Re: [tree-profiling] Fix some leftouts of previous patch

[Andrew Pinski <pinskia at physics dot uc dot edu>]

On Dec 8, 2004, at 6:43 PM, Dale Johannesen wrote:

> On Dec 8, 2004, at 3:10 PM, Andrew Pinski wrote:
> > On Dec 8, 2004, at 5:55 PM, Jan Hubicka wrote:
> > > Hi,
> > > this patch fix the frequency updating issues (I will still keep the
> > > frequency updating around as I am not quite sure how we will want to
> > > deal with the estimated profiles once available yet) and it also 
> > > avoid
> > > recursive inlining that helps some benchmarks.
> >
> > Could back port the recursive inlining part to the mainline since I 
> > see that
> > we recursive inlining functions on the mainline which is a regression 
> > from
> > 3.4.0?
>
> It's a regression in some cases and an improvement in others, as usual 
> with
> changes to heuristics.   Consider everybody's favorite CS 101 example 
> of recursion:
>
> unsigned int factorial(unsigned int x) { return  (x==0) ? 1 : 
> x*factorial(x-1); }
>
> Inlining factorial(4) enough that all the calls go away is a clear win.
> (It may be that most of the wins come in cases that shouldn't have been
> written as recursion in the first place, like this one, but users do 
> things
> like that.)
>
> I'm not saying this is necessarily a bad idea, but we should have more 
> data.
> I know of 1 case in SPEC where recursive inlining is a win (like the
> factorial case, the recursion terminates in a couple of levels, and you
> wind up with a leaf function).  I don't know of any where it's a loss, 
> but it
> sounds like Jan might.

But the simple CS 101 example of recursion for factorial is actually 
already
optimized the correct way to be a loop instead of a call of a function.
        cmpwi cr0,r3,0
        li r2,1
        beq- cr0,L6
        mtctr r3
L5:
        addi r0,r3,-1
        mullw r2,r2,r3
        mr r3,r0
        bdnz L5
L6:
        mr r3,r2
        blr

This is because we do some tail recursion, we did not do before, Zdenek
implemented back when the tree-ssa was a branch.
But with -O3 we get:
_factorial:
        stmw r22,-40(r1)
        mr. r30,r3
        mflr r0
        li r3,1
        stw r0,8(r1)
        stwu r1,-112(r1)
        beq- cr0,L4
        addic. r29,r30,-1
        bne- cr0,L31
L7:
        mullw r3,r3,r30
L4:
        addi r1,r1,112
        lwz r0,8(r1)
        lmw r22,-40(r1)
        mtlr r0
        blr
L31:
        addic. r28,r30,-2
        li r0,1
        beq- cr0,L10
        addic. r27,r30,-3
        beq- cr0,L13
        addic. r26,r30,-4
        beq- cr0,L16
        addic. r25,r30,-5
        beq- cr0,L19
        addic. r24,r30,-6
        beq- cr0,L22
        addic. r23,r30,-7
        beq- cr0,L25
        addic. r22,r30,-8
        bne- cr0,L32
        mullw r0,r0,r23
L25:
        mullw r0,r0,r24
L22:
        mullw r0,r0,r25
L19:
        mullw r0,r0,r26
L16:
        mullw r0,r0,r27
L13:
        mullw r0,r0,r28
L10:
        mullw r3,r0,r29
        b L7
L32:
        addi r3,r30,-9
        bl _factorial
        mullw r0,r22,r3
        mullw r0,r0,r23
        b L25

which still has a recursive function.

Thanks,
Andrew Pinski