Phasing Errors--Motorola Syntax

A phasing error occurs when the actual address assigned to a label on pass one and stored in the symbol table turns out to be incorrect for the context where the label must fit in the final assembly (pass two).

Most assemblers targeted for 16-bit Motorola microcontrollers assume that undefined labels (addresses) are two bytes long. During the first pass of assembly the symbol table is built using this assumption, which is reasonable since the address bus is 16 bits wide. However, if the address turns out to be "direct" or "relative" it will require only 8 bits for storage. For example, a direct address within the first 255 (decimal) or $FF bytes of the address space can be represented with only 8 bits. Likewise, a relative address of +127 or -128 (decimal) ($7F or $80) can be represented with only 8 bits. When the assembler detects that 8- bit addressing is possible, it will do that, but then the symbol table will turn out to have been incorrectly made. Later in the code the assembler will detect that the location of the label in the code does not match the symbol table. At this point a phasing error will be flagged.

              *LBL  OPCD OPERAND
              **********************
$0040               ORG  $0040
$0040 9D45          JSR  SUBR
$0042 20FF    LOOP  BRA  LOOP
##### 
Phasing error
This line was at address $0043 in pass 1, now in pass 2 it is 
$0042
$0044 39      SUBR  RTS
##### 
Phasing error
This line was at address $0045 in pass 1, now in pass 2 it is 
$0044

***************Symbol Table*********************
LOOP         $0043 
SUBR         $0045 
##### Assembly failed, 2 errors!

In the above example a phasing error occurs when the SUBR label is encountered for the second time. Notice that during the first pass the address for SUBR was found to be $0045 (16-bits by assumption). That is the value that was stored in the symbol table. When this label was actually used in the assembly (at the first occurrence of the label) only the least significant 8 bits were used (direct addressing--note the "45" in the machine code for that instruction, "9D45".) Then, when the label actually was defined in the second pass the address was $0044, which does not match the symbol table. At this point the phasing error was flagged.

The label "LOOP" suffers the same type of problem in this code. (The address of the label in the assembled code, $0042, does not match the address in the symbol table, $0043.) This is a consequence of the phasing error with SUBR.

If a label is actually used in a context where the label will be assigned an 8-bit address on the second pass, you can mark this with a "<" character in the source code the first time the label is used. Now the symbol table will be made correctly on the first pass.

              *LBL  OPCD OPERAND
              **********************
$0040               ORG  $0040
$0040 9D44          JSR  <SUBR
$0042 20FE    LOOP  BRA  LOOP
$0044 39      SUBR  RTS
                      
                      
***************Symbol Table*********************
LOOP         $0042 
SUBR         $0044 
Assembly successful

Use your browser's "back" feature to return to the page your were previously looking at.