Next thing are the unrolled:

moveq.l #0,d0
move.b (a0)+,d0
lsl.w #4,d0
lea.l 0(a5,d0.l),a1
bsr.l draw_tile

A bsr is 18 clocks. An rts is 16 clocks. Total of 34 clocks to use the routine. Looping with dbra takes 10 clocks each loop and 14 clocks the final loop. It will be smaller and faster to inline 'draw_tile' and loop.