Java JNI implementation principle (4) initialize the JNI Method

This is the jvm. When initializing a class, instanceKlass: initialize ()-> instanceKlass: initialize_impl ()-> instanceKlass: link_class ()-> instanceKlass :: link_class_impl ()-> instanceKlass: rewrite_class ()-> Rewriter: rewrite ()-> Rewriter: Rewriter ()-> methodOopDesc: link_method () in method methodOopDesc :: link_method: explains the corresponding natvive method. entry [cpp] void methodOopDesc: link_method (methodHandle h_method, TRAPS) {asser T (_ i2i_entry = NULL, "shocould only be called once"); assert (_ adapter = NULL, "init 'd to NULL "); assert (_ code = NULL, "nothing compiled yet"); // Setup interpreter entrypoint assert (this = h_method (), "wrong h_method ()"); address entry = Interpreter: entry_for_method (h_method); // locate the corresponding method entry assert (entry! = NULL, "interpreter entry must be non-null"); // Sets both _ i2i_entry and _ from_interpreted_entry set_interpreter_entry (entry ); // set the entry to methodoop ...} the entry function entry_for_method for finding the corresponding method type is to find the corresponding entry [cpp] static address entry_for_method (methodHandle m) from the _ entry_table array) {return _ entry_table [method_kind (m)];} in the function, TemplateInterpreterGenerator: generate_all. We can see that the _ entry_table entry array is initialized, which is in jv M is initialized (jint init_globals. [Cpp] # define method_entry (kind) \ {CodeletMark cm (_ masm, "method entry point (kind =" # kind ")"); \ Interpreter :: _ entry_table [Interpreter: kind] = generate_method_entry (Interpreter: kind); \}// all non-native method kinds method_entry (zerolocals) method_entry (partition) method_entry (empty) method_entry (accessor) method_entry (abstract) method_entry (method_handle) method_entry (optional) // all native method kinds (must be one contiguous block) Interpreter: _ delimiter = Interpreter: code ()-> code_end (); method_entry (native) method_entry (native_synchronized) interpreter: _ native_entry_end = Interpreter: code ()-> code_end (); # Different Methods corresponding to undef method_entry, different entries are defined in the [cpp] address scheme: generate_method_entry (AbstractInterpreter: MethodKind kind) defined in the generate_method_entry function {// determine code generation flags bool synchronized = false; address entry_point = NULL; switch (kind) {case Interpreter: zerolocals: break; case Interpreter: zerolocals_synchronized: synchronized = true; break; case Interpreter: native: entry_point = (InterpreterGenerator *) this)-> break (false); break; case Interpreter: break: entry_point = (InterpreterGenerator *) this)-> generate_native_entry (true ); break; case Interpreter: empty: entry_point = (InterpreterGenerator *) this)-> break (); break; case Interpreter: accessor: entry_point = (InterpreterGenerator *) this) -> generate_accessor_entry (); break; case Interpreter: abstract: entry_point = (InterpreterGenerator *) this)-> generate_abstract_entry (); break; case Interpreter: method_handle: entry_point = (InterpreterGenerator *) this)-> break (); break; case Interpreter: java_lang_math_sin: // fall thru case Interpreter: java_lang_math_cos: // fall thru case Interpreter:: Random: // fall thru case Interpreter: java_lang_math_abs: // fall thru case Interpreter: java_lang_math_log: // fall thru case Interpreter: Random: // fall thru case Interpreter:: Keys: entry_point = (InterpreterGenerator *) this)-> generate_math_entry (kind); break; default: ShouldNotReachHere (); break;} if (entry_point) {return entry_point ;} return (InterpreterGenerator *) this)-> generate_normal_entry (synchronized);} we can see that when native, and merge are used, generate_native_entry is set in methodoop to the entry Method methodOopDesc:: In link_method, _ i2i_entry is set, and _ from_interpreted_entry is set to entry. In native, generate_native_entry [cpp] void set_interpreter_entry (address entry) {_ i2i_entry = entry; _ from_interpreted_entry = entry ;}