Nouveau Developers Remain Frustrated By NVIDIA's Firmware Practices

[chithanh]

I don't know why the Nouveau team needs the firmware. Aren't you inspired by the old RadeonHD driver at all?

They require the firmware to do any kind of acceleration. This situation is the same with radeon and radeonhd drivers to my knowledge. What radeonhd did differently is not using the functions of the (proprietary) BIOS for modesetting.

[Hi-Angel]

No no, it is much more simple than that. They just sprinkle some magical voodoo dust on it and pray to their engineering team.

But in all seriousness they probably have corporate politics just like AMD. Most if not all of Nvidia's management wants closed source software because(insert stupid non technical reason here) Than there is all the engineers who just want to maintain all the code as it is. Then there is the tegra team who had partial management support to release FOSS(somehow probably as to gain publicity) but limited to tegra(since android is so messed up already). Then we have the nouveau team essentially decimating the non FOSS management people's ego who will be all alpha male like and harden against releasing any firmware.
In a sense it is the same revolution with AMD they had management who wanted non FOSS because(insert reason here) But their engineering team convinced management they could work faster without blob's. It worked, save engineering politics ruining it(or sabotage? you decide.). And now AMD's management team have hardened against FOSS because of trying to show alpha male superiority.

But anyway, I didn't see how that would hurt to the firmware. As I see, this is how it works: devs are either being told "do not make a firmware, we don't cooperate with FOSS in any way", or "make a firmware, so that we have most of the logic closed, but FOSS folks could build above a layer they want". But I can not imagine devs are being told by management "make a firmware blob for FOSS folks, but make sure it is as bad as possible!", that doesn't make any sense! They either cooperate, or they doesn't.

[bug77]

He is not talking of opensource on Mac or on Windows or on Android, but on Linux Desktop.

You know, the OS that isn't usually pre-installed, the OS that in many cases requires a power user to be set up properly or to troubleshoot any issue arising due to hardware support.

Here the amount of people that care about open is much much higher than average.

That's true, but companies tend to care more about their addressable market...
Well, I think we both know where Linux stands and how much push the community has, so I'm going to stop here.

[pinguinpc]

Nothing new under the sun

Nvidia dont have any interest on opensource drivers, only have minimum information for recognition cards but nothing more

Nvidia card = propietary drivers

Nvidia open source drivers dont have future, maybe as other said is better idea help other hardware manufacturers like as intel

Wayland in this moment dont have important because most apps run in X, and xwayland dont have same compatibility and performance of regular X

For moment X stay in most distributions meanwhile mir vs wayland debate must be show who be chosen

[artivision]

They require the firmware to do any kind of acceleration. This situation is the same with radeon and radeonhd drivers to my knowledge. What radeonhd did differently is not using the functions of the (proprietary) BIOS for modesetting.

Not exactly. The current driver doesn't use firmware functions for 3D acceleration, vector registers are freely accessed. The older driver didn't use any firmware functions at all.

[bridgman]

Not exactly. The current driver doesn't use firmware functions for 3D acceleration, vector registers are freely accessed. The older driver didn't use any firmware functions at all.

We need to ban the word "firmware" except for use in folder names - it leads to too much confusion between HW microcode images and VBIOS/AtomBIOS.

All AMD GPUs from Rage 128 onward require one or more microcode images in order to operate, with newer GPUs using microcode in more areas. Not sure about NVidia HW but I believe the timing and degree of usage is similar.

In most cases the microcode itself is invisible to the driver (other than maybe SMC) so perhaps you could say that "the current driver doesn't use firmware functions for 3D acceleration", but the hardware simply does not function (no acceleration among other things) if microcode has not been loaded.

[artivision]

We need to ban the word "firmware" except for use in folder names - it leads to too much confusion between HW microcode images and VBIOS/AtomBIOS.

All AMD GPUs from Rage 128 onward require one or more microcode images in order to operate, with newer GPUs using microcode in more areas. Not sure about NVidia HW but I believe the timing and degree of usage is similar.

In most cases the microcode itself is invisible to the driver (other than maybe SMC) so perhaps you could say that "the current driver doesn't use firmware functions for 3D acceleration", but the hardware simply does not function (no acceleration among other things) if microcode has not been loaded.

The ASIC doesn't work without the microcode, except if you write a program for it (probably in C).

[starshipeleven]

Sorry, I re-read a few times, but absolutely didn't get. Do you mean there's a build script which management folks are randomly running, and if it succeed, they announce a release with disregard if the actual code has bugs? It's a unlikely interpretation, but…

No, I said that NVIDIA driver devs themselves are actually friendly towards linux, but they have their hands tied by upper management guys as it's their call about when (and if) releasing firmwares or docs or even contributing to Noveau.
Really, they all have the firmwares there and ready at NVIDIA, same for docs (as they do use docs for their own development).

The same for release schedules, the management decides how much time must be allotted to driver development before a release (as in many cases the releases are also a marketing thing or must happen at specific moments), how much people is in the dev team and so on, and (indirectly) what gets most importance and what gets less.

This is true for most companies, idiot dumbfuck IT illiterates with economical or marketing backgrounds decide about things and times without keeping in mind technological hurdles or caveats, and development teams or sysadmins and techs must run and take shortcuts to get shit done in half the time they would need to do a good job.

That's large part of the reason why most companies make half-assed products of any kind in general. They might also have idiots in the design and development teams, but the most usual issue is a manager that cries if the product does not have the buzzword tech of the moment, or that thinks their people can make everything happen in a week, or that all development issues can be solved by hiring moar (cheap and crappy) developers.

[starshipeleven]

That's true, but companies tend to care more about their addressable market...
Well, I think we both know where Linux stands and how much push the community has, so I'm going to stop here.

You really think it was the community that convinced AMD and Intel and whatever else to make decent open or semi-open drivers?

Because what is true for NVIDIA is also true for everyone else.

[starshipeleven]

For moment X stay in most distributions meanwhile mir vs wayland debate must be show who be chosen

???
The "debate" never existed. Mir is Ubuntu-only, Wayland is for everyone else and possibly also for the more enlightened BSDs like FreeBSD and TrueOS, Xorg will be relegated to obsolete shit in Unix or dumb BSD land, period.