In c1_LIRGenerator_x86.cpp, towards the end of LIRGenerator::atomic_cmpxchg(..) there's this cmove:

  __ cmove(lir_cond_equal, LIR_OprFact::intConst(1), LIR_OprFact::intConst(0),
           result, type);

which should use T_INT instead of the passed in type.