Build error fluid-mantaflow-gpu branch

Hi guys,

I tried to build the fluid-mantaflow-gpu branch, unfortunately I seem to be missing some dependencies.
This is what I did:

git checkout fluid-mantaflow-gpu
git pull
make full

This is the error I get in the CMakeError.log:

C:\Program Files (x86)\Microsoft Visual Studio\2019\Community\VC\Tools\MSVC\14.29.30133\include\arm_neon.h(21,1): fatal error C1189: #error: This header is specific to ARM targets [E:\blender-git\builds\fluid-mantaflow-gpu\CMakeFiles\CMakeTmp\cmTC_28f00.vcxproj]

Im a bit stranded here. Could somebody nudge me in the right direction?

That branch has no windows support

1 Like

Oh boy.
Thats a good clue, I just naively assumed it had.
How can I check for this kind of stuff to avoid confusion in the future?

There’s really no way to tell, master we go out of our way to guarantee it will build on all platforms we support, what individual devs do in branches could quite likely break platforms they don’t develop on.

in this specific case, the way mantaflow was enabled to work on the GPU just is incompatible with the MS compilers, so it’s unlikely to ever be supported using this method unless MS drastically changes their mind on GPU support within their compilers.

1 Like

Thank you LazyDodo !
So this branch is pretty much uninteresting for Windows users indefinitely?
Or aslong as MS doesnt change up their compiler policies.

It seems like this only has a change getting merged into master with another GPU method then?
Concluding this Im questioning the point of this branch a bit. For research maybe?

Correct, unsure if clang supports it on windows, then again, even if it does we’re unlikely to switch compilers just to appease a single component in blender.

@sebbas be the one to answer that question with any authority, however judging by the code changes being made, it appears to be a “could it be done?” style experiment rather than something looking to be included in mainline blender.

1 Like

Yes, sry for my ramblings. Thank you very much for your help! :slight_smile:

@sebbas Is there a way to see the mantaflow roadmap somewhere? I found this trello board but dont know how up-to-date it is. Trello

I don’t have the same problem. Whether it has a windows support or not honestly I don’t know, but the problem that I am facing is that when I use the Cmake to build it tells me (Compiler is not supported), although I tried to build it with Clang after installing it and trying the (make -clang) option to build.

so the problem would be in general : how to compile with GPU Offloading, what compiler should a person use in Windows (gcc - clang - vs - or other), do you need to first compile the OpenMP with GPU offloading for it to work and do you need the latest OpenMP for it to work.

Honestly I don’t know if the compiler not supported message is related to (windows not supported) issue, but I am still learning. :yum:
maybe @LazyDodo can shed some light on my questions and curiosity :wink:

You’re in uncharted territory far far out of supported waters, I’m certainly not going to stop you from sailing these waters, however I can’t be there with you.