How does OpenCL being in the kernel benefit the opensource community? is there added performance or does it just become a default that is available to everyone?
Im ok for the mass public, ( and again, there we dont really speak about Linux itself, but mostly about OS based on Linux ( Unbutu, etc ).
Now, for server etc, Linux is largely used ( if not the most used )… and there we are mostly speak about professional applications ( i dont think most users dont even know what is ROCm ).
This said, ROCm 1.7 is out. supporting Caffe, Tensercore etc…
Could someone explain what exactly this means for Linux (and blender)? I’m afraid I don’t know enough about either the kernel or the code that AMD is supplying to understand the significance.
OpenCL rendering with open source driver is already possible. Even with AMDGPU PRO on Vega, it’s already the case actually as it uses RoCm. And you can also compile everything yourself. The point of this inclusion is to make it available for all distros without user needing to do anything else than install packages from official distro’s repositories.
So if you use AMDGPU pro or are using a Ubuntu-based distro (for which there are repos with RoCm), you can already benefit from all this changes without waiting for 4.15 kernel.
Thanks. I wasn’t aware of this in general or ROCm in specific. So, if I understand you correctly, the open source drivers can currently render using OpenCL, but with the new kernel they’ll be able to do so “out of the box,” so to speak.