JPH0224726A - Optimum data allocating system - Google Patents

Abstract

PURPOSE:To increase translation speed and execution speed by allocating data in the ascending order of data lengths. CONSTITUTION:A front end 2 inputs a source program 1 and outputs an intermediate language 8. At such a time, a code table forming part 3 forms a code table. Aligning means 9 obtains a data length from the code table, allocates the addresses of the code table so that the data lengths are arranged in the ascending order and stores them as allocation order information 4. A data length order data allocating means 5 executes data allocation by using the information 4 and the code table in the language 8, and returns the allocated data to the code table in the language 8. A back end 6 inputs the data-allocated intermediate language 8 and forms an objective program 7.

Description

[Detailed Description of the Invention] [Field of Industrial Application] The present invention relates to a compiler for a computer system,
In particular, data allocation when approaching data by allocating a base address as a base point at intervals of maximum displacement, setting that base address in a base register, and setting a relative address from the base address in an index register or displacement. Regarding.

[Conventional technology]

Traditionally, data allocation was based on the user's declaration layer or reference order.

Or it was done in the order of generation of Kon/9ira.

In FIG. 6, if the order of declaration or reference by the user is data A, data B, data C, and data D, the data allocation is also data A, data B.

Data C and data D were performed in this order.

To explain this with reference to FIG. 6 (1) t (2), in the example of data allocation, in the case of data B, address 1 is first output to the instruction to load base register 1, and
Next, calculate displacement 1.

Data B is approached using this displacement and modification of the pace register. The same goes for data D, which outputs an instruction to load address 2 into base register 2,
Approach after calculating Disgrace 2. In addition,
The address θ representing the beginning of the data area is set to the base register θ
The herotoss command is normally output unconditionally at the beginning.

[Problem to be solved by the invention]

However, in the conventional data allocation method described above,
If the user has previously declared or referenced data with a data length greater than the large displacement, or
The first step is to declare and reference the generated data, and even if the data length of the generated data is small, to access the data, output a load instruction to the base register, and calculate the relative address from the base register. etc. had to be dealt with.

[Means to solve the problem]

According to the present invention, the sorting means obtains a pointer value to a symbol table of each data, and sorts the pointer values in ascending order using the data length of the data as a key, and the sorted means outputs the sorted data. Data length order data allocation is a method that inputs order information for allocation that saves pointer values, and input order information for allocation generated by sorting means, and uses that information to allocate data starting from the data with the smallest data length. A data allocation method is obtained which is characterized by having the following.

〔Example〕

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

FIG. 1 shows an embodiment of a compiler for carrying out the present invention, which is a 70-year program that inputs a source program 1 and outputs an intermediate language 8. A symbol table generation unit 3 that generates a symbol table of data declared by a user belonging to the front end 2 and data generated by a compiler. Sorting means 9 of this embodiment belonging to the symbol table generation unit 3. Alignment means 9
The layout, which is the output of , is the order information 4. Intermediate language 8t outputted by front end 2 - Human output of target program 7 Hack end 6. It is located between the front end 2 and the back end 6, and performs data allocation by inputting the order information 4 outputted by the sorting means 9 and the symbol table in the intermediate language 8 outputted by the front end 2. 1 allocation consists of a data length order data allocation means for returning an address to a symbol table in the intermediate language 8.

Refer to FIG. 2 for the operation of the symbol table generation section 3.

In the reading step 21, read one piece of information about the data declared by the user or the data generated by the controller/mother 1
If it is determined in judgment step 22 that there is more information to be read, the generation step f2 is executed for the data represented by that information.
Step 3 generates a symbol table.

The operation of the alignment means of this embodiment starts from here.

Information is obtained from the currently generated symbol table, and in level determination step 24 it is determined whether this data is data with the highest level of the structure or independent data (has no dependent data). . If either condition is satisfied, control is passed to data length GET step f25 to obtain the data length from the symbol table. Next, in the layout order information generation step 26, the address of the symbol table is saved as layout order information 4 so that the data length is in ascending order.9 An example of the layout order information 4 is shown in FIG. The allocation information 31 consists of 90% allocation order, data length, and a pointer to the symbol table.

Each pointer points to a symbol table 32.33, 34.35. If next, data E symbol table 36
Assuming that the data length of data E is 10, the information 31 in 1 allocation will be placed in the position immediately before order 3. i.e. 1
For the cell in order 3, the data length is 10. The pointer value to the symbol table 36 of data E is entered as the pointer to the symbol table, the information on data C moves to order 4, and the information on data A moves to order 5.

Referring to FIG. 4, the operation of the data allocation means 5 in order of data length is performed in the initial reading Stella 7'41.

Assignment is from order information 4 to order number 1 information is read 1 judgment step 42 if it is determined that there is information to be read is 1
For the allocation, step 43 traces the pointer value in the information read, obtains a symbol table, and allocates the data represented in this symbol table.1 Allocation returns the attribution to the symbol table to complete the allocation. In the reading step 44, the information of the data specified for the 9th layout is read from the layer information 4 in order, and
1 allocation is performed until information is exhausted.

Layer information 4 does not include information on subordinate data, but the assignment of data at the highest level is completed by means 5 in the order of data length, and the assignment of subordinate data is performed using the upper level assignment. Since it depends on the previous one, it can be allocated at any time after the upper level allocation is completed, and the data allocation in order of data length may be included in the means 5 or in the pack end 6.

FIG. 5 shows an example of data allocation performed in this way. The υ tags are assigned in ascending order of data length as in Example 51.

Referring to FIG. 5 (1) and (2), since data D and data B are allocated within the maximum displacement, it is difficult to approach data B as shown in FIG. Outputting a load command to pace register 1 at address 1 and calculating displacement 1, which had been done, and outputting a load command to base register 2 at address 2, which had been performed to approach data D, and calculating displacement 2. Since there is no need for calculation, translation can be done faster.

Since the number of steps in the objective program to be output can also be reduced, execution can also be made faster.

〔Effect of the invention〕

As explained above, by allocating data in ascending order of data length, the present invention eliminates the need for outputting unnecessary target programs and calculations for accessing data, and improves translation speed and execution speed. This has the effect of increasing speed.

[Brief explanation of the drawing]

FIG. 1 is a block diagram of a conoquila employing an embodiment of the present invention, FIG. 2 is a flowchart of the symbol table generation unit 3 including the alignment means 9 in FIG. 1, and FIG. 3 is a diagram showing the layout in FIG. 4 is an example of the layer information 4, and FIG. 4 is a flowchart of the means 5 for data allocation in order of data length in FIG. 1, and FIG. 5 (1). (2) is an example of data allocation when the two embodiments are implemented, and FIG. 6 (1) and (2) are examples of conventional data allocation. 1... Original program, 2... Front end, 3
...Symbol table generation unit, 4...Layout is layer information, 5...
Data length order data allocation means, 6... back end, 7
. . . Objective program, 8. Intermediate language, 9. Sorting means. Hole: 5JLJ! I) Flowing Umbrella: Dezunogajm Fig. 2 2 (Dotted line is the symbol table generation part for the pa sub-floor [; Added alignment means of book completion) Fig. 3 Fig. 5

Claims (1)

[Claims] 1) A computer system that accesses data by allocating a base address as a base point at intervals of maximum displacement, setting that base address in a base register, and setting a relative address from the base address in an index register or displacement. In the compiler, a sorting means obtains a pointer value to the symbol table of each data, sorts each pointer value in ascending order using the data length of the data as a key, and stores the sorted pointer values output by the sorting means. Optimal data, characterized by comprising a data length order data allocation means that inputs the allocation order information and the allocation order information generated by the sorting means, and performs data allocation in ascending order of data length based on the information. Allocation method.