NVIDIA Talks About Vulkan Shader Resource Binding

[philcostin]

It's nice of them to actually go technical, this time. Step by step, we get to learn more about Vulkan, which is nice.



Couldn't someone extract the symbols to make their own header file? I don't think they have obfuscated them.

In the time it would take for someone to go and do all that, they'll have released the drivers - or at least one would have hoped that they had...

EDIT: This situation is a bit like kids who try to get a sneak peak at their Christmas gifts under the tree, trying to be a ninja about not disturbing the wrapping paper too much!

[Ancurio]

nonono - read the article!
this one https://www.phoronix.com/scan.php?pa...GDC-2016-Talks

Ok, let me rephrase. The individual functions and what they're supposed to do has been set in stone for a long time. The actual specification text describing every little detail ad nauseam was still being written in August.

[Ancurio]

It's nice of them to actually go technical, this time. Step by step, we get to learn more about Vulkan, which is nice.

Couldn't someone extract the symbols to make their own header file? I don't think they have obfuscated them.

Someone has done this, but it turned out very meager - only some windowing system extension symbols turned up. Unfortunately the forum post was deleted.

[M@yeulC]

Someone has done this, but it turned out very meager - only some windowing system extension symbols turned up. Unfortunately the forum post was deleted.

That's interesting, this is the first time I hear about it. A bit sad that it didn't end up usable, but hopefully we won't have to wait _that_ long before having an official spec, with actual official headers, and I guess that we'll have plenty of time on our hands to learn everything about it then. So, I agree with philcostin that it's rather worthless to try to do such a thing now, and that it would be more productive, say, to work on SPIR-V backends for the different drivers.
That said, IIRC, someone published some sort of reverse-engineered mantle API that could be used to actually render something on screen (maybe just triangles, but that's a good starting point).
Anyway, just wait and see.

[Ancurio]

That said, IIRC, someone published some sort of reverse-engineered mantle API that could be used to actually render something on screen (maybe just triangles, but that's a good starting point).
Anyway, just wait and see.

Yes, that was possible because 1) there are available AMD Mantle drivers and 2) the full API documentation is available; they just had to write a header off of that with clever guessing and trail/error.