AMD Goes Full Steam To Open-Source

According to Ars Technica Under the handle “GPUOpen” AMD is releasing a slew of open-source software and tools to give developers of games, heterogeneous applications, and HPC applications deeper access to the GPU and GPU resources.

In a statement AMD said that as a continuation of the strategy it started with Mantle, it is giving even more control of the GPU to developers.

“ As console developers have benefited from low-level access to the GPU, AMD wants to continue to bring this level of access to the PC space.”

The AMD GPUOpen initiative is meant to give developers the ability to use assets they’ve already made for console development. They will have direct access to GPU hardware, as well as access to a large collection of open source effects, tools, libraries and SDKs, which are being made available on GitHub under an MIT open-source license.

AMD wants GPUOpen will enable console-style development for PC games through this open source software initiative. It also includes an end-to-end open source compute infrastructure for cluster-based computing and a new Linux software and driver strategy

All this ties in with AMD’s Boltzmann Initiative and an HSA (Heterogeneous System Architecture) software suite that includes an HCC compiler for C++ development. This was supposed to open the field of programmers who can use HSA. A new HCC C++ compiler was set up to enable developers to more easily use discrete GPU hardware in heterogeneous systems.

It also allows developers to convert CUDA code to portable C++. According to AMD, internal testing shows that in many cases 90 percent or more of CUDA code can be automatically converted into C++ with the final 10 percent converted manually in the widely popular C++ language. An early access program for the “Boltzmann Initiative” tools is planned for Q1 2016.

AMD GPUOpen includes a new Linux driver model and runtime targeted at HPC Cluster-Class Computing. The headless Linux driver is supposed to handle high-performance computing needs with low latency compute dispatch and PCI Express data transfers, peer-to-peer GPU support, Remote Direct Memory Access (RDMA) from InfiniBand that interconnects directly to GPU memory and Large Single Memory Allocation support.