https://gcc.gnu.org/bugzilla/show_bug.cgi?id=125197
--- Comment #6 from H.J. Lu <hjl.tools at gmail dot com> --- (In reply to Drea Pinski from comment #5) > (In reply to H.J. Lu from comment #3) > > When -m32 is used, sparc64 GCC pass > > > > -mptr32 -mno-stack-bias -mlong-double-64 -mcpu=cypress > > > > to cc1plus. > > (In reply to Drea Pinski from comment #4) > > I am not sure compiling for sparc64 -m32 multilib with -march=v9 will valid > > thing to do because that multilib is supposed to be able to run for v8 code > > too. > > In fact it defaults to -mcpu=cypress which is a v7. > The bug is NOT in GCC here. But -m32 libatomic is built with -mcpu=v9 for sparc64. Why is there a discrepancy between -m32 libatomic and -m32 libstdc++ for sparc64?
