JPH08263263A - Data processor and compressed program generation device - Google Patents

Abstract

PURPOSE: To generate a program smaller than an original program without changing a program description by encoding and compressing an execution system program by a prescribed encoding method in a partial program unit and generating an expanding table corresponding to the encoding method. CONSTITUTION: The execution system program is compressed by encoding the execution system program by the prescribed encoding method in the partial program unit, and the compressed execution system program is stored in a memory 12 and they are sequentially read into an instruction prefetch queue 20 through a bus 16 and an instruction cache 18. The compressed execution system program which is read is read at every basic block and the expanding table 24 is generated. Then, the table 24 is referred to and the compressed execution system program is expanded. The expanded execution system progams are sequentially executed by an instruction execution pipeline 26. The program can be made smaller than the original program size without changing program description.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a data processing device and a compression program creating device, and more particularly to a data processing device for sequentially executing each instruction in a program and a compression execution form program by compressing an execution form program. And a compression program creating device.

[0002]

2. Description of the Related Art A data processing device such as a microprocessor reads and executes a program stored in a memory, and the size of the program executed by the data processing device has rapidly increased due to the recent demand for higher functionality of electronic products. It is increasing. Therefore, in the data processing device, the hardware cost of the program memory is much higher than the hardware cost of the central processing unit (CPU), and it is important to reduce the size of the program. And it is a serious issue. In particular, when a program creator creates a program in a high-level language such as C, the amount of program code in the execution format becomes larger than that in the case of creating a program in assembly language. .

[0003]

However, for the program creator, the manual program change is
This often required detailed understanding of the architecture of the data processing device, which was a very difficult task. In addition, even if a programmer who is familiar with the architecture makes a change, such a change work requires a lot of time, and even if a lot of time is spent, a large reduction in the program size cannot be expected in many cases. In addition, in the course of such modification work, new errors are often mixed in with the program, which causes a problem that the development period of the program is further extended.

Therefore, a main object of the present invention is to provide a compressed program creating apparatus capable of creating a program significantly smaller than the original program size without requiring the program creator to change the program description. Is to provide. Another object of the present invention is to provide a new, high-performance data processing device capable of performing similar processing with a memory capacity smaller than the capacity necessary for storing the original program without changing the program. It is to be.

[0005]

A first aspect of the present invention is a compression program creating apparatus for creating a compressed execution format program by compressing an execution format program, wherein the execution format program is encoded in a predetermined manner in units of partial programs. It is a compression program creation device including compression means for compressing an executable program by encoding by a method, and table creation means for creating a decompression table corresponding to the encoding method.

A second aspect of the present invention is a memory means for storing a compressed execution format program obtained by compressing an execution format program by a predetermined data compression method, a program decompression means for decompressing the compressed execution format program by a predetermined data decompression method, and decompression. The data processing apparatus includes an instruction execution unit that sequentially executes each instruction included in the executed executable program.

[0007]

According to the first aspect of the present invention, the executable program is compressed by the Huffman coding method in units of basic blocks, and the decompression table in the Huffman code table format is created. In the second aspect, the memory means stores, for example, the executable program compressed by the compression program creating device. The program decompression unit decompresses the compression execution format program stored in the memory unit for each basic block while referring to a decompression table created by the compression program creation device, for example.
While decompressing, the decompressed instruction or instruction string is given to the instruction executing means without being temporarily stored. The instruction executing means sequentially executes the expanded instructions.

[0008]

According to the first aspect of the present invention, since the executable program is compressed in partial program units, the program creator can create a program that is significantly smaller than the original program size without changing the program description. Can be created. According to the second invention, the compressed execution program is stored in the memory and is decompressed into the executable format before the instruction is executed. Therefore, it is necessary to store the original program without changing the program. It is possible to execute a program that performs the same processing with a memory amount smaller than the required capacity.

The above-mentioned objects, other objects, features and advantages of the present invention will become more apparent from the following detailed description of the embodiments with reference to the drawings.

[0010]

DESCRIPTION OF THE PREFERRED EMBODIMENT Referring to FIG. 1, a data processing device 10 of this embodiment includes a memory 12 and an I / O device 14. A compressed execution format program (compression execution format program) is stored in the memory (ROM) 12, and the compression execution format program is sequentially read into the instruction prefetch queue 20 via the bus 16 and the instruction cache 18. . The instruction decompressor 22 decompresses the compression execution format program from the instruction prefetch queue 20 for each basic block, and decompresses the compression execution format program with reference to the decompression table 24. The expanded executable program (executable program) is executed in the sequential instruction execution pipeline 26. Data necessary for executing the instruction is transferred from the memory 12 or the I / O device 14 to the bus 1
6 and data cache 28.

With reference to FIG. 2, positions (label positions) 301a, 30 at which the branch instruction in the program 30 is jumped.
If the instructions from 1b to the instruction immediately before the next label position are defined as basic blocks 302a, 302b, and the basic block is a partial program, compression is performed in units of this basic block. By the compression, for example, the basic block 301a of the program 30 becomes the basic block 322a in the compressed program 32, and the label position 301a becomes the label position 321a.
Although a program written in assembler language is shown in FIG. 2 for the purpose of explaining the label and the basic block, the executable program is actually compressed by the procedure shown in FIG.

That is, the executable program written in the memory 34 is compressed by the Huffman coding method in units of basic blocks in the compression program creating device 36, and the position of the beginning of each block from the beginning of the executable program is compressed. That is, the byte displacement to the label position is calculated, and an appropriate branch destination address is set to the branch instruction in the executable program. The generated compression execution format program is burned in the memory 12 of the data compression apparatus 10. At the same time, the decompression table 24 in the Huffman coding table format is generated. In addition to the Huffman coding method, various data coding methods such as the run length method and the modified Huffman coding method can be used for this coding / compression. The decompression table may be a memory itself, but the decompression table may be composed of a logic circuit or a software table such as a microprogram. Further, the compression program creation device 36 is a part of the program development support device, or is a general-purpose workstation or the like.

The compressed execution format program is pre-read by the instruction prefetch queue 20 via the instruction cache 18 as the program execution progresses. Since the execution format program is compressed in basic block units, the instruction prefetch queue 20 receives the pre-read compressed execution format program in basic block units to the instruction decompressor 22. The instruction decompressor 22 decompresses the received compressed execution format program by referring to the decompression table 24 and, while decompressing, decompresses the decompressed instruction or instruction sequence into the program execution flow without temporarily accumulating it. Along the sequential instruction execution pipeline 26. Similarly, the instruction execution pipeline 26 sequentially executes the received executable instructions. Necessary data is read from the memory 12 or the I / O device 14 via the data cache 28.

According to this embodiment, the compressed execution format program is stored in the memory 12 and expanded into the execution format program before being executed by the instruction execution pipeline 26. Therefore, the program can be executed without changing the program and with a memory amount smaller than the capacity required for the original storage. In this embodiment, when the execution format program is compressed, the basic block is the partial program. However, the partial program length is defined by any length of two instructions or more and the basic block or less, and the execution format program is defined in this unit. It may be encoded / compressed.

As a method of compressing an executable program, one instruction may be a partial program and encoding / compression may be performed in units of one instruction. In this case, effective compression can be achieved by checking the appearance frequency of each instruction from the entire program and assigning a code having a shorter bit length to an instruction that appears more frequently. That is, since frequently appearing instructions differ depending on the program, it is possible to perform optimal compression for each program by changing the code assignment for each program.

Further, in this embodiment, the instruction cache is used for reading the instruction and the data cache is used for reading the data. However, the present invention is not limited to this case, and the instruction cache and the data cache are not used. Of course, it can be applied to cases.

[Brief description of drawings]

FIG. 1 is a block diagram showing one embodiment of the present invention.

FIG. 2 is a diagram showing a concept of compression of an executable program according to the present invention.

FIG. 3 is a block diagram showing a procedure of compressing an executable program according to the present invention.

[Explanation of symbols]

 10 ... Data compression device 12 ... Memory 20 ... Instruction prefetch queue 22 ... Instruction decompressor 24 ... Decompression table 26 ... Instruction execution pipeline 36 ... Compressed program creation device

Claims (9)

[Claims] 1. A compressed program creating apparatus for creating a compressed executable program by compressing the executable program, wherein the executable program is encoded in a partial program unit by a predetermined encoding method. A compression program creation device comprising: a compression means for compressing a program; and a table creation means for creating a decompression table corresponding to the encoding method. 2. The compression program creating apparatus according to claim 1, wherein said compression means defines a partial program length by any length of 2 instructions or more and less than a basic block. 3. The compression means encodes one instruction of the execution format program as one partial program, assigns a code having a bit length according to the appearance frequency of each instruction in the execution format program, and encodes the code. Item 1. A compression program creation device according to item 1. 4. An address value calculating means for calculating a compressed address value of each label position included in the executable program, and an address for setting the compressed address value for each branch instruction in the executable program. The compression program creating device according to claim 1, further comprising a value setting means. 5. The compression means includes Huffman coding means for compressing the executable program by a Huffman coding method, and the table generating means generates a Huffman coding table generating a decompression table in a Huffman coding table format. The compression program creating device according to claim 1, further comprising means. 6. Memory means for storing a compressed execution format program obtained by compressing an execution format program according to a prescribed data compression method, program decompression means for decompressing the compressed execution format program according to a prescribed data decompression method, and decompressed execution. A data processing device comprising instruction execution means for sequentially executing each instruction included in a formal program. 7. The memory means stores a compression execution format program obtained by compressing an execution format program in partial program units by a predetermined data compression method, and the program decompression means stores the compression execution format program by a predetermined data decompression method. 7. The data processing device according to claim 6, wherein the instruction is expanded in units of partial programs, and the instruction executing means sequentially executes each instruction included in the expanded partial program. 8. A compression execution format program compressed by a unit of a partial program whose program length is defined by any length of two instructions or more and less than a basic block is stored in the memory means. Data processing device. 9. The memory means performs compression execution in which one instruction of the execution format program is set as one partial program, and a code having a bit length corresponding to the appearance frequency of each instruction in the execution format program is assigned and encoded. The data processing device according to claim 8, wherein the data processing device stores a formal program.