Posts: 13,295
Threads: 9,036
Thanks Received: 8,861 in 7,016 posts
Thanks Given: 9,809
Joined: 12 September 18
12 February 24, 16:24
Quote:ZLUDA enables CUDA apps on ROCm platform, no code changes required
AMD-backed ZLUDA project can now enable code written in NVIDIA CUDA to run natively on AMD hardware.
![[Image: AMD-ZLUDA-GEFORCE-CUDA-1200x615.jpg]](https://cdn.videocardz.com/1/2024/02/AMD-ZLUDA-GEFORCE-CUDA-1200x615.jpg)
AMD has reportedly taken over the project of a single developer called ZLUDA, which was originally a drop-in CUDA implementation to run through Intel OneAPI. T
his project, which has already been discontinued, allowed CUDA apps to run natively on Intel hardware. Turns out, AMD has reached out to the developer behind this project to see if he would be interested in porting this tool to run on AMD GPUs through HIP/ROCm platform.
According to the report from Phoronix, the project has restarted with efforts to enable CUDA apps to run on AMD hardware without any translation or code changes required. While not all apps run natively yet (such as NVIDIA Optix), but developers can now run binaries without changes on Radeon GPUs. The only requirement is to include ZLUDA library that will replace CUDA.
The project came back to life, but AMD had actually discontinued its support. But this gave the creator the chance to share the code as free open-source software, and that’s what’s going on now. Phoronix has tested the library, and they confirm that the CUDA-enabled software runs seamlessly on ROCm without any required modifications.
...
Continue Reading