AMD GPU Linux Driver Becoming "Really Really Big" That It's Starting To Cause Problems

[Etherman]

The hardware doesn't have to be that old. While HDDs are getting increasingly rare in laptops, I could see how a 5400RPM drive (which was relatively common) could struggle to keep up.

In any case, part of me wonders how much of this GPU code needs to be in the kernel and which could be pushed to userspace.

Isn't the module commonly in the initramfs and as such is loaded into a ramdisk by the bootloader before the kernel is run?

[Daktyl198]

Has Plymouth ever worked properly? I've tried it on at least 5 different systems, and on all of them it doesn't start fast enough to cover the initial tty output, so you see ugly text anyway. Then it quits too soon and you see even more tty output before the login manager shows up.

And that's not including the delay to startup it causes. One of the worst projects to exist by far.

[Daniel15]

This is why I like nvidia's design of putting all the code in the GPU itself and the driver is pretty much a thin wrapper over it.

The real reason they did this was so they could say they have an "open source driver" (https://github.com/NVIDIA/open-gpu-kernel-modules) when in reality all the interesting bits are in the closed-source firmware binary blob.

[NateHubbard]

Nouveau code base is a fraction of the size and supports 14 years more hardware. Just saying.

Yeah, no.

[r1348]

I was seeing the 3-dots issue on an old Kaveri laptop (HP Elitebook 755 G2) with a SATA SSD, but nothing more recent, i.e. my daily laptop has a 2700U and a NVMe SSD, and has no issues booting the Plymouth theme.

[heim]

I know that Windows has hybrid shutdown since Windows 8, which speeds up boot time after shutdown. I don't think even MacOS has something like this.

It's called FastBoot and it's fast because it's doesn't really boot the kernel. It essentially hibernates the kernel and restores it upon boot.

[fitzie]

1. With older AMD GPUs the radeon driver is actually used to drive the GPU but even though it is unused the amdgpu driver still loads slowing things down.

i do appreciate what amd's done for opensource drivers, but at the same time, it seems like they've done a 90% job. it's great that we get day one hardware support, and things are generally smooth, but somethings just seem broken. no reason for a driver to take 10 seconds to initialize when it's not going to manage the gpu, or this 6 million lines thing is really a joke. don't get me started on rocm, and the lack of fuzz testing there.

That being said, if the amdgpu driver is unused then the system shouldn't bother loading that driver. i'm all in to blame amd for having a bloated mess of a driver, but it's the distro that is trying to load it when it's not appropriate.

[scottishduck]

Nouveau code base is a fraction of the size and supports 14 years more hardware. Just saying.

[sophisticles]

All these lines of code, their coming in by the millions and millions, no one has ever seen anything like it, and it's a very bad thing. What's happening is a disgrace, it's happening in the boot up process, just look what's happening to computers all around the country with these millions and millions lines of code, it's a disgrace. Just the other day there were reports of all these lines of code eating people's boot splash screens, there's too many lines of code and the system can't handle it, people are being forced to suffer with 3 dots across the screen because all of these millions and millions of lines of code are eating the boot screens and it's a very bad thing, and I will tell you what, this never would have happened if everything wasn't open source.

[Weasel]

Michael just said the bulk of the six million lines, which he'd linked, is from the auto-generated headers. But those headers could very much be holding a large amount of config data for the wide swatch of GPUs supported that then all ends up in the compiled binary.

Ok? And you can look at the compiled output.

The size of the headers is completely irrelevant when it comes to loading the driver. You're not compiling shit when loading the kernel and its drivers modules.