[Bug fortran/17175] New: set_exponent breaks with integer*8 exponent

[Bug fortran/17175] New: set_exponent breaks with integer*8 exponent

[dje at gcc dot gnu dot org]

Fortran test intrinsic_set_exponent.f90 subroutine test_real8 invokes 
set_exponent with an integer*8 parameter.  set_exponent selects the specific 
version based on the precision of the real parameter, but not the integer 
parameter, which is assumed to be integer*4.  If the target ABI does not 
accidentally pass the integer*8 exponent parameter correctly for an integer*4 
argument, the intrinsic fails because nothing forces the parameter and function 
signature to match.

-- 
           Summary: set_exponent breaks with integer*8 exponent
           Product: gcc
           Version: 3.5.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: fortran
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: dje at gcc dot gnu dot org
                CC: gcc-bugs at gcc dot gnu dot org
 GCC build triplet: powerpc-ibm-aix5.2.0.0
  GCC host triplet: powerpc-ibm-aix5.2.0.0
GCC target triplet: powerpc-ibm-aix5.2.0.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17175

[Bug fortran/17175] set_exponent breaks with integer*8 exponent

[tobi at gcc dot gnu dot org]

------- Additional Comments From tobi at gcc dot gnu dot org  2004-08-26 16:05 -------
Indeed: 
[tobi@marktplatz tests]$ cat setexp.f90
x = 1.
x = set_exponent (x, 25_4)
x = set_exponent (x, 25_8)
end
[tobi@marktplatz tests]$ gfortran setexp.f90 -fdump-parse-tree -S | grep exponent_4
      ASSIGN x __set_exponent_4_4[[((x) (25))]]
      ASSIGN x __set_exponent_4_8[[((x) (25_8))]]
[tobi@marktplatz tests]$ grep exponent setexp.s
        call    _gfortran_set_exponent_r4
        call    _gfortran_set_exponent_r4

I.e. the information is lost when choosing the intrinsic.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|                            |1
   Last reconfirmed|0000-00-00 00:00:00         |2004-08-26 16:05:22
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17175

[Bug fortran/17175] set_exponent breaks with integer*8 exponent

[tobi at gcc dot gnu dot org]

------- Additional Comments From tobi at gcc dot gnu dot org  2004-08-26 16:16 -------
Indeed, gfc_get_intrinsic_lib_fndecl only looks at the first argument of a
function. The question is, should we rather convert the second argument to
default integer type, or if we should actually implement library functions for
all integer kinds.

I think the former is preferable, as an out of range exponent will generate an
unspecified result, so there's nothing to be lost when converting from integer*8
to integer*4 (and giving an exponent > 2*32 is calling for trouble).

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17175

[Bug fortran/17175] set_exponent breaks with integer*8 exponent

[tobi at gcc dot gnu dot org]

------- Additional Comments From tobi at gcc dot gnu dot org  2004-08-26 16:26 -------
This problem also affects the scale intrinsic:
[tobi@marktplatz tests]$ cat setexp.f90
x = 1.
x = scale (x, 25_4)
x = scale (x, 25_8)
end
[tobi@marktplatz tests]$ gfortran setexp.f90 -fdump-parse-tree -S | grep scale_4
      ASSIGN x __scale_4_4[[((x) (25))]]
      ASSIGN x __scale_4_4[[((x) (25_8))]]
[tobi@marktplatz tests]$ grep scal setexp.s
        call    scalbnf ; here an int*4 argument is passed
        call    scalbnf ; here an int*8 argument is passed
[tobi@marktplatz tests]$

And ATAN2 has a problem WRT error reporting:
[tobi@marktplatz tests]$ cat setexp.f90
double precision y
x = 1.; y =1.
x = atan2 (x, y)
end
[tobi@marktplatz tests]$ gfortran setexp.f90
 In file setexp.f90:3

x = atan2 (x, y)
            1
Error: Type of argument 'x' in call to 'atan2' at (1) should be REAL(4), not REAL(8)
[tobi@marktplatz tests]$

AFAICT these are the only intrinsics that could possibly be affected by the same
bug.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17175

[Bug fortran/17175] set_exponent breaks with integer*8 exponent

[tobi at gcc dot gnu dot org]

------- Additional Comments From tobi at gcc dot gnu dot org  2004-08-26 16:55 -------
I posted patch for the issue with ATAN2 here:
http://gcc.gnu.org/ml/fortran/2004-08/msg00234.html

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17175

[Bug fortran/17175] set_exponent breaks with integer*8 exponent

[sgk at troutmask dot apl dot washington dot edu]

------- Additional Comments From sgk at troutmask dot apl dot washington dot edu  2004-08-26 17:33 -------
Tobi,

I think you are correct that we should convert integer*8 to integer*4.
We probably want to check that the integer*8 value lies between emin
and emax for the relevant real kind, then to the conversion.

-- 
steve


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17175

[Bug fortran/17175] set_exponent breaks with integer*8 exponent

[Tobias dot Schlueter at physik dot uni-muenchen dot de]

------- Additional Comments From Tobias dot Schlueter at physik dot uni-muenchen dot de  2004-08-26 18:33 -------
Subject: Re:  set_exponent breaks with integer*8 exponent

sgk at troutmask dot apl dot washington dot edu wrote:
> I think you are correct that we should convert integer*8 to integer*4.
> We probably want to check that the integer*8 value lies between emin
> and emax for the relevant real kind, then to the conversion.

Range checking would have to happen at runtime, and therefore is costly. I
think that if the user asks for set_exponent(1.,325876358) it would be nice to
return infinity (I don't know if we currently do so) but it's not mandated by
the standard.

Actually, I think the best fix would be having a builtin do the job.
set_exponent is easily implemented: load floating point number in integer
register, replace exponent bits with the exponent the user has asked for
(after adjusting for the bias etc.), return result as float wherever it
belongs. This would also avoid inserting conversions in the frontend.

I'll see if I can do this, I assume that it would be an easy job for someone
who knows more about the frontend, though.

- Tobi


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17175

[Bug fortran/17175] set_exponent breaks with integer*8 exponent

[Tobias dot Schlueter at physik dot uni-muenchen dot de]

------- Additional Comments From Tobias dot Schlueter at physik dot uni-muenchen dot de  2004-08-26 18:52 -------
Subject: Re:  set_exponent breaks with integer*8 exponent

Er,

Tobias dot Schlueter at physik dot uni-muenchen dot de wrote:
> I'll see if I can do this, I assume that it would be an easy job for someone
> who knows more about the frontend, though.
                           ^^^^^^^^^ backend


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17175

[Bug fortran/17175] set_exponent breaks with integer*8 exponent

[paulthomas2 at wanadoo dot fr]

------- Additional Comments From paulthomas2 at wanadoo dot fr  2004-12-10 13:35 -------
You have fixed this one, haven't you? cvs 20041205 runs it fine.

However, in testing, I found another nasty:

y = 1/2 + 1/8
!y= 0.125 + 0.5                     !!works OK
x = set_exponent (y, 5_4)
print *,x
x = set_exponent (y, 5_8)
print *,x
end

gives 

$ ./a
   0.000000
   0.000000

instead of 

   20.00000
   20.00000


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17175

[Bug fortran/17175] set_exponent breaks with integer*8 exponent

[Tobias dot Schlueter at physik dot uni-muenchen dot de]

------- Additional Comments From Tobias dot Schlueter at physik dot uni-muenchen dot de  2004-12-10 13:48 -------
Subject: Re:  set_exponent breaks with integer*8 exponent

paulthomas2 at wanadoo dot fr wrote:
> ------- Additional Comments From paulthomas2 at wanadoo dot fr  2004-12-10 13:35 -------
> You have fixed this one, haven't you? cvs 20041205 runs it fine.

I don't think so, maybe it's got hidden by another change.

> However, in testing, I found another nasty:
> 
> y = 1/2 + 1/8
> !y= 0.125 + 0.5                     !!works OK

1/2 .EQ. 0 in Fortran, same for 1/8, this is integer division :-)

- Tobi


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17175

[Bug fortran/17175] set_exponent breaks with integer*8 exponent

[tobi at gcc dot gnu dot org]

------- Additional Comments From tobi at gcc dot gnu dot org  2004-12-10 14:53 -------
BTW Paul, this depends on the endianess of your target, the error won't show up
on i686.  I'm currently testing a patch which should really fix this.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17175

[Bug fortran/17175] set_exponent breaks with integer*8 exponent

[tobi at gcc dot gnu dot org]

------- Additional Comments From tobi at gcc dot gnu dot org  2004-12-10 15:32 -------
Patch here: http://gcc.gnu.org/ml/fortran/2004-12/msg00099.html

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |patch, wrong-code


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17175

[Bug fortran/17175] set_exponent breaks with integer*8 exponent

[cvs-commit at gcc dot gnu dot org]

------- Additional Comments From cvs-commit at gcc dot gnu dot org  2004-12-11 11:06 -------
Subject: Bug 17175

CVSROOT:	/cvs/gcc
Module name:	gcc
Changes by:	tobi@gcc.gnu.org	2004-12-11 11:06:31

Modified files:
	gcc/fortran    : ChangeLog iresolve.c 
	gcc/testsuite  : ChangeLog 
Added files:
	gcc/testsuite/gfortran.dg: scale_1.f90 

Log message:
	fortran/
	PR fortran/17175
	* iresolve.c (gfc_resolve_scale): Convert 'I' argument if not of
	same kind as C's 'int'.
	(gfc_resolve_set_eponent): Convert 'I' argument if not of kind 4.
	
	testsuite/
	PR fortran/17175
	* gfortran.dg/scale_1.f90: New test.

Patches:
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/fortran/ChangeLog.diff?cvsroot=gcc&r1=1.266&r2=1.267
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/fortran/iresolve.c.diff?cvsroot=gcc&r1=1.24&r2=1.25
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/ChangeLog.diff?cvsroot=gcc&r1=1.4737&r2=1.4738
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/gfortran.dg/scale_1.f90.diff?cvsroot=gcc&r1=NONE&r2=1.1



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17175

[Bug fortran/17175] set_exponent breaks with integer*8 exponent

[tobi at gcc dot gnu dot org]

------- Additional Comments From tobi at gcc dot gnu dot org  2004-12-11 11:10 -------
I'm sure this is fixed, but I don't have one of the affected platforms. I'd be
glad if you could look at your testresults closely, and close this bug, if
appropriate.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED
   Last reconfirmed|2004-11-28 02:50:31         |2004-12-11 11:10:39
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17175

[Bug fortran/17175] set_exponent breaks with integer*8 exponent

[dje at gcc dot gnu dot org]

------- Additional Comments From dje at gcc dot gnu dot org  2004-12-12 17:38 -------
The patch fixes the problem on big-endian systems.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
         Resolution|                            |FIXED


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17175

[Bug fortran/17175] set_exponent breaks with integer*8 exponent

[Tobias dot Schlueter at physik dot uni-muenchen dot de]

------- Additional Comments From Tobias dot Schlueter at physik dot uni-muenchen dot de  2004-12-12 18:53 -------
Subject: Re:  set_exponent breaks with integer*8 exponent

dje at gcc dot gnu dot org wrote:
> ------- Additional Comments From dje at gcc dot gnu dot org  2004-12-12 17:38 -------
> The patch fixes the problem on big-endian systems.
> 
Thanks!


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17175

[Bug fortran/17175] set_exponent breaks with integer*8 exponent

[pinskia at gcc dot gnu dot org]

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |4.0.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17175