Page History

Sub-topics

Method handle fundamentals

Abstractly, a method handle is simply a type and some behavior that conforms to that type. As befits an object-oriented system, the behavior may include data.

Concretely, a method handle can refer to any JVM method, field, or constructor, or else it can be a transform of any previously specified method handle. Transforms include partial application (binding), filtering, and various forms of argument shuffling.

The method handle's type is expressed as a sequence of zero or more parameter types, and an optional return type (or the non-type void). Concretely, this is a MethodType reference, and can be extracted from any method handle using MethodHandle.type.

The behavior is what happens when the method handle is invoked, using the method MethodHandle.invokeExact. The special capability of method handles is that invokeExact accepts any number of any type of arguments, and can return any type or void. A regular invokevirtual instruction performs this. (It is rewritten secretly to invokehandle, as discussed below, but this can be ignored except by HotSpot implementors.)

Uniquely to method handles, the invokevirtual instruction can specify any structurally valid type signature, and the call site will link. Technically, we say that invokeExact is signature polymorphic. Practically speaking, when linking such a call site, the JVM must be ready to deal with any type signature, which means it will have to generate adapters of various sorts. From the user's point of view, a method handle is a magic thing which can wrap and/or invoke any method, of any type.

Concretely, the behavior of a method handle depends on a object called a LambdaForm, which is a low-level description of step-by-step operations. A method handle's lambda form is stored in its form field, just as its type is stored in its type field.

A method handle's lambda form may ignore the method handle completely and do something context-independent, like throw an exception or return zero. More generally, it can consult the method handle for information. For example, it can examine the method handle's return type and convert some value to that type before returning it.

More interestingly, if a method handle's class is a subclass which contains additional data fields, the lambda form can refer to those fields as it executes.

Since method handles express behavior more than state, their fields are typically immutable. But, method handles can easily be bound to arbitrary Java objects, producing closures.

The "basic type" system

In order to implement signature polymorphism more simply, method handles internally operate in terms of basic types. A basic type is a JVM type in which many inconvenient distinctions have been "erased", so that the remaining distinctions (such as reference vs. primitive and int vs. long) can be attended to.

For starters, in the basic type system, all 32-bit types except float are erased to simple int. If a byte value is required somewhere, it must be masked down from a full int. Thus, there are only four primitive types to worry about.

Under basic typing rules, all reference types are represented by java.lang.Object. Thus, there are a total of five basic types, represented by their JVM signature characters: L, I, J, F, D. To these we add V for the non-type void.

In the bulk of Java code, the full type system is in force. In order to name reference types, a system of class loaders and type constraints must be consulted and honored. From perspective of the JSR 292 runtime, this type system is a complex mix of names and scopes. Inside the runtime, using basic types there are no names to worry about, except Object and other types on the boot class path.

If a reference of a narrower type is required somewhere, an explicit checkcast must be issued before the reference is used. In fact, the checkcast is in general a call to Class.cast, with the specialized type being a constant Class reference rather than a symbolic reference name.

Normally, all extra conversions (such as int to byte and Object to a named reference type) disappear in the optimizer, which keeps track of full type information from context.

Lambda form basics

In brief, a lambda form is a classic lambda expression with zero or more formal parameters, plus zero or more body expressions. The types of parameters and expression values are drawn from the basic type system.

Each expression is simply the application of a method handle to zero or more arguments. Each argument is either a constant value or a previously specified parameter or expression value.

When a lambda form is used as a method handle behavior, the first parameter (a0) is always the method handle itself. (But there are other uses for lambda forms.)

When a method handle is invoked, after any initial type checking, the JVM executes the lambda form of the method handle to complete the method handle invocation. This leads to some bootstrapping challenges, since the lambda form executes by evaluating additional method handle invocations.

Lambda forms are described in detail elsewhere:http://wiki.jvmlangsummit.com/Lambda_Forms:_IR_for_Method_Handles

Lambda forms will be introduced by example as various behaviors are described.

Lambda form optimization

There is one more indirection in lambda form execution which allows the system to optimize itself: A lambda form has a field called vmentry which (at long last) provides a Method* pointer for the JVM to jump into, in order to evaluate the lambda form.

(Note: Since Java cannot directly represent JVM metadata pointers, this vmentry is actually of type MemberName, which is a low-level wrapper for a Method*. So there is one more indirection after all, to hide the metadata.)

When a lambda form is first created, this vmentry pointer is initialized to a method called the lambda form interpreter, which can execute any lambda form. (Actually it has a thin wrapper which is specialized to the arity and basic types of the arguments.) The lambda form interpreter is very simple and slow. After it executes a given lambda form a few dozen times, the interpreter fetches or generates bytecode for the lambda form, which is customized (at least partially) to the lambda form body. In the steady state, all "hot" method handles and their "hot" lambda forms have bytecode generated, and eventually JIT-compiled.

Thus, in the steady state, a hot method handle is executed without the lambda form interpreter. The low-level JVM steps are as follows:

• Fetch MethodHandle.form.
• Fetch LambdaForm.vmentry.
• Fetch MemberName.vmtarget, a hidden Method* pointer.
• Fetch Method::from_compiled_entry.
• Jump to optimized code.

As noted elsewhere, if the method handle (or if the lambda form or the member name) is a compile-time constant, all the usual inlining can be done.

Invokedynamic

As defined in the JVMS, invokedynamic consists of a name, a method type signature, and bootstrap specifier.

...

Here is an example of an adapter method, taken from an actual application:

LambdaForm(a0:D,a1:L,a2:L)=>{
    t3:L=Invokers.getCallSiteTarget(a2:L);
    t4:L=MethodHandle.invokeBasic(t3:L,a0:D,a1:L);
    t4:L}

...

The body of the lambda form extracts a method handle target from the appendix using the subroutine Invokers.getCallSiteTarget. The method handle is bound to t3, and then immediately invoked on the two leading arguments, a0 adnd and a1.

As may be seen by inspecting the Java code, getCallSiteTarget expects to get a non-null CallSite argument. If this were to fail, it would mean that the trusted Java code has a bug in it, since the trusted code is responsible for returning to the JVM a consistent pair of adapter and appendix.

...

Second, it allows loose typing of its arguments and return value, according to the basic type scheme used in the JSR 292 runtime. Under basic typing rules, all reference types are represented by java.lang.Object. If a reference of a narrower type is required somewhere, an explicit checkcast must be issued before the reference is used. Also, in the basic type system, all 32-bit types are erased to simple int. Thus if a byte value is required somewhere, it must be masked down from a full int. Normally, these extra conversions disappear in the optimizer.

Method handle invocation

Internally to HotSpot (in rewriter.cpp) method handle invocations are rewritten to use a special instruction called invokehandle. This instruction in many ways is parallel to invokedynamic. It resolves to an adapter method pointer and an appendix. The appendix (if not null) is pushed after the explicit arguments to invoke or invokeExact.

The resolution is done via a call to trusted Java code, to a method called MethodHandleNatives.linkMethod. As with linkCallSite, the JVM passes all resolved constant pool references to linkMethod, and receives back a coordinated pair of values, a MemberName and an Object. After unpacking, these are plugged into the CPCE as the adapter and appendix.

The same degrees of freedom apply to invokehandle CPCE entries as apply to invokedynamic CPCE entries, and similar optimization opportunities apply.

There is one major difference from invokedynamic: Many invokehandle instructions can share a single CPCE entry, if they all have the same signature and method name ("invokeExact" vs. "invoke").

Adapter method for invokeExact

The standard semantics of an invokevirtual of MethodHandle.invokeExact are simple. The signature for the call site (which may contain any mix of any references and primitive types) is resolved (once, at link time) into a MethodType. Every time the method handle is invoked, the method type is checked against the type of the method handle being invoked. (Since the method handle is a computed value, it can of course be different every time, and the type does not necessarily match.) If the two types differ in any way, a WrongMethodTypeException is thrown. Otherwise, the method handle is invoked on the given types, and returns a value of the type expected by the caller.

An adapter method for invokeExact is correspondingly simple. It merely performs a method type check and then calls invokeBasic. The appendix is a reference to the resolved MethodType, required to make the type check.

Here is an example:

LambdaForm(a0:L,a1:L,a2:L)=>{
    t3:V=Invokers.checkExactType(a0:L,a2:L);
    t4:L=MethodHandle.invokeBasic(a0:L,a1:L);
    t4:L}

The leading argument a0 is a method handle. The trailing argument a2 is the method type appendix, computed when the call site is resolved. The middle argument a1 is the sole argument to the method handle.

First the subroutine Invokers.checkExactType is called on the method handle and the appendix, extracting the type from the method handle and comparing it with the static call site type. If no exception is thrown, control returns to the adapter method. No value is returned, and the name t3 has a pseudo-type of void.

(Since the appendix represents the static call site type, and the type of the method handle is the dynamically acceptable type, this could be viewed as a simple dynamic type check.)

Next, invokeBasic is used to jump into the method handle (which is now known to be completely safe for this call). The result comes back, briefly named t4, and is returned to the caller.

Adapter method for generic invoke

Method handles also support a more complex invocation mode, which can perform type conversions on individual arguments and return values, and even group arguments into varargs arrays.

As with invokeExact the resolution of such a call site is carried out by a call to MethodHandleNatives.linkMethod. In this case, the trusted Java code must return a more flexible and complex adapter method.

Here is an example:

LambdaForm(a0:L,a1:L,a2:I,a3:L)=>{
    t4:L=Invokers.checkGenericType(a0:L,a3:L);
    t5:I=MethodHandle.invokeBasic(t4:L,a3:L,a0:L,a1:L,a2:I);
    t5:I}

As before, a0 is the method handle and the trailing a3 is a method type. Here, there are two regular arguments, a reference and an int.

As before, the first job is to check the type, and this is done by Invokers.checkGenericType. Unlike the simpler check, this routine returns a value. (The routine can also throw WrongMethodTypeException if necessary.)

The value returned from checkGenericType is in fact a method handle t4. This method handle is immediately invoked on the original arguments, plus the original method handle a0, plus the desired call site type a3 (not in that order).

(See above.)

Example execution sequence for invokedynamic

The target of the invokedynamic instruction's call site can be any method handle. In the simplest case it could be a direct method handle connecting the method containing the invokedynamic instruction to some other Java language method.

Here is an example of the sequence of events and stack frames that would make such a connection, from a method IndyUser.m1 to a target method LibraryCls.m2:

There are two internal stack frames, one for the adapter bound to the invokedynamic call site, and one which handles invocations for the target method handle.

The special methods internalMemberName and linkToStatic are explained on the page about direct method handles.

More low-level details are explained on the page about invocation of method handlesDepending on the match or mismatch between the type of a0 and the call site type a3, the actual behavior of t4 could be very simple (basically another invokeBasic) or very complex (an arity change with type conversions. That's up to the runtime.