Avoiding or deferring data copies

• US 10,095,531 B2
• Filed: 08/14/2015
• Issued: 10/09/2018
• Est. Priority Date: 03/14/2013
• Status: Active Grant

• Current Assignee: Amazon Technologies, Inc. (Amazon.com, Inc.)
• Original Assignee: Amazon Technologies, Inc. (Amazon.com, Inc.)
• Inventors: Boynes, Jeremy
• Primary Examiner(s): Chowdhury, Ziaul A
• Application Number:
  US14/827,121
• Publication Number:
  US 20150355921A1
• Time in Patent Office: 1,152 Days
• CPC Class Codes: G06F 11/36 Preventing errors by testing or debugging software
• CPC Class Codes: G06F 12/00 Accessing, addressing or allocating within memory systems or architectures digital input from, or digital output to record carriers, e.g. to disk storage units, G06F3/06
• CPC Class Codes: G06F 16/00 Information retrieval; Database structures therefor; File system structures therefor
• CPC Class Codes: G06F 8/443 Optimisation
• CPC Class Codes: G06F 9/45504 Abstract machines for programme code execution, e.g. Java virtual machine [JVM], interpreters, emulators
• CPC Class Codes: G06F 9/45525 Optimisation or modification within the same instruction set architecture, e.g. HP Dynamo

Abstract

Methods and systems for avoiding or deferring data copies are disclosed. Using a virtual machine, it is determined whether a set of program code comprises references to a data object after an operation to generate a copy of the data object. If not, a set of optimized program code is generated in which the operation to copy the data object is replaced with an operation to update a reference. Using the virtual machine, it is determined whether the set of program code comprises an operation to generate a copy of a buffer object. If so, a set of further optimized program code is generated, comprising an allocation of one or more memory pages to store the buffer object with a copy-on-write parameter instead of the operation to generate the copy of the buffer object.

Claims

1. A method, comprising:

performing by one or more hardware computing devices;

determining that a set of program code does not comprise one or more references to an object after an operation to generate a copy of the object, wherein the set of program code is executable using a virtual machine; and

in response to determining that the set of program code does not comprise the one or more references to the object after the operation to generate the copy of the object, generating a set of optimized program code, executable using a virtual machine, comprising replacing the operation to generate the copy of the object in the set of program code with an operation to update a pointer in the set of optimized program code such that the optimized program code will not perform the operation to generate the copy of the object.

2. The method as recited in claim 1, wherein the operation to update the pointer comprises an operation to make an output object of a procedure point to the object, wherein the object is built by the procedure.

3. The method as recited in claim 1, wherein the copy of the object comprises an immutable object.

4. The method as recited in claim 1, wherein the operation to generate the copy of the object comprises an operation to generate output of a data builder function, and wherein determining that the set of program code does not comprise one or more references to the data object after the operation to generate the copy of the data object comprises determining that a state of the object is not visible within the program code outside of the data builder function.

5. The method as recited in claim 1, wherein the set of optimized program code comprises a truncation or reallocation of memory storing the object.

6. The method as recited in claim 1, further comprising:

determining that a set of program code comprises an operation to generate a copy of an object into which data is accumulated, wherein the set of program code is executable using a virtual machine; and

in response to determining that the set of program code comprises the operation to generate the copy of the object into which data is accumulated, generating a set of further optimized program code, wherein the set of optimized program code comprises an allocation of one or more memory pages to store the object with a copy-on-write parameter instead of the operation to generate the copy of the object.

7. A non-transitory computer-readable storage medium storing program instructions computer-executable to perform:

determining that a set of program code does not comprise one or more references to an object after an operation to generate a copy of the object, wherein the set of program code is executable using a virtual machine; and

generating a set of optimized program code, executable using a virtual machine, comprising replacing the operation to generate the copy of the object in the set of program code with an operation to update a reference in the set of optimized program code such that the optimized program code will not perform the operation to generate the copy of the object.

8. The computer-readable storage medium as recited in claim 7, wherein the operation to update the reference comprises an operation to make an output object of a function reference the data object, wherein the data object is built by the function.

9. The computer-readable storage medium as recited in claim 7, wherein the copy of the object comprises an immutable object.

10. The computer-readable storage medium as recited in claim 7, wherein the operation to generate the copy of the object comprises an operation to generate output of a data builder function, and wherein determining that the set of program code does not comprise one or more references to the data object after the operation to generate the copy of the data object comprises determining that a state of the object is not visible within the program code outside of the data builder function.

11. The computer-readable storage medium as recited in claim 7, wherein the program instructions are further computer-executable to perform:

determining a size of memory storing the object;

wherein the operation to generate the copy of the object is replaced if the size of the memory storing the object is larger than a threshold size.

12. The computer-readable storage medium as recited in claim 7, wherein the program instructions are further computer-executable to perform:

determining that a set of program code comprises an operation to generate a copy of an object into which data is accumulated, wherein the set of program code is executable using a virtual machine; and

in response to determining that the set of program code comprises the operation to generate the copy of the object into which data is accumulated, generating a set of further optimized program code, wherein the set of optimized program code comprises an allocation of one or more memory pages to store the object with a copy-on-write parameter instead of the operation to generate the copy of the object.

13. A method, comprising:

performing by one or more hardware computing devices;

determining that a set of program code comprises an operation to generate a copy of an object into which data is accumulated, wherein the set of program code is executable using a virtual machine; and

in response to determining that the set of program code comprises the operation to generate the copy of the object into which data is accumulated, generating a set of optimized program code, wherein the set of optimized program code comprises an allocation of one or more memory pages to store the object with a copy-on-write parameter instead of the operation to generate the copy of the object.

14. The method as recited in claim 13, further comprising:

executing the set of optimized program code using the virtual machine, comprising;

attempting to write to at least one of the memory pages storing the object; and

in response to the attempting to write, causing a copy of the at least one memory page to be generated.

15. The method as recited in claim 13, wherein the object comprises a buffer into which the data is accumulated using a plurality of modification operations resulting from an input/output operation.

16. The method as recited in claim 13, wherein the data is accumulated using a plurality of buffers.

17. The method as recited in claim 13, further comprising:

determining that the set of program code does not comprise one or more references to an additional object after an operation to generate a copy of the additional object; and

in response to determining that the set of program code does not comprise the one or more references to the additional object after the operation to generate the copy of the additional object, replacing the operation to generate the copy of the additional object in the set of program code with an operation to update a pointer in the set of optimized program code.

18. The method as recited in claim 17, wherein the operation to update the pointer comprises an operation to make an output object of a procedure point to the object, wherein the object is built by the procedure.

19. The method as recited in claim 17, wherein the copy of the object comprises an immutable object.

20. The method as recited in claim 17, wherein the operation to generate the copy of the object comprises an operation to generate output of a data builder function, and wherein determining that the set of program code does not comprise one or more references to the data object after the operation to generate the copy of the data object comprises determining that a state of the object is not visible within the program code outside of the data builder function.