Hacker News new | ask | show | jobs
by zanny 4405 days ago
Mantle is not an open spec at all. Certainly AMD has showed to Intel and Nvidia, but I see no reason for them to consider it unless we get standards publications of it.

And even then, this attitude AMD has had for the last two years since announcing it certainly dooms it to a death of irrelevance. If they were serious about it being an open API, they would have opened up proposals for modifications and considerations from other vendors before having it deployed in games.

They could still do a heel face turn of course, but I would never expect it. OpenGL5 is more likely the answer, but that requires Khronos to accept that even the Core Profile shift in 3.2 was not enough, they need to rewrite the entire shader language and standardize a modern pipeline without all this redundant backwards oldschool bullcrap that makes the implementations so slow due to state nonsense.

1 comments

> Mantle is not an open spec at all.

That kind of defeats the purpose then. If AMD wanted any kind of chances for adoption they had to open it from the start.

> But that requires Khronos to accept that even the Core Profile shift in 3.2 was not enough, they need to rewrite the entire shader language and standardize a modern pipeline without all this redundant backwards oldschool bullcrap that makes the implementations so slow due to state nonsense.

How exactly is Khronos managed? Is there someone with decisive power, or it's simply by consensus of all participants? For example can Valve or any other interested party come tomorrow with a serious overhaul of OpenGL and expect that such changes will be accepted?

It is design by committee, and a huge swathe of companies are no said committee. Valve could develop an extension branch (ARB prefix means approved language features, and vendor specific would be NV, INTEL, WIN, SGI, etc. So Valve could write their own plethora of extensions as VALVE_something, yeah, and pressure the review board into accepting them into the official implementation.
It sounds like rewriting the API with changing some underlying concepts is a deeper task than simply adding new extensions. (For example changing the threading model used in OpenGL).