JP2002304367A - Arbiter equipment and method for arbitration - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an arbiter equipment capable of preventing a processing speed form being lowered, even if the number of clients increases when arbitrating requests for one resource from plural clients, and a method for arbitration. SOLUTION: Clients 2A-2H are disposed in the lowest hierarchy of a tree structure, and a sub-arbiter to arbitrate between two clients in the hierarchy or between sub-arbiters is disposed in the tree structure. If request signal is generated from the clients 2A-2H, the request signal is sent to a sub-arbiter 1G in the highest hierarchy as it is being arbitrated by an arbiter in each hierarchy. A grant signal is sent from the sub-arbiter in the highest hierarchy through the similar passage. Thereby, the number of times to judge conditions is prevented from increasing, even if the number of clients increases.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an arbiter apparatus and a method for arbitrating access when one client shares one resource.

[0002]

2. Description of the Related Art When a single resource is shared by a plurality of clients, an arbiter is used to arbitrate access from the plurality of clients to the resource.

For example, MPEG (Moving Picture Codin)
g Experts Group) 2, a video stream and an audio packet are separated from the transport stream, and the decoded video data is subjected to image processing and OSD (OnS
(Creen Display) An image processing apparatus for superimposing an image has been developed.

[0004] Such image processing apparatuses include MPEG2
A transport stream processor for separating TS (Transport Stream) packets, a graphics engine for performing various image processing, and a display processor for generating and superimposing an OSD screen and enlarging and reducing the screen Etc. are arranged. All of these processors require a large-capacity memory.

However, if a memory is provided for each processor, the circuit scale becomes enormous. Therefore, in such an image processing apparatus, one memory is shared by each processor.

In such a case, processors such as a transport stream processor, a graphics engine, and a display processor correspond to a plurality of clients, and a memory corresponds to one resource.

In such an image processing apparatus, 1
Access to a single memory may occur simultaneously from multiple processors. The arbiter arbitrates access requests from a plurality of clients for such one resource.

Various arbiter algorithms have been proposed, but a token ring method is often used.

FIG. 5 is a block diagram of a token ring type arbiter. In FIG. 5, the arbiter
Units 101A and 101 corresponding to each client
B, 101C,... Are provided. When each client requests access to a resource, the request signal req
Issue

Each unit 101A, 101B, 101
C, the token is running in between. Only the units 101A, 101B, 101C,... Having the token can issue the grant signal. Each unit 101A,
If there is no request, 101B, 101C,... Pass the token to the next unit. The tokens are sent in the order of the units 101A, 101B, 101C,...
Is returned to any of the units 101A to 101E. Unit 1 with the token
Only when a request signal is received at 01A, 101B, 101C,..., A grant signal is issued from that unit, and the client is granted access to the resource.

[0011]

However, in such a token ring system, it is necessary to perform the same number of condition determinations as the client in order to perform one arbitration.

That is, in the token ring system, FIG.
As shown in the flowchart in FIG.
01A, it is determined whether there is a request signal req_a for the unit 101A. If there is a request signal req_a, a grant signal gnt_a is output. If there is no request signal req_a, a token is output to the unit 10A.
1B, a request signal for the unit 101B
It is determined whether or not there is req_b, and if there is a request signal req_b, a grant signal gnt b is output and the request signal
If there is no req_b, the process of transferring the token to the unit 101C is repeated.

Therefore, one arbitration requires a number of condition judgments corresponding to the units 101A to 101E.

Therefore, when the token ring system is used, as the number of clients increases, the time required for one arbitration increases, causing a problem that the operation speed decreases in proportion to the number of clients.

Therefore, an object of the present invention is to arbitrate requests from a plurality of clients for one resource, and to prevent a decrease in processing speed even if the number of clients increases, in an arbiter apparatus and method. Is to provide.

[0016]

SUMMARY OF THE INVENTION According to the present invention, when a plurality of clients share one resource, a request issued from each client is arbitrated, and use of the resource is permitted among the plurality of clients. Arbiter device that grants permission to a given client, arbitrates a plurality of clients arranged in the lowest layer of the tree structure and requests from subarbiters or clients in the lower layer of the tree structure It has multiple sub-arbiters in a tree structure, and receives requests from clients,
While arbitrating with the sub arbiter arranged in the upper layer, the request tracings to the sub arbiter at the top layer of the tree structure, and the sub arbiter at the top layer is sent along the sub arbiter of the tree structure Arbiter device, wherein upon receiving the request, the permission is returned to the client permitted to use the resource on the same route as the request has followed to the subarbiter of the highest layer.

According to the present invention, when a plurality of clients share one resource, a request generated from each client is arbitrated, and a client to which use of a resource is permitted among a plurality of clients is provided. In the arbiter method for granting permission, a plurality of clients are arranged at the lowest layer of the tree structure, and a plurality of sub-arbiters at the lower layer of the tree structure or a plurality of sub-arbiters for arbitrating a request signal from the client are arranged in a tree. In the tree structure, the request from the client is arbitrated by the sub-arbiter arranged in the upper layer, it is traced to the sub-arbiter in the top layer of the tree structure, and the sub-arbiter in the top layer is When a request is received following a subarbiter, the request signal is By the same route as that followed up to the arbiter, a arbiter method characterized by use of the resource is to return the permission to clients who are authorized.

A subarbiter that arbitrates requests from two clients or subarbiters is arranged in a tree to arbitrate requests from a plurality of clients. In this way, assuming that the number of clients is n, the number s of condition determinations is s = 2 × log ₂ n, and even if the number of clients increases, the number of condition determinations can be prevented from increasing.

[0019]

Embodiments of the present invention will be described below with reference to the drawings. In the present invention, as shown in FIG. 1, the sub-arbiters 1A to 1G are arranged in a tree to constitute an arbiter for a plurality of clients 2A to 2H. Sub arbiter 1A-1
G is mediating two subarbiters or clients in the hierarchy below it. In this way, by arranging two subarbiters 1A to 1G in the lower hierarchy or subarbiters 1A to 1G for arbitrating clients in a tree shape, two requests are selected by the subarbiters 1A to 1G in each hierarchy. The requests from the plurality of clients 2A to 2H at the lowest layer of the tree structure are finally selected by the subarbiter 1G at the highest layer of the tree structure, and the plurality of clients 2A to 2H are selected. 2A
Arbitration of ~ 2H is made.

FIG. 2 shows, for example, the structure of a subarbiter 1A in which clients 2A and 2B are provided in a lower layer.

In FIG. 2, the sub-arbiter 1A includes a token ring circuit 11 and a selector 12.
The clients 2A and 2B output request signals req_a and req_b, respectively, when requesting access to resources. The request signals req_a and req_b from the clients 2A and 2B are transmitted from the request signal input terminals 13A and 13B to the sub-arbiter 1A, respectively.
And supplied to the token ring circuit 11.

The token ring circuit 11 receives request signals req_a and req_a from the clients 2A and 2B.
b is arbitrated by the token ring method, and when request signals req_a and req_b come simultaneously from the clients 2A and 2B, one request signal req_a or req_b is selected, and the selected request signal req_a or req_b is sent to the request signal output terminal 14. And sends it to the subarbiter 1E in the upper layer.

The token ring circuit 11 also receives request signals req_a and req_a from the clients 2A and 2B.
A select signal sel indicating which request is selected among q_b is output. This select signal sel is
It is supplied to the selector 12.

A grant signal gnt is sent from the sub-arbiter 1E located at a higher layer than the sub-arbiter 1A. The grant signal gnt is supplied to the grant signal input terminal 1
5 is input to the sub arbiter 1A and supplied to the selector 12.

The selector 12 includes a token ring circuit 11
Is switched based on the select signal sel from the client 2 so that the grant signal gnt is output to the selected one of the clients 2A and 2B in the lower layer. Grant signal from selector 12
gnt is supplied to the grant signal output terminal 16A or 16B, and is supplied to the client 2A or 2B in the lower layer.

FIG. 3 shows the configuration of the token ring circuit 11. As shown in FIG. 3, the token ring circuit 11 includes units 21A and 21B.
Units 21A and 21B correspond to the two units or clients in the lower hierarchy. In this example, the client 2A is located below the subarbiter 1A.
And 2B, the unit 21A is connected to the client 2
A corresponds to A, and the unit 21B corresponds to the client 2B.

The unit 21A is supplied with a request signal req_a from the client 2A from the request signal input terminal 20A, and the unit 21B is supplied with a request signal req_b from the client 2B from the request signal input terminal 20B. Is done.

A token runs between the two units 21A and 21B, and the unit 2 having the token
Only 1A or 21B can issue a request signal to a higher layer. If there is no request, units 21A and 21B pass the token to the other unit, and either unit 21A or 21B has the token.

For example, when a token is coming to the unit 21A, the client 2A
When the request signal req_a is supplied from the
1A outputs a request signal. Otherwise,
No request signal is output from unit 21A.
When a request signal req_b is supplied from the client 2B to the unit 21B while a token is coming to the unit 21B, the request signal is output from the unit 21B. Otherwise, unit 21B
Does not output a request signal.

The request signals from the units 21A and 21B are supplied to the OR gate 22 and also to the select signal generation circuit 23. The output of the OR gate 22 is output from the request signal output terminal 25.

The select signal generation circuit 23 includes the unit 2
It generates a select signal for determining which unit is selected from 1A and 21B.
From the select signal generation circuit 23, for example, the unit 2
“1” is output when 1B is selected, and “0” is output otherwise (ie, when the unit 21A is selected). This select signal generation circuit 23
Is output from the select signal output terminal 24.

As described above, the token ring circuit 11
From the request signals req_a and req_b from the two lower clients 2A and 2B, one request signal req is selected by a token ring method,
The select signal sel indicating the selected client is output. If the operation at this time is expressed by a logical expression, req = req_a | req_b; if (b) sel = 1; else sel = 0;

As shown in FIG. 1, an arbiter is constructed by arranging the sub-arbiters 1A to 1G having such a configuration in a tree shape. In the lower layer of each sub-arbiter 1A-1G, there are two clients or sub-arbiters, and each sub-arbiter 1A-1G arbitrates a request signal from the two clients or sub-arbiters in the lower layer, and When request signals from two lower clients or sub arbiters are generated at the same time, the request signal of one client or sub arbiter is selected by the token ring method. When such sub-arbiters 1A to 1G are arranged in a tree shape, one request is selected from two requests in each layer, and the sub-arbiter 1 in the highest layer is selected.
In G, a request for one client will be selected.

The grant signal from the sub-arbiter 1G in the uppermost layer is sent to the lower layer by tracing information (select signal sel) on which sub-arbiter 1A to 1G has selected which request signal. It is sent to one of the clients 2A to 2H in the lowest layer.

For example, assume that the client 2C generates a request signal req_c to request a resource.

The sub-arbiter 1B is a client 2C
Arbitration between the request from the client 2D and the request from the client 2D. When the client 2C is selected, the request signal req_c from the client 2C is output from the sub-arbiter 1B.

The sub arbiter 1E is the sub arbiter 1
A request from A and a request from the sub-arbiter 1B are arbitrated, and when the sub-arbiter 1B is selected, a request signal req_ from the client 2C is received.
c is output from the sub-arbiter 1E.

Subarbiter 1 at the top of the tree structure
G arbitrates between the request from the sub-arbiter 1E and the request from the sub-arbiter 1F. When the sub-arbiter 1E is selected, the request signal req_c from the client 2C is output from the sub-arbiter 1G.

Subarbiter 1 at the top of the tree structure
Upon selecting one request, G sends a grant signal gnt to the sub-arbiter 1E on the side that has selected the request. The grant signal gnt is received by the sub-arbiter 1E.

The sub arbiter 1E is a sub arbiter 1
When the grant signal gnt is received from G, the grant signal gnt is sent to the sub-arbiter 1B on the side that has selected the request. The grant signal gnt is received by the sub-arbiter 1B.

The sub arbiter 1B is the sub arbiter 1
When the grant signal gnt is received from E, the grant signal gnt is sent to the client 2C that has selected the request.

The client 2C can access the resource by receiving the grant signal gnt.

As described above, in the embodiment of the present invention,
Subarbiters 1A to 1G that arbitrate requests from two clients or subarbiters are located in a tree shape, and arbitrate requests from a plurality of clients 2A to 2H. In this example, the number of clients is eight. At this time, considering the number of condition judgments until the grant signal is finally output, the number of condition judgments performed by one sub-arbiter 1A to 1G is 2. If the number of clients is 8, this will be stacked in three stages,
As a result, the condition determination is repeated six times.

On the other hand, in the conventional token ring arbiter shown in FIG. 5, when the number of clients is 8, the number of condition judgments is eight. Therefore, in the embodiment of the present invention, the number of condition judgments is reduced,
The operation speed can be improved.

That is, in the embodiment of the present invention, assuming that the number of clients is n, the number s of condition determinations is s = 2 × log ₂ n. On the other hand, the number s of condition determinations in the case of the conventional token ring system shown in FIG. 5 is s = n. Therefore, in the embodiment of the present invention, if the number n of clients is larger than 4, the number s of condition determinations can be reduced as compared with the case of the conventional token ring method, and the operation speed can be increased.

This arbiter is suitable for arbitrating a memory in an image processing apparatus as shown in FIG.

FIG. 4 shows an example of an image processing apparatus suitable for reproducing digital satellite broadcasts. In FIG. 4, the stream processor 51 separates a TS packet from an MPEG2 TS stream by referring to a PID. The video packet separated by the stream processor 51 is sent to a video decode processor 52, and the audio packet is
Sent to 3. The information packet is transmitted to the main processor 54.
Sent to

The video decode processor 52 has an MPE
Video data compressed by the G2 method is decoded. The audio decode processor 53 has an AAC
(Advanced Audio Coding) The audio data compressed by the method or the like is decoded.

The graphics engine 55 performs graphics processing based on graphics data.

The display processor 56 performs processing such as OSD screen processing, image enlargement and deformation, and the like.

SDTV (Standard Definition Televisi)
on) The encoder 57 performs image size conversion and the like on video data sent in various image formats, for example, in the NTSC format. HDTV (Hi
gh Definition Television) encoder 58 converts the video data transmitted in various image formats into, for example, an image size so as to be in the NTSC format.

The stream processor 51, the video decode processor 52, and the
The audio decode processor 53, the graphics engine 55, the display processor 56, the SDTV encoder 57, and the HDTV encoder 58 all require a large-capacity high-speed memory for processing. A memory 50 is provided as a common memory for these processors. As the memory 50, for example, an SDRAM (Sy
nchronous Dynamic Random Access Memory) is used.

Therefore, in this case, the stream processor 51, the video decode processor 52, the audio decode processor 53, and the graphics engine 5
5. The display processor 56, the SDTV encoder 57, and the HDTV encoder 58 become clients,
The memory 50 is a common resource. An arbiter 59 is provided to arbitrate requests from each client.

That is, the stream processor 51, video decode processor 52, audio decode processor 53, graphics engine 55, display processor 56, SDTV encoder 57, HDTV
In the encoder 58, when accessing the memory 50,
A request signal is output. This request signal is
It is supplied to the arbiter 59.

The arbiter 59 includes a stream processor 51, a video decode processor 52, an audio decode processor 53, a graphics engine 55, a display processor 56, an SDTV encoder 57,
Which of the requests from the HDTV encoder 58 is permitted to access the memory 50 is determined. As the arbiter 59, one having a structure in which subarbiters are arranged in a tree shape as shown in FIG. 1 can be used. It is determined by tracing the tree structure from two of the arbiter or the client. Then, a grant signal is given to one of the stream processor 51, the video decode processor 52, the audio decode processor 53, the graphics engine 55, the display processor 56, the SDTV encoder 57, and the HDTV encoder 58 to which access is permitted.

Of course, the present invention can be widely used as an arbiter when a single resource is shared by a plurality of clients. Furthermore, multiple computers can be connected to a LAN (Local Ar
ea Network).

[0057]

According to the present invention, requests from a plurality of clients are arbitrated by arranging subarbiters for arbitrating requests from two clients or subarbiters in a tree shape. In this way, assuming that the number of clients is n, the number of conditional decisions s is s = 2 × log ₂ n, and even if the number of clients increases, the number of conditional decisions can be prevented from increasing.

[Brief description of the drawings]

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

FIG. 2 is a block diagram showing a configuration of a sub-arbiter to which the present invention is applied.

FIG. 3 is a block diagram showing a configuration of a token ring circuit in a sub-arbiter to which the present invention has been applied.

FIG. 4 is a block diagram of an example of an image processing apparatus to which the present invention can be applied.

FIG. 5 is a block diagram used for explaining a conventional arbiter.

FIG. 6 is a flowchart used to explain a conventional arbiter.

[Explanation of symbols]

1A to 1G: subarbiter, 2A to 2H: client, 11: token ring circuit, 12 ...
·selector

Continuation of the front page F term (reference) 5B045 AA01 EE03 EE20 5B060 CD11 5B061 BA01 BB27 BB36 GG02 5C064 BA01 BC06 BC14 BC16 BD02 BD08

Claims (8)

[Claims] 1. A method for arbitrating requests generated from each client when one client is shared by a plurality of clients, and for a client permitted to use the resource among the plurality of clients. Arbiter device that grants permission to a client, a plurality of clients arranged in the lowest layer of the tree structure, and a plurality of subarbiters in the lower layer of the tree structure or a plurality of sub-arbiters of a tree structure for arbitrating requests from clients. Arbiter, and arbitrates a request from the client with a sub-arbiter arranged in an upper layer thereof, traces the request to a sub-arbiter in the uppermost layer of the tree structure, and When the arbiter receives the request sent following the sub-arbiter in the tree structure, the arbiter Doo is the same route as that followed until Sabuabita of the uppermost layer, the arbiter unit being characterized in that to return the permission to clients who use the resource is permitted. 2. The arbiter apparatus according to claim 1, wherein the sub-arbiter arbitrates requests from two clients or sub-arbiters in lower layers. 3. The sub-arbiter further comprises a selection means for transmitting a permission sent from a sub-arbiter in an upper layer to a selected one of sub-arbiters or clients in a lower layer. An arbiter apparatus according to claim 1. 4. The arbiter apparatus according to claim 1, wherein the sub-arbiter arbitrates a request from a client or a sub-arbiter in a lower layer by a token ring method. 5. A method for arbitrating a request issued from each client when a plurality of clients share one resource, and for a client permitted to use the resource among the plurality of clients. In the arbiter method, a plurality of clients are arranged in the lowest layer of the tree structure, and a plurality of sub-arbiters in the lower layer of the tree structure or a plurality of sub-arbiters for arbitrating a request signal from the client are arranged in a tree. In a tree structure, the request from the client is arbitrated by the sub-arbiter arranged in the upper layer, and traced to the sub-arbiter in the top layer of the tree structure, and the sub-arbiter in the top layer is arbitrated. Receives the request sent by following the tree-like sub-arbiter, Arbiter how signal is characterized in that to return the permission by the same route as that followed until Sabuabita of the uppermost layer, to client that is allowed to use the resource. 6. The arbiter method according to claim 5, wherein the sub-arbiter arbitrates requests from two clients or sub-arbiters in lower layers. 7. The sub-arbiter performs a selection to send a permission sent from a sub-arbiter in a higher layer to a selected one of sub-arbiters or clients in a lower layer. 6. The arbiter method according to 1. 8. The arbiter method according to claim 5, wherein the sub-arbiter arbitrates a request from a client or a sub-arbiter in a lower layer by a token ring method.