First i want to say this has been incredible progress. I find beta 5 stunning in many respects. To those who saw haiku in 2020 to now, it is absolutely distinct.
Yet it brings me some level of worries that the gpu acceleration and vulkan initiatives and gpu drivers are taking quite a lot of mindshare, and not because i think they are not fair and useful, but, i have some worries that this may make it the first-class rendering method and attention to devs. I know everyone generally works in what they feel pleasure or feel as useful as a lot of devs are not paid to write code, but the level of integration, the unique performance, power efficiency and general function of current software rendering is amazing, considering it and the whole system being not optimized yet.
Also it allows far many more machines to be fully compatible, it removes the need for a gpu, and it reduces bugs to users and frees more available memory to those who do not run dedicated GPUs.
How are the plans about this? are both forms of rendering systems going to be equally focused by the core devs? or is gpu rendering going to take the front when ready? Are we going to be able to turn it off?
I think the context is important here… right now no one is actually submitting changes for hardware rendering. And it’s a huge piece of work to implement, even for one type of GPU, never mind for all the major cards out there. The chances of it happening are still quite low, so the chances of it becoming the de-facto standard for most haiku machines is also very low, and the chances of those things happening and then software rendering being forgotten about as well are vanishingly small. So I wouldn’t worry about this!
As far as I know, the only developer really working on it is x512.
And I don’t think there are any plans to remove the software rendering support. After all, we still need to support some machines where there will be no accelerated drivers, as well as some where there is no acceleration hardware at all (or unsuitable for what we would need).
hardware acceleration would mostly be used for 3D things: videogames, CAD software, …
Some chnages are needed for the 2D architecture as well, in particular, it would be nice to implement a true double buffer system to eliminate screen tearing and partial redraws. Maybe later we could venture into accelerated compositing.
Anyways, outside the forum discussions, I don’t think this takes a lot of the “mindshare”, as you say. For me this is definitely a “to look into later” thing. There are a lot more important things to fix first.
I recently resumed active work on porting NVRM+NVK drivers to Haiku and it looks promising. Mesa NVK Vulkan driver added OS abstraction layer that simplified using Nvidia NVRM API instead of Linux Nouveau DRM API. I initially planned to make kernel side NVRM → Nouveau wrapper HDRM (Haiku DRM), but it is no longer needed, so I switched to adding NVRM support directly to NVK Mesa driver.
I will make announcement forum topic when I make some significant progress understandable by regular user such as running VKGears demo on NVRM+NVK drivers ported on Haiku.
This is a very refreshing view and i think it will well utilize one of the unique strengths of this platform. The software rendering is really outstanding and to hear it is going to be improved and changed and further developed to aspects i cannot anticipate is really good news.
I agree. I do not see any reason to think about this. I think i had an oversight too ahead in time. I think this is the first era i can completely switch to haiku so i think i am thinking too far ahead without much actual basis for any of the concerns.
