https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61725
Bug ID: 61725
Summary: [4.9 Regression] __builtin_ffs(0) leads to bad code
generation
Product: gcc
Version: 4.9.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: middle-end
Assignee: unassigned at gcc dot gnu.org
Reporter: jtaylor.debian at googlemail dot com
__builtin_ffs is well defined for 0 according to the gcc documentation:
Returns one plus the index of the least significant 1-bit of x, or if x is
zero, returns zero.
yet for following program gcc 4.9 and gcc-4.10 20140627 optimizes away the
conditional on the result value for zero inputs.
#include <stdio.h>
int main(int argc, char *argv[])
{
int x;
for (x = -128; x <= 128; x++) {
int a = __builtin_ffs(x);
if (x == 0 && a != 0)
printf("%#x: %u\n", x, a);
}
return 0;
}
gcc test.c -O2
./a.out
# this should not print:
0: 0
disassembly:
10: 83 c3 01 add $0x1,%ebx
13: 81 fb 81 00 00 00 cmp $0x81,%ebx
19: 74 26 je 41 <main+0x41>
1b: 0f bc d3 bsf %ebx,%edx
1e: 0f 44 d5 cmove %ebp,%edx
21: 83 c2 01 add $0x1,%edx
24: 85 db test %ebx,%ebx <<<<<<<< test on ebx
but not edx
26: 75 e8 jne 10 <main+0x10>
28: 31 f6 xor %esi,%esi
2a: 31 c0 xor %eax,%eax
2c: bf 00 00 00 00 mov $0x0,%edi
31: 83 c3 01 add $0x1,%ebx
34: e8 00 00 00 00 callq 39 <main+0x39>
39: 81 fb 81 00 00 00 cmp $0x81,%ebx
/tmp$ gcc -v
Using built-in specs.
COLLECT_GCC=gcc
COLLECT_LTO_WRAPPER=/tmp/gcc/libexec/gcc/x86_64-unknown-linux-gnu/4.10.0/lto-wrapper
Target: x86_64-unknown-linux-gnu
Configured with: ./configure --enable-languages=c --disable-bootstrap --prefix
/tmp/gcc CFLAGS='-g3 -O0' CXXFLAGS='-g3 -O0'
Thread model: posix
gcc version 4.10.0 20140627 (experimental) (GCC)
gcc 4.8.2 appears to work fine
