459a9b5949
Because lwzu or stwu moves the pointer, I can remove an addi instruction from the loop, so the loop is slightly faster. I wrote a benchmark in Modula-2 that exercises some of these loops. I measured its time on my old PowerPC Mac. Its user time decreases from 8.401s to 8.217s with the tighter loops.
20 lines
431 B
ArmAsm
20 lines
431 B
ArmAsm
.sect .text
|
|
|
|
! Set symmetric difference.
|
|
! Stack: ( a b size -- a/b )
|
|
|
|
.define .xor
|
|
.xor:
|
|
lwz r3, 0(sp) ! r3 = size
|
|
srwi r7, r3, 2
|
|
mtspr ctr, r7 ! ctr = size / 4
|
|
add r4, sp, r3 ! r4 = pointer before set a
|
|
|
|
! Loop with r4 in set a and sp in set b.
|
|
1: lwzu r5, 4(r4)
|
|
lwzu r6, 4(sp)
|
|
xor r7, r5, r6 ! symmetric difference of words
|
|
stw r7, 0(r4)
|
|
bdnz 1b ! loop ctr times
|
|
addi sp, sp, 4 ! drop last word of set b
|
|
blr
|