Programmed Controller

1. Field of the Invention

This invention relates to a programmable controller for processing finite state applications programs which control the states of operation of machines and/or processes. The invention also relates to debugging a machine and/or process which allows for easy diagnostics of the finite state applications program.

2. The Prior Art

In the 1960's, programmable controllers were developed to replace relay control systems which used electromagnetic relays to control the operation of industrial machines or processes. These programmable controllers were designed to work as if they contained relay circuits, even though the circuits were actually programs in a computer. The relay circuit designs, emulated by these programmable controllers, were called ladder diagrams because of their general appearance. Ladder diagrams are excellent for designing combination logic control problems, where outputs or actions are directly dependent on the states of inputs or conditions. However, for sequential control problems, where the control actions are time-dependent, the ladder diagram approach becomes cumbersome, difficult to design and fault-find. Lloyd, "Graphical Function Charts Programming for Programmable Controllers," Control Engineering (October 1985).

In order to emulate ladder diagrams, a programmer strings together long lists of Boolean equations in the form of software commands either in text form or graphically. Each command controls an internal or external signal. Debugging or fault-finding in such a system usually involves observing an unexpected condition, and then intuitively searching back through some or all possible combinations of factors that could have caused the unexpected condition. Because all of the ladder diagram is scanned each scan cycle, it is not possible to eliminate any part as definitely not containing the problem.

Another drawback of the ladder diagram controllers is that in order to determine the causes of the unexpected condition it is desirable to record information related to the cause of the problem. However the ladder diagram system does not have any mechanism to differentiate between relevant and irrelevant changes in system variables and conditions, and this leads to the need to store an impractically large amount of data which would be very difficult to analyze and sort.

Furthermore, ladder diagram controllers are cumbersome to program. By their very nature, ladder diagrams are difficult to understand. Additionally ladder diagrams do not clearly show the broader, macro functionality they are designed to provide, particularly sequential functions. What they show is in effect a circuit that provides that functionality. For many years, machine design and manufacturing engineers, who were unfamiliar with ladder diagrams, could not easily develop control systems for their machines or processes. Instead, they relied on specialized control engineers to program their control systems.

In an effort to simplify programmable controllers, systems were developed which enable engineers to use high level languages to emulate "state diagrams." A state diagram is a graphical representation of the states of operation and the conditions which change the status of the states of operation in a machine or process. In graphical form, a state diagram is a series of nodes with interconnecting arcs. Each node represents a particular state of operation and each connecting arc represents a transition function which causes the status of the state to change. State diagrams are easy to understand and easy to design by all types of engineers. The current systems which emulate state diagrams, however, are difficult to program, and difficult to debug.

An example of one such system is disclosed in U.S. Pat. No. 4,562,529 to Drummond. Drummond discusses a method and apparatus for controlling the states of operation of a machine or process by using "state logic variables." A "state logical variable" uniquely represents each operating state of the machine or process. The "state logical variable" can only be one of two values--either TRUE or FALSE, TRUE meaning that the state is active and FALSE meaning that the state is inactive. By using the methods as disclosed, the Drummond system operates very similarly to the ladder diagram controllers. Several problems are present in the Drummond method and apparatus.

First, the entire finite state applications program must be cyclically scanned in order for the Drummond system to properly process the "state logical variables". To make assignments to state variables, which is necessary to make changes of state, or to carry out state dependent actions, Drummond must evaluate many expressions containing state variables. The Drummond patent discusses simplified ways for doing this, however, the process is still inefficient and can be eliminated altogether.

Second, programming the Drummond controller to emulate state diagrams is a cumbersome process. The language disclosed in Drummond does not lend itself to easily converting the state diagrams into program code. This occurs because the condition statements which affect the status of the "state logic variables" cannot be lumped into specific areas of program code and require explicit inclusion of state variable terms on the left hand sides. Additionally, virtually every statement in the code requires a "state logical variable" to be assessed in order to determine whether an state assignment should be made or whether an action should be carried out. As a result, much extra code must be drafted in order to perform simple functions.

Lastly, because every statement in the finite state applications program must be scanned in order to evaluate the status of the state logical variables, the system operates very inefficiently. This drawback severely limits through put and limits the recording of system activity information for debug purposes. Also, Drummond does not disclose an adequate strategy or structure for keeping track of the multiple operating tasks to be controlled simultaneously.

Note should also be taken of the commercially available GRAFCET systems based on French standard NF C 03-190. GRAFCET does not restrict a state diagram to only having one active state at one time and so knowing the identity of one active state in a GRAFCET state diagram does not totally define the conditions and operations associated with that state diagram at that time. This has significant disadvantages during program debugging.

Thus PLC's have been programmed by a means known as ladder diagrams, or an equivalent means, from the early days. As more and more functionality has been added to the language (in the form of available ladder diagram symbols and structure), it has become apparent that there are limitations to the ladder diagram that can only be addressed by a change to the basic approach to programming. This is particularly apparent when PLC's are used for the control of machinery in flexible automation schemes. The functioning of such schemes relies significantly on the ability of the controller to sequence the machinery in a complex and flexible way with regard to both the activities carried out and the functions that step the sequence on and with regard to the ability of the PLC to control many essentially independent sequential processes at the same time. Such things can and are done with ladder diagram PLC's. However such programs require a translation process to generate a ladder diagram from the definition of the sequencing which may be in the form of a written specification or a flow chart. When the ladder diagram is inspected, the functions carried out are not obvious and the program and the process of debugging the program are cumbersome. Much of the information inherent in the spec or flow chart is lost or obscured in the translation process which converts the functional requirements into a relay circuit that fulfills those requirements.

Over the last few years there has been much interest in improving this situation. Sequencing methods have been developed which aim at improving the generation, debugging, and comprehension of such programs. The French GRAFCET effort has been notable and this has been accepted internationally as a useful technique. Related French and West German standards exist. Several PLC's using variations of the GRAFCET methods of describing sequences using standarized flow charts are on the market. However, there are further issues to address to reach an optimum form of control systems programming and these are not addressed by any currently known system.

The problems that need addressing call for improvements in

program structure and form

focuses on relevant

simplifies understanding of complex systems

suited to high levels of functionality

minimization of transformations

avoidance of side effects

cost of hardware

self-explanatory to widest range of people and user friendly

minimized training

maximum assistance with debugging

The present invention seeks to provide a control method and apparatus that reduces or eliminates the disadvantages referred to above, and is more readily and freely used by control, manufacturing and design engineers.

The following is a preliminary explanation of terms which will assist the reader to more fully comprehend the nature of the invention that is described in this specification.

1. "State(s) of operation" means a particular combination of conditions and operations which exist for a finite time interval in a machine or process; only a subset of the machine's or process' signals or values are at particular defined values. The "state of operation" of a machine or process can be thought of as being a particular task or subset of tasks presently performed by the machine or process.

2. "Controller" means a device for exercising control over a process or machine or activity, which includes means for discerning the "state of operation." For this purpose, the controller must be kept informed of variations in the values and the status of the machine or process.

3. "Program state" means a particular block of one or more statements of program code used for evaluating, maintaining and controlling a "state of operation."

4. The terms "machine" or "process" are systems which proceed in real time from one specific operating state to a next operating state. The controller reads signals from the machine or process, monitors the operation of the machine or process, tracks the necessary operating states by way of its own informal states, and provides necessary output signals to put the machine into the necessary operating states based on the algorithms embedded in the applications program.

5. The term "multi-tasking" means a method of asynchronously running separate programs or separate tasks contained in the same program by only one controller such that all appear to be running simultaneously with relation to one another.

It is the concept of a "state" that allows one to say that a machine, or a section of a machine, is in a particular state, and that allows the structuring of a program so that a state is defined in one contiguous block only within the program with some form of state delimiting statement or means. From this follows the ability to say that while a particular state exists, the only possible relevant part of the program related to that loop is the block defining the state. A class of outputs is provided as part of the language where no outputs of this type exist from a system, other than those defined in the states active in each loop. This type of output is typically used for activating motors or valves. A state can therefore also be said to exist in the sense that it produces real world outputs while it exists, and these outputs are functions of the state only, i.e., they are activated conditional on nothing other than the existence of the state.

It may be advantageous to allow qualification of those outputs based on other conditions present in the system, but this would be an optional extension of the system.

It should be noted that it may be advantageous to also provide a class of output that can be activated in one state and deactivated in another, used for instance for driving displays or registers.

It is also the existence of explicit state type statements that allow subclasses such as "break state" and "pause state" to be provided for the purpose of controlling the flow of the program.

Debugging and comprehension of the control activity can therefore be based on a "state structure" as follows

an identification of the states that exist and what major control function they provide. E.g., while a state may activate several outputs, do some timing, interrogate inputs, print, and eventually take the decision to change to another state, it can also be said to carry out some macro function; e.g. the controller for an elevator might have a state that performs the macro function "go up one floor," which consists of activities like those noted above. The state could therefore be called GoUpOneFloor as a single word symbolic name in our system, and this provides structure to the program and broad understanding without the need to consider the lower level operations necessary to carry out the macro function.

examination of the flow of control through the various states that existed previously, that ultimately led to the overall existing state of the machine (i.e., tracing).

enabling the flow of control until a particular state is reached, i.e., trapping with a break state, with subsequent diagnosis of the activity to that stage, or pausing a particular loop with a pause state.

It can therefore be seen that it is most advantageous for a loop, program, group program or whatever it is called to be able to be said to be in some state, and for that state to be defined unambiguously, and to be a function that can be said to exist or not, that can be detected and/or recorded.

For tracing to be effective, the concepts of states and loops must both exist.

Briefly, a preferred embodiment of the invention includes a programmed controller for emulating one or more state diagrams and for controlling the states of operation of at least one machine or process. The controller performs an application-type program which comprises a plurality of blocks of statements. Each of these blocks of statements when executed is a program state defining an internal control or state which corresponds to one of the machine or process states of operation. Each block of statements defines the next blocks of statements one of which may replace the existing block as the active block if the defined transition conditions are satisfied and actions to be taken during the corresponding state of operation by the machine or process.

Program states are organized into one or more program loops which represent independent sequential control schemes for an entire machine or process or for a sub-section of the machine or process. Each program loop essentially emulates a state diagram.

The controller can activate one or more program loops simultaneously, however, each program loop can only allow one of the program states within the program loop to become active. To keep track of the active program loops and the associated active program states, a variable is associated with each active program loop. The variable stores an identifier of the active program state or the address of the start of the active program state. The variable also has an identifier which indicates within which active program loop the particular program state resides. The controller also has a task table which stores the variables for the active program loops. The task table enables the controller to perform several active program loops asynchronously and it determines the sequence of execution of the active program loops. Essentially, the task table keeps track over which program loops and which program states are active and of the order and sequence in which each program state will be processed.

A method and apparatus are also provided for facilitating "debugging" of the finite state applications program. Means are provided for recording information about the activity of each of the active program states. Specifically, the controller traces a history of the order and occurrence of each program state and the conditions responsible for each occurrence of the program state. Display means are also provided for enabling the programmer to evaluate the recorded information.

Significantly, the present invention allows a control, manufacturing or machine design engineer to more freely understand and even design high-level applications programs representative of state diagrams. The specific design advantage of the present invention is that the programmer is able to organize all program statements related to a particular state of operation into a contiguous segment of program code or block of statements and none of the statements are related to any other state of operation. In this way, the programmer is able to compartmentalize all of the various conditions dependent on each state of operation without having to worry about any of the other states of operation. This procedure does away with "state logical variables" and it facilitates techniques for programming in an organized and structured format. Furthermore, program loops representative of overall state diagram structures are easily designed into code by simply connecting the compartmentalized code together with the "conditional GOTO statements" or the like contained within program state blocks.

The present invention is more efficient than the prior art because all of the statements in the applications program do not have to be scanned each time a particular state of operation needs to be controlled. Instead, only those statements in the particular program state associated with the particular state of operation which is active are scanned. This technique enables the controller to process very quickly and it allows the present invention to more efficiently process a multitude of tasks at the same time. The preferred embodiment of the invention also has a task table for keeping track of the various sequences of tasks to be performed simultaneously and/or determining the order in which each sequence of task will be performed.

Lastly, because the applications program is designed with program states, pinpointing the causes of unexpected events is greatly facilitated. Means are provided for recording the sequential order of occurrence of each active program state as they occur and one or more conditions responsible for transitions occurring for each program state to the next or responsible for initial activations of program states. By analyzing a display of the occurrences of each of the program states and by analyzing the conditional statements causing the transitions to occur, the analyst can quickly determine the location and the cause of the unexpected event in the program.

FIG. 1a is a diagram of the simple apparatus and controls which the invention is applicable.

FIG. 1b is a ladder diagram for a state of the art controller for the apparatus and controls of FIG. 1b;

FIG. 1 shows a programmable controller for controlling the operations of a machine or process in accordance wih the present invention;

FIG. 2a is a state diagram of the various states of operation of an elevator and the conditions which cause transitions to occur between states;

FIG. 2b is a diagram of blocks of statements corresponding to the states of operation of the state diagram in FIG. 2a;

FIG. 2c is a diagram emphasizing the structure of the program code;

FIG. 2d is a more complicated state diagram;

FIG. 3a is a diagram of a program loop pointer;

FIG. 3b is a diagram of the task table;

FIG. 3c is a diagram of a task table pointer;

FIG. 4 is a logical representation of a trace table shown in the form of an endless loop;

FIG. 5 is a schematic block diagram of the subroutine called START PROGRAM;

FIG. 6 is a schematic block diagram of the subroutine ADVANCE P-CODE POINTER;

FIG. 8 is a schematic block diagram of the subroutine ADVANCE TASK TABLE REGISTER;

FIG. 9 is a schematic block diagram of the subroutine NEXT OPERATOR;

FIG. 10 is a schematic block diagram of the subroutine GO TO;

FIG. 11 is a schematic block diagram of the subroutine START;

FIG. 12 is a schematic block diagram of the subroutine GO SUBROUTINE;

FIG. 13 is a schematic block diagram of the subroutine RETURN;

FIG. 14 is a schematic block diagram of the subroutine called STORE HISTORY;

FIG. 15 is a schematic block diagram of the debugging routine called DISPLAY HISTORY;

FIG. 16 is a schematic block diagram of the debugging routine called DISPLAY LINE;

FIG. 17 is a schematic block diagram of the debugging routine called DISPLAY PROGRAM STATE;

FIG. 18 is a schematic block diagram of the debugging routine DISPLAY LOOP; and

FIG. 19 is a schematic block diagram of the debugging routine FOUND LOOP.

Before considering the description and implementation of the languge interpreter software, we will describe the principles on which the language is based and give examples of how programs according to the invention are constructed.

The Loop-State structure is the single most important feature of the language system which enables simple management of the complex and changing requirements for control in complex systems.

The simple concepts are as follows. Consider an elevator such as may be used to raise/lower a platform on a building site. It has an Up button and a Down button at the top and on the ground. It can be said to be in the states of "stopped," "going up," or "going down," but only in one of those at any one point in time. Its operation can be described with the help of the state diagram in FIG. 1a.

The following should be noted:

the states with their numbers and names included in circles.

the transition conditions define in words (shown beside the arcs) which detail the conditions that cause the state to change.

the way in which a state diagram description minimises and defines what is relevant. E.g., in "Going up," what is happening to the "Down buttons," 1 or 2, is quite irrelevant.

Similarly the way all relevant activities are shown associated with the state they relate to, e.g., "Energize motordown" in "Goingdown."

The ease with which the operation of the elevator can be understood, e.g., if it is stopped and up button 1 is pushed, it goes up, but if it is going down and up button 1 is pushed, it ignores it.

The program using our language which is functionally equivalent to the above state diagram is:

Stopped

Brake on

if Upbutton1 on or UpButton2 on goto GoingUp

if DownButton1 on or DownButton2 on goto GoingDown

GoingUp

UpMotor on

if UpButton1 off and UpButton2 off goto Stopped

GoingDown

DownMotor on

if DownButton1 off and DownButton2 off goto Stopped

The only parts omitted from this program are defining statements that allocate particular names to particular variables, i.e., that define UpButton1 is connected to input terminal 21.

The ladder diagram equivalent is as shown in FIG. 1b In the ladder diagram:

] [ is a normally open contact

]/[ is a normally closed contact

() is a relay coil

Note how the circuit operation is relatively easy to follow. What it actually does functionally is far harder to deduce. There is no structure to show that it is relevant at any one point in time.

A state diagram with 100 states instead of 3 still has states as simple as the 3, and each may be understood in turn. A ladder diagram with 100 rungs is a nightmare.

The principles of our loop-state structure are:

Programs are organized into (one or) more than one separate loops that represent independent sequential control schemes for sections of the machine. A loop can be processed independently of the rest of the program, or can handshake and synchronize with other loops via flags, etc.

Loops are formed of sequences of states. I.e., they are in effect state diagrams representing state machines having outputs which may be conditional/unconditional on the inputs, etc.

Each state machine, i.e., loop, has one and only one state active at any point in time that the loop is active. If the loop is not active, then there is no state active in that loop. Active loops can therefore be said to possess a state variable, the value of which is the number/name of the state active at that point in time. This can be accessed and used for handshaking, diagnostics, etc. This is also the mechanism that focuses the observer onto a particular part of the program as being relevant for a particular purpose, and EXCLUDES the possibility that any other part is relevant. This exclusion is most important because it says "look here, and DON'T LOOK ANYWHERE ELSE."

A class of outputs is provided as part of the language where no outputs of this type exist from a system, other than those defined in the states active in each loop. This type of output is typically used for activating motors or valves. A state can therefore also be said to exist in the sense that it produces real world outputs while it exists, and these outputs are functions of the state only, i.e., they are activated conditional on nothing other than the existence of the state. The output setting statement in the language effectively says "While in this state this output is on and it will go off when this state becomes inactive. It does not say "set this output on," and does not need any additional statement to turn the output off again.

Transition conditions are a function of the state they cause to be deactivated. They say "change the active state from this to another," where the other is given as a parameter of a "go to" statement. They therefore appear in the state block of which they are apart.

State blocks are contiguous segments of program, starting with a state statement and ending at the next state statement. They contain all the code related to that state, and none related to any other.

We see therefore that:

for one section of a machine, only one loop is relevant,

for that section of the machine at a particular point in time in the sequencing of the section, only one state in that loop is relevant.

We have therefore focused down in the program to a segment representing one state, probably, e.g., 10 lines out of 2000, and we can be sure that that is the only part that is relevant.

To do this we need only determine the section of the machine associated with the problem or other event of interest--which is physical association only, and the point in the sequence that the problem occurs. Trapping and tracing provide the means for determining the point in time in the sequence, if this is not obvious from the program. The specifics of trapping and tracing are described in our other patent brief.

Documentation is made easy by the textual form of the program, as opposed to a graphical form.

FIG. 1 depicts a programmable controller and a computer program which form programmed controller 10, for emulating state diagrams. A finite state applications program is stored at 14 in the memory portion 10A of the programmed controller 10. A central processing unit ("CPU") 70 performs mathematical and logical functions for the finite state applications program ("applications program"). A task table or set of pointers 68 keeps track of active program loops. A trace table 72 contains a temporary history of the activity of the applications program. The programmable controller is equipped with a debugging monitor 74 for enabling a user to easily pinpoint errors in the applications program. The trace table 72, the task table 68, and debug monitor 74 are all stored in memory 10A. Other operating system software for controlling miscellaneous features of the programmable controller are stored at 76. Terminal 80 is provided to enable a user of the controller to program the applications program and debug the system on terminal 80.

FIG. 1 shows a programmed controller 10 connected to machine or process such as elevator 12. The programmed controller 10 emulates the applications program at 14, for controlling the operation of the elevator 12. Finite state applications programs are high-level programs representative of state diagrams (See FIG. 2a). State diagrams are pictorial representations of the states of operation of the particular machine or process to be controlled. A programmer controls the machine or process represented by the state diagram by first designing a finite state applications program and, second, processing it on the control to emulate the state diagrams and control the order and sequence of occurrence of each task or operation to be performed by the machine or process.

A simplified application of the programmed controller 10 is monitoring, evaluating and controlling the operations of elevator 12. Elevator 12 consists of a reversible motor and brake 16, pulley 20, and an elevator compartment 18. A fundamental principal of the elevator is that only one state of operation of the elevator can be activated at any one point in time. The elevator's states of operation are "Stopped", "Going Up", or "Going Down".

Buttons 22 in elevator compartment 18, cause the elevator to change from one state of operation to another state of operation. For example, when the "Up" button is closed and the "Down" button is open, the elevator compartment 18 goes up. When the "Down" button is closed and the "Up" button is open, the elevator compartment 18 goes down. When the "Up" button is pressed while "Going Down" or when the "Down" button is pressed while "Going Up" the elevator compartment 18 is stopped. FIG. 2a is a state diagram or graphical representation of these states of operation and the conditions which effect changes to occur from one state to another. More particularly, a state diagram is a series of nodes 24, 26 and 28 with interconnecting arcs 30, 32, 34 and 36. Each node represents a particular state of operation and each connecting arc represents a condition or transition function which causes the status of a particular state to change.

Specifically, the state diagrams in FIG. 2a have nodes 24, 26 and 28 which respectively represent the states of operation "Going Up", "Stopped", and "Going Down". The conditions which affect the status of the states are the arc connectors 30, 32, 34 and 36 between the nodes. Each arc defines its own set of transition conditions, origin and destination state. Arc 32 has switch condition and action 42:

If "Up" button is on and "Down" button is off then state of elevator 18 goes to "Going Up."

Arc 30 has switch condition and action 38:

If the "Up" button is off or "Down" button is on, then state of elevator compartment 18 goes to "Stopped."

The rest of the conditions and actions can be discerned from FIG. 2a.

An applications program representative of the state diagram in FIG. 2a is:

"Going Up"

"Up" Motor On

If "Up" button off or "Down" button on, go to "Stopped"

"Stopped"

Brake On

If ""Up" button on and "Down" button off go to "Going Up"

If "Up" button off and if "Down" button on, go to "Going Down"

"Going Down"

"Down" Motor On

If "Down" button off, or "Up" button on go to "Stopped"

The three blocks of statements above are called program states and each block corresponds to both a particular state of operation of the elevator 12 and to a particular node in the state diagram (FIG. 2a). Each of the program states consists of a state delimiting statement, "Going Up", "Stopped", or "Going Down", which signifies to the controller the end of one program state and the beginning of a new program state. The state delimiting statement is followed by a series of statements consisting of none, one or more compound statements which can cause actions or cause transitions to occur from a currently processing or active program state to another program state. A compound statement consists of a condition part and an action part. If the condition part is satisfied, then the action part is executed. If the action part contains a GOYTO, GOSUB or RETURN statement, a transition to a new program state occurs. If the action part contains a START statement, then a new program loop is initiated with an additional newly active state. Stated differently, the compound statements determine what the program state changes or initiations will be and the type of actions to be taken by the machine or process during the currently active state of operation. All three program states 24, 26 and 28 above have compound statements which include "GoTo" which cause the program states to shift processing to another program state in the same state diagram.

By gathering all statements which are associated with a particular state of operation into only one contiguous segment of code, the system can scan only these statements to initiate actions required by the state of operation of the machine or process and to determine if a particular set of conditions has been satisfied in order to initiate a new state of operation. Whereas, the Drummond system evaluates every statement in the applications program to determine whether to change state activity or carry out an action, and in the course of this has to evaluate many state variables. The structural difference in the program enables the present invention to operate much more efficiently by eliminating the need to evaluate state variables.

FIGS. 2b and 2c are pictorials of the program states. Block 48 in both figures refers to the sequence of code for the program state "Going Up"; block 50 represents the sequence of code for the program state "Stopped", and block 52 represents the sequence of code for the program state "Going Down". Arcs 54, 56, 58 and 60 (FIGS. 2b and 2c) correspond to the arcs 30, 32, 34 and 36 of the state diagram (FIG. 2a). Arcs 54, 56, 58 and 60 show how the flow of activity in the applications program can be affected by the compound statements.

Arc 58 of FIG. 2b demonstrates that when the condition part of compound statement:

If "Up" button off or "Down" button on, go to "Stopped" is true, then a transition from the "Going Up" program state to the "Stopped" program state occurs. Likewise, arc 60 demonstrates that when the condition part of compound statement:

If "Up" button off and if "Down" button on, go to "Going Down",

is true, then a transition from the "Stopped" program state to the "Going Down" program state occurs.

The connections between the program states "Going Up" "Stopped" and "Going Down" logically create a program loop 46. The program loop 46 controls the sequence of operations to be performed by the elevator 12. In the preferred embodiment, a program loop can only have one active program state at any one time. In this way, the controller can keep track of the sequential processing of an individual scheme of tasks. The program loop 46 in FIG. 2b exhibits the characteristic of having only one program state active at any one time in order to make the algorithms involved in program execution and debugging more efficient and to reduce ambiguity in the system.

In more complex machines, a single program loop will only represent the control of a particular independent sub-part of the machine or process. Usually, multiple program loops are required to independently control several sub-parts of a complex machine or process. FIG. 2d, shows a more complex machine represented by a plurality of state diagrams 62, 64 and 66. Each state diagram represents a program loop which depicts the sequential control scheme for a section of the machine or process. The present invention emulates these diagrams by processing program loops which control three hypothetical sub-parts of a machine asynchronously (also called multi-tasking). A particular program loop can be processed independently of the rest of the program loops or it can "hand shake" and synchronize with the other program loops using some convenient mechanism such as flags or a statement of the type "if loop N at State M," which tests the value of the active state in a different loop. This multi-tasking capability of the controller is largely transparent to the programmer and the operating system and language provide simple means to implement it.

To facilitate the execution of program states within a program loop, the preferred embodiment provides a program loop-record for each of the active program loops. Referring to FIG. 3a, the program loop-record 90 has fields 92 and 94 which hold simple variables. The first simple variable 92 is an identifier of the program loop. The second simple variable 94 is an identifier of the active program state or an address in the memory of the start of the active program state. Both the address or the identifier of the program state can be easily converted to the other. The remaining portion at 96 of the pointer 90 contains other less important status information about the program loop. The real purpose for the program loop record 90 is to keep track of which program state within a particular program loop is active at any one point in time.

When a particular program loop becomes active, a program loop record 90 is associated with the loop. As various program states became active within the program loop, simple variable 94 is overwritten with the new program state addresses or identifier. The program loop record 90 enables one program loop to "hand shake" by interrogating a second program loop to determine which state is active in the second program loop. Furthermore, the program loop record 90 can be accessed by terminal 80 to allow programmers to display or modify the active program states. In the preferred embodiment, the program loop records are all stored in a task table FIG. 3b which is stored in memory 10A (FIG. 1). The task table functions primarily to enable processing of several program loops all at the same time. The program loop records are packed in the task table so that all non-empty records are at the start of the table and they are arranged in the numerical order of their respective program loop identifiers (FIG. 3a at 92). Referring to FIG. 3b, the first program loop record 100 in the task table has the lowest loop identifier (e.g., 1) and therefore is first to be scanned during a program scan because scanning occurs in numerical order of the program loops, scanning and processing the active state only in each program loop, scanning and processing each statement in each active state in the order it appears in the state block. Loop record 102 has the second lowest loop identifier in the table and it will be processed second. Loop record 104 has the third lowest loop identifier (e.g., 3) associated with it and it will be processed third. Loop record 106 has the next lowest identifier (e.g., 34) and it will be processed fourth.

Also associated with the task table is a task table register 112 (FIG. 3c) which points to the program loop record which is currently controlling the execution of an active program loop. To scan and process the program loops in the task table, the controller writes the address of the active program loop record 90 into the task table register 112.

When the controller executes the program loop associated with record 100, only one block of statements in the program state associated with program loop 100 is executed or activated. Then, the controller will process the next program loop (e.g., 2) and only one program state associated with it will be processed. The controller repeats this procedure for every active program loop. When the task table register points to program loop record 108 (FIG. 3b) which has a zero associated with the loop identifier, the controller is alerted that it has reached an empty program loop record. When this happens, the operating system of the controller points the task table register back to the top of the task table and executes the next program state of the first program loop in the table. This procedure is an endless process unless a terminal interrupt or the like occurs.

Because only a few lines of program code are processed for each program state, the controller can jump very quickly from one program loop to the next. In this way, the controller appears as if it is processing multiple program loops or tasks all at the same time.

In order to show that a program loop has been activated, the controller will assign a program loop record to the activated program loop and insert the new program loop record (FIG. 3a at 90) into the task table. The system evaluates the numerical value of the new program loop identifier (FIG. 3a at 92) and fits the program loop pointer into position in numerical order within the task table corresponding to the numerical value. For example, if the numerical value is greater than 34, the new pointer will replace record 108 in the task table (FIG. 3b). However, if the new program loop identifier is between 3 and 34, then the program loop pointer 108 will be swapped with the pointer at 106 and the new program loop pointer will be inserted at 106. Because the present embodiment of the task table can have a maximum of 31 program loop pointers in the task table, the contents of the task table must be evaluated to ensure that there is space available for the new pointer. The controller will alert the programmer when the task table has reached capacity or when an overflow condition in the task table has occurred.

Referring to FIG. 4, a logical representation of a trace table 82 is shown. The trace table 82 is stored in the memory portion 10A of the controller (FIG. 1). The trace table is a linear table in memory which is made to function as a wrap around "endless tape" storage buffer. The trace table 82 works quite similarly to an aircraft flight recorder. An aircraft flight recorder provides a "playback" of what happened to the airplane just prior to when the airplane "crashes". Likewise, the trace table records what happens to the states of a machine or process just prior to an event requiring investigation. Stated differently, the playback can be used to find out exactly what decisions were made by the controller just prior to the time an error occurs.

A trace table register 84 stores the addrress of the last recorded location in the trace table 84. FIG. 4 shows register 84 pointing to record 74 as the last location to be filled in the trace table 82. When the trace table has been filled at 86, the system will continue to store new data and overwrite old data at 0.

The following is a detailed description of the preferred embodiment for processing finite state applications programs and for facilitating diagnostics of unexpected conditions in the finite state applications program.

FIGS. 5, 6, 7, 8, 9, 10, 11, 12 and 13 are flow diagrams that represent routines which together with the programmable controller control the processing of an applications program. Particularly, in FIG. 5, routine START PROGRAM initializes the various pointers and tables of the programmable controller. In FIG. 6, routine FIRST OPERATOR evaluates the first statement in a program state. In FIG. 7 ADVANCE P-CODE POINTER controls a standard applications program pointer (p-code pointer). This pointer points to the address in memory of the current statement being executed by the controller. ADVANCE TASK TABLE REGISTER in FIG. 8 effects which record the task table register points to. In FIG. 9 routine NEXT OPERATOR evaluates all statements other than the first statement of the program state.

GOTO in FIG. 10 controls which states are active in the program loop. START initiates a change of state, or establishes an active state in the loop specified in the start statement. GOSUB in FIG. 12 and RETURN in FIG. 13 provide a subroutine call and return facility which operate at the state level.

Referring to FIG. 5, a flow block diagram for START PROGRAM is shown. This routine initializes the program controller. Specifically, at block 122, the controller initializes the task table (FIG. 3b), the task table register (loop pointer) (FIG. 3c), and the p-code pointer. The task table is initialized with a program loop record that points in memory to the first active program loop. Also, the program loop record is initialized to the first program state in that program loop. The p-code pointer is initialized with the start address of the first active program state i.e., the first statement. The system is now ready to process the first active loop. In block 124, the system calls subroutine FIRST OPERATOR to evaluate the first statement pointed to by the p-code pointer.

The block diagram in FIG. 6 depicts routine FIRST OPERATOR. At block 128, the controller analyzes the type of statement pointed to by the p-code pointer. The first statement encountered can only be a state type statement either: a "State" statement, a "Pause" statement, or a "Break State" statement. A state type statement is a program state delimiter which symbolizes the beginning of a program state. A Pause statement also causes its own program loop to pause but allows processing of other active program loops. A Break State statement also causes the entire system to stop running and return to the monitor in order to alert the programmer that the controller has discontinued processing. If a state statement is encountered at 130, the controller advances the p-code pointer to the next statement address in the program state by entering routine ADVANCE P-CODE POINTER at 132. If the statement encountered in the program state is a pause state statement at 134, the system with enter ROUTINE ADVANCE TASK TABLE REGISTER at 136 to advance processing to the next active program loop. If the statement encountered is a Break State type, routine GO TO MONITOR at 138 is entered.

GO TO MONITOR places a message on the terminal to alert the user of the programmable controller that a break statement has been reached and that execution of the program has been halted.

Referring to FIG. 7, a detailed block diagram of the routine ADVANCE P-CODE POINTER is shown. At block 142 the p-code pointer is incremented to the next statement in the active program state. The controller processes the next statement by entering routine NEXT OPERATOR at 144.

When the Pause State statement is encountered by FIRST OPERATOR (FIG. 6), routine ADVANCE TASK TABLE REGISTER is entered. The purpose of this routine is to ignore the current loop and switch to the processing of the next program loop. FIG. 8 shows a detailed block diagram of this routine. At block 148, the task table register is advanced to the address of the next program loop pointer in that task table. Routine FIRST OPERATOR is entered at block 150 to process the active program state of the next program loop in the task table.

When a State type statement is encountered at 128 (FIG. 6) the controller proceeds to routine ADVANCE P-CODE POINTER 142 (FIG. 7). As mentioned above, this routine advances the p-code pointer to the next statement in the program state and then enters routine NEXT OPERATOR at 144. The first statement of a program state is always evaluated by routine FIRST OPERATOR.

FIG. 9 is a schematic block diagram of the routine NEXT OPERATOR. The purpose of this routine is to evaluate any type of simple statement in a program state, other than the first statement. At block 154, the controller reads the next command pointed to by the p-code pointer. If the next statement encountered is either a State statement, a Pause State statement, or a Break State statement, the controller will enter routine ADVANCE TASK TABLE REGISTER at 158 (FIG. 9) because the end of the state block has been reached. If the statement encountered is not one of the three above, then the controller will evaluate whether the statement is a GoTo statement at 160. If the statement is a GoTo statement, then the routine GOTO is entered at 162. This routine will be discussed in more detail below.

If the statement is not a GoTo statement, then the controller will advance to block 164 and determine whether the statement is a Start statement. If the statement is a Start statement, the controller will enter routine START at 166. We will discuss this routine shortly.

If the statement is not a Start statement, the controller will advance to block 168 and determine whether the statement is a GoSub statement. If it is a GoSub statement, the controller will enter routine GOSUB at 170. Assuming that the statement is not a GoSub statement, the controller will determine whether the statement is a Return statement at 172. If it is a Return statement, the controller will enter routine RETURN at 174.

If the statement is not a GoTo, Start, GoSub or Return type, then the controller will advance to block 174 and evaluate the statement as a normal control statement. Normal control statements do not affect the status of the program states, or stated differently, the program state continues to process without changing which state is active in the current or a different program loop. Whereas, the statement types: Go To, Start, Go Subroutine, and Return, all potentially affect the status of the program state and/or the corresponding active program loop. As discussed earlier, these types of statements affect the status of a particular program state if a set of conditions is satisfied. Once the controller has evaluated the normal control statement at 176 (FIG. 9), it will enter routine ADVANCE P-CODE POINTER to evaluate the next statement in the active program state.

FIG. 10 is a block diagram of the routine called GOTO. The GOTO routine determines if a particular set of conditions are satisfied, and if so changes the action state in the loop currently being processed to the new value designated in the GOTO statement and then advances processing to the next program loop in the task table. GOTO provides the basic state changing function whereby transition conditions for active states which are true cause a new specified state to become active in place of the currently active state. Once a GOTO statement is actioned in a state, then no further statements are processed in that state.

At block 182 of FIG. 10, the controller evaluates the condition flag status resulting from the evaluation of the condition part of the compound GOTO statement. If the set of conditions in the compound GoTo statement are not satisfied, the controller will enter routine ADVANCE P-CODE POINTER at 184. If the condition is satisfied, the controller will call subroutine STORE HISTORY at 186. STORE HISTORY saves a "decision point" address and the program loop identifier into the trace table of the current loop being processed (FIG. 4). The decision point is the address of the simple statement in the program at which the controller found the conditional part of the compound statement containing the GOTO to be true. At block 188, the current program loop record in the task table is updated so that simple variable 94 (FIG. 3a) points to a new program state address. ADVANCE TASK TABLE REGISTER at block 190 is then called to update the task table register so that it points to the next program loop pointer in the task table.

FIG. 11 is a block diagrm of the subroutine called START. The purpose of START is to initiate processing of any other program loop at a state specified in the START STATEMENT. By comparison, the GOTO subroutine can only advance processing to the next state stored in the current program loop.

At block 194 of FIG. 11 the condition flag status resulting from the evaluation of the condition part of the compound Start statement is evaluated. If the condition flag is false the controller enters routine ADVANCE P-CODE POINTER at 196. If the condition is true, then the controller advances to block 198 and processes subroutine STORE HISTORY. STORE HISTORY saves the address of the decision point and the program loop identifier for the current loop. At block 200 the controller determines whether the specified program loop is presently active and stored in the task table. If the program loop is already active, then at block 202, the controller overwrites the address currrently stored in simple variable 94 of the program loop pointer (FIG. 3a) with the address of a new program state to be activated. However, if the program loop is not presently active, then the system will slot a new program loop record into the task table at 204. The new program loop record will be placed in numerical order in the task table via the numerical value of its program loop identifier 92 (FIG. 3a). At block 206, the address of the new program state will be written into the program loop record. At block 202, the controller will enter routine ADVANCE P-CODE POINTER and go to the next statement.

FIG. 12 is a block diagram for GOSUB. GOSUB affects the processing of the active program loop. This routine enables a new state to be activated in the current program loop and status information to be saved so that a corresponding RETURN statement can return control to the state containing the GOSUB statement at the next statement following the GOSUB statement. This provides a subroutine call and return function at state level.

Referring to block 210 of FIG. 10 (the condition flag status resulting from the evaluation of) the condition part of the compound GoSub statement is evaluated. If the condition has not been satisfied, then the controller will enter routine ADVANCE P-CODE POINTER at 214. However, if the condition statement is satisfied, then the controller will call subroutine STORE HISTORY at 216. This subroutine will save both the decision point and the program loop identifier in the trace table. At block 218, the controller will temporarily save in memory the present address stored in the p-code pointer, the address of the currently active program state, and other necessary status information for use by a corresponding RETURN statement. At block 220, the address of the new program state is written into simple variable 94 of the program loop record 90 (FIG. 3a), and this address will also be stored in the p-code pointer. The controller will initiate activity in the new program state by entering routine FIRST OPERATOR at 222.

FIG. 13 is a detailed schematic block diagram of routine RETURN. The purpose of the Return statement is to get the controller to return to the program state which contained the GOSUB statement. In other words, once the specified program state or state diagram called by GOSUBROUTINE (FIG. 12) is completely executed, the controller will return control to the old program state which was active when GOSUB was executed. This requires that the status values stored by GO SUBROUTINE be restored.

Referring to FIG. 13 at block 224, the controller evaluates the condition flag status resulting from the evaluation of the condition part of the compound Return statement. If the condition has not been satisfied, then the controller will enter routine ADVANCE P-CODE POINTER at block 226. However, if the condition is satisfied, then the controller will call STORE HISTORY at block 228. This routine saves the address of the decision point and the loop identifier in which it occurred. At block 230, the controller will restore the values saved at block 218 of GOSUB (see FIG. 12). ADVANCE P-CODE POINTER at block 232 is then entered to continue processing in the old program state.

FIGS. 14, 15, 16, 17 and 18 are block diagrams of programs which facilitate debugging of application programs. Specifically, FIG. 14 is a block diagram of STORE HISTORY which is a subroutine for updating the trace table (FIG. 4) with information and is called by the routines, GOTO, START, GOSUB and RETURN. The remaining figures are essentially diagrams of subroutine utility which enable a programmer on his own command to view information stored in the trace table, the task table or in the program state itself. Specifically, FIG. 15 is a block diagram of a subroutine called DISPLAY HISTORY which displays the information stored in the trace table on terminal 80. FIGS. 16 and 17, respectively are subroutines DISPLAY LINE and DISPLAY PROGRAM STATE. These routines are responsible for displaying the statements of a particular program state and illustrate, e.g., where within the program state was the decision point at which the controller decided to execute the GOTO, GOSUB, START or RETURN statement that caused the change of state. FIGS. 18 and 19, respectively, are DISPLAY LOOP and FOUND LOOP which display the active program states and associated active program loops which are stored in the task table.

STORE HISTORY (FIG. 14) is called whenever the status of a program state changes. STORE HISTORY is called in subroutines GOTO (FIG. 10 at 186), GOSUB (FIG. 12 at 216), START (FIG. 11 at 198), and RETURN (FIG. 13 at 228). As discussed above, STORE HISTORY essentially saves into the trace table the address of the decision point in the active program state and the loop identifier of the program loop being processed.

At block 236 in FIG. 14, the address stored in the trace table pointer 84 (FIG. 3) is evaluated. If the address pointed to by the trace table pointer is at the end of the trace table, the trace table pointer will be automatically set to the begining of the trace table at block 240. This step makes the trace table appear as if it is an endless loop (see FIG. 4). Whether the pointer is at the end of the trace table or not, the address of the decision point will be written into the record pointed to by the trace table pointer at block 242. Additionally, at block 244, the controller will also save the program loop identifier of the loop being processed into the record and at block 246 the new address will be saved into the trace table pointer. At block 248, the controller will return to the calling routine.

Subroutine DISPLAY HISTORY enables a programmer to evaluate the history of program states which occurred during the processing of an applications program. The programmer initiates this program when he wants to see a display of what was stored by the trace table. This sub-routine displays the contents of the trace table in a two-axis table. The horizontal axis presents the program loop identifier and the program state corresponding to the GOTO, GOSUB, RETURN or START statement that made the entry in the trace table. The vertical axis represents the sequential order of the program states and program loops with the newest states at the top. DISPLAY HISTORY is a subroutine that is initiated directly by a user at terminal 80 (FIG. 1). A typical printout of the DISPLAY HISTORY might be:

Note that the command DISPLAY HISTORY is shown first on the screen. Second, across the top are the words "active program states" and "active program loops". The two columns beneath these statements refer respectively to the active program states and the active program loops. A short comment is provided in this specification on the right-hand side to explain each line. The program state which caused a state change last is shown in line 0. Program state 2 in program loop 2 was the last to make a change. The first program state to cause a change and the program loop in which it resided is shown on line 7. Line 8 shows the end of the trace table.

If an unexpected condition occurred altering the expected flow of active states in the applications program, a programmer could easily narrow down which program state the unexpected condition occurred in by viewing the sequence of program states that occurred during processing. For example, assume the program state 3 should not have occurred in loop 1 (line 3 in the table) in the applications program. The unexpected condition must have occurred in program state 2, the previous state in loop 1 which caused the status of program state 2 to change to program state 3. The programmer can more closely analyze program state 2 and the conditions which caused program state 2 to change to program state 3 by calling DISPLAY LINE (FIG. 16). This program will be discussed shortly.

Referring to DISPLAY HISTORY in FIG. 15 and at block 252, the controller prints the header which appears at the top of the screen as shown above. Next, at block 254, a copy of the trace table and a copy of the trace table pointer are stored into a buffer. By placing the trace table and trace table pointer in a buffer, a snap shot of what was stored in the trace table when the request for DISPLAY HISTORY was made is acquired and remains unchanged for as long as needed while allowing continued updating of the main trace table. The reason for this step is that the trace table is a dynamically changing table and it is constantly being updated. Only the status of the table at the time the request is made is required for purposes of analysis. At block 256, the system will get the address stored in the trace table pointer of the last record stored. At block 258, the controller will evaluate this address to determine whether the buffer is empty. If the buffer is empty, an end statement is printed and the controller returns to its original processing at block 260. If the buffer is not empty, the trace table is evaluated to determine if it points to the end of the buffer at block 262. If the pointer is at the end of the buffer, then the trace table pointer is set to the beginning of the buffer at block 264. In this way, the controller processes the contents of the trace table as if the trace table were an endless loop. Regardless of whether or not the trace table pointer is at the end of the buffer, the loop identifier and the state address are obtained from the buffer at block 266. The decision point address will be used as a starting point in locating the program state number which has been previously stored in the state delimiting statement of the program state. At block 268, the program state numbers and the program loop identifiers will be printed in their respective columns as shown above. At block 270, the trace table pointer is incremented to the next record in the buffer and the routine will continue processing at block 258. The program will iterate loop 271 until all of the program states, and the corresponding program loops, which were stored in the trace table, are printed.

In our example above, the programmer knew that program state 3 should not have occurred. He also knows that an unexpected condition must have occurred during program state 2 on line 4 of DISPLAY HISTORY. The programmer can analyze more closely what occurred in program state 2 and determine exactly what caused the program state to change to program state 3, by caling the sub-routine DISPLAY LINE. DISPLAY LINE generates a listing of the statements of the program state and indicates at which state within the program state a decision was taken which caused the status of the program state to change.

Referring to FIG. 16, a detailed block diagram for the subroutine DISPLAY LINE is shown. This program is initiated directly by the programmer via a call from the terminal 80 (FIG. 1). At block 274, the programmer will designate which line of the printout for DISPLAY HISTORY he wants to have expanded. For example, because he is interested in what occurred from line 4 to line 3, the programmer will expand line 4 via the DISPLAY LINE. At block 274, the desired line number of DISPLAY HISTORY is set equal to the variable "Line Count". At block 276, the controller will determine if the line number chosen fails within the valid range of numbers. If it does not, then the controller will print "line invalid" and return to its normal processing at block 278. Assuming that the line number is within a valid range, the last data entry in the trace table which contains data for line 7 will be pointed to by the trace table pointer at block 280. Next, the trace table pointer is incremented by one at block 290. At block 292, the controller will determine whether the trace table pointer points to the end of the trace table buffer. If it does, the trace table pointer will be set to the beginning of the trace table at block 292. Regardless of whether or not the trace table pointer points to the end of the buffer, the controller will evaluate the "line count" to determine if it is zero at block 294. If "line count" is equal to zero, then the desired program state has been uncovered, and DISPLAY PROGRAM STATE will be entered at block 296. If "line count" is not equal to zero, "line count" is decremented by one and the trace table pointer is incremented by one at block 290. Loop 301 iterates until "line count" is equal to zero at block 296, and the designated line is found.

FIG. 17 shows a detailed block diagram of the routine DISPLAY PROGRAM STATE. At block 304, the controller will determine if the record pointed to by the trace table pointer is empty. If the record is empty, then the system will print a statement on terminal 80 that the line requested is invalid and return to its normal processing at 306. Assuming that the record is not empty, the controller will extract the address of the decision point from the trace table and determine the starting address for the program state which encompasses the decision point at 308. In block 310, the controller will print the text for one simple statement of the program state. In block 312 the controller will check to see if it has encountered the decision point in the program by comparing the address of the statement printed to the decision point address stored in the trace table. If the decision point is reached, then the statement "**DECISION POINT**" will be printed at block 314. The controller will then step to the next statement in the program state at block 316. At block 318, the controller will then determine whether the next statement is a state delimiting statement, state statement, pause state statement or break state statement. If any one of these three statements are encountered, the controller will return to its normal processing at block 320 because the end of the state block has been reached. If none of the three statements are encountered, processing will continue via loop 322 until all the statements of the program state are printed.

FIG. 18 is a block diagram for DISPLAY LOOP. DISPLAY LOOP is a subroutine for displaying the contents of the task table. Referring to a typical screen display below, loop identifiers are presented in a horizontal axis and the active program state within each of the program loops is designated vertically beneath each of the program loop identifiers. The following is a typical screen display of an output for DISPLAY LOOP.

At block 324 of FIG. 18, the controller is commanded to print the loop identifiers as shown in the first line of the screen display, as shown above. Block 326 sets the task table register to the first program loop pointer in the task table. A variable called "loop count" is set equal to zero at block 328. "Loop count" is incremented by one at block 330. The controller then determines whether "loop count" is equal to the maximum number of program loops that can be stored in the task table. If this condition is true, then the controller will return to its original processing at block 334. If loop count is not equal to the maximum number of program loops, then the controller will get the loop identifier from the relevant task table record. At block 338, the loop identifier is checked to determine if it is equal to zero. If the program loop identifier is equal to zero, then the system will return to its normal processing at block 342, however, if the loop identifier is equal to the "loop count", then subroutine LOOP FOUND is called at block 344. If the program loop identifier is not equal to "loop count", then the controller will tab across the screen to the next loop position at 348. "Loop count" is incremented by one at block 350, and the blocks after block 328 will be iterated until no more active program loops are in the task table.

FIG. 19 is a detailed block diagram of subroutine FOUND LOOP. At block 354, the start address of the active program state stored in the program pointer will be accessed. The program state number is extracted from a pre-stored value in the state delimiting statement in the program state. At block 356, the task table pointer is incremented to the next program loop pointer and the controller returns to the routine DISPLAY LOOP at 364 of FIG. 18.

The invention has been described in an exemplary and preferred embodiment, but it is not limited thereto. Those skilled in the art will recognize that a number of additional modifications and improvements can be made to the invention without departure from the essential spirit and scope. For example, a number of different software techniques and any number of different software languages would be suitable for implementing the disclosed invention.

However, by way of explanation, practical applications of the invention may be useful to put the invention into effect, an interpreter now to be described is used to execute an intermediate or "pseudo code" (P code). The interpreter enables an applications program to be written essentially in the form of one or more state diagrams, each of which defines operations of a part of a machine or process whereby any individual state defines the output signals and other data transmitted to the machine or process for a finite period of time.

The following interpreter used an RCA 1802 microprocessor and is assembled using a cross assembler form 250 A.D. Inc.

The interpreter is used to execute the applications program in which there are:

(1) a task table, i.e., a table indicating the states active in each loop, and containing the loop variables

(2) a trace table which when displayed enables the state changes in the program run to be traced.

Details of the interpreter are as follows:

Using the above interpreter code, an applications program code is executed.

The invention enables the following to be effected:

1. Being able to set "traps" or "breakpoints" or "pauses" or "halts" within the program, and being able to run, pause or stop the program from the monitor:

The control program written in a finite state type language can be run or stopped from the Monitor program. The control program, or part(s) of it, can also be paused, halted or stopped by instructions within the control program itself. The instructions that do these functions in the following examples are:

PAUSESTATE causes that particular loop to pause or halt when control is passed to that particular state

BREAKSTATE causes the whole program to stop running and return to the monitor in either the "online" or the "offline" modes when control is passed to a particular state

HALT is a conditional instruction that causes all instructions after it is in its state to be ignored

2. Being able to selectively or otherwise record the history of state changes of the program and being able to display these:

The program is such that in the step of storing of the history:

During the execution of each instruction that could influence the decision to change states (any part of the transition function), a temporary record is kept of its location in the program. At such time that the transition function becomes true and control is passed to another state, the temporary information and information identifying the particular loop is transferred to the historical changes table (TT the Trace Table). An example of this is shown in the interpreter code at MORE, IIFOC and STOREHIST.

3. Displaying the historical changes and the action that caused the change:

This simply requires the contents of the Trace Table or Task Table ro be printed out or displayed in a convenient form. The examples below show three such methods.

The first method displays the identifier of the currently active state in all of the active loops by deriving the required information from the task table.

The second method displays the contents of the trace table as a two-axis table. The horizontal axis represents the loop that the change took place in, while the vertical axis represents the time order of the changes.

The third method displays the change that occurred in any selected line of the table from the second method, and the event or events that caused the change, by listing out the relevant part of the program and indicating the point within the listing where the decision to change states was finalized.

It is to be expressly understood, however, that the following listings are for illustration only and are not to be construed as defining the limits of the invention.

The following section shows how we structure a program written in our language.

The program functional structure can be defined as follows

a program consists of one or more program (logical) loops

a loop consists of one or more states, only one of which can be active at one time. The set of states making up a loop is implied by the `start`, `restart`, etc., loop initializing statements in the program and the `goto`, `gosub`, state changing statements in those states

a state consists of a state delimiting statement, followed by a statement list

a statement list consists of zero, one or more compound statements

a compound statement consists of an optional condition part and an action part. The absence of a condition part implies a true condition and the action part would then be unconditionally executed

a condition part is a boolean expression consisting of an initializing part, one term (or more terms connected by `or` operators), and terminates at the next action part

an initializing part provides the functional equivalent to the connection to the power bus in a ladder diagram, and in our language is an `if` statement.

a term consists of one condition statement (or more condition statements connected by `and` operators)

an action part consists of one or more action statements

The following diagram shows the structure of a compound statement. The meaning of the statement in the diagram is `if input 10 is on or input 11 is off and timer 3 has timed for greater or equal to 23 time units, output 14 is on, and op 17 is on`

The method of evaluation of the condition part and execution of the action part is as follows

the statement is evaluated in the same order as it is read

`if` sets a boolean condition flag to 1 and this flag is used as a running record of the evaluation as it proceeds, and as a record of the result when it has been fully processed.

a condition statement is evaluated and if 0 the flag is set to 0

an `and` operator just proceeds to the next evaluation of the next condition statement

an `or` operator looks at the condition flag

if the flag is 1, the whole condition part must be true and so the remainder of the evaluation is skipped

if the flag is 0, the preceding terms must all be untrue, so the flag is set to 1 and the evaluation proceeds with the next term

an action statement executes and carries out its function if the condition flag is true when the action statement is read

The method does not allow the use of brackets in the condition part, and a condition part is equivalent to multiple rungs on a ladder diagram, each consisting of any number or series connected input variables, all connected in parallel at the point where the rungs connect to the actuator, but with no parallel connections at any other point. This allows any necesary condition to be evaluated, although the parallel connections would have allowed this with less evaluation of input variables. However the method provides a very simply implemented way of recording which inputs resulted in a true evaluation and thus caused the decision to execute the action list.

This is achieved by maintaining a pointer to the p code representing the program that is being processed, and this is necessary anyway. Any `or` operator which detects a 1 condition flag saves the value of that pointer before modifying it to skip to the action part. As we only have complete `rungs` with no alternative or ambiguous paths in the condition part, the saved pointer value must point to the end of the first rung found to evaluate true. This is termed the `decision point`, and is used in debugging. Should the evaluation proceed to the end of the last term, and thus past the last `or` operator, this position becomes the decision point provided the last term yields a true condition, or the position of the action statement can serve equally well in this case.

Broadly a compound statement is equivalent to a rung on a ladder diagram which evaluates a condition, and if the condition is found true, carries out some action. The detail of this part of the structure as defined above is of significance mainly from the point of view of easy recording and print out of the decision point in a trace of any one particular state change. The Program-loop-state structure is the thing of major significance because it is this structure that yields the greatest benefits in terms of easy designing of error-free programs, and of understanding those programs.

It is also true from another viewpoint that a program consists of one active state diagram (logical loop), which automatically starts with state 1 active, together with one or more potentially active state diagrams which can be activated from the first, and subsequently any state diagram can activate or deactivate any other. Code describing all potentially active state diagrams exist within the body of the program at all times, but a state diagram is not executed and is therefore functionally inactive until the state diagram is activated by the assignment of one of its states to a loop variable which subsequently keeps track of which state is active in that state diagram at any one point in time. The systems software has code and variables which enable it to determine whether an assignment has been made, or whether the loop has been deactivated.

An example of a program using numeric (rather than symbolic) variable names is shown below. The program does not `do` anything, and is for the purpose of illustrating structure only.

In the program above, there is an interesting example of the time sequencing capabilities of the system stemming from the way all statements in a state are processed sequentially in one contiguous block. Part of a message is being printed in state 20, and the remainder is being printed in state 30. This is a suitable technique where there is no random printing by other loops than this one. However, if other loops have states that could possibly print between "Hold state entered at" and "@ Hrs.", this technique would cause a problem. The way around the problem is simple, just place all associated print statements together in a single state and the operating system interpreter will guarantee to process all together, avoiding hold-ups in the control functions by placing characters in a print buffer to be down loaded to the printer one at a time by an interupt routine as the printer becomes ready.

It should be noted at this point that the functional structure is different from the structure as apparent from a listing of a program. A listing does not directly show the presence of loops, and a program listing consists of one or more state listings only, including the finer detail of compound statements . . . action statements. The loop information is of course there, and can be found in the `start` statements, and the goto, gosub statements, etc.

The third statement in the example program, `start loop 2 with state 200`, defines the state to become active in loop 2 on the next scan of that loop after the execution of the start statement. Other `start` statements would normally be used to define initial states for all other loops used in the program.

It should be noted here that there are two different meanings of the word "loop" in our system, and the word "loop" could be qualified and shown as `logical loop` and `physical loop`. Broadly the difference is the same as, for instance, `logical devices`, and `physical devices` in MSDOS (TM of MicroSoft), i.e., a program can talk of printing on a `theoretical` line printer without reference to where that line printer is, but at some stage the overall software must make the link between the theoretical line printer and the fact that such a device actually exists at some particular address or location in the hardware of the system running the program. Similarly, our program is designed in terms of (logical) loops which are effectively state diagrams and are termed loops because in most practical systems the state that is first active in the diagram will become active again each time the machine or process cycles. The process activity therefore typically cycles through a series of states arriving back where it started and therefore traverses a `loop`.

In order to enable the processor to run the program in the way desired, we must assign a logical loop to a physical loop. This consists of using a start statement, `start loop n with state m` which associates state m with loop variable n, either directly storing m in loop variable n, or storing a pointer of other indirect reference. After executing one or more start statements we have one or more physical loop variables that are initialized with state references and the system software can read these in turn, process the states, changing the references in response to goto, gosub statements, etc., and performing the necessary activities to control the machine/process. As noted elsewhere, which state in the program forms part of which logical loop(s) is determined by the start, goto, gosub statements, etc. and not by any means such as associating state blocks with larger subprogram blocks.

It would, of course, by possible to define `Loop Blocks` with initial states, and use a statement such as `start loop block 3`. It is considered that the current method provides good flexibility and ease of use.

We should also note that it is often desirable to have different logical loops associated with the same physical loop (one at a time) when performing different alternative control tasks.

Normally the exact meaning of the work "loop" will be apparent from the context and so the word is not normally qualified with `logical` or `physical`.

The details of which states form parts of which loops, and the interconnections between the various states in a loop can be readily determined by following the network through via the state changing instructions in the states, i.e., the goto's or gosub/returns.

In our example we can see that loop 1 starting at state 1 also has states 2 and 20 in it because state 1 contains goto 2, and goto 20 statements. We can follow through states 2 and 20 via their goto statements etc., and find the complete network. Normally the programmer would define the structure of the network before the details of the states, and so this would be predefined. A utility program would normally be used to list any loop network structure should this be necessary, rather than manual inspection.

It can be seen from the above that the functional structure of the loops is built into the program but does not have any delimiting statements to define a contiguous block of code associated with any one loop. It would however be considered good programming practice to code loops using contiguous blocks of states, i.e., perhaps states from the numerical range 200-299 for loop 2, 300-399 for loop 3, 400-499 for loop 4, etc.

It should also be noted that this provides a mechanism whereby a loop may control different processes at different points in time, in the sense that a process is represented by a network of states. This might occur where a section of a machine can make one of a number of different parts depending on requirements. When a particular part is required, a particular loop is started at the start state for the state network which controls the making of that part. If a different part is required, the same loop is started at the start state for the different state network that controls the making of that different part. In this way a program can be structured so that different subclasses of activities are controlled by different state networks, thus modularizing the program and minimizing the extent of the program code that is relevant under any particular circumstances, and simplifying the comprehension of the operation of the machine.

This illustrates the principle that of all the combinations of input variables and states that exist in a system, only a very few simplified combinations are significant to the real operation of the machine, and at any one point in time, only a few of those few are significant.

It is also relevant to note that normal programming in this language, particularly of sequencing operations, often leads to all action statements (other than state changing statements, such as goto, gosub, etc.), being unconditional, and the decisions made in the system thus are mainly related to changes of state. The lack of conditions qualifying actions such as turning on outputs, etc., leads to a high degree of certainty as to what has or will happen when the system is in a particular state, and the system is therefore predictable. The user does not have to wonder what happened as he would if outputs were conditioned by inputs whose values were not known and are transient and not measureable later.

It should also be noted that any one state can be said to represent a ladder diagram because the content of the state is a list of compound statements, each of which consists of a condition part and an action part. From this, it follows that state changes actually represent the deactivation of a ladder diagram equivalent to the state being deactivated, and the activation of the ladder diagram equivalent to the state being activated. However it should also be noted that this is done in a high level way using abstract concepts such as states and conditional goto statements, and does not involve the manipulation of any extraneous elements such as enabling or master relay contacts or connections from such contacts to the various rungs of the dependent ladders, and in addition the activation of one state by a goto statement precludes, at the system level, the possibility of the state containing the goto statement still being active, because this action involves the overwriting of the value held in the loop variable for that loop.

It also follows from this that a program, only containing state 1 (and thus not having any state changing action statements) and where state 1 contains a list of compound statements, is in fact the functional equivalent of a ladder diagram program as used with a normal ladder diagram type PLC. Although this method of programming would not normally be used when the more powerful state-loop structure is available, it is relevent to note the equivalence because it indicates the relationship of ladder diagrams to state machines of the type that have their outputs conditioned by system inputs.

It follows that ladder diagram functionality is a sub-function of a state diagram, which is itself a subfunction of a program consisting of multiple state diagrams, and in our case where these state diagrams may be active or inactive, depending on whether they have been started by a `start loop . . . with state . . . ` type of statement. (In our system, they may also be overridden, paused, stopped, etc.). It is also worth noting that the loops could also be started by means other than statements in the program itself, by preallocation of loop starting state numbers, or a table driven system, from the system monitor, etc.

It also follows that physical loops are not directly equivalent to particular state diagrams. They are system variables, members of a list, that hold the value of an active state, and by virtue of the structure of the various state diagrams emulated by the program, define which state diagram is being controlled via that loop. Which particular state diagram is being controlled via a particular loop depends on the start statement that started it (or the subsequent gosub statements that swap in different `sub-state diagrams` until a `return` statement is encountered, etc.). Loop variables may be statically allocated in memory or dynamically allocated.

In the above sense `loop 2` refers to physical loop 2 or the value of the active state stored in loop variable 2. `Get loop 2` in our language means get the number of the active state in the state diagram currently being controlled by physical loop 2 using loop variable 2. The word loop has also been loosely used to mean state diagrams, i.e., logical loops, and we have just seen that any one physical loop can be used to control different state diagram emulations at different points in time. The precise meaning should be determined from the context. The reason the same term has been used for two different but related concepts is that this assists in the intuitive understanding of program functionality.

As we have seen, each active state is processed each scan. Statements such as `start timer` do not need processing each scan. Depending on the approach to programming, it is necessary to process such statements

either, in the first scan after activation of the state,

or, in the first scan after each time that the condition part of the compound statement involved becomes true, including the first scan through the state that it is true.

Such processing is similar to the equivalent ladder processing together with the additional state qualification, and avoids timers started repetitively each scan, thus held at 0, and continual repetition of messages printed each scan, etc. Most action statements require this type of processing, except for statements such as `op 22 on`, which means `output22 is on in this state`. This needs repetitive insertion into a pre-zeroed output array each scan so that the output will automatically deactivate when the state is deactivated.

A program using symbolic variables but otherwise identical to that is Listing 1 is shown below in Listing 2.

Note that state 1 now has a list of define statements for loops, states, etc. These may be manually inserted or automatically inserted by the operating system when a new variable name of a particular type is encountered during program entry, automatically allocating the next free number for the particular type of variable. These define statements could be in state 1 or not (they could be in a separate block), hidden or explicit (as shown).

The main point of note is that the program now has a great deal of user friendliness, and to a large extent is self-explanatory and self-commenting providing only that the reader has an understanding of the machine or process itself.

The indenting of the text should also be noted as this conveys information indicating the positions of `if`, `or`, `and` operators, compound statements, condition parts, action statements, etc., and breaks the program up into groups of statements that are functionally related, again easing the task of comprehension.

Our program is processed by the processor in a similar way to that of a standard PLC in the sense that it scans repetitively through the relevant statements, evaluating and actioning. In detail it differs. The scan only involves the statements contained in the active states. Which states these are is determined by reference to the loop variables. The processor reads the first loop variable in the list, determines from this where the beginning of the active state for that loop is in the program. It goes to that state and processes the statements in the state until it reaches the next state delimiting statement. This is the beginning of the next state in the program listing and marks the end of the current state being processed. The processor then goes to the next loop variable in turn and repeats the procedure, and ditto through the complete list of loop variables having active states. This activity together with various overhead activities common to PLC's makes up one scan. The scan repeats continually.

Goto, restart, return and gosub statements change the value of the loop variables immediately when processed and processing of the statements in that state ceases when this occurs and the processor transfers to process the next state in the list. I.e., statements past the goto statement in that state are not processed in that scan. On the next scan it, of course, processes the statements in the new state that has been swapped in, and the old state which was overwritten is not active or processed. Order of processing statements in any state is in the same order as the program listing, and loops are scanned from 1 upwards in numerical order.

Multi-tasking operating systems (MTOS) used as the basis for real time software are well known. They function in a similar way to our system in that they share processor time around the different tasks requiring attention, and wo do the same with the different active states in the different loops. MTOS however do not necessarily impose the rigid order of processing on their tasks that is imposed by our order of scanning loops, processing each active state completely before going on to the next. MTOS normally can swap the processor from one task to another on the basis of task timeout, ready task priority, interrupt, etc. This means that the actual order of processing of tasks and even statements is not necessarily known to the programmer who writes the applications program, although there are mechanisms provided to allow the programmer to control sensitive parts of the program or hardware system as deemed necessary.

In a machine control or process control environment, it can be very advantageous to know the exact order of processing, or to be able to predict or ensure that the response time of the processor to system inputs will be adequate. Machines particularly, require fast response, and with PLC's it is normal to specify the time for one full scan of the program, and this may be of the order of 1 to 20 milliseconds. Out system provides a means of ensuring that all necessary activities are always carried out sufficiently quickly.

The structure of our language thus establishes that one scan will take the time needed to process all the active states once (but not any of the inactive states), plus some additional system overhead that is predictable. This provides for predictable scan times and because inactive states are not processed, scan times are short, or looked at conversely, a particular scan time can be achieved using out system and a lower power and cost processor, than can be achieved with a system such as ladder logic in which all statements are scanned every scan.

It should be noted that there are ways for standard PLCs to avoid scanning all ladder diagram statements every scan, but these are not as convenient as those provided with our system.

The swapping in and out of active states together with the loop structure provides the programmer with facilities similar to an MTOS, but which are totally predictable. The programmer can control response time by ensuring that states are coded so that no state takes too long for processing, or by using the loop variables to examine the other loops to ensure no long-winded processing can occur in one loop while a sensitive state is active. The system provides the necessary facilities in a very simple form.

In addition, as the statement list in each active state is processed in the order of the statements in the list, and all statements are processed before going on to the next loop, the programmer has a very simple way of guaranteeing that all necessary operations have been carried out once a list of sensitive operations has been started, and that no operations associated with some other state will intervene. Examples of when this might be important could be when using global variables or writing blocks of text to a terminal or when hand-shaking operations are involved. While this can be achieved using other techniques such as an MTOS, our structure which is all built in to a high level language provides a particularly simple way of achieving this. As well as being simple from the point of view of the programmer, it also is simple from the point of view of the systems software and hardware, and this translates into more economical means for controlling machines or processes.

The applications program is stored as intermediate code or `p code`. An interpreter program, which is part of the systems software, takes this p code and interprets it carrying out the necessary evaluations, actions and overhead activities. Details of such software techniques are well known and standard techniques are used throughout.

Facilities are provided so that the applications program can have access to the loop variables and so determine and use the information as to which state is active in any loop. Statements such as `get loop` as shown above are typical.

Discussing the various types of statements, the p code for all statements consists of two general categories of code

firstly, one or more bytes constituting an operator which defines the statement type and is used by the interpreter to select the service routine to be used.

secondly, optional data as required.

The types of statements (shown with examples) that are of interest are

loop initializing statements

`start loop 3 with state 200`

The p code for start loop, consists of the operator, the loop reference, and the state reference. The loop reference may conveniently be the address of the start of the loop data in the task table. The state reference may conveniently be the address of the state in the program p code.

The service routine involves

oncetest (see ONCETEST routine in previous application) and skip remainder if necessary

transfer state address into the loop variable in the correct loop location in the task table

set oncetest flags etc as required

save decision point data

exit to process next operator

`restart 3000`

similar to above but start loop 1 at state 3000

zero flags, registers etc., as appropriate.

save decision point data

exit to start of scan

state changing statements

`goto 201`

transfer new state address into loop variable in task table for loop currently being processed

clear once flag

save decision point data

exit to process next loop

`gosub 5000`

oncetest then as goto, except save address of next operator and of address of state to be returned to in save area in task table for use by return statement.

save oncetest flag and mask

save decision point data

exit to next loop

`return`

get saved addresses from save table so can process remainder of calling state

restore oncetest flag and mask

save decision point data

set condition flag

exit to process rest of calling state

two of the statements above are once only statements, namely start loop and gosub. All the other 3 function once only also by the nature of what they do, i.e., they change the active state in the loop in which they appear. Because they do this, they are not executed on the next scan through the loop. They thus have no need to refer to a Oncetest routine and should not be classed as once only statements. The reason gosub has this need, is that although it alters its own loop, after the return from the subroutine the calling state will become active again possibly for many scans, and it would be inappropriate to call the subroutine every scan, but calling it on a `once per state activation` basis, with a return from a subroutine not counted as a state activation, or on an edge activation basis is logical.

once only statements. Most statements are once only statements. The following are some other examples which can be implemented either as edge activated or once per state activation types as required.

`start timer 5`

`increment register 17`

It will be readily apparent why it is desirable to start a timer or increment a register only once per state activation or when a condition comes true; e.g., to start the timer every scan would just hold it reset but not running while in that state, instead of resetting to 0 and then allowing it to count the time that passes.

The execution of these statements involves little more than writing a 0 into the appropriate timer which is then counted up once per time interval by an interupt routine until it eventually sticks at timer overflow (255 for 8 bits), or adding one to the number stored in the appropriate register.

Once only operation.

The ONCETEST routine in the source code disclosure. shows how to implement routines that execute their function every time the function (their state is active and their condition part is true) becomes true. There is a simplified way of providing once only operation. This involves having a single boolean flag per loop. This flag is set by any routine which alters the state that is active in that loop. It is reset at the end of the first scan through that loop just before moving on to process the next loop, and optionally for ease at the end of every scan through the loop. It is therefore true only during the first scan through a newly active state. It is used by the service routines for any once only statements, and these execute their function only if the flag is true. The net value of this technique depends on the trade-off between a simplified interpreter and more flexibility in the language.

Under some circumstances it can be advantageous to be able to have a once only routine repeat the execution of its function every scan. For instance to hold a timer reset until the state becomes deactivated and then automatically release it. For this purpose, a `repeat` statement can be provided which when executed by the interpreter causes the next, or subsequent statements up to an `end repeat` statement, to execute every scan which can be done by the interpreter selecting alternative entry points to the service routines for those repeating statements, which entry points ignore ONCETEST and the first scan flag.

every scan statements. The prime example of this is the output statement.

`op 45 on`

This statement says effectively `output 45 is on while the state the statement appears in is active and subject to any condition applying to it`. The output automatically turns off when the state is deactivated. It does not say `turn output 45 on`, and need another statement to turn it off later. This is achieved by having all the bits in the output map set to 0 at the beginning of each scan, and providing a service routine for the statement that sets the appropriate bit in the map each scan the state is active and the condition part is true.

miscellaneous statements

define

A class of define statements is provided for variables or constants. These may be manually inserted into the program or altered, or automatically inserted when a new name is recognized. They are of the form `define state 5=thisstate` and basically associate the name with the numerical value of the variable/constant, and amount to a symbol table stored as part of the program. They do nothing during program execution and are just skipped, and so do not slow the interpreter significantly but they are used during program listing by the print routines.

Statements can be provided to test whether a particular state is active in some defined loop. This involves going to the loop variable for the loop in question, extracting the loop reference, which may be the address of the p code for the state that is active, and if so going to the state p code and extracting the state number which may be stored as the second and third bytes of p code. A comparison is then carried out. The statements must be compatible in terms of variable types, and may be of the form

Get loop 3

If equal to 315

Goto 350

or

If state 315 in loop 3

Goto 350

By way of additional disclosure for the sake of clarity, the following charts show the routines already described but in greater detail, and also some of the other supporting routines necessary to make such a controller a practical tool.

Also by way of additional disclosure for the same purpose, source code is disclosed for implementing the key routines on a particular microprocessor.

These following charts follow the source code disclosed closely if not exactly and use entry and exit points named similarly to the labels in the source code to assist with understanding of the code and cross referencing.

The following charts show the broad flow of control for our implementation of the novel scanning methods disclosed.

The above flow charts depend on the following data structure and assumptions.

1. A program preferably stored in p-code, but possibly stored in some other form, consisting of at least a list of one or more state blocks.

2. Each state block consisting of a state statement delimiting the start of the block, or similar delimiting means, none, one or more statements defining control action to be carried out including control of active states and loops, and ending at the start of the next state block or at some other convenient delimiter.

3. At least some of the statements available being able to define the static conditions that shall prevail while the state is active, as opposed to a list of procedural tasks to be carried out, e.g., `op 10 on` means `output 10 is on`, and not `turn output 10 on`, (for later turning off by some other statement). Note output 10 will go off automatically when the state becomes inactive. Note that the repetitive scanning and other special operating system support combines with the language to provide this feature.

4. At least some of the statements being able to act once only when the state is activated, and to automatically skip execution on subsequent scans through that state while it stays active.

5. A list of loop variables that may be active or inactive, and if active, that associate a loop number with an active state, and where the numerical order of said loop numbers control the order of scanning of said active states.

6. Goto statements that can work from the necessary evaluation to change the active state in the loop in which the statement appears without there being any need to evaluate which state is active in that loop, and a program structure that allows this to occur.

7. Start statements to make a loop variable active and load a pointer to the active state block into the loop variable.

8. Gosub/return statements that act AT THE STATE LEVEL as described.

All Drummond's basic claims of

1. Only one state active in a set at once.

2. One becoming active makes remainder in set inactive etc., apply.

Note exception: we can make loops active and inactive, and in the latter case make active states inactive without making any other active state in the same set active in turn.

Detail Flow Charts describing Interpreter routines. Labels and variable names are shown in CAPITALS.

Program starts at STARTINTER

The flowcharts detail essential elements in the algorithms. They do not necessarily describe house-keeping activities included in the program listing, such as maintaining registers, because these activities are specific to particular implementations.

Once a decision has been made that the condition required to carry out an action is true, then it is no longer necessary to continue evaluating the remaining terms in the condition statement because they can have no effect on the net result of the statement overall. This situation arises when an `or` type statement is encountered and the condition flag is true when the statement is read. As an example, in the statement

`if ip 5 on, and ip 6 on, or ip 7 on, set op 66 on`

if both inputs 5 & 6 are on, the statement `set op 66 on` will be carried out independent of the state of input 7, and therefore the evaluation of input 7 (and any other condition evaluating statements following) may be skipped. A special algorithm is provided for this and is as follows.

This algorithm relies on having the operators coded so that their binary values are the same as the low byte of the address in the jump table at which their respective entries reside, or the value could be equal to an offset for calculating the jump address into the table.

Statements are stored in memory as p-code and these statements may conveniently be stored as 3 consecutive bytes

byte 1--operator indicating kind of statement, in this case meaning `set op` or `set flag`

byte 2--address in internal output or flag map in memory which holds the byte containing this output/flag

byte 3--mask to select actual output/flag

Note the same routine may be used for both flags and outputs and byte 1 is identical, only the address determines whether the variable is an output or flag.

`Halt` is a conditional operation which if activated causes transfer of processing to the next loop.

Note that during the processing of any one state, two state type operators are read and processed. One at the beginning of the state block and one at the end. How they are processed is different. The state type operators found at the beginning of a state block are processed as shown in the charts above. (See approx 15 lines before and after TRPST. The operators found at the end of blocks are processed as below. They simply indicate the end of processing for this state, and transfer processing to the state that is active in the next loop. The operator at the end of a block is in fact the first operator in the next state block in line in the program listing.

State type operators include state, pausestate, and breakstate.

The p-code for `goto` is--operator, new state address (i.e. address of state operator at beginning of new state block in p-code in memory).

Uses the trace table which is a linear table in memory as a wrap around `endless tape` storage buffer. A pointer to the last used/next to be used location is saved as a separate pointer variable. Once the table has been filled, older data is overwritten by new data.

Note: in our implementation, a trace enable flag is stored encoded into the loop number highest bit. This can be done because we implement no more than 7FH loops encoded into 8 bits. A separate variable could be used. This variable is used for the `tracing of this loop disabled?` decision.

The p-code for `remark` is--operator, address of next operator in program, remark character string, (next operator).

This instruction is understood more easily by understanding the following

The task table is an array of records, some of which hold data and represent active loops that have been started, and some are empty and represent loops that have not been started. The array is dynamic in the sense that all non-empty records are packed to the start of the table and are ordered in numerical order of the value of the loop variables in the records. Inserting a new loop record into the array involves inserting it in numerical order into the packed array and if necessary moving higher order loops down the array to allow this to be done.

Each record is a loop record and has fields that hold the following simple variables

TTLOOP--The number of the loop the record represents, and that also decides the order in which active states are scanned. Loop numbers are mutually exclusive. In our representation valid loop numbers are all less than 127. We store the loop number as 7 bits of an 8 bit variable, and use the 8th bit as a `loop tracing disable flag`, see STOREHIST. Both the loop number and the flag are stored in TTLOOP.

TTSTATE--the address of the active state for the loop, i.e. the address in memory of the program p-code for the state operator which defines the start of the program state block for the active state.

TTONCE--This is the onceflag variable used by the ONCETEST

TTACCUM--The an intermediate result and data transfer register for this looop. It is used by statements that manipulate numerical data.

TTEXTEN--An extension for TTACCUM

TTSTATESTR, TTONCESTR, TTRTNAD--Three variables for preserving information from TTSTATE, TTONCE, and the p-code pointer respectively for use by `gosub` and `return` statements.

The p-code for `start` is operator, loop number, new state address.