DXVA convertion

Here is what one of our DXVA experts had to say about this:

The best way to start writing a DXVA2 decoder is to study the DXVA1 decoder documentation first. The data written to the DXVA compressed buffers is identical between DXVA1 and DXVA2.

The DXVA1 documentation can be found here http://msdn.microsoft.com/library/default.asp?url=/library/en-us/Display_r/hh/Display_r/dxvaref_47b33a0a-0766-4133-8b3b-1db4e844dd1b.xml.asp?frame=true

Additional info can be found here: http://download.microsoft.com/download/1/6/1/161ba512-40e2-4cc9-843a-923143f3456c/DXVA_1.01.doc

There are some small changes between the DXVA1 and DXVA2 decoder API’s these are:

1. You know longer use an IAMVideoAccelerator interface QI’ed form a DShow Video Renderer input pin.

2. Instead you create the decoder device yourself see http://msdn.microsoft.com/library/default.asp?url=/library/en-us/medfound/html/eeb62178-b54d-45d3-a584-75865f0662fa.asp?frame=true

3. With DXVA2 you pass in the configuration information at the same time as you are creating the device – this means that the device is correctly configured and ready to use as soon as you have created it. This is different from DXVA1, here you had to first create the device and then configure it before it could be used.

4. With DXVA2 we have explicitly enforced the correct compressed buffer locking in order, which is:

1. Lock compressed buffer(s)

2. Fill compressed buffer(s) with compressed video data

3. Release compressed buffer(s)

4. Call the DXVA decoder Execute function

The same locking order should have been used with DXVA1 but it was not enforced by the DXVA1 runtime.

BeckyWeiss-MSFT at 2007-8-30 >