AMD's Modern Graphics Driver In Linux 5.14 Exceeds 3.3 Million Lines Of Code

[mdedetrich]

mdedetrich

"Thankfully unused portions of these header files... are eliminated by the compiler at build time"

Pity you didn't notice that the entire time I am not counting the headers (which I have repeated about 3 times and you have conveniently ignored constantly). But yes, keep on fighting that strawman

[tomas]

bridgman

"SOMETHING," bellowed Steve Dallas at a hastily called town meeting, "SHOULD BE DONE!"

Heh heh. Best comment ever. Spot on. 😁

[illwieckz]

Wow some people here can really be a bit thick/dense and not read between the lines. Phoronix articles are either made because the author/s think

1. There is a new interesting feature (not this case)
2. There is a some bug/critical issue/security problem (not this case)
3. Some benchmarking of hardware (not this case)
4. Deprecation of something (not this case)
5. Some concerning treand/issue (which is this case). Other examples including Steam Linux usage charts, distro/desktop usage charts/, increasing/decreasing hardware requirements for software etc etc

I mean the article headlines is literally "there is a huge bunch of AMD graphics code in the kernel". You can argue till the cows come home whether this is just an "observation" but

1. Its not the first article on this point
2. No one is going to make an article about an observation unless its pointing some positive (or negative trend) or that its entertaining (not sure about you, but the line count of the kernel isn't exactly entertaining in itself).

Someone being there from some year may have noticed Michael has some pet subject and may like to focus on this or that just for pleasure, and that's very good, really. I hope that will not change. This is also a part of the Phoronix color, part of why people like Phoronix. The fact someone tends to put more attention on this or that can also just be about passion, taste. You cannot assume it's necessarily a concern, this can also just be for pleasure or because that's just his current mood, which is super fine. The would would be insipid if everything was driven by outside forces, this is super good Phoronix has its own freedom to study his own topics just for passion, fun, pleasure or anything I don't know, also, that means Phoronix sill has the freedom to decide what's the topic in town, not being a follower, this article shows Phoronix can decide what's to be the trendy topic for the day, whatever your opinion on the topic.

It's funny that in the 5 points you list, 3 are about issues, deprecation or concern, while you completely miss the point it can be for passion, pleasure, or just an own Michael-specific interest we can be glad he has. If no one had specific interest, then there would even not be any generic interest… It's funny you quote the sentence “there is a huge bunch of AMD graphics code in the kernel”, there is absolutely no good and no bad in that sentence, at all. For example the “huge” adjective is not negative per se. For example some people are free to see this as awesome. For example if one day Phoronix publishes an article named “Nvidia just freed its GPU source code, prepare for huge amount of lines to land in kernel repository”, this would be seen as awesome by many people and no one could assume that hugeness to be bloat until someone says it's bloat by looking at the code shape itself and just not the size. Same if someday it happens that Phoronix announces that Steam survey displays a “huge” Linux gaming market share, that may not be seen as a necessarily negative…

[mdedetrich]

Someone being there from some year may have noticed Michael has some pet subject and may like to focus on this or that just for pleasure, and that's very good, really. I hope that will not change. This is also a part of the Phoronix color, part of why people like Phoronix. The fact someone tends to put more attention on this or that can also just be about passion, taste. You cannot assume it's necessarily a concern, this can also just be for pleasure or because that's just his current mood, which is super fine. The would would be insipid if everything was driven by outside forces, this is super good Phoronix has its own freedom to study his own topics just for passion, fun, pleasure or anything I don't know, also, that means Phoronix sill has the freedom to decide what's the topic in town, not being a follower, this article shows Phoronix can decide what's to be the trendy topic for the day, whatever your opinion on the topic.

You are kinda scraping the bottom of the barrel here, this is obviously not an article about Michaels dog. Right now you are arguing for the sakes of arguing

It's funny that in the 5 points you list, 3 are about issues, deprecation or concern, while you completely miss the point it can be for passion, pleasure, or just an own Michael-specific interest we can be glad he has. If no one had specific interest, then there would even not be any generic interest… It's funny you quote the sentence “there is a huge bunch of AMD graphics code in the kernel”, there is absolutely no good and no bad in that sentence, at all. For example the “huge” adjective is not negative per se. For example some people are free to see this as awesome. For example if one day Phoronix publishes an article named “Nvidia just freed its GPU source code, prepare for huge amount of lines to land in kernel repository”, this would be seen as awesome by many people and no one could assume that hugeness to be bloat until someone says it's bloat by looking at the code shape itself and just not the size. Same if someday it happens that Phoronix announces that Steam survey displays a “huge” Linux gaming market share, that may not be seen as a necessarily negative…

Well Linus, the creator of the Linus kernel disagrees as he has publicly stated that he is concerned about bloat/size of the Linux kernel and he said its a problem. So you can take that as you will, but I trust his statements more than commentators using weasel word tactics to try an argue about an obvious common sense trend.

BTW I am not blaming AMD on this, they had no choice in the matter. If they want an open source driver in Linux, they are pretty much forced to mainline it because of how its mandated.

[illwieckz]

To be frank as a software engineer I would disagree here, having giant monolithic systems with so many lines of code invariably ends up causing problems. Even Linus himself agrees that he is concerned about the code bloat in the Linux kernel, and that was in 2009

If such big repository may require specific care (for example, the cost of cloning is not low, that can be a problem), if one day some Linux big head decide to turn big parts of it (like such GPU code) as submodule, that would be a worst concern. At this point there is no better solution than submodule for repository repository, and I have hard time to write “no better” in such sentence because it may sound good while submodules are a deep pain to deal with.

At this point, dealing with multiple modular repositories and keeping them in sync would be so painful… That would not solve any problem, just bring more painful ones…

Some aspect of Linux repository may not be perfect, but it looks like other solutions are just worst…

Cost of submodules is higher than “the big monolitic repo”.

Cost of stable ABI is higher than “the big monolitic repo”.

[illwieckz]

You are kinda scraping the bottom of the barrel here, this is obviously not an article about Michaels dog. Right now you are arguing for the sakes of arguing

You just noticed you missed a very strong thing about how a business is done (it's driven by people with people behaviors), so you decide to ignore it. =)

[mdedetrich]

If such big repository may require specific care (for example, the cost of cloning is not low, that can be a problem), if one day some Linux big head decide to turn big parts of it (like such GPU code) as submodule, that would be a worst concern. At this point there is no better solution than submodule for repository repository, and I have hard time to write “no better” in such sentence because it may sound good while submodules are a deep pain to deal with.

At this point, dealing with multiple modular repositories and keeping them in sync would be so painful… That would not solve any problem, just bring more painful ones…

Some aspect of Linux repository may not be perfect, but it looks like other solutions are just worst…

Cost of submodules is higher than “the big monolitic repo”.

Cost of stable ABI is higher than “the big monolitic repo”.

The whole point of lets say making a stable ABI and moving the graphics driver outside of the kernel is that the Linux developers dont actually need to care about the driver. They need to care about the ABI, and do frequent testing (to of course make sure there aren't regressions as new drivers are rolled out) and every now and then also update the API to add more features (ergo VR or HDR).

Do you think Microsoft is the primary developer of AMD/NVidia's graphics driver for Windows? Or in Android that Google is the primary maintainer/developer for the various proprietary graphics drivers on ARM phones?

This isn't about moving code around (if that wasn't already clear by now).

You just noticed you missed a very strong thing about how a business is done (it's driven by people with people behaviors), so you decide to ignore it. =)

Congrats on making the most meaningless wishy washy statement that provides absolutely nothing to the conversation.

[bridgman]

You are kinda scraping the bottom of the barrel here, this is obviously not an article about Michaels dog.

Michael has a dog ? Dammit, there should have been an article.

BTW my take on why the article was written is really simple - it's interesting how large & complex modern graphics drivers are. When I joined ATI after 20 years of working on graphics hardware & software I was astounded by the size of a modern Windows GPU driver.

[mdedetrich]

Michael has a dog ? Dammit, there should have been an article.

BTW my take on why the article was written is really simple - it's interesting how large & complex modern graphics drivers are. When I joined ATI after 20 years of working on graphics hardware & software I was astounded by the size of a modern Windows GPU driver.

Agreed, GPU drivers are insanely massive hence why I think its important to ask and question whether we should treat them the exact same way as keyboard drivers. The only two things they have in common is that they are a "driver" but in any other conceivable way there is a world of difference between them.

Hence why I am also highly confused why people are making claims that maintaining a supposed stable ABI instead of a full fledged driver (like Windows does) is going to mean more maintenance and work for the Linux kernel devs. The type of work is obviously a bit different, but thats honestly just unsubstantiated hyperbole/FUD.

[tomas]

mdedetrich

Pity you didn't notice that the entire time I am not counting the headers (which I have repeated about 3 times and you have conveniently ignored constantly). But yes, keep on fighting that strawman.

Well, it is you that claim that the mere existence of this article must imply that there is an issue regarding the code size of the AMD GPU driver in the Linux kernel. Or that it implies having the drivers as part of the Linux kernel being a problem. Even though you can still not point to anything specific in the article to support that claim. Just lots of hand waving. In fact, if it's not about "counting the headers", then your whole argument about the article falls flat, since it is exactly about the headers constituting the majority of the code for the AMD GPU drivers.

Now, you are of course perfectly entitled to your opinion that Linux should have a stable driver ABI and not ship drivers as part of the kernel. But don't try to support that opinion based on this article, because it is not there.

BTW, you are confusing having an ABI for drivers with "modularity" and the kernel being "monolithic" or not. Having the drivers shipped separately from the kernel or not does not change the "monolithic" aspect of the kernel. They would still run in kernel space with the same privileges as the existing in-kernel drivers. And as you are well aware of, drivers for Linux are built as kernel objects that are loaded on-demand, so in that sense it is not a classic "monolithic" kernel.

I understand that the BSDs have a stable driver ABI, no? How is that working out for them? Do they enjoy a lot of driver support from commercial vendors because of this? 😊