JP2834194B2 - Hardware monitor - Google Patents

Description

Description: TECHNICAL FIELD The present invention relates to a hardware monitor for collecting measurement data relating to performance characteristics of a data processing system.

[Conventional technology]

[What is a hardware monitor?] There is a hardware monitor as one means for monitoring the performance of a processor. The hardware monitor obtains meaningful information by extracting internal signals of the monitored processor, and has a feature that the operation of the monitored processor is not disturbed. The hardware monitor is used to measure the performance of the processor and to obtain information necessary for analyzing the processor's overhead factor.

From the viewpoint of the measurement function, the hardware monitor can be classified into a counting method and a recording method. The counting method is a method of counting the number of times an event has occurred and the time during which the event has occurred.
The counting method is used when the counted value itself, such as the total number of times or the total time, becomes meaningful information. On the other hand, the recording method is a method in which when an event occurs, data relating to the event is collected and recorded on a recording medium. The process of editing the recorded data into meaningful information is performed after recording on the recording medium. The recording method is used when meaningful information cannot be obtained by the counting method because an event to be analyzed is complicated.
An example is analysis of the operation history of a program.

[Examples of events that require analysis of processor performance degradation factors]

One example of a factor that degrades processor performance is overhead due to resource contention. Examples include resources that require exclusive control, such as data shared by multiple processes and storage devices shared by multiple processors. When a resource conflict occurs, the requestor of the resource must wait until the resource can be acquired, which is overhead. In a system in which a plurality of processes and processors operate in parallel, it is necessary to eliminate this overhead as much as possible, and its analysis is important.

[Analysis of overhead associated with exclusive control and examples of countermeasures]

In order to analyze the overhead associated with such exclusive control, information such as the total number of times of waiting and the total time, as well as the distribution of the time allotted and the distribution of the time during which resources have been acquired, are required.

First, the ratio of the overhead of the event of interest to the total overhead is examined by using the total number of times and the total time during which waiting has occurred, and it is determined whether or not the overhead is a countermeasure to be taken. If you decide that you need to take action, then examine the time distribution. For example, it is possible to check the distribution of the waiting time and take measures such as changing a timing pattern for sending a resource acquisition request. Also, by using the distribution of the time during which the resources have been acquired, it is possible to take measures such as changing the exclusive control unit such as dividing or integrating the exclusive control part.

As described above, the overhead factor can be analyzed using the information such as the number of times and the time distribution, and countermeasures can be taken.

[Example of time distribution measurement]

The usefulness of the information, such as the number and time distribution, has been described. Next, a conventional method for measuring a time distribution without disturbing the operation of the monitored processor using a hardware monitor will be described. Since the number of times can be easily measured by the counting method classified above, the description of the number of times measurement method is omitted.

The time distribution measures the time when an event has occurred each time the event occurs, records the measured value on a recording medium, and after the measurement is completed, reads out the data recorded on the recording medium and classifies the data according to time. It is required by That is, it can be obtained by the recording method described above.

For example, as shown in FIG. 3A, it is assumed that a total of 20 events have occurred in the order of 16 cycles at the first time and 12 cycles at the second time. In the recording method, each time an event occurs, the occurrence times of these events are obtained, and the occurrence times are sequentially recorded on a recording medium and stored in a format as shown in FIG. 3 (B). In FIG. 3 (B), Seq (n) indicates the occurrence time of the n-th occurring event. When the measurement is completed, these data are read out one by one and are obtained by classifying them by time as shown in FIG. 3 (C). In FIG. 3 (C), Time (n) indicates the number of occurrences of an event that has occurred in n cycles.

Here, the recording medium is appropriately selected in accordance with the total number of events to be measured and the intervals at which it occurs, and is connected to a hardware monitor. For example, when events occurring at short intervals are measured, a high-speed memory or the like is used as a recording medium, and when a number of occurring events are measured, a magnetic recording medium or the like is used.

[Problems to be solved by the invention]

When measuring the time distribution by the conventional recording method, there is the following problem.

(1) Since the recording method requires a recording medium, the upper limit of the total number of events that can be measured is determined by the capacity of the recording medium. Further, since the recording method requires transfer processing of measured values, the lower limit of the interval between occurrences of measurable events is determined by the transfer speed. For this reason, in order to measure a large number of events occurring at short intervals, a recording medium having a high writing speed and a large capacity must be connected. That is, there is a problem that the physical quantity of the recording medium increases in proportion to the total number of events to be measured.

(2) Further, in the recording method, every time measurement is completed, data must be read from the recording medium and classified according to time. That is, there is a problem in that post-processing takes time.

SUMMARY OF THE INVENTION It is an object of the present invention to provide a hardware monitor capable of recording an occurrence time distribution of a large number of occurring events at short intervals. In the previous example, Fig. 3 (A)
When the events occur in the order shown in Fig. 3 (C)
An object of the present invention is to provide a hardware monitor capable of measuring the classification by occurrence time as shown in FIG.

[Means for solving the problem]

In order to achieve the above object, the hardware monitor of the present invention includes a measurement value conversion device that receives a measurement value of a time when an event has occurred and generates a storage address of a value at which an addition command is applied.

[Action]

Since the measurement value is converted and recorded by the measurement value conversion device, the physical quantity of the recording medium can be reduced, and the classification work after the end of measurement, which is conventionally required, is not required. In the previous example, a recording medium for recording the measurement values in the format shown in FIG. 3 (B) becomes unnecessary, and the format shown in FIG. 3 (B) is changed to the format shown in FIG. 3 (C). Eliminates the need to classify by format.

〔Example〕

FIG. 1 shows a hardware monitor (2) according to the present invention.
, Monitored processor (1), and measurement value editing device (3)
An example of connection is shown. In FIG. 1, the monitored processor (1) is described as being constituted by one unit, but the monitored processor (1) may be constituted by a plurality of units. In this case, the processor internal signal group increases by the number of processors. In FIG. 1, the monitoring processor (1) and the measurement value editing device (3) are described as separate devices.
The monitored processor (1) and the measurement value editing device (3) may be the same. In the following, for simplicity of description, a case will be described in which each of the monitoring processor (1) and the measurement value editing device (3) is constituted by one.

As shown in FIG. 1, the hardware monitor (2) is composed of the following four types of devices. Next, the functions of these devices will be described in numerical order with reference to FIG. In the following description, the hardware monitor is abbreviated as HM.

HM control device (21) The following six types of control are performed according to the HM control command sent from the measurement value editing device (3), and a response to the HM control command is returned. (1) setting of measurement method, (2) setting of measurement value conversion method, (3) control of start and end of measurement, (4) reading of value stored in event occurrence time distribution storage device, (5) event Clear the measurement value of the occurrence time measuring device, (6) Initialize the event occurrence time distribution storage device.

Event occurrence time measurement device (22) Based on the measurement method specified by the monitoring event designation device (21), an event to be measured is selected from a processor internal signal group sent from the monitored processor (1), and an event occurrence signal is generated. Generate Further, the time when the event occurrence signal is logically 1 is counted, and a measurement value is generated. These functions are valid as long as the measurement start / end control signal (omitted in FIG. 1) sent from the monitoring event specifying device (21) is logically 1. Further, it has a function of clearing the measured value in response to a time measured value clear command (omitted in FIG. 1) sent from the HM control device (21).

A configuration example of the present apparatus will be described in detail in a first embodiment of the present invention described later.

Measurement value conversion device (23) Based on the measurement value conversion method specified by the monitoring event specification device (21), an address signal is sent to the location specified by the address from the measurement value sent from the event occurrence time measurement device (22). And an addition command which means to increment the stored value.

A configuration example of the present apparatus will be described in detail in a first embodiment of the present invention described later.

Event occurrence time distribution storage device (24) In response to the address signal and the addition command sent from the measured value conversion device (23), the value stored in the location specified by the address signal is incremented. Although not shown in FIG. 1, it also has a function of clearing the value stored in the memory in response to the initialization command sent from the HM control device (21). Furthermore, in response to a read command and a read address of a measurement value sent from the HM control device (21), a function of returning a value stored at a specified address and a signal indicating the significance of the value (eg, an overflow signal) is also provided. Have both.

FIG. 2 shows an example of realizing a function corresponding to the present invention using a conventional technique. In the prior art, based on a measurement method specified by an HM control signal (21), an event selection device (100) in an event occurrence time measurement device (22) transmits an event to be measured from a monitored processor (1). An event occurrence signal is generated by selecting from a group of processor internal signals to be sent. Further, the occurrence time measuring device (200) counts the time when the event occurrence signal is logically 1 and generates a measured value.
The measured values thus obtained are sequentially recorded on the recording medium (31) inside the measured value editing device (3). After the measurement, the data recorded on the recording medium (31) is read out and classified by time to obtain a time distribution.

The operation of the embodiment shown in FIG. 1 will be briefly described. As described above, the event to be measured is selected from the processor internal signal group, the event occurrence signal is generated, the time when the event occurrence signal is logically 1 is counted, and the point of generating the measurement value is as follows. Same as before. In the present invention, the address generation device (300) generates an address signal from the measured value, and the addition command generation device (400) generates an addition command. The event occurrence time distribution storage device (24) specifies the address signal by the address signal. Is different from the conventional method in that the value stored in the stored location is incremented. Conventionally, the measured values are stored in the order in which the events to be measured have occurred, but in the present invention, the value stored in a designated location corresponding to the measured value is incremented. That is, the number of times (frequency distribution) when the measured value has a certain value is measured. As described above, in the present invention, the classification of the measured values, which has been performed after the end of the conventional measurement, can be performed in parallel with the measurement.

[First Embodiment of the Present Invention] First, a first embodiment of the present invention will be described with reference to FIGS. 1 and 4 to 8.

FIG. 4 shows a configuration example of the event occurrence time measuring device (22) shown in FIG. In the figure, = outputs a logical value of 1 when the two inputs are equal, and outputs a logical value of 0 when the two inputs are different. L indicates a latch, A indicates an AND gate, and O indicates an OR gate. Also, one of the input signals of A
A circle attached to one means an inverter. In the present configuration example, the event occurrence time measurement device (22) includes an event selection device (100) that selects a specified event and an occurrence time measurement device (200) that measures the occurrence time of the selected event.

The event selection device (100) executes the instruction stored at the address specified by Start Adr.
Measures the time until the instruction stored at the address specified by Adr.Reg. Is executed. The values of these registers are specified by the measurement method specifying signal sent from the HM control device (21). For example, when measuring the execution time distribution of the exclusive control unit described above, the start address and the end address of the exclusive control unit are specified. The device internally generates an event start signal, an event end signal, and an event occurrence signal. When the measurement start / end control signal sent from the HM control device (21) is logically 1, the event occurrence signal and the event end signal are generated. Send a signal.

The occurrence time measuring device (200) counts the time during which the event occurrence signal sent from the event selection device (100) is valid (logically 1). The counting is performed in one clock pitch, and the measured value is output after one clock. The expression form of the measured value is binary data consisting of n bits. Event sorting device (100)
When the event end signal is input from, the input of the adder is switched to '0'. When the time measurement value clear signal sent from the HM control device (21) logically becomes 1, the measurement value is cleared.

FIG. 5 shows an example of the configuration of the measurement value conversion device (23) shown in FIG. The measurement value conversion device (23) receives the measurement value generated by the event occurrence time measurement device (22) as an input and generates an address according to a designated method (300).
And an addition command generation device (400) that receives an event occurrence signal or an event end signal and generates an addition command.

The address generation device (300) converts the measurement value generated by the event occurrence time measurement device (22) according to the designation of the decrement value designation signal and the shift value designation signal sent from the HM control device (21), and Is output. A subtractor (301) outputs a value obtained by subtracting a value designated as a subtracted value from the measured value from the measured value. The shifter (302) receives the output of the subtractor as input, and outputs the result of arithmetic right shift by the value specified as the shift value. The subtracter (301) and the shifter (302) are used to extend the range of the measurement time. Ratchi (30
3) is used to match the output timing with the addition command. The address upper limit determination comparator (304) detects that the upper limit of the storage area in the occurrence time distribution storage device (24) has been exceeded, and sends an address upper limit over detection signal to the HM controller (21). Address upper limit comparator (304)
Will be described later.

The addition command generation device (400) generates an addition command based on the event end signal generated by the event occurrence time measurement device (22). The AND gate (401) acts to generate an addition command only when no borrow occurs in the subtractor (301) (the output of the subtractor is not negative). Ratchi (40
2) is used to match the output timing with the address. The AND gate (403) operates to generate an addition command only when the upper limit of the storage area in the address occurrence time distribution storage device (24) is not exceeded. The AND gate (404) detects that there is an event that cannot be measured because a borrow has occurred in the subtractor (301),
An address lower limit over detection signal is sent to the HM controller (21).

The event occurrence time distribution storage device (24) shown in FIG. 1 responds to the address output by the measurement value conversion device (23) and the addition command to store the value stored in the location specified by the address as one.
Increment. It is assumed that the address signal is performed in m bits and the addition is performed in one clock pitch. If the memory access time or the addition time is insufficient, the event occurrence time distribution storage device (24) is apparently placed in the location specified by the address within the minimum transmission interval of the addition command by making a multi-bank configuration or the like. Configure the stored value to increment. Since the event occurrence time distribution storage device (24) seems to be easily configured by those skilled in the art, a configuration example is omitted.

FIG. 6 shows an example of the configuration of the address upper limit determination comparator (304). In this configuration example, when no borrow has occurred and all the signal lines of the measurement value consisting of m bits have become 1, it is determined that the upper limit of the storage area in the occurrence time distribution storage device (24) has been exceeded, and the address is determined. Generate an upper limit over detection signal. The event end signal clears the detection signal in order to detect an address upper limit over for the next event.

The operation of the hardware monitor according to the first embodiment of the present invention based on the circuit configuration described above will be described with reference to a time chart shown in FIG. For example, the event to be measured is 2 clocks for the first time, 3 clocks for the second time, and 1 clock for the first time.
How the measured value is stored in the event occurrence time distribution storage device (24) when it occurs with a clock time width will be described. In FIG. 7, DATA (i) means data stored at address i. Occurrence time measurement device (20
It is assumed that the initial value of the measured value which is the output value of (0) is 0, and that the data in the event occurrence time distribution storage device (24) is all stored as 0 as the initial value. The value of the decrement value designation signal is 0, and the value of the shift value designation signal is also 0.

First, when the first event starts, the occurrence time measurement device (20
0) starts measurement of the time when the event occurs, and the event end signal becomes valid when the measured value is 2, that is, when the address becomes 2. Then, the addition command is sent to the measured value storage device,
1 is added to the location where the address is 2, that is, the value stored in DATA (2). The measured value is cleared by the event end signal and is initialized for the next measurement. The above operation corresponds to recording that an event having an occurrence time of two cycles has occurred once. Next, when the second event occurs, 1 is added to the value stored in DATA (3), and 3
When the first event is completed, 1 is added to the value stored in DATA (1).
Is added. These have an onset time of 2 cycles and 3
This corresponds to recording that each of the events in the cycle has occurred once.

Finally, DATA (1) is 1, DATA (2) is 1, DATA
1 is stored in (3), which means that the occurrence time is 1 cycle, 2
This indicates that the event that occurred in one cycle and three cycles occurred once. Assuming that the memory address of the event occurrence time distribution storage device (24) is the event occurrence time and the memory data is the frequency, the event occurrence time distribution storage device (24) stores the occurrence time distribution of the event just measured.

Next, the operation when the value of the decrement value designation signal is 2 and the value of the shift value designation signal is 1 will be described using the time chart shown in FIG.

First, when the first event starts, the event end signal becomes valid when the measured value becomes 2, that is, the address becomes 2. At this time, since the borrow signal is not valid, an addition command is sent to the measured value storage device, and 1 is added to the value stored in DATA (0). When the second event occurs, the DA is the same as the first event
1 is added to the value stored in TA (0). Since the borrow signal is valid for the third event, the addition command is suppressed.

The value of the decrement value designation signal is 2 and the value of the shift value designation signal is 1
Under the condition (2), the address = (count value−2) / 2 ¹ (rounded down after the decimal point), so that the event to be measured is stored in DATA (i) in 2 × i + 2 cycles to 2 × i + 3.
The number of times a cycle has occurred is counted. In this example, it is recorded that an event having an occurrence time of 2 to 3 cycles has occurred twice.

In the first embodiment of the present invention, the subtracter (301) can be omitted when the decrement value is fixed at 0, and the shifter (302) can be omitted when the shift value is fixed at 0. Of course, when the decrement value is fixed at 0 and the shift value is fixed at 0, the subtracter (301) and the shift (3
02) can be omitted. In this case, the measured value is used as it is for the address.

Further, in the present embodiment, the address generation device (300)
Is constituted by the subtractor (301) and the shifter (302), but may be constituted so that an address can be uniquely converted for a certain measured value. One method is to use a conversion table that generates an address using a measured value as an index.

Further, the event end signal is generated from the event end signal sent from the event occurrence time measuring device (22), but can be generated from the measured value. The measured value is cleared to 0 when the event ends. Therefore, the end of the event can be determined by detecting that the measured value is equal to or smaller than a certain value other than 0.

With respect to the distribution of the occurrence time of the specified event, the memory address signal is m bits, the decrement value is j, the shift value is k, and the measurement range of the occurrence time distribution is 2 ^{k + m at} intervals of j cycles to 2 ^k cycles. + J-1 cycles. When the counted value exceeds the upper limit of the memory address, the address upper limit judging comparator sends an address upper limit over detection signal. When a borrow occurs and falls below the lower limit of the memory address, an address lower limit over detection signal is transmitted. With these, the measurement leakage (the event that the time distribution cannot be measured) is detected.

[Second Embodiment of the Present Invention] Next, a second embodiment of the present invention will be described with reference to FIGS. This embodiment is different from the first embodiment in that an addition command generation device (40
0) is different. Therefore, the event occurrence time measurement value (2
2), the description of the configuration of the address generation device (300) inside the measurement value editing device (23), and the configuration of the event occurrence time distribution storage device (24) will be omitted. This embodiment is different from the first embodiment in that the number of occurrences is also recorded at an address corresponding to the measured value during the measurement.

FIG. 9 shows a configuration example of the address generation device (300) and the addition command generation device (400). In the figure, ≠ logically outputs a value of 1 when two inputs are equal and
If they are equal, a logical value of 0 is output. Ratsch (41
0) is used to match the input timing of the event occurrence signal. The AND gate (411) generates an addition command when an event is occurring and the address output from the address generation device (300) is switched. The address switching is determined by comparing the internal address of the address generation device (300) with the address sent to the event occurrence time distribution storage device (24), and determining that the address has been switched when they are not equal. The reason why the event occurrence signal is input is to suppress the case where the address is switched due to the end of the event and the clearing of the measured value. A borrow signal is input to the AND gate (411). This signal acts to generate an addition command only when the output of the subtractor (301) is not negative. The latch (412) is used to match the output timing with the address. The AND gate (413) operates to generate an addition command only when the address does not exceed the upper limit of the storage area in the event occurrence time distribution storage device (24). The AND gate (414) detects that there is an event that cannot be measured because a borrow has occurred in the subtractor (301), and sends an address lower limit over detection signal to the HM controller (21).

The operation of the hardware monitor according to the present invention based on the circuit configuration described above will be described with reference to the time chart shown in FIG. As in the above description, the initial value of the measured value as the output value of the occurrence time measuring device (200) is 0, and the data in the event occurrence time distribution storage device (24) are all 0 as the initial value.
Is stored. The value of the decrement value designation signal is 0, and the value of the shift value designation signal is also 0.

First, when the first event starts, the occurrence time measurement device (20
0) starts measuring the time when the event occurs. The event occurrence signal is logically 1 and the address is from 0 to 1,
In order to change from 1 to 2, the values stored in DATA (1) and DATA (2) are sequentially added by one. When the event ends, the measured value is cleared by the event end signal, and is initialized for the next measurement. At the same time, it is detected that the measured value has changed from 2 to 0, but the addition command is suppressed because the event occurrence signal is logically 0. The above operation corresponds to recording that an event having an occurrence time of one cycle or more is recorded once and an event having an occurrence time of two cycles or more is recorded once. Next, when the second event of 3 cycle width occurs, DATA
The values stored in (1), DATA (2), and DATA (3) are sequentially added by one, and when the third event of one cycle width occurs, the value stored in DATA (1) is incremented by one. .

Finally, 3 for DATA (1), 2 for DATA (2), DATA
1 is stored in (3). Since the number of times that the event to be measured has occurred at least i cycles or more is counted in DATA (i), the frequency of the event having occurred i cycles is given by the difference between DATA (i) and DATA (i + 1). . In other words, an event whose occurrence time is one cycle occurs once (= DATA
(1) -DATA (2) = 3-2) Similarly, the occurrence time is 2
It is recorded in the event occurrence time distribution storage device (24) that the event having a cycle is once and the event having an occurrence time of three cycles is once.

In this example, the number of occurrences of the event corresponds to the value stored in DATA (1). In the first embodiment, since the number of occurrences of an event is given by the sum of DATA (1) to DATA (2 ^m -1), it must be read from the memory 2 ^m times. However, in this example, it is only necessary to read the value stored in DATA (1). Furthermore, DATA (i) and DATA (i
+ J) are equal, the occurrence time is i cycle to (i + j)
-1) It means that there is no event in the cycle. By utilizing this, unnecessary memory reading can be eliminated.

In this example, the address upper limit over detection flag can be ignored. This is because the frequency of an event whose occurrence time is equal to or longer than 2 ^m -1 cycles is counted in DATA (2 ^m -1).

In the above, the measurement range has been described in the case where the value of the decrement value designation signal is 0 and the value of the shift value designation signal is also 0. This can be generalized as follows. In this example, when the memory address signal is m bits, the decrement value is j, and the shift value is k, the measurement range of the occurrence time distribution is 2 ^k × (2 ^m
-1) Classification can be performed at intervals of 2 ^k cycles up to + j-1 cycles, and the number of events occurring 2 ^k × (2 ^m -1) + j cycles or more can be measured.

Third Embodiment of the Present Invention A third embodiment of the present invention will be described with reference to FIG. In this embodiment, an example will be described in which the HM based on the first embodiment and the HM of the conventional recording method are realized by switching the measurement mode. This embodiment is partially different from the first embodiment in the configuration of the event occurrence time distribution storage device (24). Therefore, description of the other components is omitted.

As shown in FIG. 11, in the present embodiment, the HM control device (21)
The input of the address of the memory of the event occurrence time distribution storage device (24) and the input of the data are switched by the measurement mode signal transmitted from. When the measurement mode signal is 0, the configuration of the HM according to the present invention is obtained. When the measurement mode signal is 1, it is possible to operate as the HM of the conventional recording method. The operation when the measurement mode signal is 1 will be described below.

In the present embodiment, the conversion method command signal sent from the HM control device (21) specifies that the measured value is used as an address as it is, and that an event end signal is used as an addition command. In the first embodiment, the value of the decrement value designation signal is 0,
The value of the shift value designation signal is designated as 0. It is also assumed that the initial value of the counter in the event occurrence time distribution storage device (24) is 0.

The configuration of the counter is shown in FIG.
The configuration may be the same as in (0). That is, the event occurrence signal input may be used as the addition command input, and the measured value output may be used as the output of the counter. As the time measurement value clear signal, a counter initialization command sent from the HM control device (21) is input.

When the event occurs, the occurrence time is measured, and when the event ends, the addition command and the measured value of the occurrence time are sent to the event occurrence time distribution storage device (24). Then, the measured value is DATA
It is stored in (0), and the value of the counter is incremented by one and becomes one. When the next event occurs, the measured value is stored in DATA (1), and the value of the counter is further incremented by one. In this way, the occurrence time is stored sequentially each time an event occurs.

After the measurement is completed, the measurement value editing device (3)
To 1), a command to read a value stored in the memory of the event occurrence time distribution storage device (24) is sent. By reading the value stored in the memory and returning it to the HM control device (21) in response to the HM control device (21), the measurement value editing device (3) can edit the measurement value according to the purpose. . Thus, when the measurement mode signal is 1, it can operate as the HM of the recording method.

As described above, only the two selectors and one counter make up the first
The HM based on the embodiment and the HM of the conventional recording method can be realized by switching the measurement mode. Thereby, the occurrence times of the events can be grasped in the order of occurrence.

In this embodiment, the case where the occurrence times of events are sequentially recorded has been described. However, information accompanying the occurrence events may be recorded together with the occurrence times.

〔The invention's effect〕

In the present invention, since the measured values are classified and recorded by time, when measuring events that occur quickly and in large quantities,
The physical quantity of the recording medium of the measured value can be reduced as compared with the related art. In addition, the classification work after the end of measurement, which is conventionally required, becomes unnecessary.

In the method of recording using values during measurement,
When the data stored in the locations specified by two certain addresses are equal, it is possible to eliminate the reading of the data stored in the location between them. In addition, since the total number of occurrences of the event is the data of the address corresponding to the occurrence time of 1, the reading can be performed only once.

By adding two selectors and one counter, the function of the conventional hardware monitor can be realized with a small quantity.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of a hardware monitor according to the present invention, FIG. 2 is a block diagram showing a configuration when the occurrence time distribution is measured by a conventional hardware monitor, and FIG. 3 is an occurrence of an event to be measured. FIG. 4, FIG. 5, FIG. 6, FIG. 9, and FIG. 9 show a method of storing measured values when measuring a time distribution, and FIG. 8, 8 and 10 are timing charts showing the operation of the hardware monitor according to the present invention.
FIG. 1 is a block diagram showing a configuration when a function of a conventional hardware monitor is realized by a hardware monitor according to the present invention.

Claims (6)

(57) [Claims] An apparatus for measuring the duration of a specified event in a processor and comparing the duration of the event with a predetermined duration to determine whether a predetermined condition is satisfied. A hardware monitor, comprising: a memory for storing the number of times a predetermined condition is satisfied; and means for increasing the number of times held in the memory when a predetermined condition is satisfied. 2. The memory comprises a plurality of storage areas, wherein the storage areas are assigned different predetermined durations, and the determining device is assigned the duration of the event and the individual storage areas. 2. The hardware according to claim 1, wherein the increasing unit increases the number of times stored in the corresponding storage area when a predetermined condition is satisfied. monitor. 3. The hardware according to claim 1, wherein the determination device determines that the condition is satisfied when the duration of the event is equal to a predetermined time. Wear monitor. 4. The hardware according to claim 1, wherein the determination device determines that the condition is satisfied when the duration of the event is equal to or longer than a predetermined time. monitor. 5. The method according to claim 1, wherein the predetermined duration is specified by a range, and the determination device determines that the condition is satisfied when the duration of the event is included in the range of the predetermined duration. The hardware monitor according to claim 1 or 2, wherein 6. The memory has the following two measurement modes: (1) Every time a specified event occurs in the processor, an increment of a value stored in a storage area determined according to a duration of the event occurs. Time distribution measurement mode (2) Every time an event specified in the processor occurs, the counter is incremented, and the occurrence time history measurement mode for storing the duration of the event in the storage area specified by the counter is switched. The hardware monitor according to claim 2, wherein the hardware monitor can be used.