I think he comes online only when he is in work... i'd quess by 2nd january. Send him a PM.

I do have that code, but i am not sure if I am supposed to give it away

MiniDos is a very basic Dos clone, many function are not implemented, i was hope coder would add to it, to make it better, but so far its not happened and i have little time.

When you set up vesa you need to call a function that fills out a Mode_Info, you can get the LFB address from eg: ModeInfo_PhysBasePtr

See the example from cdpod

For programming the Xbox see here: http://www.xbdev.net/non_xdk/nonxdk.php

> Sinsi - How can I get the LFB address from Vesa?

Here's one of my first fasm programs - a bit crap, but shows some stuff

Thank you all for the info!

I'm thinking if we each with our different video cards can use Fenryo's method of figuring out the base pointer for a certain mode, then we can compile a open-listing of different lfb addresses for different modes for different graphics cards.

But it would be nice for us all to figure out how to change modes without VESA, but through ports (if possible, which I think it would be.)

> But it would be nice for us all

YES.

> to figure out how to change modes without VESA,
> but through ports (if possible, which I think it would be.)

1. It definitely is possible.

2. There is no quick hack to discover. It's card specific, and messy. Linux reportedly has some open source graph "drivers" - compiled into the kernel (???) - OTOH the guys are "open" for closed source drivers as well (provided by card manufacturers, as for "Windows") - check Linux source (C mixed with GAS, not what I personally prefer)

Maybe some Linux experts can reveal more details ?

Dos386 is right. It *is* possible, info how to do it already is in linux sources. It's just to find it.

Her is a website that has linux open source drivers...

http://www.viaarena.com/default.aspx?PageID=2&OSID=25&CatID=2580

just click on the ones avail and download..

Though its all in C style code.. It still can be helpfull!

rhyno, you might like to have a look at this:
http://web.inter.nl.net/users/be-hold/BeOS/NVdriver/index.html

Thanks Raedwulf!

I'm going to email this guy and ask him how he got his info, because it seems he knows something we all don't.

Apparently before linux was too popular with video drivers, there was some agreements to release various info.

I'm however toying with another idea, what if you can wrap the windows drivers and use it? Something like ndiswrapper but for video drivers. Is this even possible? I wonder...

Tasks would include:
Windows 2000 Display Driver Interface OR WDF (for vista drivers)
ABI translation (esp. for 64-bit)
Host operating system Display Interface...

The benefits of starting a wrapper, is any hobbyist OS could maybe in the far distant future support 3d graphics cards... and games .

To make a wrapper you only have to know which OS functions the drivers use and emulate them. So I think it would be possible although not necessarily easy.

However for Vista, with it's DRM requirements. you may be SOL on that count It would be interesting to see if a Vista driver could be successfully fooled into believing it was in a secure environment.

I think Windows 2000 drivers are the way to go because:
1.) Better compatibility
2.) Lighter/better performance
3.) Less api to deal with as you said revolution

But thanks for the vote of confidence, at least im not a mad hatter after all hehe
Basically the inspiration for this is, there is Windows, Linux and MacOSX (and a bit of FreeBSD) which take up the majority of the userbase, and all other OSes are ignored because there is no 3d acceleration. (This may be true or untrue ) but with 3d acceleration, who knows? At least for people like us, it would be fun to play quake3 on something like menuetos.

http://www.reactos.org/en/index.html - they aim to be binary compatible for windows drives. Oh, and check http://www.tinykrnl.org/ as well.

Edit: oh bollocks, tinykrnl has been shut down

For 64bitness then Vista or XP drivers will be needed.

windows 2000 and xp share the same api according to microsoft, so i guess 64-bit api is similar

2000 and XP are very similar yes (NT5.0 and NT5.1), there's kernel changes on XP though, and some changes in userland win32 API as well... but I think drivers written for one should basically run out of the box for the other.

The ISSUE, however, is not 2000 vs. XP, it's 32 vs 64 bits. You can't just use drivers for one mode in the other.

Also, XP64 is actually based on 64bit win2k3 server code, not 32bit XP code... but again, 2k3 server is still NT5.x afaik.

1. method: int 10h - vesa call - returns dword of framebuffer memory
; ---
2. method: PCI scan - this method is able to handle qword address of framebuffer (perhpas good for computers with 4 GB of RAM or more but I have an idea where frambuffer is still under 4 GB limit in such configs... needs test on such machine)
this method uses allmost every new OS, including win - you can see various memory and port ranges in device manager
;---
After obtraining the framebuffer memory, YOU MUST MAP THAT MEMORY into virtual memory space if your OS runs under protected mode !!! You mustn't write directly into framebuffer memory under protected mode without mapping it somewhere... It is e.g. 1 API under win - less than 10 instructions or about 20-40 instructions under OS you are developping
I'll post win64 sample driver how to do it, but I'll manage it not earlier than the day after tomorrow.
Don't worry, you'll be able to improve it further, e.g. not to obtain only the start of framebuffer, but to obtain the end of framebuffer too (as you can see it in win device manager).

Okay - the question remains - are there ANY graphics cores broken until today? Is is possible (of course it is, when you've got infinite amount of time) to reverse-engineer only ONE card, say the upcoming N9x core from nVidia and test the reverse-engineered sample say, on MenuetOS? I'd like to see how much difference does VESA have in speed compared to flipping frames in that fast hardware and having all those fancy Blit operations with AntiAliasing and Transparency

Is it too much to ask (rhetorical)...

Madis, we're trying to figure out a way around reverse engineering graphics drivers.

NOT true. You must map if you have paging enabled ... you can "directly" write to $E800'000 or whatever if you don't use paging



Maybe ... but still very insufficient Writing into the LFB won't have any (useful) effect without setting the mode before ... and there is no universal method to do except VESA