The Most Interesting Google GSoC 2018 Projects: QEMU Vulkan, Virtual KMS, Nautilus GTK4

[phoronix]

Phoronix: The Most Interesting Google GSoC 2018 Projects: QEMU Vulkan, Virtual KMS, Nautilus GTK4

Google has announced the accepted student projects for this year's Google Summer of Code. As usual, there is an interesting mix of open-source software projects across the hundreds (or rather thousands) of applicants. Here's a look at the most interesting initiatives we found when going through the list...

http://www.phoronix.com/scan.php?page=news_item&px=GSoC-2018-Projects

[uid313]

I would really love to see GIMP get some love.

[Marc Driftmeyer]

Sorry, but I'll take Blender's Further Cycles Rendering improvements for Volumes, with Cycles integrating fully OpenVDB for the win.

https://summerofcode.withgoogle.com/projects/#6423025180737536

This project alone will save me several hours, per frame, when dealing with all volumetric renderings of fluids, smoke, fire, etc.

[Brophen]

Since we are casting votes, Portage on Android sounds interesting. I don't quite understand the ramifications, but anything that brings Android closer to Gnu/Linux is a big plus in my book

[oiaohm]

virtual KMS module is going to be interesting to follow on X.org. This really is a missing piece to allow thinclients/remote desktop without requiring modified X11 server or modified wayland compositor particularly if it supports PRIME https://devtalk.nvidia.com/default/t...nchronization/

The ability to use normal opengl drivers on thin-clients will be a big plus.

[Brophen]

Do I understand this correctly:

"QEMU

- Vulkan support for guest VMs along the same lines as VirGL to OpenGL. "

Does that mean games on a guest will be able to run as well as on the host? Or do you still have to do passthrough to get that kind of performance?

[Terrablit]

It's a really thankless, ugly job, but I'd like to see someone help backport wireguard to older kernel versions (3.4 and then 3.0). It's not terribly applicable to mainstream linux users, but it opens up wireguard support on some really old SoCs that have driver issues with modern kernels because their vendors are asshats. Project leads state that they'd accept patches on this sort of thing (example 1, example 2), but you'd need to hit all versions in between the current oldest (3.10) and your target. So you'd be patching for 6 versions to hit 3.4, and 10 to hit 3.0.

Somebody who could "talk" vendors into better wireless and switch driver support could help, too. And I hear b43 could use some new blood. Someone asked about mainline support for AC PHY and it got shot down due to lack of resources.

[oiaohm]

Do I understand this correctly:

"QEMU

- Vulkan support for guest VMs along the same lines as VirGL to OpenGL. "

Does that mean games on a guest will be able to run as well as on the host? Or do you still have to do passthrough to get that kind of performance?

Its never as well as the host. You still have pass of data from guest to host then from host to card. This is still way faster than dropping back to pure software rendering also does not require you to have video cards with passthough support. Like Nvidia and AMD are attempting to lock passthrough support only to high end cards.

Until we get it we will not know how bad the overhead is. It stands a chance of being acceptable level of performance loss vs software rendering that is totally not acceptable performance loss.

[OneTimeShot]

"Fedora Happiness Packets" sounds like something from a 1970s B-movie about a dystopian future. Be well.

[EduRam]

Vulkan on Qemu will allow to run Google Fushia OS GUI as guest. I look forward for this.