Re: Display Device Mirror Driver - Build or Buy?
- From: Lev Kazarkin <mirrdrv@xxxxxxxxxxxxx>
- Date: Sat, 5 Nov 2005 09:09:02 -0800
> I'm not sure there is any network capable of transferring data at the
> speed of the composition engine. If you have seen the rodolex effect
> at the PDC, then, you should have the idea.
> Remote control programs will have inherently no physical bandwidth to
> send over the wire 1280x1024@32bpp 100Hz frames, no matter
> if they can get them from the PCI-Express 16X bus at maximum speed.
(skipped)
> The implementation of BitBlt would do exactly the same,
> with the difference that the mapping of the section in the address space
> would be done in Kernel Mode. You can set a breakpoint
> in nt!MmMapViewSection to conform.
> While you are doing this, the GPU would have naturally recalculated
> shading and effects based upon user interaction, invalidating your results.
I've come out with somewhat unexpected and maybe paradoxal conclusion
about all these tendencies.
Upcoming Windows Vista does not render mirror drivers useless at all.
Vice versa, Vista increases enormously the need for mirror drivers for
RC applications. Indeed, visual experience without desktop composition
is acceptable for RC. And, in the same time, there is no another way
in Vista to get screen update information efficiently, consistently and
keep a traffic at the decent rate for network interface.
Screen composition disabled - this is what is needed exactly for RC app.
Aero generates screen updates at a rate of 100 Hz.
So it leaves no choice. Aero must be switched off.
--
Best regards,
Lev Kazarkin
http://www.demoforge.com
"Ivan Brugiolo [MSFT]" wrote:
> I can understand your disappointment, but, I've just been trying to explain
> what was announced this spring at the Driver Developer Conference
> (and many time in the past) about the LDDM display driver model.
> LDDM was designed to minimize the number of 3rd-party
> components running in session space, and to eliminate
> the old-school display drivers. Not my decision.
> I'm just explaining the consequences of that decision.
>
> --
> This posting is provided "AS IS" with no warranties, and confers no rights.
> Use of any included script samples are subject to the terms specified at
> http://www.microsoft.com/info/cpyright.htm
>
>
> <me@xxxxxxxxxxxxxxxxxxxxxxx> wrote in message
> news:1129500688.279084.166020@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
> > So, the user mode driver talks with dxgkrnl.sys, which in turn talks
> > directly to the hardware, using PCI, PCI Express, or AGP bus. In the
> > software package there could be included a PCI card which just has a
> > "skeleton" configuration, i.e. the absolute minimum to get Windows to
> > accept it as compatible with Aero (through trickery, not actual specs).
> > The driver dxgkrnl.sys would communicate with this skeleton card, which
> > would in turn pass all rendering commands back to a 3rd party kernel
> > mode driver, for rendering (or logging) in system RAM. Including such a
> > PCI card in one's software package is a hassle, but possible - in the
> > unlikely case that a workaround via software is not figured out.
> >
> > For most purposes this would be overkill. The current way of
> > programatically disabling DirectDraw seems to work fine for mirror
> > drivers in Windows XP. There are some complaints, but for 75% of
> > customers it works fine. People who do serious work via remote control
> > do not have much need for games and movies anyhow. As for Vista, I
> > question whether most corporate customers will turn Aero on. Of those
> > that do turn it on, only a small percentage will demand that it remain
> > active when they are sharing their desktop remotely.
> >
> > Still, I am disappointed. This task was easy with SetDDIHook in Windows
> > 98, and since then has gotten progressively more difficult. You must
> > walk a fine line between giving freedom to developers and ensuring a
> > stable OS. I believe that you have taken away too many freedoms, and I
> > question whether it will increase stability in this case.
> >
>
>
>
.
- Prev by Date: Re: Signing Win32-LibUSB
- Next by Date: Re: Need driver
- Previous by thread: Re: Question about DDK
- Next by thread: Re: Help! Accessing an IO port location in Win2k driver
- Index(es):
Relevant Pages
|
