Any plans to implement d3d 9 on top of vulkan ?

[kalin]

I think is good idea to be start this kind of project. It will make easier to port games/apps from windows to linux,mac and any other os that support vulkan.
I hope someone will contact with valve, wine and sdl guys and share this idea. I also hope that the community will hug this idea and start the project itself.
For regret im not very experienced in 3d programing and can't start this project as stand alone.
pros:
1 open source
2 user space
3 os independent
4 vendor independent
5 widly available
6 give access to d3d9 renderer to our favorite os
cons:
1 require d3d10/opengl 3.3 hardware

fill free to share your opinion.

[Kano]

I think that would be counter productive to do so. Vulkan/DX12 will be needed for next generation engines/games - especially in the mind that you can control more things in way it works faster with YOUR app. If you create a generic wrapper how would you optimize it? There are definitely things useful to implement, like GLSL or maybe HLSL to SPIR-V but i think the rest should be done directly. Too many abstraction layers do not really increase speed - MS talks about 20% speed increase on average for DX12. Most likely Vulkan can provide the same if done properly, but I doubt that you will see so many current games using Vulkan in the end. But the future may look much britgher if in 1-2 years all most multiplatform engines/games will have a backend for it. The reason is simple, you can not sell games full price for a lifetime, only at the beginning, that means you have to focus first on new things and not tune old ones first. If you can update to a new engine easyly you could expect some updates for a bunch of older games however.

[kalin]

Performance is not the most important thing when you don't have other options. And vulkan to d3d is a lot more reasonable from opengl to d3d (wine eON way)
Also performance lost will be not so great (10 -20% I think) and old games (write with d3d9) are write for older hardware from those that support vulcan.
The Only native alternative (gallium3d d3d9) work only with open source driver only with amd wich will not be respected from game porters.
I think the idea is very good.

[enigma101]

I think that would be counter productive to do so. Vulkan/DX12 will be needed for next generation engines/games - especially in the mind that you can control more things in way it works faster with YOUR app. If you create a generic wrapper how would you optimize it? There are definitely things useful to implement, like GLSL or maybe HLSL to SPIR-V but i think the rest should be done directly. Too many abstraction layers do not really increase speed - MS talks about 20% speed increase on average for DX12. Most likely Vulkan can provide the same if done properly, but I doubt that you will see so many current games using Vulkan in the end. But the future may look much britgher if in 1-2 years all most multiplatform engines/games will have a backend for it. The reason is simple, you can not sell games full price for a lifetime, only at the beginning, that means you have to focus first on new things and not tune old ones first. If you can update to a new engine easyly you could expect some updates for a bunch of older games however.

From what I saw it looks like a 66% performance increase. Intel did a demo and in OpenGL (on Windows) the app got about 30fps and with Vulkan it went up to just over 50fps. I was thinking the same thing. Maybe converting the DX calls to Vulkan would give better performance if done right. Mainly because Vulkan uses so much less CPU, that leaves more CPU to convert more calls or do more computing for the application.