Re: Intel compilers performance over itanium 2 (madison)

Paul Pluzhnikov <[EMAIL PROTECTED]> wrote:
> Terje Mathisen <[EMAIL PROTECTED]> writes:
> 
>> I'd love to see the generated IA64 asm!
> 
> I don't have Intel compilers, but here are 'g++ -O3' and (HP) 'aCC +O3'
> results from 

I've got the Intel 7.1 compiler set for Itanium. For the same code,
I get the following assembly (using "ecc -S -O3"):

_ZN3Foo4funcEj:
 {   .mii
        alloc   r10=ar.pfs,2,0,0,0                       //0:  6   30
        add     r8=0,r0                                  //0:  7    6
        nop.i   0 ;;                            
 }
 {   .mib
        cmp4.gtu.unc    p7,p0=1,r33                      //1:  7    4
        nop.i   0                               
  (p7)  br.cond.dpnt    .b1_4 ;;                         //1:  7    5
// Block 2:  Pred: 0     Succ: 4 3  -G
// Freq 5.0e-01, Prob 0.50
 }
 {   .mmi
        ld4     r2=[r32] ;;                              //0:  7    7
        cmp4.gtu.unc    p6,p0=r33,r2                     //1:  7    8
        nop.i   0                               
 }
 {   .mib
        nop.m   0                               
        nop.i   0                               
  (p6)  br.cond.dpnt    .b1_4 ;;                         //1:  7    9
// Block 3:  Pred: 2     Succ: 4  -G
// Freq 2.5e-01, Prob 1.00
 }
 {   .mii
        add     r8=1,r0                                  //0:  7   10
        nop.i   0                               
        nop.i   0 ;;                            
// Block 4: exit  Pred: 2 0 3     Succ:  -GO
// Freq 1.0e+00, Prob 1.00
 }
.b1_4: // emit lab 1
 {   .mib
        nop.m   0                               
        nop.i   0                               
        br.ret.sptk.many        b0 ;;                    //0:  7   12
 }

-- 
Bjørn-Ove Heimsund
Centre for Integrated Petroleum Research
University of Bergen, Norway