https://gcc.gnu.org/bugzilla/show_bug.cgi?id=82318
Bug ID: 82318
Summary: -fexcess-precision=standard has no effect on a libm
function call
Product: gcc
Version: 8.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: c
Assignee: unassigned at gcc dot gnu.org
Reporter: vincent-gcc at vinc17 dot net
Target Milestone: ---
Consider the following C program:
#include <math.h>
int main (void)
{
double r = 5.0;
return (double) log2 (r) == (double) log2 (r) ? 0 : 1;
}
When compiled on x86_64 with -O0 -m32 -fexcess-precision=standard, the program
returns 1 instead of 0.
The generated asm code contains:
[...]
fldl .LC0
fstpl -16(%ebp)
subl $8, %esp
pushl -12(%ebp)
pushl -16(%ebp)
call log2
addl $16, %esp
fstpl -32(%ebp)
subl $8, %esp
pushl -12(%ebp)
pushl -16(%ebp)
call log2
addl $16, %esp
fldl -32(%ebp)
fucomi %st(1), %st
[...]
If I understand correctly, the result of the first log2 is stored into memory,
but not the second one.
