JP5084094B2 - Data processing time prediction method, data processing apparatus, and program - Google Patents

Description

  The present invention relates to a data processing apparatus that performs data processing for developing and synthesizing pattern data such as LSI mask patterns, and a method for predicting the data processing time.

  In recent years, circuit patterns formed on LSIs and the like have been miniaturized, and the data amount of the pattern data has been steadily increasing.

  Usually, when pattern data is recorded (saved) on a recording medium, it is often recorded using a predetermined compression format in order to reduce the data amount. For example, for an area including the same pattern among all the patterns, the same pattern (this is referred to as “repeated figure”) is registered and used repeatedly, which is necessary for actual drawing. Some compress the amount of data. In this case, the pattern of the area including the same pattern can be formed by the data indicating the shape of the “repeated figure” and the “coordinate data” indicating the arrangement position of the repeated figure. The pattern in the other area is recorded as a unique shape pattern (this is referred to as “single figure”). The pattern of the area recorded by the shape data and the “coordinate data” of the “repeated figure” is formed by developing and arranging the “repeated figure” based on the “coordinate data”.

  In creating an LSI mask pattern or the like, the pattern data is often designed for each block such as a CPU section or a memory section. That is, pattern data may be designed by different manufacturers for each block, or data may be registered in different compression formats. Even in such a case, final pattern data is generally created by developing and synthesizing layout data prepared for each block.

  By the way, data processing for developing and synthesizing data may take several hours to several tens of hours depending on the type and format of the data. In the future, it is expected that there will be many cases where the processing time is, for example, 100 hours or more due to miniaturization of circuit patterns. In such data processing over a long period of time, it is important to make a work plan to predict the processing time more accurately.

  Conventionally, a simple proportional calculation formula has often been used to predict such processing time. For example, when processing data of a predetermined size, if the time for processing 10% of the total size of the data to be processed is A time, the remaining processing time is simply predicted based on the size ratio. , Predicted to be 9 times the A time.

Further, as another example of the processing time prediction method, a technique has been proposed in which an execution history of data processing executed in the past is made into a database and more accurate prediction is performed based on data in the database (for example, (See Patent Document 1). According to this document, for example, in data processing in which a plurality of data items are processed in one data processing, the number of data processing items for each data processing and the total processing time required for the data processing are registered in advance. A technique for predicting the processing time by calculating the data processing time per unit as (total processing time / total number of data) and multiplying this by the number of data processing of the current data processing has been proposed.
JP-A-9-305556

  However, with the conventional prediction method, the processing time may not be accurately predicted. That is, in the first example, if data processing that takes a relatively long time is concentrated in 10% of all data processing, the processing time per unit size to be calculated becomes long. As a result, the overall prediction processing time is estimated to be long. Conversely, if the processing time per unit size is estimated to be short, the overall prediction processing time is also estimated to be short. In such a prediction method, generally, the predicted processing time varies greatly in the first half of the data processing, and the prediction accuracy often increases only in the second half. However, when making a work plan, it is desirable to obtain an accurate prediction processing time immediately after the start of data processing, and even if the accuracy is improved in the second half of the data processing, it is hardly useful.

  SUMMARY OF THE INVENTION An object of the present invention is to provide a data processing time prediction method and a data processing apparatus that accurately predict data processing time when creating pattern data such as LSI mask patterns.

In order to achieve the above object, a data processing time prediction method according to the present invention is a data processing time prediction method in data processing for creating pattern data based on layout data, and the data processing time is calculated from the layout data. A step of reading numerical information for predicting time, and a multiple regression equation for predicting the data processing time per unit file size are read, and the numerical information read from the layout data is substituted into the multiple regression equation Calculating the data processing time per unit file size, and multiplying the data processing time per unit file size by the sum of the file sizes stored in the layout data to substantially reduce the data processing time. and a step of calculating manner.

  According to the present invention, the data processing time is calculated by substituting the numerical information read to predict the data processing time into the correlation equation (multiple regression equation) created by performing multiple regression analysis in advance. Therefore, the prediction accuracy of the data processing time is improved as compared with that predicted using a conventional simple proportional calculation formula.

Here, the layout data has a format including single figure data, coordinate data indicating the arrangement position of the single figure, data of the repeating figure, and coordinate data indicating the arrangement position of the repeating figure. In the step of reading the numerical information, the data read as numerical information includes the number T _{S of} single figures, the number T _r of repeated figures after expansion, the file size S _s of the data of all the single figures in the layout data, and , The file size S _r of all the repeated figure data after the expansion in the layout data. In this case, the multiple regression equation is used as a variable of (T _S / S _s ) and (T _r / S _r ). It was created.

  The above-described method for predicting the data processing time according to the present invention includes a step of measuring an actual time required for data processing performed in the past, the data of the actual time, and the data performed in the past. And a step of performing a multiple regression analysis using the numerical information used in the processing for predicting the processing time of the data processing and re-determining coefficients necessary for determining a multiple regression equation. It may be.

  In this way, by using the actual time taken for data processing performed in the past and the numerical information of the layout data used in the data processing, multiple regression analysis is performed and the coefficient of the multiple regression equation is updated. The accuracy of the coefficient is further improved.

The data processing apparatus of the present invention is a data processing apparatus that performs data processing for creating pattern data based on layout data,
A data input unit that is connected to the layout data and reads numerical information for predicting a data processing time in the data processing from the layout data;
Recording means for pre-stored multiple regression equations for predicting the data processing time per unit file size ;
The data processing time per unit file size is calculated by substituting the numerical information read from the layout data into the multiple regression equation, and stored in the layout data at the data processing time per unit file size. and that by multiplying the total file size, provided with a processing time calculation unit which substantially calculates the data processing time,
The layout data is a format including single figure data, coordinate data indicating the position of the single figure, data of the repeated figure, and coordinate data indicating the position of the repeated figure,
Wherein the data input unit, the number T _S of the sole figure, the number T _r of the repeating graphic after expanding the file size S _s of the data of all the single figure in the layout data, and, in the layout data Read the file size S _r of all the repeated figure data after expansion ,
The processing time calculation unit uses the multiple regression equation with (T _S / S _s ) and (T _r / S _r ) as variables when calculating the data processing time per unit file size. Features.
According to the data processing apparatus having such a configuration, the prediction method of the data processing time by the present invention are well performed. In addition, the data processing apparatus may include a display unit that displays the predicted data processing time.

The program of the present invention also includes a step of reading numerical information for predicting data processing time in data processing for creating pattern data from layout data, and an overlap for predicting the data processing time per unit file size. The regression processing is read, the data processing time per unit file size is calculated by substituting the numerical information read from the layout data into the multiple regression formula, and the data processing time per unit file size is calculated as the data processing time. A program for causing a computer to execute a process including a step of substantially calculating the data processing time by multiplying a total of file sizes stored in layout data ,
The layout data is a format including single figure data, coordinate data indicating the position of the single figure, data of the repeated figure, and coordinate data indicating the position of the repeated figure,
The step of reading the numerical information includes the number T _{S of the} single figure, the number T _{r of the} repeated figure after expansion, the file size S _s of the data of all the single figures in the layout data, and the layout data Including reading the file size S _r of the data of all repeated figures after expansion in
The step of substantially calculating the data processing time includes the step of calculating the data processing time per unit file size by using (T _S / S _s ) and (T _r / S _r ) as variables. Including using regression equations.

  According to the data processing time prediction method and data processing apparatus of the present invention, as described above, since the data processing time is predicted using the multiple regression equation, the data processing time can be accurately predicted. In addition, if the information obtained in the past data processing is used and multiple regression analysis is performed to update the coefficient of the multiple regression equation, the accuracy of the multiple regression equation coefficient will be further improved. In addition, the data processing time can be predicted more accurately.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram showing a configuration of a data processing apparatus according to an embodiment of the present invention.

  A data processing apparatus 1 shown in FIG. 1 is provided as a part of an electron beam drawing apparatus that draws an electron beam on a mask to form a circuit pattern (mask pattern), for example, and includes a data input unit 2 that reads data. Based on the data read / synthesized by the data input / output unit 2 and the data development / synthesizer 6 for creating final drawing data for electron beam drawing by developing / combining the data read by the data input unit 2. And a processing time calculation unit 4 for predicting the data processing time. The illustrated state shows a state in which a plurality of layout data A, B, and C prepared as circuit data for the CPU unit and the memory unit are connected to the data input unit 2, for example.

  Each of the layout data A to C includes “repetitive graphic” shape data, “coordinate data” for arranging the repetitive graphic, and “single graphic” shape data. The layout data A to C include information on the number of “repeated figures” after expansion, information on the number of “single figures”, and data on each of “repeated figures” and “single figures”. It also includes information on the amount (file size). In the present embodiment, the three layout data A to C are connected. However, the number of connected layout data is not particularly limited. For example, only one layout data is connected. May be.

  For the sake of simplicity, it is assumed that both the “repetitive graphic” and the “single graphic” have a substantially rectangular shape. Further, the above-mentioned “number” of repetitive figures or single figures is the number when one substantially rectangular figure is counted as one.

  The data input unit 2 has a function of reading data from the layout data A to C. Further, the layout data A to C can be read even if they are delivered from different manufacturers or have different compression formats.

  The data development / synthesis unit 6 has a function of reading data from the data input unit 2 and creating final pattern data (drawing data) by developing and synthesizing the data. A recording unit 6 a is connected to the data expansion / combination unit 6 as necessary.

  The recording unit 6a stores, for example, final pattern design data. In this case, the data processing apparatus 1 may combine and expand and synthesize the data read from each of the layout data A to C based on information such as the design data stored in the recording unit 6a. .

  The output unit 5 is a part to which final pattern data created by data processing by the data development / synthesis unit 6 is output. For example, the output unit 5 is connected to a drawing drive unit (not shown) of an electron beam drawing apparatus. Yes. Thus, an electron beam drawing apparatus (not shown) can draw a mask pattern based on the pattern data.

  The measurement unit 8 is connected to the data development / synthesis unit 6 and has a function of measuring the time of data processing by the data processing apparatus 1. More specifically, the time (actual processing time) from the start of one data processing until the data development / synthesis unit 6 creates final pattern data is measured, and the time obtained as a result is measured. A function of outputting data to the recording unit 4a (details will be described later) is provided. Further, the measuring unit 8 has a function of reading the data of the layout data A to C used for data processing together with the time data and outputting the data to the recording unit 4a.

  The processing time calculation unit 4 is a component that predicts the processing time of data processing by the data processing device 1 based on the layout data A to C, and can read and write data to the recording unit 4a. It has become.

  The display unit 7 is connected to the processing time calculation unit 4 and includes, for example, a display. As a result, the processing time predicted by the processing time calculation unit 4 is displayed on the display unit 7.

  In the recording unit 4a, a calculation formula for predicting the processing time, coefficient data used in the calculation formula, and data of time of data processing performed in the past, if necessary, Data such as the number of figures in the used layout data is stored. The “data such as the number of figures in the layout data used in the data processing” specifically refers to the number of repetitive figures and “single figures” of the layout data A to C and the data amount ( File size) information, and actual processing time data output from the measurement unit 8.

  The data processing apparatus 1 includes a control unit (not shown) for controlling each of the above-described components, and this control unit (not shown) causes each component to execute a predetermined process. Has a program.

  The data processing apparatus 1 uses a multiple regression equation as a calculation formula for predicting a processing time, and predicts a time required for data processing for each layout data A to C.

  Hereinafter, calculation of the data processing time using this multiple regression equation will be described in detail.

In the data processing for each layout data A to C by the data processing apparatus 1, the data processing time per unit file size is set to t _A0 , t _B0 , and t _C0 corresponding to the layout data A to C, respectively.

Here, the “file size” of “per unit file size” means the file size (data amount) of “single figure” and the file size of “ repetitive figure” stored in each layout data A to C ( Data amount).

Stored in each layout data A through C, "alone figure" number of the T _SA, T _SB, T _SC respectively, the number of each T _rA of "repetition figure" after expansion, T _rB, and T _rC To do. Similarly, the file sizes of “single figures” stored in the layout data A to C are respectively S _SA , S _SB and S _SC, and the file sizes of “repetitive figures” are S _rA , S _rB and S _SC , respectively. _{Let rC} . When the regression coefficients of the multiple regression equation are a ₀ , a ₁ , and a ₂ , the data processing times t _{A0 to} t _C0 per unit file size are represented by the following relationship.
t _A0 = a ₀ + a ₁ × T _SA / S _SA + a ₂ × T _rA / S _rA (1)
t _B0 = a ₀ + a ₁ × T _SB / S _SB + a ₂ × T _rB / S _rB (2)
t _C0 = a ₀ + a ₁ × T _SC / S _SC + a ₂ × T _rC / S _rC (3)
The regression coefficients a ₀ , a ₁ , and a ₂ are obtained by performing multiple regression analysis in advance using data of the execution history of data processing executed in the past.

From the above, the times t _A , t _B , t _C required for the respective data processing on the layout data A to _C are
t _A = t _A0 × (S _SA + S _rA ) (4)
t _B = t _B0 × (S _SB + S _rB ) (5)
t _C = t _C0 × (S _SC + S _rC ) (6)
Furthermore, the total time t required for data processing on the layout data A to C is
t = t _A + t _B + t _C (7)
It becomes. The total time t predicted in this way is displayed on the display unit 7. If necessary, in addition to the total time t, the time t _A , t _B , t _C required for all data processing for each layout data A to _C may be displayed.

Referring again to FIG. 1, to supplement the function of the processing time calculation unit 4, the processing time calculation unit 4 updates the above-described regression coefficients a _{0 to} a ₂ by adding the number of “single figures” T _SA to T _SC , the number of “repetitive figures” T _{rA to} T _rC , the file sizes S _{SA to} S _SC and S _{rA to} S _{rC of} both figures, and the actual processing output from the measurement unit 8 to the recording unit 4a It also has a function of performing multiple regression analysis using time to calculate new regression coefficients a ₀ ′, a 1 ′, a ₂ ′ and updating them in the recording unit 4 a.

The operation of the data processing apparatus 1 configured as described above will be described below with reference to the flowchart shown in FIG.
(Create drawing data)
The data processing operation in which the data processing apparatus 1 develops and synthesizes the layout data A to C to create final drawing data is performed as follows.

  First, when data processing by the data processing apparatus 1 is started, the data input unit 2 reads data relating to each “single figure”, “repetitive figure”, and the like from the respective layout data A to C. At the same time as the data processing starts, the measurement unit 8 starts measuring the time required for the data processing.

  Next, the data development / combination unit 6 reads the data read from the data input unit 2 into the layout data A to C, and collates the data with the design data stored in the recording unit 6a. Is expanded and combined to create the final “drawing data”. The “drawing data” created in this way is output from the data development / synthesis unit 6 to the output unit 5.

  The measurement unit 8 stops measuring time when the data development / combination unit 6 creates final drawing data, and outputs the time actually taken by this data processing to the recording unit 4a.

Through the series of steps described above, the layout data A to C are combined and developed, and final drawing data is created.
(Prediction and display of processing time)
The operation in which the data processing device 1 predicts the data processing time is performed as follows.

First, the data input unit 2 determines the number of “single figures” T _{SA to} T _SC , the number of “repeated figures” T _{rA to} T _rC from the layout data A to C, and the file sizes S _SA of both figures. to S _SC, read S _rA to S _rC (see step S1 in FIG. 2).

Next, the processing time calculation unit 4 reads each of the data read into the data input unit 2 and the equations (1) to (3) stored in the recording unit 4a, and the data is converted into the equations (1) to (1) to (1) to (3). By substituting in (3), the processing times t _{A0 to} t _C0 per unit file size in the data processing corresponding to the layout data A to _C are calculated (S2).

Next, the processing time calculation unit 4 substitutes the processing times t _{A0 to} t _C0 per unit file size calculated in the previous step (S2) into the formulas (4) to (6), so that each layout data A to Times t _{A to} t _C required for data processing corresponding to _C are calculated (S3).

The processing time computing unit 4 calculates the total time t of the data processing time by substituting the time t _{A to} t _C necessary for the data processing calculated in this way into the above-described equation (7) (S4).

Next, the total time t (or times t _{A to} t _C necessary for each data processing as necessary) of the data processing time calculated in this way is displayed on the display unit 7 (S5).

Through the series of steps described above, the processing time of the data processing is predicted, and the predicted time is displayed on the display unit 7.
(Coefficient calculation by database construction and multiple regression analysis)
The actual time data measured by the measuring unit 8 is used to update the regression coefficients a _{0 to} a ₂ . Hereinafter, an operation of updating the regression coefficients a _{0 to} a ₂ will be described.

As described above, the measurement unit 8 uses not only the measured time data but also the data of the layout data A to C used in the data processing, more specifically, as the variables of the above-described equations (1) to (3). Data used (number of “single figure” T _{SA to} T _SC , number of “repetitive figure” T _{rA to} T _rC , file size S _{SA to} S _{SC of} “single figure”, file size S of “repetitive figure” _{rA to} S _rC ) are also output to the recording unit 4a. In this way, the data used in one data processing (the number of “single figures” T _{SA to} T _SC etc.) and the actual time data required for the data processing are associated with each other. Thus, a database comprising these data is stored in the recording unit 4a.

Processing calculating unit 4 uses the data in the database that are configured in the recording unit 4a as described above, again, calculate the coefficients a ₀ ~a ₂ equation by performing multiple regression analysis (1) - (3) Then, the new coefficient obtained as a result is updated in the recording unit 4a.

The coefficients a _{0 to} a ₂ are updated through the series of steps described above. The updating of the coefficients a _{0 to} a ₂ may be automatically performed every time data processing is performed.

According to the data processing apparatus 1 described above, the multiple regression equation is used as the calculation formula for predicting the data processing time, and the regression coefficients a _{0 to} a _{2 of the} regression formula are “single figure” and “repetitive figure”. Therefore, the data processing time can be predicted more accurately. In addition, since the speed of data processing often varies depending on the attributes of the data processing target, it is more difficult to handle layout data separately as “single figure” and “repetitive figure” as in this embodiment. This is preferable in that an accurate prediction result can be obtained. Furthermore, the measurement unit 8 is provided to measure the actual processing time required for one data processing, and based on the data, the multiple regression analysis is performed again to update the regression coefficient. The accuracy is further improved.

It is a block diagram which shows the structure of the data processor by one Embodiment of this invention. It is a flowchart for demonstrating an example of the data processing by the data processor of FIG.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Data processing apparatus 2 Data input part 4 Processing time calculating part 4a Recording part 5 Output part 6 Data expansion | deployment / synthesis | combination part 6a Recording part 7 Display part 8 Measurement part AC Layout data

Claims (6)

A method for predicting data processing time in data processing for creating pattern data based on layout data,
Reading numerical information for predicting the data processing time from the layout data;
Multiple regression equation reads, the data processing time of the unit file per size by substituting the numerical information read from the layout data in the multiple regression equation to predict the data processing time per unit file size calculates, by multiplying the sum of the file size stored in the layout data to the data processing time of the unit per file size, has a step of substantially calculates the data processing time,
The layout data is a format including single figure data, coordinate data indicating the position of the single figure, data of the repeated figure, and coordinate data indicating the position of the repeated figure,
The step of reading the numerical information includes the number T _{S of the} single figure, the number T _{r of the} repeated figure after expansion, the file size S _s of the data of all the single figures in the layout data, and the layout data Including reading the file size S _r of the data of all repeated figures after expansion in
In the step of substantially calculating the data processing time, when calculating the data processing time per unit file size , the weight with (T _S / S _s ) and (T _r / S _r ) as variables is used. A method for predicting data processing time, including using a regression equation. Measuring the actual time taken to process data in the past;
A multiple regression analysis is performed using the data of the actual time and the numerical information used for predicting the processing time of the data processing used in the data processing performed in the past, and a multiple regression equation is obtained. The method for predicting data processing time according to claim 1, further comprising the step of re-determining a coefficient required for determination. A data processing apparatus that performs data processing for creating pattern data based on layout data,
A data input unit that is connected to the layout data and reads numerical information for predicting a data processing time in the data processing from the layout data;
Recording means for pre-stored multiple regression equations for predicting the data processing time per unit file size ;
The data processing time per unit file size is calculated by substituting the numerical information read from the layout data into the multiple regression equation, and stored in the layout data at the data processing time per unit file size. and that by multiplying the total file size, provided with a processing time calculation unit which substantially calculates the data processing time,
The layout data is a format including single figure data, coordinate data indicating the position of the single figure, data of the repeated figure, and coordinate data indicating the position of the repeated figure,
Wherein the data input unit, the number T _S of the sole figure, the number T _r of the repeating graphic after expanding the file size S _s of the data of all the single figure in the layout data, and, in the layout data Read the file size S _r of all the repeated figure data after expansion ,
The processing time calculation unit uses the multiple regression equation with (T _S / S _s ) and (T _r / S _r ) as variables when calculating the data processing time per unit file size. Processing equipment. It further includes a measurement unit that measures actual time taken for data processing performed in the past and outputs data of the actual time,
The processing time calculation unit uses the data of the actual time and the multiple regression analysis using the numerical information for predicting the processing time of the data processing used in the data processing performed in the past. 4. The data processing apparatus according to claim 3, wherein the coefficient necessary for determining the multiple regression equation is obtained again.   The data processing apparatus according to claim 3, further comprising a display unit that displays the predicted data processing time. Reading numerical information for predicting data processing time in data processing for creating pattern data from layout data;
Multiple regression equation reads, the data processing per unit file size by substituting the numerical information read from the layout data in the multiple regression equation to predict the data processing time per unit file size Calculating a time, and multiplying the data processing time per unit file size by the sum of the file sizes stored in the layout data to substantially calculate the data processing time, A program for executing
The layout data is a format including single figure data, coordinate data indicating the position of the single figure, data of the repeated figure, and coordinate data indicating the position of the repeated figure,
The step of reading the numerical information includes the number T _{S of the} single figure, the number T _{r of the} repeated figure after expansion, the file size S _s of the data of all the single figures in the layout data, and the layout data Including reading the file size S _r of the data of all repeated figures after expansion in
The step of substantially calculating the data processing time includes the step of calculating the data processing time per unit file size by using (T _S / S _s ) and (T _r / S _r ) as variables. A program that involves using regression equations.

Images