JPS6132155A - Buffering system of unspecified length continuous area - Google Patents

Abstract

PURPOSE:To reduce wasteful memories by classifying an unspecified length of each request in advance to a buffer request of unspecified length of continuous area and managing an idle continuous area in the unit of nearly the greatest common measure of the length. CONSTITUTION:A buffer management section 3 classifies buffer length as to bufferlength information of a buffer required in each job program processed by a job program processing section 2 in advance at the start of system operation. Then the greatest common measure M is calculated, the M is taken as the size of each element of a buffer pool 5 to manage the buffer pool in the unit of elements. When a buffer request comes, a continuous idle area satisfying the request buffer length is segmented in the unit of elements and the elements are returned in a pointer chain of the idle buffer management table 6 at return.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a data buffering method in a data processing system, and more particularly to an extraction control method for efficiently acquiring continuous areas of unspecified length from an area pool. .

[Conventional technology]

When executing an input/output command in a data processing system, an input/output buffer of a necessary size is secured in memory to buffer the input/output data.

In the conventional cover sofa control, a continuous area for the requested input/output cover sofa is carved out from a buffer pool on the virtual memory. By the way, when the size of the cutout area in each cutout request is unspecified, a method is adopted in which a contiguous area of the maximum required buffer length is cut out for each request.

Figure 2 conceptually shows such a buffer pool. For example, when buffers of various sizes from a minimum of 12 bytes to a maximum of 256 bytes are required, the size of the cutout element of the buffer pool can be changed. maximum 25
It is set to 6 bytes, and an element of the same size is cut out for each buffer request. For this reason, when a request to cut out a buffer in a small area of less than 256 bytes continues, a large amount of wasteful unused area as shown by the shaded area ρ in the figure occurs, resulting in a decrease in memory utilization efficiency.

[Problem that the invention seeks to solve]

The present invention is a series of unspecified lengths! This is to improve the problem that conventionally, in response to a request to acquire an area buffer, wasted area was generated when extracting a buffer from the buffer pool.

Means for Solving Problem c] The present invention classifies in advance the unspecified length of each request for buffer requests in continuous areas of unspecified length, and further calculates approximately the greatest common divisor of each classified length. This length is used as a unit to manage free continuous areas in the area, and to cut out a buffer of the requested length as necessary.

As a specific means for that purpose, the present invention provides a data processing system equipped with a buffer management means that performs a process of cutting out a buffer of a necessary size from a buffer pool when a buffer request for a continuous area of unspecified length is made. , the buffer management means checks in advance the sizes of unspecified lengths in all buffer requests, calculates the approximate greatest common divisor among them, and calculates the size of the value of the greatest common divisor in units of elements. The continuous free area of the buffer pool is managed in a table, and when there is a buffer request for a continuous area of unspecified length, a continuous free area sufficient to satisfy the requested buffer length is cut out in element units.

〔Example〕

The details of the present invention will be explained below based on examples.

FIG. 1 is a block diagram of one embodiment of the present invention, in which 1 is a processing device, 2 is a business program processing section, 3 is a buffer management section, 4 is a virtual memory, 5 is a buffer pool, and 6 is a main Management table 7 represents an empty buffer management table.

The buffer management unit 3 classifies buffer lengths by size based on the buffer length information of the buffers required for each business program processed by the business program processing unit 2 in advance at the time of starting operation of the system, etc. Then, the greatest common divisor (M) of these is calculated, and the value of M is set as the size of each element of the buffer pool 5, and the buffer pool is managed in element units.

For example, the length of the buffer requested to be cut out from the business program processing unit 2 is 24.
Assuming that there are four types of 36.96.120, the greatest common divisor obtained is 1;2. Therefore, the size of each element of buffer pool 5 is determined to be 12 bytes. Note that the size of the entire area of buffer pool 5 is Q.
and the number of elements is n, then n is Q = 12
It is given as the integer part of the quotient divided by .

The buffer management unit 3 manages unused elements using an empty buffer management table 7. The head position of the free buffer management table 7 is pointed to by the main management table 6.

The free buffer management table 7 represents blocks of consecutive unused elements as a pointer chain, and the buffer management unit 3 responds to each buffer 7 request by
A necessary number of consecutive empty elements sufficient to satisfy the breadth size are cut out and removed from the pointer chain of the empty buffer management table 7.

On the other hand, if the buffer that was being used is returned due to the completion of processing of a business program, the element is
It is incorporated into the pointer chain of the free buffer management table 6 again.

In the illustrated example, elements ■ and ■ and after [phase] are in use, so elements ■ to ■ and element ■
. . . . . . . . . . . . . . . . . . . . . . . . . , . . . . . . . . . . . . . . . . . . . . . . .

Note that the pointer is placed only between blocks, and within a block, it only manages the number of consecutive blocks starting from the first block (for example, 3 for ■ and 4 for ■).

FIG. 3 shows an example of buffer management processing by changing the state of the free buffer management table 7. Indicates the table state that is connected to the chain as an empty block, and (in the figure) is 2
There is a buffer request for the element, so the table state after cutting out the elements and ■ from the free block consisting of elements ■ to ■, and fcl is returned with the elements ■ and This shows the table state when ■ to ■ become one continuous area block.

Next, a basic control algorithm for buffer extraction and return processing by the buffer management section 3 will be explained using FIG. 4.

FIG. 4 shows an example of the state of the buffer pool at a certain point in time, where a and b represent the addresses of the upper and lower limits of the buffer pool, and ρρ represents the length of a continuous area of arbitrary length. As shown, ρρ=X, ∼X
There are five consecutive free blocks of l, and the blocks between these free blocks are in use.

Here, it is assumed that buffer requests for Qρ=L and QQ=L occur successively, and thereafter, buffer returns for Qρ=L and QQ=L occur successively. Xl to X
S and L1 to L4 may have any length.

(i) In the buffer cutout process for the buffer request (Qρ-ri), a block that satisfies the length Ll is checked from the free blocks X1 to X5 and cut out.

If X5 <Xa <Xl <L, l<Xl <Xa
Assuming that the block is larger than Ll and is closest to Ll, L1 is cut out from Xl, which is set as being in use, and the remaining (Xz Ll) is managed as a free block.

(ii) The buffer extraction process for the next buffer request (ρρ=L2) is performed in the same manner as in (i).

If (Xz Ll) <XS <Xa <Lz
If <Xl<Xl, then 2 minutes of X, which is larger than Lz and closest to Lz, is cut out and used, and then the remaining (Xl Lz) is managed as a free block.

(iii) In the return processing for buffer return (UQ, -Lr), check whether the address of the L3 area is consecutive to any free block, and if so, manage them as one free block. .

If L3 is continuous to the tail of Xl and also continuous to the head of Xa, then ((Xz - Ll ) + L
z +X3) new empty blocks are created.

(iv) The return process for the next buffer return (12Q=L4) is performed in the same way as in (iii).

Assuming that L4 is continuous to the tail of Xl and not continuous to the head of X, new empty blocks for (Xl + L4) will be created.

By the way, let the length of each buffer request be Lr, Lz, L3.
..., L7, the above, L4, X, and X can be expressed as follows using the greatest common divisor M.

Ll −QI M X+'=XINMLz
=+2z M Xz = Xl ML3 =
Q 3 M Xa = X3MLa = Q
a M Xl =*a MX5=x5M Here, Ql to Q4 and Xl to X are each integers. In this case, the new free blocks created in (i) to (iv) above are respectively: (i), (xz -Q+)XM (ii), (Xl -Qz)XM(iii), ( xz -Q+ +03+X4 )xM(1v) is expressed as (Xl +u4)XM.

However, it is extremely difficult in terms of software technology to manage the actually required buffer size as an arbitrary length. Therefore, as the greatest common divisor of Ll to Ln, an approximate value M is used depending on the situation. In this case, L, ≦u'l M' X r = X'+
M'L2≦, Q'2 M' Xz = X'
Z M'L3≦, Q'3 M' X 3 =
X'3 M'L4 <ll'4 M' X4
= X'4 M'Xv, = X'5 M', and the buffer is extracted using the value on the right side. For this reason, there may be a slight difference between the true buffer length requested and the buffer length to be cut out, which may result in wasted space. It can be made negligibly small compared to the example case.

[Effects of the Invention] According to the present invention, the larger the difference in buffer length between each buffer request, the more efficient buffer extraction can be performed compared to the conventional buffering method, reducing wasted memory. can be done.

[Brief explanation of the drawing]

FIG. 1 is a configuration diagram of one embodiment of the present invention, FIG. 2 is an explanatory diagram of a conventional buffering method, FIG. 3 is an explanatory diagram of a processing example in the embodiment configuration shown in FIG. 1, and FIG. FIG. 2 is an explanatory diagram of buffer management according to the present invention. In the figure, 1 is a processing device, 2 is a business program processing unit, 3 is a buffer management unit, 4 is a virtual memory, 5 is a buffer boolean, 6 is a main management table, and 7 is an empty buffer management table.

Claims (1)

[Claims] In a data processing system that is equipped with a buffer management means that performs a process of cutting out a buffer of the necessary size from a buffer pool when there is a continuous buffer request of unspecified length, the buffer management means Check the size of the length, find the approximate greatest common divisor between them, manage the continuous free area of the buffer pool in a table in units of elements of the value of the greatest common divisor, and create a continuous area of unspecified length. 1. A buffering method for a continuous area of unspecified length, characterized in that when a buffer request is made, a continuous free area sufficient to satisfy the requested buffer length is cut out in element units.