https://gcc.gnu.org/bugzilla/show_bug.cgi?id=66703
--- Comment #3 from H.J. Lu <hjl.tools at gmail dot com> --- (In reply to Yulia Koval from comment #2) > Why zero-extend with and is better than zero-extend with movz? Why it's ok > to clobber the flags? According to Intel Pentium optimization guide, zero-extend with and is faster than zero-extend with movz. It is OK to clobber EFLAGS to do unsigned int x = (unsigned int) b; where b is a boolean value.
