Page 121 - Introduction to Microcontrollers Architecture, Programming, and Interfacing of The Motorola 68HC12

98 Chapter 4 Assembly Language Programming

We have also been using the inequality symbols with index addressing to indicate
whether the constant offset is to be described with eight or sixteen bits. The 6812
actually has another choice for the offset that we have not discussed before now because
there is no special symbol for it. This is the 5-bit offset option. In this case, one can
actually squeeze the offset into the post byte as described in the instruction set summary
of the CPU12RG/D manual. The assembler chooses between the three offset options in
exactly the same way that it chooses between page zero and direct addressing. On the first
pass, if it knows the values in all the labels used in an expression for the offset, it will
choose the shortest possible offset option or, if the expression is zero, it will take the
zero offset option, which is pointer addressing. If it does not know some of the labels
used in the expression for the offset, the assembler will default to the 16-bit offset
option, determining these bytes on the second pass. From now on, we will drop the use
of inequality signs in all addressing modes, except where it is needed in the relative mode
to designate a short forward reference. Generally, it is best to let the assembler choose the
appropriate option.
We sometimes observe an error message "phasing error," or "labels changed values
on second pass." Such an error occurs when an instruction's length is computed on the
first pass but is computed to have a different value on the second pass. Following such
an instruction, successive labels will have a different value on the second pass than they
had on the first pass. To fix such an error, read backward from the first instruction with
the line that has such an error until you see an instruction whose length changes in. the
second pass, due to its using a different addressing mode. Put a prefix or suffix on its
operand to force the instruction's first pass length to its second pass length,
ORG $800
K: DS M
M: EQU 2
*
LDD K
ADDD #3
STD K
SWI
Figure 4.6. Program with Illegal Forward Reference
As we have discussed earlier, an assembler does several things for us. It allows us to
use instruction mnemonics, labels for variable locations, and labels for instruction
locations while still providing the machine code for our program. As Figure 4.6 shows,
however, we must be careful with forward references when using assembler directives.
The assembler reads the assembly-language program in Figure 4.6 twice. In pass
one, the symbol table is generated and, in pass two, the symbol table, the instruction
set, and assembler directive tables are used to produce the machine code and assembly
listing. On each pass, each line of assembly language is processed before going to the
next line so that some undetermined labels may be determined on the second pass. For
example, in the program in Figure 4.6 the assembler will not determine the length of M
on the first pass because the DS directive makes a forward reference to M, that is, uses a
symbol in the expression for K that has not been determined yet. Suppose now that we
change the program a little bit. See Figure 4.7.