US 4,649,470 AGrant

Data Processing System

Issue Date:1987-03-10

•4 Claims

•26 Drawing Sheets

Abstract

A data processing system using microcode architecture in which a two-level microcode system comprises one or more first, or "horizontal", microinstructions and a plurality of second, or "vertical", microinstruction portions in a vertical microcontrol store. In a preferred embodiment the vertical microinstruction portions include one or more "modifier" fields, a selection field for selecting a horizontal microinstruction and a sequencing field for selecting the next vertical microinstruction portion of a sequence thereof, one or more fields of the horizontal microinstructions being capable of modification by the vertical modifier fields in order to form output microinstructions for performing data processing operations. Unique bus protocol signals are generated to prevent simultaneous access to the system bus by two competing system components and to permit substantially immediate control of the systems bus by a component without requiring a CPU decision thereon. Further, a unique system I/O interface unit permits access to certain I/O components via other I/O buses, such unit utilizing a unique polling technique to identify on an updated basis, all components present on one of such other buses. The system I/O interface unit also includes a unique frequency synthesizer unit for providing at least one clock signal having a substantially constant frequency which can be generated in response to any one of a plurality of input clock signals each having a different frequency.

Metadata

Assignee

Data General Corporation

Inventors

David H. Bernstein
Edward M. Buckley
Roger W. March
Ronald I. Gusowski, deceased

Application Information

Application Number:US 4353854

Filing Date:1982-10-20

Priority Date:1980-02-11

Art Unit:237

Classifications

IPC:

G06F 1338

Field of Search:

364370200 MS File;900 MS File85

Patent Drawings (26 sheets)

Description

This application is related to the following applications:

Ser. No. 120,272 now U.S. Pat. No. 4,394,736--Filed Feb. 11, 1980

Ser. No. 120,271 now U.S. Pat. No. 4,371,925--Filed Feb. 11, 1980

Ser. No. 120,291 now U.S. Pat. No. 4,464,772--Filed Feb. 11, 1980.

Background of the Invention

1. Field of the Invention

This invention relates to data processing systems using micro-processors and, more particularly, to systems using two-level microcode architecture.

2. Description of Related Art

Data processing systems have generally been developed to provide system configurations which range from compact, singleboard microcomputers to more complex, high performance minicomputers. Such systems use microcode architecture in which macroinstructions are suitably decoded so as to provide access to a microinstruction or to a sequence of more than one microinstruction obtained from a suitable data store thereof.

Generally, in such systems, for example, a macroinstruction is appropriately supplied from a macroinstruction register to suitable decoding logic so as to provide a starting address for access in the microinstruction data store (sometimes referred to as the microcode store) of an initial microinstruction of a sequence thereof. The accessed microinstruction includes control information for performing the instruction designated and sequence information for determining the microaddress of the next microinstruction of the sequence. Each sequential microinstruction contains the same kind of information until the last microinstruction of the sequence has been accessed at which point the microinstruction routine having been completed, the system is ready to decode the next macroinstruction.

Such systems normally require a relatively large microcode data store utilizing microinstruction words which are relatively wide (i.e., they contain a relatively large number of bits) so as to contain the required control and sequencing information. While the use of relatively wide microinstruction words provides higher speed operation (i.e., a large number of bits are simultaneously available in parallel to provide the control and sequencing operations) such systems tend to be more costly not only because the number of storage bits in microcode data store becomes relatively high but the data paths for handling a wide microinstruction word become more complex and the system requires more expensive components and data path configurations.

In order to reduce the data storage space required for the microinstructions and to avoid handling a large number of "wide" instruction words, certain microcode systems have utilized "two-level" microcode store techniques as opposed to one level microcode stores as discussed above. Such two-level configurations arise from the recognition that control information in the microinstruction words are often common to a large number of the microinstructions. Therefore, in order to avoid the repetitive storage of the same relatively large number of data bits required to store all of the control and sequencing information for each microinstruction separately, control information, which is common to many microinstructions, is stored in one ROM store separately from sequencing information which is stored in a different ROM store. At the "first level" of operation the sequencing process is performed at the sequence microcode store ROM to produce sequential addresses to access control information in the control microcode store ROM which at the "second level" of operation provides the control information required to sequentially perform the particular microinstruction involved, which latter information may be common to many microinstructions.

Such two-level approach tends to reduce the microcode storage space required in comparison with one-level microcode systems. Such two-level technique, however, may tend to provide only a limited microcode operational capability since the limited control store capacity may permit the system to be designed for a specified set of microinstructions with no real flexibility for providing for expansion of the basic microinstruction set in order to increase the operational capability of the overall system.

Summary of the Invention

Although not limited thereto, the invention claimed herein can be particularly useful in a system which utilizes an extension of the two-level concept which increases the power of a two-level microcode system in that it provides a capability for greatly expanding the microinstruction set without increasing the size of the microcode control store. In accordance therewith, the system includes what has been termed an "orthogonal" microcontrol store in which a first, or "vertical", microcontrol store provides a "narrow" microinstruction word portion having one field comprising a selected number of vertical microinstruction bits for selecting one of a plurality of second or "horizontal" microinstructions from a second level, or horizontal, microcontrol store, one or more "modifier" fields, as described more fully below, and a sequencing field for presenting the address of the next (i.e., each successive) vertical microinstruction in a sequence thereof. Each horizontal microinstruction represents a basic function to be performed but, unlike prior systems, one or more of the fields thereof are subject to modification by the modifier field of the vertical microinstructions. For example, prior two-level systems use a vertical sequence store which merely specifies the address of a particular control word in a horizontal control store, the control word requiring the movement of data from a first specified register (e.g., general register GR1) to a second specified register (e.g., general register GR2). In contrast the two-level microcode system of the invention might, for example, specify, with a vertical microinstruction, a horizontal instruction which requires the movement of data from an unspecified location to an unspecified location. Such microinstruction would be common to a large number of data movement instructions. The vertical microinstruction could then also contain in its modifier fields the specific source and destination locations for such data movement and thereby modify the basic horizontal microinstruction so as to identify such locations therein.

Further, the two-level microcode system of the invention can be arranged so that the vertical microinstructions can be fetched from control stores which are located directly on the microprocessor chip or, alternatively, from external microcontrol sources. The horizontal microcontrol store can also be located directly on the microprocessor chip.

The power of each horizontal microinstruction is effectively multiplied by the modifications of such microinstructions which can be achieved by the vertical microinstructions, such technique being sometimes designated herein as a two-level "micro-modification" technique. Since each horizontal microinstruction can be modified by a large number of vertical microinstructions to represent thereby a large number of overall unique microinstructions, a relatively small number of horizontal microinstructions can be used to provide a general, but efficient set of basic control operations, sometimes referred to as "control primitives".

The particular system of the invention which is claimed herein includes the capability for providing communication with peripheral devices via more than one input/output (I/O) bus. For example, the system in a preferred embodiment provides a system I/O interface unit which includes means for identifying all devices which are currently present on a selected one of the I/O buses using unique "polling" techniques for such purposes as discussed below. Further, the system I/O interface unit is arranged to provide direct control of the transfer of information through the system I/O interface unit to the selected bus which has been polled and indirect control of an auxiliary interface unit for providing transfer of information through the auxiliary unit to another selected bus which has not been polled.

The system I/O interface unit also utilizes one or more counter signals, each of which has a substantially constant frequency and is derived from a system input clock signal. The system I/O interface unit, however, may receive a system input clock signal which has any one of a plurality of different, but known, frequencies. Accordingly, such unit includes a unique frequency synthesizer unit which can respond to any of the plurality of different frequency input signal and can still derive the desired one or more counter signals therefrom at the desired substantially constant frequencies involved.

Further, the system of the invention as claimed herein can also be particularly useful in a system which includes a unique system bus protocol which prevents simultaneous access to the system bus by two competing system components. Moreover, the system bus protocol permits a system component to obtain control of the system bus directly without the requirement for a bus control decision to be made by the central processor unit, so long as all other system components have indicated that the system bus is ready for access and so long as no other system component has asserted a signal indicating its refusal to relinquish prior control of the system bus.

Description of the Invention

The various aspects of the invention can be described more fully with the help of the accompanying drawings wherein:

FIG. 1 shows a broad block diagram of an overall system in accordance with the invention;

FIG. 2 shows a block diagram of an exemplary central processor unit (CPU) of the system of FIG. 1;

FIG. 3 shows a more specific block diagram of the vertical sequence unit of the CPU of FIG. 2;

FIG. 4 shows a more specific block diagram of the micromodification unit of the CPU of FIG. 2;

FIGS. 5-7 show block diagrams of exemplary logic useful in controlling system bus protocol for the system of FIG. 1;

FIGS. 8 and 8A show a block diagram of the system I/O interface unit of FIG. 1;

FIGS. 9 and 9A show more specific logic diagrams of the master control finite state machine and control logic of FIG. 8;

FIGS. 10, 10A and 10B show more specific logic diagrams of the microNOVA.RTM. finite state machine and control logic of FIG. 8A;

FIGS. 11 and 12 show more specific logic diagrams of the microNOVA.RTM. polling logic of FIGS. 8 and 8A;

FIG. 13 shows a more specific block diagram of the frequency synthesizer unit of FIG. 8;

FIG. 14 shows a chart which depicts the parsing algorithm used for the operation of the frequency synthesizer of FIG. 13;

FIG. 15 shows an exemplary timing diagram of the frequency synthesizer output signal as generated from either of two exemplary input signals thereto;

FIG. 16 shows a logic diagram for generating the .mu.EXST signal of FIG. 8A;

FIGS. 17 and 17A show a state flow diagram for the master control finite state machine of the system in FIG. 8;

FIGS. 18 and 18A show a state flow diagram for the microNOVA.RTM. finite state machine of the system in FIG. 8A;

FIG. 19 shows a logic diagram of the interface unit 14A of FIG. 1;

FIG. 20 shows a logic diagram of a programmable interval timer;

FIG. 21 shows a logic diagram of the instruction register and decoder of FIG. 8A;

FIG. 22 shows a logic diagram of the system bus control of FIG. 8; and

FIG. 23 shows a timing diagram for an exemplary operation controlled by the system I/O interface unit of FIG. 1.

An overall system using the techniques of the invention is shown in FIG. 1 wherein a microprocessor chip 10 (identified as a central processor unit or CPU in the figure) is interconnected via an appropriate system bus 11 with a memory system 12, one or more system I/O units 17, one or more external microcontroller chips 13 (identified as "XMC" chips) and a system input/output interface unit 14 (identified as the system I/O interface) and a further interface unit 14A (identified as ECLIPSE.RTM. I/O interface) which latter units interface with appropriate buses 15 and 15A for external input/output devices. For example, in a particular embodiment the system is designed to operate with peripheral (I/O) units of the microNOVA.RTM. type as used with microNOVA.RTM. computer systems as made and sold by Data General Corporation of Westboro, Mass. and with peripheral (I/O) units of the ECLIPSE.RTM. type as used with ECLIPSE.RTM. computer systems made by Data General Corporation.

In a particular embodiment thereof the system bus 11 is a 16-bit parallel system bus, as shown, while microcode bits from the external microcontroller chips 13 are transmitted to the CPU in a time-multiplexed fashion via a suitable dedicated eight-bit microcode bus 16.

FIG. 2 depicts a more detailed block diagram of CPU 10, the eight-bit external microcode bus 16 supplying its microcode bits to an appropriate external microcontroller chip (XMC) interface unit 17A and the system bus 11 being connected to a suitable bus transceiver unit 18 capable of accepting or of supplying appropriate data for transmission to and from the CPU.

The CPU utilizes macro-instruction logic which includes macro-instruction register (IR) 19 and program counter (PC) register 20, together with a suitable register file 21, comprising appropriate general registers and accumulators, and a suitable arithmetic logic unit (ALU) and shifter unit 22 having "A" and "B" inputs and a "C" output, such registers, accumulators, ALU and shifter components being capable of implementation using suitable configurations well known to the art.

A plurality of four internal buses 23, 24, 25 and 26 provides for appropriate internal data paths among such units, as shown in FIG. 2. The C-output of the ALU and shifter logic 22 is supplied to the other units as well as to the system bus via bus transceiver on internal C-bus 23. The A and B inputs to the ALU and shifter logic 22 communicate with the other units via internal A-bus 24 and B-bus 25, respectively. Internal bus 26 provides a direct path from the registers and accumulators to the system bus 11 via bus transceiver 18. The use of such multiple internal data paths permits several concurrent operations to occur within a single microcycle, which in the particular embodiment discussed here is 400 nanoseconds. In accordance therewith, 16-bit register-to-register operations are performed in one microcycle (sometimes referred to as a T-period) and memory-to-register transfers are performed in two microcycles (two T-periods).

The two-level microcontrol stores are depicted as vertical control read-only-memory (ROM) 31 and horizontal control ROM 32. In a particular embodiment of the invention, the first level, or vertical, control ROM 31, provides an 18-bit vertical microinstruction, 6 of such bits being used to select one of a plurality of second level, or horizontal, microinstructions at horizontal microcontrol ROM 32, there being 64 of such instructions stored therein in the particular embodiment shown here. Horizontal control ROM 32 provides horizontal microinstructions having 33 bits, in a particular embodiment, the horizontal control ROM, for example, having a 64.times.33 bit configuration. In the embodiment described, eight bits from the vertical control store 31 are used to provide two 4-bits modifier fields which can be used to modify the horizontal microinstruction which has been selected from the horizontal microcontrol ROM as discussed below. The remaining four bits of each vertical microinstruction are used to control the sequencing of the vertical microinstructions, i.e., to specify the next address mode from vertical control ROM 31 via appropriate vertical sequencing logic 33. The modification of specified fields of the selected horizontal microinstruction by the two modifier fields of the vertical microinstruction is accomplished via appropriate micro-modification logic 34 which thereupon supplies a 35-bit microinstruction word which is suitably decoded by horizontal decode logic 35 to provide the necessary control signals and sequencing signals required for performing the function of the selected microinstruction.

Vertical microinstructions can be fetched either from the CPU-resident vertical control ROM 31 or from an external microcontroller chip 13 (see FIG. 1) via external microcontroller chip interface unit 17A, while horizontal microinstructions are fetched from the CPU-resident horizontal control ROM 32. External microcode can be obtained from an XMC unit 13 by the use of suitable time-multiplexing techniques, the external vertical microcode comprising 16 bits supplied in time-multiplexed 8-bit microcode bus 16.

The sequence of microinstructions required to be decoded by horizontal decode logic unit 35 is begun by a suitable decoding of a macroinstruction from instruction register 19 by a decode control programmed logic array (PLA) unit 40. Decoding of each macroinstruction by the decode control PLA unit 40 provides an initial vertical microinstruction on internal bus 39, as well as supplying on internal bus 41 the address of the next microinstruction which is to be fetched from the vertical control ROM 31. The sequencing of vertical control ROM 31 is appropriately controlled by the 9-bit sequence control signal supplied from vertical sequencer logic 33 via internal bus 43. Once the decode control PLA unit 40 supplies the initial microinstruction and the address for the next microinstruction from vertical control ROM 31, the sequencing of subsequent vertical microinstructions from ROM 31 is determined by the sequencing bits supplied to vertical sequencer unit 33 at internal 4-bit bus 44. If the microinstruction from vertical control ROM 31 requires a branch, or jump, to another vertical microinstruction rather than the selection of a horizontal microinstruction, the modifier bits are used to specify the branching operation and the 6-bits normally used to select a horizontal microinstruction comprise a "no-operation" (NOOP) code which inhibits the micromodification unit 34 from using the 8-bit modifier fields for modification of a horizontal microinstruction.

Thus, in the micro-modification process, two 4-bit fields, for example, of the vertical microinstruction can be substituted for two or more 4-bit fields of the selected horizontal microinstruction. The technique of substituting vertical modifier fields into selected horizontal fields effectively multiplies the power of each horizontal microinstruction so that the overall orthogonal, two-level microcontrol store approach described provides a general, yet bit-efficient, set of basic horizontal microinstruction operations, sometimes referred to as control primitives having powerful capabilities for expanding the number of microinstructions in the microinstruction set of the overall system.

It is helpful in understanding the micro-modification process to consider a complete horizontal microinstruction set stored in horizontal control ROM 32 as set forth in Appendix A.

In a particular embodiment, for example, the horizontal microinstructions each include the following nine fields as shown below: ##STR1##

The A-BUS field defines the source of the data for the A-input of ALU/shifter 22 on A-bus 24 while the B-BUS field defines the source of data for the B-input of the ALU/shifter 22 on B-BUS 25. The ALU field defines the operational function to be performed by the ALU while the SHF field defines the shifter function. The DEST field defines the destination for the data which is placed on C-BUS 23 from the ALU/shifter 21. The ADR field defines the source of the memory address while the MEM fields provides control for the interconnection between the system bus 11 and the C-BUS 23. The TEST field is used to identify various system conditions to be tested, while the RAND (random) field provides for other special control functions. Appendix A sets forth in more detail the meanings of each of the above fields.

The vertical microcontrol ROM 31 is a ROM configuration which is a particular embodiment, for example, includes 288 microinstruction words, each 18 bits wide, one horizontal microinstruction being executed for each vertical microinstruction that is executed. Sequences of vertical microinstructions interpret macro-instructions (i.e. machine instructions as decoded by decode control PLA unit 40).

The vertical microinstructions each involve the following four fields as shown below: ##STR2##

The 6-bit ADRH field defines the address of the horizontal microinstruction which is to be selected in the horizontal main control store 32. The 4-bit V1 field defines the first modifier (sometimes referred to as modifier 1) for the selected horizontal microinstruction while the 4-bit V2 field (modifier 2) defines the second modifier for the selected horizontal microinstruction. A 4-bit NAM field which is supplied to the vertical sequencer logic 33 identifies the next address mode for selecting the next sequential vertical microinstruction, such mode bits being used to generate a new vertical program counter 9-bit address. Appendix B sets forth in more detail the meaning of each of the above fields.

As can be seen in Appendix A, in many of the horizontal microinstructions certain fields are specified as requiring one of the vertical microinstruction modifier fields, identified as either the V1 vertical modifier field or the V2 vertical modifier field. The remaining fields of each of the horizontal microinstructions are fixedly specified as shown. In a particular horizontal microinstruction, for example, identified as a "write memory" (WMEM) microinstruction, all fields thereof are specified except the ABUS field, which requires the use of the V1 vertical modifier field, and the ADR field, which requires the use of the V2 vertical modifier field, the V1 and V2 modifier fields being capable of identifying one of a plurality of sources of the data to be written (V1) and one or more addresses at which said sourced data is to be written (V2). Thus the basic and general write memory (WM) microinstruction can, with modification by the modifier fields, be used to generate a large number of specific "write" microinstructions with regard to a large combination of data sources and address destinations. In the particular embodiment discussed the use of two modifier fields permits a large number of microinstructions to be generated from only 64 basic horizontal microinstructions.

The decode control PLA unit 40 has a configuration which contains microinstructions each having 24 bits. Such unit is addressed with a 16-bit address from the instruction register 19 and provides the initial vertical microinstruction decoded from the macro-instruction and a 9-bit pointer to the next vertical microinstruction in the vertical control ROM 31 when a sequence of more than one microinstruction is required, as well as a flag which may invoke a subsequent macro-instruction for decoding where only a single microinstruction is required, as discussed below. There are as many entries in the decode control PLA unit 40 as there are macro-instructions implemented by the system of the invention.

The five fields of a starting microinstruction of decode control PLA unit 40 are shown below: ##STR3##

The 6-bit ADRH field is the same as that described above with reference to the vertical microcontrol ROM 31, the 4-bit V1 field and the 4-bit V2 field also being the same as those described above with reference to the vertical microcontrol ROM. In addition, the decode control PLA unit supplies a nine-bit ADRV field which is the address in the vertical control ROM 31 of the second microinstruction of the microinstruction sequence required for the decoded macro-instruction. A single bit (the D field) indicates that the single microinstruction provided by the decode control PLA suffices to interpret the macro-instruction. In this case the ADRV field is ignored. A more detailed summary of the five fields of a starting microinstruction is set forth in Appendix C.

The vertical and horizontal control ROMs and the decode control PLA unit are conventional logic units well known to the art and need not be shown in greater detail. Logic for the vertical sequence unit 33 and micromodification unit 34 are shown in FIGS. 3 and 4, respectively.

As can be seen in FIG. 3, for example, the vertical sequencer in a particular embodiment can comprise a vertical program counter unit 100 and a multiplexer unit 101 which selects any one of five input signals to provide (1) for the decoding of the 9-bit output from decode control PLA unit 40, (2) for a jump (branch) or a call operation, identified by the zero-bit (VPC.0.) of the program counter and the vertical modifier fields V1 and V2, (3) for a SKIP operation, i.e., a skip of the next program counter (PC+1) output to the next subsequent program counter output (PC+2), (4) the next program counter (PC+1) output, or a RETURN operation wherein the next program counter output (PC+1) is saved (as when an interrupt operation occurs) and then, following the interrupt routine, the vertical sequencer returns to the (PC+1) output.

The multiplexer operation is controlled by switching control logic 102 which provides for the next address mode of operation (wherein one of the above five operations is selected) when the test conditions required are present.

FIG. 4 shows the micromodification unit 34 which comprises a plurality of 2, or 3, input multiplexers 105 each of which includes either a 4-bit horizontal input (for horizontal fields ABU, BBUS, ALU, SHF, DEST, TEST and RAND), a 3-bit horizontal input (for horizontal ADR field) or a 2-bit horizontal input (for horizontal MEM field) and either a V1 input (horizontal fields ABUS, ALU, RAND), a V2 input (horizontal fields BBUS, SHF, ADR, MEM and TEST) or both V1 and V2 fields (horizontal DEST field), as shown in Appendix A. The 4-bit horizontal inputs (HROM inputs) control the selection of the multiplexer input as either from the HROM bits or the V1 or V2 bits for suuply to the horizontal decode logic 35.

The above discussed embodiment utilizes the particular fields specified for the horizontal microinstructions and set forth in Appendix D. In most cases, as can be seen, the horizontal fields which use the vertical modifier fields are those which specify source and destination locations. For example, for an INCH (increment) microinstruction, the A-BUS field is modified by the V1 vertical modifier field, the data at the location specified by V1 to be incremented being placed on the A-BUS, and the DEST field is modified by the V2 vertical modifier field, the incremented data then being placed in the location specified by V2.

The memory field (MEM) in the particular embodiment described bears special scrutiny. Up to 16 memory operations can be defined by using a 4-bit MEM field. However, as is noted in Appendix D only four operations are specified, ie., NOOP (no operation), RM (read memory), WM (write memory), and a V2 field. In this case the V2 field must be used to specify memory operations other than RM and WM. Thus, as can be seen in specific horizontal microinstruction micro-order encodings of Appendix E, the V2 modifier can specify the other memory operations (i.e., WHM, RMOD, WLM, RHYP, etc.). Thus, the MEM field need only use 2 bits (rather than 4 bits) with a vertical modifier used to specify all operations other than an RM and a WM. Thus, the horizontal control ROM 32 needs to provide only 33 bits, as shown in FIG. 2. However, as seen in Appendix E, and as shown in FIG. 4, 35 bits can be supplied to the micro-modification unit 34, two of the MEM bits 23-26 having no operating significance.

In each case where the memory operation is defiend by the V2 modifier, the V2 modifier cannot be used to specify any other field and only the V1 modifier is free to modify one or more other fields, as appropriate. However, for both RM and WM operations both V1 and V2 modifiers are so available. Since most memory operations are RM or WM operations (where both V1 and V2 modifications are available), the modifier limitation placed on the other (not RM or WM) memory operations is not severe inasmuch as such latter operations are not often used. However, the overall flexibility of the MEM field operations is considerably increased since only four microinstructions are needed to specify memory operation horizontals. The use of the V2 modifier field increases the range of micro-order (control states) for memory operations to 15 additional micro-orders, as shown in Appendix E.

A similar arrangement may be used for other horizontal microinstruction fields, if desired, and the overall horizontal control ROM can be used to specify relatively few microinstructions, the range of micro-orders thereof being considerably increased by defining additional control states by using a vertical modifier field. While the use of one of the two vertical modifiers for such purpose limits the ability to modify more than one other field, in many instances the need to modify more than one other field may not be critical. Obviously, if one increases the number of vertical modifier fields, flexibility increases in this regard at the cost of a larger vertical microinstruction word.

In the limit, it may be possible to use a single horizontal microinstruction with a single bit for each field thereof which bit in each case specifies whether a modifier is to be used for such field or whether a default operation is to occur. In such a system, the number of available vertical modifiers should be sufficient to permit modification of all fields which require it for each definable micro-order. The range of horizontal microinstruction micro-orders can thereby be considerably increased at the expense of the requirement for relatively large vertical microinstruction words having the desired number of modifier fields.

System Bus Protocol

Information transfer (i.e., addresses, data, instructions, etc.) to and from the CPU from system components other than the CPU, such as one or more memory units and one or more input/output (I/O) devices, takes place on the 16-bit system bus 11 either directly or via interface units 14 and 17. An effective system bus protocol must be utilized in order to provide for the correct transfer of such information. In this connection, each bus transaction comprises two segments, one of which can be identified as a "specifier" segment and the other of which can be identified as a "data" segment.

The timing of the system operation is synchronized by two externally generated clocks, identifiable as phase-1 (.phi.-1) and phase-2 (.phi.-2) clocks. The clocks divide a microcycle time period (sometimes referred to as a "T-period") into a .phi.-1 portion and .phi.-2 portion.

During the specifier segment the bus is used to describe the type of bus cycle operation which is occurring (e.g., a memory reference operation or an I/O operation) and information describing the target of the transaction (e.g., an address for a memory reference operation). The specifier segment always occurs during the first .phi.-1 portion of a bus transaction and cannot be extended. During the data segment the bus is used to transfer data as, for example, to or from the addresses or I/O devices specified in the specifier segment. The data segment begins in the first .phi.-2 portion of a bus transaction and may be extended by an integral number of additional microcycles (each having .phi.-1 and .phi.-2 portions). The .phi.-1 and .phi.-2 portions of the T-period may generally be separated by an essentially fixed time gap (e.g., 15 nanoseconds) and, at any rate, must be set up so as to provide no overlapping of the phases.

In operation, a particular system component may be transferring data on the system bus during the data segment of its bus transaction while another component waits to transfer an address during the .phi.-1 portion of the specifier segment of the next bus transaction. In time-multiplexed systems of the type described here, frequent bus "turn-around" operations of this nature give rise to problems. Because of delays in various logic components used in the operation of the system, the transfer of data during the data segment of one bus transaction may not be fully completed by the start of the .phi.-1 portion of the specifier segment of the next bus transaction. Accordingly, the second component may attempt to place its address on the system bus before the data transfer for the previous component has been completed.

Such a condition will produce a relatively high instantaneous surge of current which will tend to have an adverse effect on the system. For example, if system components are formed as TTL circuitry, the current surge will tend to cause undesirable noise effects, i.e., electromagnetic interference, to occur throughout the system as well as tending to provide a long-term adverse impact on the TTL components themselves. Further, if the CPU is formed as an integrated circuit chip, for example, by MOS techniques, such a current surge will also adversely impact upon at least the long term, and possibly the short term, reliability of the MOS chip. Therefore, it is necessary to devise an appropriate technique for avoiding any overlapping of the system component bus driver operations which may result from the action of two different components at a bus turnaround operation in attempting to access the bus simultaneously (a "bus fighting" condition).

One previous solution to such problem has been to utilize a "dead" cycle (i.e., a non-operating cycle) at a bus turnaround operation, e.g., between the completion of each data transfer operation and the start of the next operating cycle, such as an address transfer operation, or vice-versa. During such dead cycle no information can be driven on to the bus and the bus in effect is idle. However, the presence of an extra dead cycle under such conditions either will require a basic microcycle clock of much higher resolution or will cause the operation of the overall system to be slower than desired. Accordingly, the use of a dead cycle is generally an undesirable approach to the problem.

Another proposed solution to the problem is to widen the separation between the two time phases, .phi.-1 and .phi.-2, of the T-period. However, the use of a wider gap lengthens the overall T-period and, accordingly, slows the system down to an even greater extent.

In accordance with the system discussed here the bus access overlap problem is avoided by the suitable generation of an address enable signal (ADREN) and a data enable signal (DATEN). The assertion of an ADREN signal indicates that a valid address is present on the system bus, while the assertion of a DATEN signal indicates that data is present on the system bus. If a DATEN signal is asserted, the address drivers of all other units which are capable of placing an address on the system bus are inhibited from operation. In a similar manner all data drivers are inhibited so long as an ADREN signal is asserted.

The ADREN signal is generated by the system component which has requested the current bus cycle to be started and indicates that information describing the type of bus cycle operation which is occurring has been placed on the bus. The particular system as described herein, for example, can initiate one of four different types of bus cycle operations, namely, a program memory reference, a console memory reference, an I/O operation, or a local memory reference operation. When the ADREN signal is asserted the system identifies which of the above four operations is to occur by placing an appropriate 16-bit word on the bus together with an additional bit as described below.

Thus, for a program memory reference the 16-bit word comprises a 15-bit address in bit positions 1-15, while bit .0. of the word is used in conjunction with an additional bit supplied at an additional pin, designated as the memory cycle (MEMCYC) pin, as follows: ##STR4## In the system described, for example, the memory references may be to a standard NOVA/Eclipse logical address space to be used for program execution in NOVA and Eclipse computers made by Data General Corporation, Westboro, Mass., the operation of which is well known to the art. For such program memory reference, the states of the MEMCYC and bit .0. pins identify the program memory reference operation, while the remaining 15 bits comprise the logical address for the memory reference. The combination of the MEMCYC and bit .0. states are used to define each of the four system operations involved, as follows:

In the case of console memory operation which relates to address space used to store software for console operations, for example, the MEMCYC and bit .0. pins define the console memory reference operation while again the remaining 15 bits define a console address. The local memory reference relates to address space which contains instructions for implementing inter-system communications (e.g., communications between the system CPU and other processors, etc.) and accordingly the system operation word includes a 15 bit local memory address. The I/O operation format which is placed on the system bus is discussed in more detail later.

When the bus type operation descriptor (i.e., the appropriate address or I/O function definition) is placed on the bus the ADREN signal is asserted. When whatever data involved is to be transferred on the bus, for example, during .phi.-2 of the bus cycle, the ADREN signal is no longer asserted. The DATEN signal is asserted to indicate that a data transfer is taking place on the system bus so that all address drivers on the system are inhibited from operation, as mentioned above. The DATEN signal is supplied by the particular system component which is supplying the data which is being placed on the system bus.

The use of ADREN and DATEN signals for such control protects against the bus fighting condition which occurs during bus turnaround operations because of time delays in the system logic and, accordingly, improves the reliability of the system whether used with TTL logic or MOS chips.

The system can be utilized with memories having different operating time cycles. If data at the address signified in .phi.-1 is to be transferred between the CPU and a memory unit external to the CPU, for example, the data must be available for driving on to the bus during the subsequent .phi.-2 so that the transfer can be completed therein. If the data is ready, the memory unit asserts a READY signal at the end of the .phi.-2 portion of the current cycle during which the data has been driven on to the bus and the data transfer has been completed. If the data is not available and has not been driven on to the bus for transfer by the end of the .phi.-2 portion of the current T-period, the READY signal is not asserted and the bus cycle operation is extended causing the DATEN signal to remain asserted for one, or more, T-periods, as necessary, until the data is available and ready for transfer. Logic for generating the READY SIGNAL, the ADREN signal and the DATEN signal is shown in FIGS. 5, and 6.

Further, if the system component which has control of the system bus for a data transfer operation does not intend to relinquish such control, such component must assert a bus locking signal (B/LOCK) in order to maintain control of the bus until the operation it is performing has been completed. For example, a system component may wish to read data, modify the data in some way, and return the modified data to the component for use or for storage, an operation sometimes referred to as a read-modify-write (RMW) operation. Accordingly, such component can not give up control of the bus until the data has been modified and returned to the component, an operation which may require one, or more, T-period cycles. In such situation the system component which is performing the RMW operation asserts a B/LOCK signal until the overall RMW operation has been completed. Such assertion prevents any other unit from obtaining control of the bus. An exemplary logic circuit for generating the B/LOCK signal is depicted in FIG. 7.

Any unit which is in communication with the system bus can be made master of the bus whenever a current bus cycle operation, if any, terminates, control of the bus being relinquished at the beginning of the next T-period, (i.e., so long as a READY signal has been asserted by any unit utilizing the current bus cycle and so long as no B/LOCK signal has been asserted by another unit). Thus, a unit which desires master bus control monitors the READY signal and the B/LOCK signal, requests control of the bus by asserting a bus request (BREQ) signal and is permitted to gain control of the bus under the above described conditions of the READY and B/LOCK signals. Such unit maintains control until it no longer asserts the BREQ signal, so long as a device having a higher bus priority does not also request control of the bus at the end of any bus cycle.

In this way bus control is handled by the unit which desires the control rather than by the CPU. In previous systems bus control is normally handled by the CPU which must entertain a bus request from an external unit and subsequently decide whether or not to provide a "bus grant" signal, or its equivalent. Such a decision-making process by the CPU requires a relatively long time due to the propagation delays of the signals required between the CPU and the various system components with which the CPU must communicate in order to make its decision. In accordance with the procedure used in the system of the invention, i.e., where the unit itself generates a BREQ signal and under the proper conditions thereupon receives control of the bus, bus control requests are handled in a much faster manner and bus control can be handed over to the requesting unit much sooner.

As mentioned above, each of the different types of bus operation require the utilization of the MEMCYC pin which, together with bit .0. of the bus cycle description word (i.e., an address or I/O operation word), defines the type of operation which is to be performed. In accordance with the operation chart set forth above, I/O operations are defined when the MEMCYC and bit .0. of the address word are both zero. For such operating condition the following word format for I/O operations is placed on the system bus: ##STR5## The above I/O instruction effectively represents a re-encoded version of the standard NOVA/Eclipse I/O instruction format received at the CPU, the re-encoded instruction being placed on the bus during the cycle description portion of the current bus cycle. The original microNOVA.RTM./ECLIPSE.RTM. I/O instructions are re-encoded as above in order to facilitate the design of the interface logic between the CPU and the I/O devices.

As can be seen therein, bit .0. is utilized in conjunction with the MEMCYC pin to define the I/O operation. Bit 1 is a non-used bit which has been reserved for possible future use. Bits 2-4 are utilized to define the system, I/O and data channel (DCH) functions as follows:

Bits 5 and 6 define the I/O port which is utilized as follows:

Bit 7 defines the direction of the transfer as follows:

Bits 8 and 9 define the control option which is being utilized, i.e., either a start, a clear, or a pulse operation in accordance with normal ECLIPSE.RTM. operations as follows:

Bits 10-15 define the 6-bit device code which identifies a particular I/O device which is to be utilized during the I/O operation. The above operations are as defined, for example, with previously known ECLIPSE.RTM. computers and, being known to those in the art, need not be defined in further detail.

The use of the above I/O instruction format requires less external logic for handling I/O instructions than was normally required for the handling of I/O instructions using the known previous ECLIPSE.RTM. instruction format. The type of operation, i.e., a memory reference or a I/O operation, is appropriately flagged (by the MEMCYC and bit .0. states) and, if it is an I/O operation, the I/O instruction is sent in accordance with the above format in order to ease the interface operation.

In this connection one particular combination of bits in an I/O instruction is of some significance since it extends the capability of the I/O operation over that previously available to ECLIPSE.RTM. systems. Thus, as can be seen, the I/O ports include not only the normal A, B and C ports utilized in ECLIPSE.RTM. systems but also identifies an additional port capability defined as the "Status" port, the use of which is described below.

Thus, if the I/O instruction identifies a STATUS port in the port field (bits 5,6 are 0,0) and an IN direction in the direction field (bit 7 is 1), such instruction operates as a request to the identified device (identified by device code bits 10-15) to provide information as to its status. The device then returns data in the following format: ##STR6## The device indicates its presence on the system bus by the state of the poll bit 2, and further indicates its "busy" or "done" state by bits 0 and 1, respectively. In addition the device has 13 reserved bits for use in providing further information.

Such format contrasts with the previous ECLIPSE.RTM. instruction format which did not permit the inclusion of such status information (thereby requiring two extra wires for separately generated "busy" and "done" flags) and which was not adaptable for "polling" (i.e., obtaining an indication of the presence of a device on the system bus) as is helpful in the present system described herein.

Accordingly, the revised I/O instruction format, which is supplied to the system bus and which does not require the inclusion of accumulator bits (which are already identified for the CPU by the standard I/O instruction received thereat) provides for the further ability to transfer data not provided for by the standard microNOVA.RTM./ECLIPSE.RTM. I/O instruction format.

System I/O Interface Unit

As can be seen in FIG. 1, the system CPU is arranged to be capable of communicating with I/O devices on a plurality of different buses. For example, the CPU may communicate with I/O devices directly on the system bus 11, with I/O devices on bus 15A designated in a particular embodiment, for example, as the ECLIPSE.RTM. I/O bus for ECLIPSE.RTM. I/O devices, or on bus 15B designated in a particular embodiment, for example, as the microNOVA.RTM. bus for microNOVA.RTM. devices. In accordance with the system described herein, data can be transferred directly to and from the system I/O devices via the system bus 11, as described in more detail below, while data for ECLIPSE.RTM. I/O devices are transferred to and from such devices via ECLIPSE.RTM. bus 15A through a separate ECLIPSE.RTM. I/O interface unit 45, the transfer thereof being controlled by control signals from the system I/O interface unit 14. Further, data for the microNOVA.RTM. I/O devices are transferred to and from such devices on the microNOVA.RTM. bus 15B through the system I/O interface unit. Moreover, the interface unit may include local devices which can be used by the CPU, communication with such local devices also being made via appropriate logic forming part of the interface unit 14.

In summary, the particular embodiment of the I/O interface unit described herein permits communication from the system bus having its own timing and operating characteristics to two other I/O buses 15A and 15B having timing and operating characteristics which differ from each other and from those of the system. Moreover, in a particular embodiment, for example, one I/O bus may handle parallel data while one other may handle serial data. Further, the system can be arranged to communicate with one or more local devices located at the interface unit itself.

A more detailed block diagram of the system I/O interface unit 14 is shown in FIGS. 8 and 8A. As can be seen in FIGS. 8 and 8A, the system I/O interface unit can include a plurality of local, or internal, devices and various control logic circuits therefor, shown as block 50, one exemplary device being a programmable interval timer (PIT) as discussed in more detail below, which devices are of a type already known and used with respect to previously available systems. Thus, the PIT logic has been utilized in the Data General Corporation ECLIPSE.RTM. and microNOVA.RTM. computer systems, as shown and discussed with reference to the following publications:

ECLIPSE.RTM. M/600 Principles of Operation No. 014-000092

ECLIPSE.RTM. S/250 Programmer's Reference Manual No. 014-000611

Interface Designer's Reference NOVA.RTM. and ECLIPSE.RTM. Line Computers No. 015-000031

microNOVA.RTM. Integrated Circuits Data Manual No. 014-999974

The above publications are representative of publications available from Data General Corporation, Westboro, Mass. concerning NOVA.RTM., ECLIPSE.RTM., and microNOVA.RTM. systems made and sold by Data General Corporation. Such publications also contain information useful as background information in understanding not only the aforementioned internal devices but also various other aspects of the system being described herein.

The PIT unit, for example, provides a programming timing signal which produces clock signals at a counter clock rate at one of four selectable frequencies. In a particular embodiment, for example, the programmable interval timer can be arranged to produce counter clock intervals at 1 MHz, 100 KHz, 10 KHz or 1 KHz from a constant input clock signal of 1 MHz through well-known frequency divider techniques. In the system I/O unit, such timer responds to a pseudo-constant 1 MHz signal which is derived from a frequency synthesizer unit 51, as discussed in more detail below.

Other exemplary local devices (not shown) may include other well known circuits, such as, for example a power monitor unit or a "real time clock" (RTC) unit which generates timing pulses that are independent of the system clock rate. The real time clock, for example, can be used to interrupt at any one of four selectable frequencies, for example, at the AC line frequency, or at 10 Hz, 100 Hz or 1000 Hz the latter three clocks all derived from the psuedo-constant 1 MHz signal through well-known frequency divider techniques. The power monitor unit can be used to detect any transition of a power monitor pulse which indicates that AC power has failed or that AC power has returned after a momentary power failure, or that a start up power condition exists.

Further, for convenience, a suitable local random access memory (RAM) storage unit 61 may also be present on the system I/O interface unit for providing temporary storage space which may be used, for example, when performing console operations. Such storage unit is of a well-known RAM type which can be formed to provide storage space of an appropriate size for the desired console use. For example, a particular embodiment may be capable of storing a selected number of 16-bit words for such purpose. As such, the RAM unit 61 does not form a part of the inventive aspects of the system I/O interface unit 14 and need not be discussed in more detail.

Addresses and data are supplied to and from the system bus 11 at the system I/O interface unit 14 in the form of 16-bit words (MB 0-15) supplied at the physical address/data (PADS) pins 63. Suitable driver and multiplexer circuits 64 are utilized for such purpose, the address and data words which are being transferred to or from the system bus being supplied to internal U bus 65 for use in the system I/O interface unit 14 either directly via an internal bus 66 or indirectly via temporary address/data (A/D) register 67 which may be used for temporary storage thereof before such addresses and data are supplied to the system I/O interface unit or to the system bus depending on the direction of transfer thereof.

The I/O buses which must be used to transfer addresses and data between the CPU and the I/O devices with which the CPU may desire communication differ in their timing and operating characteristics and the selection of which bus is to be used depends on which type of I/O devices are involved (e.g., ECLIPSE.RTM. devices which communicate via ECLIPSE.RTM. bus 15A using parallel data, microNOVA.RTM. devices which communicate via microNOVA.RTM. bus 15B using serial data, and all other systems devices which communicate via system bus 11 which in a particular embodiment uses parallel data). When communicating with I/O devices via buses 15A or 15B, the system I/O interface unit 14 must decide which I/O device is involved and, hence, which I/O bus is to be used for the transfer of such data so that the interface unit can control such transfer to and from the CPU. In accordance with the particular system of the invention disclosed in FIGS. 8 and 8A, the system I/O interface unit 14 contains logic circuitry for identifying all of the devices which are on a selected one of the two buses which it controls. Such identification can be achieved by appropriately "polling" the selected one of such buses and storing suitable information as to which devices are on the selected bus.

In accordance with the specific embodiment of the invention described herein, which utilizes either a microNOVA.RTM. I/O bus 15B or an ECLIPSE.RTM. I/O bus 15A, the system I/O interface unit is arranged to "poll" the microNOVA.RTM. I/O bus 15B to determine which devices are on such bus. A suitable microNOVA.RTM. poll counter 68 supplies each of the expected device code words which identify all possible microNOVA.RTM. devices, which device code words, for example, are 6-bit device codes supplied to the microNOVA.RTM. I/O bus 15B via "low-byte" (a byte is 8 bits) shift register 69 from poll counter 68 to the output microNOVA.RTM. bus pin 70 identified as IOD2 in FIG. 8A. The 6-bit device codes are supplied in parallel to shift register (S/R) 69 during the polling process and are in turn supplied serially to the microNOVA.RTM. bus from shift register 69, which bus in accordance with its operating characteristics handles serial-bit information.

When microNOVA.RTM. I/O devices, which communicate with the CPU on the microNOVA.RTM. bus 15B, are interrogated with a device code signal for identification thereof, such devices supply a return identifiable I/O clock (IOCLK) signal at pin 71 to microNOVA.RTM. finite state machine (FSM) and control logic 72 if a device matching a particular device code is present on the microNOVA.RTM. bus. Such operation is in contrast with devices on the ECLIPSE.RTM. bus 15A for which no such identifiable I/O clock signal would be present if such devices were interrogated with the device codes. Because of such contrasting operations microNOVA.RTM. devices can be "polled" to determine their presence on microNOVA.RTM. bus 15B and a record of their presence kept in the system I/O interface unit 14.

Thus, receipt of an IOCLK signal at pin 71 from a particular I/O device on the microNOVA.RTM. bus is supplied to control logic in the microNOVA.RTM. finite state machine and control logic unit 72. Such logic generates a microNOVA.RTM. exist signal (sometimes referred to as .mu.EXST) for supply to a microNOVA.RTM. poll store register 73 to indicate that an I/O device having such device code is present on the microNOVA.RTM. bus.

The microNOVA.RTM. poll store register 73, for example, is a 65-bit register, each bit corresponding to a particular device code (in a particular embodiment there being up to 64 possible microNOVA.RTM. devices which can be present on the microNOVA.RTM. bus). If a 6-bit device code representing a specified device which has been interrogated is provided at the input thereto from microNOVA.RTM. poll counter 68, the particular bit associated therewith will be placed in a state which indicates the presence or absence of such device on the microNOVA.RTM. bus, depending on whether a .mu.EXST signal is so generated. Accordingly, when all of the device codes have been polled, the microNOVA.RTM. poll register 73 contains information as to the presence of all devices which are on the microNOVA.RTM. I/O bus 15B.

Polling of the microNOVA.RTM. bus is first performed at "start-up", the complete set of 64 possible devices in a particular embodiment, for example, being polled during one overall polling time interval before the overall system is put into operation. If, for example, the polling of each device takes 10 microseconds, the complete poll can be accomplished in 640 microseconds. Once the system is in operation, updating of the poll store can be accomplished via periodic polling to provide updated poll information concerning devices which were previously indicated as not being present but which have now been turned on and devices which were previously indicated as being present but which have now been turned off. Such updated polling can be accomplished at a 100 Hz rate, a single device being polled each 0.01 seconds (a completed update polling of all devices occurs every 0.64 seconds, for example).

In some cases a device will be turned on and could, therefore, be present on the microNOVA.RTM. bus but the poll store may not indicate its presence (the updating for such device may not yet have occurred since the device came on line after the last update poll). If, under such condition, the device requests an interrupt operation, the interrupt request itself indicates the presence of the device and the microNOVA.RTM. poll register is automatically updated by appropriate gating logic in the microNOVA.RTM. finite state machine which provides a .mu.EXST signal to the poll register.

All external devices which are not identified as being on the microNOVA.RTM. I/O bus, but which are in communication with the system, will by the process of elimination be either on the ECLIPSE.RTM. I/O bus 15A or on the system bus 11. System I/O devices which are internal to interface unit 14 (e.g. PIT, etc.) can also be appropriately identified as existing in the unit itself by suitable logic. As will be noted, a jumper register 74 is arranged so that an ECLIPSE.RTM.ENABLE and/or a microNOVA.RTM.ENABLE is provided. The presence or absence thereof is controlled by the presence or absence of a suitable "jumper" connection being reflected by the state of the "jumper" register 72. The following chart summarizes the conditions for identifying the devices involved:

In the first condition where neither jumper connection is enabled the only devices identifiable are "system" devices. Where the ECLIPSE.RTM. jumper is enabled but the microNOVA.RTM. jumper is not, the system generates the ECLIPSE.RTM. timing signals (signified as "E") but the interface unit 14 does not determine whether the devices are ECLIPSE.RTM. or system devices. Where the microNOVA.RTM. jumper is enabled but the ECLIPSE.RTM. jumper is not, both microNOVA.RTM. and system devices are identifiable. Where both jumpers are enabled, microNOVA.RTM. devices can be identified but the system interface unit does not determine whether the devices are ECLIPSE.RTM. or system devices (the ECLIPSE.RTM. timing signals are generated). Accordingly, the polling technique permits the system I/O interface unit 14 to determine whether a particular I/O device must be communicated with either on the microNOVA.RTM. I/O bus or on one of the other buses. If a microNOVA.RTM. device is present on the microNOVA.RTM. bus, its presence is indicated by the microNOVA.RTM. poll store unit 73 which provides a microNOVA.RTM. sensor signal, .mu.SNSR, to modify the operation of a master control finite state machine 75 so that its next state of operation provides the necessary control signals to control the operation of the microNOVA.RTM. finite state machine 73 for microNOVA.RTM. operation.

Control signals for controlling the operation of the poll counter and poll store register are obtained from microNOVA.RTM. FSM and control logic 72 and are formed from suitable combinations of "action line" bits therefrom (in this case as action line bits .mu.AL7-.mu.AL1.0. and .mu.AL12--see Chart D below) not shown in FIG. 8A but discussed more fully with reference to the specific logic diagrams of the microNOVA.RTM. FSM and control logic 72 shown and discussed below, with reference to FIGS. 10, 10A and 10B and with reference to the state flow diagram and action line designations and encodings discussed in connection with FIGS. 17 and 17A and Charts C and D. Thus, the low byte shift regiser 69 is loaded with the device code at the assertion of a selected action line bit by the microNOVA.RTM. FSM 72. The poll count signal POLCN (see FIG. 11) received from the master control FSM 75 (see FIG. 9A) increments the poll counter to place the counter in condition to begin the next poll. The high byte shift register is loaded with a suitable instruction which can be selected to produce the least effect on the system (e.g., an IOSKP instruction in which data in all registers remain unchanged, etc.) at the assertion of other selected action line bits. If a return IOCLK is received at pin 71 in response to a poll sequence, a .mu.EXST signal (FIG. 8A) is asserted by suitable control logic (see FIG. 16) which signal, together with the device code (identified as the .mu.N POLL ADDRESS in FIG. 8A) applied to the address latch of the poll store 73, provides an indication of the presence of the particular device involved. When the complete poll count is completed at start-up, for example, an SPOLLDNE signal is asserted by poll counter 68 to indicate that system initiation is complete (FIG. 11).

Specific logic for the microNOVA.RTM. poll counter 68 is depicted in FIG. 11, while specific logic for the microNOVA.RTM. address latch and poll store register 73 is shown in FIG. 12, the latter providing the .mu.SNSR signal when a device is indicated as present on the microNOVA.RTM. bus 15B. Specific logic for providing the .mu.EXST signal when a IOCLK signal is received by the microNOVA.RTM. FSM as shown in FIG. 16.

The PLWR (FIG. 12) signal is used on "interrupts" by a device when the CPU acknowledges the interrupt request by such device by asserting a suitable interrupt acknowledge signal. At the assertion of a suitable interrupt acknowledge signal the device itself updates the poll store, via the suplying of its device code and the supplying of the PLWR signal by the master control FSM to the poll store logic.

For a poll store read operation, when an ADREN signal occurs, the poll store is accessed by using the device code as an address if a microNOVA.RTM. I/O instruction is present. The poll store then provides a .mu.SNSR signal to the master control FSM if such device is present (see FIG. 12 and the master control "action line" encoding discussed below with reference to the master control FSM state flow diagram of FIGS. 16 and 16A and the action line charts and encodings of Charts A and B).

A poll update request signal is supplied to microNOVA.RTM. FSM 72 at a 100 Hz rate as determined by a suitable 100 Hz input counter signal.

The required microNOVA.RTM. instruction register and decode logic 76 and the microNOVA.RTM. low byte and hi-byte registers 69 and 77 for supplying the high byte data at microNOVA.RTM. bus pin 78 (IOD1) are shown in FIG. 8A, as well as D-register 79 for shifting data to and from the U bus 65. The microNOVA.RTM. shift registers 69 and 77 are in accordance with standard microNOVA.RTM. operation and are of the type which are already well known to the art.

More specific logic for the master control FSM and control logic unit 75 and the microNOVA.RTM. FSM and control logic unit 72 are shown in FIGS. 9, 9A, and FIGS. 10, 10A, and 10B, respectively. As mentioned above, the poll counter 68 and associated circuitry is shown in FIG. 11, while the poll store register 73 and associated circuitry is shown in FIG. 12.

The master control FSM and control logic 75 shown in FIGS. 9 and 9A provides a plurality of action line bits AL0-AL18 which control the operation of the system so as to provide the desired communication with the microNOVA.RTM. bus 15B, the ECLIPSE.RTM. bus 15A, and with local devices internal to the interface unit 14.

Operation of the master control FSM can be better understood with the help of the state flow diagram depicted in FIGS. 17 and 17A, the state actions being set forth in the state flow Chart A (below) and the state flow encodings set forth in Chart B (below).

The master control FSM 75 state flow diagram of FIGS. 17 and 17A (at the top thereof) shows the various mutually exclusive machine states into which the FSM can be placed in order to perform its control of the operation of interface unit 14. As can be noted, once the master control FSM has completed a selected state flow for a particular operating machine state, it is always returned to its "idle" state to await the next instruction which will place it into its next operating machine state.

The designations used to depict the various operating machine states required for the state flows of FIGS. 17, 17A are explained by the master control FSM state flow descriptions (i.e. the state actions) shown in Chart A and the master control FSM word encodings set forth in Chart B, the latter encodings explaining the meanings of selected combinations of the action line bits AL.0.-AL18 in FIG. 9 which produce such actions. Chart B includes specific exemplary binary encodings for the actions listed therein as well as the signal designations corresponding to the signal descriptions in FIGS. 8 and 8A, with reference to the interface unit block diagram and in FIGS. 9 and 9A with reference to the specific logic diagrams of the master control FSM 75. The latter logic diagrams, together with the state flow diagram of FIGS. 17 and 17A, the state actions of Chart A, and the enclodings of Chart B, provide a complete understanding of the structure and operation of the master control FSM 75 for the particular embodiment described herein.

In a similar manner the specific logic used in the microNOVA.RTM. FSM and control logic 72 is depicted in FIGS. 10, 10A and 10B. The state flow diagram for the microNOVA.RTM. FSM is depicted in FIGS. 18 and 18A. A fuller explanation of the states depicted in FIGS. 18 and 18A is set forth in Appendix F which provides a listing of all events and actions associated with the states set forth in the state flow diagram of FIGS. 18 and 18A. The state actions associated with specific combinations of microNOVA.RTM. FSM "action line" bits (.mu.AL.0.-15) are identified in Chart C, while the specific encodings and action explanations are shown in Chart D. As can be seen in the latter chart, appropriate combinations of action line bits .mu.AL7-1.0. control the loading of the hi and low byte registers 77 and 69 and the loading of the poll counter by a command signal, as described above. Charts C and D are set forth below.

The control signals generated by the microNOVA.RTM. FSM permit appropriate communication to occur between the CPU and microNOVA.RTM. devices which are present on microNOVA.RTM. bus 15B (the presence of such devices being indicated as mentioned above by the microNOVA.RTM. poll store 73), the interface unit 14 effectively translating the system bus control signals into suitable microNOVA.RTM. bus control signals so that the required bus protocol (i.e., timing and control) required for microNOVA.RTM. bus operations can occur.

The master control FSM also provides the required TP and CO control signals, as shown in FIG. 9A, which operate the ECLIPSE.RTM. I/O interface logic unit 14A which interface unit then transfers data and control signals from the system bus to the ECLIPSE.RTM. I/O bus 15A. Since in the embodiment disclosed herein the ECLIPSE.RTM. I/O bus handles parallel words, as does the system bus 11 (in contrast with the microNOVA.RTM. bus which handles serial data) the interface logic unit 14A merely includes appropriate registers and decode logic for converting the time multiplexed address and data supplied by system bus 11 to the proper ECLIPSE.RTM. I/O form for use by bus 15A. Specific logic for such purpose is shown in FIG. 19.

In addition to permitting the system bus to communicate with devices on external buses 15A and 15B, the CPU may wish to utilize certain internal (local) devices on the interface unit 14 itself. One such exemplary device, as mentioned above, is the programmable interval timer (PIT) which is shown more specifically in FIG. 20. As can be seen in the exemplary embodiment thereof, a plurality of decade counters 100, 101, 102, and 103 are supplied with a clock signal having one of four different values (e.g., 1 MHZ, 100 KHZ, 10 KHz and 1 KHz) which are supplied from a multiplexer 104. The JR7, JR8 signals for controlling MUX 104 represent the PIT RATE SELECT signal shown in FIG. 8 at the jumper register 74. The decade counters supply the selected one of the four frequencies, the desired input (2's complement) count for the programmable timer being loaded into the counters (U.0.-U15). When the counters overflow, a carry output determines that the PIT count is completed and an interrupt is generated. Output lines (D.0.-D15) are provided from the PIT register shown in FIG. 8A via tri-state drivers 105, 106.

The interface unit 14 may include other "local" devices (not shown), as mentioned above, for use by the CPU, e.g., a real time clock, interface registers for use with teletype machines, etc., and suitable logic and control can be provided for such use in a similar manner.

The instruction register and decoder unit 76 is shown in more detail in FIG. 21 wherein the input instruction bits U5-U15 from the U bus are latched to provide the instruction register bits IR5-IR15 while bits U0-U4 are used in selected combinations to provide further instruction signals. The latched signals are supplied to programmable logic array unit 115 to produce the various decoded instruction signals relating to local device operation, i.e., the LOCAL COMMANDS depicted in FIG. 8A. ROM's 116 and 117 use the latched signals to produce the branch control state signals shown on the state flow design of FIGS. 17 and 17A, which state signals are used at ROM 118 to produce CPU 1-4, i.e., CPU COMMANDS to the master control FSM depicted in FIG. 8A. The IR7 bit is merely a "directional" bit used at many places throughout the interface logic wherever direction must be specified. The RAM R/W signals of FIG. 8A correspond to the HWRIT and HREAD signals shown in FIG. 21 and relate to the writing into or the reading from RAM unit 61 of FIG. 8A. The PIT bit from FPLA 115 in combination with the DOA or DIA bits from ROM 116, for example, are used to form the DOAPIT or DIAPIT signals when respectively loading or reading the PIT counters shown in the PIT logic of FIG. 20.

The system bus control logic 75A shown in FIG. 8 as providing the required signals for connecting between the system bus 11 and the master control FSM 75 is depicted in more detail in the specific logic diagram thereof in FIG. 22.

In connection with the timing required for the above operation an exemplary timing diagram is depicted in FIG. 23 for an ECLIPSE.RTM. programmed I/O operation, for example, (which operation is explained more fully in the aforesaid ECLIPSE.RTM. documents). Such diagram depicts the progression of selected states into which for the master control FSM 75 is placed in order to provide the required state flow for such operation. The particular state flow required is shown at the left in FIG. 17 as the EXT machine state, the flow representing the states into which the control FSM must be placed in order to perform the exemplary programmed I/O operation.

The above explanation and the information in FIGS. 8, 8A, 9, 9A, 10, 10A, 10B, 11, 12 and 16-23 provide a description of the structure and operation of interface unit 14 which permits communication between system bus 11 and either ECLIPSE.RTM. bus 15A or microNOVA.RTM. bus 15B, all of such buses having different operating and timing characteristics and which further permits communication between system bus 11 and local devices internally located on interface unit 14.

The system I/O interface unit 14 is designed so as to be capable of operation at more than one input system clock frequency and is, therefore, adaptable for use with a system wherein the system clock is not expected to remain at a single fixed frequency for all operations. The frequency synthesizer unit 51 is provided so as to be responsive to one of a plurality of different selected input frequencies to produce the desired internal clock frequency required for operation of the internal devices of the interface unit and the desired master baud rate signal for operation of the asynchronous terminal system. In a particular embodiment, for example, the internal clock signal must have a substantially constant frequency of 1 MHz so as to provide for correct operation of the internal, or local, devices (such as the PIT and RTC devices mentioned above) and a substantially constant selected baud rate derived from the substantially constant master baud rate signal for the asynchronous terminal input/output logic.

A unique design for such frequency synthesis is disclosed in FIG. 13 wherein a more detailed block diagram thereof is depicted for generating the 1 MHz signal, for example. In a particular embodiment thereof, the clock rate of the selected input clock signal to the frequency synthesizer unit 51 may be one of eight different clock signals having clock periods which vary from 300 nanoseconds (nS) to 600 nanoseconds (see chart of FIG. 14). For example, in a particular embodiment eight different signals having clock periods, T, of 300 nS, 325 nS, 350 nS, 375 nS, 400 nS, 450 nS, 500 nS and 600 nS are utilized. The input frequency is identified by a 3-bit input signal (shown as U-bus bits U 9,10,11) which is supplied to clock rate latch and decode ROM 80, which identifies which of the eight clock rates must be latched for supply to the frequency synthesizer. The selected clock rate so latched is supplied to a high-count position and parsing ROM 81, to a denominator select ROM 82, and a high-low integer select ROM 83. The functions thereof can be understood better with the help of an example explained with reference to the chart of FIG. 14.

Such chart depicts an example of the above frequency synthesizer technique for generating a pseudo-constant 1.0 MHz output clock for each of eight different input clock signals. As can be seen therein, except for the 500 nS clock signal, the clock periods for each of the eight frequencies involved are uneven multiples of the clock period for the desired 1.0 MHz signal. For example, an input clock signal having a 350 nS clock period completes 2 6/7 periods in the total 1000 nS clock period of a 1.0 MHz signal. Thus, for every 2 6/7 clock pulses of the 350 nS input clock signal, one clock pulse must be produced for the 1.0 MHz output clock signal, i.e., for any 20 input clock pulses, 7 output clock pulses are produced.

The chart of FIG. 14 depicts the parsing algorithm associated with each of the input clock signals. As can be seen therein, groups of input clock pulses are produced over repeatable or recycle time intervals the number of pulses in each group being selected as one of two numbers. Thus, for an input clock signal having a 350 nS period, seven groups of input pulses are utilized having either 3 pulses (high count) or 2 pulses (low count). The position of the high count groups is depicted by the "high-count position" column in accordance with the lettered column of the algorithm. Thus, the high count (3-pulse) groups are at the A and B positions of the seven group cycle (the low count group being at position C) so that the seven groups are as follows:

It should be noted that the uneven count (i.e., 2 6/7) identifies the algorithm to be used. Thus for the 350 nS input clock, the low/high counts vary between "2" (the whole number) and "3" (the next highest integer). The numerator of the fraction identifies the number of high-count position groups (i.e., "6" high count position groups), and the denominator identifies the total number of position groups ("7" position groups), sometimes referred to as the recycle time period. As another example the 300 nS clock signal requires a count of 3 3/9 and utilizes low/high counts of "3" and "4", a recycle time period (the denominator of the fraction) of "9" (i.e., 9 groups), three of which groups (the numerator of the fraction) are "high count" groups. The 3 high-count groups are placed in positions B and C. The other algorithms can be similarly described. In FIG. 13 the denominator select ROM 82 selects the appropriate denominator, or recycle time period, depending on which input clock RATE has been latched, the selected denominator (i.e., "7" for the 350 nS input clock) being identified by a 4-bit output of ROM 82. The ROM 81 identifies which of the eight parsing algorithms shown in the chart of FIG. 14 is to be used. In the example being discussed the parsing algorithm A.B representing seven groups of 350 nS pulses having 3 3 3 2 3 3 3 pulses in each group is selected.

It should be noted that in each parsing algorithm one of two numbers of pulses is used in each group. Thus, for the 350 nS algorithm either 3 pulses (the high count) or 2 pulses (the low count) are used, in the 300 nS algorithm either 4 pulses (the high count) or 3 pulses (the low count) are used. The parsing ROM 81 effectively determines for each demoninator count from denominator counter 84 which of the pulse counts (high or low) is to be used. Thus, for the 350 nS clock the parsing ROM 81 determines that the high count (3) is selected for the first three and the last three of the seven denominator counts and that the low count (2) is selected for the middle denominator count in the overall recycle counter period.

The high or low integer is then selected by high-low integer select ROM 83 to identify which of the counts on integer counter 85 must count so in order to supply one output pulse therefrom for each of the seven groups (i.e., whether the group contains 3 350 nS pulse counts or 2 350 nS pulse counts) so that for every 7 groups (making up 20 350 nS pulses in total) seven output pulses are supplied. Accordingly, as seen in the timing diagram of FIG. 15, the output of integer counter 85 is a 1 MHz clock signal, i.e., every 7000 nS (20.times.350 nS) the output counter 85 produces 7 pulses which is, of course, equivalent to one pulse every 1000 nS. A similar analysis is shown in FIG. 15 for the 450 nS input clock and can be made for each of the eight input clocks and associated parsing algorithms shown in the chart of FIG. 11.

As can be seen in FIG. 15 for the 350 nS input clock, while the pulse spacing of the output signal is not uniform, the number of pulses over each seven microseconds (i.e., 7000 nanoseconds) recycle time period is the same. Thus, while a timing error may occur during each recycle time period the error reduces to zero at the end thereof. For the 350 nS clock signal the maximum error which occurs during the recycle period is .+-.150 nS, as shown in FIG. 15, and the error at the end of the recycle period is zero. For the 450 nS input signal the maximum error is 300 nS and at the end of the recycle time period of 9000 nS, the error is zero. The maximum error for all of the input clock signals are shown in the chart of FIG. 14. Such errors are of little significance over the time periods required for operation of the internal units which utilize the 1 MHz signal so that, for all practical purposes, the 1 MHz signal has an essentially constant frequency. It should be noted that in no case is the maximum error greater than the time period of the input clock signal involved.

The above frequency synthesis unit provides an extremely adaptable system I/O interface unit 14 which can supply the desired master clock signal for the local or internal devices and the desired master baud rate signal for asynchronous terminal operation even when the system input clock signal has one of a plurality of different selectable frequencies. As mentioned above, the overall frequency synthesizer circuitry includes substantially two versions of the general unit shown in FIG. 13, one for the 1.0 MHz output signal and one for the master baud rate output. A baud rate selection circuit can be utilized to provide one of a plurality of 16 different baud rates from the master baud rate by suitable baud selection circuitry which provides integer multiples of the master baud rate as desired. The circuitry for dividing the master baud rate to obtain each of 16 different baud rates is well known to those in the art. Moreover, the 1.0 MHz clock rate can also be appropriately divided down to produce clock rates which are integer submultiples of the 1.0 MHz master clock rate for various internal devices as required. Lists of the exemplary baud rates which can be derived from a master baud rate of 614,400 Hz and exemplary local counter output signals (decode counter clocks) which can be derived from a master 1.0 MHz clock signal are provided below.

While specific implementations of the two-level micro-modification technique, the system bus protocol techniques, the I/O bus polling techniques, and the frequency synthesizing technique are disclosed above for the particular embodiment of the invention, the invention is not limited thereto. Modifications thereof within the spirit and scope of the invention will occur to those in the art. Hence, the invention is not to be construed as limited to the particular embodiments described except as defined by the appended claims.

Appendix F ##Spc1##

Claims

What is claimed is:

1. A data processing system comprising a central processor unit connected to a system bus means having system operating timing and control characteristics for providing communication with one or more system units connected to said system bus; a first input/output bus means having first operating timing and control characteristics which are different from those of said system bus means; a second input/output bus means having second operating timing and control characteristics which are different from those of said first input/output bus means and from those of said system bus means; interface means interconnecting said system bus means with said first input/output bus means and with said second input/output bus means for providing further communication with one or more first units via said first input/out bus means and with one or more second units via said second input/output bus means, said interface means including logic means responsive to an address of a unit with which it is desired that said system bus means communicate for determining which input/output bus means is required to be used in order to provide communication between the addressed unit and said system bus means and for permitting the direct transfer of data therebetween without intermediate storage thereof; said interface means further includes polling logic means for identifying each of said one or more second units which are present on said second input/output bus means; poll store means responsive to said polling logic means for storing information which indicates the presence of each of said identified second units; and means responsive to the information stored in said poll store means for controlling the operation of said interface means to provide access from said system bus means to an addressed unit via said second input/output bus means when said addressed unit is identified as present on said second input/output bus means and for controlling the operation of said interface means to provide access from said system bus means to an addressed unit via said first input/output bus means when said addressed unit is not identified as present on said second input/output bus means.

2. A data processing system in accordance with claim 1 wherein said interface means includes one or more internal units with which it is desired that said system bus means communicate located at said interface means; and means for providing communication between said system bus means and said one or more internal units.

3. A data processing system in accordance with claim 1 wherein said polling logic means includes shift register means for supplying coded information with respect to all of said second units on said second input/output bus means; means connected to said second input/output bus means and responsive to a first signal generated in response to said coded information by each of said second units which is present on said second input/output bus means for producing a second signal which indicates that said second unit is so present on said second input/output bus means; and further wherein said poll store means is connected to said second signal producing means and is responsive to said second signal and is connected to said shift register means and is further responsive to said coded information for storing information indicating the presence of said second units on said second input/output bus means.

4. A data processing system in accordance with claim 3 wherein said poll store means provides a sensing signal when access is requested to a unit which is identified by said poll store means as present on said second input/output bus means; and said interface means further includes bus control means connected to said poll store means and responsive to said sensing signal for controlling access by said system bus means to said second input/output bus means via said interface means.

Patent Citations (13)

Patent	Date	Inventor
US3742457	1973-06-01	Calle et al.
US3967246	1976-06-01	House
US3999163	1976-12-01	Levy et al.
US4000485	1976-12-01	Barlow et al.
US4025906	1977-05-01	Riikonen
US4038644	1977-07-01	Duke et al.
US4041472	1977-08-01	Shah et al.
US4047157	1977-09-01	Jenkins
US4047162	1977-09-01	Dorsey et al.
US4245307	1981-01-01	Kapeghian et al.
US4261033	1981-04-01	Lemay et al.
US4354225	1982-10-01	Frieder et al.
US4418382	1983-11-01	Larson et al.

Non-Patent Literature (3)

IEEE Trans. on Nuclear Science, vol. NS-26, No. 4, Aug. 79, R. W. Downing, "Fastbus--Details of Addressing and Bus Mastership".
IEEE Trans. on Nuclear Science, vol. NS-26, No. 1, Feb. 1979, R. S. Larsen: "Status of the Inter-Laboratory Development of a High-Speed Standard Data Bus-Fastbus".
IEEE Computer Society International Conference, COMPCOPN 75, 25th-27th, Feb. 1975, A. Baum et al.: "Hardware Considerations in a Microcomputer Multiprocessing System".

Abstract

Description