Comments (17)
I think that we should focus on Linux. The Dev team is smaller than for the Windows version of the driver so supporting multiple operating systems seems like a bad idea. Nobody really uses Haiku or any other different OS (for daily driving) than Linux/Windows/MacOS so we should focus on developing the best drivers we can for Linux.
from open-gpu-kernel-modules.
I think that we should focus on Linux. The Dev team is smaller than for the Windows version of the driver so supporting multiple operating systems seems like a bad idea.
On the contrary: Linux users have been arguing for the importance of such support e.g. by pointing to the fact that there's not just Windows and macOS. If the community would now rush to just be part of the "big three" and be ok with starving the rest, this is a betrayal of the ideal of Open Source. And this would hurt the whole ecosystem - including Linux.
Nobody really uses Haiku or any other different OS (for daily driving) than Linux/Windows/MacOS so we should focus on developing the best drivers we can for Linux.
You are very wrong, sir (This comment was written on my FreeBSD workstation).
from open-gpu-kernel-modules.
Thank you for your interest!
Traditional NVIDIA drivers use the same binary blob (nv-kernel.o
) on FreeBSD and Solaris and only have a minor OS-specific adapter layer that is available in source form when you download the drivers. The new "GSP" driver architecture is also very amenable to this approach and we do hope to support them eventually.
Thus far, our focus has been on Linux only, and that is not likely to change for the next few releases while we add support for more consumer facing features (such as power management).
We will gladly work with the community if they decide to pursue this ahead of our own schedules, though. If anyone is planning on tackling this, I would strongly recommend reaching out to us first for guidance.
Thank you.
from open-gpu-kernel-modules.
I am also interested in Haiku operating system support that I use and contribute.
from open-gpu-kernel-modules.
Nobody really uses Haiku
Citation needed.
from open-gpu-kernel-modules.
We will gladly work with the community if they decide to pursue this ahead of our own schedules, though. If anyone is planning on tackling this, I would strongly recommend reaching out to us first for guidance.
Hello from the FreeBSD Foundation. We are very interested in seeing FreeBSD support in these open source drivers.
Although we don't have any plans to directly support this (with staff time) in the immediate future, we do entertain project grant proposals from the community. If someone is capable and interested in working on FreeBSD support in the near future please get in touch.
Foundation staff can also help with guidance and code review, either as part of a project grant or in collaboration with upstream.
from open-gpu-kernel-modules.
Nobody really uses Haiku or any other different OS (for daily driving) than Linux/Windows/MacOS...
You are very wrong, sir (This comment was written on my FreeBSD workstation).
I respect anyone who uses Haiku or other Unix like systems. What I meant is that Windows, MacOS and Linux are mainstream, where mostly only tech geeks use other systems. I'm not saying there shouldn't be support in other systems, I'm just saying that we should mostly focus on polishing current drivers on Linux.
from open-gpu-kernel-modules.
Nobody really uses Haiku or any other different OS
The funny thing is that I recognize in the comments at least 3 HAIKU excellent develpers full of potential,
One of these has already done the entire porting of HAIKU on RISC-V, he ported the AMD GPU drivers himself to run them on his RISC-V platform and since it was not enough, he ported the x86 AMD gpu drivers bypassing all the "junk" of DRI linux to collect better performance and better integration with the HAIKU philosophy, and directly supporting Vulkan and run Zink on top of it to have good OpenGL support and at the same time save coding and work.
I am sure he would give a big boost to the improvement of these nvidia drivers for all platforms..
I'm not a developer, but I'm a deep observer of this "no one used HAIKU"
I tell you just one thing, this HAIKU used by no one, is full of pioneers who look far ...
do not be obtuse.
from open-gpu-kernel-modules.
I respect anyone who uses Haiku or other Unix like systems. What I meant is that Windows, MacOS and Linux are mainstream, where mostly only tech geeks use other systems. I'm not saying there shouldn't be support in other systems, I'm just saying that we should mostly focus on polishing current drivers on Linux.
Fair enough. There certainly is more demand for great Linux drivers, I also have no doubts there. Especially with efforts from Valve and Google's Stadia. But I wanted to bust that myth that "nobody" was seriously using other operating systems. I've come a long way from my first Riva TNT card (didn't like the 3dfx hype back then) when I still ran Windows to kind of fighting with nvidia cards on Linux to finally finding the proprietary driver in good shape on FreeBSD when I made that switch several years ago.
I don't have the time to play a lot of games anymore and am more interested in other features like supporting multiple screens. And going with nvidia has worked very well for me in that regard. While it's not an ideology for me, I do prefer open source drivers and would love to see other systems supported, too. However on platforms where leadership didn't try as hard to piss off everybody who doesn't love the GPL (there are very valid reasons for that) and nvidia specifically, things worked well before. It's great that things are moving into the right direction now and I hope that this marks the end of certain hostilities. But it would be very unfortunate if that lead to support for other operating systems to fall short. But according to mtijanic's answer that does not seem to be the plan which is great!
from open-gpu-kernel-modules.
I hope that NVidia will keep current architecture with most driver part OS independent and per-OS support layers. That approach may be broken if NVidia decides to integrate driver to Linux upstream code base because Linux kernel maintainers may force to drop OS abstraction layer.
from open-gpu-kernel-modules.
The driver is now open source, the FreeBSD guys can do the same thing they are doing for AMDGPU/Mesa drivers and just write a layer so the driver will run as expected. I think this is the best choice for the "not mainstream OSes" since this is what happens since the begin. Linux is growing(thanks VALVe/Community) and the win that linux acquired will benefit all the other open source OSes by no doubt.
from open-gpu-kernel-modules.
Traditional NVIDIA drivers use the same binary blob (
nv-kernel.o
) on FreeBSD and Solaris and only have a minor OS-specific adapter layer that is available in source form when you download the drivers. The new "GSP" driver architecture is also very amenable to this approach and we do hope to support them eventually.
@mtijanic This is not at all my area of expertise, but wondering if an approach similar to how the traditional drivers are handled would be feasible for bringing CUDA (and other libraries like cuDNN) to FreeBSD? It’d be awesome to run AI workloads on FreeBSD in the future! 😄
from open-gpu-kernel-modules.
"may" is an understatement. As a FreeBSD user I would be very pissed off about the termination of an (official) support.
if an approach similar to how the traditional drivers were handled would be feasible for bringing CUDA
Tested with mixed results. Works for less ambitious things like OpenCL, though.
from open-gpu-kernel-modules.
"may" is an understatement. As a FreeBSD user I would be very pissed off about the termination of an (official) support.
I think this was in reference to “were handled”. It was not my intention to imply a discontinuation of official support for the existing, proprietary drivers. I’ve edited the response to hopefully reflect this. Apologies about any confusion.
from open-gpu-kernel-modules.
Eh, that part refers to the post containing the word "may". Hence the quotation marks.
from open-gpu-kernel-modules.
Ah misunderstood that, my bad!
from open-gpu-kernel-modules.
Great news about modules, but UNIX (*BSD/OpenSolaris/IllumOS--'genetic Unix') is mainstream for scientists/programmers (and still networking, at a time was vast majority) because, as originally designed according to plan (instead of 'make up as you go along') has cleaner/stabler software such as kernels. IllumOS is the new Open Solaris, but I couldn't care less about plain Solaris nor other proprietary Unix (though I hope it and Sys7 to Sys10 to UnixWare become Free/Libre/Opensource Software (FLS, OSS, FOSS, FLOSS.))
from open-gpu-kernel-modules.
Related Issues (20)
- Nvidia gpu not used
- Please Update Changelog for huge 550.90.07 code dump HOT 2
- PRIMUS - Card remains acrive - NVRM: krcWatchdogCallbackVblankRecovery_IMPL: NVRM-RC
- Periodic stutters and “NVRM: RmCheckForGcxSupportOnCurrentState” kernel warnings on Ubuntu 22.04 RTX 4070
- Strip: Assertion failed Thin LTO nvidia-open-dkms HOT 1
- Suspend sometimes causes a crash when using the open 555.52.04 drivers HOT 9
- NO SHARED MEMORY FOR YEARS [NVIDIA_UVM] - BASIC FEATURE HOT 2
- Failed to load module on boot HOT 3
- Failures to resume when sleeping (s0ix) on newer kernels HOT 9
- The screen show incorrect image when using custom EDID file on ubuntu
- can't install cuda (530 | 535 | 555) in my system HOT 4
- Very slow 3D performance after a while
- "NVRM RmInitAdapter: Cannot initialize GSP firmware RM" error found HOT 1
- Cufft GPU memoey increase HOT 1
- black background after certain time HOT 2
- Integrated GPU broken from environment variable. X Error of failed request: BadValue (integer parameter out of range for operation)
- IOMMU error on ppc64le HOT 3
- Vboost Print Output is Wrong HOT 1
- gpuHandleSanityCheckRegReadError_GM107: Possible bad register read HOT 4
- CUDA fails with CUDA call "cu->cuInit(0)" failed with CUDA_ERROR_UNKNOWN (999): unknown error
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from open-gpu-kernel-modules.