Otoy is cross-compiling CUDA for non-Nvidia GPUs
Otoy has developed a cross-compiler to enable CUDA code to run on non-Nvidia hardware, including AMD, ARM and Intel GPUs, VentureBeat is reporting.
According to the story, the work will be complete “by this summer” and will be made available as part of the 3.1 release of OctaneRender, Otoy’s popular GPU-based renderer.
If so, that would mean that users of AMD GPUs will be able to take advantage of OctaneRender for the first time.
So far, there’s no information from Otoy itself, but the company has retweeted the news, and promises that an official post is “coming soon”.
So why not just use OpenCL?
Although we already have an AMD-compatible alternative to CUDA in the shape of OpenCL, the open standard has never quite made the headway within the DCC market that its supporters originally hoped.
VentureBeat reports Otoy as believing that CUDA is “superior” to OpenCL and that it enables “much richer graphics software”, hence its efforts to develop a CUDA cross-compiler for non-Nvidia hardware.
According to the story, the work enables OctaneRender to run on other GPUs at “the same speed as it runs on Nvidia cards”.
Interesting stuff. We’ll update if Otoy releases any further information directly.