WO2000010090A1

WO2000010090A1 - Method, apparatus, and article of manufacture for facilitating resource management for applications having two types of program code

Info

Publication number: WO2000010090A1
Application number: PCT/US1999/018321
Authority: WO
Inventors: Ole Agesen; David L. Detlefs; Derek R. White
Original assignee: Sun Microsystems, Inc.
Priority date: 1998-08-17
Filing date: 1999-08-12
Publication date: 2000-02-24
Also published as: US6253215B1; EP1105804B1; AU5557499A; EP1105804A1

Abstract

Methods, systems, and articles of manufacture consistent with the present invention provide a program component including a set of instructions native to the system, include in the set of native instructions an intruction to maintain information on use of a particular object, and permit reuse of memory resources corresponding to the particular object based on an indication from a source that the particular object is no longer being used, the source being different from any source used to provide information on use of objects associated with non-native instructions of the program component. Additionally, garbage collection is not permitted during native code operations to read or write data in object fields because during such operations an indication exists that such collection may be inaccurate and could possibly reclaim or relocate objects referenced by native code though not specified as such in the native code stack and global variables.

Description

METHOD, APPARATUS, AND ARTICLE OF MANUFACTURE FOR

FACILITATING RESOURCE MANAGEMENT FOR APPLICATIONS

HAVING TWO TYPES OF PROGRAM CODE

BACKGROUND OF THE INVENTION

A. Field of the Invention

This invention generally relates to memory management for computer

systems and, more particularly, to a methodology for managing memory resources

for an application program having two types of program code, native code

executing directly in an operating environment and target code for execution by an

abstract computing machine associated with the operating environment and

responsible for memory management for both types of code.

B. Description of the Related Art

Object-oriented programming techniques have revolutionized the computer

industry. For example, such techniques offer new methods for designing and

implementing computer programs using an application programming interface

(API) associated with a predefined set of "classes," each of which provides a

template for the creation of "objects" sharing certain attributes determined by the

class. These attributes typically include a set of data fields and a set of methods

for manipulating the obj ect.

The Java™ Development Kit (JDK) from Sun Microsystems, Inc., for

example, enables developers to write object-oriented programs using an API with

classes defined using the Java™ programming language. The Java programming

language is described, for example, in a text entitled "The Java Language

Specification" by James Gosling, Bill Joy, and Guy Steele, Addison-Wesley, 1996.

The class library associated with the Java API defines a hierarchy of classes with a child class (i.e., subclass) inheriting attributes (i.e., fields and methods) of

its parent class. Instead of having to write all aspects of a program from scratch,

programmers can simply include selected classes from the API in their programs

and extend the functionality offered by such classes as required to suit the

particular needs of a program. This effectively reduces the amount of effort

generally required for software development.

The JDK also includes a compiler and a runtime environment with a

virtual machine (VM) for executing programs. In general, software developers

write programs in a programming language (in this case the Java programming

language) that use classes from the API. Using the compiler, developers compile

their programs into "class files" containing instructions for an abstract computing

model embodied by the Java VM; these instruction are often called "bytecodes."

The runtime environment has a class loader that integrates the class files of the

application with selected API classes into an executable application. The Java

VM then executes the application by simulating (or "interpreting") bytecodes on

the host operating system/computer hardware. The Java VM thus acts like an

abstract computing machine, receiving instructions from programs in the form of

bytecodes and interpreting these bytecodes. (Another mode of execution is "just

in time" compilation in which the VM dynamically compiles bytecodes into so-

called native code for faster execution.) Details on the VM for the JDK can be

found in a text entitled "The Java Virtual Machine Specification," by Tim

Lindholm and Frank Yellin, Addison Wesley, 1996.

The Java VM also supports multi-threaded program execution. Multi¬

threading is the partitioning of a computer program or application into logically independent "threads" of execution that can execute in parallel. Each thread

includes a sequence of instructions to carry out a particular program task, such as

a method for computing a value or for performing an input/output function. When

employing a computer system with multiple processors, separate threads may

execute concurrently on each processor.

Thus, object-oriented facilities like the JDK assist both development and

execution of object-oriented systems. First, they enable developers to create

programs in an object-oriented programming language using an API. Second,

they enable developers to compile their programs, and third, they facilitate

program execution by providing a virtual machine implementation.

However, object-oriented programs may not be suitable for all functions of

a system or it may not be economically feasible to convert all of the programs in

an existing legacy system into object-oriented programs. It may also be

necessary, for a system having primarily object-oriented programs, to use features

of a platform's operating system that are not available in implementations using a

VM like the Java VM. Finally, the virtual machine implementation itself is

generally not written in the language it executes but rather in the native code of

the host machine. Thus, it is not uncommon for systems to have programs with

"native" and "non-native" code.

For purposes of this description, native code includes code written in any

programming language that is then compiled to run on a compatible operating

system/hardware configuration. For example, native code in this context includes

program code written in the C or C++ programming language and compiled by an

appropriate compiler for execution on a particular platform, such as a computer having the Windows 95 operating system running on an Intel Pentium processor.

Native code is distinguishable from the non-native code, which will be referred to

as "target code," because while non-native code is foreign to a platform's

operating system/hardware configuration, its target for purposes of this description

is an abstract computing machine, such as a VM, operating on any compatible

platform configuration. For example, target code for the Java VM is generally

written in the Java programming language. This combination of native and target

code in the same application tends to complicate the management of memory

resources (i.e., the allocation and deallocation of memory) for such systems.

In practice, when an application seeks to refer to an object, the computer

must first allocate or designate memory for the object. Using a "reference" to the

allocated memory, the application can then properly manipulate the object. One

way to implement a reference is by means of a "pointer" or "machine address,"

which uses multiple bits of information, however, other implementations are

possible. Objects can themselves contain primitive data items, such as integers or

floating point numbers, and/or references to other objects. In this manner, a chain

of references can be created, each reference pointing to an object which, in turn,

points to another object. When no chain of references in an application reaches a

given object, the computer can deallocate or reclaim the corresponding memory

for reuse.

Memory reclamation can be handled explicitly by the application program.

This method, however, requires programmers to design programs to account for

all allocated objects and to determine when the objects are available for

reclamation. The alternative is to assign responsibility for memory management to a runtime system responsible for controlling program execution. The Java VM,

one such system responsible for controlling program execution for example,

includes a "garbage collector" to manage available memory resources used during

execution of Java code.

"Garbage collection" is the term used to refer to a class of algorithms used

to carry out memory management, specifically, automatic reclamation. Garbage

collection algorithms generally determine reachability of objects from the

references held in some set of roots. When an object is no longer reachable, the

memory that the object occupies can be reclaimed and reused. There are many

known garbage collection algorithms, including reference counting, mark-sweep,

and generational garbage collection algorithms. These, and other garbage

collection techniques, are described in detail in a book entitled "Garbage

Collection, Algorithms For Automatic Dynamic Memory Management" by

Richard Jones and Raphael Lins, John Wiley & Sons, 1996.

To be effective, garbage collection techniques should be able to, first,

identify references that are directly accessible to the executing program, and,

second, given the reference to an object, identify references contained within that

object, thereby allowing the garbage collector to transitively trace chains of

references. Unfortunately, many of the described techniques for garbage

collection have specific requirements which cause implementation problems,

particularly when a garbage collector is charged with managing memory for a

system having programs written in both native and target code. For example, the

Java VM's garbage collector manages resources for Java code with relative ease; however, it requires additional facilities to manage resources for other native code

and even then the garbage collector has significant limitations.

In most language implementations, including the implementation of the

Java programming language embodied in the JDK, stacks form one component of

the root set. A stack is a region of memory in which stack frames may be

allocated and deallocated. In typical object-oriented systems, each method

executing in a thread of control allocates a stack frame, and uses the slots of that

stack to hold the values of local variables. Some of those variables may contain

references to heap-allocated objects. (The heap is an area of memory designated

for resources associated with objects.) Such objects must be considered reachable

as long as a method is executing. The term stack is used because the stack frames

obey a last-in/first-out allocation discipline within a given thread of control.

There is generally a stack associated with each thread of control, and when a

thread involves both native and target program code, there are often two stacks,

one for each type of code. Another component of the root set includes global

variables used to hold references to objects outside a stack frame, which makes

the objects available to multiple methods.

A garbage collector may be exact or conservative in how it treats different

sources of references, such as stacks. A conservative collector knows only that

some region of memory (e.g. , a slot for a local variable in the stack frame or a

memory location holding a global variable) may contain references, but does not

know whether or not a given value in that region is a reference. If such a collector

encounters a value that is a possible reference value, it must keep the referenced

object alive. Because of the uncertainty in recognizing references, the collector is constrained not to move the object, since that would require updating the

reference, which might actually be an unfortunately-valued integer or floating¬

point number. The main advantage of conservative collection is that it allows

garbage collection to be used with systems not originally designed to support

collection. For example, the collectors described in Bartlett, Joel F., Mostly-

Copying Collection Picks Up Generations and C++, Technical Report TN-12,

DEC Western Research Laboratory, October 1989, and Boehm, Hans Juergen and

Weiser, Mark, Garbage Collection in an Uncooperative Environment. Software-

Practice & Experience, 18(9), p. 807-820, September 1988, use conservative

techniques to support collection for C and C++ programs.

In contrast, a collector is exact in its treatment of a memory region if it can

accurately distinguish references from non-reference values in that region.

Exactness has several advantages over conservatism. A conservative collector

may retain garbage referenced by a non-reference value that an exact collector

would reclaim. Perhaps more importantly, an exact collector is always free to

relocate objects since it is able to identify references exactly. In an exact system,

one in which references and non-references can be distinguished, this enables a

wide range of useful and efficient garbage-collection techniques that cannot easily

be used in a conservative setting. For example, the ability to relocate objects

enables an exact collector to compact used memory during a collection cycle.

However, a drawback of exact systems is that they must provide the information

that makes them exact, i.e., information on whether a given value in memory is a

reference or a primitive value. A VM can do this effectively for its target code

using techniques such as stack maps that distinguish references from primitive values in the target code's stack. However, there is no known implementation that

uses exact garbage collection for programs including both native and target code

and allows the same level of flexibility and convenience in writing natuve code.

Sun Microsystems, Inc. also developed an interface, called the Java™

Native Interface (JNI), for native program code executing within the Java VM.

The JNI is comprised of a library of functions, i.e., an API, and developers of

native code call upon these functions with references to them by name in the

native code. The JNI functions enable the Java VM's garbage collector to obtain

certain information concerning the native code for purposes of garbage collection.

Using JNI functions, for example, the native code can reference objects in a heap

managed by the Java VM's garbage collector. While the interface itself allows an

implementation supporting exact garbage collection, in the most common

implementation exact garbage collection is not possible. This is because

references are maintained in the same stack used to hold references for the Java

code and the Java VM uses an indicator in a special frame of Java code stack to

control garbage collection of the native code objects. This implementation is

satisfactory for conservative garbage collection but it does not prevent the

"leaking" of direct object references outside the JNI stack frame. In other words,

direct references to objects may be lost during a garbage collection cycle when all

of the references may not be located in the JNI stack frame. Consequently, such

an implementation of the JNI does not support an exact collection algorithm.

There is, therefore, a need for a mechanism that facilitates flexible garbage

collection for memory resources for an application having two types of program

code, native code familiar to an operating environment and target code for execution by an abstract computing machine associated with the operating

environment.

SUMMARY OF THE INVENTION

Methods, systems, and articles of manufacture consistent with the present

invention, as embodied and broadly described herein, manage memory resources

corresponding to objects in a system, by providing a program component

including a set of instructions native to the system. These instructions include an

instruction to maintain information on use of a particular object, and permit reuse

of memory resources corresponding to the particular object based on an indication

from a source that the particular object is no longer being used, the source being

different from any source used to provide information on use of objects associated

with non-native instructions of the program component. The system includes a

runtime environment for executing the program component and a garbage

collector of the runtime environment is invoked to permit reuse of memory

resources. The garbage collector may implement an exact garbage collection

algorithm. The source may be a stack or linked list associated with the native

instructions.

In another implementation, memory resources corresponding to objects in

a system are managed by providing a program component including a set of

instructions native to the system. These instructions include an instruction to

synchronize a garbage collector with the set of native instructions, and prevent, in

response to the synchronize instruction, the garbage collector from permitting

reuse of memory resources corresponding to a particular object until after

operation of certain native instructions. The instruction to synchronize a garbage collector with the set of native instructions may include setting an inconsistency

bit in a data structure associated with the program component.

In yet another implementation, memory resources corresponding to objects

in a system are managed by receiving a program component including instructions

native to the system and instructions targeted to the abstract computing machine,

wherein the instructions include references to objects representing memory

resources, managing object references for the target instructions is a data structure,

and managing object references for the native instructions in a linked list distinct

from the data structure for managing object references for the target instructions.

A garbage collection process is invoked to reclaim memory resources

corresponding to objects based on information from both the data structure for

object references for the target instructions and the linked list for object references

for the native instructions. The garbage collection process reclaims the memory

resources for a particular object when an indication exists that memory resources

must be reclaimed to implement an instruction to allocate another object and it is

determined that the target instructions and the native instructions no longer require

the memory resources for the particular object.

BRIEF DESCRIPTION OF THE DRAWINGS

The accompanying drawings, which are incorporated in and constitute a

part of this specification, illustrate an implementation of the invention and,

together with the description, serve to explain the advantages and principles of the

invention. In the drawings, FIG. 1 is a block diagram of an exemplary system with which methods,

systems, and articles of manufacture consistent with the invention may be

implemented;

FIG. 2 is a block diagram showing data structures for a multi-threaded

application consistent with the present invention;

FIG. 3 is a block diagram showing a target stack and a native stack in

accordance with the principles of the invention;

FIG. 4 is a block diagram showing a stack map for target code;

FIG. 5 is a block diagram showing a linked list of local roots created in

accordance with the principles of the present invention;

FIG. 6 is a flowchart of the procedure for creating the linked list of FIG. 5

in a manner consistent with the principles of the present invention;

FIG. 7 is a flowchart illustrating a first method for synchronizing

inconsistent threads with garbage collection in a manner consistent with the

principles of the present invention; and

FIG. 8 is a flowchart showing the processing performed by a garbage

collector in a manner consistent with the principles of the present invention.

DETAILED DESCRIPTION

Reference will now be made in detail to an implementation consistent with

the present invention as illustrated in the accompanying drawings. Wherever

possible, the same reference numbers will be used throughout the drawings and

the following description to refer to the same or like parts.

Introduction Methods, systems, and articles of manufacture consistent with the present

invention facilitate a flexible approach for garbage collection associated with the

execution of systems having both native and target code by tracking objects

referenced by each type of code in separate stacks. A stack obeys the Last-In-

First-Out (LIFO) model and holds local variables, including references to objects

in the heap. In contrast, "static" or global variables, which may also include

references to objects in the heap, are managed outside the stack. The target code

objects are identified using a map of object references in the stack for the target

code, whereas objects referenced by native code are identified in the native stack

using a linked list.

Additionally, garbage collection is not permitted during native code

operations to read or write data in object fields because, during such operations, an

indication exists that such collection may be inaccurate and could possibly fail to

find and update object references in native code but not specified as such in the

native code stack and global variables.

System Architecture

Figure 1 depicts an exemplary data processing system 100 suitable for

practicing methods and implementing systems and articles of manufacture

consistent with the present invention. Data processing system 100 includes a

computer system 110 connected to a network 170, such as a Local Area Network,

Wide Area Network, or the Internet.

Computer system 110 contains a main memory 120, a secondary storage

device 130, a central processing unit (CPU) 140, an input device 150, and a video

display 160, each of which are electronically coupled to the other parts. Main memory 120 contains an operating system 128, a virtual machine (VM) 122, and a

multi-threaded program 124. An exemplary VM 122 for purposes of this

description is the Java VM described above. In such exemplary implementations,

the Java VM is part of a runtime system, which also includes an API and other

facilities required for running applications using the Java VM.

One skilled in the art will appreciate that although one implementation

consistent with the present invention is described as being practiced using the Java

VM, systems and methods consistent with the present invention may also be

practiced in different environments, including those compatible with the Java VM.

Also, although aspects of one implementation are depicted as being stored in

memory 120, one skilled in the art will appreciate that all or part of systems and

methods consistent with the present invention may be stored on or read from other

computer-readable media, such as secondary storage devices, like hard disks,

floppy disks, and CD-ROM; a carrier wave received from a network such as the

Internet; or other forms of ROM or RAM. Finally, although specific components

of data processing system 100 have been described, one skilled in the art will

appreciate that a data processing system suitable for use with the exemplary

embodiment may contain additional or different components.

VM 122 includes a garbage collector 122a. In one implementation,

garbage collector 122a implements an exact garbage collection algorithm,

although other algorithms may be implemented without departing from the

principles consistent with the present invention.

For purposes of simplifying the illustration program 124 is shown with

more than one thread of execution T,, T₂ and T_n, although those skilled in the art will understand that each thread represents a process consisting of a set of

program instructions executing in CPU 140. A single thread such as T,, may

execute portions of native code 124a and target code 124b. Consequently, VM

122 and garbage collector 122a must manage memory resources for both types of

code operating in the same thread.

API 126 includes native interface 126a and class libraries 126b. Class

libraries 126b includes a set of classes, and developers can select classes for target

code 124b. Native interface 126a includes a set of functions, and developers can

include in native code 124a calls to the native interface 126a to effect various

functions, including memory resource management in a manner consistent with

principles of the present invention.

In general, native interface 126a includes instructions to perform two types

of functions. The first concerns managing a native stack associated with a thread

of control including certain portions of native code 124a. These functions build

and maintain a linked list structure within the native stack to identify all stack

entries containing references to objects. Garbage collector 122a traverses the list

to identify all such referenced objects and, if an object in the heap is not

referenced in the native stack or elsewhere, for example, in a global variable or a

stack for the target code, then garbage collector 122a reclaims the object's

resources. The second set of functions in native interface 126a enable native code

developers to specify periods of time during execution of native code 124a for

which garbage collection is not permitted because, for example, execution of a

garbage collection cycle may destroy objects for which references may exist

outside of the known reference sources, including the native stack, global variables, and target code stack, or move such objects, updating only the known

but not the unknown references. If garbage collection were permitted during such

periods, there is a risk of loss of objects having valid references and, potentially,

later program execution errors.

Appendix A contains a document entitled the "LLNI User's Guide," which

details how to use an exemplary interface consistent with the principles of the

present invention with an implementation of the Java VM having a collector that

uses an exact garbage collection algorithm.

Stack Structures

VM 122 is responsible to executing program 124 using CPU 140 in

conjunction with an operating system 128. In alternative configurations, all or

some of the functions of VM 122 may be incorporated in the operating system or

CPU 140. To facilitate program execution in a multi-threaded fashion, VM 122

maintains a thread element for each thread. As shown in Fig. 2, a thread element

246 includes two stacks 250 and 252 for managing execution of each type of code.

For example, target stack 250 holds references to objects in memory used by

target code 124b and native stack 252 holds references to objects in memory used

by native code 124a. Those skilled in the art will recognize that implementations

consistent with the principles of the present invention may involve intermingling

the two stacks in a memory structure or use a single stack with appropriate

designations to distinguish between portions of the stack used for target code

versus those portions used for native code.

Thread data elements are typically linked together by thread data

structures, such as thread data structure 248. For example, thread 1 data structure 248 identifies or points to a location for the data structure for a next thread. By

linking all thread data elements together, VM 122 may step from thread to thread,

and access the data elements of each thread.

VM 122 also maintains global roots 260 and native global roots 262

separate from the thread elements. Global roots 260 and native global roots 262

contain variables accessible by all threads. These variables may include

references to stored objects.

Figure 3 is a block diagram showing a portion of thread data element 246

of Fig. 2 in greater detail. Thread data element 246 comprises thread data

structure 248, target stack 250 and native stack 252. Thread data structure 248

includes fields of information for managing the data elements of the thread. For

example, thread data structure 248 stores inconsistency bit 310, target code stack

pointer 312, native code stack pointer 314, and next thread pointer 315. Thread

data structure 248 may also store other information about the thread.

Inconsistency bit 310 is set whenever a thread is starting to enter a region of

program code that may result in pointers being used in a manner inconsistent with

the implemented garbage collection algorithm.

Target code stack pointer 312 points to target stack 250. Target stack 250

stores information used during execution of target code of a particular thread.

Similarly, native code stack pointer 314 points to native stack 252, which stores

information used during execution of native code of a particular thread. Target

stack 250 and native stack 252 have similar structures. Target stack 250

comprises frames 316, 318 and 320. Native stack 252 comprises frames 322 and 324. Finally, next thread pointer 315 points to the thread data structure of the next

thread, as explained with reference to Fig. 2.

In one implementation, each stack frame 316, 318 and 320 for target code

includes an area for holding local variables, an invoking frame pointer, a method

pointer, a program counter (PC), and an operand stack, as shown in Fig. 3. When

a target or native method starts executing, a stack frame for the method is added

on the appropriate stack. The stack frame holds variables used during execution

of the method, including references to stored objects.

A target method uses values in either the operand stack or the local

variables of the stack frame from which it is executing. For example, a target

method might add two integers by pushing the integers on the operand stack and

then performing an add bytecode which pops the top two items off the operand

stack, adds them, and pushes the answer back on.

Each target code stack frame 320, 318, and 320 also contains an invoking

frame that points to the previous stack frame on the stack, a method pointer that

points to a method block associated with the method, and a program counter (PC)

that points to the current line of the method being executed.

In contrast, native code stack 252 comprises a series of linked frames 322

and 324, each of which holds local variables used by the native method executing

out of the frame. As in target stack 250, native stack 252 contains local variables

that reference stored objects.

As target or native methods execute, VM 122 manipulates local variables

referencing stored objects, deleting references to stored objects when no longer

needed by the method. Even though an object is no longer needed, it still occupies space in the heap. As more object references are deleted, the space

occupied by unused objects grows until there is no space left in the heap for

allocating new objects. To provide more space for object allocation, garbage

collector 122a periodically determines which objects are being referenced, and

reclaims the remaining space in the heap. The reclaimed space can then be used

for allocating space for more objects. Garbage collector 122a determines which

objects are being referenced by stepping through each thread data element 246

using the thread data structure 248 of each element.

An object is usually considered referenced if there is some path of pointers

leading to the object from roots located in variables of the program. The roots of a

program at a given point in execution are comprised of the global (i.e. static)

variables of the program together with the local variables of any procedure or

method currently being executed at that execution point. This includes, for

example, global roots 260 and native global roots 262 of Fig. 2, and the local

variables of stack frames 316, 318 and 320 of Fig. 3.

It is generally not difficult to identify the static variables of a program, and

trace objects from those containing pointers because static variables implementing

pointers generally remain pointers throughout execution. It can, however, be

difficult to identify which local variables contain pointers to objects as opposed to

primitive values.

To save stack space, for example, the slots in stack frames are sometimes

reused. Consider a method "m" that has two subparts: a pointer-containing

variable "p" that is used only during the first part, and an integer-containing

variable "i" that is used only during the second part. Since "p" and "i" are never in use at the same time, a single slot "s" in a stack frame for "m" might be used for

both. In such a situation, garbage collector 122a has difficulty determining

whether to consider slot "s" a pointer or a primitive. If it does not consider "s" a

pointer, and "s" actually does contain a pointer, the garbage collector risks

incorrectly recycling the object to which "s" points.

Garbage Collection for Target Code

To address this problem for target code, thread element 246 has a

corresponding stack map, as shown in Fig. 4. To create a stack map for a method,

the method is first scanned to find safe points for garbage collection. Typically,

these garbage collection safe points are times of transition, such as at a call or

backward branch instruction. Once a safe point is found, the stack map defining

all of the pointer locations is generated and associated with that particular

instruction. Therefore, when a safe point is reached during execution, a garbage

collector can determine from the stack map where each pointer is located in the

stack frame at the time the respective instruction is executed. Using this

information, the garbage collector knows exactly where all pointers are located.

Stack maps can be generated at any point before garbage collection. For example,

they can be generated when the program is compiled or during program execution.

Figure 4 is a block diagram illustrating an example of a stack map. In the

stack frame 410 associated with a method of thread n, method pointer 412 points

to method block 414. Method block 414 points to the method 416, which is the

code of the method. Method block 414 also points to stack map data structure

418. Stack map data structure 418 comprises program counter values

corresponding to particular lines of the code in method 416. Each program counter value is associated with a respective map in the set 420. Each map in the

set of stack maps 420 specifies the stack slots or memory registers containing

references to heap-allocated objects. For purposes of illustration, each stack map

is divided into two sections, indicated by a heavy vertical line. Locations to the

left of the heavy line indicate slots (SI, S2, S3, S4) in the stack frame of the

method, and locations to the right indicate registers (Rl, R2).

As described above with reference to Fig. 3, a PC is stored in each frame,

and is used to track the line currently being executed in the method corresponding

to the stack frame. When the PC in stack 410 (not shown) equals line 10, stack

map 422 defines which slots and registers have object pointers at that particular

point in execution of method 416. Stack map 422 indicates that slots SI and S4,

and register Rl, each marked by a "1," have a pointer when execution of method

416 is at program counter value 10. Slots S2 and S3, and register R2, each

marked by an "O," do not contain pointers. Therefore, the garbage collector can

determine precisely where each pointer is located for method 416 by using the set

of stack maps 420.

When garbage collector 122a begins a garbage collection cycle, each

thread is stopped and advanced to a safe point. Once execution reaches the safe

point, garbage collector 122a uses the stack map associated with each method to

determine pointer locations with certainty.

To find the stack map associated with a particular method, garbage

collector 122a first steps through each thread data structure to access the target

stacks, and uses the method pointer in the stack frame to access the corresponding

set of stack maps. Garbage collector 122a then uses the stack map corresponding to the line of code at which the method was stopped to determine the stack frame

locations having pointers referencing objects. Further details on the use of a stack

map in this fashion for garbage collection can be found in O. Agesen, D. Detlefs,

J.E.B. Moss, "Garbage Collection and Local Variable Type-Precision and

Liveness in Java™ Virtual Machines," Proceedings of the ACM SIGPLAN '98

Conference on Programming Language Design and Implementation, ACM, 1998,

pp. 269-279, which is incorporated herein by reference.

Garbage Collection for Native Code

In contrast, garbage collection for code operating out of a native stack in a

manner consistent with the principles of the present invention involves including

in the native code calls to certain functions of native interface 126a. Developers

can modify existing native code to include the function calls; alternatively, they

can write new native code with the function calls. For purposes of this

description, there are two types of function calls. The first concerns creating and

maintaining a linked list in the stack for the native code identifying all of the slots

for local variables containing references to objects in the heap. The second

involves setting inconsistency bit 310. When set, this bit prevents garbage

collector 122a from executing during an "unsafe" period when not all local

variables containing references to objects are identifiable.

Linked List

Figure 5 is a block diagram showing an implementation of a per-thread

linked list of local roots created in accordance with the principles of the present

invention. Stack 514 contains stack frames having slots used by native methods

during execution of the methods. Slots B, D and F point to objects. Slot E points to slot D, and slot C points to slot B. Native stack pointer 512 points to stack 514,

and local variables pointer (LVP) 510 points to the first local variable in the linked

list.

The linked list is formed by grouping each object pointer with a pointer to

the previous object pointer in the stack. Therefore, by using the value of LVP

510, garbage collector 122a can determine exactly the location of the object

pointer in slot F, go to the next location and trace slot E to the object pointer in

slot D, and go to the next location and trace slot C to the object pointer in slot B.

Slot B is followed by A, which has a null value, indicating the end of the linked

list. In this way, the garbage collector can determine the exact location of each

pointer in the stack.

Figure 6 is a flowchart showing how the linked list of Fig. 5 is created.

For purposes of this description assume LVP 510 is pointing to slot D. Upon

receiving a routine call from native code 124a that requires creation of a new

object pointer, VM 122 uses code in native interface 126a to set the value of slot E

to the value of LVP 510 (step 610). Thus, slot E now points to slot D. VM 122

then initializes the slot for the pointer to the new object in the heap (i.e., slot F) by

setting the slot to a null value (step 612). VM 122 then sets the value of LVP 510

to the address of slot F (step 614), so that LVP 510 continues to point to the

uppermost object pointer in the stack. Finally, VM 122 sets the value of slot F to

the address of the object in heap 524 (step 616).

When the native routine is done with a local root, VM 122 pops it off the

top of the local roots linked list by setting LVP 510 to the address of the next object pointer downward in the stack. In the example shown, LVP 510 will be set

to the value stored in slot E, i.e., the address of slot D.

Any global variables containing object references are also manipulated

using a similar linked list structure.

GC Synchronization

Objects in the heap can be accessed by means of direct and indirect

pointers. A direct pointer is the same as a reference in a local or global variable

and can be used in a program to access an object. In contrast, an indirect pointer

is a pointer to a direct pointer. Consistent regions of program code use only

indirect pointers to reference objects by means of direct pointers. To access an

object, for example, to write a value in a field of the object, the indirect pointer is

"dereferenced," obtaining access to the direct pointer and thus access to the object

itself. Regions of program code during which objects are accessed by using direct

pointers are "inconsistent" regions because the dereferencing of an indirect pointer

may copy a direct pointer value into a location not known by the garbage collector

to contain such a pointer. Thus, garbage collection is not permitted during

inconsistent regions of program code because it is not possible to determine

exactly which slots in the stack frame are pointers to objects in the heap. If the

garbage collector relocates an object (as is often the case with a compacting

garbage collector, for example), the collector may fail to update direct pointers

that were obtained by dereferencing indirect pointers to the new location of the

relocated objects. Thus, to access an object for read or write purposes in a manner

consistent with the present invention, garbage collection must be postponed until

the access operation is completed. This is accomplished by including in the native code a call to a GC synch routine of interface 126a to set inconsistent bit 310 in

thread data structure 248, indicating that the thread is now in an inconsistent

region and garbage collection is not permitted. For efficiency, the call may be "in¬

line," meaning the program code for the called routine is actually included in the

calling program instead of requiring the routine to be loaded from another

location. The GC synch routine synchronizes native code with garbage collector

122a.

Figure 7 is a flowchart illustrating a method for synchronizing inconsistent

threads with garbage collection consistent with the principles of the present

invention. When an inconsistent region of code is entered, the native code

declares the thread inconsistent by calling a routine of interface 126a to set the

inconsistent bit 310 (step 710).

VM 122 then dereferences the indirect pointer received in the routine call

using interface 126a (step 712) to obtain the value pointed to by the indirect

pointer (i.e., the direct pointer). After native code is through using the direct

pointer associated with an indirect pointer, the thread consistency bit is reset (step

718). A global flag is then checked, to determine whether a garbage collection

was requested while in the inconsistent region (step 720). If the global flag

indicates that no garbage collection was requested, the process is exited. If the

global flag indicates that garbage collection has been requested, VM 122 waits

until collection is complete (step 722).

Figure 8 is a flowchart showing the processing performed by garbage

collector 122. For example, when a method requests to allocate an object and the

request cannot be granted because the heap is full, the thread must run a garbage collection to find memory space to satisfy the allocation request. In response to a

garbage collection request, garbage collector 122 first stops all threads currently

being executed (step 810). Garbage collector 122a then determines whether all

threads are consistent by checking the inconsistent bit 310 of each thread data

structure 248 (step 812). If all threads are consistent, garbage collection is

performed (step 820), after which the threads are restarted.

If some threads are not consistent, garbage collector 122 raises a global

flag (step 814) indicating to threads coming out of inconsistent regions that the

thread should synchronize with garbage collector 122 when it comes out of an

inconsistent state. Garbage collector 122 then restarts the inconsistent threads

(step 816) and waits until all threads are consistent (step 818). Upon all threads

becoming consistent, garbage collection is performed (step 820) and the threads

are restarted.

Therefore, threads may temporarily enter an inconsistent region, which

prevents garbage collector 122 from starting a collection cycle.

Conclusion

Methods, systems, and articles of manufacture consistent with the present

invention therefore facilitate a flexible approach for garbage collection associated

with the execution of systems having both native and target code and permit

implementations using either a conservative or exact collection algorithm.

The foregoing description of an implementation of the invention has been

presented for purposes of illustration and description. It is not exhaustive and

does not limit the invention to the precise form disclosed. Modifications and

variations are possible in light of the above teachings or may be acquired from practicing of the invention. For example, the described implementation includes

software but the present invention may be implemented as a combination of

hardware and software or in hardware alone. The invention may be implemented

with both object-oriented and non-object-oriented programming systems. The

scope of the invention is defined by the claims and their equivalents.

Claims

WHAT IS CLAIMED IS:

1. A method for managing memory resources corresponding to objects in a

system, comprising:

executing a program component including a set of instructions native to

the system, the set including an instruction to maintain information on use of a

particular object; and

permitting reuse of memory resources corresponding to the particular

object based on an indication from a source that the particular object is no longer

being used, the source being different from any source used to provide

information on use of objects associated with non-native instructions of the

program component.

2. The method of claim 1 , wherein the system includes a runtime

environment for executing the program component, and wherein permitting reuse

of memory resources includes

invoking a garbage collector of the runtime environment.

3. The method of claim 2, wherein invoking a garbage collector includes

implementing an exact garbage collection algorithm.

4. The method of claim 1 , wherein permitting reuse of memory resources

includes

using a stack associated with the native instructions.

5. The method of claim 1 , wherein permitting reuse of memory resources

includes

using a linked list associated with the native instructions.

6. A method for managing memory resources corresponding to objects in a

system, comprising:

executing a program component including a set of instructions native to

the system, wherein the set of native instructions includes an instruction to

synchronize a garbage collector with the program component; and

preventing, in response to execution of the synchronize instruction, the

garbage collector from permitting reuse of memory resources corresponding to a

particular object until after operation of certain native instructions.

7. The method of claim 6, wherein the instruction to synchronize a garbage

collector with the set of native instructions includes

setting an inconsistency bit in a data structure associated with the program

component.

8. The method of claim 6, further comprising:

signaling the garbage collector to permit reuse of memory resources

corresponding to a particular object after operation of certain native instructions.

9. The method of claim 8, wherein the instruction to synchronize a garbage

collector with the set of native instructions includes

setting an inconsistency bit in a data structure associated with the program

component.

0. The method of claim 8, wherein signaling the garbage collector includes

resetting the inconsistency bit.

11. A method for managing memory resources corresponding to objects in a

system, comprising:

providing a program component including a set of instructions native to

the system, wherein the set of native instructions includes an instruction to

synchronize a garbage collector with the program component; and

signaling the garbage collector to postpone a start of a collection process

until after operation of certain native instructions in response to the synchronize

instruction.

12. A memory for managing memory resources corresponding to objects in a

system, the memory comprising:

a set of instructions native to the system, the set including an

instruction to maintain information on use of a particular object; and

an interface with code for permitting reuse of memory resources

corresponding to the particular object based on an indication from a

source that the particular object is no longer being used, the source

being different from any source used to provide information on use of

objects associated with non-native instructions.

13. A computer-implemented method for managing resources corresponding

to objects in a memory, comprising:

executing a program component including a set of instructions native to

particular object; and

relocating another object in the memory based on an indication from a

source that the particular object is no longer being used, the source being different

from any source used to provide information on use of objects associated with

non-native instructions of the program component.

14. The method of claim 13, further comprising:

updating a data structure referencing the relocated object with its new

location.

15. A method for managing memory resources corresponding to objects in a

system having an abstract computing machine, comprising:

executing a program component including first instructions native to the

system and second instructions targeted to the abstract computing machine,

wherein the first and second instructions include references to objects representing

memory resources;

managing object references for the second instructions in a data structure;

and

managing object references for the first instructions in a linked list distinct

from the data structure for managing object references for the second instructions.

16. The method of claim 15, further comprising:

invoking a garbage collection process to reclaim memory resources

corresponding to objects based on information from both the data structure for

object references for the second instructions and the linked list for object

references for the first instructions.

17. The method of 16, wherein invoking the garbage collection process

includes

reclaiming the memory resources for a particular object when an indication

exists that memory resources must be reclaimed to perform an instruction to

allocate another object and it is determined that the second instructions and the

first instructions no longer require the memory resources for the particular object.

18. A system for managing memory resources corresponding to objects,

comprising:

a memory having a program component including a set of instructions

native to the system, the set including an instruction to maintain information on

use of a particular obj ect; and

a processor configured to permit reuse of memory resources corresponding

to the particular object based on an indication from a source that the particular

object is no longer being used, the source being different from any source used to

provide information on use of objects associated with non-native instructions of

the program component.

19. The system of claim 18, wherein the processor is further configured to

invoke a garbage collector.

20. The system of claim 19, wherein the garbage collector implements an

exact garbage collection algorithm.

21. The system of claim 18, wherein the source is a stack associated with the

native instructions.

22. The system of claim 18, wherein the source is a linked list associated with

the native instructions.

23. A system for managing memory resources corresponding to objects,

comprising:

a memory having a program component including a set of instructions

native to the system, wherein the set of native instructions includes an instruction

to synchronize a garbage collector with the set of native instructions; and

a processor configured to prevent, in response to the synchronize

instruction, the garbage collector from permitting reuse of memory resources

corresponding to a particular object until after operation of certain native

instructions.

24. The system of claim 23, wherein the processor is further configured to set

an inconsistency bit in a data structure associated with the program component.

25. The system of claim 23, wherein the processor is further configured to

signal the garbage collector to permit reuse of memory resources corresponding to

a particular object after operation of certain native instructions.

26. The system of claim 25, wherein the processor is further configured to set

an inconsistency bit in a data structure associated with the program component.

27. The system of claim 25, wherein processor is further configured to reset

the inconsistency bit.

28. A system for managing memory resources corresponding to objects,

comprising:

a memory having a program component including a set of instructions

native to the system, the set of native instructions including an instruction to

synchronize a garbage collector with the set of native instructions; and

a processor configured to signal the garbage collector to postpone a start of

a collection process until after operation of certain native instructions in response

to the synchronize instruction.

29. A system for managing memory resources corresponding to objects,

comprising:

a memory having a program component including a set of instructions

use of a particular object; and

a processor configured to relocate another object based on an indication

with non-native instructions of the program component.

30. The system of claim 29, wherein the processor is further configured to

update a data structure referencing the relocated object with its new location.

31. The system of claim 29, wherein the processor is further configured to

invoke a garbage collector.

32. A system for managing memory resources corresponding to objects and

having an abstract computing machine, comprising:

a memory having a program component including first instructions native

to the system and second instructions targeted to the abstract computing machine,

wherein the first and second instructions use references to objects representing

memory resources; and

a processor configured to manage object references for the second

instructions in a data structure and object references for the first instructions in a

linked list distinct from the data structure for managing object references for the

target instructions.

33. The system of claim 32, wherein the processor is further configured to

invoke a garbage collection process to reclaim memory resources corresponding

to objects based on information from both the data structure for object references

for the second instructions and the linked list for object references for the first

instructions.

34. The system of 33, wherein the processor is further configured to reclaim

the memory resources for a particular object when an indication exists that

memory resources must be reclaimed to implement an instruction to allocate

another object and it is determined that the second instructions and the first

instructions no longer require the memory resources for the particular object.

35. A computer-readable medium containing instructions to perform a method

for controlling a data processing system to manage memory resources

corresponding to objects in the data processing system, the method comprising:

executing a program component including a set of instructions native to

particular object; and

permitting reuse of memory resources corresponding to the particular

being used, the source being different from any source used to provide

information on use of objects associated with non-native instructions of the

program component.

36. The computer-readable medium of claim 35, wherein the data processing

system includes a runtime environment for executing the program component, and

wherein permitting reuse of memory resources includes

invoking a garbage collector of the runtime environment.

37. The computer-readable medium of claim 36, wherein invoking a garbage

collector includes

implementing an exact garbage collection algorithm.

38. The computer-readable medium of claim 35, wherein the source is a stack

associated with the native instructions.

39. The computer-readable medium of claim 35, wherein the source is a linked

list associated with the native instructions.

40. A computer-readable medium containing instructions to perform a method

for controlling a data processing system to manage memory resources

corresponding to objects in the data processing system, the method comprising:

providing a program component including a set of instructions native to

the system; including in the set of native instructions an instruction to synchronize a

garbage collector with the set of native instructions; and

preventing, in response to the synchronize instruction, the garbage

collector from permitting reuse of memory resources corresponding to a particular

obj ect until after operation of certain native instructions.

41. The computer-readable medium of claim 40, wherein the instruction to

synchronize a garbage collector with the set of native instructions includes

setting an inconsistency bit in a data structure associated with the program

component.

42. The computer-readable medium of claim 40, wherein the method further

comprises:

signaling the garbage collector to permit reuse of memory resources

43. The computer-readable medium of claim 42, wherein the instruction to

synchronize a garbage collector with the set of native instructions includes

setting an inconsistency bit in a data structure associated with the program

component.

44. The computer-readable medium of claim 42, wherein signaling the

garbage collector includes

resetting the inconsistency bit.

45. A computer-readable medium containing instructions to perform a method

for controlling a data processing system to manage memory resources

corresponding to objects in the data processing system, the method comprising:

providing a program component including a set of instructions native to

the system;

including in the set of native instructions an instruction to synchronize a

garbage collector with the set of native instructions; and

signaling the garbage collector to postpone a start of a collection process

instruction.

46. A computer-readable medium containing instructions to perform a method

for controlling a data processing system to manage memory resources

corresponding to objects in the data processing system, the method comprising:

executing a program component including a set of instructions native to

particular object; and

relocating another object based on an indication from a source that the

particular object is no longer being used, the source being different from any

source used to provide information on use of objects associated with non-native

instructions of the program component.

47. The computer-readable medium of claim 46, wherein the method further

comprises:

updating a data structure referencing the relocated object with its new

location.

48. A computer-readable medium containing instructions to perform a method

for controlling a data processing system to manage memory resources

corresponding to objects in the data processing system, the method comprising:

. executing a program component including instructions native to the system

and instructions targeted to the abstract computing machine, wherein the

instructions include references to objects representing memory resources;

managing object references for the target instructions in a data structure;

and

managing object references for the native instructions in a linked list

distinct from the data structure for managing object references for the target

instructions.

49. The computer-readable medium of claim 48, wherein the method further

comprises:

invoking a garbage collection process to reclaim memory resources

corresponding to objects based on information from both the data structure for

for the native instructions.

50. The computer-readable medium of 49, wherein invoking the garbage

collection process includes

reclaiming the memory resources for a particular object when an indication

exists that memory resources must be reclaimed to implement an instruction to

allocate another object and it is determined that the target instructions and the

native instructions no longer require the memory resources for the particular

object.

51. A system for managing memory resources corresponding to objects,

comprising:

means for executing a program component including a set of instructions

use of a particular object; and

means for permitting reuse of memory resources corresponding to the

particular object based on an indication from a source that the particular object is

no longer being used, the source being different from any source used to provide

information on use of objects associated with non-native instructions of the

program component.

52. A memory device encoded with a data structure used for managing system

resources for a thread executing in a computer and including a set of instructions

native to the computer and a set of instructions targeted to an abstract computer

machine operating in the computer, the data structure comprising:

a first field with an identifier for a memory area for holding references to

objects used in the native instructions;

a second field with an identifier for a different memory area for holding

references to objects used in the target instructions; and

an inconsistency bit that, when set, prevents a garbage collector from

permitting reuse of memory resources corresponding to a particular object until

after operation of certain native instructions.