AW: -O2 problem / GCC 3.3.x

AW: -O2 problem / GCC 3.3.x

[Beschorner Daniel]

The point is the unused array Q.
If I remove it O2 works fine. But this must have a reason.
Strange...

-----Ursprüngliche Nachricht-----
Von: Bossom, John [mailto:John.Bossom@Cognos.COM]
Gesendet: Montag, 15. November 2004 23:07
An: Beschorner Daniel; gcc-help@gcc.gnu.org
Betreff: RE: -O2 problem / GCC 3.3.x


I would not have expected this to fail with -fgcse
According to the documentation it performa a global common subexpression
elimination pass as well
as global constant and copy propagation.

Then again, you are going to get rounding errors (I believe the results
overflow, hence, one could
argue that the results you are getting to begin with are suspect - I guess
it would be nice if
they failed the same way)

-----Original Message-----
From: Beschorner Daniel [mailto:Daniel.Beschorner@facton.com]
Sent: Monday, November 15, 2004 4:51 PM
To: Bossom, John; gcc-help@gcc.gnu.org
Subject: RE: -O2 problem / GCC 3.3.x


Thank you!

I tested on your advice (i386 hardware)...

-O2 -fno-strict-aliasing, still wrong result
but
-O2 -fno-gcse, correct result

Daniel

-----Ursprüngliche Nachricht-----
Von: Bossom, John [mailto:John.Bossom@Cognos.COM]
Gesendet: Montag, 15. November 2004 21:53
An: Beschorner Daniel; gcc-help@gcc.gnu.org
Betreff: RE: -O2 problem / GCC 3.3.x


Read up on -fstrict-aliasing (which is on with -O2).

Try building again with -O2 -fno-strict-aliasing

(Not saying this is the problem, however, if it goes away with the
-fno-strict-aliasing
 then you'll know)


-----Original Message-----
From: gcc-help-owner@gcc.gnu.org [mailto:gcc-help-owner@gcc.gnu.org]On
Behalf Of Beschorner Daniel
Sent: Monday, November 15, 2004 3:42 PM
To: 'gcc-help@gcc.gnu.org'
Subject: -O2 problem / GCC 3.3.x


Any idea why results differ between -O and -O2 with GCC 3.3.x/Linux??? GCC
2.95 works fine.

results:
-O	500000
-O2	300000

Thanks
Daniel



#include <stdio.h>

int A,B,C;
int i;
int Q[1];

int H(int X, int Y)
{
        return X^Y^100000;
}

int main()
{
        A=B=C=100000;

        A+=H(B,C);
        C+=H(A,B);
        A+=H(B,C);
        C+=H(A,B);

        for (i=0;i<1;i++)
                Q[i] = 0;

        printf("%i\n",A);
        return 0;
} 
  
       This message may contain privileged and/or confidential information.
If you have received this e-mail in error or are not the intended recipient,
you may not use, copy, disseminate or distribute it; do not open any
attachments, delete it immediately from your system and notify the sender
promptly by e-mail that you have done so.  Thank you.