There’s a lot of data. There’s a lot of need for compute. [Which means] there’s a huge need for different architectures to handle that type of compute. — Jeff McVeigh, GM of Visual Computing Software Products, Intel Corporation.
In other words, when it comes to the diversity of hardware and the growing need for specialized workloads, one size no longer fits all. Development and deployment must necessarily span more than one architecture—from scalar (like a CPU) to vector or spatial (like GPUs and FPGAs) and other accelerators.
It’s a universal problem and, according to Jeff, the impetus and motivation for oneAPI. In this 6-minute video, he discusses the new industry initiative, including:
- What it is in a nutshell
- The industry-wide and developer-specific issues it solves
- How Intel is driving—and delivering to—this open strategy