AMD has recently requested the removal of ZLUDA, an open-source CUDA translation layer project, from its GitHub repository. This development has sent ripples through the tech community, especially given the project's potential to support NVIDIA's CUDA on AMD Radeon hardware.
 |
|---|
| AMD's request to remove ZLUDA highlights the complexities of running NVIDIA CUDA on AMD hardware |
The Rise and Fall of ZLUDA
ZLUDA, originally created by developer Andrzej Janik for Intel GPUs, was later adapted to work on AMD GPUs with support from the company. The project allowed various CUDA applications to run smoothly on AMD hardware, bridging a significant gap in the GPU ecosystem.
Earlier this year, AMD ceased funding for ZLUDA. Following this, and in line with a previous agreement, Janik released the code as open-source in February. However, AMD's legal team has now claimed that this release was not legally binding, leading to the current situation.
Legal Concerns and NVIDIA's Stance
While AMD cited legal concerns for the takedown request, it's worth noting that NVIDIA recently banned the use of translation layers with CUDA-based software. This move by NVIDIA could potentially cause legal troubles for AMD, which may have influenced their decision to distance themselves from ZLUDA.
Interestingly, Janik stated, I have received no legal threats or any communication from NVIDIA, suggesting that the pressure is coming solely from AMD at this point.
The Future of ZLUDA
Despite this setback, Janik remains determined. He has announced plans to rebuild ZLUDA from its pre-AMD codebase, stating, At this point, one more hostile corporation does not make much difference.
The rebuilt version will have a different scope and will not include certain features, such as the planned support for NVIDIA GameWorks. Janik had successfully implemented this in Batman: Arkham Knight but never completed it, and now that code will never see the light of day.
Implications for the GPU Market
This development highlights the ongoing competition and legal complexities in the GPU market, especially concerning software ecosystems. The removal of ZLUDA could potentially impact developers and researchers who were relying on this translation layer to run CUDA applications on AMD hardware.
As the story unfolds, the tech community will be watching closely to see how Janik's efforts to rebuild ZLUDA progress, and whether this will lead to a broader conversation about interoperability in the GPU space.