Deep Learning has established itself to be a common occurrence in the business lexicon. The unprecedented success of deep learning in recent years can be attributed to: abundance of data, availability of gargantuan compute capabilities offered by GPUs, and adoption of open-source philosophy by the researchers and industry. Deep neural networks can be decomposed into a series of different operators. MIOpen, AMD's open-source deep learning primitives library for GPUs, provides highly optimized implementations of such operators, shielding researchers from internal implementation details and hence, accelerating the time to discovery. This paper introduces MIOpen and provides details about the internal workings of the library and supported features. MIOpen innovates on several fronts, such as implementing fusion to optimize for memory bandwidth and GPU launch overheads, providing an auto-tuning infrastructure to overcome the large design space of problem configurations, and implementing different algorithms to optimize convolutions for different filter and input sizes. MIOpen is one of the first libraries to publicly support the bfloat16 data-type for convolutions, allowing efficient training at lower precision without the loss of accuracy.

NVIDIA, which dominates the machine learning market, provides drivers with their proprietary license so that they can control the rules freely. For example, they can freely change their EULA and restrict the data center deployment and commercial use of some products. On the other hand, drivers called ROCm contains the Kernel driver and runtime, and libraries for machine learning called MIOpen developed by AMD are subject to open source licenses. Therefore, it is unlikely that commercial use will be restricted even for consumer products. The MIOpen that runs on ROCm supports two programming models of OpenCL and HIP. And HIP which is a component of ROCm allows developers to convert CUDA code to portable C .

The new Radeon Instinct is composed of hardware and software aspects that have the ability to deliver a full machine intelligence platform. AMD has unveiled a new GPU, the Radeon Instinct, but it's not for gaming. The Instinct is designed for high-performance machine learning, and uses a brand new open-source library for GPU accelerators called MIOpen. Together, the Instinct and MIOpen create what AMD hopes will be an open software ecosystem for machine intelligence. "Radeon Instinct is set to dramatically advance the pace of machine intelligence through an approach built on high-performance GPU accelerators, and free, open-source software in MIOpen and ROCm," said AMD President and CEO Dr. Lisa Su.