Solution to never-ending driver woes

[nullbit]

Relatively long time lurker, first time poster.

Excuse me for jumping into this topic out of nowhere, but we desperately need an open source driver.

The problem with s3d is we are at the total mercy of companies liable to abandon us whenever we no longer fit their cost/benefit analysis. s3d will never keep up with bleeding edge gaming if we don't free it from corporate shackles.

A wrapper to directX + openGL is what I mean by a driver. KindDragon has shown this possible.

If every programmer in the community just contributes a few hrs a week*, and everybody else did what they could to help (e.g. ideas, testing, research, documentation, spreading word of mouth) in no time, we could have something that puts nvidia's stereo 3d driver to shame.

We can have better compatibility (software + hardware), better performance, and better 3d. Forever.

An open source driver** can never die, because as long as the community lives, there will be members to maintain it. By spreading word of mouth, if one programmer leaves, another two will replace him.

We do desperately need an open source driver.

You have two choices:
1. Keep waiting for a miracle from nvidia or some other 3rd party
2. Do what ever you can to help (see todo list below) to make true s3d support happen now

Things that need doing***:
- spread word to others to maximize help
- setup sourceforge project -- DONE
- setup project management/collab software
- setup wiki for documentation
- do research to establish essential feature requirements (e.g. what 3d display types to support) based on demand
- compile list of helpful resources relevant to development
- anything else you can think of

I think the most important thing is speed to "market," once we have a proven prototype a lot more people will jump on board to help. Some mere suggestions that might help speed things up:
- focus should be on simplicity
- don't bother to support legacy systems. e.g. start at win xp + dx9
- non-essential features should not be addressed until core functionality stable
- work exclusively on direct3d (as the dominant standard), and only then address openGL

------------------

* I for one will add myself to the list of programmers. I'm not at all oriented in directX/openGL, but I can learn.

** Preferably licensed under most tried+tested solution, GPL.

*** I plan on doing some of these things myself tomorrow if people haven't beat me to it.

(Lastly, I don't pretend to be any authority on any matter above. I'm pretty much a laymen on all counts, I just have a real desire to see this thing happen.)

[nullbit]

I've created a sourceforge project, pending staff approval.

I had to make educated guesses about some of the fields, but these can be changed later.

[Likay]

Great effort! This could very well be the salvation for S3D the way we know it! The latest drivers from the Nvidia site is released over a year ago and that tells something about the trend... I'll do what i can when i can even so i'm quite busy most of the time. You're really welcome and i hope we could get some more cunning people into this (obs: Not counting myself as cunning though) .

[sharky]

I personally think that it is a brilliant idea. BUT i think that we should NOT create something wich seem to go against nvidia or other producers. we should simply create an alternative where everybody is free to take its own decision. its not about put invidia to a shame, but to show ourself that we can create something and MTBS and its comunity have something to share and give to users for free....

if it stays focussed this way i am happy to help as much as i can, but it has to give a good image of us, and not a bad image of others. Remember that untill now the comanies wich you are talking about have given us the less drivers we have. without them it wouldn't even exist..

keep up the good work..

sharky

[kapi]

I'm not a programmer but i could help for beta test if needed, i've a lot of games and hardwares/OS, it's a good idea but don't need a driver a hook would be great too like kinddragon started to do with a panel like E-D or Nvidia to setup.

[nullbit]

Pleased to see there is support for the idea.

Kapi, yes that's along the lines I was thinking. This would be a alternative to the Nvidia stereo 3d driver, not the the base GPU driver. I guess wrapper would be a more appropriate descriptor.

I'm going to get start on some actual proof of concept code today. Initially I'll be focusing on page-flipping mode (for shutter glasses).

In _theory_ basic implementation should be easy:

1. This page demonstrates how to hook into Direct3D processes.

2. By overriding IDirect3DDevice9::EndScene I can alternate the viewport left and right for each frame.

[peter64]

Hey this sounds like a great idea! I've been mulling over it for a while!
I'll be happy to lend a land and I do have some programming experience. I think I'll need an updated video card as my geforce2 may present problems with testing shaders and textures which i believe you have to modify for different perspectives (don't remember why).

The only issue with supporting frame sequential stereo is that nvidias 3d drivers as far as I know actually hack it to support a dual frame buffer which is then alternated by default. However if we are coding this ourselves you won't have access to such a dual frame buffer. As such you must have a frame rate higher than the frame rate the glasses will be refreshing in order for this to work. If it drops below you will have the incorrect eye displaying on the screen. I was really hoping some how we could use information from the neavau project to find out where the pointer is stored that directs media to these various frame buffers. Once you know the memory address of this pointer and how it works perhaps we could emulate it in such a way that we can use the nvidia dual frame buffer hack but actually render the 3d content ourselves. I think this will be a big help.

Once i get my next pc up and running with an nvidia card i'll try to use a machine level debugger to see if i can find out whats going on with that. No promise tho

[sharky]

so the idea seems dead.. what happens??

[peter64]

hey sharky well i dunno about mr nullbit but I aint dead here yet hahahah .... I just got my DepthQ on ebay the other day and hopefully it's on its way over to canada about now. Next i'm debating about putting together a pc with 8800 gts 320 mb ... and installing some debug softice poop and trying to figure out if theres some way to hack the pageflipping part of the 8800. Then if i can find a way I'm gona go and harass kinddragon lol and try to get him to implement that feature in his code. Then I'm gona start trying to implement a similar thing to kind dragon I have big dreams hahahaha and soon I'll have a lot of free time. Those two things go well together... We'll see what happens tho as there isnt much information available to me hahaha. I really wish there was more of a colaborative effort taking place. I get the impression people think anyone who wants to make a stereo driver is directly attacking Nvidia and thus threatening the 3d communities sacred relationship with nvidia. I really don't see the harm in a little nvidia driver slagging, (to try and get motivated) after all anger is normally a catalist for change . And we arent half as bad as other 3d communities out there.
I think if i buy an 8800 gts and i can't use it for poop. The motivation will definately be there to do something about making it work.

[sharky]

personally i think that gerenally everything that has been "stolen" from others do not bring anything good. at work i learned one thing:

different persons come to different solutions! if you start from zerowith your plan, your idea, you probably are going to solve the problem in a different way wich could be better or worst. but it is something more wich could be used for a working driver. AND, if you crate something alone maybe a BIG contacts you.. if you copy it you stay alone just for spearing some months of work.

bye

sharky