JPH0436420B2 - - Google Patents

Description

[Detailed description of the invention]

[Industrial Application Field] The present invention relates to a hang-up timer method for detecting a hang-up of a microprogram or the like. [Prior Art] In a microprogram central processing unit (CPU), a loop may be formed during program execution, and it may become impossible to escape from the loop (hangup).This is detected and recovery processing such as initialization is performed. A hang-up timer is provided so that a certain part of the program can be executed, and when a certain part of the program is executed, an appropriate time determined according to the time required for execution of the part is set in the hang-up timer and the time is started to be measured. It is checked whether the timer times out before the execution of the part ends (if it times out, the program hangs up). However, in the conventional system, there is only one means for setting the timeout detection time of the hang-up timer. For example, a register for setting a timeout detection time is provided, and the determination of the timeout detection time is changed by changing the contents of the register. After a change is made, a timeout is detected at the changed time, and the contents of the register must be changed again to restore the original value. Therefore, in the conventional method, the timeout time is set to match the machine instruction or microinstruction that takes the longest processing time, and the timeout setting time is not dynamically changed for each machine instruction or microinstruction. [Problems to be Solved by the Invention] However, since some machine instructions or microinstructions require a particularly long processing time, it may be desirable to set a particularly long timeout for that particular machine instruction or microinstruction. Of course, if you set a long timeout setting overall in such a case, the setting time will be too long for machine instructions or microinstructions that have short processing times, and hangup detection will be delayed. requires two processes: modification and restoration. Therefore, in addition to the conventional timeout detection time setting means, the present invention provides means for temporarily changing the timeout detection time, so that the setting of the timeout detection time can be dynamically changed in units of machine instructions or microinstructions. This is what we are trying to do. [Means for Solving the Problems] The hang-up timer method of the present invention includes register means for setting a basic timeout detection time in a timer for detecting hang-up of a program;
It uses a counter means that is cleared at the end of execution of an instruction or is set with write data when an external write signal is input, and is incremented by a clock, and a comparison means that compares the contents of these register means and counter means. The comparison means compares the contents of the counter means, which is cleared at the end of execution of the instruction, and the register means having the basic timeout detection time, and a timeout is detected at the basic timeout detection time, and the program writes the predetermined write data. When the timeout detection time is changed by writing to the counter means, timeout is detected using the changed detection time, and when the counting means is cleared by the completion of instruction execution, timeout is automatically detected using the original basic timeout detection time. It is characterized by returning to the action of [Operation] In addition to the hang comparison register that sets the basic timeout detection time, the use of a hang count register that is cleared upon completion of instruction execution makes it easy to temporarily change the timeout setting time. [Example] FIG. 1 shows an example of the present invention. This example is an example of a timer that detects a hang-up in units of machine instructions. NCNTR is the hang count register,
HCMP is a hang comparison register that holds data to be compared with register HCNTR, INC is a register
Incrementer that adds 1 to the contents of HCNTR, OR ₁
is an OR gate for the clock enable condition of register HCNTR, _A2 is an AND gate that synchronizes the output of OR gate _OR1 with the system clock, and MPX is a multiplexer for the input data of register HCNTR. Multiplexer MPX and gate
A _3, A ₄ and an OR gate OR ₂ are provided for the number of data bits, and when the clock ISE generated every second, which is the basic timeout detection time, is input through the AND gate A ₄ , the output e of the incrementer INC is
is selected, micro write data d is selected when the external write signal EXTW is input, and when the return signal RTN is input, AND gates A3 _{and A4} are closed and data "0" is output from the OR gate _OR2 . Also, COMP is register HCNTR and HCMP
FFa is a flip-flop that is set by the match comparison result of comparator COMP and reset by return signal RTN. FFb is an AND gate _A1 when flip-flop FFa is set and clock 1SEC is input.
This is a flip-flop that is set by the output of , and the output h at this setting becomes a timeout signal. Clock 1SEC is a developing device as shown in Figure 2.
Obtained by dividing the OSC output using a frequency divider circuit FD.
The external write signal EXTW is output (becomes "1") when the microprogram writes data to the register HCNTR. A return signal RTN is also generated at the end of a machine instruction. In other words, the machine instructions are D (decode cycle) and A as shown in Figure 3.
(address calculation cycle), an i-cycle consisting of T (TLB access cycle), and a plurality of E
It is executed in an E-cycle consisting of (execute; microinstruction execution), and a flag is added to the last E to indicate that one machine instruction is finished, and this becomes the return signal RTN.
Note that τ is one machine cycle, and one microinstruction is executed within this time. A CPU that uses a pipeline system prefetches instructions, and as shown in Figure 3, as soon as the execution of machine instruction 1 is completed, the next machine instruction 2 is executed.
execution begins. At the end of execution of machine instruction 1, a return signal RTN is generated, and the flip-flop shown in Figure 1 is reset and the register is reset.
HCNTR is cleared to 0. register HCMP
A basic timeout detection time is written in advance by a microprogram. The basic timeout detection time is 1 second in this example, and data 0001 (micro write data or SVP setting data g) indicating this 1 second is taken into the register HCMP. Machine instruction 2 is executed in this state,
The execution ends before 1 second elapses and a return signal is sent.
If RTN occurs, the flip-flop is reset and the register is cleared, and no timeout signal is generated. On the other hand, if clock 1SEC occurs before return signal RTN is generated, register HCNTR is
The output e of the incrementer INC, which adds 1 to the content 0000, ie, 0001, is loaded into the register HCNTR through the multiplexer MPX, and the output a of the register becomes 0001. Since this is the same as the output b of the register HCMP, the comparator COMP sets the output c to "1", which sets the flip-flop FFa. When another second passes in this state, the next clock 1SEC occurs, and at this time, the output of AND gate A becomes "1", and flip-flop FFb
Set. Therefore, the output h of the flip-flop FFb becomes "1", indicating timeout. The reason why the flip-flop is configured in two stages is to ensure hang-up detection with a margin of one second. The above is the normal (basic) operation, but if you want to change (lengthen) the timeout setting time, prepare the required data d, generate the signal EXTW, and load it into the register HCNTR. Specifically, the hang count register HFNTR is a 4-bit counter in this example that can be preset.
After 1111, it returns to 0000. Therefore, if 1111 is prepared as micro write data d, the register
The contents of HCNTR are initially 1111, clock 1SEC
When clock 1SEC is input, it becomes 0001, and when clock 1SEC is input again, it becomes 0001, which is now equal to the contents of register HCMP. In this way, the timeout setting time has been changed to 2 seconds. If 0010 is prepared as the micro write data d, a match will be found when it reaches 0001 after one cycle, so the timeout setting time has been changed to 15 seconds. Moreover, if the signal EXTW is not generated when the next machine instruction is executed, the register HCNTR will be cleared to 0 by the return instruction generated at the end of the execution of the instruction, returning to the original basic timeout setting time, and special processing will be performed for this return. is not necessary. The following table shows an example of how to use this circuit.

〔Effect of the invention〕

As described above, according to the present invention, it is easy to temporarily change the timeout setting time, and the timeout detection time can be dynamically changed in units of machine instructions or microinstructions.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing an embodiment of the present invention, FIG. 2 is an explanatory diagram of a clock generation circuit, and FIG. 3 is an explanatory diagram of machine instructions. In the drawing, HCMP is a register means for setting a basic timeout detection time, HCNTR is a counter means, and COMP is a comparison means.

Claims (1)

[Claims] 1 A register means for setting a basic timeout detection time in a timer for detecting a program hang-up, and a counter means that is cleared at the end of execution of an instruction or set with write data when an external write signal is input, and incremented by a clock. , a comparing means for comparing the contents of the register means and the counter means, and comparing means for comparing the respective contents of the counter means, which is normally cleared upon completion of execution of the immediately preceding instruction, and the register means having a basic timeout detection time. Detect a timeout with the basic timeout detection time by comparing
When the program writes predetermined write data to the counter means and changes the timeout detection time, a timeout is detected using the changed detection time, and when the counter means is cleared by the end of instruction execution, the original basic timeout is automatically performed. A hang-up timer method characterized by returning to the operation of detecting a timeout based on the detection time.