Debugging and non-byte-oriented architectures

LLDB and the GDB remote serial protocol beneath it are quite strongly byte-oriented, for example the RSP “read memory” command has arguments “addr,length-in-bytes”. I’m experimenting with a debugger target for an architecture that is based upon 16-bit-words, and while it’s simple to map between words and bytes (e.g. when reading code from $PC double the address and the number of units retrieved) this gets confusing and raises the risk that its doesn’t get done universally.

My impression is that the code base is quite strongly wedded to 8-bit bytes addressed directly, but maybe I didn’t notice some kind of transparent scaling (e.g. declare a register class as being 16-bit-word, then all indirect access through it gets automagically scaled)? Are there existing non-byte-based architectures supported by LLDB that I could look at?