[FFmpeg-devel] [PATCH][0/8] VA API patches summary

Gwenole Beauchesne gbeauchesne
Thu Feb 5 17:47:35 CET 2009


On Fri, 30 Jan 2009, Jean-Michel Pour? wrote:

> Silly question: is this accelerated display or do you provide
> accelerated hardware decompression. For kdenlive needs, we need full
> access to decoded frame content. Do you see any solution to speed up
> decoding and access full frames?

It should be possible to fetch decoded frames off GPU memory. But other 
changes would be needed, e.g. an extra flag to signal we want the decoded 

Off the xvmc_acceleration replacement idea, we could have it named 
hardware_acceleration with the following flags:

HWACCEL_XVMC_THING_1	0x00101	///< current xvmc_acceleration==1
HWACCEL_XVMC_THING_2	0x00102	///< current xvmc_acceleration==2
HWACCEL_VAAPI_VLD	0x00201	///< VA API VLD level acceleration
HWACCEL_VAAPI_IDCT	0x00202	///< VA API IDCT entry-point
HWACCEL_VAAPI_MOCO	0x00203	///< VA API MoComp entry-point
HWACCEL_VDPAU		0x00301	///< VDPAU VLD level acceleration
HWACCEL_READBACK	0x10000	///< indicates decoded frames are needed

Note: VA API is a display independent API. You may want to extend it to 
handle display-less accelerators. e.g. OpenMAX, davinci, whatever.


More information about the ffmpeg-devel mailing list