Loop Unrolling; Loop Unrolling In N Array Scale Routine - Motorola DSP56600 Manual

Pipeline Interlocks
Data ALU Pipeline Interlocks
6-4 Optimizing DSP56300/DSP56600 Applications
move x:(r4)+,AA,y:(r1)
move B,y:(r7)
move y:(r0)+,B
The parallel source moves that caused the pipeline interlocks were
shifted to the following instructions. This example illustrates the
importance of ordering the arithmetic instructions and the parallel
read operations. Taking this approach when writing a program can
shorten the execution time by preventing unnecessary pipeline
interlocks.
6.1.2.2

Loop Unrolling

The usage of two accumulators can avoid arithmetic pipeline when
combined in loop unrolling techniques. The following two
examples demonstrate possible applications for this method.
6.1.2.2.1

Loop Unrolling in N Array Scale routine

The following code segment is used for scaling an array of N
positive numbers:
clr A
rep #N
max B,A
clb A,B
move
do #N,_end
normf B1,A
move x:(r1)+,AA,y:(r4)+
_end
The read operation of accumulator A in the eighth instruction
causes an arithmetic pipeline interlock in the critical loop, causing
the loop to execute 3N cycles instead of 2N. Using two accumulators
can avoid this to happen, as demonstrated in the modified code:
clr A
rep #N
max B,A
clb A,B
move
move
do #N/2,_end
normf y0,A
normf y0,B
move
move
_end
x:(r0)+,B
x:(r0)+,A
x:(r1)+,A
x:(r0)+,B
x:(r0)+,A
x:(r1)+,A
x:(r1)+,BB,y0
x:(r1)+,AA,y:(r4)+
x:(r1)+,BB,y:(r4)+
;A=next a,PUT d'
;PUT b', A=next a
;B=next c
;Largest value of N numbers
;Count leading bits of the
;largest number
;Scaling block of N numbers
;Largest value of N numbers
;Count leading bits of the
;largest number
;Scaling block of N numbers
MOTOROLA