Neither I am an expert in graphics hardware, but my understanding is this: Newer graphics card can do something called off-screen rendering, which means that the card create the graphics, and the puts it back to the system main memory, there applications can use it however they like.
(I did a quick google search on the term, and it gave back this: http://www.mesa3d.org/osmesa.html) If I am not mistaken, for examples compositing window managers use this feature. For example, my mainboard has integrated graphics hardware (intel g965), which uses a part of the system memory as video memory. So if off-screen rendering is possible, the rendered image should just end up in the memory. So, the data is surely available to the CPU, it just has to be transferred to the video memory if the other card. (At least, this is what I would like to believe.) Kristof -----Eredeti üzenet----- Feladó: [EMAIL PROTECTED] meghatalmazó: Z F Küldve: 2007.03.22., Cs 17:52 Címzett: Discussion list for development of the IVTV driver Tárgy: Re: [ivtv-devel] Displaying (other) HW-rendered data in the ivtv Xserver? --- Csillag Kristof <[EMAIL PROTECTED]> wrote: > Hi there! > > Hi! > > I have a weird idea, and I am not familiar enough with graphics > hardware to decide > whether it's possible or not. So, the question is the following: > > Is it possible to use one graphics chip-set to generate image > (ie. to render OpenGl, etc), and an other chip-set to output > the image to a physical output device? I am not an expert on graphics hardware, but my understanding is the following. In principle yes. In practice no. The situation is the following. When a video card (GPU) makes calculations, its output has to be delivered to another device. Unfortunately, old video cards have no output at all, other than the monitor connector. New cards can output data via PCI/AGP bus back to the main CPU, but the speed is so slow that is mostly useless for these types of applications. Why did I say "yes" in principle. Old video cards have a "feature" connector where data prepeared to be displayed can be pushed into the card much faster than PCI bus allows. For AGP cards, this is simply done via AGP bus. So, for these old PCI video cards which could not do calculations special hardware was needed (basically a co-processor). And such hardware existed in the form of hardware DVD decoders (for example). If one has such special hardware, it is possible to push precomputed images into the video card. I do not know if such hardware is commercially available and what the price of it would be. Anyway, my understanding that there is no way to get the results of GPU calculations (from a consumer video card) back to the main CPU fast enough to be interesting for such application. ZF ____________________________________________________________________________________ Need Mail bonding? Go to the Yahoo! Mail Q&A for great tips from Yahoo! Answers users. http://answers.yahoo.com/dir/?link=list&sid=396546091 _______________________________________________ ivtv-devel mailing list [email protected] http://ivtvdriver.org/mailman/listinfo/ivtv-devel
<<winmail.dat>>
_______________________________________________ ivtv-devel mailing list [email protected] http://ivtvdriver.org/mailman/listinfo/ivtv-devel
