CN1307278A

CN1307278A - Method for Calibrating Real Time Clock Using Time Stamp Counter Inside Central Processing Unit

Info

Publication number: CN1307278A
Application number: CN 00101812
Authority: CN
Inventors: 蔡俊男
Original assignee: Mitac International Corp
Current assignee: Mitac International Corp
Priority date: 2000-01-25
Filing date: 2000-01-25
Publication date: 2001-08-08
Anticipated expiration: 2020-01-25
Also published as: CN1151433C

Abstract

A method for calibrating real-time clock by using time-stamp counter in CPU of computer system features that the time-stamp counter in CPU of computer system is used as reference, the real-time clock periodically triggers a calibration procedure, and the error between the time of real-time clock and the time corresponding to the counting times of time-stamp counter is kept in a certain range.

Description

Calibrate the method for real-time clock with the time stamp counter of central processing unit inside

The present invention relates to real-time clock (Real Time Clock; RTC) calibration steps is particularly with inner time stamp counter (the Time Stamp Counter of the central processing unit (CPU) of computer system; TSC) be benchmark, in order to the method for calibration real-time clock.

Real-time clock in the general computer system cooperates the R-C circuit to constitute with crystal oscillator mostly.Yet because component ageing or problem such as do not match, the real-time clock in the system also can be subjected to related influence, and loses its accuracy.For server system (Server System), on network, need the time of broadcast date information to give user side, or for computer system, performed Application Software Program may need the precise time of date and time information; If intrasystem real-time clock loses its accuracy, then the overall efficiency of system will be had a greatly reduced quality, even makes a mistake and can't move.Therefore, be necessary really to propose a kind of effective calibration steps, to guarantee the accuracy of real-time clock at real-time clock.

In view of this, it is benchmark that the present invention utilizes the inner time stamp counter of system CPU (for example Intel Pentium II), make real time clock period ground trigger a calibration process, and allow the time of real-time clock and time stamp rolling counters forward number of times error between the two of pairing time remain within certain scope, and reach the purpose of proofreading and correct real-time clock.

For achieving the above object, the present invention proposes a kind of method of using the time stamp counter calibration real-time clock of computer system CPU inside, and this method comprises the steps.

Set real-time clock, make its periodically every M trigger a calibration process second and carry out N time; And every execution calibration process once, just first count value added 1.

When calibration process is carried out N time, carry out following steps: the actual count value of obtaining interior time stamp counter M second with and nominal count value; Wherein, nominal count value is Z * M, and Z is the standard meter number of times of above-mentioned time stamp counter per second; Obtain actual count value and nominal count value error amount between the two; Use above-mentioned error amount to obtain corresponding error time, so that the time value of calibration real-time clock.

For above-mentioned purpose of the present invention, feature and advantage can be become apparent, preferred embodiment cited below particularly, and conjunction with figs. elaborates.Wherein:

Fig. 1 is shown as enforcement the inventive method, and the process flow diagram of the assignment procedure of carrying out in advance.

Fig. 2 is the process flow diagram that shows calibration process method of the present invention.

Fig. 3 is presented in the calibration process, the process flow diagram of employed error adjustment process.

According to the present invention, the time stamp counter that uses computer system CPU inside is with the calibration real-time clock Method may further comprise the steps.

Set real-time clock, make its periodically every M triggers the calibration process of execution N time second; That is work as the school Positive process is carried out after N time, the time-out take the time value of real-time clock as benchmark M second.

Every execution calibration process once just adds 1 with first count value.

When calibration process is carried out N time, then carry out following steps:

By the standard meter number of times Z of time stamp counter per second, can obtain nominal count value Z * M (because of For whether M accurately waits to verify second); Again the count value of present time stamp counter being deducted the 1st time holds The count value of stamp counter during the row trimming process, it is actual in time second at M namely can to draw the stamp counter On the counts value.

Obtain actual count value and nominal count value error amount between the two; If the actual count value greater than nominal count value, represents that promptly the time of real-time clock is slow, if the actual count value less than nominal count value, represents that promptly the time of real-time clock is fast.Therefore, the use error value can be obtained corresponding error time, and the time value of being calibrated real-time clock.

Hereinafter will be for an embodiment, and describe running of the present invention in detail with reference to Fig. 1～3.

Assignment procedure

Fig. 1 is the process flow diagram that is shown as the assignment procedure of implementing the inventive method and carrying out in advance.

At first, the clock cycle that draws its internal processor by system's employed CPU model speed (for example CPU is Pentium II 233 MHz); That is obtain internal system time stamp counter, required time when counting one time.Again by the standard counts Z that draws above-mentioned time stamp counter per second above-mentioned clock cycle.Then, running is interrupted in shielding, and this moment, any interrupt request all can not accepted, to guarantee can not be subjected to the interference of other interrupt request when importing new interrupt procedure again.Then, storage INT 70h interrupt vector process pointed, and replace with calibration process of the present invention.Afterwards, again real-time clock is set at every M and produces N interrupt request (IRQ 8) second, so that trigger calibration process.At last, setting the first count value count1 is 0, recovers the normal interruption running again, thereby finishes assignment procedure.In the present embodiment, setting real-time clock is that per second produces 1024 interrupt request.

The interrupt request that real-time clock sends is IRQ 8, and it can trigger INT 70h interrupt vector process pointed; Therefore, in assignment procedure promptly with calibration process of the present invention as a new INT 70h process, to replace the original process pointed of INT 70h interrupt vector.So, when real time clock period when sending interrupt request, just trigger and carry out calibration process of the present invention.

Calibration process

Fig. 2 is for showing the process flow diagram of calibration process method of the present invention.

Before calibration process of the present invention was described in detail in detail, the meaning of the storing value representative of each storage buffer of explanation was as follows earlier.

A: after the adjustment action in the calibration process startup back first time or the calibration process takes place, the counts value that read by the time stamp counter first time.

B: through M after second, the counts value of being read by the time stamp counter.

C: for the first time after the time for reading stamp counter, the actual count number of times at the interior time stamp counter of M second promptly equals B-A.

D, E: M is after second for process, actual count number of times and nominal counts difference between the two.

Z: the standard counts in the time stamp counter 1 second.

Count1: behind the 1st the time for reading stamp counter, trigger the number of times (count1 is every promptly to be re-set as 1 through M second) of carrying out calibration process by real-time clock.

Count2: after the 1st time for reading stamp counter, real-time clock think own itself the second number of process.

Count3: after the count value of time stamp counter is compared, the time deviation of real-time clock.

Use the present invention, computer system one is carried out after the above-mentioned setting, and the real-time clock of system will send 1024 interrupt request (IRQ 8) by per second, and periodically carries out following calibration process.As shown in Figure 2, calibration process comprises the steps.

(whether step 1), calibration process are triggered by the periodic interruptions request of real-time clock, can learn by the buffer status from real-time clock to read the state of real-time clock.If judging calibration process is not that periodic interruptions request by real-time clock is triggered (step 2), then finish calibration process.

Triggered by the periodic interruptions request of real-time clock if judge calibration process system, then the metering value TSC_count (step 3) of time for reading stamp timer.Then, judge whether the first count value count1 is 0 (step 4).

If be judged to be 0 (expression calibration process be the 1st time be performed), then setting second and third count value count2, count3 is 0, and the counts value TSC_count of the above-mentioned time stamp timer that will read is stored to register A (step 10).Then, again the value of the first count value count1 is added 1 (step 11); Afterwards, finish calibration process.

If judge that the first count value count1 is not 0, continue then to judge whether count1 equals 1024 (step 5); That is whether the time of judgement real-time clock is through 1 second.If judge that count1 is not equal to 1024, then the value with the first count value count1 adds 1 (step 11); Afterwards, finish calibration process.

If it is 1024 (time of expression real-time clock is through 1 seconds) that step 5 is judged the first count value count1, the action below then carrying out in step 6: the time value RTC_time that reads present real-time clock; The value of present time stamp timer is stored among the register B; Value among the register B is deducted value among the register A, and the value of B-A is stored among the register C; And, the second count value count2 is added 1.Wherein, the value representation real-time clock of count2 thinks that institute elapsed time own is 1 second, the value representation among the register C from the 1st time for reading stamp counter through after 1 second, the actual count number of times of time stamp counter.

The second count value count2 be multiply by the standard counts Z of time stamp counter per second, and obtain nominal count value Z * count2, try to achieve between the two difference of the storing value of actual count number of times among the register C and nominal count value (C-Z * count2), and it is stored in (step 7) among register D and the E again.

Utilize storing value and nominal count value difference between the two among the above-mentioned register C, carry out error adjustment process (step 8), obtain the time error value that to be revised (just, the time deviation value of real-time clock), as the content of above-mentioned the 3rd count value count3, in order to the time value of calibration real-time clock.

At last, with the time value that is worth the above-mentioned real-time clock of replacement correction real-time clock time of adjustment, replace the storing value of register A with the storing value of register B; And first count value is set at 1, second, third count value is set at 0 (step 9).So finish calibration process.

The error adjustment process

The error adjustment process mainly comprises the steps.

(1) storing value of comparand register C and nominal count value if difference value (being stored among the register D) between the two is not less than a certain errors allowable value, are then carried out next step; Otherwise, first count value is set at 1, finish calibration process.

(2) difference value is converted to the time deviation value of real-time clock.

(3) if register C storing value is not less than nominal count value, the time of expression real-time clock more normally is slow, then the time deviation value is added the time value of real-time clock, and obtains through the value real-time clock time of adjustment; If register C storing value is less than nominal count value, the time of expression real-time clock more normally is fast, then deducts the time deviation value in the time value by real-time clock, and obtains through the value real-time clock time of adjustment.

The process flow diagram that hereinafter will cooperate error adjustment process shown in Figure 3, the action of specification error adjustment process.

(during step 8), at first, whether the value among the criterion register D is more than or equal to 0 (step 80) when calibration process proceeds to the error adjustment process; If more than or equal to 0, then execution in step 82; If less than 0, then the value among the register D is taken absolute value (step 81) afterwards, execution in step 82 again.

Whether actual count number of times and the nominal counts error amount (being the value of register D) between the two of judging the time stamp counter exceed the preset range (step 82) of allowing.In this embodiment, error permissible range system is set at the standard meter number of times Z (that is to say that the error permissible range of corresponding real-time clock is within 1 second) of time stamp counter per second.

If D is less than Z, then owing to still in error range, needn't adjust the time of real-time clock, so the first count value count1 is reset to 1 and finish calibration process.

If D is more than or equal to Z, then repeated execution of steps 83 and step 84, the value in register D less than Z till.Wherein, every execution in step 83 once, just the value with register D deducts Z and deposits back register D, and the value of the 3rd count value count3 was added for 1 (second).So the error amount of counts is converted to the represented time deviation value by count3.

Afterwards, in step 85, if the actual count number of times of judging the time stamp counter is more than or equal to nominal counts (being E 〉=0), expression real-time clock calibration reason shape is slow, so the time deviation value of count3 is added on the time RTC_time of real-time clock, and reaches the purpose of adjustment real-time clock.

In step 85, if the actual count number of times of judging the time stamp counter is less than nominal counts (being E＜0), expression real-time clock calibration reason shape is fast, thus deduct the time deviation value of count3 by the time RTC_time of real-time clock, and reach the purpose of adjustment real-time clock.

Though the present invention discloses as above with preferred embodiment; but be not in order to qualification the present invention, the people of any skilled, without departing from the spirit and scope of the present invention; can make and change and modification, so protection scope of the present invention is determined by the scope that appending claims defined.

Claims

1. A method for calibrating a real-time clock using a time stamp counter inside a central processing unit of a computer system, characterized in that the method comprises the following steps:

Set the real-time clock so that it periodically triggers a calibration process every M seconds and executes it N times;

Every time the calibration process is executed, the first count value is increased by 1;

When the calibration process is performed N times, the following steps are performed:

Find the actual count value and its nominal count value of the time stamp counter in M seconds; wherein, the nominal count value is Z×M, and Z is the standard number of counts per second of the time stamp counter;

finding an error value between said actual count value and said nominal count value;

The corresponding error time is obtained by using the error value, so as to calibrate the time value of the real-time clock.

2. The method according to claim 1, characterized in that, the time required for counting one time of the time stamp counter is obtained by the central processing unit model of the system used, and then the standard of each second of the time stamp counter is obtained accordingly Count Z.

3. The method according to claim 1, wherein the real-time clock periodically sends an interrupt request IRQ 8 to trigger the calibration process.

4. The method according to claim 3, characterized in that, the process pointed to by the computer system INT 70h interrupt vector is stored in advance, and replaced by the calibration process.

5. The method according to claim 1, wherein the actual count value of the time stamp counter in M seconds is obtained by subtracting the initial count value from the count value of the time stamp counter at M seconds; The count value is the first time the calibration process is executed, or the number of counts read by the time stamp counter for the first time after the real-time clock is calibrated.

6. A method for calibrating a real-time clock using a time stamp counter inside a central processing unit of a computer system, wherein the calibrating process includes the following steps:

reading the contents of said timestamp counter;

Determine whether the first count value is 0; if it is not 0, then execute the next step; if it is 0, then the second count value and the third count value are all set to 0, and the read time The content of the stamp counter is stored in device A, and after adding 1 to the first count value, the calibration process is ended;

Determine whether the first count value is equal to a preset number of times: if it is equal to the value, proceed to the next step; if it is not equal to the value, add 1 to the first count value, and end the calibration process;

Read the time value of the real-time clock;

Adding 1 to the second count value, recording the content of the timestamp counter at this time in a device B, and calculating the difference between the stored value in the device B and the stored value in the device A difference, and store it in device C;

multiplying the second count value by a standard count count per second of the timestamp counter to obtain a nominal count value;

Using the difference between the stored value of the device C and the nominal count value, the error adjustment process is carried out, and the time error value that should be corrected is obtained as the content of the third count value to calibrate the the time value of the real-time clock;

Replace the time value of the real-time clock with the adjusted real-time clock time value, and replace the device A storage value with the device B storage value;

The first count value is set to 1, and the second and third count values are set to 0.

7. The method according to claim 6, wherein the error adjustment process comprises the steps of:

Comparing the stored value of the device C with the nominal count value, if the difference between the two is not less than a specific error tolerance value, then perform the next step; otherwise, set the first count value to 1, ending the calibration process;

Converting the difference value into the time error value that should be corrected, wherein:

If the storage value of the device C is not less than the nominal count value, the time error value is added to the time value of the real-time clock to obtain the adjusted real-time clock time value; if the device C stores If the value is smaller than the nominal count value, the time error value is subtracted from the time value of the real clock to obtain the adjusted real time clock time value.

8. method as claimed in claim 6, is characterized in that, also comprises setting process, in said setting process, calculates the time stamp counter required for counting once by the model speed of the central processing unit of the system used. time, and then obtain the standard number of counts per second of the timestamp counter.

9. The method according to claim 6, further comprising a setting process, before performing the calibration process, the standard number of counts per second of the time stamp counter is obtained by the setting process, and The first count value is set to zero.

10. The method according to claim 6, further comprising a setting process, in which a real-time clock is set to periodically trigger the execution of the calibration process, and the number of executions per second Then it is the preset number of times.

11. The method according to claim 10, wherein the real-time clock periodically sends an interrupt request IRQ 8 to trigger the calibration process.

12. The method according to claim 11, characterized in that, the process pointed to by the INT 70h interrupt vector of the computer system is stored in advance, and replaced by the calibration process.

13. The method according to claim 6, wherein when performing the calibration process, the state of the real-time clock is first read to determine whether the calibration process is triggered by a periodic interrupt request of the real-time clock Start; if yes, continue to execute the calibration process, otherwise end the calibration process.