Well I tried posting this to the gcc listserv and was met with deafening
silence so will try cross-posting this to the 390 list (sorry).
Compiling with gcc/g++ for s390x, register constraints don't allow
specification of specific registers as with x86. But, because certain
instructions work on p
While I believe things are misbehaving, is there any reason not to use the
_builtin_ functions to do the atomic exchange?
If you need to use your own then in the asm block unload %0 etc. into the
registers you want, e.g.: lgr 2,%0 and code the cdsg with those registers. In
the constraints
Hi,
this looks like a problem I've debugged last year:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=33661
I've also proposed a fix for that but unfortunately forgot to ping this after a
while:
https://gcc.gnu.org/ml/gcc-patches/2015-06/msg00838.html
I'll try to get this in after GCC 6 release.
Whoops, I guess I wasnt patient enough on the gcc listserv. I got an answer
on the gcc listserv from Andreas Krebbel which indicates that the bug is
well known: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=33661 and that
there is a fix in the works. Sorry about the noise.
From: Alex Kodat [mail