8333154: RISC-V: Add support for primitive array C1 clone intrinsic

Reviewed-by: fyang

Author: zifeihan

src/hotspot/cpu/riscv/c1_LIRAssembler_arraycopy_riscv.cpp

@@ -341,7 +341,9 @@ void LIR_Assembler::emit_arraycopy(LIR_OpArrayCopy* op) {
     __ call_VM_leaf(entry, args_num);
   }
 
-  __ bind(*stub->continuation());
+  if (stub != nullptr) {
+    __ bind(*stub->continuation());
+  }
 }
 
 

src/hotspot/cpu/riscv/c1_LIRAssembler_riscv.cpp

@@ -1023,7 +1023,8 @@ void LIR_Assembler::emit_alloc_array(LIR_OpAllocArray* op) {
                       arrayOopDesc::base_offset_in_bytes(op->type()),
                       array_element_size(op->type()),
                       op->klass()->as_register(),
-                      *op->stub()->entry());
+                      *op->stub()->entry(),
+                      op->zero_array());
   }
   __ bind(*op->stub()->continuation());
 }

src/hotspot/cpu/riscv/c1_LIRGenerator_riscv.cpp

@@ -733,7 +733,13 @@ void LIRGenerator::do_ArrayCopy(Intrinsic* x) {
   assert(x->number_of_arguments() == 5, "wrong type");
 
   // Make all state_for calls early since they can emit code
-  CodeEmitInfo* info = state_for(x, x->state());
+  CodeEmitInfo* info = nullptr;
+  if (x->state_before() != nullptr && x->state_before()->force_reexecute()) {
+    info = state_for(x, x->state_before());
+    info->set_force_reexecute();
+  } else {
+    info = state_for(x, x->state());
+  }
 
   LIRItem src(x->argument_at(0), this);
   LIRItem src_pos(x->argument_at(1), this);
@@ -766,6 +772,9 @@ void LIRGenerator::do_ArrayCopy(Intrinsic* x) {
   int flags;
   ciArrayKlass* expected_type = nullptr;
   arraycopy_helper(x, &flags, &expected_type);
+  if (x->check_flag(Instruction::OmitChecksFlag)) {
+    flags = 0;
+  }
 
   __ arraycopy(src.result(), src_pos.result(), dst.result(), dst_pos.result(), length.result(), tmp,
                expected_type, flags, info); // does add_safepoint
@@ -844,7 +853,13 @@ void LIRGenerator::do_NewInstance(NewInstance* x) {
 }
 
 void LIRGenerator::do_NewTypeArray(NewTypeArray* x) {
-  CodeEmitInfo* info = state_for(x, x->state());
+  CodeEmitInfo* info = nullptr;
+  if (x->state_before() != nullptr && x->state_before()->force_reexecute()) {
+    info = state_for(x, x->state_before());
+    info->set_force_reexecute();
+  } else {
+    info = state_for(x, x->state());
+  }
 
   LIRItem length(x->length(), this);
   length.load_item_force(FrameMap::r9_opr);
@@ -861,7 +876,7 @@ void LIRGenerator::do_NewTypeArray(NewTypeArray* x) {
   __ metadata2reg(ciTypeArrayKlass::make(elem_type)->constant_encoding(), klass_reg);
 
   CodeStub* slow_path = new NewTypeArrayStub(klass_reg, len, reg, info);
-  __ allocate_array(reg, len, tmp1, tmp2, tmp3, tmp4, elem_type, klass_reg, slow_path);
+  __ allocate_array(reg, len, tmp1, tmp2, tmp3, tmp4, elem_type, klass_reg, slow_path, x->zero_array());
 
   LIR_Opr result = rlock_result(x);
   __ move(reg, result);

src/hotspot/cpu/riscv/c1_MacroAssembler_riscv.cpp

@@ -282,7 +282,7 @@ void C1_MacroAssembler::initialize_object(Register obj, Register klass, Register
   verify_oop(obj);
 }
 
-void C1_MacroAssembler::allocate_array(Register obj, Register len, Register tmp1, Register tmp2, int base_offset_in_bytes, int f, Register klass, Label& slow_case) {
+void C1_MacroAssembler::allocate_array(Register obj, Register len, Register tmp1, Register tmp2, int base_offset_in_bytes, int f, Register klass, Label& slow_case, bool zero_array) {
   assert_different_registers(obj, len, tmp1, tmp2, klass);
 
   // determine alignment mask
@@ -308,7 +308,9 @@ void C1_MacroAssembler::allocate_array(Register obj, Register len, Register tmp1
 
   // clear rest of allocated space
   const Register len_zero = len;
-  initialize_body(obj, arr_size, base_offset, len_zero);
+  if (zero_array) {
+    initialize_body(obj, arr_size, base_offset, len_zero);
+  }
 
   membar(MacroAssembler::StoreStore);
 

src/hotspot/cpu/riscv/c1_MacroAssembler_riscv.hpp

@@ -101,7 +101,8 @@ using MacroAssembler::null_check;
   // base_offset_in_bytes: offset of first array element, in bytes
   // f                   : element scale factor
   // slow_case           : exit to slow case implementation if fast allocation fails
-  void allocate_array(Register obj, Register len, Register tmp1, Register tmp2, int base_offset_in_bytes, int f, Register klass, Label& slow_case);
+  // zero_array          : zero the allocated array or not
+  void allocate_array(Register obj, Register len, Register tmp1, Register tmp2, int base_offset_in_bytes, int f, Register klass, Label& slow_case, bool zero_array);
 
   int  rsp_offset() const { return _rsp_offset; }
 

src/hotspot/share/c1/c1_Compiler.cpp

@@ -235,7 +235,7 @@ bool Compiler::is_intrinsic_supported(vmIntrinsics::ID id) {
   case vmIntrinsics::_counterTime:
 #endif
   case vmIntrinsics::_getObjectSize:
-#if defined(X86) || defined(AARCH64) || defined(S390)
+#if defined(X86) || defined(AARCH64) || defined(S390) || defined(RISCV)
   case vmIntrinsics::_clone:
 #endif
     break;

src/hotspot/share/c1/c1_LIR.cpp

@@ -351,7 +351,7 @@ LIR_OpArrayCopy::LIR_OpArrayCopy(LIR_Opr src, LIR_Opr src_pos, LIR_Opr dst, LIR_
   , _tmp(tmp)
   , _expected_type(expected_type)
   , _flags(flags) {
-#if defined(X86) || defined(AARCH64) || defined(S390)
+#if defined(X86) || defined(AARCH64) || defined(S390) || defined(RISCV)
   if (expected_type != nullptr && flags == 0) {
     _stub = nullptr;
   } else {