TWI425363B - Adjustable Priority System Resource Arbitration Method - Google Patents

Description

System resource arbitration method with adjustable priority

A system resource arbitration method with adjustable priority, in particular, by monitoring latency parameters and slice parameters, to obtain the status of the master device requesting system resources, so that system resources are fully utilized.

In the conventional computer system, when various peripherals require system resources (such as CPU, bus width, occupied address, input port, memory, etc.) for the computer system, there must be an arbitration mechanism (Arbitration) for the finite system. The resource is effectively used, as shown in the first figure, in which there are a plurality of master devices (Master), such as the first master device M1, the second master device M2, and the third master device M3, etc. When using and sharing system resources 12, an arbitration device (Arbitrator) 11 is used for timing allocation, so that each of the main control devices M1, M2, and M3 can use the system resources.

Generally, the computer system allocates the main control devices M1, M2, and M3 to share the same weight of the system resources 12, that is, W1=W2=W3, when the main control devices M1, M2, and M3 need to utilize When the system resource is 12, the request signal is sent to the computer system. Because the weights are the same, the computer system only uses the arbiter 11 to perform the sequential allocation according to the request signals, and the main control device that requests the system resources for each request is used. The ownership of the system resource 12 is obtained. The concept of U.S. Patent No. 5,987,549 is to use the above-mentioned Round-Robin method to arbitrate the method of occupying shared resources, wherein it is disclosed that each peripheral device that needs to occupy resources provides an arbitration unit to judge resource occupation. In order of priority, the resources are sequentially allocated to the peripheral devices of each request according to the priority order.

For the application of finite system resource allocation, the above-mentioned round-robin method is commonly used, as disclosed in US Pat. No. 5,072,363, which discloses an arbiter composed of various logic circuits (arbiter). ) to prioritize the occupancy of system resources.

U.S. Patent No. 5,088,024 discloses a cyclical averaging method for arbitrating access buses, whereby methods are used to arbitrate priority usage rights for limited resources in a computer system, such as resource access for arbitration buses. Please refer to the schematic diagram of the main control device and the bus bar shown in the second figure, which shows a parallel contention arbitration scheme, in which a plurality of master devices 21, 23, 25, 27 request the system using the bus bar 20. The resource has a request signal line 24 connecting the main control devices 21, 23, 25, 27 and the control signal line 26, and the arbitration data bus bar 22 connects the main control devices 21, 23, 25 in a parallel manner. 27.

When one of the above-mentioned master devices sends a request for system resource occupation by requesting the signal line 24, after the arbitration, the control signal line 26 sends a signal requesting the grant to the master device to occupy the system resource. . Under normal circumstances, resources are allocated in an equal round-robin manner (such as fair Round-Robin), so that the master devices that request resources occupying resources take turns to occupy resources; and this patent further discloses another embodiment, which is additionally added to the bus bar. Setting a signal bit as a control signal, determining the priority of the occupied resource by the identity value of each host device occupying the resource, and when any of the master devices requests the system resource, the signal is used. The bit causes the arbitration data bus 22 to allocate system resources for use; and an embodiment records the identity value of each system occupied by the master device as a reference for the next arbitration.

The above-mentioned conventional techniques for allocating system resources by means of cyclic averaging have the following disadvantages: (1) each master device uses a fixed priority order, and does not consider the amount of data of each master device; (2) frequency Wide use efficiency is not good, it can not properly use the time slot (time slot) remaining after timing distribution during data transmission; (3) because of the waste of time slot, other host devices waiting to occupy system resources generate packet delay (4) The system resources are allocated in a cyclical average manner, and there is no priority according to the importance of the master device.

A system resource arbitration method with adjustable priority is to monitor a latency parameter and a slice parameter by means of a timeout monitoring mechanism between the master device and the arbitration unit, so as to learn the master control of the system resource occupied by the request. The device's latency timeout and slice timeout can prevent the master with the highest priority from rejecting the delay or error caused by the host device with lower priority weight occupying system resources. phenomenon.

The method steps of the preferred embodiment include: monitoring a latent time signal or a piece of timeout signal, and recording a counter, and comparing the count with a threshold to determine the request to occupy system resources. The priority of the master device is heavy, and the occupation right of the system resource is allocated to the master device having the highest priority.

The preferred embodiment of the operation mechanism of the arbitration mechanism of the present invention is: first, one or more master devices make a request for use of a system resource, and monitor the master device having the highest priority, and then determine the master device. Whether there is a latent timeout, whether the master device requests to occupy the system resources, and whether the master device requesting the use of the system resource has a slice timeout, and finally according to whether the master device has a latent timeout or The slice is overdue to judge the priority of each master device, and the master device with the highest priority is allowed to occupy system resources.

The present invention provides an adaptive arbitration priority order arbitration mechanism, which solves the problem in the prior art that the limited system resources are occupied by the average allocation of the master devices, or the system required for each master device cannot be effectively considered. The problem of resources causes a problem of system resource waste and transmission packet delay.

The preferred embodiment of the present invention is applied to the use of a bus in a computer system, and the latency of each of the main control devices is considered by the adjustment of a latency parameter (Latency) and a slice of a sub-parameter (Slice). For example, the data transmission bandwidth is set, and the priority order of data transmission of each main control device is set, and the system resource sharing of the main control device that does not request the system resource is excluded, so as to adapt to various system resource occupation conditions, Effectively use and save system resources (such as bus).

The invention sets the basic priority value of each main control device at the beginning, and in the cycle of each signal processing, borrows a timeout monitor mechanism to the above-mentioned latency parameters. The slice parameters are monitored, and then the priority value of each master device is compared, and then the request for the system resource occupation by the master device having the highest priority value is agreed. The latency parameter is the number of cycles in which the master device requests to occupy system resources but is not granted. The slice parameter is that the system resources are still occupied or agreed to occupy after continuous arbitration. The number of cycles of this system resource.

The period during which the master device requesting the system resources is still not authorized to occupy the system resources is called the Latency timeout. At this time, the master device in the arbitration mechanism of the present invention has a higher priority. If the master device that has occupied the system resources continues to occupy or is granted to occupy system resources, this period is called slice timeout. In the arbitration mechanism of the present invention, the system resources are occupied. The priority of the master device is thus reduced, so that other master devices can be allowed to occupy the system resources.

By means of the above-mentioned monitoring parameters such as latency parameters and slice parameters, it is possible to prevent the master device with the highest priority from refusing the delay or error caused by the system resources with lower priority weights occupying system resources.

The third figure shows a schematic diagram of the system resource arbitration mechanism of the present invention, wherein the three main control devices M1, M2, and M3 are respectively represented as the first main control device M1 and the second main control device. The M2 and the third main control device M3 are disposed between the arbitration unit 30 and the timeout monitor, such as the first timeout monitoring unit 301 and the second timeout monitoring unit 302, which are proposed to count the signals. The third time-out monitoring unit 303 or the like, when the master control device makes an occupancy request to the system resource 32 via the arbitration unit 30, the first master device M1 in the figure as the request signal r1 sent to the arbitration unit 30, the first The timeout monitoring unit 301 and the arbitration unit 30 analyze the parameters of the above-mentioned latency parameter (L1), the slice parameter (S1) and the current priority weight (P1) of the first master device M1, so as to understand the requirements of each master device. The weight of the system resources used, such as the occupied bandwidth of the bus, the processing time of the processor, etc., by monitoring the above-mentioned latent time and the time-out of the slice to determine the priority of the master device, and then borrowing the consent signal G1, g2, g3, etc. The occupation rights of the resource 32 are allocated to the master device with the highest priority.

Each of the above-mentioned timeout monitoring units 301, 302, 303 further includes a plurality of analysis modules for important parameters such as latency parameters, slice parameters, priority weights, etc., thereby obtaining the occupancy of the system resources by the master device, as shown in the fourth figure. Schematic diagram of the time-out monitoring unit.

The figure shows the function modules and signal transmission diagrams of the time-out monitoring units 301, 302, 303, and the request signal 401 is used between the main control devices (not shown in the figure) and the arbitration unit (not shown in the figure). The consent signal 402, the sliced timeout signal 403, the latent timeout signal 404, the priority retransmission number 405 and the slice counting module 41, the latent counting module 42, the system resource usage record module 43 and the like are used as the main control device and Analysis of system resource occupancy.

For example, the slice counting module 41 is configured by the master device to perform the counting process by using the request signal 401 of the system resource and the consent signal 402 of the arbitration unit through the arbitration unit to obtain the slice parameter and the arbitration unit. A timeout signal 403 is issued; the Latency counting module 42 performs a counting procedure with the reason that the master device submits the request signal 401 using the system resource and the consent signal 402 replied by the arbitration unit via the arbitration unit, and obtains the latency. Parameters and a latent timeout signal 404 is issued to the arbitration unit.

The latency counting module 42 issues a latent timeout signal 404 to the arbitration unit when the master device requesting the system resources is still not authorized to occupy the system resources. The message is recorded by a system resource usage record module 43. The priority of the master device requesting to occupy the system resources becomes higher; if the master device that has occupied the system resources continues to occupy or is agreed to occupy the system resources, the slice counting module 41 issues a slice to the arbitration unit. The timeout signal 403 is also recorded by the system resource usage record module 43. The priority of the master device occupying the system resources is thus lowered, so that other master devices can be agreed to occupy the system resources. The system resource usage record module 43 records the request signal 401, the consent signal 402, the slice timeout signal 403 and the latent time signal 404, and finally issues an adjusted priority weight 405 to the arbitration unit, which completes the adjustment of the present invention. Priority system resource arbitration method.

The fifth figure shows a preferred embodiment of the operation flow of the arbitration mechanism of the present invention. At the beginning, the counter of the system resource usage record module is reset (reset) to zero (counter=0) (step S501). The system resource usage record module is used to record the usage status of the system resources, such as the bus bandwidth usage, the processor load, the memory usage, the network traffic, etc., and then adjust the system resource usage record module by monitoring the latent time signal. The counter is based on the latency parameter to determine whether there is a latent timeout. If the latent time signal is detected, the counter accumulates its count (counter=counter+1). If the latent timeout signal is not monitored, the counter remains unchanged. (Step S503); adjusting the counter of the system resource usage record module by using the monitoring time-sharing signal, that is, determining whether there is a time-out phenomenon according to the slice parameter, and if the slice timeout signal is detected, The counter subtracts the count amount (counter=counter-1), and if the slice timeout signal is not monitored, the counter does not change (step S505).

Then, the counted amount is compared with a threshold after each signal cycle (step S507), thereby determining the priority weight of the master device that requests the system resource to occupy (step S509), if the count quantity is greater than the threshold value, the master device that requests the system resource has the highest priority weight, and if the threshold value is not higher than the threshold value, the monitoring of the latent time and the slice timeout is repeated. And continue to be occupied by the master device that originally occupied the system resources. Finally, the system resource occupation right is allocated to the master device having the highest priority (step S511).

The preferred embodiment of the system resource allocation process of the arbitration mechanism based on the adaptive adjustment priority of the present invention is as shown in the sixth figure: initially, one or more master devices make use requests for system resources, as shown in the fourth figure. The timeout monitoring mechanism monitors the highest priority heavy master device (step S601), and the timeout monitoring mechanism provides the request signal, the consent signal and the timeout signals between the master device of the arbitration mechanism of the present invention and the system resources; When judging whether the main control device has a latent timeout (Latency timeout)? (Step S603), if the latency of the master device is delayed by the latency parameter, that is, the master device requesting the system resource is still not authorized to occupy the system resources, and the master device is in the present invention. In the arbitration mechanism, the priority of the arbitration becomes higher, and then step S607 is performed to agree that the master device occupying the system resource with the highest priority; if the latency parameter indicates that the master device does not have the latency timeout Then, continue to determine whether the master device continues to make requests? (Step S605); If the main control device does not request to use the system resource, proceed to step S609 to determine whether there are other main control devices requesting to occupy system resources.

At this time, if no other master device requests to use the system resource, the system resource usage right consent is continuously allocated to the master device having the highest priority (step S613), and conversely, if other master devices request to use the device The system resource agrees that the master device having the highest priority and requesting the system resource occupies the system resource (step S615); if the determination in step S605 is that the master device continues to make the request, then step S611 is performed, The master device that uses the system resource for this request has a sliced timeout? That is, it is known from the slice parameter that the master device that has occupied the system resources continues to occupy or is allowed to occupy system resources. In the arbitration mechanism of the present invention, the priority of the master device occupying the system resources is reduced. Allowing other master devices to be allowed to occupy the system resources; if the master device using the system resources does not have a slice timeout, then the master device currently occupying the system resources continues to be used (step S617), if any If the fragmentation timeout occurs, step S619 is performed, that is, in addition to the main control device that currently occupies the system resources, the master control device that has the highest priority and requests to occupy the system resources occupies the system resources.

In the process shown in the sixth figure, it is judged whether the main control device has latent timeout and slice timeout by using the latency parameter and the slice parameter, thereby judging whether the system resource is wasted and whether it is used for the main need to use On the control device, and according to various conditions: (1) continue to assign the system resource usage right agreement to the master device with the highest priority (step S607 or S613); (2) agree to have the highest priority and request the occupancy system The master device of the resource occupies system resources (step S615); (3) agrees that the master device currently occupying the system resources continues to use (steps S617) and (4) agrees to occupy the highest priority and requests the master device occupying system resources to occupy The system resources (step S619) and other results.

The invention monitors the latent parameters and the slice parameters by using a timeout monitoring mechanism between the master device and the arbitration unit, and can adjust the priority system resource arbitration method to make full use of the system resources and avoid the master with the highest priority. The device rejects the delay or error caused by the lower priority heavy master occupying system resources.

The above drawings are provided for reference and description only, and are not intended to limit the invention.

In summary, the present invention is an indispensable invention item, and has industrial applicability, novelty and progress, and fully meets the requirements for invention patent application, and submits an application according to law, so please check and give permission. The patent in this case is to protect the rights and interests of the inventor.

However, the above description is only a preferred embodiment of the present invention, and thus the scope of the present invention is not limited thereto, and equivalent structural changes made by using the present specification and the illustrated contents are equally included in the present invention. Within the scope of the agreement, Chen Ming.

The first master device. . . M1

Second master control device. . . M2

The third master control device. . . M3

Weights. . . W1, W2, W3

Arbitrator. . . 11

system resource. . . 12

Main control device. . . 21,23,25,27

Busbars. . . 20

Request signal line. . . twenty four

Control signal line. . . 26

Arbitration data bus. . . twenty two

First timeout monitoring unit. . . 301

Second timeout monitoring unit. . . 302

Third timeout monitoring unit. . . 303

Arbitration unit. . . 30

system resource. . . 32

Request signal. . . R1,r2,r3

Agree the signal. . . G1, g2, g3

Latent parameters. . . L1

Slice parameters. . . S1

Priority is heavy. . . P1

Request signal. . . 401

Agree the signal. . . 402

The film is divided into time-out signals. . . 403

Latent overtime signal. . . 404

Priority resignation number. . . 405

Slice counting module. . . 41

Latent counting module. . . 42

System resource usage record module. . . 43

The first diagram is a schematic diagram of the use of system resources by the arbitration mechanism in the conventional technology; the second diagram is the schematic diagram of the usage of the master control device and the busbar of the conventional technology; the third diagram is the system resource of the arbitration mechanism of the prioritized order of the present invention. Schematic diagram of allocation; the fourth figure is a schematic diagram of the internal function module and signal transmission of the overtime monitoring unit in each main control device of the present invention; the fifth figure is a flow chart of the timeout monitoring mechanism of the present invention; A flow chart of a preferred embodiment of the system resource arbitration method of the present invention.

Claims (29)

A system resource arbitration method capable of adjusting priorities, wherein one or more master devices occupy a system by using a latency parameter and a number of cycles that continue to occupy system resources or are allowed to occupy the system resources after arbitration The resource situation, in order to achieve the arbitration method, the steps include: monitoring a latent timeout signal, and recording a count amount, determining whether there is a latent timeout according to the latency parameter; monitoring a minute timeout signal, and Recording the count amount according to the number of cycles of continuing to occupy system resources or being allowed to occupy the system resources after arbitration, determining whether there is a fragmentation timeout; comparing the count amount with a threshold value; determining the request One of the master devices occupying the system resources has a priority; and the occupation rights of the system resources are allocated to the master device having the highest priority. The system resource arbitration method of the prioritized priority according to claim 1, wherein the count value is reset before the counting amount is calculated. The system resource arbitration method of the prioritized priority according to claim 1, wherein the counting amount is accumulated when the latent time signal is detected. The system resource arbitration method of the prioritized order of claim 1, wherein the count is not monitored when the latent time signal is not monitored. change. The system resource arbitration method of the prioritized priority according to claim 1, wherein the count amount is decreased when the slice timeout signal is detected. For example, the system resource arbitration method of the prioritized priority described in claim 1 wherein the count amount does not change when the slice timeout signal is not monitored. The system resource arbitration method of the prioritized priority according to claim 1, wherein the master device that uses the system resource has the highest priority weight when the count is greater than the threshold. The system resource arbitration method of the prioritized priority according to claim 1, wherein if the count is not higher than the threshold, the monitoring of the latent time and the slice timeout is repeated, and It continues to be occupied by the master device that originally occupied the system resources. A system resource arbitration method capable of adjusting priorities according to a latency parameter and a number of loop cycles that continue to occupy system resources or are allowed to occupy the system resources after arbitration, and determine that one or more master devices occupy a system The resource situation, in order to achieve the arbitration method, the steps include: one or a plurality of master devices requesting use of the system resource; monitoring the highest priority heavy master device; using the latency parameter to determine whether the master device is There is a latent timeout; it is judged whether the master device requests to occupy the system resources; if the system continues to occupy the system resources after being arbitrated or is allowed to occupy the system resources The number of cycles of the source determines whether the master device that requests the system resource has a slice timeout; and the master device that has the highest priority weight occupies the system resource. The system resource arbitration method of the prioritized priority according to claim 9, wherein the monitoring of the highest priority heavy control device, wherein the monitored signal comprises between the main control device and the system resource A request signal, a consent signal, a latent time signal, a time-out signal and a priority re-signing number. The system resource arbitration method of the prioritized priority according to claim 9, wherein in the step of determining whether the master device has a latency occurrence, if the latency parameter is known to be latent by the master device When it is agreed that the master device has the highest priority, the system resources are occupied. The system resource arbitration method of the prioritized priority according to claim 11, wherein the master device that has occupied the system resource is still not approved to occupy the time when the host device has a latent timeout System resources, and their priority will become higher. The system resource arbitration method of the prioritized priority according to claim 9, wherein when the main control device does not have a latency, the determining whether the master device requests to occupy the system resource step. The system resource arbitration method of the prioritized priority according to claim 9, wherein if the master device does not request to use the system resource, then it is determined whether another host device requests to use the system resource. The system resource arbitration method of the prioritized priority according to claim 14, wherein if no other master device requests to occupy the system resource, the system resource continues to be occupied by the master device. The system resource arbitration method of the prioritized priority according to claim 14, wherein if another master device requests to occupy the system resource, the master device having the highest priority and requesting to occupy the system resource occupies the system resource. The system resource arbitration method of the prioritized priority according to claim 9, wherein if the master device does not have a slice timeout, the master device continues to occupy or is agreed to occupy the system resource. The system resource arbitration method of the prioritized priority according to claim 9, wherein if the master device has a slice timeout, in addition to the master device currently occupying the system resource, agree to have the highest priority and The master device requesting to occupy the system resources occupies the system resources. For example, the system resource arbitration method of the prioritized order described in claim 18, wherein the priority of the master device occupying the system resource is reduced, so that other master devices can be agreed to occupy the system resources. The system resource arbitration method of the prioritized priority according to claim 9, wherein if the master device that uses the system resource does not have a slice timeout, then the master that currently occupies the system resource is agreed. The device continues to be used. A system resource arbitration method capable of adjusting priorities according to a latency parameter and a system that continues to occupy system resources after being arbitrated or is allowed to occupy the system The number of cycles of the resource, determining whether one or a plurality of master devices occupy a system resource, thereby achieving the arbitration method, the steps comprising: one or a plurality of master devices requesting use of the system resource; monitoring The highest priority heavy master device; determining, by the latency parameter, whether the master device has a latent timeout; determining whether the master device requests to occupy the system resource; determining whether another master device requests to occupy the system resource And agree that the master device with the highest priority and requesting to occupy the system resources occupies the system resources. The system resource arbitration method of the prioritized priority according to claim 21, wherein the step of monitoring the highest priority heavy control device, wherein the monitored signal comprises between the main control device and the system resource A request signal, a consent signal, a latent time signal, a time-out signal and a priority re-signing number. The system resource arbitration method for adjusting the priority according to claim 21, wherein in the step of determining whether the master device has a latency occurrence, if the latency parameter is known to be latent by the host device When it is agreed that the master device has the highest priority, the system resources are occupied. The system resource arbitration method of the prioritized priority according to claim 23, wherein when the host device has a latent timeout, the master device requesting to occupy the system resource is still not agreed to occupy the System resources, and their priority will become higher. Systematic resources for adjustable prioritization as described in claim 21 The source arbitration method, wherein when the master device does not have a latency excess, the step of determining whether the master device requests to occupy the system resource is performed. The system resource arbitration method of the prioritized priority according to claim 21, wherein if the master device requests to use the system resource, then it is determined whether the master device that requests the system resource uses a slice. Time. The system resource arbitration method of the prioritized priority according to claim 26, wherein if the master device does not have a slice timeout, the master device continues to occupy or is agreed to occupy the system resource. The system resource arbitration method of the prioritized priority according to claim 26, wherein if the master device has a slice timeout, in addition to the master device currently occupying the system resource, agree to have the highest priority and The master device requesting to occupy the system resources occupies the system resources. The system resource arbitration method of the prioritized priority according to claim 21, wherein if it is determined that no other master device requests to occupy the system resource, the system resource usage right consent is further allocated to the highest priority. The weight of the master device.