The documentation is part of the Media SDK package:
The release notes are here:
You can review the code and documentation for the samples here:
Here are the answers to your questions:
1) Does this SDK supports encoding on cross fire ?
2) How the context is defined for encoders ? Is one encoder bind with one GPU or one GPU can support multiple encoders ?
One encoder is bound to one GPU. A GPU should also support multiple encoders, but we do not have a test case to showcase this.
3) What mechanism does it use for giving input frames to encoder ? Does it accepts single frame( blocking ) or it can take multiple frames (non-blocking) ?
4) How it will generate output frames ?
We will get back on this. Could you please provide some context? What will you pass the output to?