[A83] Z88DK vs. SDCC

[Prev][Next][Index][Thread]

[A83] Z88DK vs. SDCC


Consider the following code:
main ()
{
     int i, j, k;

     i = 23;
     j = 45;
     k = i * j;
     return k;
}

Z88DK produces:
     push bc
     push bc
     push bc
     ld   hl,4
     add  hl,sp
     push hl
     ld   hl,23
     pop  de
     call l_pint
     ld   hl,45
     pop  de
     pop  bc
     push hl
     push de
     ld   hl,4
     add  hl,sp
     call l_gint
     push hl
     ld   hl,4
     add  hl,sp
     call l_gint
     pop  de
     call l_mult
     pop  bc
     push hl
     ld   hl,0
     add  hl,sp
     call l_gint
     pop  bc
     pop  bc
     pop  bc
     ret

SDCC produces:
    push    bc
    push    de
    push    ix
    ld  ix,#0
    add ix,sp
    ld  hl,#0x040B
    pop ix
    pop de
    pop bc
    ret

Note how much better this is, even though it saves BC & DE and creates a
stack pointer for no reason. In a normal routine, saving BC & DE and
creating a stack pointer would all be necessary. This shows how much it
matters for the compiler itself to be good, and not just some sort of
peephole optimizer. Note that the SDCC code wasn't run through a peephole
optimizer (doesn't seem it would've helped this out much), whereas the Z88DK
code was. It was even worse before the peephole optimizer.