If I'm going to use a call for an immediate table reference, I write the subroutine to pop the CALL's return address as the table pointer. Loading pointers consume a lot of time.

Example:
. call sendthis
. db "Serial Link Test.",0h
. jmp somethingelse

After stepping through the table's data, the pointer is pointing to the proper return address. Usually a traversal ends one increment past the 00h terminator but you don't need that increment because returning to the 00h byte will be interpreted as a NOP.

Since the pops have restored the call pushes, you can merely JMP to the pointer address as a flat return. Most feel more comfortable pushing the new return address and then issuing a RET, and I admit I do this to improve readability though it wastes run time.

Note that these techniques would probably sabotage disassemblers, code checkers, and maybe simulators. I can live with that. :)

aka j