[Bug rtl-optimization/60763] [4.9 Regression] ICE in extract_insn starting with rev 208984

[rsandifo at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60763

--- Comment #4 from rsandifo at gcc dot gnu.org <rsandifo at gcc dot gnu.org> ---
Created attachment 32553
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=32553&action=edit
Patch for 60772 testcase

OK, I could reproduce it with Andreas's instructions after configuring
with a real powerpc64-linux-gnu assembler.  I think the lack of an
assembler was causing some important configure test to fail.

One fix for the PR60772 case is attached.  It takes the same approach
as rs6000_split_multireg_move and IMO is preferable to allowing the
subreg in the movdi pattern.  The output is the same as before the
general_operand change.