AMD Starts Rolling Out New Linux Driver Model, But Many Issues Remain
An anonymous reader writes: With the upcoming Linux 4.2 kernel will be the premiere of the new "AMDGPU" kernel driver to succeed the "Radeon" DRM kernel driver, which is part of AMD's long talked about new Linux driver architecture for supporting the very latest GPUs and all future GPUs. Unfortunately for AMD customers, there's still much waiting. The new open-source AMDGPU Linux code works for Tonga/Carrizo GPUs but it doesn't yet support the latest R9 Fury "Fiji" GPUs, lacks re-clocking/DPM for Tonga GPUs leading to low performance, and there are stability issues under high-load OpenGL apps/games. There's also the matter that current Linux users need to jump through hoops for now in getting the code into a working state with the latest kernel and forked versions of Mesa, libdrm, new proprietary microcode files, and the new xf86-video-amdgpu user-space driver.
Not sure why AMD and nVidia keep dragging their foot. It makes no businesses sense
There's also the matter that current Linux users need to jump through hoops for now in getting the code into a working state with the latest kernel and forked versions ...
Seems the usual way Linux works.
Yeah, running "apt get update" which breaks the install and requires a full rebuild of a web server is such a good idea for a OS. It happened to me. I have yet to do an install or upgrade on a Linux server that does not refer to packages that I do not have or which may be incomparable with packages I do have.
I use Linux only because I have to (IIS is even worse).
Isn't it sad that Linux is getting ROFLstomped by both Windows 8 (the most hated Windows since ME) and Windows 10 ( a fricking beta that isn't even RTM yet) as well as fricking "other"?
Of course shit like this is EXACTLY why the Hairyfeet challenge has lasted 8 years and why every other FOSS OS has a stable driver ABI including BSD and Android, because if you don't? You get shit drivers and updates break drivers, simple as that. I mean what did you THINK would happen? How many kernel devs are devoted full time to drivers? 30? 40? lets be generous as hell and say 200. So you have 200 devs and over 100,000 drivers with hundreds more coming every.single.quarter. and because there is no stable ABI all it takes is a pointer change in the right spot and thousands of those drivers are now worthless, they DO NOT WORK.
It don't take Hawking to do the math folks, Torvalds and his "let the devs handle it" is a classical mythical man month situation that will never ever improve, it just can't because you will NEVER 1.- Have enough devs with low level driver experience that are, 2.- intimately familiar enough with the hardware to make more than generic (at best) drivers even if you give them the code, 3.- Nor enough hours to go through the tens of thousands of drivers when a major change causes breakage and finally 4.- No way will they have access to all that hardware to test.
Of course all that goes away with an ABI, which is why Apple (OSX and iOS),BSD,Haiku, Windows, hell even OS/2 Warp has a fricking ABI folks! And the one argument they always bring up, the "they won't give us the code precious" BS? Yeah what is the GPU that every.fricking.Linux.article. advises people to use? Nvidia which don't give you the code, never will give you the code, yet its always the best driver of the lot!
Its been 22 damned years folks, and Linux is lower than "other"...doesn'ty that tell you something? And you can't even bring up android as Google fricking HAS an ABI for Android! If Torvalds model worked then others would have adopted it...they haven't, not even the other FOSS OSes. Its time to face the facts, which is the kernel devs should be focused on the kernel and the ones that actually make the hardware should be focused on the drivers! But that isn't gonna happen if they have to rewrite the damned things yearly because of Torvalds fiddling, not when a driver written for Vista will run in Windows 10. You can't dictate terms when your OS is at 1.6% and trying to has kept the OS dead last for 22 years, isn't it about time to stop sticking with a failed model?
ACs don't waste your time replying, your posts are never seen by me.
Is it worth investing hundreds of thousands of dollars on software to support tens of billions of dollars of research and development in a hardware products evelotion. Why yes, yes it is.
If you dont understand the benefits of open source why are you even posting here ?
Stick to what you know, hairyfeet. Linux has no ABI because it does not want to encourage having random binary blobs on the users' systems. It's an ideological principle that will never change.
You clearly only understand the desktop arena, which is fine, but most Linux users are pretty happy with its current niche. The Linux ecosystem is probably better described as a set of tools for building an OS, and so you see things made from it like Android, Maemo/Meego/Sailfish/Whatever, SteamOS, various container-style projects, and the normal run of distributions. It's a development platform, mostly for server applications. Yeah, it would be nice in some ways if everyone was programmer enough to use Linux, but it's essential to no one, and the minority of people who have aftermarket video cards is not sufficient justification to bend over backwards for closed development models. Gamers may keep you in business, but they are a tiny part of the computing market. Now if you could point to something like ABI compatibility being an issue with GPU supercomputing, that might be more compelling.
There are arguments for a stable ABI. They are never going to get traction in a very successful open development paradigm, and desktop market share is not one of them. I'm sorry you've been riding this hobby horse for however many years, and I hate to tell you how silly you look doing it, but if that's your thing I guess you can keep it up until doomsday.
Those who advocate genocide deserve every protection afforded by law, and none afforded by common human decency.