Computing System Interface Using Common Parallel Bus and Segmented Addressing

This invention relates to memory systems in general, and more particularly to a memory system that includes a central processor unit having an internal memory system defined by a random access memory for providing the respective data registers required by the CPU, and to a method of sequentially transmitting input/output data and address data over a common parallel external bus.

Numerous combinations of CPUs and external memory are available in industry, respectively offering a variety of advantages such as size, speed, cost, etc. Recently it has become particularly advantageous, especially from a cost viewpoint, to utilize a memory circuit that includes predominantly insulated-gate-field-effect-transistor devices. For most applications the memory must be compatible with transistor-transistor-logic (TTL) circuits. Problems typically associated with such a memory system relate to speed of operation, flexibility of use, and general size, it being desirable to reduce the area of semiconductor material required for the systems. From a fabrication and reliability viewpoint, it is desirable to minimize the number of external interconnections between various chips in the system.

It is particularly advantageous to provide a central processor unit on a single chip. A major problem associated with this, however, involves reducing the area of respective functional units of the CPU and the interconnects required therebetween. Another desirable feature of a computing system is a simplified external interconnection system between the CPU and external memory units. In many applications it would be advantageous to be able to increase the size of external memory without providing additional interface circuitry.

One technique that has been proposed for incrasing memory capability involves multiplexing techniques for increasing address capability. For example, the CPU of a computing system has a fixed bit processing capability determined by design characteristics. This capability limits the number of external memory units that can be addressed. For example, an eight bit processor can address 2.sup.8 separate locations in memory. A proposed multiplexing technique involves storing eight low order address bits during one portion of a cycle, then storing eight high order bits in a separate register during a different portion of a cycle, and then addressing external memory with these sixteen bits. Such a technique involves increased complexity in design, more storage devices and relatively slow operation.

Accordingly, an object of the invention is to provide a computing system having an improved central processor having an internal memory section defined by a dynamic random access memory.

Another object of the invention is to provide a memory section for a central processor unit which defines a mutli-level program address stack.

An additional object of the invention is to provide a data register section for a central process that is defined by a single random access memory matrix.

Yet another object of the invention is to provide a method for operating a computing system that includes sequentially transmitting input-output, high order and low order address bits over a common external parallel bus.

In accordance with the present invention, a computing system includes a central processor unit in combination with a plurality of external memory units. In the preferred embodiment, the CPU is formed on a single chip and includes a parallel arithmetic logic unit, an instruction register and an internal dynamic random access memory. These functional elements are interconnected by a common parallel bus, sequential use of which is controlled by a CPU timing section. The CPU is connected to a plurality of external memory units by a common parallel bus.

In one aspect of the invention a method for operating a computing system includes multiplexing information over the common external bus to maximize utilization of the CPU elements and addressable memory capability. The method includes sequentially transmitting low order address bits. WRITE data, high order address bits, and READ data over the bus. The low order address bits are decoded by the memory decoder interface during transmission of the high order address bits. This has the advantage of increasing speed of operation. The method also conserves external hardware requirements.

In another aspect of the invention a computing system includes a central processor having data registers defined by a single dynamic random access memory matrix. The matrix is divided into first and second sets of data registers respectively coupled together to form pairs. In this manner increased data address capability is provided. In the preferred embodiment, the registers are each eight bits in length thus a pair of registers stores sixteen bits of address information. One pair from the two sets of registers is selected by a bidirectional static counter to function as the program address register. The remaining pairs of registers define a multi-level last-in-first-out program address stack. This provides means in hardware for absolute sixteen bit addressing for subroutine calls. For subroutine call instructions, the counter counts in one direction, selecting a new pair of registers for the program address register. The previous register pair stores the address for a return to the previous program. In response to a return instruction the counter counts the opposite direction.

The memory matrix also includes a third set of data registers which function as general purpose registers. Thus, all of the data memory registers of the CPU are formed as a part of a single matrix. This has the advantage of reducing space required on the chip. Further logic means are provided for selectively coupling the matrix to the internal bus. The logic means can select either the first, second, or third sets of registers for connection to the bus. In the event that none of the registers are selected, a refresh counter is enabled for refreshing one row of the memory cells. The refresh counter is enabled at least once during each cycle of operation of the CPU.

Other advantages and objects of the invention will be apparent upon reading the following detailed description of an illustrative embodiment of the invention in a computing system in conjunction with the drawings wherein:

FIG. 1 is a functional block diagram illustrating a single chip central processing unit interconnected with external memory systems.

FIG. 2 is a functional block diagram of a one chip CPU which may be used in accordance with the present invention.

FIG. 3A is a block diagram illustrating the internal bus interface of the CPU.

FIG. 3B is a logic diagram of the CPU illustrating internal bus interconnects to the various functional elements of the CPU.

FIG. 4A is a schematic of a dynamic random access memory cell that may be used in the present invention.

FIG. 4B is an integrated circuit layout of the dynamic random access memory cell of FIG. 4A.

FIG. 5 is a logic diagram of one bit of the arithmetic logic unit of the single chip central processing unit of the present invention.

FIG. 6 is an instruction map of the various classes of instructions executed by the CPU in accordance with the present invention.

FIG. 7 is a functional block diagram of the CPU sequence and control.

FIG. 8 is a logic diagram of a state timer circuit which may be used in the CPU sequence and control.

FIG. 9 is a logic circuit which may be used for the CPU input/output.

FIG. 10 is a logic diagram of the instruction register of the CPU of the present invention.

FIG. 11 illustrates the logic of the instruction decode portion of the CPU.

FIG. 12 is a logic diagram of the cycle timer shown in block form in FIG. 7.

FIGS. 13a, 13b and 13c contain a logic definition of the internal control of the CPU.

FIG. 14 illustrates the logic of the status decode of arithmetic operations of the ALU.

FIG. 15 is a logic diagram of the restart operation.

FIG. 16 is a functional block diagram of the arithmetic logic unit of the CPU.

FIG. 17 is a logic diagram of the arithmetic control section of the ALU.

FIG. 18 illustrates the logic of the temporary storage register, shift circuitry, and increment logic.

FIG. 19 is a logic diagram of the arithmetic unit.

FIG. 20 is a schematic of an 8-bit parity precharge circuit used in accordance with the present invention.

FIG. 21 illustrates the logic associated with the arithmetic flags of the ALU.

FIG. 22 is a logic diagram illustrating the arithmetic operation add.

FIG. 23 is a logic diagram illustrating the arithmetic logic for subtract.

FIG. 24 is a logic diagram illustrating the logic of exclusive OR.

FIGS. 25a, 25b and 25c are a logic and schematic diagram illustrating operation of the random access memory of the one chip CPU.

FIG. 26 is a schematic diagram of a clocked inverter that may be used in the RAM logic circuit of FIG. 25.

FIG. 27 is a logic diagram illustrating operation of the stack pointer logic.

FIG. 28 illustrates the logic associated with the refresh counter of the random access memory of the one chip CPU.

FIG. 29 illustrates the logic of the flip-flop utilized in the logic of circuit 25.

FIG. 30 is a functional block diagram illustrating operation of the interface between the one chip CPU and the external memory of the present invention.

FIGS. 31a, 31b and 31c schematically and in logic format illustrates the interface elements shown in FIG. 30.

FIGS. 32A and 32B illustrate multiplexing of the external 8-bit parallel bus for 1K byte of external memory.

FIG. 32C illustrates the bus system conventionally required in nonmultiplex systems for 1K byte of memory.

FIG. 33 schematically and in block diagram form depicts the external memory bus.

FIG. 34 is a logic diagram of the external timing.

FIG. 35 is a logic diagram of the external timer.

FIG. 36 is a logic diagram of the external memory interface.

FIG. 37 is a functional block diagram of the external random access or serial memory and associated control elements.

FIG. 38 is a logic diagram of the external bus slave timer.

FIG. 39 is a logic diagram of the chip select sample and hold circuit.

FIG. 40 is a logic diagram of the external memory input/output circuit.

FIG. 41 is an illustration of the address register logic of the external random access memory.

FIG. 42 is a logic diagram of the refresh counter logic of the external random access memory.

FIGS. 43A and 43B logically illustrate the X and Y address decode logic of the external memory.

FIG. 44 illustrates typical clock waveforms that can be used in accordance with the present invention; and

FIG. 45 illustrates an embodiment of the invention that incorporates two single chip central processing units in combination with common external memory elements.

The present invention is directed to a central processing unit (CPU) integrated on a single chip in combination with external RAM and ROM memory units. The invention will be described hereinafter first as it functions as a system. The functional units of the CPU will then be described. This description will include definition of the instruction set that may be utilized in conjunction with the CPU. For purposes of explanation, the CPU is functionally described as comprising a sequence and control logic unit, and arithmetic logic unit, and a CPU random access memory. These functional elements are interconnected by an 8-bit parallel bus. The various logic circuits associated with the sequence and control logic, the arithmetic logic unit, and the internal RAM, will then be described in detail, along with illustrative examples of operations for specific examples. Finally, the interface logic for interconnecting the CPU with the external memory will be described.

FIG. 1 illustrates in block diagram format a central processing unit 10, an external memory unit 12, and a read-only memory 14. These three units 10, 12, and 14 are interconnected by a common eight-bit parallel bus 18. An input/output system interface is indicated generally by block 16. This interface system enables external inputs and outputs to and from the CPU and memory unit 12.

In accordance with the present invention, the CPU 10 is integrated on a single chip. This has the advantage in that it enables fast execution time and a minimum number of leads necessary for interconnect to other elements of a computing system. The external memory 12 may be either a random access memory or a serial memory. As will be explained hereinafter with reference to the detailed description of FIG. 37, the logic of the external memory is designed so that either random access memory or a shift register type memory may be utilized.

The read-only memory (ROM) 14 is used in the system to store fixed subroutines or control programs. The CPU 10, memory 12, and read-only memory 14 are interconnected with each other and with the input/output interface 16 by a common eight-bit parallel bus 18. On one phase of the clock, the CPU or memory will output data; on the other phase of the clock, the CPU and memory will accept an input.

FIG. 2 is a functional block diagram of the CPU organization. The CPU consists basically of three blocks, the control decode 20, the ALU 32, and the internal RAM 40. The control section 20 controls CPU operation and synchronization such that communications between various blocks of the CPU may occur over the common eight-bit bus 25. The control section 20 includes a control decoder 26. This block has as inputs an interrupt request (INT REQ) and a ready (READY) signal. Outputs of the control decoder 26 include a SYNCH, FETCH, CYCLE, interrupt acknowledge (INT ACK), and MEMORIZE signals. A master system timer 22, and a cycle timer 24 (which enables variable instruction lengths) are connected to the control decode block 26. An instruction register 28 also inputs information into the control decoder 26. The control decoder has 18 outputs which control the internal RAM, the ALU, and the system interface and external memory units. An input/output block 30 forms a part of the control section 20 and is interconnected to the internal bus 25. The detailed logic circuits associated with the various blocks of the control section 20 will be described with reference to FIGS. 8 - 15 hereinafter.

Block 32 generally depicts the ALU section of the CPU. The arithmetic logic unit includes a temporary storage register 34 which contains the shift right and shift left circuits. Block 36 refers generally to an eight-bit arithmetic unit. This unit can execute eight different functions, add, add with carry, subtract, subtract with borrow, NAND, exclusive or, OR, and compare. A code P corresponds to each of these arithmetic operations. As will be explained hereinafter with reference to the instruction set of the CPU, bits 5, 4, 3 of the instruction register contain binary information corresponding to these arithmetic operations. By way of example, subtract with borrow has a code P equal to 3. This would be a binary code of 011.

Block 38 represents the four arithmetic flags which indicate the status of data of an arithmetic operation. As may be seen, the four flags are carry (C), zero (Z), sign (S), and parity (P). The condition code (cc) corresponding to each of these condition flags is shown as 0, 1, 2, and 3, respectively. As understood by those skilled in the art, two bits of binary data may be used to uniquely select one of the four flags. The condition flag codes, and also the arithmetic operation codes P are shown in Table V hereinafter.

The internal RAM of the CPU is shown generally at 40. This RAM contains 26 eight-bit registers. Two of these registers are selected for program address. These two registers are indicated at 42 and 44, corresponding to the lower order address bits (P.sub.L) and the high order address bits (P.sub.H) respectively. In combination, these two registers enable absolute 16-bit addressing of a location in memory. Using 16-bit memory addressing, it is possible to address up to 64K bytes of data in memory. The RAM also contains data registers, A, B, C, D, E, H, L, and M'. Data register A is used as the accumulator. Registers, B, C, D, and E are general purpose registers, and registers H and L are combined and contain the location of memory address. Data register M' is used internally only. Fourteen of the data registers in the internal RAM define a seven-level last-in-first-out stack (STACK). This has the advantage in that it facilitates calling subroutines.

The detailed logic associated with the arithmetic unit 32 will be described in detail with reference to FIGS. 16 through 24. The detailed logic circuits associated with the internal RAM 40 will be explained with reference to FIGS. 25 through 29.

As will be explained hereinafter with respect to the instruction set of the CPU, one of the data registers, A, C, B, C, E, H, or L of the internal RAM 40 may be selected by source or destination codes in the instruction. Three bits of data are required to select one of the registers as a source or destination register. For example, to select register D, a binary coding of 011 (3) would be required. As mentioned previously, the data register M' is used only for internal operation of the CPU. Thus, a coding of seven; that is binary 111, is used in the present invention to refer to external memory.

FIG. 3A is a block diagram showing the various interconnects to the internal bus 25 of the CPU. It may be seen that the instruction register 28, the internal RAM 40, storage register 44 and the ALU 32 all interconnect with the bus 25. Selection of the various registers of the internal RAM 40 is also illustrated at 41. The eight-bit registers of the internal RAM generally are of three types. There are general purpose data registers, A, B, C, D, E, H, L, and M', high order 8-bit address registers (P.sub.H) and low order 8-bit registers (P.sub.L). In other words, sixteen of the eight-bit registers are used to define address storage registers. An eight-bit low order register combined with an eight-bit high order address register provide means in hardware for absolutely addressing a sixteen-bit memory address. Two of these sixteen eight-bit address registers are selected by an up-down counter to function as the program address register. These are illustrated at 41 as P.sub.H and P.sub.L. The other fourteen registers form a seven level last-in-first-out pushdown STACK. Whether one of the general purpose registers, a high order address register, or a low order address register is selected for access to the bus 25 depends on the binary coding of input signals U and V. Which level of the registers is selected depends upon the address register coding. By way of example, if the input signals U and V are both logic 1, and the address register coding is 010, general purpose register C will be selected. In another example, consider U to be a logical 1 and V to be a logical 1, and the address register to have a coding of 001. Such a situation would result in level six of the high order address bits being selected. Similarly, if the signal U is a logical 1 and V is a logical 1 and the address register coding is 011, the low order address register P.sub.L would be selected. For the situation where U and V are both logical 1's; that is, for the situation where none of the data address registers of the internal RAM are selected for operation, then one row of the dynamic random access memory cells are automatically refreshed. This will be described in more detail with reference to FIG. 25.

FIG. 3B illustrates the logic gates of the CPU bus interconnections illustrated in FIG. 3A. Block 46 refers generally to one of the eight internal precharged buses referenced generally by the numeral 25 in FIG. 3A. The precharge enables a large capacitance to be driven in an extremely fast time. During phase 1 of the clock, the bus is precharged to a negative voltage level (assuming P channel type insulated gate field effect transistors) by transistor 53. During phase 2 of the clock, the bus 46 will be conditionally discharged. Inputs to the bus are generated by control signals prefaced with an asterisk (*). An example of such a signal is the control signal *M which generates the bus from the input/output buffer 45 of the CPU. The reference symbol $ indicates signals which sample the bus and allow data to be sampled into the various sections of the CPU. By way of example, a bus signal is generated by the NOR gate 47. For purposes of explanation, positive logic will be used in the examples described hereinafter. Thus, when the signal *M is a logic 1, the input signal on the input/output 39 will be transferred to the gate 47 when phase 1 of the clock becomes a logic 0. During phase 2 of the clock, the signal will be gated onto the bus 46. If the signal $M becomes a logical 1, the bus will be sampled and the output thereof transferred to the input/output 39.

Other sections which are connected to the bus include the instruction register shown generally at 28, which is operated by control signals *I and $1 The inputs *I.sub.3, 4, 5 enable bits 3, 4, and 5 of the instruction register (I) to be transferred to the bus when a restart instruction is executed. Operation of the instruction register will be described in detail hereinafter with reference to FIG. 10.

The instruction register is coupled to the control and timing section 20 and generates the various control signals that control the RAM, the arithmetic unit and the bus * terms and $ terms. The control and timing block receives two input signals, interrupt request (INT REQ) and ready (READY). Five output signals include SYNCH, FETCH, CYCLE, interrupt acknowledge, (INT ACK), and MEMORIZE. Eighteen outputs are generated by the control and timing section 20. Seven of these outputs are RAM control signals, three are $; i.e., sample enable signals, and eight are * or generate enable signals. Logic operation of the control and timing section 20 will be described hereinafter with reference to FIG. 7.

The bus is also coupled to a storage register (R), one bit of which is illustrated at 51. A temporary storage register goes directly on the bus and enables a right or left shift, or provides an input to the arithmetic unit 32. The logic performing a shift left and a shift right is illustrated generally by the blocks 57A and 57B respectively. The temporary storage register is described hereinafter with respect to FIG. 18.

An arithmetic unit 32 receives an output both from the temporary storage register 51 and the bus 46. When a valid result from an arithmetic operation is contained in the arithmetic unit, the signal *F will be a logic 1. This signal will generate the bus from the arithmetic unit. Operation of the arithmetic unit logic will be described in detail hereinafter in the detailed description of FIGS. 19 and 21.

The internal random access memory of the CPU also samples the bus. On phase 2 of the clock, the bus is sampled and depending on the status of two control signals, U and V, to the RAM, either register P.sub.L or P.sub.H (the low order program address bits or the high order program address bits respectively), the general purpose data registers, or refresh is selected. A typical RAM storage cell is illustrated at 48. When one of the data registers of the RAM is not being accessed, the RAM will automatically refresh through controls from the control section 20. The signal *RAM generates the bus from a RAM storage cell. The circuit cell and transistor logic of a typical RAM storage cell 48 will be described with reference to FIGS. 4A and 4B. Operation of the RAM will be described in detail hereinafter with respect to FIG. 25.

One bit of the arithmetic unit 32 is illustrated in FIG. 5. The arithmetic unit comprises inverters, shown generally at 59, NAND gates 60, NOR gates 62, complex gates 61, exclusive OR gates 58, and MOS transfer gates 63, interconnected in such a manner that in response to selected control signals, eight separate arithmetic operations may be executed. Operation of the logic to effect arithmetic operations add, subtract, and exclusive OR is described hereinafter in the detailed description of FIGS. 22, 23, and 24, respectively.

FIG. 4A is a schematic drawing of an insulated gate field effect transistor random access memory cell 48 that may be used in accordance with the present invention. In operation, the WRITE lines 15 is activated and IGFET device 17 becomes conductive resulting in the information present at the IN line 19 being transferred to capacitance 21. When the WRITE means becomes inactive, the information previously transferred to the capacitance 21 remains stored for a period of time depending only upon the capacitance-leakage resistance product of the storage node. This time constant will not be less than on the order of one millisecond for conventionally fabricated insulated gate field effect transistor devices under normally expected environmental conditions. The IGFET device 23 will be either conductive or non-conductive dependent on the state of the information stored. When the READ means 27 is activated the IGFET device 29 becomes conductive and consequently the state of the information present at the capacitance 21 may be determined by measuring the presence or absence of the conduction path from output line 31 to V.sub.SS through devices 23 and 29.

Data input line 19 may be activated by a normal ratio type IGFET element or precharge/discharge type of device.

The data output line 31 may go to a current sensing device or become the driver path of an IGFET ratio device with a discharge path in a precharge/discharge/IGFET element.

FIG. 4B shows a plan view of an insulated gate field effect transistor layout for the schematic of FIG. 4A. The devices may be formed using conventional photolithigraphic masking and etching techniques typically used in fabrication of insulated gate field effect transistor circuits.

The CPU can be divided generally into four sections: a data section, an address section, a control section, and an arithmetic logic unit. The control section is shown generally by the block 20 in FIG. 2, while the data and address sections are labelled 40. The data and address sections are defined by data registers which are included as a part of the internal RAM of the CPU. In addition, block 32, refers generally to the ALU section of the CPU.

As described previously, the internal RAM of the CPU contains 24 8-bit registers. Seven of these registers are data registers, including an accumulator labelled A and four general purpose registers, B, C, D, and E, and a memory register labelled H,L. The general purpose registers, B, C, D, and E can be used as index registers or secondary accumulators by programmer subroutine definition. All seven of these registers may be arithmetically combined with the accumulator, including the memory address register H,L. As will be explained in more detail hereinafter with respect to the CPU instruction set, the desired source and destination (S, D) is specified by three bits of the instruction to select one of the data registers A, B, C, D, H, or L or external memory. The binary codes for these various registers are shown in Table V.

The address section of the CPU is defined by 16 of the 8-bit data registers of the internal RAM. A pointer from an up-down counter selects two of these data registers to serve as a program address register or program counter P. The remaining fourteen registers define a seven level last-in-first-out program address stack (STACK). The purpose of the STACK is to provide means in hardware for accommodating absolute 16-bit addressing and subroutine address storage for 64K byte memory system.

In addition to the data and address registers, an instruction register (I) and a temporary storage register (R) are also accessible on the internal bus of the CPU. The control section of the CPU is based on sequential use of the parallel 8-bit bus between internal functional elements. To facilitate this control, there is a state counter 22 (FIG. 2) having four states, S1, S2, S3, and S4 and a cycle counter (C) labelled 24, having 1, 2, 3 cycles. The CPU is characterized as having two control states, WAIT and STOP. WAIT is induced by the control input READY to the control decoder 26. STOP is induced by the command HALT in either program or interrupt mode. Both of these control states break the normal chain of S1, S2, S3, S4 S1 circulation. The instruction register, INTERRUPT and READY inputs, state counter, and cycle counter are combined in the programmable control decoder 26 to mechanize controls which operate the arithmetic unit 32, the RAM 40, bus 25, and excite the state and cycle counters 22 and 24.

The CPU is designed to execute five separate classes of instructions namely, moves, arithmetic, jump, input/output and control. All instructions execute in 1, 2, or 3 machine cycles. Each machine cycle consists of one fetch and one execute. Each fetch and execute requires five microseconds for execution. The instruction word format is shown in Table I. As may

With respect to an arithmetic instruction, the P in the column for bits I.sub.5, I.sub.4, and I.sub.3 of the instruction refers to arithmetic operand codes. These three bits may be coded to select one of eight arithmetic operations that may be performed. These codings and the corresponding arithmetic operation are also shown in Table V. By way of example, a coding of 010 would refer to a subtract operation. An example of the logic associated with performing a subtract operation in response to such an instruction will be described hereinafter with reference to FIG. 23. An X in the column I.sub.5, I.sub.4, I.sub.3 refers to a "don't care" situation. These bits may be used by the programmer as desired.

FIG. 6 graphically illustrates an instruction map of the instruction set that may be utilized in accordance with the CPU of the present invention. With reference to FIG. 6, it may be seen that the instruction map includes four quadrants. These quadrants are respectively identified by the binary coding of instruction bits I.sub.6 and I.sub.7. For example, the upper righthand quadrant labelled move (3DS) corresponds to instruction bits I.sub.6 and I.sub.7, both being binary 1's; hence, the 3. Similarly, the upper lefthand quadrant of the map corresponds to a binary 2, which results from instruction bit I.sub.7 being a binary 1 and instruction bit I.sub.6 being a zero. It may be seen that each quadrant of the instruction map is 8 bits by 8 bits square. With respect to the move instructions in the upper righthand quadrant, the vertical registers 0 through 7 labelled I.sub.2, 1, 0, respectively correspond to source (S) registers, such as data registers A, B, C, D, E, H, L, or M' of the RAM of the CPU. The source destination S may take on any of the values 0 through 7. The horizontal axis of the quadrant is labelled I.sub.5, 4, 3 and may take on any of the values 0 through 7 as the destination (D) of a move instruction. Since the source and destination locations of a move instruction may respectively vary from 0 to 7, the entire upper righthand quadrant is required with a move class of instructions. Additionally, in the lower lefthand quadrant identified by I.sub.7 and I.sub.6, both being logic 0, there is a move instruction designated as 0D6. The D may take on any of the values 0 through 7 and thus requires one entire row in the lower lefthand quadrant. It is noted, however, that the source destination code is a binary 6. Thus, only one 8-bit unit is required for this instruction. This, coupled with the 8 8-bit inputs required for the class of move instructions in the upper righthand quadrant of the instruction map, yield a result that the move class of instructions occupy 9/32nds of the instruction map, there being 32 8-bit blocks illustrated in the instruction map.

With respect to the jump class of instructions, the tcc in colums I.sub.5, I.sub.4 and I.sub.3 refers to a conditioned true jump. For example, if the code cc (which is one of the flags carry, zero, sign, or parity, associated with the ALU section of the CPU) is of a value equal to t, a jump will occur. The binary codes for the respective condition flag codes are also shown in Table V.

Again with reference to the move class of instruction, the moves are specified by a three-bit source code S and a three-bit destination code D; hence, it is possible to move from register to register, from memory to register, and from register to memory. Memory, of course, refers to the contents of the location specified by the memory address register H,L. In addition to the above operation, a separate instruction is provided for load immediate or load literal. This instruction codes and executes in two bytes. The first byte specifies only destination code; the second byte is the literal source data.

The arithmetic codes are similar to the move codes occuping 5/16 of the instruction map shown in FIG. 6, except that the three-bit destination field is instead a 3-bit op-code field P. The destination is implied to be the accumulator A. The source is specified in the same manner as above for the move instructions, including the register, memory and immediate formats. The eight operation codes are add (AD), add with carry (AC), subtract (SU), Subtract with borrow (SB), and (ND), or (OR), exclusive or (XR), and compare (CP). In all of the arithmetics, except compare, the accumulator is combined with the source and the result replaced in the accumulator.

In addition to the above arithmetics, shift right circular (SRC) and shift left circular (SLC) are provided in separate codes. The shift instructions operate on the accumulator and the carry flag and provide a mechanism for provisional branching of specific bits of the accumulator.

The arithmetic, logical, and shift instructions all implicity update the four hardware flags associated with the ALU. These flags are used by conditional jump instructions as the condition code. Compare updates the flags like subtract.

The jump instructions occupy 3/8 of the instruction map in FIG. 6. One 3-bit field is used to distinguish eight different types of jumps. Another 2-bit field of the instruction is used to select a particular condition code from the four hardware flags. The jump can be conditional or unconditional. If conditional, it can be conditional true or conditional false. Finally, it can be a subroutine or not subroutine jump. For all of the above jumps, the address is taken as the literal two bits immediately following the jump instruction. If the jump is executed, these two bytes are inserted in the program counter and the program jumps to that location. If a subroutine jump is executed, the previous program counter is stored into the program address stack. In addition to the above jumps, a separate code is used to allow a return from subroutine jumps. The return can also be conditional or unconditional with the true or false condition. Of course, the return address is the last program counter address stored into the program address stack. Since the program address stack is seven levels deep, nested subroutine software becomes a convenient, efficient and effective alternative to indirect addressing.

Input/output instructions occupy 1/8th of the instruction map shown in FIG. 6. The external instruction contains a five bit "dont't care" which has no meaning to the internal operation of the CPU. It is for the programmer's utilization and the peripheral hardware designer's design of external op codes to be executed by the peripheral system. The external instruction does nothing more than put out the internal instruction and accumulator registers into external latches. This is used to create an efficient command and control system for the communication for peripherals to and from the CPU and peripherals to and from memory. The input instruction is the subset of the external instruction with only three don't cares remaining. In this case, the CPU loads the internal accumulator register with selected data. Hence, there is a direct hardware provision for input and output of eight bit characters under program control.

The control instructions occupy very little of the instruction map, but are very important in terms of the operator and programmer convenience they represent. The three important control instructions are HALT, RESTART and CONTINUE. Of these, only restart requires mutiple codes. A 3-bit don't care in the restart op-code is loaded into the three highest order bits of the program address register. Hence, there are actually eight restarts to eight distinct locations at 8K byte increments around the 64K byte memory system. All of these control instructions are available to be used under normal program control. However, in practice, their use in interrupt mode is much more significant.

Insertion of an INTERRUPT instruction into the normal stream of program execution can be accomplished very simply. First, the INTERRUPT key (or more generally, INTERRUPT peripheral) must encode the desired instruction onto the 8-bit data selector. Secondly, it must raise the INTERRUPT line which is a direct input to the instruction control decoder. The decoder will then recognize the INTERRUPT at the completion of the current instruction execution.

Halt and continue do not disturb the operation of the executing program. Restart, however, is the direct abortion of the current program flow. It is not a subroutine call; hence, any desired protection of the current program with respect to desired recognition of INTERRUPTS must be handled by the INTERRUPT program at the location indicated by the restart command code. A simple restart program would store out the current contents of all the CPU registers and store the return address of a program which would recover them. Then, on completion of servicing the INTERRUPT, the INTERRUPT program would terminate and return to the normal program flow. Both hardware (fast) and software (slow) means of priority (INTERRUPT recognition are feasible. The hardware scheme would use an external priority encoder to select the highest order INTERRUPT present. The software scheme would involve a software decision tree at the location of the restart command.

Table II includes a list of the CPU instructions of the present invention. Instructions include register to register load instructions, memory reference load instructions, load immediate instructions, arithmetic and logical register instructions, arithmetic and logical memory reference instructions, arithmetic and logical immediate instructions, shift instructions, jump instructions, subroutine instructions, return instructions, input/output instructions, restart instructions, and halt instructions.

Table III includes the CPU arithmetic/logical Mnemonics and condition flags.

Table IV includes instruction mnemonics and register mnemonics of the CPU.

Table V includes the instruction coding of the CPU.

The truth table of the instruction set is shown in Table VI. In the truth table, a horizontal row represents time and the input/output or internal nodes are printed in a column. Terms printed are listed at the start of the truth table. The instruction table illustrates the changes in the internal registers, one location of the program stack, and the arithmetic/logical flags for each type of instruction. The numbeer of bytes or cycles per each instruction is shown by the number of lines printed for an instruction. A row is printed at the end of the cycle. The inputs, outputs or registers printed in the instruction truth table are listed below. Names listed from top to bottom are respectively the names from left to right in the tables. For the instruction set, there is listed the following:

Ready

Interrupt

Execute

State 1

State 2

State 3

State 4

Input

Instruction Register

A register

B register

C register

D register

E register

H register

L register

Program address

Carry

Zero

Sign

Parity

The instruction set truth table, Table VI, shows the operation of the groups or classes of instructions possible. For each horizontal line in the truth table, one cycle is printed. Time flows in a vertical column. Instructions are labelled in the lefthand column. For purposes of illustration, the restart instruction will first be described. It is assumed that the CPU is in a HALT state. With reference to Table VI, at time 1, it may be seen that there is a transition of logic from 0 to 1 at the interrupt input and ready input. This transition interrupts the CPU HALT state. During the first cycle, the restart instruction will be fetched at the input. This is shown at the input bits 7 through 0. With reference to Table V, it will be noted that the binary code on inputs I.sub.7 through I.sub.0 corresponds to the code for restart operation. The instruction register also shows that the instruction is transferred from the input terminals to the register during the first cycle. During the second cycle, line 3 on the truth table, it may be seen that the three center bits of the instruction bits 5, 4, and 3, are transferred to the high order program address bits 15, 14, and 13 of the program address counter. During time 4 an ADA instruction is executed. The input shows the ADA instruction. With reference to Table V, it may be seen that the instruction 10000000 corresponds to the operand code AD; i.e., bits 5, 4, and 3 are respectively 000. With respect to Table I, it may be seen that the instruction class of arithmetic operations is characterized as 2PS. Bits 7 and 6 of the instruction are respectively 1 and 0 corresponding to the 2. The source destination (bits 2, 1, and 0) are 000 corresponding to Register A (reference Table V). The instruction is transferred to the instruction register. No other operation of the registers is shown except the program counter is incremented by 1. The purpose of this instruction is to update the carry, zero, sign and parity flags. This may be seen by the change in the parity flag logic level. The flags now reflect the status of the A register.

The next instruction is load to B register from memory. Bits 2, 1, and 0 of the instruction (the source of data) are respectively 1,1,1; i.e., 7 which corresponds to memory. The instruction is present at the input during the first cycle time, number 5, during which it is transferred to the instruction register. The program counter is also incremented. During the second cycle time, time 6, the data that is to be transferred to the B register is present at the input. The instruction register does not change until the next instruction is received. In time period 6, it may be seen that the input is transferred to the B register.

The program address counter was not incremented because the instruction was a memory instruction which does not use the program address, but rather uses the H and L registers of the internal RAM for the location.

The fourth instruction executed in the program is an input instruction. The instruction is transferred to the instruction register during the first cycle time, frame number 7. The B register does not change. During the second cycle of the input, the data on the input terminal is transferred to the A register which is shown on time frame 8. It will be noted that the flags are not affected by the transfer of the input to the A register. The flags are updated only by an arithmetic or shift instruction. The fifth instruction return false parity shows that a return instruction occurs. The instruction is transferred by the instruction register. By looking at the program address register, one can see the location of the program counter. Because the program address counter is a fixed location in the RAM, the change in address location will not be shown. The address location will stay the same until a call instruction is executed. For example, during the frame 15. The instruction shift right circular at time frame 10 shows that the A register shifts right one bit and the carry flag is set from bit A.sub. 7 after the shift.

In a manner similar to that above described, it is possible to trace through the instructions of the instruction set, and observe the changed binary data in the various registers of the CPU.

FIG. 7 is a functional block diagram of the sequence and control logic of the CPU. Each block contains a reference to a figure wherein detailed logic circuits suitable for performing that function is described. The CPU logic names and function thereof are described in Table VII.

The state timer, the detailed logic of which is described with reference to FIG. 8 hereinafter, functions as the master timer for the CPU/external memory system. It controls all timing in the CPU, the interface timer and the slave timer on the external RAM. (Reference FIG. 35). There are four state outputs to the state timer, namely S1, S2, S3, and S4. The state timer generates an automatic index output P which updates the address register after an instruction execution. The state timer receives the inputs READY and interrupt (INT REQ) which enable an interrupt instruction to be inserted. These signals are also used in accordance with the present invention to enable use of either a serial type external memory or a random access memory. This feature of the invention will be explained hereinafter with respect to the detailed description of FIG. 8.

As previously explained, a cycle includes both a fetch and execute, each of which is characterized as having four states, S1-S4. Each state has two phases, phase 1 and phase 2. During the fetch cycle time, an instruction is fetched from the external memory. A logic diagram of the input/output (I/O) circuit is illustrated generally in FIG. 9. During the execute time, an instruction is executed. The state timer also contains a programmable logic array (PLA) which enables different amounts of state time or subcycle tine execute/fetch to be programmed by changing a gate mask only. A more versatile processor can be made by using this technique. A PLA is described in more detail in U.S. Pat. No. 3,541,543 entitled "Binary Decoder" issued to R. H. Crawford and assigned to the assignee of the present invention, and incorporated herein by reference.

The input/output section contains the interface to the common 8-bit external bus. During the fetch subcycle of an instruction cycle, the program address location, i.e., the location of the desired instruction in external memory is output through the CPU I/O interface. During state 1, the low address bit P.sub.L are output from the internal RAM and during state time S3 the high order address bits P.sub.H are output. This allows sixteen bits to be output through the common 8-bit bus, enabling use of a memory system of up to 64K words. During phase 2 of state 4 of the fetch subcycle, the external memory location addressed by the 16 bits (FIG. 1, block 12) will output the instruction.

During state 1 of execute, the instruction will be clocked into the CPU through the I/O. The instruction will be stored in the instruction register (FIG. 10). During the four states of execute, the instruction fetched will be executed. If the instruction requires more than one cycle, the address will be put out during the next fetch from either the program address counter or the internal RAM registers H, L. The data will be output from the RAM during the end of the fetch cycle for the second or third instruction cycle.

An instruction stored in the instruction register is input into a programmable logic array which defines the instruction decode. The use of the PLA in the instruction decode allows the instructions decoded to be changed by reprogramming the gate mask.

The cycle timer receives inputs from the instruction decode and the state timer. The cycle counter determines if an instruction is one, two, or three cycles in length. Instruction cycles may be varied by the use of a PLA in the cycle timer circuitry.

The cycle timing information, instruction decode information, and state timing information are combined in an internal control block which includes a read-only memory (ROM), which generates all of the internal CPU timing. The outputs of this ROM go to either the bus, the internal RAM or the arithmetic unit. Outputs of the internal control labelled with an asterisk (*) generates a bus signal while control signals labelled with a $ enable the bus dara to be sampled. Two other outputs of the internal control are *I.sub.3, 4, 5, and *RS. These two controls are used when a restart instruction is executed. During one clock, one state time, the signal *RS will discharge the bus, allowing all zeros to be put in the program stack. This may be seen with reference to FIG. 15 wherein the signal *RS produces a logic 0 at the output of NAND gate 71. This discharges to ground the internal bus lines 0 through 7. The signal *I.sub.3,4,5 transfers the instruction bits I.sub.3,4,5 to the upper order address locations P.sub.H, 5, 6, and 7 respectively. Operation of the restart instruction and transfer of bits 3, 4, and 5 to P.sub.H5, 6, and 7 may be seen in the instruction truth table, Table VI, time frame 3. The signal $I is used to sample the instruction into the instruction register.

Three of the outputs of the internal control A1, A2, A3 go to the internal RAM. These signals define the location as the A, B, C, D, E, H, L or M' registers. Two other controls to the RAM, U and V, select either the registers previously mentioned, the low order address registers P.sub.L, the high order address registers P.sub.H, (reference FIG. 2) or when none of the registers of the RAM are being addressed, the signals U, V enable RAM refresh circuitry. Another RAM control signal is *RAM. This signal generates the bus when a RAM output is desired. Two other controls which go to the RAM are PUSH and POP. These controls operate the pushdown stack in the RAM. The pushdown stack will be explained in further detail with reference to explanation of FIG. 25.

The control signal CZSP and -W are controls that go to the arithmetic unit. The signal CZSP is the signal which samples or updates the carry, zero, sign and parity flags. The output of these flags is combined with the instruction bits I.sub.3,4,5 which are decoded with the arithmetic outputs to determine whether a conditional call, return, or jump is to be executed when these instructions are fetched.

The restart circuitry enables a restart instruction to be executed. This circuitry is effective to discharge the bus and allow all zeros to be put into the program stack. Subsequently, instruction bits I.sub.3,4,5 are enabled for insertion into the three highest order bits of the register P.sub.H.

The input/output logic circuitry for the CPU of the present invention is illustrated in FIG. 9. The intrnal eight-bit parallel bus of the CPU is shown at 81, having bus lines 7 through 0. CPU inputs and outputs occur over the lines A.sub.0 through A.sub.7. The logic interface between the internal bus 81 and the output lines comprises a series of NAND gates shown generally at 83, and NOR gates shown generally at 85. Insulated gate field effect transistor transfer gates 87 connect the output lines to corresponding logic gates which are connected to the internal bus 81. System ground is shown at 89. Operation of the input/output circuit is as follows. When the control signal $M is logic 1, the data on the internal bus 81 will be sampled. Consider, for example, that the data on the internal bus line 1 is a logic 1. When the control signal $M becomes a logic 1, both of the inputs to the NAND gate 83A are 1's, producing a logic 0 as the output. This logic zero output is effective to bias on the MOS transfer gate 87A, connecting the output line A.sub.1 to circuit ground. This transistor will output a current onto the external line A.sub.1 which will be sensed as indicative of a signal on the bus line 1. As a further example, consider that the internal bus line 2 has a signal that is a logic 0. For this situation, when the control signal $M becomes a logic 1, the inputs to the NAND gate 83B are 0 and 1, respectively. This produces an output signal which is a logic 1 and therefore the transfer gate 87B is not energized. Thus, no current flows through the output lines A.sub.2 and a logic 0 for the data bus line 2 is indicated.

During the phase 1 time of a state immediately following sampling of the data on the internal bus 81, an input will be sampled from the same line such as A.sub.1 or A.sub.2 in the previous example. This will occur if the signal *M is a logic 1. For example, when *M becomes a logic 1 during the time that .phi.1 becomes a logic 0, *M logic 1 will be transferred to one of the inputs of the NAND gate 91. As soon as .phi.1 becomes a logic 1, the output of the NAND gate 91 becomes a logic 0. Referring, by way of example, to the NOR gate 85A, the logic 0 resulting at the output of the NAND gate 91 produces a logic 0 for one of the inputs of the NOR gate 85A. Depending upon the level of the input information sampled on the line A.sub.1, a logic 0 or a logic 1 will be transferred by the NOR gate 85A to the internal bus 1. For example, if A.sub. 1 input data to be sampled is a logic 0, both of the inputs to the NOR gate 85A are logic 0's. This will produce a logic 1 output which will be sampled onto the bus line 1. For the situation, however, where the input A.sub.1 is a logic 1, then the NOR gate 85A will produce a logic 0 output. The NOR gates 85 transfer the input information onto the bus lines 81 which are precharged in order to enhance speed of operation.

FIG. 10 illustrates the instruction register of the CPU. The instruction register is a sample and hold register and operates as follows. For clarity of explanation, only one bit, block 54, of the instruction of the 8-bit instruction register will be described. When the control signal $1 is a logic 0, the bus bit 7 will be sampled into the storage register. It is sampled through the complex gate which is an AND-OR invert gate. During phase 2 of the clock, this input will be transferred to the input on the inverter 63 between the phase 1 and phase 2 transfer gates. The output of inverter 63 will be sampled on phase 1. If the control signal $I is now a 1, the bit will be sampled back through the complex gate through the other input of the AND-OR invert. This will allow the bit to be circulated until a new bit is sampled into the instruction register. More particularly, when the signal $I becomes a logic 0, one input of the NAND gate 35 is a logic 1. Assume now, for purposes of illustration, that the data to be sampled on line 7 of the internal bus is a logic 1. Since both inputs to the NAND gate 35 are 1's, the output of the NAND gate 35 is also a logic 1. This ensures that the output of the NOR gate 39 is a logic 0, since it would have a logic 1 output only if both of its inputs were logic 0;s. The logic 0 output of the NOR gate 39 is transferred by the phase 2 transfer gate to the input of the inverter 63. The logic 1 output of the inverter 63 is transferred to the input of the inverter 65 by the phase 1 clock. This logic 1 signal is a feedback signal to the NAND gate 37. If the sample signal $I now becomes a 1, both inputs to the NAND gate 37 will be logic 1 setting the latch, since the output of the NAND gate 37 is a logic 1, ensuring that the output of the NOR gate 39 is a logic 0. This data will be circulated until the sample signal $I again becomes a logic 0. Similarly, it may be shown that a logic 0 on the data bus 7 will be sampled by the NAND gate 35. With this situation, the output of the NAND gate will be a logic 0. Since one of the inputs to the NAND gate 37; i.e., $I, is a logic 0, the output of NAND gate 37 will also be a 0. This will ensure that the output of the NOR gate 39 is a logic 1. This signal will be transferred to the input of the inverter 63 by phase 2 of the clock. At phase 1, the transfer gate will transfer this inverted signal to the input of the inverter 65. Also, the signal will be fed back to one of the inputs of the NAND gate 37. For the situation where the control signal $I has again become logic 1, a signal corresponding to the logic 0 level of the input bus line 7 will be circulated through the register until a subsequent sample signal indicating that new data is to be sampled. Since the internal bus line 81 contains inverted signals of the desired data information, the output from the instruction register, such as 64, will correspond to the true value of the input data.

The instruction decode portion of the CPU is illustrated in FIG. 11. The instruction decode comprises two NAND matrixes 65 and 67. These matrixes are defined by a programmable logic array such as described in the aforementioned Crawford patent. Operation of the decode may better be understood by an illustrative example. Consider the instruction signal JMP. This signal is obtained when signals are present on the output lines from the instruction register I.sub.0, I.sub.6, and I.sub.7. The output of the various instructions such as JMP, HALT, and etc., are coded on the matrix 65. By way of example, the instruction HALT requires a combination of two terms in the matrix 65. These two terms are shown respectively at gates 73 and 75. The NAND matrixes 65 and 67 form an AND-OR matrix. It may be seen that since the instruction decode comprises programmable logic arrays, a great deal of flexibility is afforded the CPU of the present invention. By merely programming the gate mask for the programmable logic arrays, new functions, sequences of information, etc., may be accomplished.

A cycle timer that may be used with the CPU of the present invention is shown in FIG. 12. The timer comprises a NAND matrix 81, the inverting function of which is shown generally by the symbol at 69. The output of the NAND matrix is applied to one terminal of a phase 2 transfer gate shown generally at 83. On phase 2 of the clock, the outputs of a NAND matrix 81 are combined by NAND gates 85A through 85F to form cycle control signals C1, C2A, C2B, C2C, and C3 which are output on phase 1 of the clock by transfer gates 87. The cycle information is fed back on phase 1 of the clock onto the input matrix 81. The only time a change in the cycle information occurs is if the instruction decode described in FIG. 11 has a new or if the state counter described in FIG. 8 has an output of S4 EX.

An example is illustrated by cycle 1 (C1). If the signals EX and S4 are both logic 1, the NAND gate 80 will have a logic 0 output. This will be inverted by the inverter 91 to form a signal S4 EX that is a logic 1. If the control signal Z is also a logic 1, the cycle timer will produce a control signal C1. It may be seen that these two gates, that is, the gate formed by the control line Z and the gate formed by the control signal S4 EX are all the gates in the programmable logic array 81 required to produce a logic 1 on one line of the matrix to the NAND gate 85A. This logic 1 is inverted by gate 69 producing a logic 0 input to NAND gate 85A, thereby assuring a logic 1 output of the NAND gate. The cycle C1 will circulate on itself as long as C1 remains a logic 1 and the signal S4 EX is a logic 1, i.e., when the signals S4 and EX are not logic 1's.

The next change in cycle timing will occur when the signal S4 EX becomes a 1 and a new instruction from the instruction decode, shown generally in FIG. 11, becomes a 1. An example of the second cycle instruction would be cycle C2A. If cycle C1 is a 1 so that a 1 is present at the gate 93, the control signal Z is a logical 0, so that the NAND gate 85A does not produce a logic 1 output, and the signal S4 EX becomes a 1, cycle C2A will have an output of a logic 1 if the instruction line from the instruction decode decoding the control signal EXT + LrM = OM + RST is a 1 (external or load memory to r or arithmetic memory or restart). Cycle C2A will continue to circulate until the next S4 EX logic 1 state since C2A is a 1 and the signal S4 EX is a logic 1, making one input to the NAND gate 85B a logic 0, assuring a logic 1 output therefrom.

The internal control portion of the CPU, in accordance with the present invention, comprises one level of NAND logic shown generally at 95, clocked on phase 2 of the clock signal to discrete MOS NAND gates 97A through 97K. An example of operation of the internal control circuit for producing the output *RAM will be described. If the control signals S4 and execute (EX) are both logical 1's during phase 2 of the clock, they will be gated into NAND gate 97D to produce the *RAM control.

A further example of generation of a control signal by the internal control may be seen by reference to the programmable logic arry 99 and the control signal *RS. This is a signal that is required to generate a restart operation. For the situation where the instruction restart (RST) is a 1, the signal cycle C2A is a 1, EX is a 1, and state 3 (S3) is a 1, the output signal *RS will be true; i.e., a logic 1. This will be clocked from a NAND gate 101 into an inverter 103 on phase 2 of the clock, allowing a restart instruction to be executed. The control signal *RS is combined with phase 1 of the clock through a NAND gate 71. The output of this NAND gate, a logic 0, is connected to the gates of IGFETS 105, biasing these transistors into a conductive state. This discharges the internal bus lines O-7 to ground, enabling all 0's to be inserted into the program address stack. The other control signal for restart instruction is the signal *I.sub.3,4,5. When this signal is a 1, bus bits 4, 3, 2, 1, and 20 will be discharged. Instruction bits 3, 4, and 5 will be transferred to the bus bits 5, 6, and 7 respectively, for storage in the three highest order bits of the high order program address register, as was explained previously with reference to the description of the instruction set of the CPU.

The condition decode for the arithmetic flags of the ALU is shown in FIG. 14. The condition decode comprises the NAND array 111 combined into a 9-input NAND gate 113. By way of example, if instruction bits I.sub.3, I.sub.4, and I.sub.5 are logic 1's, the condition matrix 111 will decode the inverse of the carry flag. The condition output will be combined with a call instruction, jump instruction, or return instruction to determine if the instruction should be executed. If parity is true and a conditional call occurs, bits 3, 4, and 5 will be a 1 and the instruction will be executed.

The state timer of the CPU in the present invention is illustrated in FIG. 8 and is used to control the master timing of the CPU. Control signals used by the CPU and their function are shown in Table V11. The state timer comprises a 4-bit shaft register having outputs S1, S2, S3, and S4. The outputs of the shift register are combined with cycle information and status information from READY and INT (interrupt) to determine if execute or fetch is to be executed. These outputs are programmed into a programmable logic array 604 which enables changing the state operation. The interrupt circuit debalances an interrupt input and synchs it with the state cycle information to determine when an interrupt can be instituted. Another

In general, the state timer comprises an edge detector 600, a programmable logic array 604, a storage circuit 602 for storing an interrupt request until the interrupt request has been acknowledged, a storage register 606 for storing an interrupt acknowledge for several states, and a series of shift register bits.

The edge detector 600 detects a 0 to 1 transition of the interrupt request signal. If this signal makes a 1 to 0 transistion, no effect will occur in the circuit. This, of course, can be changed so that the inverse occurs by changing the one-shot detection circuit.

Synchronization at the interrupt request signal with CPU timing is as follows. When a logic 0 to 1 transition of interrupt control signal occurs, the NAND gate 601 will output a 1 state, phase 1 to phase 1 pulse to storage circuit 602. This is accomplished by the edge detector circuit. For example, if .phi.1 or if interrupt was previously at a 0 level, then the output of the inverter 608 will be a logical 1. This signal will become one input to the NAND gate 601 during phase 2 of the clock. The other input to the NAND gate 601 is a logic 0; i.e., the same as the input to the inverter 608. For this situation, the NAND gate output 610 will be a logic 1. When the interrupt request signal makes a transition to a logical 1 during phase 1, the input to the NAND gate 601 will change and since a logic 1 was stored previously on the other input to the NAND gate 601, the output 610 will become a zero pulse. This pulse will be transferred to the phase 1 transfer gate 611 to the NAND gate 612, producing an output therefrom that is a logic 1. During the phase 2 of the clock, the logic 0 stored on gate 608 will be transferred to the NAND gate 601, returning the output to a logic 1.

During the zero pulse output from the NAND gate 601 in response to the 0 to 1 transition of the interrupt signal, the logic output of NAND gate 612 becomes a logic 1. This logic 1 will continue to circulate through the NAND gate 613 back through the NAND gate 612 during subsequent phase 1 and phase 2 of the clock until the input 614 to NAND gate 613 changes to a logic 0. The input 614 was previously at a logic 1 level.

Tracing through the circuit, it can be seen that prior to the signal 614 becoming a logic 1, the output of the inverter 616 is a logic 1. This enables the gate 619 in the programmable logic array 604 to a logic 1. During the next time that X (labelled node 620) becomes a logic 0, the output of the NAND gate 622 will change to a logic 1. This will enable the gate 624 in the PLA 604 to a logic 1. During the next time that the signal EX (execute) is true, gate 626 of the PLA will become a logic 1. Similarly, when the signal S4 becomes a logic 1, this will enable gate 628 of the PLA. When the control signal HALT is a logic 0, the inverter 621 will assure that the gate 619 becomes a logic 1. With this combination of signals, the interrupt will be acknowledged. An output is provided through two levels of NAND logic such that the output 632 is applied to the shift register which has the four states S1, S2, S3, and S4 as outputs. To synchronize the control when the interrupt is recognized, gates 634 and 636 will become logic 1's. This will enable the NAND gate 638 to provide an output which will shift through a two-bit delay to allow the RAM addressing to be set up properly. During the start of this two-bit delay, the flag, shown generally at block 606, will be set. The inverter 641 will invert the logic 1 output of the NAND gate 638 to a logic 0 which will form one input to the NAND gate 640. The output of the NAND gate 640 will then become a logic 1 which will be transferred to the NAND gate 643 by the phase 1 transfer gate. The NAND gate 640 will set the interrupt recognize latch so that the external control timing can accept interrupt data through the data terminal on the interface logic. During the next time of this two-bit delay, the interrupt recognize latch block 602 will be set to logic 1 because line 614 is transformed to a logic 0 by inverters 641, 645, and 647 operating on the output of the NAND gate 638. This enables the latch 602 to be reset. This circuitry, it may be seen, completely synchronizes the interrupt request and interrupt acknowledge and the state operation of the CPU.

Table VIII illustrates the control signal timing for the bus action of the control cycle. As an example, reference the instruction RST (restart). The first signal P.sub.L shown as being present during state 1 of FETCH is the control that ensures that the low order address bits are transferred from the RAM onto the internal bus for use in fetching the restart instruction. During the FETCH subcycle, state S1, phase 1, the control signal P.sub.L must be present. This signal is effective to discharge the bus on phase 2 of the clock, the bus being precharged during the phase 1. Another control signal is the signal M'. During the execute subcycle, state 1, phase 1, M' must be present to ensure a transfer of the restart instruction from the external bus to the internal bus. When the signal is present on the internal bus, it is sampled in the instruction register by the control signal I. (Reference the description of FIG. 10). The signal I is generated on execute, state 1, phase 2.

This section will describe the operation and logic of the arithmetic logic

Operation of the arithmetic control section, FIG. 27, is as follows. By way of example, consider instruction bits I.sub.3, I.sub.4, and I.sub.5 to be true, i.e., logic 1. This coding corresponds to the instruction compare (reference Table V). A control signal -P which is generated by the state timer of the CPU (described in FIG. 8) to update the address register after an instruction is applied as one input to NAND gates 88, 98, and 102. The other input to these NAND gates is respectively bits I.sub.3, I.sub.4, and I.sub.5 of the instruction register. When the control signal -P is a 1, the output of gate 88 will be a logic 0 because -P and I.sub.5 are both 1's. The output of the inverter 90 will be a logic 1. The logic 0 of NAND gate 88 ensures a logic 1 at the output of the NAND gate 94 since at least one of its inputs is a 0. The NAND gate 96 thus has one input that is a logic 1. The second inpupt to NAND gate 96 is from the control -W. This control must be a 1 to assure that the bootstrapped load on the NAND gate 96 will continually refresh its capacitance. It is understood by those skilled in the MOS art that a bootstrapped load may be used to drive a large capacitance on the output of a device. The capacitance must be continually refreshed or the logic value will drift from its true value.

Because the output of gate 94 and -W are both logic 1's, the output of NAND gate 92 will determine the logic level output of gate 96 (for the present example). The NAND gate 92 receives inputs from both the inverter 104 and 100 which, in turn, respectively receive inputs from NAND gates 102 and 98. For the present example where instruction bits I.sub.3, I.sub.4, and I.sub.5 are logic 1's, the output of NAND gates 98 and 102 are 0's, producing an output from inverters 100 and 104 that are respectively logic 1's. These logic 1 signals control the NAND gate 92, producing a logic 0 output therefrom. Thus, the output of NAND gate 96 is controlled to be a logic 1 selecting the control signal "SU or SB or CP or W" to be a logic 1, causing the compare instruction to be executed. Similar examples can be shown for the other seven arithmetic operations; i.e., by changing the logic of instruction bits I.sub.3, I.sub.4 and I.sub.5 the other arithmetic operations may be selected.

The arithmetic unit control and operation will continually do the instruction code present in the instruction register (bits 3, 4, and 5) although an arithmetic instruction may not be executed. The only time the results of the arithmetic operation will be sampled will be when a *F control signal from the control decode is present. This may be seen by reference to FIG. 22, which illustrates one bit of the arithmetic unit. Gate 86 is the control for generating the bus from the arithmetic unit. If the control signal *F is a logic 1, the bus will be generated during phase 2 of the clock. During the logic 0 of phase 1, the transfer gate 106 will transfer the *F logic 1 command to the input of the NAND gate 86. Since phase 1 is a logic 0, the output of the gate 86 will remain at logic 1 as long as phase 1 remains logic 0. When phase 1 goes to logic 1, however, gate 86 is enabled such that a logic 0 is outputed. A logic 0 on the input of NOR gate 84 (which is part of the precharged bus) will transfer the output 108 (F.sub.n) of the arithmetic unit to the bus during phase 2; that is, when F.sub.n is logic 1, the signal F.sub.n is logic 0, producing a logic 0 at the input of NOR gate 84. Since both inputs to NOR gate 84 are now 0's, a logic 1 output is produced on the bus.

Since the signal *F will be a logic 0 during the next phase 1, the gate 84 will not be enabled until further *F signals occur. The times at which the signal *F occur are shown in Table VIII. It may be seen that the signal F occurs only during phase 1 of states 2 and 4, for both execute and fetch subcycles. As an example, consider subcycle fetch, state S4, phase 1. To simplify the control, the signal *F occurs for every instruction at this time. At some times, such as cycle C2A, the results of the arithmetic unit are not desired. This is indicated by a blank spot in the Table VIII during the phase 2 time of that state. The result of the arithmetic unit will not be stored in any register at that time.

The temporary storage register of the CPU is shown in FIG. 18. The logic of one bit of the storage register is illustrated; other bits of the temporary storage register are shown generally in block diagram form at 114, 116, 118, 120, 122, 124, and 126. The internal bus is shown at 25 as having lines 0 through 7. Signals on these lines are inverted signals as indicated by the label BUS.

Operation of the storage register is as follows. If the control signal $R becomes a logic 1, the complex gate 110 which includes NAND gate 110A, 110B, and NOR gate 110C will be enabled through the BUS input line 0. The output will be stored at the output node of the logic gate 110 until the following phase 1 clock signal. During phase 1, it will be transferred through the inverter gate 112. The output of gate 112 will be transferred back through gate 110B on phase 2 if the control signal $R is now a logic 0. That is, if the signal $R is a logic 0, both inputs to the NAND gate 110B are logic 1's producing a logic 1 output therefrom. This logic 0 is inverted by gate 112 and recirculated. This transfer will continue until the control signal $R becomes a logic 1 again. The inverter 113 inverts the BUS signal on the internal bus 25 so that a true signal is applied to the logic gate 110.

The temporary storage register is also used for the shift right and shift left instructions, or for normal operation. This is controlled by the control signals *R, *RGT and *LFT. If a right shift is desired, the data on bus line 0 will be shifted to line 1. During phase 2 of the clock, after the signal *RGT becomes a logic 1, the shift will occur as follows. The bus line 0 is shifted to line 1 by shifting the logic value of the output of the inverter 112 onto the bus line 1 when control signal 130 is a logic 0. By way of example, if a true signal were present on bus line 0, this would be represented there as a logic 0 since the bus 25 is inverted. The logic 1 would be present at the output of inverter 112 representing the true data value. The NOR gate 134A thus has an input 130 of logic 0 and an input from inverter 112 of logic 1, producing an output on inverted bus line 1 of logic 0, effecting the shift right.

Similarly, if the shift left were required, a logic 1 input *LFT would provide an output 132 that would be applied to one of the inputs to the NOR gate 134B. The other input to the NOR gate 134B would be the output of the inverter 112. The output of the NOR gate 134B is connected to bus line 7. Thus, for a shift left, the signal on bus line 0 is shifted to bus line 7.

If normal operation were desired, the input *R going to a logic 1 would produce an output 128 that would be applied to the NOR gate 134C. This would recirculate the data on the bus line 0 back to bus line 0.

Circuitry for incrementing the program address is also illustrated in FIG. 18. The signal for accomplishing a program address increment is the control signal No. P. As mentioned previously, this signal is generated by the state timer circuitry described with reference to FIG. 8. If the control signal No. P is a logic 1, the output of NOR gate 136 will be a logic 0. The output of NAND gates 138, 140, 142, 144, 146, 148, and 150 will be a logic 1 since the No. P signal is inverted by the inverter 139. This will ensure that the complement of 1 is added through the arithmetic unit. This is done because the input to the arithmetic unit, gates 136 through 150, are inverted inputs. The incrementing occurs during the fetch subcycle, state 1, and the fetch subcycle, state 3. The incremented output occurs during state times S2 and S4 of the fetch subcycle.

With reference to FIG. 19, there is illustrated one bit of the arithmetic unit in logic form generally at 67A. The other seven bits of the arithmetic unit are shown in block diagram form at 67B through 67I. The arithmetic unit comprises inverters shown generally at 59, complex gates 61, NAND gates 60, NOR gates 62, ring dot gates 58 (which are the inverse of exclusive ORs) and transfer gates 63. These logic gates are interconnected such that in response to a preselected code on instruction bits I.sub.5, I.sub.4, and I.sub.3, eight separate arithmetic operations may be effected. Operation of the logic in performing an add instruction is shown in FIG. 22. The logic associated with the subtract instruction and exclusive or instructions are shown respectively in FIGS. 23 and 24.

With reference now to FIG. 22, description of the logic of one bit of the arithmetic logic unit in performing an add instruction will be described. If an add instruction is desired, the control signal 152 will be a logic 0. This signal is labelled SU + SB + CP + W. It will be recalled that this signal is generated by the arithmetic logic unit control circuit described previously with respect to FIG. 17. With reference to FIG. 17, it may be seen that if the instruction bits I.sub.5, I.sub.4, and I.sub.3 are respectively 000, the code corresponding to add (reference Table V) then the output of AND gate 96 will be a logic 0. Hence, the control signal 152 is 0 when an add instruction is desired. Similarly, the inverse of the control signal 152 labelled 154 in FIG. 22, is a logic 1.

Signal 154 is applied to the OR gate 155A and the inverter 73A. The output of the inverter 73A is applied to one input of the OR gate 155B. The input signal X.sub.n which is outputed from the temporary storage register illustrated in FIG. 18, is applied directly to the other input of the OR gate 155B. The signal X.sub.n is inverted by inverter 73B to produce the true signal X.sub.n which is applied to the other input of the OR gate 155A and also to an input of the ring dot gate 75. The logic 1 level of the control signal 154, i.e., SU + SB + CP enables operation of the arithmetic unit. The output of the NAND gate 74 will be bit X.sub.n. For example, consider a situation where the bit X.sub.n from the temporary storage register such as 114 in FIG. 18, is a logic 1. For this situation, when the control signal 154 goes to a logic 1, the inputs to the OR gate 155B are logic 0 from the inversion 73A, and 0 from the signal X.sub.n. This produces an output of the OR gate 155B of a logic 0, which forms one input to NAND gate 74. Further, one input to the OR gate 155A is a logic 1 corresponding to the control signal 154. The other input to the OR gate 155A is a logic 1 corresponding to the true value of the bit X.sub.n. This produces an output of the OR gate 155A of a logic 1, which is also an input to the NAND gate 74. Thus, NAND gate 74 has inputs of 0 and 1 corresponding to an output of a logic 1 which, in turn, is the logic level assumed for the bit X.sub.n. In a similar manner, it may be shown that where the logic bit X.sub.n is a 0, a logic 0 will be produced at the output of NAND gate 74.

The inverse of the exclusive or of bit X.sub.n and Y.sub.n will be formed by the circle dot gate 75, one input of which is the signal X.sub.n taken from the output of the inverter 73B, and the true signal Y.sub.n. The output of the gate 75 is labelled X.sub.n .sym. Y.sub.n. This output is part of the sum and carry of the sum X.sub.n and carry C.sub.n. The carry will first be traced. The output of gate 70 will be the NAND sum of X.sub.n and Y.sub.n. This output is shown as X.sub.n . Y.sub.n. If these bits X.sub.n and Y.sub.n are both logic 1, a carry must be generated. This is accomplished since logic 1 inputs to the NAND gate 70 effecting a logic 0 output which is effective to operate the transfer gate 158, generating a carry (C.sub.n) signal since clock phase 2 is normally a logic 1 level. If a carry is not generated, then a carry may be propagated by the inverse of the exclusive or of X.sub.n or Y.sub.n, shown at the output of gate 75. This occurs because the output X.sub.n .sym. Y.sub.n is transferred through gate 176 to logic gate 160. The carry will be propagated from node 92 through gate 160 to the next bit C.sub.n. The input to gate 82 will be either C.sub.n.sub.-1 or a generate from the NAND term X.sub.n.sub.-1 and Y.sub.n.sub.-1.

The sum F.sub.n is expressed as F.sub.n = X.sub.n .sym. Y.sub.n + C.sub.n.sub.-1 ; C.sub.n = X.sub.n . Y.sub.n + C.sub.n.sub.-1 (X.sub.n .sym. Y.sub.n). The sum F.sub.n will be generated through the inverse exclusive OR gate 78 and will be the exclusive or of the carry C.sub.n.sub.-1 and excuslive or of X.sub.n or Y.sub.n. The sum term at the output of gate 78 generates the bus if a *F occurs for that time frame at NAND gate 86.

FIG. 23 shows the operation of the ALU logic to effect subtraction. As may be seen from the sum term F.sub.n = X.sub.n .sym. Y.sub.n .sym. C.sub.n.sub.-1, the difference is the same for subtract as add. The only difference between operation of the two circuits is that the X.sub.n input to the carry equation is inverted. This may be seen at the output of the NAND gate 70, where the output is shown as Y.sub.n . X.sub.n. Otherwise, operation of the subtract logic is the same as for add as described with reference to FIG. 22.

With respect to FIG. 24, operation of an exclusive or instruction will be described. For this example, the sum F.sub.n is equal to X.sub.n .sym. Y.sub.n. For an exclusive or operation, the signal XR must be a logic 1. This signal will be generated when the instruction bits I.sub.3, I.sub.4, and I.sub.5 are respectively 101. (Reference Table V). With reference back to FIG. 17, it may be seen that such a coding for instruction register bits I.sub.3, I.sub.4, and I.sub.5 will produce a logic 1 at the output of the NAND gate 96A, producing an output signal XR that is a logic 1. The signal ND + XR is applied to the transfer gate 164. This signal will be a logic 0 when the signal ND + XR + W is a logic 1. The transfer gate 164 will be activated by the signal 162 such that a logic 1 is transferred to the input of gate 168. The signal 166 is the inverse of exclusive or of the two terms X.sub.n and Y.sub.n. The inverse of this exclusive or will be combined with the logic 1 present at the input of gate 168. The output of gate 168 forms the exclusive or of the terms X.sub.n and Y.sub.n. This output will be gated onto the bus when the control signal *F is a logic 1 at the NAND gate 86.

In a similar manner, the logic associated with the other arithmetic operations may be traced through the logic of the arithmetic unit.

A precharged parity circuit in accordance with the present invention includes the advantage of increased speed of operation. This circuit will be described with reference to FIG. 20. The parity circuit includes precharged insulated gate field effect transistors shown generally at 174, having gate inputs labelled phase 1. Inputs from the bus are labelled F and F. For the illustrative 8-bit parity circuit of FIG. 20, there are eight F signals, F.sub.0 through F.sub.7, and eight F signals, F.sub.0 through F.sub.7. These inputs are selectively applied to gates of interconnected insulated gate field effect transistors to produce odd parity and even parity outputs.

Operation of the precharged parity circuit is as follows. During phase 1 of the clock, nodes shown generally at 170 are precharged to a reference voltage V.sub.DD. During the high part of the clock; that is, when transistors 174 are non-conductive, nodes 170 are conditionally discharged depending on the inputs F and F to the parity circuit. By way of example, node 170A would be discharged if F.sub.1 and F.sub.0 were both logic 0's, of if F.sub.1 and F.sub.0 were both logic 0's. With reference to node 170B, the inverse occurs; i.e., node 170B discharges if inputs F.sub.1 and F.sub.0 are both logic 0's or if inputs F.sub.1 and F.sub.0 are both logic 0's. Thus, the functions of the inputs at 178 may be expressed as F.sub.0 .F.sub.1, the function at 180 As F.sub.1 .F.sub.0 ; the function at 182 as F.sub.1 . F.sub.0 ; and the function at 184 as F.sub.1 . F.sub.0. The functions 178 and 180 are combined at node 170A to result in the exclusive or function F.sub.0 .sym. F.sub.1. Similarly, the functions 182 and 184 are combined at node 170B to yield F.sub.0 .sym. F.sub.1. In other words, the node 170A will be discharged only if signals F.sub.0 and F.sub.1 are opposite logic levels. If both inputs are logic 1's, or if both are logic 0's, the node will not discharge. Similarly, the inverse is true for node 170B; i.e., the node will not discharge for the situation where the signal F.sub.0 and F.sub.1 are opposite. Parity is the exclusive or of all of the bits that parity is to be checked for. The circuit can be expanded for as many bits as desired. Parity is the exclusive or term. The inverse of exclusive or is called even parity.

The arithmetic unit flags carry (C), zero (Z), sign (S) and parity (P) will be described with reference to FIG. 21. The sign flag indicates the status of bit 7 of an arithmetic sum. If bit 7 is a logic 1, the sign flag will be true; if bit 7 is a 0, the sign flag will be false. Operation is as follows. Consider first the example where bit 7 is a logic 1. It will be noted that the bus shown generally at 25 is inverted. Thus, a logic 1 for bit 7 will be present on line 7 as logic 0. This logic 0 is inverted by the inverter 700 to produce a true data bit signal. This logic 1 will be transferred on phase 2 of the clock by the transfer gate 701 to form one input of the AND gate 702. The other input of the AND gate 702 is formed by the control signal No. CZSP. When this control signal goes to a logic 1, the output of AND gate 702 is a logic 1. This ensures that the output of the NOR gate 706 is a logic 0. On the subsequent phase 1 of the clock, this logic zero is inverted by the inverter 708 to form the true or logic 1 output of the sign flag. This logic 1 is transferred back through the transfer gate 709 on phase 2 of the clock to form one input of the AND gate 704. If the control signal No. CZSP goes to a logic 0, the other input to the AND gate 704 will be a logic 1. This will ensure that the logic 1 level of the sign flag will recirculate until BUS 7 is again sampled. Similarly, if the data on bus bit 7 is a true logic 0, the inverted signal logic 1 will be present on the inverted bus. This logic 1 will be inverted by the inverter 700 to form a logic 0 as an output of the inverter 700. This logic 0 is transferred by the transfer gate 701 during phase 2 of the clock to one input of the AND gate 702. A 0 input to the AND gate ensures that the output of the AND gate 702 is also a 0. Similarly, when the control signal No. CZSP goes to a logic 1, it is applied as one of the inputs of AND gate 704, ensuring a logic 0 output therefrom. Thus, both inputs to the NOR gate 706 are logic 0's, producing an output that is a logic 1. The logic 1 output is transferred during phase 1 by the transfer gate to the inverter 708 which produces a logic 0 output as the sign flag corresponding to the zero level of bit seven of the bus.

The zero flag indicates that all of the inputs of the arithmetic unit are logic 0's; that is, bits 0 through 7 are all logic 0's. For example, for the situation where all of the bits 0 through 7 are 0's, a logic 1 will be present on these lines since the bus 25 is inverted. This will enable the NAND gate 710 producing a logic 0 output therefrom. This logic 0 output is inverted by the inverter 712 to provide a logic 1 as an input to the AND gate 714. The other input to the AND gate 714 is also a logic 1 when the control signal No. CZSP is true. Thus, the output at the AND gate 714 is a logic 1 ensuring that the output of the NOR gate 716 is a logic 0. During phase 1 of the clock, the logic 0 output of the NOR gate 716 is inverted by inverter 718 such that the Z flag is a logic 1. Similarly, it may be seen if any of the bus bits 0 through 7 are logic 1's rather than 0's, the output of the NAND gate 710 will be a logic 1, and the level of the Z flag will be 0.

The parity flag indicates that there is an odd number of 1 bits in the 8-bit output of the arithmetic unit. The details of the parity circuit itself were described with reference to FIG. 20. The logic for producing the parity output (P) in response to the control signal No. CZSP going to a logic 1 is similar to that described with reference to the sign and zero flags.

The flag carry is updated if a carry out of bit 7 of the arithmetic unit occurs. The carry flag will also be updated for a shift right or shift left instruction. The other flags are not affected. A shift left operation will indicate the least significant bit of the 8-bit output after the shift. A shift right will indicate the most significant bit of the 8-bit output. By way of example, the control signal No. SLC is the inverse of the control signal *LFT which controls a shift left command. Operation of this signal was described with reference to FIG. 18. When the signal *LFT becomes a logic 1, this generates shift left command. Thus, the control signal SLC is a logic 0. This logic 0 is applied as an input to the NOR gate 722. As may be seen, the other input to the NOR gate 722 samples the least significant bit, bit 0, of the bus 25. If this bit is a logic 1, it will be represented on the bus bit 0 as a logic 0 since the bus is inverted. The logic 0 will enable the NOR gate 722 to provide a logic 1 output. This will be sampled through the OR gate 724 providing a carry sign of logic 1. Similarly, it may be seen that the NOR gate 726 samples the most significant bit; i.e., bus bit 7 after a shift right instruction.

The carry flag is also set if a carry out of bit 7 in the arithmetic unit occurs. For this situation, a logic 1 is applied to the input of the inverter 728. This is clocked during phase 1 to an input of the NOR gate 730 producing a logic 0 at this input. The output of NOR gate 730 will be a logic 1 if the other input to the gate is a logic 0. With reference back to FIG. 17, it may be seen that the other input to the NOR gate 730 will be a logic 0 for the situation where the instruction bits I.sub.3, I.sub.4, and I.sub.5 are respectively 110, 010, 100, or 111, the codes corresponding to the arithmetic operations OR, SU, ND, and CP. The logic 1 output of NOR gate 730 is inverted at 732 to provide a logic 0 input to the NOR gate 734. The other input to the NOR gate 734 is an inverted control signal No. CZSP. Thus, when this control signal goes to a logic 1, its inverse, logic 0, is applied to the NOR gate 734 producing a logic 1 output therefrom. This logic 1 output is sampled through the OR gate 724 to produce a logic 1 carry flag. This signal is connected to the control portion of the ALU illustrated in FIG. 17 to generate the signal C.sub.in. It may be seen that the value of C.sub.in will be a logic 0 when the carry flag is a logic 1 and instruction bit I.sub.3 is a 1 and the instruction bit I.sub.5 is a 0. The signal C.sub.in is applied as an input to the arithmetic unit described with reference to FIG. 19.

The status of the carry flag will be recirculated by NAND gate 734 for the following situation. Assume, for purposes of illustration, that the carry flag is a logic 1. The output of the NOR gate 734 will recirculate this logic 1 when both of its inputs are logic 0's. One of its inputs is an inverted output from OR gate 724. The output of the OR gate 724 is a logic 1 for the situation where the carry flag is a logic 1 so its inverted level will be the required logic 0. The other logic 0 input to the NOR gate 734 is produced by the NAND gate 736. The output of the NAND gate 736 will be a logic 0 when all of its inputs are logic 1's. One of its inputs is the inverted control signal No. CZSP. For the situation where this control signal is a logic 0; that is, when it is not desired to update the status of the flags, then its inverted signal, logic 1, will be one of the inputs to the NAND gate 736. The other two inputs to the NAND gate 736 are respectively No. SLC and No. SRC, the signals corresponding to shift left and shift right instructions. For the situation where no shifts are desired, these signals will have a logic level of 1 since they are the inverse of the control signal specifying the shift instructions. Thus, when it is not desired to update the status of the flags, and when no shift instructions are being executed, the NAND gate 736 will have an output of logic 0, allowing the status of the carry flag to be recirculated.

The internal RAM of the CPU in accordance with the present invention is illustrated in FIG. 25. The RAM contains 192 bits of data storage organized into 24 8-bit registers. Eight of these registers are the low order address registers (P.sub.L); eight are the high order address registers (P.sub.H); and eight form general purpose memory registers, seven of which are generally available and one of which is used only internally. The sixteen registers used for program address, P.sub.L and P.sub.H, enable 16-bit addressing. Only one program address register is used at a time, the other seven are used for a pushdown stack subroutine call operation.

With reference now to FIG. 25, one section of the RAM is illustrated generally at 200 and illustrates one bit of each of the three types of registers, i.e., general purpose registers (A, B, C, D, E, H, L and M'), high order address registers P.sub.H and low order address registers P.sub.L. There are eight sections similar to block 200. Operation of the RAM is as follows. the RAM control signals U and V, shown generally at 202, are coded to select either the low address registers P.sub.L, the high address registers P.sub.H, the general purpose registers, or a refresh counter. By way of example, if the control signal U and V are both logic 1's, during phase 1 of the clock the transfer gates shown at 201 will provide logic 1 inputs to the NAND gate 224. The output of the NAND gate 224 is a logic 0. This signal is inverted at 226 to a logic 1 level. This logic 1 is applied as an input to the inverting buffer shown generally at 230 and also to the inverter 228. A logic zero output of the inverter 228 is applied to the gate of the MOS devices shown at 216, enabling these devices for address line selection, as will be described hereinafter.

The logic 1 output of the inverter 226 is gated by phase 2 of the clock to the input of the inverter 230A. The output of this inverter is a logic 0. This logic 0 is gated during phase 1 of the clock to the input of inverter 230B. Thus, it may be seen that two delays are imparted to the output of the inverter 226. The logic 0 input to the inverter 230B is applied to one input of the complex logic shown generally at 220, and in particular, to one input of the NOR gate 220A. The other input to the NOR gate 220A is the BUS line 221.

The logic 0 on line 234 (which is the input to the inverter 230B) will transfer the data on the bus line 221 onto the line labelled 236 which is the output of the OR gate 220B. This line 236 accesses one column of the internal storage cells of the RAM. Thus, when a logic 0 is present on the line 234, it is possible to write data into the general purpose registers selected by the control signals U and V, both being logic 1's. It is to be understood, of course, that other selections of U and V would have addressed the high order address registers (P.sub.H) or the low order address registers (P.sub.L).

It may be seen that the data present on the inverted bus 221 will be reproduced on the line 236. For example, consider that a true logic 1 signal is present on the bus line 221. Since the bus is inverted, it will be present on the bus as a logic 0 level. This logic 0 level will combine with the logic 0 input from line 234 of the NOR gate 220A to produce a logic 1 output of that NOR gate. This logic 1 output will be sampled by the OR gate 220B to produce a logic 1 output on the line 236, which would enable storage of a logic 1.

For purposes of illustration, let us assume that it is desired to store information into the internal storage cell of the RAM shown generally at 232 which is bit D.sub.i of the D register. The i in the present example, of course, may be any of the bits 0 through 7. In response to a logic 0 level of the line 234, the data on the bus 221 will be transferred to the line 236. This line will access all of the registeres A, B, C, D, E, H, L, and M'. To select the block 232 for storage of data, the input lines A.sub.1, A.sub.2, and A.sub.3 must respectively be 1, 1, and 0. This code, for example, corresponds to the source and destination codes for selecting the D register as set forth in Table V; that is, for a destination of a D register, the instruction bits 2, 1, and 0 must respectively be 0, 1, and 1. This coding is effective to select the D register as follows. The output of inverter 228 is a logic 0 which enables the transfer gates 216. The A.sub.1, A.sub.2, and A.sub.3 logic levels are thus respectively applied to the inverters 212A, 212B, and 212C, resulting in respective outputs of logic 0, logic 0, logic 1. The signals A.sub.1, A.sub.2 and A.sub.3, or the complement of these signals, are respectively applied as inputs to NAND gates shown generally at 215. It may be seen that for the situation A.sub.1, A.sub.2, and A.sub.3 are respectively logic 1, logic 1, and logic 0, only the NAND gate 215A has inputs that are all logic 1's. The output of this NAND gate, logic 0, is inverted by inverter 217 to a logic 1. This logic 1 is transferred on phase 1 of the clock to the input of inverters 219, labelled .phi.1' and .phi.2' which are coupled to the storage cell 232. The output of the inverter .phi.1' is labelled 244 and the output of the inverter .phi.2' is labelled 242. As will be explained hereinafter (FIG. 26), the inverters labelled .phi.1' and .phi.2' are clocked inverters which are used advantageously to address the memory cells.

During the time phase 2, the write line 242 of the memory cell will be enabled. The read line of the memory cell is shown at 244, while the output line is shown at 248. Detailed description of the operation of the dynamic random access memory cell 232 is explained in the description of FIGS. 4A and 4B. During phase 1 of the clock, if the signal 234 is low, indicating the selection of one of the registers, the register output will be selected through the complex gate 260 to node 252. At this point, if the generate signal *RAM becomes a logic 1, the NAND gate 254 will be activated and the output transferred to the BUS. During phase 2, information may be written into the storage cell, such as 232, from the bus bar via the line 236.

Consider, for example, that a 0 is stored in the memory cell 232 and that it is desired to read this data. Thus, the line 244 is biased to a logic 0 during phase 1 and the logic 0 data stored in the cell is transferred to the output line 248. This logic 0 forms one input to the NOR gate 250A. The pther input to the NOR gate 250A is the line 234. This signal is also a logic 0. Thus, the output of NOR gate 250A is a logic 1. This logic 1 is transferred through the OR gate 250B and is transferred to one of the inputs of the NOR gate 254 during phase 1. This ensures that the output of the NOR gate 254 is a logic 0. This logic 0 is stored on the bus line 221. Similarly, if a logic 1 were stored in the memory cell 232, the output of the NOR gate 250A would be a logic 0. Thus, the output of the OR gate 250B would also be a zero forming one input to the NOR gate 254. In response to the signal *RAM becoming a logic 1, the other input to the NOR gate 254 becomes a logic 0. This enables a logic 1 output to be distributed on the bus.

In a similar manner, the high order address registers P.sub.H and the low order address registers P.sub.L could be addressed by a control signal U V combination of UV and UV respectively.

Other operational portions of the RAM include a stack pointer, a refresh counter, and the program locations P.sub.L or P.sub.H. The stack pointer will continually point to one location in the pushdown stack. This location will be the current program address. If the RAM input signals U and V are codes 01 or 10, this will produce an output from the NAND gate 255 of logic 1. This logic 1 is inverted by inverter 257 and enables the transfer gate shown generally at 256. These transfer gates enable outputs from the stack pointer S1, S2, and S3. These outputs are respectively connected to inverters 212A, 212B, and 212C. Depending on the logic levels of S1, S2 and S3, one of the levels, that is rows, in the RAM of the P.sub.H registers or P.sub.L registers will be selected. Whether it is the P.sub.H register or the P.sub.L register depends upon whether the coding of U and V was 10 or 01. If a call instruction or return instruction is executed, the stack pointer address will be changed by changing the count in the stack pointer.

The logic of the stack pointer is illustrated in FIG. 27. The stack pointer comprises an up-down counter and has two inputs, pop and push. For every pop signal, the counter will increase by one count. A push will decrease the count by one. This counter will store the new location of the program address until another return or call is executed. Calls will cause the counter to count one direction and returns will make the counter count in the other direction. From this, it may be seen that the stack pointer provides a convenient and advantages method for subroutine addressing.

Operation of the stack pointer may better be seen by reference to the truth table, Table IX. The inputs, outputs, or registers printed in the stack truth table are as follows, when names listed from top to bottom are ##SPC2## respectively the names from left to right on the table.

Ready

Interrupt

Input

Instruction Register

A Register

Address level 0

Address level 1

Address level 3

Address level 6

Address level 7

Carry

Zero

Sign

Parity

With reference to Table IX, the first two instructions are RST and ADA. These instructions are operative to set the program address level to zero and set the carry flag to reflect status of the A register. The next instruction is a jump true zero (JTZ). The instruction is transferred during time 5 since the zero flag is true. During the next two times, first the low address bits will be input and then the high order address bits. These bits are shown in the program address level zero during time 7. During time 8, a jump true carry (JTC) is executed. Because carry is in a false state, this instruction is not executed. The next instruction is a call true parity (CTP). This instruction will not be executed since parity is not true. The following instruction call is an unconditional call and will be executed. The program counter will continue to increment for the three cycles of the call. This is shown in address level zero. Because it is a call, the address level will be changed to address level 1 in the STACK. The inputs during times 15 and 16 are transferred to address level 1 and are shown during time 16. During time 52, a jump false zero is instituted. The jump false zero charge can be seen in address level 1. During the next instruction time 53, an unconditional return is executed. It may be seen that the address level 1 is updated but that program address controls returned to stack level 0. This is shown by looking at the address counter during time 54. The address level 0 counter was updated and address level 1 remains the same. The next instruction is a return true zero. This instruction will not be executed because the zero flag is at a zero state and control will remain at level zero. The next instruction is a return false parity. This instruction returns control from address level zero to address level 7 as may be seen during time 57 wherein the address level 0 is not incremented but address level 7 is. Because the stack is an up-down counter, when address level 0 is reached, an additional return will transfer control back to level 7.

Again with reference to FIG. 25, the other combinations of U and V input signals is the situation where U and V are both logic 0's. For this situation, the NAND gate 227 is enabled providing a logic 1 input signal to the refresh counter and a logic 0 input to the gate of the transfer devices shown generally at 258. This signal enables the output to the refresh counter R1, R2, and R3 to be coupled respectively to the inverters 212A, 212B, and 212C. This will enable one entire row of the RAM to be refreshed. The refresh counter will increment one every time U and V are both selected as logic 0's. The counter counts from 0 to 7. This allows the eight rows in the RAM to be refreshed after eight count pulses. The instruction control is designed so at least one refresh occurs during an instruction. If no instructions are being executed because of a wait state or a halt condition in the CPU, the refresh counter will continually refresh the dynamic random access memory, ensuring that all data will remain valid. Table VIII illustrates when the refresh of each instruction cycle occurs. Referring to Table VIII, it may be seen that during fetch subcycle times S1, S2, S3, and S4, the registers P.sub.L or P.sub.H are always being accessed. Thus, refresh cannot occur at this time. During state 1 of execute, however, it may be seen that the RAM is never accessed. This is the time that refresh occurs.

FIG. 26 illustrates a clocked inverter utilized in accordance with the random access memory of the present invention. The inverter is used for the read and write lines of the RAM storage cells. The clock signal itself is used as the low voltage. When the clock is low, the output will be valid and will reflect the inverse of the input signal. If the clock is high, that is, logic 1, the output will always remain a 1 and the memory cell will not be addressed. The clocked inverter of the present invention provides several advantages. Illustrative of the advantages are the fact that the precharge conditional discharge technique of the present invention does not load the clock with as much capacitance as conventional precharge techniques. Further, since the clock does not carry discharge current, the clock noise is reduced. This is an advantage over conventional techniques where circuits are very sensitive to clock noise.

With reference to FIG. 27, the detailed logic of the stack pointer is illustrated. As may be seen, there are three outputs labelled S1, S2, S3. These outputs come from blocks labelled T which represents a toggle flip-flop, the logic of which is illustrated in FIG. 29. Operation of such a flip-flop is understood by those skilled in the art and need not be explained in further detail herein. The toggle flip-flop is also used in the refresh counter illustrated in FIG. 28.

This section describes the input/output interface shown in block diagram form at 16 in FIG. 1. The functional block diagram showing the various elements in the system timing is illustrated in FIG. 30. FIG. 31 shows the logic elements of the functional blocks illustrated in FIG. 30. The system interface includes interconnection of the CPU and, by way of example, external random access memory units. This interconnect is by a parallel 8-bit external bus. The interconnect of the CPU with 1k of memory of the external random access memory is illustrated in FIG. 32. As has been previously described, the external random access memory may be up to 64K bytes since 16 bits are used for addressing the memory. As understood by those skilled in the art, when an external memory of this size is used, the memory will be formed on a plurality of chips. A chip select signal from a memory interface circuit shown in FIG. 36 is supplied as an input to the external random access memory to select the required chips.

The CPU controls the system timing; thus, timing signals from the CPU are applied to an external timer which is illustrated in FIG. 35. The external timer has an output that is applied to external timing logic illustrated in FIG. 34. The external timing logic also receives an input from the CPU. The output of the external timing is connected to the interface control and timing block which synchronizes CPU/RAM/peripheral operation. External system inputs are applied to this block illustrated in FIG. 33, and system outputs are obtained therefrom. This circuitry supplies outputs to the external memory and to the memory interface.

An output truth table, Table XI, shows the five output states corresponding to changes in the READY or INTERRUPT input. Inputs lines A.sub.7 - A.sub.0 are separated into input and output terminals for the truth table. ##SPC3##

With reference now to FIGS. 32A and 32B, interconnection of the CPU and 1k of memory of the random access memory is illustrated. As may be seen, only eight external bus wires are needed for this interconnection. The CPU input/output section was described previously with respect to FIG. 9. As described therein, the CPU input/output lines occur along lines A.sub.0 through A.sub.7. These eight bus lines are interconnected with the various units of the external RAM. These various RAM memories are referenced generally by the numeral 301. Preferably, these memories are 1,024 .times. 1 dynamic random access memories. Techniques for fabricating RAMs are well known to those skilled in the art and need not be explained in detail herein. Circuitry for interfacing the RAM, refreshing it, etc., are described hereinafter with respect to FIGS. 37-44.

As may be seen, a chip select signal is applied to each of the units 301 to enable selection of the proper unit. An advantage of the interconnect system illustrated in FIGS. 32A and 32B is the fact that multiplexing the address, input and output simplifies interconnection since only eight memory bus lines are all that are required. If multiplexing is not used, it is necessary to route 26 bus lines. This is illustrated in FIG. 32C, which is a conventional way of accessing external memory.

The memory chip select is illustrated in FIG. 36. By using four 16 - output decode packages, it is possible to select from 1K to 65K of memory chips. The chip select input on the RAM is clocked so that the chip select will be sampled at the appropriate time. At all other times, the chip select output will be invalid.

FIG. 34 illustrates the external timing for the input/output control. The timing is generated from the output of the CPU and from the output of an external state timer described in more detail in FIG. 35. These signals are combined to select either one of the external storage registers or data inputs to the system from an external periphery.

FIG. 35 is the logic of the external timer. This timer counts the four states of the CPU. To ensure that the external counter runs in synch with the CPU, an output CPU is a synch signal which resets the timer every state 1. The external counter also will synch the external memory to the same frame as the CPU. This assures that the external memory will input and output on the correct state.

FIG. 33 shows the interface logic used with the current sense/voltage input of the CPU. The connection to the CPU is shown at node 300 (A.sub.1). For example, this connection may go to any of the A.sub.0 through A.sub.7 input lines to the CPU. Eight of the circuits illustrated in FIG. 33 are required for an 8-bus system. Node 300 is connected to the A outputs on the external memory and the A outputs on the CPU. During a low phase 1, the data selector 302 is enabled. Either input DATA, DMAH, DMAL, or M' is selected. The input DATA is used to load information from peripherals to the CPU or RAM. If the processor is halted, information may be loaded directly into the memory. The data must be present during state 3, and the control signal store must be logic 1. When the processor is operating the input DATA is selected during the instruction time execute state 1 of an interrupt acknowledge, or the data time execute, state 1 of an external instruction, or store = 1, state 3.

The signal DMAL corresponds to eight low order address bits being selected for a direct memory access through the DMAL match. The input is selected during execute state 2, or ready = 0, state 2, or store = 1, state 2.

The input DMAH corresponds to the eight high order bits for a direct memory access through the DMAH latch. Selection of the input is during execute, state 4, or ready = 0, state 4, or store = 1, state 4.

To load the memory when the CPU is halted, the input STORE is used. STORE must be a logic 1 from the start of state 2 to the end of the following state 1. The four state store will allow the location of memories addressed by DMAL and DMAH to memorize the byte present at the DATA input.

The output of the data selector 302 is amplified through the transistor 304. This enables all of the A lines of the external RAMs to be driven. During the low phase 2 of the clock, the CPU or RAM will output a current. This current is sensed by the sense amplifier shown generally at block 314. Such amplifiers are known to those skilled in the art and they need not be described in more detail herein. The low current is amplified to a TTL voltage level which is input to latches 306, 308, 310, and 312. These are latches which will contain valid information as to the output of the CPU. The M' register closes the current sense/voltage in loop of the CPU. The CDMA register is shown at 308. This register is a TTL latch for direct memory access and contains the byte of information located in the last byte of memory addressed by DMAL or DMAH. The register I' is shown at 310 and contains the last instruction fetched from the memory. The register A' is shown at 312. For each external instruction, this A' register will be updated with the contents of the internal CPU A register.

An example of the current sensing will be described by using the timing charts in Table XII. This table shows the times at which outputs from the CPU or inputs to the CPU will occur. Outputs always occur on phase 2 of a state and inputs occur on phase 1. By way of example, reference the cycle 2 of the EXT instruction. During a fetch subcycle, state S1', phase 2, the CPU will output the contents of the A register as a current. This current will be amplified by the sense amplifier, 314, FIG. 33, and will be an input to the latch 312, labelled CACC and also will be an input to M'. At the end of phase 2, when the data is valid on the sense amplifier 314, the signal CACC will clock in the results of the A register. This register will be updated for every external instruction.

An example of the closed loop current sense and TTL voltage in may be seen for the SHIFT instruction during S1' fetch. During phase 2, the CPU will output a current which corresponds to the contents of the P.sub.L register. This current will be an inversion of the true value. The loop will invert the output at point M'. During phase 1, the RAM will expect the contents of the address. During state 2, phase 1 time, E.sub.1 and E.sub.2 select the input C.sub.0 for the data selector 302. This, it may be seen, is the contents of the M' register or the location desired for low order address and memory. During the phase 1 clock time, the data selector 302 will output the contents of the M' register, and it will be amplified and applied to node 300. This will be the input to the RAM as shown in Table XII fetch S2', phase 1. During the time S4' of a recall cycle, the high order address bits are input to the RAM. During phase 2 times, state S4' executes the data in the RAM address location P.sub.H P.sub.L is output on the current sense lines. During the following execute state, phase 1, the instruction shift will be input into the CPU. During the execute phase of the instruction cycle, the CPU will not request information from the RAM. For purposes of programming, it is possible to have direct memory access while the CPU executes the instruction. This is done by using the DMAH and DMAL inputs. For example, during a recall condition, DMAL will occur during execute state S2', phase 1. This will be input to the data selector 302 to the A.sub.1 line. The memory will accept this address. During the state time S4', the DMAH input to the data selector will be selected and the high order address will transfer to the memory. During the S4' execute phase 2 time, the RAM will output the location of the direct memory address location. The output B will be stored in the latch labelled DMA' (FIG. 33). The clock CDMA will clock in the output of the RAM. It will be stored there for use in the external system.

A truth table illustrating the logic status of the system signals Ready, Interrupt, Store, Execute, Synch, S1 = External State 1, S2 = External

A block diagram of the basic elements of the external memory is shown in FIG. 37. The external memory system in accordance with the present invention provides several advantages. The memory includes a slave timer (FIG. 38) which enables multiplexing of address/data/input/output information. The circuitry also includes a refresh counter for use in the external memory such that refresh will be done automatically. An additional advantage of the circuit is that as a part of the external memory, an address register latch (such as illustrated in FIG. 41) is included. This type of circuitry is conventionally provided external to the memory, thus requiring more interconnects, more space, and resulting in lower reliability. The output of the address register latch is applied to an address decode such as shown in FIGS. 43A and 43B. The decode receives an input from the external bus and also from a chip enable circuit shown in FIG. 39. Input/output logic described in FIG. 40 receives an input from the slave timer and also receives recall instructions. The external storage may be either random access or serial access memory and receives its inputs from the decode circuitry, the refresh counter (if a random access memory is utilized) and from the input/output logic.

With reference to FIG. 38, there is illustrated the slave timer of the external memory. The timer receives an input signal S1 which is a synch input ensuring that the slave timer operates in synchronism with the CPU master timer. The slave timer counts the four states of, by way of example, the random access memory. During state 1, the RAM will automatically refresh. The refresh circuitry will be described in more detail hereinafter with reference to FIG. 42. During state 2, the RAM will accept low order address bits. During state 3, it will accept data and during state 4, it will accept high order address and output the data location. When the synch signal S.sub.1 is a logic 0, nodes 412 and 414 will be set at a logic 1. This is the first state time. The counter illustrated is a conventional Johnson counter and will count through the Johnson states. Such counters are well known to those skilled in the art and detailed operation need not be included herein.

FIG. 39 illustrates the chip select sample and hold circuitry utilized in accordance with the external memory of the present invention. This circuit is necessary since the chip select signal will continually change and it is necessary to clock in the correct chip signal at the appropriate time. During the counter time C.sub.0 C.sub.1, shown at node 416, the input chip select (CS) will be clocked if the signal at node 416 is a logic 0. A logic 0 at node 416 will provide a logic 1 at one input of the AND gate 417. If the chip select signal CS is a logic 1, the output of the AND gate 417 will also be a logic 1. This will ensure that the output of the NOR gate 419 is a logic 0. This logic 0 will be inverted by the inverter 421 after phase 2 of the clock providing a logic 1 output signal CS'. If the signal C.sub.0 C.sub.1 now becomes a logic 1, AND gate 423 will recirculate the chip select signal CS' until a new sample signal is received; i.e., when the signal C.sub.0 C.sub.1 again becomes logic 0.

The input/output logic is illustrated in FIG. 40. One of the inputs to the input/output interface is the recall control signal R. If the recall signal R is a logic 0, and the chip select signal CS' (reference FIG. 39) is a logic 1, then the output of NAND gate 420 will be a logic 0. This will enable the AND gate 422. The output 406 of the shift register will be transferred to the data in line 424 of the external memory. If the signal CS' is a logic 0, then the output of the NAND gate 420 will be a logic 1. This will select the output path 400 and the output will be transferred through the AND gate 425 and the NOR gate 427 back to the input line 424. If the control signal recall R is a logic 1 and the signal CS' is a logic 1 and the signal C.sub.0 C.sub.1 is a logic 1, then the output of the NAND gate 403 will be a logic 0. This will form one input to the OR gate 431. The other input to the OR gate 431 is an inverted signal of the output line 400. For this situation, the signal on line 400 will be transferred to the input/output pin of the RAM. The input line 244 and the output line 400 come from the data storage cell selected in the external memory.

From the above, it may be seen that data will be output if the signal C.sub.0 C.sub.1 is a 1, R is a 1, and CS' is a 1. Data will be input when signal R is a 1 or a 0 and CS' is a 0.

FIG. 41 shows the address latch used in conjunction with the external memory of the present invention. It is a sample and hold latch and operation of this latch is similar to the temporary storage register which was described in detail with reference to FIG. 18. The first address bits A.sub.XO through A.sub.X4 will be held in the sample and hold registers until the address bits change again. The data will be clocked in during the time C.sub.0 C.sub.1 = logic 1, and will be clocked through RAM pins A.sub.0 -A.sub.4. Bits A.sub.5, A.sub.6, and A.sub.7 will be clocked through a series of inverters 431 so that they are delayed and will be present as outputs at the same time as the outputs A.sub.X0, A.sub.X1, etc. The low order address bits A.sub.0 - A.sub.7 will be clocked in during state 2. During the state 4 time, the remaining two high order address bits necessary to address the 1024 .times. 1 bit external memory will be clocked through logic inputs A.sub.0 and A.sub.1. They will be transferred through a phase 1 delay to address lines A.sub.Y3 and A.sub. Y4. Thus, it may be seen that using only an eight-bit bus, the ten bits necessary to address the external memory are provided.

FIG. 43A shows the address multiplexer associated with the external memory. The address lines A.sub.X0 through A.sub.X4 will be decoded as soon as they are sampled in the sample and hold storage register. This enables a faster decoding for the first and second level Y decode illustrated in FIG. 43B. The first level Y decoder will decode bits A.sub.Y0 through A.sub.Y2 while the second level Y decoder will decode the bits A.sub.Y3 and A.sub.Y4.

FIG. 42 shows the RAM refresh counter logic. The refresh counter is used to automatically refresh the 32 rows of memory present in the RAM. During every state 1, one row is refreshed. This has the advantage in that it is unnecessary to halt the CPU in order to refresh the dynamic external memory. During state 2, the memory will expect low order address bits. During state 3, data stored in the memory must be present. During state 4, the memory will expect high order address bits. The inputs will all be on phase 1. During state 2, the memory will output if the location requested at the chip select input at a low logic level. During state 1, the memory will neither accept inputs nor will it output data, as this is the refresh state. With reference to table X, the logic names assigned to the external memory and function of each is described.

FIG. 44 illustrates typical MOS clocks .phi..sub.1 and .phi..sub.2 and TTL clocks .phi..sub.1, and .phi..sub.2.

The external read only memory illustrated in FIG. 1 by block diagram is preferably a 1024 .times. 8 memory. Conventional read only memories may be utilized. Such memories are well known in the art and need not be explained in greater detail herein. Typically, the ROM would contain fixed subroutine programs.

With reference to FIG. 45, a different embodiment of the present invention is described. In this embodiment, there is described a system which includes simultaneous operation of two CPU's such as described in accordance with the present invention commonly sharing the external memory and external memory interface circuitry in accordance with the present invention.