Patent Application
20060212862
The invention relates generally to computer systems, and more specifically to a program tool to optimize program code for execution.
Computer programs often include conditional evaluations based on the value of a variable, such as “If X=A, then go to step 10000; otherwise go to step 10010. The condition can be true or false depending on the current value of the variable “X”. For example, a variable “debug” is set to “true” when a program is in a debugging stage. There is a computer program which conditionally evaluates the variable “debug” to determine whether certain data, useful for debugging purposes, should be logged. In other words, the state of the debug variable is used in a decision step to determine whether to log the data. The decision step can be: “If debug=true, then write XYZ data to log; otherwise jump to next step”. Thus, if the debug variable is set to “true”, then the specified data is written to the log. However, if the program is not in the debug stage, for example, the program has already complete debugging, then there is no need to write the specified data to the log. So, after the debugging stage, the program developer uses a known program tool to set the variable “debug” to “false” so the XYZ data will not be written to the log. Also, the program developer will use this tool to declare that the “debug” variable is now fixed (or immutable) as “false”. In Sun Microsystems Java (™) programming language and other runtime languages, this declaration will be entered by the programmer as a command written into the source code. The command means that once the variable is set for the first time to the desired fixed state during execution, it will never be changed. During compilation, the compiler records this declaration, monitors this variable, and returns an error code if the value of this variable is attempted to be changed from its initial value.
U.S. Pat. No. 6,728,952 discloses a computer system for identifying predicates that always are true or always are false (i.e. “vacuous”), and states that if the expression on which an IF statement depends is identified as a vacuous predicate, the code for the IF statement may be optimized. This is because the expression on which the IF statement depends will always be true or will always be false. This patent also states that specialized computer systems already exist which are able to determine whether a predicate is vacuous.
A publication entitled “Fast, Effective Dynamic Compilation”, by Auslander et al., SIGPLAN Notices, vol. 31, no. 5, pages 149-159, May 1996, discloses various run time optimization techniques. For example, run-time constants can become instruction immediates rather than memory loads, constant propagation and folding can be applied to them, conditional branches based on them can be eliminated, and loops they control can be fully unrolled.
A publication entitled “Value Profiling” by Calder et al., Proceedings of 13th Annual IEEE/ACM International Symposium on Micro architecture, pages 259-269. 1997 discloses that identification of variables as invariant or constant at compile-time allows the compiler to perform optimizations including constant folding, code specialization and partial evaluation.
An object of the present invention is to further optimize execution of a program.
The invention resides in a system, process and program product for optimizing a computer program during run time. During run time, a determination is made whether the computer program calls a method which includes a conditional evaluation of a variable which is fixed and the evaluation of the fixed variable does not result in anything productive being accomplished except to return to execution of the computer program. If so, the call to the method which includes the conditional evaluation is deleted from the computer program for subsequent iterations of the computer program. Consequently, the subsequent execution of the computer program will yield a same result as if the conditional evaluation was executed. If the evaluation of the fixed variable results in something productive being accomplished, the call to the method which includes the conditional evaluation is not deleted from the subsequent execution of the computer program.
According to features of the present invention, the call to the first said method is contained in a second method within the computer program, and the evaluation of the fixed variable results in a direct return to the second method. The conditional evaluation of the fixed variable does not result in a call to another method outside of the computer program, or any computations or other evaluations needed by the computer program. During a first iteration of the computer program, the value of the variable is set to the fixed value and an argument associated with the variable is computed.
The present invention will now be described in detail with reference to the figures.
While the operation performed by the method which is called by program 32 or any method called by this method is not important to the present invention, the following is an example. In this example, program 32 defines a class C containing two functional methods. The following is the pseudo code for class C:
The foregoing definition of Class C indicates in the first three lines of the body that variable “X” is fixed or Post Construction Immutable (“PCI”). The foregoing definition of Class C indicates in the middle six lines of the body that there are no parameters for the constructor for Class C, and the value of the variable “X” is “true”. The foregoing definition of Class indicates in the last four lines of the body that Method A and Method B are contained within Class C.
In this example, the following is the pseudo code for Method A:
The foregoing definition of Method A indicates in the method header section that the method returns no value and accepts no arguments. The foregoing definition of Method A indicates in the method body section that the Method A contains a call to invoke Method B. The foregoing definition of Method A indicates in the method body section that the call to invoke Method B requires an argument formed by joining character strings (defining a user of program 32), i.e. joining String1+String2+String3. Before calling Method B, Method A joins the foregoing character strings prior to invoking Method B.
In this example, the following is the pseudo code for Method B:
The foregoing definition of Method B indicates in the method header section that Method B returns no value and accepts one argument of type String labeled “String1”. The foregoing definition of Method B indicates in the method body section that Method B performs a conditional evaluation of variable “X”, i.e. “If X is not equal to True, then perform operation involving String1. If X is equal to True then immediately/directly return to the invoker” (without anything productive happening, for example, without invoking another method, without performing a useful computation, without expression evaluation and without statement evaluation).
During execution of program 32 the run-time interpreter XX produces a new instance of Class C. During creation of this instance of Class C, the run-time interpreter XX records that the variable “X” is fixed (as noted by the compiler), and supplies program 32 with the addressing information for Method A. When program 32 is subsequently invoked and executes, it invokes Method A which begins to execute in step 112. In response, Method A initially computes the argument for calling Method B, by joining String1+String2+String3 in step 116. After computing the argument, Method A invokes Method B in step 120. While Method A and Method B are being executed, program 30 monitors the instruction statements of Method A and Method B in step 122. During this monitoring, program 30 notices that in Method B, a variable “X” is fixed and is being compared to another fixed variable or a fixed value, and the result is an immediate/direct return to the invoker, Method A, with nothing productive occurring (decision 130, yes branch). Consequently, program 30 optimizes program 32 by removing from Method A in program 32 the call to Method B and the associated instructions for computing the argument for the call to Method B. After program 30 optimizes Method A, the following steps enclosed in brackets [ ] are removed from Method A in step 134:
Thus, during the next iteration of program 32 the foregoing steps enclosed in brackets [ ] are not executed when program 32 calls Method A. This reduces the processing time of program 32.
Based on the foregoing, a system, method and program product to optimize a computer program during run time have been disclosed. However, numerous modifications and substitutions can be made without deviating from the scope of the present invention. Therefore, the present invention has been disclosed by way of illustration and not limitation, and reference should be made to the following claims to determine the scope of the present invention.
