Re: [PATCH] Remove CC0

[Segher Boessenkool <segher@kernel.crashing.org>]

Hi!

Whoops, I forgot to reply to this...

On Wed, May 05, 2021 at 06:25:49PM +0000, Koning, Paul wrote:
> > void g(void);
> > void h(void);
> > void i(void);
> > void f(long a, long b)
> > {
> >        if (a < b)
> >                g();
> >        if (a == b)
> >                h();
> >        if (a > b)
> >                i();
> > }
> 
> FWIW, that also works on pdp11, so it seems the general mechanism is in place and working.  Well, with one oddity, an unnecessary third conditional branch:
> 
> _f:
> 	mov	02(sp),r1
> 	mov	04(sp),r0
> 	cmp	r1,r0
> 	blt	L_7
> 	beq	L_4
> 	bgt	L_5
> 	rts	pc
> L_5:
> 	jsr	pc,_i
> 	rts	pc
> L_4:
> 	jsr	pc,_h
> 	rts	pc
> L_7:
> 	jsr	pc,_g
> 	rts	pc

We have that on PowerPC as well (-m32 because it is shorter code with
fewer distractions):

f:
        cmpw 0,3,4
        blt 0,.L9
        beq 0,.L3
        blelr 0
        b i
        .p2align 4,,15
.L3:
        b h
        .p2align 4,,15
.L9:
        b g

The "blelr" (branch to LR if less than or equal) can never be taken.
This is there in the Gimple already:

===
void f (long int a, long int b)
{
;;   basic block 2, loop depth 0
;;    pred:       ENTRY
  if (a_4(D) < b_5(D))
    goto <bb 3>; [33.00%]
  else
    goto <bb 4>; [67.00%]
;;    succ:       3
;;                4

;;   basic block 3, loop depth 0
;;    pred:       2
  g (); [tail call]
  goto <bb 8>; [100.00%]
;;    succ:       8

;;   basic block 4, loop depth 0
;;    pred:       2
  if (a_4(D) == b_5(D))
    goto <bb 6>; [30.21%]
  else
    goto <bb 5>; [69.79%]
;;    succ:       6
;;                5

;;   basic block 5, loop depth 0
;;    pred:       4
  if (a_4(D) > b_5(D))
    goto <bb 7>; [70.57%]
  else
    goto <bb 8>; [29.43%]
;;    succ:       7
;;                8

;;   basic block 6, loop depth 0
;;    pred:       4
  h (); [tail call]
  goto <bb 8>; [100.00%]
;;    succ:       8

;;   basic block 7, loop depth 0
;;    pred:       5
  i (); [tail call]
;;    succ:       8

;;   basic block 8, loop depth 0
;;    pred:       5
;;                7
;;                6
;;                3
  return;
;;    succ:       EXIT

}
===

So, it also exists in the RTL -- and it isn't optimised away later.
That would be quite hard to do in fact, and if it is handled in Gimple
all will be fine, and it should be much easier to do in Gimple.

Can you open a PR please?


Segher