Improve matching of loadNKlassCompactHeaders on aarch64

Author: rkennke

src/hotspot/cpu/aarch64/aarch64.ad

@@ -5512,6 +5512,10 @@ opclass memory(indirect, indIndexScaled, indIndexScaledI2L, indIndexI2L, indInde
                indirectN, indIndexScaledN, indIndexScaledI2LN, indIndexI2LN, indIndexN, indOffIN,
                indOffLN, indirectX2P, indOffX2P);
 
+opclass memory_noindex(indirect, indOffI, indOffL,
+                       indirectN, indOffIN,
+                       indOffLN, indirectX2P, indOffX2P);
+
 // iRegIorL2I is used for src inputs in rules for 32 bit int (I)
 // operations. it allows the src to be either an iRegI or a (ConvL2I
 // iRegL). in the latter case the l2i normally planted for a ConvL2I
@@ -6448,17 +6452,18 @@ instruct loadNKlass(iRegNNoSp dst, memory mem)
   ins_pipe(iload_reg_mem);
 %}
 
-instruct loadNKlassCompactHeaders(iRegNNoSp dst, memory mem)
+instruct loadNKlassCompactHeaders(iRegNNoSp dst, memory_noindex mem)
 %{
   match(Set dst (LoadNKlass mem));
   predicate(!needs_acquiring_load(n) && UseCompactObjectHeaders);
 
   ins_cost(4 * INSN_COST);
-  format %{ "ldrw  $dst, $mem\t# compressed class ptr" %}
+  format %{ "load_nklass_compact  $dst, $mem\t# compressed class ptr" %}
   ins_encode %{
-    __ load_nklass_compact_c2($dst$$Register, $mem$$base$$Register, $mem$$index$$Register, $mem$$scale, $mem$$disp);
+    assert($mem$$index$$Register == noreg, "must not have indexed address");
+    __ load_nklass_compact_c2($dst$$Register, $mem$$base$$Register, $mem$$disp);
   %}
-  ins_pipe(pipe_slow);
+  ins_pipe(iload_reg_mem);
 %}
 
 // Load Float

src/hotspot/cpu/aarch64/c2_MacroAssembler_aarch64.cpp

@@ -2558,7 +2558,7 @@ bool C2_MacroAssembler::in_scratch_emit_size() {
   return MacroAssembler::in_scratch_emit_size();
 }
 
-void C2_MacroAssembler::load_nklass_compact_c2(Register dst, Register obj, Register index, int scale, int disp) {
+void C2_MacroAssembler::load_nklass_compact_c2(Register dst, Register obj, int disp) {
   // Note: Don't clobber obj anywhere in that method!
 
   // The incoming address is pointing into obj-start + klass_offset_in_bytes. We need to extract
@@ -2569,11 +2569,6 @@ void C2_MacroAssembler::load_nklass_compact_c2(Register dst, Register obj, Regis
   // When that happens, we need to lea the address into a single register, and subtract the
   // klass_offset_in_bytes, to get the address of the mark-word.
   int offset = oopDesc::mark_offset_in_bytes() + disp - oopDesc::klass_offset_in_bytes();
-  if (index == noreg) {
-    ldr(dst, Address(obj, offset));
-  } else {
-    lea(dst, Address(obj, index, Address::lsl(scale)));
-    ldr(dst, Address(dst, offset));
-  }
+  ldr(dst, Address(obj, offset));
   lsr(dst, dst, markWord::klass_shift);
 }

src/hotspot/cpu/aarch64/c2_MacroAssembler_aarch64.hpp

@@ -179,6 +179,6 @@
   void vector_signum_sve(FloatRegister dst, FloatRegister src, FloatRegister zero,
                          FloatRegister one, FloatRegister vtmp, PRegister pgtmp, SIMD_RegVariant T);
 
-  void load_nklass_compact_c2(Register dst, Register obj, Register index, int scale, int disp);
+  void load_nklass_compact_c2(Register dst, Register obj, int disp);
 
 #endif // CPU_AARCH64_C2_MACROASSEMBLER_AARCH64_HPP