Code performance optimization based on an information-enhanced diffusion model

Patent metadata

Field	Value
Publication number	US-12217032-B2
Application number	US-202318129288-A
Country	US
Kind code	B2
Filing date	Mar 31, 2023
Priority date	Mar 31, 2023
Publication date	Feb 4, 2025
Grant date	Feb 4, 2025

Abstract

A given segment of computer code is obtained and modified to produce one or more inefficient versions of the given segment of computer code in comparison to the given segment of computer code. A code parse tree is generated for the given segment of computer code and each inefficient version of the given segment of computer code. Model embeddings are generated based on the generated code parse trees and a diffusion model is trained based on the generated model embeddings.

First claim

What is claimed is:

1. A method comprising:
obtaining, using a hardware processor, a given segment of computer code;
modifying, using the hardware processor, the given segment of computer code to produce one or more inefficient versions of the given segment of computer code in comparison to the given segment of computer code;
generating, using the hardware processor, a code parse tree for the given segment of computer code and each inefficient version of the given segment of computer code;
generating, using the hardware processor, model embeddings based on the generated code parse trees; and
training, using the hardware processor, a diffusion model based on the generated model embeddings.

2. The method of claim 1 , further comprising ordering the given segment of computer code and the one or more inefficient versions of the given segment of computer code by run time and wherein the training the diffusion model is based on the ordered segments of computer code.

3. The method of claim 1 , further comprising performing inferencing using the trained diffusion model to generate a more efficient version of an input computer code segment.

4. The method of claim 3 , further comprising deploying the more efficient version of the input computer code segment.

5. The method of claim 4 , further comprising executing the deployed more efficient version of the input computer code segment.

6. The method of claim 1 , wherein the modifying the given segment of computer code comprises integrating one or more FOR loops into one or more of the given segment of computer code and the one or more inefficient versions of the given segment of computer code.

7. The method of claim 1 , wherein the modifying the given segment of computer code comprises one or more of integrating code that refrains from performing a timely release of memory occupied by variables, integrating code for variable circular references, integrating code variable names that violate a program specification, and integrating code that causes a memory leak into one or more of the given segment of computer code and the one or more inefficient versions of the given segment of computer code.

8. A computer program product, comprising:
one or more tangible computer-readable storage media and program instructions stored on at least one of the one or more tangible computer-readable storage media, the program instructions executable by a processor, the program instructions comprising:
obtaining a given segment of computer code;
modifying the given segment of computer code to produce one or more inefficient versions of the given segment of computer code in comparison to the given segment of computer code;
generating a code parse tree for the given segment of computer code and each inefficient version of the given segment of computer code;
generating model embeddings based on the generated code parse trees; and
training a diffusion model based on the generated model embeddings.

9. The computer program product of claim 8 , further comprising ordering the given segment of computer code and the one or more inefficient versions of the given segment of computer code by run time and wherein the training the diffusion model is based on the ordered segments of computer code.

10. The computer program product of claim 8 , further comprising performing inferencing using the trained diffusion model to generate a more efficient version of an input computer code segment.

11. The computer program product of claim 10 , further comprising deploying the more efficient version of the input computer code segment.

12. The computer program product of claim 11 , further comprising executing the deployed more efficient version of the input computer code segment.

13. The computer program product of claim 8 , wherein the modifying the given segment of computer code comprises integrating one or more FOR loops into one or more of the given segments of computer code and the one or more inefficient versions of the given segment of computer code.

14. A system comprising:
a memory; and
at least one processor, coupled to said memory, and operative to perform operations comprising:
obtaining a given segment of computer code;
modifying the given segment of computer code to produce one or more inefficient versions of the given segment of computer code in comparison to the given segment of computer code;
generating a code parse tree for the given segment of computer code and each inefficient version of the given segment of computer code;
generating model embeddings based on the generated code parse trees; and
training a diffusion model based on the generated model embeddings.

15. The system of claim 14 , further comprising ordering the given segment of computer code and the one or more inefficient versions of the given segment of computer code by run time and wherein the training the diffusion model is based on the ordered segments of computer code.

16. The system of claim 14 , further comprising performing inferencing using the trained diffusion model to generate a more efficient version of an input computer code segment.

17. The system of claim 16 , further comprising deploying the more efficient version of the input computer code segment.

18. The system of claim 17 , further comprising executing the deployed more efficient version of the input computer code segment.

19. The system of claim 14 , wherein the modifying the given segment of computer code comprises integrating one or more FOR loops into one or more of the given segments of computer code and the one or more inefficient versions of the given segment of computer code.

20. The system of claim 14 , wherein the modifying the given segment of computer code comprises one or more of integrating code that refrains from performing a timely release of memory occupied by variables, integrating code for variable circular references, integrating code variable names that violate a program specification, and integrating code that causes a memory leak into one or more of the given segment of computer code and the one or more inefficient versions of the given segment of computer code.

Assignees

• IBM

Inventors

• Yuan Zhong Fang
• Liu Tong
• Yu Han Qiao
• Feng Lin
• Yang Xiang Yu
• Zou Hai Bo

Classifications

• G06F8/71

  Version control (security arrangements therefor G06F21/57); Configuration management · CPC title

• G06F8/4441 Primary

  Reducing the execution time required by the program code · CPC title

• G06N20/00

  Machine learning · CPC title

• G06N3/044

  Recurrent networks, e.g. Hopfield networks · CPC title

• G06N5/01

  Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound · CPC title