https://gcc.gnu.org/bugzilla/show_bug.cgi?id=127244
--- Comment #3 from Uroš Bizjak <ubizjak at gmail dot com> --- (In reply to Drea Pinski from comment #2) > The question becomes should the constraint on this inline assembly be "=r,m" > or "=rm". The former gives the decent code. With "=r,m", the compiler creates the following sequences in the kernel: 2d720c: 9c pushf 2d720d: 41 5c pop %r12 2d720f: 4c 89 64 24 18 mov %r12,0x18(%rsp) 2d7214: fa cli In this case %r12 temporary is not needed, but this might be PR124209. Otherwise, "=r,m" variant looks much better. But... shouldn't these two be equal as far as RA is concerned?
