DOWHILE Loop General Form

1
DO-WHILE Loop General Form
2
Example Multiply 2 Numbers (in 2000 and 2002)
3
Which program is correct and why?

• If the value in 2002 is 0
• -gt DO-WHILE is incorect
• The sub.w instruction is executed before the bne
  test
• Therefore the loop executes 65536 times, not 0
  times
• Important to decide which loop construct fits
  your problem domain

4
How would you fix the DO-WHILE version?

• Need some way to test the value in 2002 first
  and skip the DO-WHILE loop if value0
• -gt We need Conditional Statements

5
IF THEN ELSE General Form
6
Example
If contents of 2000 is 0 then set 2002 to 0
otherwise set it to 1.
value (2000) if (value 0) result 0
else result 1 (2002) result
7
Assembly Implementation
8
Fix the problem with the DO-WHILE loop
Loop only executed if count is not 0
9
Example Normalise a 16-bit Number

• -gt Bits are shifted left until MSB 1, shift
  count stored as the exponent
• 0001 0010 -gt 1001 0000, exponenet 11

10
Solution in English

• Set exponent to 0.
• Load number to normalise.
• Shift the number to the left until MSB 1 (i.e
  number is negative).
• Increment exponent at each shift.
• Store number and exponent.

11
Pseudo Code
WHILE loop or the DO-WHILE loop? Number might
already be negative (normalized) -gt execute loop
0 times -gt use WHILE loop.
exp 0 num (2000) while (MSB(Num) !
1) num num ltlt 1 exp exp 1 (2002)
num (2004) exp
12
Convert to Assembly Language
How do we determine if the number is negative? -gt
Examine the N flag, bmi and bpl branch depending
on the value of the N flag
move.w 0,d0 exp 0 move.w 2000,d1
load num while bmi endwhile while(ve)
add.w 1,d0 exp1 lsl.w 1,d1
numltlt1 bra while endwhile
move.w d1,2002 store num move.w d0,2004
store exp
13
Note

• Always ensure that the last instruction which
  affects the CCR, prior to conditional branching,
  is testing the appropriate value.
• In the example above only 2 instructions which
  affect the CCR execute before the bmi endwhile.
• move.w 2000,d1
• lsl.w 1,d1
• Both of these generate CCR values based on the
  current value of num. This is precisely the
  behaviour we want.