Intel Vulkan Driver Now Handles PRIME-Style Rendering, Raven Ridge Lands VCN JPEG Decode

[phoronix]

Phoronix: Intel Vulkan Driver Now Handles PRIME-Style Rendering, Raven Ridge Lands VCN JPEG Decode

With just one week of feature development remaining for the in-development Mesa 18.3, the race is on for landing the remaining feature work ahead of this next quarterly Mesa3D stable version...

http://www.phoronix.com/scan.php?page=news_item&px=Mesa-18.3-ANV-PRIME-VCN-JPEG

[schmidtbag]

I'm not sure why anyone would want to render on an Intel GPU and output to a discrete GPU. But, I guess there's nothing wrong with adding the feature, anyway.

[yurikoles]

schmidtbag powersaving? Notebooks with MUX switch for output?

[schmidtbag]

schmidtbag powersaving? Notebooks with MUX switch for output?

Wouldn't it make more sense to just render and display on the Intel chip then? To my recollection, most multi-GPU laptops use the Intel chip for the display output.

Just to clarify, I see the benefit of rendering on a dGPU and displaying on an Intel chip. What I don't understand is doing the inverse of that.

[tomtomme]

Wouldn't it make more sense to just render and display on the Intel chip then? To my recollection, most multi-GPU laptops use the Intel chip for the display output.

Just to clarify, I see the benefit of rendering on a dGPU and displaying on an Intel chip. What I don't understand is doing the inverse of that.

the use case is described in the article. just read it.

[schmidtbag]

the use case is described in the article. just read it.

The use case described still doesn't make that much sense. Like I said, if you're using a laptop, the display tends to be connected to the IGP in the first place. If you're using a desktop, you can have both the IGP and dGPU active simultaneously and you'll have display connectors for both. You can do proper testing with no performance deficit by accessing the IGP directly; there's no need for PRIME.

At this rate people are going to be like "what's your gripe with this? It's not a big deal" and I don't have a gripe with it and I don't think it's a bad feature, I just find it weird, because even the proposed use-case for it isn't the easiest or most logical solution.

[edoantonioco]

Wouldn't it make more sense to just render and display on the Intel chip then? To my recollection, most multi-GPU laptops use the Intel chip for the display output.

Just to clarify, I see the benefit of rendering on a dGPU and displaying on an Intel chip. What I don't understand is doing the inverse of that.

The use case is not for end users, is only for game developers devs that needs to test the Intel graphics without having to change display connections.

[aejsmith]

The use case is not for end users, is only for game developers devs that needs to test the Intel graphics without having to change display connections.

Pretty much this. I wanted to be able to test things on Intel more easily without changing connections around. I thought this should already work and found that it didn't, looked into why and found that it was a 2 line change to make it work, so submitted a patch. When I actually care about performance, then sure, I'll change connections around to avoid the PRIME overhead, but this just makes my workflow a little bit easier if I want to investigate a bug on Intel or check that something works there.

FWIW, this did already work for some Vulkan apps that use the other API to query for presentation support, just it didn't for Feral games which use the method where my change was made.

[Zan Lynx]

Wouldn't it make more sense to just render and display on the Intel chip then? To my recollection, most multi-GPU laptops use the Intel chip for the display output.

Just to clarify, I see the benefit of rendering on a dGPU and displaying on an Intel chip. What I don't understand is doing the inverse of that.

Don't forget desktop computers, often loved by game and graphics designers. A hefty Vega 56 perhaps, connected to a monitor. But the humble Intel iGPU still lurks within.

[coder]

I'm not sure why anyone would want to render on an Intel GPU and output to a discrete GPU.

Maybe the Intel GPU stack has some particular feature or fixes some bug, and you want the benefit of display on your FreeSync monitor.