src/hotspot/cpu/aarch64/macroAssembler_aarch64.cpp

rev 50099 : 8203157: Object equals abstraction for BarrierSetAssembler

*** 3645,3654 ****
--- 3645,3659 ----
    adrp(rscratch1, src2, offset);
    ldr(rscratch1, Address(rscratch1, offset));
    cmp(src1, rscratch1);
  }
  
+ void MacroAssembler::cmpoop(Register obj1, Register obj2) {
+   BarrierSetAssembler* bs = BarrierSet::barrier_set()->barrier_set_assembler();
+   bs->obj_equals(this, IN_HEAP, obj1, obj2);
+ }
+ 
  void MacroAssembler::load_klass(Register dst, Register src) {
    if (UseCompressedClassPointers) {
      ldrw(dst, Address(src, oopDesc::klass_offset_in_bytes()));
      decode_klass_not_null(dst);
    } else {
*** 5034,5043 ****
--- 5039,5050 ----
      // if (a==null || a2==null)
      //     return false;
      // a1 & a2 == 0 means (some-pointer is null) or
      // (very-rare-or-even-probably-impossible-pointer-values)
      // so, we can save one branch in most cases
+     cmpoop(a1, a2);
+     br(EQ, SAME);
      eor(rscratch1, a1, a2);
      tst(a1, a2);
      mov(result, false);
      cbz(rscratch1, SAME);
      br(EQ, A_MIGHT_BE_NULL);
*** 5117,5127 ****
      mov(result, false);
      // on most CPUs a2 is still "locked"(surprisingly) in ldrw and it's
      // faster to perform another branch before comparing a1 and a2
      cmp(cnt1, elem_per_word);
      br(LE, SHORT); // short or same
!     cmp(a1, a2);
      br(EQ, SAME);
      ldr(tmp3, Address(pre(a1, base_offset)));
      cmp(cnt1, stubBytesThreshold);
      br(GE, STUB);
      ldr(tmp4, Address(pre(a2, base_offset)));
--- 5124,5134 ----
      mov(result, false);
      // on most CPUs a2 is still "locked"(surprisingly) in ldrw and it's
      // faster to perform another branch before comparing a1 and a2
      cmp(cnt1, elem_per_word);
      br(LE, SHORT); // short or same
!     cmpoop(a1, a2);
      br(EQ, SAME);
      ldr(tmp3, Address(pre(a1, base_offset)));
      cmp(cnt1, stubBytesThreshold);
      br(GE, STUB);
      ldr(tmp4, Address(pre(a2, base_offset)));