8353237: [AArch64] Incorrect result of VectorizedHashCode intrinsic on Cortex-A53

src/hotspot/cpu/aarch64/c2_MacroAssembler_aarch64.cpp

@@ -108,14 +108,20 @@ address C2_MacroAssembler::arrays_hashcode(Register ary, Register cnt, Register
   assert(is_power_of_2(unroll_factor), "can't use this value to calculate the jump target PC");
   andr(tmp2, cnt, unroll_factor - 1);
   adr(tmp1, BR_BASE);
-  sub(tmp1, tmp1, tmp2, ext::sxtw, 3);
+  // For Cortex-A53 offset is 4 because 2 nops are generated.
+  sub(tmp1, tmp1, tmp2, ext::sxtw, VM_Version::supports_a53mac() ? 4 : 3);
   movw(tmp2, 0x1f);
   br(tmp1);
 
   bind(LOOP);
   for (size_t i = 0; i < unroll_factor; ++i) {
     load(tmp1, Address(post(ary, type2aelembytes(eltype))), eltype);
     maddw(result, result, tmp2, tmp1);
+    // maddw generates an extra nop for Cortex-A53 (see maddw definition in macroAssembler).
+    // Generate 2nd nop to have 4 instructions per iteration.
+    if (VM_Version::supports_a53mac()) {
+      nop();
+    }
   }
   bind(BR_BASE);
   subsw(cnt, cnt, unroll_factor);

src/hotspot/cpu/aarch64/stubGenerator_aarch64.cpp

@@ -5496,14 +5496,20 @@ class StubGenerator: public StubCodeGenerator {
     __ andr(rscratch2, cnt, vf - 1);
     __ bind(TAIL_SHORTCUT);
     __ adr(rscratch1, BR_BASE);
-    __ sub(rscratch1, rscratch1, rscratch2, ext::uxtw, 3);
+    // For Cortex-A53 offset is 4 because 2 nops are generated.
+    __ sub(rscratch1, rscratch1, rscratch2, ext::uxtw, VM_Version::supports_a53mac() ? 4 : 3);
     __ movw(rscratch2, 0x1f);
     __ br(rscratch1);
 
     for (size_t i = 0; i < vf - 1; ++i) {
       __ load(rscratch1, Address(__ post(ary, type2aelembytes(eltype))),
                                    eltype);
       __ maddw(result, result, rscratch2, rscratch1);
+      // maddw generates an extra nop for Cortex-A53 (see maddw definition in macroAssembler).
+      // Generate 2nd nop to have 4 instructions per iteration.
+      if (VM_Version::supports_a53mac()) {
+        __ nop();
+      }
     }
     __ bind(BR_BASE);