Having spent ~1hr getting acquainted with the MoltenVK SDK, I have to say that it’s quite straightforward to use, if verbose (because Vk). Here’s a cheeky little cube (3.2kLOC):pic.twitter.com/oCVYJmaqfv
-
Metal also does not let you manage the memory pools yourself and you have no good control over when uploads and sychronization happen. There is also no way to do bindless resources. You see, all of this "easy to use" has its downsides.
Mhm. Give me a few weeks with the id Tech 6 code base and I’ll show you what performance with a terse API looks like. It’s not as dramatic as you’re representing.
-
-
There is no way you would even get close to the performance on AMD GPUs. Async compute is a must.
-
First missing async compute what slowdowns are we talking? Less than 20% gains normally.. second (may be wrong) but you should Epic UE4 source code.. they mentioned support for async compute on Metal on release notes long time ago maybe even before Metal2..
-
Yes. Also, this presumption that drivers can’t schedule work onto multiple queues intelligently is mostly ego protectionism. Fact is, there’s more future in APIs that do the work for you than in APIs that make you hand-hold through every step.
-
We tried having that for a long time on Win. It was called DX11 and it was always a nightmare and in the way for AAA. DX12 and Vulkan both went into the opposite direction than what you are claiming. You can have that opinion, but I don't see it happening in relevant markets.
End of conversation
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.