[Gstreamer-openmax] 回复： Discussion on the hardware accelerator solution in GstOpenMAX project.

[Felipe Contreras]

2008/8/1 wangfei <bjwf2000 at gmail.com>:
> Hi All,
>
> I'm working on a project based on several different hardware platforms. TI
> 34xx is one of my option. I'm very interest on topic 2: how to integration
> GStreamer with OMX on GPU/DSP hardware solutions. My thoughts are as
> follows:
>
> 1.  Supper source/sink element or tunneling solution is something depends
> how OMX was implemented on a specific hardware. OMX IL spec didn't specify
> how to partition different processing functions into OMX IL components
> (Refer to openmax_il_spec_1_1_1.pdf page 21, Figure 2-1). For example, in TI
> OMAP3, a C64x DSP was equiped, a super OMX component was provided by TI  for
> audio streaming. http://omapzoom.org/gf/project/openmax/wiki   The super OMX
> component includes audio decode, PP, rendering functions which were
> processed by C64x DSP. In such case, seems a supper GstOpenMax sink element
> is the only choice?  In another OpenMAX IL implementation, decoder, render
> provided by separate OMX complents, in such case, maybe separate GstOpenMAX
> elements support OMX tunneling is better.

Right, if the OpenMAX IL component is an mp3sink, then it makes sense
to write a gst-openmax mp3sink element wrapper.

> 2. I guess GstOpenMAX didn't want to give a recommendation or specification
> for OMX IL provider on how to paration there OMX IL components. Mybe I'm not
> right?

Right. The idea is to map whatever OpenMAX IL is doing into GStreamer.
It should be fairly easy to create a new mp3sink element if that's
what OpenMAX IL is providing. With the time it will become even
easier.

Best regards.

-- 
Felipe Contreras