Dolphin, the GameCube and Wii emulator - Forums

Mantle seems to be developed to utilize the CPU better which in theory could benefit dolphin?

Nvidia seems to be pushing "Modern" OpenGL as an alternative to Mantle:

http://www.dsogaming.com/news/nvidia-tal...rn-opengl/

http://de.slideshare.net/CassEveritt/beyond-porting

Speed wise not likely. If draw calls were a problem D3D9 would've been the slowest renderer not usually the fastest.

We are usually bottlenecked by buffer updates. That's why OGL got this huge performance gain with pinned memory or buffer storage. I guess mantle also has such a buffer update way, so they are both as good as each other.

im not sure why you would need mantle its such a new api. none of the old games on gamecube would ever get any performance gain from it if they even would start with it. bf4 only got a few frams extra mainly because the devs work with amd to get mantle to work with bf4

Theoretically it gives lower level access to the GPU than OGL/D3D, and Wii/GC games like lower level access to the GPU than OGL/D3D provide. However it's more or less turned out that mantle doesn't give low level enough access to the bits we'd need to justify writing a whole new graphics backend.

(05-05-2014, 12:36 AM)AnyOldName3 Wrote: [ -> ]Theoretically it gives lower level access to the GPU than OGL/D3D, and Wii/GC games like lower level access to the GPU than OGL/D3D provide. However it's more or less turned out that mantle doesn't give low level enough access to the bits we'd need to justify writing a whole new graphics backend.

What about using CUDA or OpenCL as a graphics backend? That should be lower level than Mantle.

It's potentially possible to write a graphics backend in either of those, but again, it'd likely be a load of extra work for minimal gains in accuracy.

(05-05-2014, 09:43 PM)AnyOldName3 Wrote: [ -> ]It's potentially possible to write a graphics backend in either of those, but again, it'd likely be a load of extra work for minimal gains in accuracy.

Sure, but both can be used with OpenGL, so if parts of the OpenGL api are insufficient to properly do the emulation, those parts could be done in OpenCL or CUDA.

And if that becomes the best option to make certain things work, I'm sure the devs will implement it.