optimization/10667: Optimizer confused when operating on floats as ints.

optimization/10667: Optimizer confused when operating on floats as ints.

[jkain]

>Number:         10667
>Category:       optimization
>Synopsis:       Optimizer confused when operating on floats as ints.
>Confidential:   no
>Severity:       critical
>Priority:       medium
>Responsible:    unassigned
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Wed May 07 16:36:00 UTC 2003
>Closed-Date:
>Last-Modified:
>Originator:     Joseph Kain
>Release:        gcc version 3.2.3
>Organization:
>Environment:
Linux 2.4, i686.
>Description:
The program below was built as:
  gcc -v -save-temps -O3 -c -o bug.o bug.c

$ gcc -v
Reading specs from /home/joseph/t/gcc/lib/gcc-lib/i686-pc-linux-gnu/3.2.3/specs
Configured with: ../gcc-3.2.3/configure --prefix=/home/joseph/t/gcc
Thread model: posix
gcc version 3.2.3


If I examine the assembly output I find that the compiler has completely removed the multiply.  Code for the compare is generated but the compare result may be wrong since f.x was never computed.
>How-To-Repeat:
# 1 "bug.c"
# 1 "<built-in>"
# 1 "<command line>"
# 1 "bug.c"


struct vector {
    float x, y, z, w;
};

int func (float f1, float f2) {
    struct vector f;

    f.x = f1 * f2;

    if ((*(int *) &(f.x)) > 0) {
        return 1;
    }

    return 0;
}
>Fix:

>Release-Note:
>Audit-Trail:
>Unformatted:

Re: optimization/10667: Optimizer confused when operating on floats as ints.

[ebotcazou]

Synopsis: Optimizer confused when operating on floats as ints.

State-Changed-From-To: open->closed
State-Changed-By: ebotcazou
State-Changed-When: Wed May  7 17:01:24 2003
State-Changed-Why:
    Not a bug. Your code breaks the ISO C aliasing rules. GCC
    automatically turns on -fstrict-aliasing at -O2 so you need
    to fix your code or compile it with -fno-strict-aliasing.

http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&database=gcc&pr=10667