http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53190
--- Comment #4 from Richard Guenther <rguenth at gcc dot gnu.org> 2012-05-03 07:59:40 UTC --- (In reply to comment #3) > (In reply to comment #2) > > (clobber (reg:QI 18 fpsr)) > > (clobber (reg:QI 17 flags)) > > > > > > It might clobber those but that is not considered an use. So CSE is doing > > the > > correct thing as there is no use based on fpsr. > > The asm that depends on some unknown-to-gcc global processor state should be > marked as volatile. > > This is a bug in glibc. > > As a side note, mathinline.h really needs some serious TLC, there are many > functions that are much better implemented with gcc builtin functions, not to > mention that x87 asm in 32bit case interferes *badly* with -mfpmath=sse. > > Fortunately, there is -D__NO_MATH_INLINES, used in my projects from day one. > > Really, there is no justification for mathinline.h to live, as far as gcc is > concerned. So you say that it's correct for CSE to CSE asm()s if they have the same asm string and the same asm operands in case the asm is not volatile? I was not aware we would do that ;) (but yes, it sounds like a reasonable thing)
