JPH0488435A - Measurement system for total execution clock quantity by instruction - Google Patents

Abstract

PURPOSE:To measure the total clock quantity from the execution start to the end of a couple of instructions and the clock quantities by the instructions by adding the output of a clock measuring means to an integral value storage area only when an instruction decision means decides a coincidence. CONSTITUTION:The instruction decision means 13 decides whether an instruction being executed currently is the same as an instruction in a measurement instruction storage area 13 in an instruction emulation figure 17. A clock quantity measuring means 14 measures the clock quantity from the execution start of an instruction in an instruction emulation part 11 to the execution start of a next instruction at all times. Then a clock quantity integrating means 15 adds the output value of the clock quantity measuring means 14 to an integral value storage area 16 only when the instruction decision means 13 decides that the instruction being executed currently is an instruction to be measured. Consequently, the load rate of a CPU, i.e. the time of effective processing by the CPU and the execution time of stand-by instructions in each unit time are found by the measurement.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a method for measuring the total number of execution clocks per instruction, and particularly to a method for measuring the total number of execution clocks per instruction in an in-circuit emulator for microprocessor system development.

[Conventional technology]

In the prior art, there was a means to measure the total number of clocks from the start of execution to the end of effective execution of instructions in the - group, but it was impossible to measure the number of clocks on the instruction side.

[Problem to be solved by the invention]

In the development of microprocessor systems, system performance, especially processing speed, is a very important evaluation factor. - When it comes to making a deal, it is not just about performance that is fast; the optimum speed for the system is determined by the trade-off with cost. A commonly used method for optimizing processing speed is to measure the various load distributions of the current system and improve the weak points. At this time, measurement of the following values is very important.

(1) CPU load factor, that is, the time during which the CPU performs effective processing within a unit time.

(2) Load distribution among coprocessors in a system using one or more coprocessors.

That is, the processing time of each coprocessor within a unit of time.

The above (1) can usually be obtained by measuring the execution time of a standby instruction (hold instruction, etc.), and the above (2) can be obtained by measuring the processing time of an instruction for each coprocessor. However, in-circuit emulators used in the development of microprocessor systems have traditionally lacked the ability to measure the total processing time on the instruction side, and have had the disadvantage that such measurements cannot be easily performed. .

[Means to solve the problem]

The method for measuring the total number of execution clocks on the instruction side of the present invention includes: measurement instruction registration means for registering measurement target instructions in a measurement instruction storage area; instruction determination means for determining whether the currently executed instruction is the measurement target instruction; a clock number measuring means for measuring how many clocks are required for execution of the instruction; and a clock number integrating means for adding the output of the clock measuring means to an integrated value storage area only when the instruction determining means determines that there is a match. It consists of:

〔Example〕 FIG. 1 is a block diagram showing one embodiment of the present invention.

When the measurement instruction registration means 11 receives an instruction to be measured from the console of an in-circuit emulator, it converts it into a bit string of a machine language instruction of the target CPU and stores it in the measurement instruction storage area 12. This conversion is a well-known conversion called "assembly."

The value of each bit of the converted bit string is "1" or "0"
or "1 or 0 is fine." ``1,
0 or 0 is used when the value of the address part of the instruction can be arbitrary.

The measurement instruction storage area 12 is one area on the memory, and stores the bit pattern of the machine language of the instruction to be measured.

The instruction determining means 13 determines whether the instruction currently being executed in the instruction emulation diagram 17 and the instruction in the measurement instruction storage area 13 are the same. The determination is a comparison of two bit strings, and the method is known.

The clock count measuring means 14 constantly measures the number of clocks from the start of execution of an instruction in the instruction emulation unit to the start of execution of the next instruction.

The clock number accumulating means 15 only when the instruction determining means 13 determines that the currently executed instruction is the measurement target instruction.
The output value of the clock number measuring means 14 is stored in the integrated value storage area 16.
Add to.

The integrated value storage area 16 is an area reserved on the memory, and the data therein represents the total number of clocks of the instruction to be measured at the present time.

〔Effect of the invention〕

The present invention has the advantage that the following performance evaluation regarding a microprocessor system can be easily performed.

(1) CPU load factor. In other words, within a unit time, the CPU
is the time during which it is performing effective processing. It is obtained by measuring the execution time of a standby instruction (hold instruction, etc.).

(2) Load distribution among coprocessors in a system using one or more coprocessors.

In other words, the processing time of each coprocessor within a unit time. This is determined by measuring the processing time of instructions for each coprocessor.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing one embodiment of the present invention. DESCRIPTION OF SYMBOLS 11... Measurement instruction registration means, 12... Measurement instruction storage area, 13... Instruction determination means, 14... Clock number measuring means, 15... Clock number accumulating means, 16...
integrated value storage area, 17...instruction emulation section;
18...Clock.

Claims (1)

[Claims] Measurement instruction registration means for registering the instruction to be measured in a measurement instruction storage area, instruction determination means for determining whether the instruction currently being executed is the instruction to be measured, and a clock for measuring the number of clocks required to execute each instruction. and a clock number accumulating means that adds the output of the clock measuring means to an integrated value storage area only when the instruction determining means determines that there is a match. method.