[Bug tree-optimization/23452] Optimizing CONJG_EXPR (a) * a

[Bug tree-optimization/23452] Optimizing CONJG_EXPR (a) * a

[sayle at gcc dot gnu dot org]

------- Comment #2 from sayle at gcc dot gnu dot org  2006-05-30 21:34 -------
Subject: Bug 23452

Author: sayle
Date: Tue May 30 21:34:04 2006
New Revision: 114246

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=114246
Log:

        PR tree-optimization/23452
        * fold-const.c (fold_mult_zconjz): New subroutine of fold_binary,
        to optimize z * conj(z) as realpart(z)^2 + imagpart(z)^2.
        (fold_binary) <MULT_EXPR>: Call fold_mult_zconjz for integral
        complex values and with -ffast-math for FP complex values.

        * gcc.dg/fold-mulconj-1.c: New test case.


Added:
    trunk/gcc/testsuite/gcc.dg/fold-mulconj-1.c
Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/fold-const.c
    trunk/gcc/testsuite/ChangeLog


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23452

[Bug tree-optimization/23452] Optimizing CONJG_EXPR (a) * a

[roger at eyesopen dot com]

------- Comment #3 from roger at eyesopen dot com  2006-06-01 02:41 -------
This is now fixed on mainline provided the user specifies -ffast-math.
There are some complications where imagpart(z*~z) can be non-zero, if
imagpart(z) is non-finite, such as an Inf or a NaN.  It's unclear from
the fortran-95 standard whether gfortran is allowed to optimize this
even without -ffast-math.


-- 

roger at eyesopen dot com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED
   Target Milestone|---                         |4.2.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23452

[Bug tree-optimization/23452] Optimizing CONJG_EXPR (a) * a

[tkoenig at gcc dot gnu dot org]

------- Comment #4 from tkoenig at gcc dot gnu dot org  2006-06-01 19:13 -------
(In reply to comment #3)
> This is now fixed on mainline provided the user specifies -ffast-math.
> There are some complications where imagpart(z*~z) can be non-zero, if
> imagpart(z) is non-finite, such as an Inf or a NaN.  It's unclear from
> the fortran-95 standard whether gfortran is allowed to optimize this
> even without -ffast-math.

I'll ask con comp.lang.fortran :-)


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23452

[Bug tree-optimization/23452] Optimizing CONJG_EXPR (a) * a

[pinskia at gcc dot gnu dot org]

------- Comment #5 from pinskia at gcc dot gnu dot org  2006-11-01 04:30 -------
(In reply to comment #3)
> This is now fixed on mainline provided the user specifies -ffast-math.
> There are some complications where imagpart(z*~z) can be non-zero, if
> imagpart(z) is non-finite, such as an Inf or a NaN.  It's unclear from
> the fortran-95 standard whether gfortran is allowed to optimize this
> even without -ffast-math.

Well Fortran 95 does not have non-finite except with a TR and the IEEE fp
types, I forgot the number and the name


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23452

[Bug tree-optimization/23452] Optimizing CONJG_EXPR (a) * a

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-08-18 23:33 -------
Confirmed, there are two ways of fixing this, one in fold which folds CONJ_EXPR<a>*a, the other way is 
to do it later on via somthing like like a combiner or even in complex exand.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|                            |1
   Last reconfirmed|0000-00-00 00:00:00         |2005-08-18 23:33:48
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23452