https://gcc.gnu.org/bugzilla/show_bug.cgi?id=69343
--- Comment #1 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
libitm configury really should check if compiling something with -march=z13 on
s390{,x}-* succeeds. The question is what to do if it fails, whether we create
an ABI incompatible version of libitm in that case, or if it is possible to
provide the 128-bit vector entrypoints even without assembler support.
Related to PR69339 which shows perhaps similar need for powerpc -maltivec.
BTW, I'm surprised there is no vector256.cc and vector512.cc too, and while
libitm.h defines _ITM_TYPE_M256 on x86, it doesn't define _ITM_TYPE_M512.
