This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

[PATCH rs6000] Fix PR 37640, __sync_lock_test_and_set on PPC64 causes ICE


Hi,
  The issue here is that the rs6000 back-end tries to take the lower
part (via gen_lowpart_common) of a plus rtl.  This causes an ICE as it
is not handled.  The way to correct this is to force the address (the
plus RTL) to a register and then take the lower part of that register.

OK? Bootstrapped and tested on ppc64-linux-gnu with no regressions.

Thanks,
Andrew Pinski

ChangeLog:
* config/rs6000/rs6000.c (rs6000_expand_compare_and_swapqhi): Force
address to a register before taking the lower part.

testsuite/ChangeLog:
* gcc.c-torture/compile/sync-3.c: New testcase to check that addresses
of non zero offset works.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]