[FFmpeg-devel] [Hardware Accelerator for Video and Audio]

[Andreas Setterlind]

On 2/21/08, kenye ye <kenyeby at gmail.com> wrote:
> > Will this be a PCIe card? Will there be PCI versions too?
> In our opinion, USB version would be better in first stage becasue it is more popular.

Understand that USB could also lead to huge bandwidth issues for
high-resolution and high-bitrate when transferring uncompressed video
streams, however I can understand that at first thought it would be
appealing for you to use the USB interface, especially during
developing and testing phases, but for retail sale I think that having
both PCI and PCI-Express cards available for internal installation in
computer chassis, (and also Mini-PCI and Mini-PCIe adapter modules for
internal installations in laptops and laptop-based desktops, like the
AOpen MiniPC from your competition).

On 2/21/08, kenye ye <kenyeby at gmail.com> wrote:
> > Does it provide video output ?
> Yes, It will provide video output.

Will it still be possible to do what Benjamin asked for; send it a
video stream, the hardware decodes that stream and gives it back to
you, and then you are in charge of providing the result to a
third-party video card for output? Or will it only be possible to
output the decoded video to a display via your hardware's own
video-output?


On 2/21/08, kenye ye <kenyeby at gmail.com> wrote:
> On 2/21/08, Andreas Setterlind <gamester17 at xboxmediacenter.com> wrote:
> > Another new question; will your video accelerator processor chip be
> > software programmable? I mean in a way similar to how you can write
> > custom pixel shaders, OpenGL Shading Language (GLSL, a.k.a. GLslang),
> > and Cg (C for Graphics) can do many different things on a GPU that it
> > may not have been hardcoded for, (I am thinking of example running
> > video post-processing filters, such as sharpen, blurring, denoising,
> > deringing, requantization, etc. after the normal video decoding
> > process).
>
> Do you need a GPU-like accelerator? Why?

No, that is not what I am asking. I am asking if it will be software
programmable similar to pixels shaders act like a high-level
programming interface for GPUs (again, not the same, but similar). The
reason is that it could make the hardware a little more future-proof
if not everything is hardcoded but instead you allow a programmer to
use the C programming language to write new custom code algorithms for
execution on your hardware accelerator. NVIDIA CUDA and ATI CTM are
two other examples of programming interfaces that allows those GPU to
execute new custom code algorithms that is not hardcoded into the
hardware, (again I am not asking your hardware to act like a 'GPU-like
accelerator', I am only asking about the concept of being
programmable).
http://en.wikipedia.org/wiki/CUDA
http://en.wikipedia.org/wiki/Close_to_Metal

Verses fully hardcoding it which these days can many times be
considered a bad thing, see:
http://en.wikipedia.org/wiki/Hardcoded


On 2/21/08, Igor <goga777 at bk.ru> wrote:
> Already exists the Extension HD PCI card from ReelMultimedia with hardware h.264 decoder
> http://www.reel-multimedia.de/shop/product_info.php?products_id=223&language=en
> http://www.reel-multimedia.com/rmm-english/pdf/produkt-flyer/extension_hd.pdf

Broadcom also have a few similar video decoder hardware chips which do
provide, (see Broadcom BCM70010, BCM70012, and BCM70020) and they are
designed for PCI, PCIe, mini-PCI, mini-PCIe, and ExpressCard (PCMCIA
replacement), or onboard form factors.
http://www.broadcom.com/collateral/pb/70010-PB01-R.pdf
http://www.broadcom.com/collateral/pb/70010_70012-PB01-R.pdf
http://www.broadcom.com/collateral/pb/70020-PB01-R.pdf


Best regards / Andreas Setterlind (a.k.a. Gamester17)