Code execution method and data processing device and system

Patent metadata

Field	Value
Publication number	US-12474897-B2
Application number	US-202318502085-A
Country	US
Kind code	B2
Filing date	Nov 6, 2023
Priority date	Dec 16, 2022
Publication date	Nov 18, 2025
Grant date	Nov 18, 2025

Abstract

A code execution method and a data processing device and system are provided. The method includes storing, through a data processing device, multiple executable files corresponding to multiple functions and multiple call information obtained by parsing a target model through a compile into a main memory and running, through the data processing device, the target model. The running, through the data processing device, the target model includes the following: loading an executable file from the main memory into an application code area of a local memory based on a running sequence of a plurality of functions and loading corresponding call information into a stack area of the local memory; and running the executable file in the application code area based on the call information in the stack area and transferring a running result from the local memory to the main memory after the running is completed.

First claim

What is claimed is:

1 . A code execution method, comprising: storing, through a data processing device, a plurality of executable files corresponding to a plurality of functions and a plurality of call information obtained by parsing a target model through a compiler in a host computer into a main memory, wherein the data processing device comprises the main memory and a local memory, and the local memory comprises an application code area and a stack area, the application code area is divided into a first buffer area and a second buffer area, and the stack area is divided into a first stack area and a second stack area; and
running, through the data processing device, the target model, comprising:
based on a running sequence of the plurality of functions, loading a first executable file among the plurality of executable files from the main memory into the first buffer area and loading first call information among the plurality of call information corresponding to the first executable file into the first stack area;
running the first executable file stored in the first buffer area based on the first call information stored in the first stack area, and during a period of running the first executable file, loading a second executable file among the plurality of executable files from the main memory into the second buffer area, and loading second call information among the plurality of call information corresponding to the second executable file into the second stack area; and
running the second executable file stored in the second buffer area based on the second call information stored in the second stack area after the running of the first executable file is finished; and
transferring a running result from the local memory to the main memory after running of all the executable files of the application code area is completed.

2 . The code execution method according to claim 1 , wherein the running, through the data processing device, the target model, further comprises:
during a period of running the second executable file, loading and overwriting a third executable file among the plurality of executable files from the main memory into the first buffer area and loading and overwriting third call information among the plurality of call information corresponding to the third executable file into the first stack area; and running the third executable file stored in the first buffer area based on the third call information stored in the first stack area after the running of the second executable file is finished.

3 . The code execution method according to claim 1 , wherein the local memory further comprises a firmware area for storing firmware,
step of running the executable file loaded into the application code area based on the call information loaded into the stack area comprises: running, through the firmware, the executable file loaded into the application code area based on the call information loaded into the stack area.

4 . The code execution method according to claim 3 , wherein the firmware is a dynamic overlay manager, and
before the step of running, through the data processing device, the target model, the method further comprises: loading the dynamic overlay manager into the firmware area.

5 . The code execution method according to claim 3 , wherein the step of running, through the data processing device, the target model, further comprises:
loading, through a first direct memory access engine provided to the firmware, the executable file and the call information from the main memory to the local memory; and transferring, through a second direct memory access engine provided to the firmware, the running result from the local memory to the main memory.

6 . The code execution method according to claim 1 , wherein the target model is an artificial intelligence model, each of the executable files comprises a code segment and a data segment corresponding to each of the functions, and each of the call information comprises a call variable corresponding to each code segment.

7 . A data processing device, comprising:
a main memory, configured to store a plurality of executable files corresponding to a plurality of functions and a plurality of call information obtained by parsing a target model through a compiler in a host computer; a local memory comprising a stack area and an application code area, wherein the application code area is divided into a first buffer area and a second buffer area, and the stack area is divided into a first stack area and a second stack area; and a processor, wherein the processor is configured to:
based on a running sequence of the plurality of functions, load a first executable file among the plurality of executable files from the main memory into the first buffer area and load first call information among the plurality of call information corresponding to the first executable file into the first stack area,
run the first executable file stored in the first buffer area based on the first call information stored in the first stack area, and during a running period of the first executable file, load a second executable file among the plurality of executable files from the main memory into the second buffer area, and load second call information among the plurality of call information corresponding to the second executable file into the second stack area, and
run the second executable file stored in the second buffer area based on the second call information stored in the second stack area after the running of the first executable file is finished; and
transfer a running result from the local memory to the main memory after running of all the executable files of the application code area is completed.

8 . The data processing device according to claim 7 , wherein the processor is configured to:
during a running period of the second executable file, load and overwrite a third executable file among the plurality of executable files from the main memory into the first buffer area and loading and overwrite third call information among the plurality of call information corresponding to the third executable file into the first stack area, and run the third executable file stored in the first buffer area based on the third call information stored in the first stack area after the running of the second executable file is finished.

9 . The data processing device according to claim 7 , wherein the local memory further comprises a firmware area for storing firmware,
the processor is configured to run, through the firmware, the executable file loaded into the application code area based on the call information loaded into the stack area.

10 . The data processing device according to claim 9 , wherein the firmware is a dynamic overlay manager, and the data processing device is configured to load the dynamic overlay manager into the firmware area before running the target model.

11 . The data processing device according to claim 9 , wherein the data processing device loads, through a first direct memory access engine provided to the firmware, the executable file and the call information from the main memory to the local memory, and
the data processing device transfers, through a second direct memory access engine provided to the firmware, the running result from the local memory to the main memory.

12 . The data processing device according to claim 7 , wherein the target model is an artificial intelligence model, each of the executable files comprises a code segment and a data segment corresponding to each of the function, and each of the call information comprises a call variable corresponding to each code segment.

Assignees

• Montage Technology Co Ltd

Inventors

• Pan Yong
• Xu Kang

Classifications

• G06F8/427 Primary

  Parsing · CPC title

• G06F9/445

  Program loading or initiating (bootstrapping G06F9/4401; security arrangements for program loading or initiating G06F21/57) · CPC title

• G06F13/28

  using burst mode transfer, e.g. direct memory access {DMA}, cycle steal (G06F13/32 takes precedence) · CPC title

• G06F9/44521

  Dynamic linking or loading; Link editing at or after load time, e.g. Java class loading · CPC title

• G06F8/43

  Checking; Contextual analysis · CPC title