Hi, I logged in on purpose to ask this question:
How would the Vulkan way be, and then zink for the opengl on it? but also consequently the direct x with DXVK , the DXVK dll are so good, that dlls compiled to work in wine even work on windows, running direct x games and applications on Vulkan. My knowledge ends here, but I think given the existence of wrappers to run openGL and directx on vulkan, the best choice would be to develop Vulkan directly and leave everything else, besides being a much younger Vulkan, I believe there is much less code to manage and definitely better rewritten. Lately vulkan has been written for the Raspberry Pi 4 GPU V3DV Vulkan Mesa driver, and I think that if you should follow a porting route, the best thing would be to invest time on this gpu/Raspberry pi, being cheap, there would be a transfer of real users (and developers) who would use the Raspberry Pi 4 with accelerated Haiku …
There are so many people who owns a Pi but nobody stand up to finish the Haiku port. Maybe it is just a HackerNews thing, but it seems everybody want a Pi port, but nobody want to make their hands dirty.
The last Pi I bought was a version 2. The 4 is much better but I already have several ARM machines sitting around already including a Cubox and an ODroid XU4 that I don’t see the need for another one unless it happens to need the 64-bit architecture and 4-8 GB of RAM.
I tried to build an ARM version of Haiku but the steps in the process didn’t get me close enough to figure out what was wrong.
Those graphics drivers for the Pi are for Linux. They likely cannot be ported to Haiku without a significant effort. It’s been said before, time and time again, that Haiku is not compatible with Linux. Especially not at the kernel level.
@extrowerk
I think we are in a sort of dog that bites its tail here, developers do not invest their time on haiku because it lacks that minimum to make it palatable, a good acceleration and a good complete browser to make it fluent.
@SamuraiCrow I know that haiku is very different, but if I were to invest in the development of a driver, at least I would start from what there is, and I would invest in the raspberry pi for two reasons, it is a complete machine, it is very cheap and there are already a lot of developers who would immediately start working on it … well as soon as the porting of haiku to rasbperry pi 4 is complete, I suppose this thing should start happening anyway … but not having a ready acceleration will make everything less attractive and slower.
Having a well-functioning cheap platform will consequently speed up the development of code suitable for other more powerful platforms such as x86 etc … in short, personally I think that the support for raspberry pi 4 is a potential gate for opening new energies.
(and having seen the performance of vulkan on an accelerated raspberry pi 4, I already know without a doubt that Haiku with Vulkan would become wonderful and attract a lot of people)
Believe it or not I actually agree with you. The fact is that drivers are not the only problem. Needing a framework for the drivers to plug into is as desperately needed as drivers themselves. Vulkan alone is sufficient for the needs and performs well. The problem with ARM devices is that the drivers are written either for Android or Linux. Until or unless Fuchsia replaces Android with a newer, simpler kernel than what Linux has, writing a cross-compatibility layer for existing drivers won’t be available because Android uses the Linux kernel internally.
Unfortunately my technical knowledge, especially on HAIKU, ends here, so I don’t know how to answer, but I hope someone knowledgeable will take my suggestions positively.
I get to think of this for practicality, having a well working Rasberry ready, put on HAIKU on an SD on which to evolve gradually, it’s like it would be in the old days with an Amiga, I put on the SD card with HAIKU and work on it without worrying about doing any damage (both hardware, if I break something, who cares, I buy a new one. both for the software) I can experiment deeply, it seems trivial, but it is not at all it is even better than a dual boot, because one experiments without fear of having a broken system. (at booting level etc …) I turn on the machine and I am inside the system … as if I changed “floppy”. This as you obviously guess makes everything extremely fun and productive … it is already the secret of raspberry success and with Haiku, it would be just the icing on the cake, I have been using linux for years, but honestly I have never been really happy with it, while never ceasing to thank that it exists.
I do this thought, but I think it echoes in the minds of many people.
We can talk here hour about the pros and cons, but we cannot say anything new what havent said earlier. Use the search function to see.
There are simply just the wishful thinking available to support the pis. Nobody ever stood up, and nobody ever invested money into it.
You csn call it as Catch of 22, or anything, it still doesnt change the fact that they only want to enjoy it “i would boot Haiku on my Pi which collecting dust in the drawer” (check the announcements, they using exactly this words on other forums).
So it boils down to nobody care.
But feel free to start a bounty, pay a developer or motivate somebody to do it.
With a super cheap machine fully supported with acceleration and and a good browser things will change fast. Because when you have that minimum that makes the environment comfortable enough, then you can appreciate what Haiku has of itself, of its completely unique and comfortable user experience, then you can work more on it, you grow in it, you can invest time in it , and you can also invest resources.
Currently, especially a browser that does not make life on the internet easier for you, is a severe limitation, it is the real bottleneck that prevents everything else from flowing spontaneously. The browser is not the real problem, as much as the accelerated driver that allows good use on the web, and here we are on this thread. There are millions of raspberry pi 4 just waiting to mount “a good conscience” and the rest will be a sea that opens up … I am not a rasberry fun, I would much prefer to see Haiku blow like the wind on my PC with a latest generation NVIDIA gpu … but I realize that it is far less likely to happen …
I am pragmatic.
I could also do it, but my few resources available and the current state will not encourage anyone else to contribute, much less any developer to take the job, which will be different when the platform is ready that needs to be optimized, or even accelerated.
I’ve been working a lot on ARM64 in past few months, it currently builds all bootstrap packages. So there is progress on ARM64, just not as quick as RISC-V or in any interesting parts yet.
I know that, and sorry if i formulated my words too harsh. I mentioned the people outside of the Haiku developer team.
This seems to be also a idealized way of thinking. I can’t recall any huge roadblock to contribute even on halfway supported hw. If somebody want to do something, will do something.
Also now we are talking about not just 3D but fully supported webbrowser is mandatory. But Haiku have limited manpower even for the webbrowser.
Maybe things will change in the future, maybe won’t. We can blame everything, ourselves and others and linux for gatekeeping as an mediocre but mostly working trainwreck. Nobody cares, otherwise we would use linux or other OS only for retrocomputing to see how silly we were back then.
Do you by any chance post updates of the progress somewhere? I’d love to keep updated for ARM as well…
Also, how can I create a currently upto-date image I can boot on a pi (which type? 4 only?).
I’ve been fiddling with the mmu some time ago (32bit) specifically (on real hardware) but I couldn’t get it to respond back then. I’m more of a bare metal person so to speak (never use emulation or so).
Thanks for putting time in this! It’s very much appreciated 
I did some tweets long ago, but once the efi binary compiles and some code is in I hope to share info on progress and build instructions in a topic of its own.
the rpi lacks in compute power, Arm, largely lacks compute power.
theres the RadV amd driver and while linux specific, I believe it probably could be leveraged to haiku if all tge nix crudt was chopped out.
RadV stripped with just a vulkan IR, could be a way. iirc memory management for shaders was a big todo
Sorry but I have the impression that you have completely missed the point of what I am trying to highlight in my comments.
The focus of trying to invest in the full support of a single platform, is a strategic issue, with the development of a single driver, such as the raspberry 4 gpu (in addition to the support of the arm platform in general which, as far as I see, is in the process of being advancement) could be a good investment by the HAIKU Team and it’s supporters, because with a relative minimum effort (compared to having to try to support as many other drivers as there are gpus on the market) you can reach thousands of users and developers who have raspberry surely they have PCs and Macs, and when they become more deeply passionate about HAIKU, they might think about collaborating in the project in some way. I have seen that even RISC-V support is proceeding with greater intensity, and this is certainly excellent news, especially for the future, but to see today, it would be more profitable to invest in a Raspberry pi 4 for the simple fact that there are already now millions of devices already marketed and owned.
Maybe not being able to explain myself well, I don’t know.
The raspberry pi 4 is powerful enough to allow decent web browsing and a decent desktop for everyday use, and this is a good start, especially for the spread it has, and because it can be purchased by anyone.
RPi couldn’t be used for actual desktop type pc work, like cad/cam, vudeo editing etc
The problem is that nobody has quite stepped up to port Haiku to any of the RPi models. Since Haiku is largely a volunteer-based project dependent on donations, the things that will be worked on are those which contributors are interested in. Forcing contributors to work on something that they’re not passionate about will only lead to driving them away, undermining Haiku as a project.
For what it’s worth, the RISC-V work may benefit ARM ports in the future as well.
i use my rpi 4 to control my cnc and i also installed freecad and it works fine, obviously you can’t expect to make a workstation-like use with gadgets of 50 buchs. But but this is not the point, the point is to be able to create a minimally comfortable environment for general everyday use so that you can try to prefer to use HAIKU instead of any other OS, at present this is not currently possible, at present you can access HAIKU for a few HOURS just for fun, you see some progress, but then it ends here, you are forced to turn off HAIKU and switch to the serious OS/machine, because there is not that minimum that is web browsing.
I have quite clear how the situation is, I agree with what you say, mine is just an attempt at strategic suggestion for the Haiku Team, at least for me it would seem a vision that could allow significant advantages.
I had a look at compiling things for the RPi about 2 years ago I think… I’d absolutely love to have Haiku work on my RPi 400, for example, as that would give me a whole retro “keyboard computer” feel to it. I could quite happily use it for day to day stuff, and I have a sense that Haiku (thanks to awesome developers) is generally super fast, so it would feel better than a big old Linux distro.
On the everyday work perspective, I used RPi 3B+ and 4s in a youth group to do day to day web browsing, online exams, word processing, and coding. We had a computer club that culminated in them getting code running on the International Space Station. (Astro Pi programme). So there’s plenty of value to be gained there even if it cannot perform higher end tasks - not many people do that, and if they do I get a sense most people will reach for a Mac anyway. (Controversial I know)
Again another thing I’d like to work on, if only I had the time.