL
Decoding, encoding, and processing
Tools
HPCToolkit
Profiling toolkit from Rice University