Parallella Community

The TRAP instruction supports a trap code that can be used to perform a particular task. This works fine in a simulator because it has easy immediate access to it when decoding the instruction. However, it doesn't seem like there's an easy way to grab the trap code when running on real hardware. Of course, the host could read PC, translate it to a host-side address, fetch the 16-bit instruction, and decode the trap code field, but this seems a bit hairy.

Is there a better way to do this, or should I go with what I described above?

What you described is more or less the way we have it implemented today using the host device to do the somewhat tricky instruction inspection. In the case of GDB it's downright nasty, but it works.
Andreas