JP2000122877A - Method for arranging program code - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve the execution speed of a program by arranging a function and data which are actually and frequently used in a high speed built-in RAM. SOLUTION: This method compiles a source program 1 to generate an object program 3, executes the program 3 to acquire profile information 5 where dynamic information for the case the program is executed is recorded, analyzes the information 5 to decide a speeding-up object section, selects a function and data which are frequently referred to in the speeding-up object section, generates a source program 2 where a replacement code is inserted to a part being the head of the speeding-up object section, which code is for transferring the selected function and data to the high speed built-in RAM so as to replace the contents of the RAM and compiles the program 2 to generate an object program 4.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method of arranging a program code, and more particularly, to a high-speed built-in R which can perform high-speed data reading, writing and instruction execution.
The present invention relates to a method for arranging a program code in a computer having an AM.

[0002]

2. Description of the Related Art In a current computer, a CPU is used.
Many of the memory spaces have a high-speed built-in RAM capable of reading and writing data and executing instructions at high speed. In a computer provided with a high-speed internal RAM, memory management is automatically performed, and all variables are assigned to the high-speed internal RAM, or memory management is performed by a user, as long as there is free space in the high-speed internal RAM. By explicitly specifying a variable to be allocated to the high-speed internal RAM, the specified variable is allocated to the high-speed internal RAM.

A user who develops a computer program having a high-speed internal RAM in a part of the memory space of the CPU is expected to improve the execution speed of an object program by effectively using the high-speed internal RAM. Have been.

Japanese Patent Application Laid-Open No. 9-44363 discloses a method for effectively utilizing a high-speed internal RAM, in which a program input code is statically analyzed, and a high-speed internal RAM is analyzed from the analysis result.
The figure shows a method of selecting code and data to be placed in the storage device and inserting a code for dynamically placing them. Its configuration is as shown in FIG.

First, a source file 101 is input, and an input unit 104 and a syntax analysis unit 105 statically analyze the source file. Using the analysis result, the arrangement designation information output unit 106 selects codes and data that can be speeded up by arranging them in a high-speed internal RAM (arrangement data discrimination unit 109).
The position to be transferred to the AM is determined (exchange processing output position discriminating unit 110), and as a final result, the arrangement designation information is output as the arrangement designation information file 108 (the arrangement designation information file output unit 111).

Finally, an input unit 104, a syntax analysis unit 105
Code output unit 1 using the analysis result of the source file obtained by
At 07, an object file 103 is generated.

In the object code generation unit 107, first, the obtained arrangement designation information file 108 is inputted to the arrangement designation information file input unit 112, and the replacement output unit 113 uses this to output the transfer code to the high-speed internal RAM. Insert and convert to final object file.

[0008]

SUMMARY OF THE INVENTION The above-mentioned Japanese Patent Application Laid-open No. Hei 9-4
In the method disclosed in Japanese Patent No. 4363, only static analysis of a source program is performed in order to select code and data to be arranged in a high-speed internal RAM, and dynamic analysis such as how the program is actually executed is performed. Since the profile information is not used, there is a possibility that codes and data that are not actually used often are arranged in the high-speed internal RAM. As a result, the high-speed built-in RAM is wasted, and there is a possibility that the opportunity to increase the speed is missed.

An object of the present invention is to provide a program data arranging method which can improve the execution speed of a program by arranging functions and data that are actually frequently used in a high-speed internal RAM.

[0010]

SUMMARY OF THE INVENTION The present invention relates to a method of arranging a program in a computer system in which a high-speed internal RAM is arranged in a memory space of a CPU. To obtain profile information recording dynamic information at the time of program execution, determine a speed-up target section using the profile information, a function frequently referred to in the speed-up target section, A source program in which data is selected and a replacement code for transferring the selected function and data to the high-speed internal RAM and exchanging the contents of the high-speed internal RAM is generated at the beginning of the high-speed target section. Compiles the source program with the replacement code And generating a preparative program.

[0011]

Next, an embodiment of the present invention will be described with reference to the drawings.

FIG. 1 is a flowchart showing an embodiment of the program arrangement method of the present invention. In FIG. 1, first,
The object program 3 is generated by compiling the source program 1, and then the object program 3 is executed to obtain profile information 5 which records information at the time of execution of the program.

Next, by using the profile information 5, functions and data that are frequently referred to in the section to be accelerated are detected, and these are stored in a high-speed internal RAM to speed up the execution.
And a source program 2 in which a code for replacing the contents of the high-speed built-in RAM is inserted immediately before each section at the time of execution, with the function and data to be arranged.

Finally, the source program 2 into which the code for replacing the contents of the high-speed internal RAM has been inserted is compiled, the arrangement information 6 is given to the linker, and the final object program 4 is obtained.

Next, the operation of the program arrangement method of the present invention will be described in detail with reference to FIGS. FIG. 2 is a flowchart for explaining in detail the functions, data arrangement information creation, and replacement code insertion processing in FIG.

First, a source program 1 is created, and an object program 3 is obtained from the source program 1 using a compiler.

Next, by executing the object program 3 on a simulator or a target machine, profile information 5 which records dynamic information at the time of program execution, such as data access frequency and function execution trajectory, is recorded.
Get.

The profile information is, specifically, a function execution transition information obtained by tracing a function execution transition for a function. For data, the execution time is divided into certain intervals, and the number of times of reference to each variable during the interval is divided. This is data reference information for recording (access count). FIG. 3 is a diagram illustrating an example of function execution transition information, and FIG. 4 is a diagram illustrating an example of data reference information.

Here, a method of acquiring profile information will be described. To get profile information,
There is a method of embedding the probe code and a method of using a simulator.

In the method of embedding the probe code, in the case of a function, a call code of the probe code is inserted at the entry and exit of all functions, and when the probe code is called at the entry, the execution start of the function is recorded, At the exit, execution transition information of the function is obtained by recording the execution end of the function as data or a file in the memory together with the time at that time.

In the case of data, a call code of a probe code is inserted immediately before a reference instruction to each data (these instructions are always executed at the time of referring to data), and the number of times the probe code refers to the data in each period. To obtain data reference information.

In the method using a simulator, in the case of a function, the execution of a function call instruction and a return instruction from a function are monitored by using the simulator function relating to the instruction, and the same information as described above is recorded to record the function. Obtain execution transition information. A function can be specified by recording its call address.

In the case of data, access (reading and writing) to the address where the data is located is monitored by using the function of a simulator for monitoring access to a specific address. , The probe code is called by this interrupt, and the same information as in the above case is recorded to obtain data reference information.

Next, a section to be speeded up is determined using the profile information 5. To determine the section to be speeded up, the user looks at the profile information 5 and specifies a section that needs to be speeded up, or a specific function (about 1 to 4) determines a certain percentage of the section execution time. This is performed by automatically designating a section that is occupied and a section in which the number of reference times of a variable exceeds a certain number.

Next, the profile information 5 of each speed-up target section is analyzed to detect functions and data that are frequently referred to in the speed-up target section, and these are stored in a high-speed internal RAM for high-speed execution. Select the function and data to be placed,
At the beginning of the section, the selected function and data are transferred to the high-speed internal RAM, and a source program 2 in which a replacement code for replacing the contents of the high-speed internal RAM is inserted is generated.

Here, a method for selecting a function to be allocated in the high-speed internal RAM using the profile information will be specifically described with reference to FIG. FIG. 5 is a diagram showing the execution status of a program by execution transition information of a function.

From FIG. 5, in section 1, the functions Func_A,
It can be seen that Func_B is executed alternately, and in section 2, only the function Func_Z is executed. Therefore, in section 1, the functions Func_A, Func_A
It is determined that c_B is frequently used, and that the function Func_Z is determined to be frequently used in the section 2, and is set as a target of a function to be transferred to the high-speed internal RAM immediately before each section. Insert the exchange code for transfer into the part of the program that

Since the insertion of the transfer replacement code involves the overhead of the transfer to the high-speed built-in RAM, it is determined whether or not the speed can be increased corresponding to the transfer cost, and it is determined that the sufficient speed can be obtained. Only do it if. Further, a block transfer instruction code or a DMA (direct memory access) transfer instruction code can be used as the replacement code. FIG. 6 shows a specific example of the replacement code insertion.

In the above description, the relocation of functions has been described, but the same applies to data.

Next, there is provided a function and data area in which functions and data are arranged collectively, respectively, and arrangement information 6 is created which enables the functions and data to be exchanged together at a time. I do.

The functions and data placed together are assigned an index number for each unit. When replacing the high-speed internal RAM, the index number is designated as a parameter, and a replacement routine is called. The replacement routine also includes the following table data that records which area in the memory the index number corresponds to.

[0032] At the time of replacement, the function or data arranged in the area indicated by the designated index number is transferred to the high-speed internal RAM.

The arrangement information 6 gives an instruction to the linker so that functions and data to be arranged are arranged in the area indicated by the designated index number.

FIG. 7 shows an index (I) in the function area.
FIG. 2 is a diagram showing a relationship between D: 1, ID: 2, ID: 3) and a high-speed internal RAM.

Finally, the source program 2 into which the exchange code for transfer to the high-speed internal RAM has been inserted is compiled, the arrangement information 6 is given to the linker, and the final object program 4 is obtained.

[0036]

As described above, according to the present invention, functions and data that need to be accelerated are arranged in a high-speed internal RAM at the time of execution by using profile information obtained from a dynamic execution state of an object program. Accordingly, access to functions and data can be speeded up, and the execution speed of a program can be improved.

[Brief description of the drawings]

FIG. 1 is a flowchart showing an embodiment of a program arrangement method according to the present invention.

FIG. 2 shows functions, data arrangement information creation, and high-speed built-in RA
It is a flowchart explaining in detail the process of M transfer code insertion.

FIG. 3 is a diagram illustrating an example of execution transition information of a function;

FIG. 4 is a diagram illustrating an example of data reference information.

FIG. 5 is a diagram showing a program execution status by function transition information;

FIG. 6 is a diagram showing a specific example of insertion of a replacement code.

FIG. 7: Index in function area and high-speed internal RAM
FIG.

FIG. 8 is a diagram showing a conventional method for effectively utilizing a high-speed internal RAM.

[Explanation of symbols]

 1, 2 Source program 3, 4 Object program 5 Profile information 6 Placement information

Claims (7)

[Claims] In a method of arranging a program in a computer system in which a high-speed internal RAM is arranged in a memory space of a CPU, it is necessary to increase the speed by using profile information which records dynamic information at the time of executing an object program. A code arrangement method for a program, comprising selecting a function and data, and inserting, into a source program, an instruction code for arranging the selected function and data in the high-speed internal RAM when an object program is executed. 2. A method for arranging a program in a computer system in which a high-speed internal RAM is arranged in a memory space of a CPU, wherein a source program is compiled to generate an object program, and the object program is executed to execute the Obtaining profile information in which dynamic information is recorded; determining a speed-up target section using the profile information; selecting a function and data that are frequently referred to in the speed-up target section; The selected function and data are transferred to the high-speed internal RAM at the beginning of the target section, and the high-speed internal RA
A code placement method for a computer program, comprising: generating a source program in which a replacement code for replacing the contents of M is inserted; and compiling the source program in which the replacement code is inserted to generate an object program. 3. A function and data area in which functions and data to be transferred to and exchanged with the high-speed internal RAM are collectively arranged, respectively, so that the functions and data can be exchanged all at once at the time of exchange. 3. The code arrangement of the program according to claim 2, wherein the arrangement information to be created is compiled, the source program into which the replacement code is inserted is compiled, the arrangement information is given to a linker, and a final object program is obtained. Method. 4. The profile information is, in the case of a function, execution transition information of a function obtained by tracing the execution transition of the function. In the case of data, the profile information divides the execution time into fixed sections and indicates the number of times of reference to a variable in each fixed section. 4. The method according to claim 2, wherein the information is data reference information to be recorded. 5. The determination of the high-speed target section is performed by a user
The method according to claim 2, wherein the profile information is analyzed to specify a section that can be executed at high speed even when transfer overhead is included. 6. The determination of the high-speed target section is performed by analyzing the profile information to determine a section in which a specific function occupies a certain percentage of the section execution time and a section in which the number of times of reference to a variable exceeds a certain number of times. The method according to any one of claims 2 to 4, wherein the method is automatically specified. 7. The method according to claim 2, wherein the replacement code is a block transfer instruction code or a DMA transfer instruction code.