Gdev Open-Source CUDA Runtime Is Still Running

Gdev, the GPGPU run-time and resource management engines that provides an open-source NVIDIA CUDA run-time, is still being worked on at the University of California Santa Cruz in conjunction with PathScale.

Back in March I wrote about Gdev: A Competitive Open-Source CUDA Implementation. Gdev is being led by Shinpei Kato, the University of California developer who had last year written TimeGraph as an open-source GPU command scheduler and then an open NVIDIA compute driver.

See the earlier Gdev article for more information on what this open-source CUDA/GPGPU implementation can provide. The project is being mentioned this weekend just since it's been a while since last looking at Gdev but it's certainly an interesting project.

Gdev is a runtime-unified operating system module that manages GPUs as first-class computing resources. Currently it supports only NVIDIA's Fermi GPUs, but the concept of Gdev is also applicable to generic "compute devices". Gdev coordinates with a DRM-based GPU device driver (pscnv/nouveau) in the operating system, providing APIs for application programs. Gdev API is a low-level primitive that allows programmers to control the details of GPU resource parameters, while Gdev also supports a high-level API, such as CUDA. Gdev is available for GPGPU and graphics applications. It is self-contained for GPGPU, though graphics applications require additional packages, such as OpenGL, LIBDRM, and DDX.

Gdev is open-source. We believe that this open-source implementation facilitates further research and development of GPU technology.

While some months have passed without any major announcement, the code is still being advanced. As can be seen from its GitHub page, Gdev is still actively being maintained with the last commit from Shinpei being just three days ago. Earlier in June, Gdev's CUDA parser (Cubin), was upgraded for NVIDIA's CUDA version 4.1 implementation too. While its advancing, the CUDA driver API is still not fully supported at this time.

Besides needing the Git of Gdev to build, Envytools is needed to compile/decompile NVIDIA GPU program/firmware/macro code, and the PSCNV/Nouveau driver is needed on a Fermi-class graphics card (GeForce 400/500). There's also other manual steps needed to be carried out at this time for setting up Gdev.