Bastian Ballmann <balle@xxxxxxxxxxx> writes: > If I disassemble the test function I see that gcc 4.5.2 produced the > following assembly: > > 0x08048374 <+0>: push %ebp > 0x08048375 <+1>: mov %esp,%ebp > 0x08048377 <+3>: sub $0x20,%esp > 0x0804837a <+6>: leave > 0x0804837b <+7>: ret > > I would expect that 0x1c or 0x18 byte will be substracted from esp, but > it's 0x20. See the -mpreferred-stack-boundary option. Ian