The Qt 3D Story With Vulkan Should Be Quite Compelling For Qt 6.0

[phoronix]

Phoronix: The Qt 3D Story With Vulkan Should Be Quite Compelling For Qt 6.0

With the soon to be released Qt 5.14 is the start of their new high-level 3D API that itself is graphics API independent for being able to target the likes of Apple Metal and Vulkan as well as Direct3D and still falling back to OpenGL. The start of the graphics API independent scenegraph renderer is turning out well for Qt 5.14 but there will be more to come in the spring with Qt 5.15 while at the end of next year with Qt 6.0 should be a much more compelling story...

http://www.phoronix.com/scan.php?page=news_item&px=Qt-3D-Vulkan-Qt-6.0-Compelling

[Baguy]

I wonder when/if plasma will take advantage of vulkan considering QT is pushing it.

[cl333r]

I'm too old for this type of news, the PR statement doesn't compare old vs new FPS numbers, it simply states vague things that seem good, like in quantum computing.

[3diStan]

The way I see it, Qt Quick 3D is unfortunately another glimpse of the ugly direction in which things are headed more and more with Qt, that is: the tendency, since Qt was bought by Digia, to part from the nice LGPL road and release new parts only as GPL or commercial (the latter with an absolutely exorbitant pricing).
Back then, for Nokia (well, at least before the Microsoft shill Elop got injected there and destroyed the company)… Qt was a strategic asset and it made sense to open it up. For Digia on the other hand (and the newer "The Qt Company", subsidiary of Digia), Qt is seen as a core revenue source, and they are shifting the balance in that "freemium" model ever more towards the "pay to play" side.

My doubts about that strategy (from a user perspective, which of course differs from the perspective of those who make their money selling Qt) are such that I find myself starting to hope for a replacement whenever I see news about new UI libraries. Unfortunately lately the news about new UI libs were not cross-platform but platform-restricted walled gardens (e.g. SwiftUI for new Apple OSes, Jetpack Compose for Android)… or otherwise unsatisfactory (Flutter). The need for a new modern and liberally licensed declarative cross platform UI library with scenegraph based rendering engine is growing.

[zanny]

I wonder when/if plasma will take advantage of vulkan considering QT is pushing it.

The parts that are built on top of QML will benefit whenever Qt supports it. KWin and the desktop effect shaders will almost certainly stay OpenGL based and targeting 3.1 because there isn't anything to add switching APIs despite that effort costing thousands of hours. Every platform you would want to target will support the GL extensions for the current KWin effects.

The advantage of Vulkan is in heavy rendering workloads where the stateful overhead of GL and its validation layers costs a lot of CPU time. Nothing a desktop shell or window manager does is ever that computationally intensive.

[tului]

The parts that are built on top of QML will benefit whenever Qt supports it. KWin and the desktop effect shaders will almost certainly stay OpenGL based and targeting 3.1 because there isn't anything to add switching APIs despite that effort costing thousands of hours. Every platform you would want to target will support the GL extensions for the current KWin effects.

The advantage of Vulkan is in heavy rendering workloads where the stateful overhead of GL and its validation layers costs a lot of CPU time. Nothing a desktop shell or window manager does is ever that computationally intensive.

I agree but the optimist in me hopes that the reason UIs dont use anything but the basics is because of the previous limitations and now that they're lifted someone will make use of the additional capabilities.

The pessimist says that 3D UIs suck, and all the ones tried so far were meh at best. I suppose we'll see

[zanny]

I agree but the optimist in me hopes that the reason UIs dont use anything but the basics is because of the previous limitations and now that they're lifted someone will make use of the additional capabilities.

The pessimist says that 3D UIs suck, and all the ones tried so far were meh at best. I suppose we'll see

There aren't "capabilities" than Vulkan has over OpenGL. Both enable as fine grained pixel manipulation as you want. If anything, the lack of high level geometric functions make Vulkan much less useful for a window manager because you don't want to have to rewrite from scratch simple transforms commonly used throughout window system effects.

[lowflyer]

The way I see it, Qt Quick 3D is unfortunately another glimpse of the ugly direction in which things are headed more and more with Qt, that is: the tendency, since Qt was bought by Digia, to part from the nice LGPL road and release new parts only as GPL or commercial (the latter with an absolutely exorbitant pricing).
Back then, for Nokia (well, at least before the Microsoft shill Elop got injected there and destroyed the company)… Qt was a strategic asset and it made sense to open it up. For Digia on the other hand (and the newer "The Qt Company", subsidiary of Digia), Qt is seen as a core revenue source, and they are shifting the balance in that "freemium" model ever more towards the "pay to play" side.

My doubts about that strategy (from a user perspective, which of course differs from the perspective of those who make their money selling Qt) are such that I find myself starting to hope for a replacement whenever I see news about new UI libraries. Unfortunately lately the news about new UI libs were not cross-platform but platform-restricted walled gardens (e.g. SwiftUI for new Apple OSes, Jetpack Compose for Android)… or otherwise unsatisfactory (Flutter). The need for a new modern and liberally licensed declarative cross platform UI library with scenegraph based rendering engine is growing.

I quote the whole post because I'm not able to say it in any better way!

Our company adopted Qt about 10 years ago in order to move away from proprietary GUI libraries and today we see ourselves being pushed back to that place. We've stopped updating our builds with the latest Qt versions because of subtle license changes.

[carewolf]

I quote the whole post because I'm not able to say it in any better way!

Our company adopted Qt about 10 years ago in order to move away from proprietary GUI libraries and today we see ourselves being pushed back to that place. We've stopped updating our builds with the latest Qt versions because of subtle license changes.

Do remember though that Qt is bound by the FreeQt Foundation agreement with KDE. Qt can not change away from LGPL without consent from the open source community. Only new modules can be put under new license. This is specifically designed so that existing users do not need to fear license changes.

[lowflyer]

Do remember though that Qt is bound by the FreeQt Foundation agreement with KDE. Qt can not change away from LGPL without consent from the open source community. Only new modules can be put under new license. This is specifically designed so that existing users do not need to fear license changes.

carewolf , thanks. I was only half-way aware of that. But it is not the solution to our problems. The changes in 2016 - when they changed from LGPLv2 to LGPLv3 caused major headaches on our side and is the major roadblock for us to upgrade beyond Qt 5.9. I'm aware that you might say "that's not supposed to be a problem" - the issue here is that our lawyer needs to come to this conclusion.

That agreement makes me understand why the Qt company puts so little effort into the development of the existing Qt infrastructure and is, on the other hand, all bells and whistles when they announce a new module - which we can't use.

Going with the commercial option is also not a solution for us. We tried that already. As 3diStan already mentioned they have a quite stiff pricing plan (It's comparable to the likes of Presagis, Carmenta and Rockwell-Collins). With the commercial licensing you are bound to the Qt version "you bought". The Qt company is equally lazy with the development of Qt whether you have a commercial license or not. Their support however is OK - not great - but OK. But they ignore you when your license runs out. It would be nice if they would at least politely move your issues to the public bug-list.

Add these problems to the constant problems we have with Qt Creator and qmake and you'll see that I'm fighting a loosing battle against all the others in the company that push for Microsoft.