JPH09305447A - Method and device for predicting job execution time, data acquiring method for table update, and table updating method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To predict the job execution time in a short time even when there is a parameter which is not proportional to the job execution time. SOLUTION: The relation between parameters (CPUX, MC, RN, and FN) which are not nearly proportional to the job execution time among parameters relating to the job execution time and a unit predicted execution time τ is stored as a table in an external storage device; and those parameters are inputted 30-32 and the unit predicted execution time τ is obtained on the basis of the values of the parameters which are not nearly proportional to the job execution time by referring to the table. All the products of the values of the parameters (PN, GN, and TN) which are nearly proportional to the job execution time among those parameters are multiplied to obtain the predicted execution time (34). A timer is actuated, data needed for job execution are inputted, and characteristic values (RN, FN, PN, ON, and TN) of data are obtained on the basis of the input data; and this job is executed, the execution result of this job is outputted, and the elapsed time of this timer is acquired as a job execution time, thereby outputting the characteristic values and execution time.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a job execution time prediction method and apparatus, a table update data acquisition method for automatically obtaining table update data used in the job execution time prediction method, and this table. The present invention relates to a table updating method for updating a table using update data.

[0002]

2. Description of the Related Art With the increase in the scale of LSI circuits, job execution times such as logic simulation and failure analysis tend to increase more and more. This job execution time is relatively long, from several days to several months. When multiple jobs with different performances are executed in parallel, for each job, the job execution time according to this performance is predicted, and by assigning hardware resources to the job based on the result, It is necessary to efficiently execute multiple jobs as a whole.

In the conventional job execution time prediction method, the total execution time of each instruction of the source program is obtained.
The branch destination of the branch instruction and the number of times the loop is repeated differ depending on the input data, which causes the job execution time to differ. Therefore, in the conventional method, it is necessary to follow the instruction according to the input data to obtain the job execution time. Therefore, it takes a long time to predict the job execution time, and the time is almost the same as the actual job execution time, which is not practical.

Therefore, a method of predicting the job execution time by obtaining the unit prediction execution time in advance and multiplying the unit prediction execution time by a parameter proportional to the job execution time such as the number of input data is used. There is.

[0005]

However, this method cannot be applied when there is a parameter that is not proportional to the job execution time. In view of such a problem, an object of the present invention is to predict a job execution time in a short time even if there is a parameter that is not proportional to the job execution time, a job execution time prediction method and apparatus, and a table update. A data acquisition method and a table updating method are provided.

[0006]

In the method for predicting job execution time according to the first aspect of the present invention, among the parameters related to the job execution time, those that are not approximately proportional to the job execution time, and the unit predicted execution time The relationship of is stored in a storage unit in advance as a table, the parameter is input, the table is referred to, and the unit predicted execution time is acquired based on the value of the parameter that is not substantially proportional to the job execution time. The predicted execution time is calculated by multiplying the unit predicted execution time by the total product of the values of the parameters that are substantially proportional to the job execution time.

According to the first aspect of the present invention, with the above configuration,
Even if there is a parameter that is not substantially proportional to the job execution time, the job execution time can be easily and quickly obtained. In the first aspect of the first invention, the parameter includes a characteristic value of input data necessary for executing the job, and hardware resource information for executing the job, and the hardware resource information indicates a job execution time. This parameter is not approximately proportional.

According to the first aspect, when a plurality of computers having different performances are provided and a plurality of jobs are executed in parallel, the job execution time according to the performance is predicted for each job, and the result is calculated. By allocating the hardware resource to the job based on the above, it is possible to efficiently execute the plurality of jobs as a whole.

In a second aspect of the first aspect of the invention, the storage means stores a first table in which the relationship between the hardware resource information and the addresses is tabulated, and is stored at the position of the address. A second table in which the relation between the value not substantially proportional to the execution time and the unit predicted execution time is tabulated is stored.

According to the second aspect, since the structure of the table is easy to understand, it is possible to easily manage the table. In the table updating data acquisition method of the second invention, a timer is started, data necessary for executing a job is input, a characteristic value of the data is acquired as the characteristic value based on the data, and the job is executed. Then, the execution result of the job is output, the elapsed time of the timer is acquired as the job execution time, and the characteristic value and the execution time are output as table update data.

According to the second aspect of the present invention, the table updating data is automatically obtained each time the job is actually executed. In the table updating method of the third invention, the table updating data and the hardware resource information related to the job execution time are input, and the execution time of the table updating data is determined by the characteristic of the table updating data. Divide by the total product of the values that are approximately proportional to the job execution time,
The quotient is obtained as a unit predicted execution time, a value that is not approximately proportional to the execution time among the input data characteristic values related to the job execution time, hardware resource information related to the job execution time, and the unit predicted execution time. The table of the relationship is updated based on the input and the unit predicted execution time.

According to the third aspect, the table is enhanced each time the job is executed, the applicable range is widened, or the unit predicted execution time is more accurate. In the job execution time prediction apparatus according to the fourth aspect of the present invention, a table storage in which the relationship between the unit execution time and a parameter that is not substantially proportional to the job execution time among the parameters related to the job execution time is tabulated in advance and stored Means, input means for inputting the parameter, and the table, obtains the unit predicted execution time based on the value of the parameter that is not substantially proportional to the job execution time, and Search for calculating the predicted execution time by multiplying the total product of the values of the parameters that are approximately proportional to the job execution time
Computing means.

According to this fourth invention, the above-mentioned first invention is implemented.

[0014]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to the drawings. FIG. 6A shows a schematic configuration of the job execution time prediction device. In this device, an input device 11, an external storage device 12, and an output device 13 are connected to a computer 10. The input device 11 is operated to input the hardware resource information and the input data characteristic value as parameters relating to the job execution time to the computer 10. A table described later is stored in the external storage device 12, and the table is referred to based on the parameter input value from the input device 11 to obtain the unit predicted execution time. The predicted job execution time is calculated using this time and output to the output device 13.

FIG. 6B shows a schematic configuration of a logic simulator as an example of a job execution time prediction target. In this logic simulator, a computer 20 is connected to external storage devices 21, 22 and an output device 23. The external storage device 21 is used as input data necessary for job execution.
The logic circuit design data is stored in, and a large number of test patterns are stored in the external storage device 22. The test pattern consists of test input data to be given to this logic circuit and an expected value of output data when this is given. The computer 20 inputs the logic circuit design data and the test pattern, performs a simulation based on the data, and outputs the result to the output device 23. In addition, based on the logic circuit design data and the test pattern, the input data characteristic value related to the job execution time is obtained, and the input data characteristic value is obtained.
Output to Further, the timer is started to obtain the actual job execution time, and this is output to the output device 23. The data characteristic value and the job execution time are stored in the external storage device 12.
Used to update the table stored in.

Next, the job execution time prediction processing by the computer 10 will be described with reference to FIGS. Hereinafter, numerical values in parentheses are step identification numbers in the figure. (30) Among the parameters related to the job execution time,
The CPU type CPUX and the main memory capacity MC are input as hardware resource information that is not substantially proportional to the job execution time.

(31) Of the parameters related to the job execution time, the number of RAMs RN included in the logic circuit to be simulated as the characteristic value of the input data of the job,
Scan campus flip-flop (scan FF)
Number FN, number of pins (number of external terminals) PN and number of gates GN,
In addition, the number TN of test patterns supplied to this logic circuit is input.

(32) Two tables, table TB1 and table TB2, are used to obtain the unit predicted execution time τ.
The table TB2 has a hierarchical structure and exists in correspondence with each set of the CPU type CPUX and the main memory capacity MC. Referring to the table TB1 shown in FIG. 2, the CPU type C
Based on the set of PUX and main storage capacity MC, the start address of the table TB2 corresponding to this set is acquired. For example, when CPUX = CPUA and MC = 512 MB, the head address A3 of the table TB2 is acquired.

(33) In the table TB2, the number of RAMs R
N, number of scan flip-flops FN, number of pins PN, number of gates GN, and number of test patterns TN (lower limit and upper limit)
And the value of the unit predicted execution time τ for a set of values of the number of RAMs RN and the number of scan FFs FN are stored. It is determined whether the characteristic value input in step 31 is within these ranges, and if it is within the range, the unit prediction execution time τ corresponding to this is based on the set of the RAM number RN and the scan FF number FN. To get. For example, when RN = 1 and FN = 2, τ = 137 ps is acquired.

The reason why the table TB2 is configured to acquire the unit predicted execution time τ based on the set of the RAM number RN and the scan FF number FN is that the RAM number RN and the scan FF number FN are both job executions. This is because the number of pins PN, the number of gates GN, and the number of test patterns TN are not substantially proportional to the time, but are substantially proportional to the job execution time.

FIG. 3 is a graph showing a part of the table TB2 of FIG. 2, and FIG. 3A shows the relationship between the number of RAMs RN and the unit predicted execution time τ when FN = 0,
(B) shows the relationship between the number FN of scan FFs and the unit prediction execution time τ when RN = 0. (34) The estimated job execution time TP is TP = τ · PN ·
Calculated as GN · TN.

(35) The calculated predicted job execution time TP is output. In this way, the predicted job execution time TP can be easily and quickly obtained even if there is a parameter that is not substantially proportional to the job execution time. The table of FIG. 2 is created based on the actual job execution time. The more data the table has, the wider the applicable range or the more accurate the estimated job execution time TP obtained. Therefore, it is preferable to automatically acquire this data each time an actual job is executed, add this data to the table, and update the table.

Next, the processing by the computer 20 of FIG. 6B for automatically acquiring this data when executing a job will be described with reference to FIG. (40) The timer TR in the computer 20 is started to measure the job execution time. (41) The design data of the logic circuit to be simulated is input from the external storage device 21, and the external storage device 2
From 2, the test pattern supplied to this logic circuit is input.

(42) Based on the data input in step 41, the RAM number RN, the scan FF number FN, the pin number PN, the gate number GN, and the test pattern number TN as input data characteristic values are acquired and stored. Keep it. (43) A logic simulation is executed based on the data input in step 41. (44) Output the result of the simulation.

(45) The contents of the timer TR started in step 40 are read out, and steps 40 to 4 are read.
The elapsed time up to 5 is acquired as the job execution time TR1. (46) The RAM number RN, the scan FF number FN, the pin number PN, the gate number GN, the test pattern number TN, and the job execution time TR1 obtained in step 45 obtained in step 42 are output as table update data.

In this way, table updating data is automatically obtained each time a job is actually executed. Next, a table updating method by the computer 10 of FIG. 6A using this data will be described based on FIG. (50) Input the table updating data output in step 46 of FIG. This input is performed by reading a file or manually.

(51) Input the CPU type CPUX and the main storage capacity MC of the computer 20 for executing the job of the logic simulation. (52) The unit predicted execution time τ is calculated as τ = TR / (PN · GN · TN).

(53) The head address of the table TB2 is acquired based on the CPU type CPUX and the main memory capacity MC, and the RAM number RN, the scan FF number FN, the pin number PN,
The table TB2 is updated based on the number of gates GN, the number of test patterns TN, and the unit predicted execution time τ. That is, when the update data is the data that has not been input in the table, this is added to the table, and when the update data is the data that has already been input in the table, it is averaged with the already input data to make the data more accurate. . Further, when the applicable range is expanded by this, the applicable range in the table TB2 is expanded.

In this way, the table is enriched for each job execution, the applicable range is widened, or the unit predicted execution time τ is more accurate. The present invention also includes various modified examples. For example, the present invention can be applied to the case where one computer processes a plurality of jobs in parallel. In this case, the hardware resource information is the job allocation partial capacity of the main storage capacity, and the parallel multiplicity is one of the parameters related to the job execution time that is not substantially proportional to the job execution time.

[Brief description of drawings]

FIG. 1 is a flowchart illustrating a job execution time prediction method.

FIG. 2 is an explanatory diagram of processing of steps 32 and 33 in FIG.

FIG. 3 is a graph showing a part of a table TB2 in FIG.

FIG. 4 is a flowchart showing a method for acquiring table updating data.

FIG. 5 is a flowchart showing a table updating method.

FIG. 6 is a block diagram showing a schematic configuration of a job execution time prediction device and a logic simulator to be predicted.

[Explanation of symbols]

 10, 20 Computer 11 Input device 13, 23 Output device 12, 21, 22 External storage device

Claims (6)

[Claims] 1. A relationship between a parameter that is not substantially proportional to the job execution time and a unit predicted execution time among parameters related to the job execution time is tabulated in advance and stored in a storage unit, and the parameter is input. Then, referring to the table, the unit predicted execution time is obtained based on the value of the parameter that is not substantially proportional to the job execution time, and the unit predicted execution time is substantially proportional to the job execution time of the parameters. A job execution time prediction method, characterized in that the predicted execution time is calculated by multiplying the total product of the product values. 2. The parameter includes a characteristic value of input data necessary for executing the job and hardware resource information for executing the job, and the hardware resource information is not substantially proportional to the job execution time. The job execution time prediction method according to claim 1, wherein: 3. A first table in which the relationship between the hardware resource information and an address is tabulated in the storage means, and is stored at the position of the address, and is not substantially proportional to the execution time among the characteristic values. The job execution time prediction method according to claim 2, further comprising: storing a second table in which a relationship between a value and the unit predicted execution time is tabulated. 4. A timer is started, data required for executing a job is input, a characteristic value of the data is acquired as the characteristic value of claim 2 based on the data, the job is executed, and the job is executed. Is output, the elapsed time of the timer is acquired as a job execution time, and the characteristic value and the execution time are output as table update data. 5. The table update data of claim 4 and hardware resource information related to job execution time are input, and the execution time of the table update data is set to the characteristic value of the table update data. Of the input data characteristic values related to the job execution time and those of the job that are not substantially proportional to the execution time. A table updating method comprising: updating a table of a relationship between hardware resource information related to execution time and the unit predicted execution time based on the input and the unit predicted execution time. 6. A table storage means in which a relation between a unit estimated execution time and a parameter which is not substantially proportional to the job execution time among parameters related to the job execution time is stored in advance in a table, and the parameter is input. Input unit for obtaining the unit predicted execution time based on the value of a parameter that is not substantially proportional to the job execution time, and the unit predicted execution time includes the job execution time among the parameters. And a calculation unit for calculating a predicted execution time by multiplying a total product of values substantially proportional to the job execution time prediction apparatus.