NVIDIA Posts Tegra Gallium3D Patch For K1+ Support

[necro-lover]

Eww, who the hell wants that code anyway? It would most likely be completely unusable within the free stack anyway. What's missing is manpower to develop noveau further, not hw docs.

Say said the same for catalyst code but in fact
there is AMD-Intermediate-Language (IL) code translated to TGSI-IL/Gallium3D
code in the opensource radeon driver R600g/radeonSI.
http://amd-dev.wpengine.netdna-cdn.c...ication_v2.pdf
Bridgman talks about AMD-IL convert to TGSI : http://www.phoronix.com/forums/showt...Catalyst/page8
So it is a fact that some parts of the closed source driver can be usefull for the opensource/foss driver.

So you really think something what AMD did in the past converting some catalyast code into radeon code is not posible for nvidia and the nouveau driver?

In fact amd saved a lot of manpower by just using the AMD-IL catalyst code in the radeon driver.

[Ancurio]

Say said the same for catalyst code but in fact
there is AMD-Intermediate-Language (IL) code translated to TGSI-IL/Gallium3D
code in the opensource radeon driver R600g/radeonSI.
http://amd-dev.wpengine.netdna-cdn.c...ication_v2.pdf
Bridgman talks about AMD-IL convert to TGSI : http://www.phoronix.com/forums/showt...Catalyst/page8
So it is a fact that some parts of the closed source driver can be usefull for the opensource/foss driver.

So you really think something what AMD did in the past converting some catalyast code into radeon code is not posible for nvidia and the nouveau driver?

In fact amd saved a lot of manpower by just using the AMD-IL catalyst code in the radeon driver.

I don't think that's correct.. Mesa parses the shader source into a high level AST and then passes it into Gallium3D where it is converted to optimized TGSI. There is no AMD IL anywhere in this pipeline. Or are you talking about the final LLVM stage?

[necro-lover]

I don't think that's correct.. Mesa parses the shader source into a high level AST and then passes it into Gallium3D where it is converted to optimized TGSI. There is no AMD IL anywhere in this pipeline. Or are you talking about the final LLVM stage?

In the orginal radeon R600(mesa) there was no (converted)AMD-IL code
because it was 100% incompatible.
But in the R600g (Galium) driver is some converted AMD-IL code.

converted means: it is now TGSI code.

AMD in fact favored the Gallium version of the driver because of the similar architekture to there own AMD-IL/Catalyst driver architecture
what makes it possible to convert AMD-IL Code into TGSI code.

[bridgman]

Bridgman talks about AMD-IL convert to TGSI : http://www.phoronix.com/forums/showt...Catalyst/page8
So it is a fact that some parts of the closed source driver can be usefull for the opensource/foss driver.

So you really think something what AMD did in the past converting some catalyast code into radeon code is not posible for nvidia and the nouveau driver?

In fact amd saved a lot of manpower by just using the AMD-IL catalyst code in the radeon driver.

We used LLVM IR to AMDIL code from the Catalyst OpenCL compiler as a starting point for Tom's LLVM IR to R600/GCN HW ISA back end, which in turn was used in two ways :

- graphics path, where TGSI was converted to LLVM IR then sent through the R600 back end to generate HW ISA

- clover compute path, where LLVM IR from Clang was sent directly to the Gallium3D pipe driver (bypassing TGSI) and through the back end to generate HW ISA

Tom would have to comment on how much time was saved in the end. My impression was that it definitely helped us get started more quickly (starting with something that runs is always a big help) but not sure how much actual time was saved overall.

[necro-lover]

We used LLVM IR to AMDIL code from the Catalyst OpenCL compiler as a starting point for Tom's LLVM IR to R600/GCN HW ISA back end, which in turn was used in two ways :

- graphics path, where TGSI was converted to LLVM IR then sent through the R600 back end to generate HW ISA

- clover compute path, where LLVM IR from Clang was sent directly to the Gallium3D pipe driver (bypassing TGSI) and through the back end to generate HW ISA

Tom would have to comment on how much time was saved in the end. My impression was that it definitely helped us get started more quickly (starting with something that runs is always a big help) but not sure how much actual time was saved overall.

Thank you for helping me here.

It really has a long history that people pointing the finger on me
and say something like i talk bullshit.

I still read phoronix articles and forum posts and sometimes i can not hold me back.

Can i have a forum tag like this?:
"this person is aproved to had read 80+% of all phoronix forum posts
Plus this person writes more than 10000 forum posts in the past"

Isn't it True?

[Gusar]

Michael: "NVIDIA has out a wonderful Thanksgiving surprise"

Surprise ? are you kidding? I wrote this over a year ago: 07-16-2013

http://www.phoronix.com/forums/showt...rietary-driver

LOL! You really believe this code here has anything to do with what you wrote back then?

Nvidia will never open the proprietary driver. Even if there wasn't third-party code in it, it's a huge codebase with layers upon layers of micro-optimizations and application-specific optimizations and hacks. It also uses a whole different architecture from Mesa. So the code is very difficult to penetrate (it would take a lot of time to try and make sense of it) and in the end you wouldn't be able to use much of it anyway. Not worth the effort to go through the code and sanitize it for release, which would be a huge undertaking.

What Nvidia could do is contribute a lot more to nouveau. That means much larger drops of information than the snippets they've released so far, or large code contributions for new GPU enablement. But this code isn't even that! It's just glue code to hook up the K1 with nouveau. Definitely a valuable contribution in and of itself, but it's not some big thing you're alluding to.

It really has a long history that people pointing the finger on me
and say something like i talk bullshit.

But you do talk bullshit. You talk as if AMD opened significant parts of catalyst. But in reality they used one small part of it as a starting point to write some parts of the open driver. That's quite a different thing.

[necro-lover]

But you do talk bullshit.

A good starting point to talk isn't it?

Just tell me why is there all the hate inside of your babbling?

The effort to support FOSS for big companies viewpoint startet back
In 2007 with the words: we want to support foss with a opendriver
Then a learning prozess starts and this goes over puplish hardware informations
And the community do the rest this failed then they change there way into code first
And documentation later back then all said something like no we
don't use catalyst source code and again there change there way and now they
Sucessfully used parts of the Catalyst.

NVIDIA can do all the mistakes amd did in the past or they can be clever and learn from
AMDs history and follow the golden way amd found out.
This means they pay foss dev's and open up eome closedsource code as foss and release some
Documentation for there hardware or they will lose huge marketshare to amd because they are
Allready done with the golden way

[Gusar]

Just tell me why is there all the hate inside of your babbling?

Pointing out that what you're writing is very different from reality is not "hate". And can you please point to which parts of my post are "babbling" and in what way there are supposedly "babbling"?

[necro-lover]

Pointing out that what you're writing is very different from reality is not "hate". And can you please point to which parts of my post are "babbling" and in what way there are supposedly "babbling"?

I will not do it because i never do such things but if i block your forum Account
how many usefull informations will i miss? I check it later because now i am on my smartphone
But i think i will miss nothing.

And this is wbat i call babbling "no usefull information" no offence but it is my definition of the word "babbling"

And to your point that my worlds have nothing to do with reality here and now
You are right and i never want to point something out you can check in reality here and now.
My point is the Future nvidia is game over if they don't open up and support foss drivers.

I do have a hd6870 and a hd7850 my smartphone do not have nvidia
So who cares about nvidia anyway?