I am excited to share the wonderful news that some brilliant engineers have joined Intel’s oneAPI team.
This week, the team behind ArrayFire has joined Intel to work on our shared vision of openness for accelerated computing (includes GPU computing). Some of their recent work was highlighted in an article ArrayFire Interoperability with oneAPI, Libraries, and OpenCL published in Issue 47 of Parallel Universe Magazine.
At ArrayFire, this team pioneered productive approaches to accelerated computing. They created the ArrayFire library with fast, easy-to-use math routines resting on an ingenious runtime that optimizes low-level performance details through clever JIT compilation of kernel code and intelligent memory management.
Because oneAPI is a foundational effort to ensure accelerated computing is open, multivendor, and multiarchitecture; the teams instantly knew we were kindred spirits.
With their pioneering track record and demonstrated skill at bringing open solutions to accelerated computing, we are all excited at the opportunity to join forces in helping make open accelerated computing a reality for everyone.
A Brief History of ArrayFire
With a 15-year history in accelerated computing, the experts behind ArrayFire are accelerated computing pioneers.
ArrayFire's predecessor AccelerEyes, built the first high-level language approach to GPU computing with its Jacket product which offloaded MATLAB* functions to the GPU, initially using OpenGL shaders. Jacket’s user base grew until, in 2012, it became part of MATLAB through their Parallel Computing Toolbox and MATLAB Distributed Computing Server as part of a licensing deal with MathWorks.
A few years later, in 2014, ArrayFire was released as an open-source project. The company pivoted from a licensing business model to providing custom code acceleration and training services.
Importance of Open
“The move to open source was the most significant decision in our 15-year history,” says John Melonakos, CEO & Co-Founder of ArrayFire (now with Intel). “It was tough to walk away from the licensing revenue we had built over seven years. But we trusted that a broader community-centric and open approach would be more meaningful than closed proprietary work. Intel’s vision with oneAPI for an open and collaborative building experience nicely aligns with the ArrayFire mission.”
Thankfully, developers responded by building incredible applications using the ArrayFire library, impacting every technical computing domain.
oneAPI Future for ArrayFire
This team at Intel is now building a oneAPI backend allowing library users to migrate to oneAPI. This exciting work will further promote our vision for open accelerated technical computing.