JPH09319856A - Structure data editing system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a structure editing control system which can increase the speed of a structure editing function in a graphic system. SOLUTION: In a host 1, structure management parts 8 and 9, adapted to a conventional server 10, are transferred to libraries 6 and 7 coupled to clients 4 and 5. In an accelerator 13, not only a server queue 16 but also a queue table 18 and client queues 19 and 20 are provided on a shared memory 15. Clients directly transfer a structure editing request to the shared memory, without passing the server. Thus, inter-process communication between clients and the server is omitted, to increase the speed of the structure editing function.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a structure data editing control system in a graphic system which uses an accelerator (high-speed arithmetic processing unit) and has graphic data and attribute data as a structure.

[0002]

2. Description of the Related Art In recent years, three-dimensional CAD has become widespread,
There is a demand for higher speed graphic systems. In particular, in the field of construction, a structure is often constructed by a huge amount of data to express a three-dimensional model, and there is a demand for speeding up the structure editing process.

A structure editing control system in a conventional graphic system will be described with reference to FIG. A graphic system that constructs a structure and manages data has a hardware configuration including a host (computer main body) 1 and an accelerator 13 connected to the host.
In the conventional graphic system, the server (window system) 10 receives the structure data from the clients (applications) 4 and 5 coupled to the libraries 6 and 7 of the host 1 via the socket 3 of the OS 2, and the structure management unit 8 After managing the headers of the structure and the element, the structure data is transferred to the shared memory 15 in the accelerator 13 via the host bus 12. Therefore, it is necessary to transfer the data twice between the clients 4 and 5 and the server 10 and between the server 10 and the accelerator 13. In addition,
The head address of the server queue 15 that stores the structure data is a fixed address.

[0004]

In the structure data editing control method in the above-mentioned conventional graphic system, the above-mentioned two data transfers are required until each client 4, 5 constructs the structure in the shared memory 15. Therefore, the time required for the structure editing process is wasted.

It is an object of the present invention to provide a structure edit control system capable of realizing a high speed structure edit function in a graphic system.

[0006]

To achieve the above object, the present invention provides a host having a library and a server,
The accelerator control unit is connected to the host and includes an accelerator control unit and an accelerator having a shared memory, and the accelerator control unit is requested by a structure edit request transferred from a client coupled with the library to a data communication area on the shared memory. In the structure data edit control method of a graphic system for constructing a structure, a structure management unit is provided in the library, and the client stores the structure edit request directly in the shared memory in the accelerator by the structure management unit. To do.

As described above, by migrating the structure management section of the conventional server to the library,
The client transfers the structure edit request directly to the shared memory in the accelerator without passing through the server. Therefore, it is possible to omit the conventional inter-process communication between the client and the server via the socket of the OS, and it is possible to realize the speedup of the structure editing function.

Further, according to the present invention, the accelerator control unit stores a structure edit request independently for each client by providing a client queue dedicated to the structure edit request for each client on the shared memory. Can be done. As a result, each client can independently perform efficient structure editing.

Further, in the present invention, a server queue for storing requests other than the structure edit request is provided on the shared memory of the accelerator,
The accelerator control unit can perform a synchronization process between the server queue and the client queue. This makes it possible to maintain the sequence of the entire graphic system.

[0010]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Embodiments of the present invention will be described with reference to the drawings. FIG. 2 is a diagram showing a hardware configuration of the graphic system. In FIG. 2, reference numeral 1 is a host, which includes an OS 2, clients 4 and 5, libraries 6 and 7 connected to the clients 4 and 5, and a server 10.
have. Reference numeral 12 is a host bus. An accelerator 13 is connected to the host 1 via the host bus 12, and has an accelerator control unit 14 and a shared memory 15. Reference numeral 17 is a CRT that is sequentially drawn by the accelerator 13.

FIG. 3 is a diagram for explaining the principle of the graphic system of this embodiment. In the host 1, the structure management unit, which was in the server 10 in the past (FIG. 2), is moved to the libraries 6 and 7, and each library 6,
Structure management units 8 and 9 are provided for each of the seven. Also,
In the accelerator 13, on the shared memory 15,
In addition to the conventional server queue 16, a queue table 18 and client queues 19 and 20 for each client are acquired.

The operation of the system shown in FIG. 3 will be described below. (1) When the system is activated, the server queue 16 is acquired at a fixed address on the shared memory 15. (2) Each client 4, 5 is
A connection request request with the server 10 is issued to the server 10 via the socket 3 of S2.

(3) The server 10 issues a client queue creation request to the server queue 16 in the request processing of (2) above. (4) When the accelerator control unit 14 receives the request of (3) above, it creates the client queues 19 and 20 in the shared memory 15 for each of the clients 4 and 5 in the order in which the connection request is made, and sets their addresses. Cue table 18
The address is notified to the clients 4 and 5 via the server 10 as well.

(5) On the host 1, the driver 11
The address of the accelerator 13 is memory-mapped. (6) Thereafter, the clients 4 and 5 store the structure edit request in the client queues 19 and 20 directly from the structure management units 8 and 9 via the host bus 12 based on the returned address. Requests other than the structure edit request (eg, drawing request) are stored in the server queue 16 via the server 10 as in the conventional case.

(7) The accelerator control unit 14 registers the address in the queue table 18 every time the client queues 19 and 20 are created, and polls the queue registered at that time from the top of the queue table 18,
Process a fixed number of requests one by one. Here, in FIG.
Details of the queue table 18 are shown. The accelerator control unit 14 receives the client queue 19,
The area of the queue management information 21 and the area of the request queue 22 are acquired as 20, and the address of the area of the queue management information 21 is registered in the queue table 18.

The search flag of the queue management information 21 is
This queue is used to perform a later-described synchronization process. When the value of this search flag is other than 0, this queue is ignored at the time of polling and no request process is performed. The pointer information of includes the following. Indicates the start address of the request queue 22, and indicates the next address of the request queue. Advances this pointer after the library 6, 7 or server 10 stores the request. Is advanced by the accelerator control unit 14 after the request processing is completed. The accelerator control unit 14 can know the presence or absence of an unprocessed request based on the difference between and.

As described above, the structure management unit of the conventional server 10 is moved to the library,
Since the structure management units 8 and 9 are provided for the respective libraries 6 and 7, the clients 4 and 5 can directly use the structure management units 8 and 9 in the accelerator 13 without passing the structure edit request through the server 10. The shared memory 15 is transferred to the shared memory 15. As a result, inter-process communication via the socket 3 of the OS 2 between the clients 4 and 5 and the server 10 which is conventionally performed can be omitted, and the speed of the structure editing function can be realized.

At this time, each client 4, 5 requests a structure edit request to the client queue 19, 20 for each client 4, 5 provided on the shared memory 15 by the accelerator control unit 14 when a connection request with the server 10 is made. Is stored, each of the clients 4 and 5 can independently perform efficient structure editing.

Next, in this example, since the client queues 19 and 20 are created separately from the server queue 16, it is necessary to synchronize the two queues. This synchronization processing will be described with reference to FIGS. 1, 5, and 6. When the clients 4 and 5 want to synchronize between the server queue 16 and the client queue 19, the server queue 16 includes a synchronization request to which the address of the client queue 19 is added, and the client queue 19 includes the server queue 16 with the synchronization request. Issue the synchronous request with the address added at the same timing. The details of this synchronization request are shown in FIG.

The accelerator controller 14 polls the queue table 18 (step S1 in FIG. 6),
When a synchronization request is encountered by dispatching (step S2) (step S3), the search flag of the queue designated by the synchronization request is compared with the address of its own queue (step S4). When these two are equal, it indicates that the queue specified by the synchronous request is not currently being processed. In this case, 0 is assigned to the search flag of the designated queue (step S5), and the request processing of that queue is restarted.

If the two values are different, it indicates that the request processing of the own queue must be interrupted, and the address of the designated queue is substituted for the own search flag (step S6). After that, this queue will be ignored. Therefore, when a synchronization request is encountered during polling of the queue table 18, the accelerator control unit 14 processes the queue until the corresponding synchronization request is encountered in the queue of the address added to the synchronization request. Is not done at all.

Then, when a corresponding synchronization request is encountered in the queue of the address added in the above-mentioned synchronization request, the request processing of the queue that first encounters the synchronization request is restarted. By these operations, the accelerator control unit 14 performs a synchronization process between the server queue 16 and the client queues 19 and 20 as needed by the clients 4 and 5, thereby maintaining the sequence of the entire system. It will be possible.

[0023]

According to the present invention, since the number of data transfers is reduced, the structure editing function can be speeded up. In addition, since the structure can be edited independently for each client, the waiting overhead between them is reduced.

[Brief description of drawings]

FIG. 1 is a diagram for explaining a conventional structure edit control method.

FIG. 2 is a system configuration diagram of a graphic system.

FIG. 3 is a diagram for explaining a structure edit control method according to the embodiment of the present invention.

FIG. 4 is a diagram illustrating details of a queue table in FIG.

FIG. 5 is a diagram for explaining the principle of synchronization processing according to the embodiment of the present invention.

FIG. 6 is a flowchart illustrating a synchronization process according to the embodiment of the present invention.

[Explanation of symbols]

 1 ... Host 2 ... OS 3 ... Socket 4,5 ... Client 6,7 ... Library 8,9 ... Structure management unit 10 ... Server 11 ... Driver 12 ... Host bus 13 ... Accelerator 14 ... Accelerator control unit 15 ... Shared memory 16 ... Server queue 17 ... CRT 18 ... Queue table 19, 20 ... Client queue 21 ... Queue management information 22 ... Request queue

Claims (3)

[Claims] 1. A data communication area on the shared memory from a client coupled to the library, comprising: a host having a library and a server; and an accelerator connected to the host and having an accelerator control unit and a shared memory. In the structure data edit control method of the graphic system in which the accelerator control unit constructs a structure by the structure edit request transferred to the client, the structure management unit is provided in the library, and the client directly by the structure management unit. A structure data edit control method, characterized in that the structure edit request is stored in a shared memory in an accelerator. 2. The accelerator control section provides a client queue dedicated to a structure edit request for each client on the shared memory so that the structure edit request can be stored independently for each client. Characterized by that
The structure data editing control method according to claim 1. 3. On the shared memory of the accelerator,
3. The structure data editing according to claim 2, wherein a server queue for storing a request other than the structure editing request is provided, and the accelerator control unit performs a synchronization process between the server queue and the client queue. control method.