Any updates on Hardware Accel? It’s so hard not even being able to play basic games like half life (original) due to low frame rates. Major issue for haiku being generally use able (I am still daily’ing it). Just wondering if any works been coming along or if anything is being looked forward too on the GPU accel side. I don’t have much programming experience but do have Linux and BSD experience, if anyone needed help on a port I’d gladly help with a little guidance. Be on guys!
The devs are working on it it’s not quite ready, yet NVIDIA is kind of a pain to make drivers for to since they’re a bit reluctant to open their source code, but I think intel drivers are fully supported
That was true a few years ago,nowadays it’s quite the opposite.
Nvidia provides open-source drivers for Linux for their newer cards and they’re comparably easy to port since the largest part of the functionality runs directly in the firmware of the card and the open-source driver only needs to interface with that firmware.
Intel drivers (and also AMD) are better documented,I think,but all code needs to run in the driver rather than the firmware,so it’s more code that needs to be written,much work for such a small team.
Intel drivers in Haiku allow setting the screen resolution and brightness for a large part of GPUs (not all of them),but they don’t have hardware acceleration at all.
I know, but It’s still somewhat of a pain in the neck from what i heard there are some things that still need tweaking with
It is right that after Nvidia released source code of their kernel driver, Nvidia become easiest target to support 3D acceleration on Haiku, but closed firmware thing is unrelated. Most important factor is that Nvidia kernel driver is written in a way that makes it easily portable. So all interfaces for communicating with OS are abstracted and collected in single place. Code itself is OS-independent at is trivially buildable on any UNIX-like OS. Even if Nvidia released kernel driver with all hidden code that later moved to GSP, it will change nothing in portability.
Rumors about bad compatibility of Nvidia GPUs with open source are largely Linux-specific thing and are unrelated to alternative OSes such as Haiku. It is Linux ecosystem fault that they do not provide proper stable APIs and specifications for GPU drivers. They also make hard to develop GPU drivers that are out of tree of Linux kernel and Mesa.
oh so thats why
This may be the case now, but a few years ago the situation was:
- Intel: provides full documentation of the hardware, making it possible to write our own driver
- AMD: a lot of the hardware initialization and modesetting is done in “atombios”, that is bytecode stored in the videocard ROM. The driver mainly needs to provide aneinterpreter for the bytecode and call the relevant functions. Also, AMD engineers had a public support channel where we could ask for help from their engineers
- nvidia: no documentation, no open source driver published, nothing. You have to reverse engineer the hardware or closed source drivers to understand what they do.
This info is obsolete now, especially on nvidia side since there is reference code in the linux driver and it is clean and portable. But the rumors were fully justified a few years ago, and explain why Haiku didn’t get an nvidia driver at the time.
I think that there are multiple factors but alternate systems users including linux users are expecting hardware to have a longer life span. Closed source drivers are a no go; a company can’t maintain half a dozen of drivers eternally.
If it is so difficult to provide drivers for linux, companies should provide open source drivers for Haiku. Linux devs would have to adapt them but they have a lot more manpower that we have.
This do not help a lot with 3D acceleration, only modesetting. There are no detailed 3D engine initialization and operation documentation at all and only basic documentation only for old Cayman/Evergreen GPUs. AMD employees just commit code to Linux kernel without documentation.
Where such support channel exist? #radeon
IRC channel is mostly silent. I experience much more help from Nvidia engineers.
When AMD took over ATI, the state of linux drivers was well … a complete disaster. As their plan was to follow intel lead and integrate the chipset in cpu, something had to be done to restore faith. They reached their goal, taking few customers to NVidia, it’s no more a priority.
NVidia had to react, and offers probably better support now. IMHO, it won’t last. This is cyclic and depends of company goal at a time.
You will have to ask kallisti5. But, that was the situation a few years ago, things are different now on both nvidia and amd sides. So now is the right time to make progress on the nvidia driver
Yeah, I’ve heard Nvidia’s support for FreeBSD has been historically pretty good, which is weird they worked better with FreeBSD well before Linux.
Not only that,also Solaris/OpenIndiana.
Nvidia provides first-class proprietary drivers for almost any card for Solaris,and since it’s binary-compatible with Illumos,those also work on OpenIndiana.
Works better than Intel or AMD there,with hardware acceleration and stuff.