It's not really "Gallium vs Mesa", it's "Mesa over Gallium vs Mesa over the existing driver model". Mesa is still the actual 3D driver; Gallium just plugs into Mesa and encapsulates the HW-specific code. You can also write other drivers which directly use the Gallium API. Gallium basically defines a low-level API (at the TGSI level) which you can build other drivers on -- since Mesa already existed it wasn't "built on Gallium" as much as Gallium is being "built into Mesa".
Eventually I imagine that Gallium-based Mesa will outperform the current driver model, if only because that's where all the ongoing development is going to be done. I don't think we are there yet though.
re: Wine, I don't know. There were some design assumptions made in Wine which seem to have been based on the internals of the NVidia driver (specifically the memory management) but I think those dependencies are going away over time.
Eventually I imagine that Gallium-based Mesa will outperform the current driver model, if only because that's where all the ongoing development is going to be done. I don't think we are there yet though.
re: Wine, I don't know. There were some design assumptions made in Wine which seem to have been based on the internals of the NVidia driver (specifically the memory management) but I think those dependencies are going away over time.
Comment