Cycles OpenCL rendering on macOS is no longer supported

[BTolputt]

@bsavery

The subject above and Ton’s message was about a Metal backend for UI, not Cycles.

True, but I was responding to your query specifically asking about the cuda backend of cycles. You know, as per my quoting of you at the start of the post.

@bsavery

As someone who actually manages large projects, I’d actually disagree on that “if it ain’t broke don’t fix it” mentality.

Just so we can stop this from getting out of hand - I too am “actually” someone that has managed large projects. At one point managing over ten other developers whilst still having to get my hands dirty in code. So, with that in mind, we can focus on the arguments and not any relative merit from our career backgrounds.

@bsavery

If it takes a bunch of work to rewrite something which makes it more maintainable and capable in the future, that is a good investment.

Two things to note about that.

Firstly, the investment has already been made. Had you made the argument that it would have been better to move to QT before the changes to OpenGL core profile, the argument would have more merit. As it stands now, the cost to be considered is both the future cost of reimplementing the core of Blender (remembering QT is not “just” a UI) and the cost of what’s been done already.

Secondly, you make the assumption that the benefits offered by QT outweigh the advantages of a slimline core like that of Blender GHOST library. Currently, what Blender’s UI core offers is what Blender needs - it’s not missing what they need outside the “possibility” that Apple will remove OpenGL completely from OSX. There are advantages to lightweight UI libraries (which is why FLTK, GHOST, and the like still exist). There are also advantages to being able to direct when that UI library fixes a given problem. So in order for moving to QT to be a “good investment”, there needs to be a relative weighting given to each of the benefits QT provides vs what Blender’s dev team loses.

@bsavery

This is the part that is “broke” in that it took person years to maintain. Why not make this use an abstraction layer so QT maintainers are the ones to worry about OS compatibility?

It took person years to maintain over the past four years (he dis say “since 2014”). For a project of Blender’s scope - that’s not as big a deal as you seem to think it is. Certainly not something I would consider “broken”.

Right now, Blender has a UI that does everything it needs to do. It “might” need something in the future in regards to Apple’s deprecation of OpenGL or it “might” have someone step in to solve the problem for ALL Apple developers through a library that serves as an OpenGL/Vulkan facade to Metal (-cough-MoltenVK-cough-). That is indeed a risk, but whether it’s worth jumping to a new UI core at this point would mean putting a lot of weight on that risk and the need to cater for Apple users. I don’t think that focus exists in BF and I’m not sure I disagree with them on that front (and I’m an Apple user/developer).