Profiling typically involves measuring execution time and memory consumption in runtime (dynamically). In case of driver code you would most likely only be interested in execution time.
I'm far from an expert of JTAG, but I don't believe you can get accurate realtime trace with just a simple JTAG interface, you would need a full in-circuit emulator for that. Since you seem to be using Lauterbach, check out their FIRE emulator.
But... the best way to measure execution time on embedded systems is to set an I/O pin before and after the code you want to measure. Measure the time between pin toggle with an oscilloscope. You won't be getting much more accurate timing than that. Also, if you are attempting to develop embedded systems without an oscilloscope, you are desperately in need of one.
