x86 inline assembly should provide an alternative for Intel syntax

[lhmouse]

Instead of

thread/include/boost/thread/win32/interlocked_read.hpp

Line 158 in 52b62ee

__asm__ __volatile__ ("movl %1, %0" : "=r" (res) : "m" (*x) : "memory");

we should have

 __asm__ __volatile__ ("{ movl %1, %0 | mov %0, %1 }" : "=r" (res) : "m" (*x) : "memory");