Program optimizing apparatus, program optimizing method, and program optimizing article of manufacture

• US 8,990,786 B2
• Filed: 11/30/2011
• Issued: 03/24/2015
• Est. Priority Date: 12/15/2010
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Nakaike, Takuya
• Primary Examiner(s): Chen, Qing
• Assistant Examiner(s): Lyons, Andrew M
• Application Number:
  US13/307,731
• Publication Number:
  US 20120159461A1
• Time in Patent Office: 1,210 Days
• Field of Search: 717/139, 717/140, 717/151, 717/159
• US Class Current: 717/139
• CPC Class Codes: G06F 8/443 Optimisation
• CPC Class Codes: G06F 8/4441 Reducing the execution time required by the program code
• CPC Class Codes: G06F 8/4443 Inlining
• CPC Class Codes: G06F 8/447 Target code generation

Abstract

An apparatus having a transactional memory enabling exclusive control to execute a transaction. The apparatus includes: a first code generating unit configured to interpret a program, and generate first code in which a begin instruction to begin a transaction and an end instruction to commit the transaction are inserted before and after an instruction sequence including multiple instructions to execute designated processing in the program; a second code generating unit configured to generate second code at a predetermined timing by using the multiple instructions according to the designated processing; and a code write unit configured to overwrite the instruction sequence of the first code with the second code or to write the second code to a part of the first code in the transaction.

Claims

1. A program optimizing apparatus including a non-transitory transactional memory enabling exclusive control to execute a transaction, comprising:

a first code generating unit configured to interpret a program, and to generate a first code in which a begin instruction to begin the transaction and an end instruction to commit the transaction are inserted before and after an instruction sequence including a plurality of instructions to execute a designated processing in the program;

a second code generating unit configured to generate a second code at a predetermined timing by using the plurality of instructions according to the designated processing; and

a code write unit configured to overwrite the instruction sequence of the first code with the second code or to write the second code to a part of the first code in the transaction,

wherein the designated processing is a function call processing, and the plurality of instructions include a set instruction to set an argument, a call instruction to call a function, and a return instruction to read a return value,

wherein the predetermined timing is a timing of an interpretation of a called function during execution of the program,

wherein in response to the interpretation of the called function, the second code generating unit generates the second code by replacing the call instruction in the instruction sequence with an instruction to read a value in the called function, by changing a parameter in the instruction to read the value, which is placed by the replacement, and then by deleting an unnecessary instruction group including the set instruction and the return instruction in the instruction sequence.

2. The program optimizing apparatus according to claim 1, further comprising a size judgment unit configured to judge whether a code size of the second code is not greater than a code size of the instruction sequence of the first code, wherein if the size judgment unit judges that the code size of the second code is not greater than the code size of the instruction sequence of the first code, the code write unit overwrites the instruction sequence of the first code with the second code.

3. A program optimization method executed by a program optimizing apparatus including a non-transitory transactional memory enabling exclusive control to execute a transaction, the method comprising the steps of:

interpreting a program and generating first code in which a begin instruction to begin the transaction and an end instruction to commit the transaction are inserted before and after an instruction sequence including a plurality of instructions to execute designated processing in the program;

generating second code at a predetermined timing by using the plurality of instructions according to the designated processing; and

overwriting the instruction sequence of the first code with the second code or writing the second code to a part of the first code in the transaction,

wherein the designated processing is function call processing, and the plurality of instructions include a set instruction to set an argument, a call instruction to call a function, and a return instruction to read a return value,

wherein the predetermined timing is a timing of interpreting a function called during execution of the program,

wherein in the step of generating the second code, the second code is generated in response to interpreting the function called, the second code being optimized by replacing the call instruction in the instruction sequence with an instruction to read a value in the called function, by changing a parameter in the instruction to read the value, which is placed by the replacement, and then by deleting an unnecessary instruction group including the set instruction and the return instruction in the instruction sequence.

4. The program optimization method according to claim 3, further comprising:

judging whether a code size of the second code is not greater than a code size of the instruction sequence of the first code, wherein

in the step of writing the second code, if the code size of the second code is judged to be not greater than the code size of the instruction sequence of the first code, the instruction sequence of the first code is overwritten with the second code.

5. A computer readable non-transitory article of manufacture tangibly embodying computer readable instructions which, when implemented, cause a computer to carry out the steps of a method of claim 3.

6. The article of manufacture according to claim 5, tangibly embodying instructions which, when implemented, cause a computer to carry out the steps of:

judging whether a code size of the second code is not greater than a code size of the instruction sequence of the first code, wherein

if the code size of the second code is judged to be not greater than the code size of the instruction sequence of the first code, overwriting the instruction sequence of the first code with the second code.

7. A program optimizing apparatus including a non-transitory transactional memory enabling exclusive control to execute a transaction, comprising:

a first code generating unit configured to interpret a program, and to generate a first code in which a begin instruction to begin the transaction and an end instruction to commit the transaction are inserted before and after an instruction sequence including a plurality of instructions to execute a designated processing in the program;

a second code generating unit configured to generate a second code at a predetermined timing by using the plurality of instructions according to the designated processing; and

a code write unit configured to overwrite the instruction sequence of the first code with the second code or to write the second code to a part of the first code in the transaction,

wherein the designated processing is processing involving a branch, and the plurality of instructions include at least one branch instruction and a branch destination instruction to be executed at a branch destination,

a frequency judgment unit configured to judge, when the program is interpreted, whether an execution frequency of the branch destination is lower than a threshold, by using execution frequencies obtained from a counted number of positive results and a counted number of negative results of the branch instruction, responding to, as the predetermined timing, a judgment made by the judgment unit that the execution frequency of the branch destination is lower than the threshold, wherein responding includes the second code generating unit generating a trace for forming the second code wherein the trace is a sequence of a control flow up to the branch instruction immediately before the branch destination where the execution frequency is judged to be lower than the threshold.

8. The program optimizing apparatus according to claim 7, further comprising:

an abort judgment unit configured to judge whether a number of aborts caused by a branch destination instruction exceeds a set value during execution of the program; and

a trace search unit configured to search a plurality of traces generated by the second code generating unit to find a null trace wherein the null trace does not include the branch destination instruction having the number of aborts exceeding the set value,

wherein, if the abort judgment unit judges that the number of aborts caused by the branch destination instruction exceeds the set value, the code write unit writes the second code, including the null trace, to a part of the first code in the transaction.

9. A program optimization method executed by a program optimizing apparatus including a non-transitory memory enabling exclusive control to execute a transaction, the method comprising the steps of:

interpreting a program and generating first code in which a begin instruction to begin the transaction and an end instruction to commit the transaction are inserted before and after an instruction sequence including a plurality of instructions to execute designated processing in the program;

generating second code at a predetermined timing by using the plurality of instructions according to the designated processing, wherein the designated processing is processing involving a branch, and the plurality of instructions include at least one branch instruction and a branch destination instruction to be executed at a branch destination;

overwriting the instruction sequence of the first code with the second code or writing the second code to a part of the first code in the transaction;

judging, when the program is interpreted, whether an execution frequency of the branch destination is lower than a threshold, by using execution frequencies obtained from a counted number of positive results and a counted number of negative results of the branch instruction; and

responding to a judgment made by the judgment unit that the execution frequency of the branch destination is lower than the threshold, by generating a trace for forming the second code wherein the trace is a sequence of a control flow up to the branch instruction immediately before the branch destination where the execution frequency is judged to be lower than the threshold.

10. The program optimization method according to claim 9, further comprising the steps of:

judging whether a number of aborts caused by a branch destination instruction is larger than a set value during execution of the program; and

if the number of the aborts caused by the branch destination instruction is judged to be larger than the set value, searching a plurality of traces generated by the second code generating unit to find a null trace wherein the null trace does not include the branch destination instruction having the number of aborts larger than the set value, wherein

in the step of writing the second code, the second code including the null trace is written to a part of the first code in the transaction.

11. A computer readable non-transitory article of manufacture tangibly embodying computer readable instructions which, when implemented, cause a computer to carry out the steps of a method of claim 9.

12. The article of manufacture according to claim 11, tangibly embodying instructions which, when implemented, cause a computer to carry out the steps of:

judging whether a number of aborts caused by a branch destination instruction is larger than a set value during execution of the program; and

if the number of the aborts caused by the branch destination instruction is judged to be larger than the set value, searching a plurality of traces generated by the second code generating unit to find a null trace wherein the null trace does not include the branch destination instruction having the number of aborts larger than the set value, and

writing the second code including the null trace to a part of the first code in the transaction.