moving parts of code to FPGA

I am trying to port OpenOCD to an eCos powered board that also contains a FPGA for general purpose I/O.

The first approach was to have the JTAG pins directly mapped as I/O, which turns out to be very slow (JTAG TCK running at aprox 10 KHz).

The next step is to move some parts to FPGA. I have the TAP controller and the ICE access moved to the FPGA now, and I am trying to integrate that into OpenOCD.

My questions are:

  1. What’s the best approach to implement this in OpenOCD?

  2. Do I need direct access to TAP, or access to ICE is enough?

Regards,

Edgar