The carrier tracking has been improved. However I am still limited to
symbol rates below 2 MS/s. This is due to the overhead of the adaptive equaliser.
I have plans to accelerate that part of the code. The equaliser is quite agressive
as it needs to track the symbol rate error, reducing that error will reduce the
equaliser CPU overhead. Also moving the equaliser onto the GPU and
equalising 128 frames in parallel should hide the processor overhead too.
The problem at the moment is how to take the single thread CPU code and
move it to the multi-processor GPU.
No comments:
Post a Comment