Zink OpenGL-On-Vulkan Hitting ~95% Speed Of Native OpenGL Driver Performance

[phoronix]

Phoronix: Zink OpenGL-On-Vulkan Hitting ~95% Speed Of Native OpenGL Driver Performance

Zink as the Mesa Gallium3D implementation putting OpenGL 3.x/4.x on top of the Vulkan API is now offering near-native performance...

http://www.phoronix.com/scan.php?page=news_item&px=Zink-95-OpenGL-Performance

[uid313]

This is amazing! 👍️

Would it be possible with further improvements and optimizations to Zink to reach 100% performance parity?
And would it be possible for Zink to be faster than Intel's OpenGL driver?

[marlock]

given an extensively optimized OpenGL driver vs. an extensively optimized Zinc solution, in my humble non-specialist view it's unlikely to have better performance...

...but vulkan drivers are easier to maintain (especially if OpenGL stops sucking up driver developer time) and Zinc will be more and more optimized for efficiently mapping over vulkan...

...so IMHO there is at least hope it might exceed real performance against sub-optimal / legacy drivers

[schmidtbag]

69% improvement!
Nice...

[andre30correia]

given an extensively optimized OpenGL driver vs. an extensively optimized Zinc solution, in my humble non-specialist view it's unlikely to have better performance...

...but vulkan drivers are easier to maintain (especially if OpenGL stops sucking up driver developer time) and Zinc will be more and more optimized for efficiently mapping over vulkan...

...so IMHO there is at least hope it might exceed real performance against sub-optimal / legacy drivers

this is a good thing for future hardware and some hardware with no big resourges teams and for soc from android

[user1]

I guess this is going to be useful in the far far future when no new software will support OpenGL and new hardware will stop supporting OpenGL as well.

[shmerl]

I'd take 95% of radeonsi through radv/aco to avoid dealing with llvm bugs.

[kbios]

It's absolutely amazing. If a single developer can reach 95% of native performance in a few months there is really no reason to develop hardware-specific opengl drivers anymore.

Just think about what would happen if amd and intel helped zink with 1-2 developers each...

[Ladis]

I guess this is going to be useful in the far far future when no new software will support OpenGL and new hardware will stop supporting OpenGL as well.

With 95% performance it will be ok even for new software. And the future doesn't have to be far far away. ARM MacBooks are coming in weeks and that will accelerate ARM adoption in PC notebooks and budget desktops. Funny thing is, we will probably use Zink to get OpenGL on the all OSes: macOS (after removing the already deprecated OpenGL we will have Zink+MoltenVK combo), Windows 10 on ARM (MS is working with Zink to implement translation directly to the Direct3D driver API), and Linux (Vulkan driver at the bottom is already at a better shape than desktop OpenGL driver on ARM devices - as those have SoC coming from phones and there is no desktop OpenGL).

[cl333r]

Mike Blumenkrantz - what a productive guy! Someone please pour a bucket of water onto his keyboard before it catches fire