AMD Contributes 8.5x More Code To The Linux Kernel Than NVIDIA, But Intel Still Leads

[Ardje]

I love kernel stats. But they aren't the only thing that count. So always keep that in mind reading kernel stats. AMD does a lot of code and opening up for their rivals to use on other fronts as well.
The vulkan driver is not part of the kernel for instance.
And sometimes it is not always about code, but also financial support. Code support is usually when you want something to work on your hardware. Financial support is when you want somebody else work on whatever they work on, and maybe fix things for you (like dxvk).
So yeah, while kernel commit and line stats are generally nice to read, it doesn't paint the complete gnu system.

[pal666]

Lines of code is hardly a metric you should be proud of. If anything, with mature codebase, you should be removing lines, rather than adding them.

mature driver codebase for new hardware? what are you smoking?

[pal666]

NVIDIA does seldomly contribute to nouveau here and there as far as I know.

you are mistaken. those "contributions" are trivial and not related to desktop

[pal666]

As for Nvidia, their focal point is their blob driver and that lives outside the upstream Linux kernel,

that means "nvidia does not support linux"

so why compare their lines of code to AMD or Intel?

because that is what matters here, not whether nvidia decided that it will only do windows driver

[perpetually high]

pal666 thank you for the correction. Then I see no problem with people calling out NVIDIA. Doesn’t even matter discussing LOC, code quality, or number of commits

[pal666]

You can add all the code in the world but it won't mean anything if it's bad code.

code in question works for majority of users very well, so it is infinitely better than no code at all from nvidia

And yes, the AMD stuff is crap like it always has been since it was ATI.

As I type this right now FreeCAD, Slic3rPE, Wings3D, and Blender are all flickering and glitching all over on my other monitors (using Radeon 280x on Ubuntu).

my previous card was 280 and it had one visual glitch: small square with randomly shifted subparts which could appear on some surface, but since it was random, it was easy to fix by restarting app(it would appear at the start and stay there until exit). on new card i've never seen it so far, so i assume it is either hw bug on si or driver bug on si only. or maybe only my card was broken. but i never used apps you mention, maybe there are more issues with them. or maybe you use wrong driver(amdgpu-pro) ?
i thought about it a bit more and i remember another issue: flickering in hitman, which went away on new card. so now we have two issues if they are unrelated(though since hitman is only one app, it can be issue with hitman, not with driver)

They can't write OpenGL drivers worth anything!

well, their driver is better than nvidia windows blob, so they are certainly worth something.

I myself have developed OpenGL apps against AMD cards and what a steaming pile. For anything serious I always revert back to nVidia.

this is stupidiest thing to do as a developer, nvidia driver is both closed-source and non-conformant

[pal666]

Why are we acting like shitty code is just routinely merged in?

because most participants have no clue and just parrot something they heard somewhere

[pal666]

There's absolutely no fallacy in saying that LoC is almost not a metric. The fact that AMD has to add headers every now and then means they have a lot of hardware design changes, and a lot of different SKU's - how is that nVidia's fault, or AMD's benefit in sense of contribution to kernel, it's hard to derive.

what are you trying to imply here - nvidia has less skus than amd or nvidia has less design changes than amd? all i'm getting from you is you are butthurt nvidiot

Adding header specs can hardly be called anything more than a boilerplate effort, but it's more than welcome.

imbecile, people who develop drivers for nvidia hardware, would be happy to see header specs from nvidia. it is not boilerplate, it is priceless information. boilerplate is code you have to repeat, but which brings no information

Now, in 200k of C code you can say really a lot. And 200k is hard-ish to maintain as well. I've never seen a production code that couldn't benefit from refactoring. And I've worked on code for repositories ranging from 10k, 100k to several millions lines.

so you've worked on project with several million lines but are bitching about project with 200k lines? why didn't you refactor your project from several million lines to less than 200k lines? your stupidity is funny. btw, linux kernel is 20 million loc

And yes, it's modern to s*** on nVidia, but while you're at it, don't use LoC. Use number of features added, percentage of SKU gamma covered by os driver, test coverage, performance results, review issues count... Almost anything but LoC, I'd say.

what makes you think any of that is better metric than loc?

[pal666]

It's more than possible. I've made a POS terminal support another ECR protocol by consolidating other supported protocols with a total negative LoC outcome. No support was broken, we have unit tests for that

it's only possible because your existing code was shit. now consider possibility that linux existing code is better than that

[pal666]

AMD has changed a lot of architecture styles in a last couple of years VLIW, GCN, you name it.

shader compiler is not part of kernel driver (what you named are isas)

It might be that nVidia wouldn't require so much code to properly support their GPU's.

it might be that nvidia would require even more code to properly support their gpus. how does that improve your argument?

There is more than enough proper metrics that could show that

i'm sure michael will take your article based on proper metrics, then we can discuss its fallacies