JPH0240721A - Parallel compiling system - Google Patents

Abstract

PURPOSE:To shorten the time required for processing by providing a task assignment control part, assigning individual tasks which can be executed in parallel, and utilizing a processor effectively. CONSTITUTION:The source program selection part 11 of a compiler 1 passes a source program 2 where source programs 21-23 are present to the task assignment control part 12. The control part 12 generates source program analytic object generation tasks 15-17. Then information from the selection part 11 and information required for compilation from an object program 31 are passed to the task 15. When a processor other than a processor where the control part 15 runs is usable by the system, the task 15 begins to run on the processor and the source programs 22 and 23 begin to be compiled by another processor. Thus, the respective programs 21-23 are distributed to an processed by different processors, so the time required for processing is shortened.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a compiler processing method, and particularly to a method for shortening the processing time of a compiler having a function of collectively translating a plurality of programs.

[Conventional technology]

This type of compiler is sometimes used to compile multiple programs at once. In such batch translation of multiple programs, in the conventional compilation method, the processing of "target program selection, analysis, and target program generation" is performed for each program based on the parameters and control statement specifications given to the compiler. This was achieved by repeating the steps in order.

[The problem that the invention attempts to solve]

In the conventional method described above, the translation of the next program is started after the translation of one program is completed, so there are the following drawbacks.

■ Even though a system that can use multiple processors should be able to translate separate programs at the same time in parallel, with conventional compilation methods, multiple programs are translated all at once using only one processor. As a result, the processor cannot be used effectively.

■ Similarly, the time required to translate an entire program cannot be reduced by increasing the number of processors.

[Means to solve the problem]

The parallel compilation method of the present invention is a compilation method that analyzes a source program and generates a target program for a system having a plurality of processors and a synchronization function between tasks executed in parallel on those processors. The present invention includes means for allocating the translation of each program to separate tasks that can be executed in parallel when a plurality of programs are translated at once.

〔Example〕

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

Figure ya1 is a functional block diagram of a compiler used in an embodiment of the present invention. The compiler 1 includes a source program selection section 11. Task allocation control unit 12, post-processing unit 13
．． Source program analysis purpose program generation task 1
” It consists of a task group 14 consisting of n.

The source program selection unit 11 passes the file name and file address in which the source program 21 to be compiled exists to the task assignment control unit 12. The task allocation control unit 12 generates a source program analysis purpose program generation task 15 and assigns it to the source program selection unit 11.
information passed from , and the generation position 1 of the target program 31
Pass information necessary for compilation, such as the location of the work area.

If a processor other than the main task, ie, the processor on which the task allocation control unit 12 is running, is available in the system, the source program analysis purpose program generation task 15 starts running on that processor and starts compiling the source program 21. Task allocation control unit 1
2 then returns control to the source program selection section 11.

The source program selection unit 11 selects the next source program 22 and passes the file name and file address where it exists to the task assignment control unit 12. The task allocation control unit 12 generates a source program analysis target program generation task 16, and as in the case of the source program 21, starts compiling the source program 22 on yet another processor.

After repeating the above process until a source program to be compiled is found, the task allocation control unit 12
waits for the completion of the generated tasks, that is, the completion of compilation, and when all tasks are completed, the post-processing unit 1
Transfer control to 3.

The post-processing unit 13 performs a process of summarizing each compilation result (for example, a process of summarizing the list outputted from each compilation or the target program 3) as necessary.

Figure 2 shows processor A, processor B, processor C,
In a system equipped with a processor D and a processor E, program 1. Program 2. Program 3. FIG. 4 is a diagram showing an aspect when programs 4 are compiled all at once. In the conventional method,
Only one processor can be used at the same time, so Program 1, Program 2. Program 3. Program 4
The compilation time of jl*j, respectively, is ! S*j4
Then, the total compilation time t = tl + tz + 1
It becomes 3+14.

FIG. 3 is a diagram showing a state in which the same program is compiled all at once by a compiler using the present invention in a system having the same processor configuration as that in FIG. 2. In the compiler using the method of the present invention, program 1. Program 2゜Program 3. Since program 4 is compiled in parallel on processor B, processor C, processor D, and processor E, the overall compilation time + [ t
= 3α+ts+β. Therefore, assuming that the compile time t3 of program 3 is sufficiently long, the entire compile time will be approximately equal to the compile time of program 3.

- If the number of processors in the membrane is greater than the number of programs to be compiled, it can be said that the longest compilation time is approximately equal to the total compilation time.

〔Effect of the invention〕

As explained above, according to the parallel compilation method of the present invention, when multiple programs are compiled at once, the translation of each program is distributed to separate processors, thereby making effective use of the processors. This reduces the time required to translate the entire program. The present invention has such effects.

...Post-processing unit, 14...Task group, 15-1
6...Source program analysis/purpose program generation tasks 1-n121-23...Source programs 1-n131-33...Objective programs 1-n0 Agent Patent attorney Honjo Shinsuke

[Brief explanation of the drawing]

FIG. 1 is a functional block diagram of a compiler used in an embodiment of the present invention, and FIG. 2 is a diagram showing how multiple programs are compiled using a conventional method on a system equipped with multiple processors. , FIG. 3 is a diagram showing an aspect when a plurality of programs are compiled on a system equipped with a plurality of processors according to the method of the present invention.

Claims (1)

[Claims] For systems that have multiple processors and a synchronization function between tasks that are executed in parallel on those processors, multiple programs can be compiled at once using a compilation method that analyzes source programs and generates target programs. A parallel compilation method characterized by having means for assigning the translation of each program to separate tasks that can be executed in parallel.