[Bug c/29538] New: Incorrect interpretation of floating-point unary minus

[Bug c/29538] New: Incorrect interpretation of floating-point unary minus

[guillaume dot melquiond at ens-lyon dot fr]

Testcase:
void f(float);
void g(double x) { f(-x); }

Using -fdump-tree-original shows that GCC understands the above code as
"f(-(float)x)", although it should really be "f((float)(-x))". As a
consequence, the generated code is not efficient (on x86), as GCC will have to
store x to memory as a float before loading it again and changing its sign.
Moreover, the computed value is not the expected one when the rounding mode is
a directed rounding, as the conversion happens before the sign change.

Tested with Debian GCC 3.3.6, 3.4.6, 4.0.4, and 4.1.2. Version 3.3 generates
the correct code, the other ones don't.


-- 
           Summary: Incorrect interpretation of floating-point unary minus
           Product: gcc
           Version: 4.1.2
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: guillaume dot melquiond at ens-lyon dot fr


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29538

[Bug target/29538] Incorrect interpretation of floating-point unary minus

[rguenth at gcc dot gnu dot org]

------- Comment #1 from rguenth at gcc dot gnu dot org  2006-10-21 18:19 -------
Unary negation does not affect rounding, so I fail to see how it shall matter
what type the negation is done on.

It is true that for x87 math the -(float)x sequence is slower:

g:
        pushl   %ebp
        movl    %esp, %ebp
        subl    $4, %esp
        fldl    8(%ebp)
        fstps   -4(%ebp)
        flds    -4(%ebp)
        fchs
        fstps   8(%ebp)
        leave
        jmp     f

so, confirmed as missed-optimization.


-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |rguenth at gcc dot gnu dot
                   |                            |org
             Status|UNCONFIRMED                 |NEW
          Component|c                           |target
     Ever Confirmed|0                           |1
 GCC target triplet|                            |i?86-*-*
           Keywords|                            |missed-optimization
   Last reconfirmed|0000-00-00 00:00:00         |2006-10-21 18:19:02
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29538

[Bug target/29538] Incorrect interpretation of floating-point unary minus

[pinskia at gcc dot gnu dot org]

------- Comment #2 from pinskia at gcc dot gnu dot org  2006-10-21 20:17 -------
Though minus cannot change the rounding but the rounding mode can be different
for rounding from double to float in that we could have a negative number with
round towards zero which causes us to get different answers.  Now the default
rounding we are fine.

*** This bug has been marked as a duplicate of 21032 ***


-- 

pinskia at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |DUPLICATE


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29538