The generous interpretation is they just don't want to invest into that if it isn't "broken" but the cynic in me can't help but think at least part of the motivation is being different and maintaining the wall around the garden.
The generous interpretation is they just don't want to invest into that if it isn't "broken" but the cynic in me can't help but think at least part of the motivation is being different and maintaining the wall around the garden.
Metal is the reason Vulkan exists: Apple announced Metal in June 2014 and the first Khronos meeting about a replacement for OpenGL was hosted at Valve in July 2014 so they could rush to announce an "open standard" under development at SIGGRAPH later that year. The only reason it managed to ship a final spec in 2016 was because AMD donated Mantle to boot-strap the process.
When you make your own GPUs and have sufficient scale to get developer adoption owning your own API can let you move much faster than others while avoiding vendor extension hell (that turns into a bespoke API-within-an-API at the extreme end).
Besides which MoltenVK effectively allows Vulkan software to run against Metal anyway. In that sense it's also better for Vulkan because it can ship API updates whenever Khronos feels like it.
That seems to give way too little credit to Mantle. Which was released in 2013 and became the basis for Vulkan.
I also noted that AMD donated Mantle as the starting point which is why Vulkan didn't take 5 years to finalize.
In any case claiming that Apple is somehow at fault for not adopting Vulkan doesn't make sense.
I just said that it's severely overstating the case.
> In any case claiming that Apple is somehow at fault for not adopting Vulkan doesn't make sense.
Why not? They could have adopted Mantle instead, for example. It came out a year earlier. Or it could have replaced Metal with Vulkan later (as it is often a reasonable choice to replace your custom solution with the industry's standard). Or it could have chosen to officially support Vulkan alongside it.
Apple has lots of money in the bank, and with that, lots of options.
It's also gaining some nice features that are usable out of the gate, rather than waiting on Vulkan extension standardization and adoption. It simply makes the most sense for Apple.
Sure, but that was hardly the point of Vulkan in the first place. The point is to have cross-platform 3D grahipcs, but since Apple never gave a shit about anything cross-platform, they chose to write their own stuff.
And if they actually did care, they could have modeled the same API as they have with Metal, but over Vulkan, so the engine is the same but the API is "much friendlier".
It simply makes the most business sense for Apple, but as is tradition, the whole software ecosystem is worse off because of their choices.
Opinion time: WebGPU will become the standard graphics API to target for cross-platform. gfx-rs/wgpu already works over Vulkan/Metal/D3D12/D3D11/OpenGL/WebGL & WebGPU in browsers. Much better than simply Vulkan, unless performance is your primary sticking point.
The only cross platform alternative that actually works is OpenCL, but Blender is moving away from OpenCL, mostly because literally every GPU vendor is doing so as well.
The more annoying variation between the two is the need to refactor the code from using "kernels" to "compute shaders", but that's just a difference in the underlying programming model.
Yes, there is SPIR-V, yet few care to target it in practice, besides typical shading languages.