Dynamic compilation control

• US 6,993,755 B1
• Filed: 10/24/2001
• Issued: 01/31/2006
• Est. Priority Date: 10/24/2001
• Status: Active Grant

• Current Assignee: Oracle America, Inc. (Oracle Corporation)
• Original Assignee: Sun Microsystems Incorporated (Oracle Corporation)
• Inventors: Ungar, David M.
• Primary Examiner(s): Chaki, Kakali
• Assistant Examiner(s): Vu, Tuan A.
• Application Number:
  US10/044,112
• Time in Patent Office: 1,560 Days
• Field of Search: 717/114, 717/140, 717145-153, 717/160, 712/222
• US Class Current: 717/158
• CPC Class Codes: G06F 9/45516 Runtime code conversion or optimisation

Abstract

Modern programming languages have stimulated work on systems that dynamically compile or optimize frequently executed portions of programs. In practice, such systems typically rely on ad hoc heuristics. For example, a system may optimize (or compile) some code once its execution count exceeds a given threshold. An analytical model has been developed that expresses performance of such a system. In one embodiment, the model is based on a bytecode frequency histogram, which indicates (for a given program) how many bytecodes run for how many times. It predicts that the optimal compilation threshold will occur where the hazard rate falls through the reciprocal of the break-even point, the number of times a compiled bytecode must be executed to recoup its compilation time. Based on the insight provided by the model, a dynamic compilation control technique has been developed.

Claims

1. A method of selecting certain portions of a computer program for compilation, the method comprising:

computing an execution frequency threshold at which a decreasing hazard rate corresponds to a reciprocal of a break-even number of executions that recoup computational costs of compilation; and

during execution of the computer program, dynamically compiling individual ones of the portions based on the execution frequency threshold

wherein the hazard rate, hr(x), for a particular one of the computer program portions at least approximates a probability that the particular portion will stop being executed in the computer program given that the particular program has executed x times.

2. A method as recited in claim 1,

performing the computation of the compilation threshold coincident with execution of the computer program and using runtime information from the execution.

3. A method as recited in claim 1, wherein the break-even number of executions is a function of:

time or other execution resource to execute an uncompiled version of a particular one of the computer program portions;

time or other execution resource to compile the particular portion; and

time or other execution resource to execute a compiled version of the particular portion.

4. A method as recited in claim 1,

wherein the portions correspond to functions, procedures, methods or routines of the computer program.

5. A method as recited in claim 1,

wherein the portions correspond to bytecodes executable in an execution environment.

6. A method as recited in claim 1,

wherein those computer program portions not selected for compilation are interpreted.

7. A method as recited in claim 1,

wherein a particular one of the computer program portions is interpreted for a first subset of executions thereof;

and

wherein, after the dynamic compilation, subsequent executions are of a compiled version of the particular portion.

8. A method as recited in claim 1,

wherein a first subset of executions of a particular one of the computer program portions is of a first compiled version thereof;

wherein, after the dynamic compilation, subsequent executions are of a second compiled version of the particular portion.

9. A method as recited in claim 8,

wherein the second compiled version is substantially optimized as compared with the first compiled version.

10. An computer-implemented execution environment for a computer program including execution codes that may optionally be executed in either a first or a second form thereof, the execution environment comprising:

a dynamic compilation mechanism that transforms an implementation of a particular one of the execution code in a first form to the second form thereof, wherein the second form is substantially optimized as compared to the first form; and

a transformation threshold computation mechanism that computes, at least for the particular execution code, an execution frequency threshold at which a decreasing hazard rate corresponds to a reciprocal of a break-even number of executions that recoup computational costs of transformation to the second form,

wherein the dynamic compilation mechanism is responsive to the computed execution frequency threshold;

wherein the hazard rate, hr(x), for a particular one of the computer program portions at least approximates a probability that the particular portion will stop being executed in the computer program given that the particular program has executed x times.

11. An execution environment as recited in claim 10,

wherein the first and second forms respectively include uncompiled and compiled version of the execution code.

12. An execution environment as recited in claim 10,

wherein the first and second forms both include compiled versions of the execution code, but the second form is substantially optimized as compared to the first form.

13. An execution environment as recited in claim 10,

wherein the execution code is a bytecode.

14. A computer program product encoded in at least one computer readable medium, the computer program product comprising:

first instructions executable on a processor to instrument execution of a computer program executing thereon, the first instructions providing data indicative of execution frequency for at least a particular portion of the computer program; and

second instructions executable to identify a particular execution frequency threshold in the execution of the computer program at which a decreasing hazard rate calculated from the execution frequency data for the particular portion of the computer program corresponds to a reciprocal of a break-even number of executions thereof that recoup computational costs of transformation to an optimized form

wherein the hazard rate, hr(x), for a particular one of the computer program portions at least approximates a probability that the particular portion will stop being executed in the computer program given that the particular program has executed x times,

wherein the execution frequency threshold identified by the second instructions for a particular portion of the computer program indicates an opportunity to dynamically compile the particular portion of the computer program.

15. The computer program product of claim 14,

wherein the particular portion is dynamically compiled to the optimized form coincident with identification of the particular point.

16. The computer program product of claim 14, further comprising:

a dynamic compiler.

17. The computer program product of claim 14, embodied as part of an execution environment for the computer program.

18. The computer program product of claim 14, embodied as part of the computer program.

19. The computer program product of claim 14,

wherein the at least one computer readable medium is selected from the set of a disk, tape or other magnetic, optical, or electronic storage medium and a network, wireline, wireless or other communications medium.

20. An apparatus comprising:

means which is encoded in at least one computer-readable media, for measuring execution frequency for at least a particular execution code and, based thereon, determining in an execution of computer code that includes the particular execution code a execution frequency threshold at which a decreasing hazard rate corresponds to a reciprocal of a break-even number of executions that recoup computational costs of transformation to the optimized form; and

means which is encoded in at least one computer-readable media, for dynamically transforming an implementation of the particular execution code, based on the execution frequency threshold determined by the measuring means, to an optimized form thereof;

wherein the hazard rate, hr(x), for a particular one of the computer program portions at least approximates a probability that the particular portion will stop being executed in the computer program given that the particular program has executed x times.