[LLVMbugs] [Bug 1804] New: APFloat wrong handling of smallest normalized positive 32 bit float

[bugzilla-daemon at cs.uiuc.edu]

http://llvm.org/bugs/show_bug.cgi?id=1804

           Summary: APFloat wrong handling of smallest normalized positive
                    32 bit float
           Product: new-bugs
           Version: unspecified
          Platform: Other
        OS/Version: Linux
            Status: NEW
          Severity: enhancement
          Priority: P2
         Component: new bugs
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: baldrick at free.fr
                CC: llvmbugs at cs.uiuc.edu


Here's a C testcase showing the problem (originally spotted in the
Ada testsuite):

float F(void) {
  const int i = 8388608;
  const float f = *(float *)&i;
  return f;
}

$ llvm-gcc -S -O4 af.c
->
define float @F() pure  {
entry:
        ret float 0.000000e+00
}

The float has become zero, but 8388608 = 2^23 corresponds to
2^(-126), the smallest positive normalized float, and is non-zero.
It is also output as zero in the final assembler:
        .long   0       # float 0


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.