US 4,379,328 AGrant
Linear Sequencing Microprocessor Facilitating
Issue Date:1983-04-05
•11 Claims
•19 Drawing Sheets
Abstract
A microprocessor system in which a microprocessor with a linear sequencing circuit works with arithmetic logic, program memory, registers, and other support circuitry to provide control lines and bus connections to an external application dependent logic module which has control logic, external registers and external memory and is oriented to handle the specific requirements for data transfers to and from a particular type of peripheral device. Means are provided in said microprocessor for selecting the number of times an instruction word is to be repeated and for the halting of a repeated instruction by internal or external means.
Metadata
Assignee
- Burroughs Corporation
Inventors
- Robert D. Catiller
- Brian K. Forbes
Application Information
Application Number:US 216681&
Filing Date:1980-12-15
Priority Date:1979-06-27
Art Unit:237
Classifications
IPC:
G06F 930G06F 1300G06F 304
Field of Search:
364200 MS File;900 MS File
Patent Drawings (19 sheets)
Description
BACKGROUND
Field of the Invention
This invention relates to a microprocessor system that employs a program PROM for instruction storage, an arithmetic logic unit for processing data, associated registers and circuitry, working in conjunction with an external application dependent control logic module having external memory, registers and control logic for communicating to peripheral devices.
CROSS REFERENCES TO RELATED APPLICATIONS
This application is related to the following patents and patent applications.
A patent application entitled "Microprocessor System Operating with Normal or Slow Memories", inventors Robert Catiller and Brian Forbes, filed June 27, 1979, U.S. Ser. No. 052,821 now abandoned.
A patent application entitled "Microprocessor System With Source Address Selection", inventors Robert Catiller and Brian Forbes, now U.S. Pat. No. 4,290,106.
A patent application entitled "Microprocessor Having Word and Byte Handling", inventors Robert Catiller and Brian Forbes, now U.S. Pat. No. 4,301,505.
A patent application entitled "Digital System for Data Transfer Using Universal Input-Output Microprocessor", inventors Robert Catiller and Brian Forbes, now U.S. Pat. No. 4,293,909.
A patent application entitled "Dual Mode Microprocessor System", inventors Brian Forbes and Robert Catiller, now U.S. Pat. No. 4,287,560.
A patent application entitled "Microprocessor System with Specialized Instruction Format", inventors Brian Forbes and Robert Catiller, now U.S. Pat. No. 4,291,372.
Prior Art:
Microprocessors and microprocessor systems will be seen to generally follow the architectural formulas and interrelationships which were established from general computer technology. However, microprocessors making use of large scale integrated techniques can be made into small packages with the use of integrated circuit chips and will generally follow the pattern of using a central processing unit, memory input-output circuits and various other control and support circuitry.
It is typical that the architecture of microprocessors will generally use a program counter which is used to select some portion of memory to be addressed for retrieving program instruction codes to be executed.
In the usage of systems (for example, such as in U.S. Pat. No. 4,174,352) where a central or main host processor operates in conjunction with a number of remotely located "base modules" which base modules support a group of peripheral-controllers which service and control external peripheral units, a considerable amount of circuitry has been required to provide data transfer capability on an efficient basis and also to cater to the specific needs of each type of peripheral unit. Heretofore each peripheral-controller which was situated in a base module was burdened with many processor and control circuits for handling the specific requirements of each peripheral terminal unit leading to a heavy burden of circuitry in each peripheral-controller. The microprocessor described herein alleviates much complexity and cost of earlier peripheral controllers by the provision of a "universal-type" microprocessor which can cooperate with an application-dependent logic module to establish a faster, more efficient peripheral-controller for handling data transfers between a variety of peripherals and a main host computer.
SUMMARY OF THE INVENTION
To alleviate the burden of costly circuitry for each peripheral controller, a universal or architecturally standard microprocessor was developed which would provide a multiplicity of processing functions for each peripheral-controller so that the peripheral-controller would only need a limited amount of application dependent logic and circuitry tailored to the particular peripheral which it serviced. Thus, the universal processor, or as it is called here the "universal I/O state machine", could serve as a standard unit for all types of peripheral-controllers so that each of the peripheral-controllers would only need to have the minimum amount of circuitry required for application to a particular peripheral.
Thus, with the use of the universal I/O state machine as a general purpose microprocessor in these applications, it became possible to reduce the cost, the quantity of circuitry, the space requirements and the system of architectural arrangements so that a more economical and more efficient architectural system could be established. Thus, with the advent of the described universal I/O state machine, there has been devised a generalized type of microprocessor which can serve with any type of peripheral-controller in a base module thus to handle any variety of peripheral terminal units.
The universal microprocessor working in conjunction with the peripheral oriented application logic permits the development of a variety of peripheral-controllers whose advantage is in respect to faster data transfers and the minimal amount of circuitry needed to adapt to the pecularities of any given peripheral device. Since the hardware and programming of the microprocessor remains standard and constant for each of the peripheral-controllers, it is thus possible to accomplish not only great versatility in data transfer and control features but also to provide great economies in circuitry and cost.
In digital systems where many instructions and control signals are required for the transfer of data to a variety of different types of peripheral units, generally a specifically tailored and designed peripheral-controller is provided which is particularly suited to handle the data transfers to and from a particular type of peripheral unit. It was found that by providing a uniform microprocessor for all types of peripheral-controllers which only varied in the type of application dependent logic required for specific types of peripheral devices, that great economies of space and cost could be realized. Thus, the combination of a universal microprocessor working in conjunction with an application dependent module could serve the function of a peripheral-controller in an optimum fashion.
The microprocessor of the present system provides a linear sequencing circuit which sequences an instruction set in a predetermined fashion so as to eliminate any requirement for a next address generator. The microprocessor includes: a processor or arithmetical logic unit, a program memory, memory address means which includes a program counter and a memory reference register for addressing external memory and program memory in said application dependent logic module. An instruction register receives instructions from said program memory or external memory on an instruction bus which is conveyed to a memory operand register to said arithmetic logic unit. A plurality of accumulator registers connects to an input-output bus and temporarily stores data for input to said arithmetic logic unit. A stack memory saves data from the program counter during interrupt actions while a memory-save register saves data from said memory reference register during interrupt operation. An instruction decoder-controller generates and receives control signals to and from said application dependent logic module, in addition to providing control signals to the universal microprocessor.
The instruction decoder-controller is provided with a linear sequencing circuit which sequences an internal instruction set in a predetermined fashion so as to eliminate the requirement for a next-address generator. The microprocessor includes a repetition counter which, when loaded with a number N, will set a repeat-mode flip-flop in the decoder-controller to cause iterative operation of the same instruction until the repeat mode flip-flop is reset.
BRIEF DESCRIPTION OF THE DRAWINGS
FIG. 1 is an overall block diagram illustrating the use of a main host system computer with a plurality of base modules which support peripheral-controllers which communicate with peripheral devices.
FIG. 2 is a diagram of the elements of the universal input-output microprocessor state machine and its communication lines to the application dependent logic module. FIG. 2 is subdivided into two portions designated as FIG. 2A and FIG. 2B such that by juxtaposing FIG. 2A to the left of FIG. 2B (and FIG. 2B to the right of FIG. 2A) the continuity of connections between the two drawings may be seen. FIG. 2C is a schematic drawing of the decoder-controller with its linear sequencer circuitry; FIG. 2C-1 shows the state counts; FIG. 2D illustrates the addressing of selected accumulator registers and their relation to the arithmetic logic unit and I/O bus; FIG. 2E shows how source and destination addresses are accessed in one clock period; FIG. 2F is a schematic drawing illustrating the relationship between the repetition counter and repeat-mode flip-flop.
FIG. 3 is a schematic diagram showing certain memory addressing features of the UIO-State Machine, particularly the use of a dedicated bit designated as the "byte-swap" bit.
FIG. 4A is a perspective drawing of the input-output base module and showing major elements such as the distribution control card, the peripheral controller (line control processor) card and certain connections from the front plane (fore plane) to the host computer and to peripheral units. FIG. 4B is a schematic drawing showing circuit connections to the four front plane connectors. FIG. 4B is oriented such that it shows the first front plane connector while FIG. 4B-1 shows the second front plane connector. Orientation-wise the FIGS. 4B-1 can be juxtaposed underneath the FIG. 4B in order to provide continuity to the connections shown between the two front plane connectors. FIG. 4B-2 illustrates connections to the front plane connector No. 3 and FIG. 4B-3 illustrates the connections to the front plane connector No. 4.
FIG. 5 is a timing diagram showing the clocking of output control registers of the UIO-state machine.
FIG. 6 is a circuit drawing showing how the program memory is enabled.
FIG. 7 is a timing diagram for the clear line.
FIG. 8 is a timing diagram showing the WAIT line timing.
FIG. 9 is a timing diagram showing timing for non-repeated GET/PUT functions.
FIG. 10 shows the timing diagrams for the repeated GET/PUT functions.
FIG. 11 is a diagram showing the correlated timing features for reading data from external memory on the memory data bus. FIG. 12 is a timing diagram showing the timing correlations between the clock, the memory address signal and the memory write data on the I/O bus.
FIG. 13 is a timing diagram showing the relationship of the halt repetition function signal to the clock.
FIG. 14 is a timing diagram showing the relationship of the parity error signal to the clock and memory address signal.
DESCRIPTION OF THE PREFERRED EMBODIMENT
Referring to FIG. 1, there is seen a system block diagram of a typical Burroughs host computer system, with central processing unit cabinet 100 having basically a series of main memory modules 101, a memory control module 102, a data processor module 103, working in cooperation with a host dependent port module 104. The host dependent port module provides a series of communication cables 105 designated as MLI or message level interfaces. These message level interfaces provide communication lines to a base module cabinet 106 (or a plurality of such cabinets) which are designated as the Universal Input/Output Base Module Cabinet. The base module cabinet is built to support a plurality of universal input/output (UIO) base modules 107. Each of these UIO base modules provides communication, control, and data transfer lines to specific types of peripheral devices.
It should be noted that the host dependent port module can have message level interfaces to a plurality of UIO base module cabinets. And likewise, each UIO base module cabinet may have message level interface communication lines to other host systems 200 (main memory and processor) in addition to the first described host system of FIG. 1.
In the preferred embodiment of the system to be described herein, it will be seen that a base module cabinet can support from one to four UIO base modules. Each of the UIO base modules may contain up to eight peripheral-controllers 108 known as "line control processors" and each of the line control processors will be dedicated to data transfer and control with a particular type of peripheral device 109.
Another provision of the system will allow for a unit designated as a line expansion module (LEM) which can be inserted in the UIO base module so that a single message level interface can be extended to communicate with up to eight UIO base modules.
The host system as represented by the central processing unit cabinet 100 can communicate with up to 64 peripheral controllers (line control processors) for each message level interface (MLI), if the base module is provided with the unit known as an LEM or line expansion module.
As seen in FIG. 1, the host system 100 can have eight message level interfaces (MLI) and can thus communicate with up to 512 peripheral controllers (line control processors).
The peripheral controllers or line control processors (LCP) are controllers which are dedicated to data transfer and control between a peripheral terminal and the main host system or a plurality of host systems.
The peripheral controllers known as line control processors have been described, both in their system interrelationships and in their internal workings, in a number of previously filed patent applications and issued patents. These include the following United States Patents: U.S. Pat. No. 4,174,352 entitled "Modular Block Unit for I/O Subsystem", inventors Darwen J. Cook and Donald A. Millers, II; U.S. Pat. No. 4,106,092 entitled "Interface System Providing Interfaces to Central Processing Units and Modular Processor-Controllers for an Input-Output Subsystem", inventor Donald A. Millers, II; U.S. Pat. No. 4,162,520 entitled "Intelligent Input-Output Interface Control Unit for Input-Output Subsystem", inventors Darwen J. Cook and Donald A. Millers, II; and U.S. Pat. No. 4,189,769, entitled "I/O Subsystem for Digital Data Processing System", inventors Darwen J. Cook and Donald A. Millers, II, and these patents are included herein by reference.
The peripheral controllers known as line control processors (LCP's ) may generally fall into types of categories. There is the 2-card LCP and the 3-card LCP.
The 2-Card LCP:
This peripheral controller has a first card which is designated as the CFE or common front-end on which there are supplied a plurality of read-only memories (ROM's); hardware-wise these common front-end cards are of the same construction and nature except that in individual cases the ROMs are made to hold different programs which are suitable to the application at hand or the peripheral device with which it is used; thus, hardware-wise all the front-end cards will be found to be the same physically but only differ in the fact that different program material is inserted in the individual ROMs. The second card in the 2-card LCP is the PDC or peripheral dependent card. This card is uniquely built and adapted to suit the particular type of peripheral unit to which it communicates. It may be designated as an "application dependent logic module".
The 3-Card LCP:
Here, again, the first card is the CFE or common front-end card which is physically the same except for different programming inside the ROM, this program being varied according to the application required. The second card is the CDC or common data card which is tailored to suit the requirements of different peripheral units; however, it may be the same for such a group of items as magnetic tape, punched card-tape, disk pack, and console controls. The third card of the 3-card LCP is designated as the PDC or peripheral dependent card. This card is unique and is designed to handle one particular type of peripheral device.
In FIG. 4A there is seen a more detailed perspective drawing of a typical I/O base module 107. A plurality of these base modules are housed and serviced by a base module cabinet 126 of FIG. 1.
The base module 107 provides housing whereby printed circuit cards holding integrated circuitry may be inserted on sliding tracks in order to connect with the backplane connectors for automatic connection into the system. A peripheral controller (or line control processor) 108 may require, as previously described, two printed circuit or three printed circuit cards. The base module may support up to eight such peripheral controllers in its housing. Communications between the base module and the host system are controlled by distribution card 110. Other specialized handling printed circuit cards may be inserted, such as the previously mentioned line expansion module, common front-end card, common data card; or a second distribution control card, etc., can be inserted for connection into said base module 107.
The frontplane or foreplane permits jumper cable connections 111 (FIG. 4A) between printed circuit cards and also permits connection cables such as message level interface 105 to the host computer or ribbon cable 109c to permit connection from the base module to external devices such as peripheral device 109.
An interface panel 106 may be provided on the base module cabinet 120 to provide for cable connectors such as 104.sub.cc, 109.sub.cc to facilitate disassembly and reconnection.
As was described in the previously mentioned patent references regarding line control processors, each base module unit is provided with a Distribution Control Card 110, FIG. 4A, which connects the message level interface from the main host system to a selectively addressed line control processor (peripheral controller) in that base module. Further, the base module has a common maintenance card and a common termination card which acts in common to service the entire group of up to eight line control processors in a base module.
The UIO base module may also support a line expansion module or LEM. This feature expands the usage of a UIO base module by adding another Distribution Control Card which permits connection of that base module to yet another main host system in addition to the one already connected to the first Distribution Control Card. Thus, communication from a main host system can be expanded to cover two or more UIO base modules and their attendant line control processors. Likewise, two or more host systems can be made to communicate to a UIO base module or a plurality of such base modules.
When it is desired to have one UIO base module connected to two main host computers, then a control card is used which is called the BCC or base control card. The base control card serves two basic functions; (a) to provide an interlock when multiple host computers try to access a base module, so that line control processors can thus be locked out from use by unauthorized host computers, and (b) to provide assorted features such as a base identification (ID) number and for maintenance control. The BCC is not mandatory when multiple hosts are connected to a base; however, it is useful. The base control card, however, can also be used when only a single host computer is connected to the UIO base module. Thus, the BCC is an independent and optional module, as is the line expansion module LEM.
Thus, in reference to FIG. 1, it should be stated that when a host computer is connected to a UIO base module, the base module must have at least one distribution card 110, FIG. 4A, for each host computer which is connected to it. Now, when a UIO base module is connected to a multiple number of host computers, then there is need for the base module to be provided with a unit known as the PSM or path selection module. The path selection module is the control unit which regulates the communication to and from the multiple number of distribution cards in the UIO base module.
In the development of line control processors, such as indicated in the prior cited patents and references, it was found that the amount of logic required for adaptability to any given type of peripheral unit or to a variety of peripheral units required that expensive and space consuming logic circuitry be developed which would take large quantities of hardware and space requirements on printed circuit cards such that sometimes four or five extra logic cards were needed in order to complete a particular line control processor. Soon the cost and space requirements of these developments became prohibitive and better means were required for handling the variety of functions required of a line control processor.
In this situation a microprocessor-controller was developed which could handle all the basic logic for any variety of peripheral terminal units and which could be built with a minimum amount of components and placed on a single printed circuit card. This device was designated as the universal input/output state machine or UIO-SM. Thus, the UIO state machine could be built typically on a 112 chip circuit board and be placed within the slide-in card rack which is supported by the LCP base module (now called UIO base module). The UIO state machine can interface easily to any application dependent logic through its front plane and, except for power connections and clock connections, no connections needed to be made to the LCP or UIO base backplane. The universal I/O state machine is basically implemented in transistor type logic (TTL).
General Overview of the Universal I/O State Machine
The UIO-SM has been made to implement a useful group of general purpose operators which include:
Arithmetic OP's
Logical OP's
Read/Write Memory OP's
PUT/GET OP's
Program Stack OP's (Branch, Call and Return)
At FIG. 2 there is seen an overall block diagram of the UIO state machne. The input/output lines at the bottom of the drawing are all front plane lines which connect to the peripheral's application dependent control (PDC) logic since it is desired that the UIO state machine itself be independent of any application type logic or of the special requirements of particular peripheral units.
Referring to the state machine block diagram of FIG. 2, there will be seen a series of interconnecting lines 16, 17.sub.1, 17.sub.2, 18, 10, 19, 11, 12 and 20, which connect to the application dependent logic 60.sub.L. These lines are labeled, such as the external memory data output bus 12, the I/O and memory write parity line 18, the memory address bus 16, the memory write enable 19 (and other functions listed on FIG. 2), the direct memory access (DMA) request line 11 (and other functions on FIG. 2), in addition to the main I/O bus 10. Line 15 is the instruction output bus from the program PROM 50.
A program counter 41, having input lines from stack memory 45 and I/O bus 10, via 10.sub.f, provides an output line which feeds an input signal to the program PROM 50 (internal memory 50). A memory reference register 40 (MRR) receives its input signals via I/O bus 10 from accumulator registers 30, memory operand 31, and the save-MRR 47 (memory reference register-save). Repetition counter 42 also receives an input line via I/O bus 10 for signals from accumulator registers 30 and memory operand 31.
The stack memory 45, controlled by stack pointer 46, provides an output line to the program counter 41. The repetition counter 42 provides an output which feeds back around line 10.sub.f over to the I/O bus 10 and also around to the accumulator registers 30.
Attached and receiving input from the I/O bus 10 is a control register 1 designated as item 37, a control register 2 designated as item 38 and a parity register 39. The control registers 37, 38 provide buffering of signals to application control register lines 17.sub.1, 17.sub.2. The parity generator 39 has an output line 18 (part of I/O bus 10) and provides "I/O and memory write" parity.
The memory output data bus 12 from RAM external memory 90 connects to the memory operand register 31 whose outputs feed to ALU 32 and parity check circuitry 21 which has an output line 20 to provide memory read parity to external memory 90. Data bus 12 also provides an input to the instruction register 22 which has an output line connecting to the input of the instruction decoder-controller 23. Situated within instruction decoder-controller 23 are external flag registers 23.sub.e, interrupt mask 23.sub.i, and state counter 23.sub.s.
Accumulator registers 30 actually comprise two sets of eight registers, one set designated 30.sub.f (foreground accumulator registers) and the other set as 30.sub.b (background). Each of the 16 registers are addressed and selected by an address line 23.sub.a from the instruction decoder-controller 23.
The I/O bus 10 provides a channel to the accumulator registers 30 which have an output to the B input of the arithmetic logic unit 32. Another input A to the arithmetic logic unit 32 comes from the memory operand register 31 which gets its input from external memory via the memory output bus 12, or from internal memory via bus 15.
The arithmetic logic unit 32 (ALU) has one set of outputs which feed to the "foreground" flag register 35 and another set of outputs to the "background" flag register 36. These flag registers 35 and 36 form an output which is fed to the instruction decoder-controller 23. They are used to signal "foreground mode" conditions for normal operation and "background mode" conditions for interrupts or emergency operation.
The F output of the arithmetic logic unit 32 feeds on a bus 10.sub.i to the shift logic circuitry 33 and to the byte-swap circuitry 34. The outputs of these circuits 33, 34 connect to the I/O bus 10.
Summary: General Operation
The PUT OP (operator) writes a 16-bit word (two bytes symbolized as AB) from the I/O bus 10 into a selected application dependent register 60 (external register), the particular register being addressed via GET-PUT address lines 19. The PUT OP can address any one of 32 application dependent registers. The GET OP reads a 16-bit word from a selected application dependent register 60 into an accumulator register 30 on the UIO-SM or into RAM memory 90 through the I/O bus 10. The GET OP can also address a selected one of the 32 application dependent registers 60 (external registers).
The UIO-SM has the ability to repeat certain OP's (PUT's, GET's and logical OP's) by loading a number N (from the software) into a repetition counter 42 to control the memory reference register (MRR) 40 which is used as a counter after having been loaded with the starting address of the data block to be used in the "repeated" OP. A more detailed discussion of this operation is presented hereinafter in connection with FIG. 2F.
In order to hold "PUT Data" for a longer period of time than can occur directly off of I/O bus 10, there are two 8-bit registers (control registers 37, 38) which have been placed in the UIO-SM. The strobing of these registers 37, 38 is under the control of the application dependent logic 60.sub.L (FIG. 2).
A "WAIT line" (on bus 11) has been built into the UIO-SM such that when an external "slow memory" is addressed, the "slow memory" can force the UIO-SM to wait however long required for the Read or the Write to be valid. This is seen on FIG. 2 at bus 11 which feeds signals to the instruction decoder-controller 23. This line can also be used to halt the machine. A clock enable line on bus 11 (CLKENB) has been included so that the application dependent logic 60.sub.L may control the state machine clock for single pulse operations and also direct memory access (DMA) operations. The timing for this signal is the same as the WAIT signal on bus 11.
As seen in the state machine diagram of FIG. 2, the UIO-SM can check via circuitry 21 for odd parity on each memory fetch. All memory fetches from external memory 90 come on the memory output data bus which is shown as bus 12 and which connects to instruction register 22 and to the memory operand register 31. All memory writes go out on the I/O bus 10 along with odd parity on line 18. The UIO-SM contains 8-K words of user-defined instruction programs in PROM 50 which is capable of being expanded to 64-K words.
With reference to FIG. 3, it will be seen that memory addresses can come from two sources--the program counter 41 and the memory reference register (MRR) 40. The UIO-SM provides the ability to do condition and unconditional Branches, Calls and Returns. "Calls" can be nested up to 16 levels from stack memory 45. The memory reference register 40 stores addresses and is used (at certain times specified in instruction words) to address RAM data memory 90 which will provide programs applicable to the peripheral device connecting to the application dependent logic module 60.sub.L. This external memory 90 also serves to store data which is being transferred between the host computer and the peripheral terminal device. Since memory 90 and 50 are organized two bytes wide (16-bits plus parity) but the data for transfer is sometimes required in single byte form, as for peripherals such as card readers, the memory reference register (MRR) 40 has been designed such that only the high order 15-bits are used to address memory 90. The low order bit (BYTESWP) is used by the byte-oriented read/write OP's to determine whether or not to byte-swap the two byte data word before storage into accumulator registers 30 on an "external data memory read" or before storage into external RAM data memory 90 on a "data memory write". This particular feature permits the UIO-SM to easily handle byte-oriented data in that the desired byte is automatically read into the "low order" byte position of the accumulator registers 30 on "data memory reads", as well as being automatically written from the "low order" byte position into the correct byte position on "data memory writes". The "high order" byte position of the accumulator registers 30 must contain the last byte written, if any is present.
In Table I there is shown the instruction format for the UIO-state machine operators. The external memory output data bus 12, or internal memory instruction bus 15, provides 17 bits designated 0-16. The bus bits 0-7 (low order) constitute the "B" byte while the bus bits 8-15 constitute the "A" byte (high order).
As will be seen from Table I certain fields are specified such as the P field for odd parity, the basic function field F1, F0, and function variant field by V4-V0. The destination specifier field is designated D3-D0; the source accumulator field is designated S2, S1, S0. The field which causes selection of the memory address source to be used in fetching the operand of the instruction is designated M; while the function field for copying the arithmetic logic unit output into the memory reference register 40 is designated by the field C.
The odd parity bit is set such that an odd number of 1's will exist across the entire 17-bit instruction (or the operand). The basic function field F1, F0 is described in Table II such that various binary digit combinations of F1, F0 provide certain functions as shown.
For example, the GET function provides for the transfer of the contents of an external register 60 to the accumulator 30 or to RAM memory 90. The digital bits 0, 1 provide for the transfer of the contents of an addressed register in accumulator 30 (or an addressed portion of external RAM memory 90) to the external registers 60. This is called the PUT function. The F field bits 1, 0 provide for execution of the arithmetic logic unit operators; while bits 1, 1 provide for the functions of the program stack (45) operators such as Branch, Call, Return.
Table III shows the function variant field V4-V0 and its relation to the function field F1, F0.
For example, the F-bits 0, 0 specify a GET operation from an external register address while the F-bits 0, 1 specify a PUT operation into an external register of a specified address.
Table IV shows the destination specifier field, D3, D2, D1, D0.
This D3-D0 field is subdivided into (i) GET and ALU OP's and (ii) PUT and program stack OP's. In the case of (i) there may be no destination specified; or only a miscellaneous control field may be used to update the flag register; or the destination may be the RAM memory 90.
In the case (ii) for the destination field, the operand source may come from RAM memory 90; or the operand source may be from the source accumulator 30; or the destination may be a load into the repetition counter 42.
The Source Accumulator Field S2, S1, S0 is shown in Table V below:
The source accumulator field specified as S2, S1, S0 (Table V) may apply (i) to the ALU or the PUT Op where the operand from the source accumulator A(S) can be the same as the destination accumulator A(D); or (ii) where the GET Op can function so that the source is an external application register 60 or that the source is the repetition counter 42; or wherein (iii) the program stack Op's can provide a return operator or an ICall operator.
M Field Specifier: Selecting Memory Address Source for the Operand of an Instruction:
The M field specifier (as originally noted in Table I) may occur as one of two binary states, that is as a 0 or as a 1. When M is 0, then the operand for this instruction is located one address beyond this instruction (PC+1, program counter plus 1) or there is no operand. If the instruction does not require a memory operand, then PC+1 is not used for an operand; the next "instruction" will be at PC+1 (no space is allowed for an operand). The operators that do not use an operand are listed as follows:
1. All "GET" operators;
2. "PUT" operators with D3=1;
3. "CALL" or "BRANCH" with D3=1;
4. ICALL, RETURN, IRETURN, ROTR, SHFL, SHFR, CMA, ONES, ZERO, SWAC, WRTW, WRTB, DRTL, EI, DI and RSTK Operators. (These operators are discussed subsequently hereinafter).
When the M field bit is "1", then the operand for this instruction is addressed by the memory reference register 40 (MRR) or else there is no memory operand for this instruction.
C Field: For Copying ALU Output to the Memory Reference Register 40:
The C field originally shown in Table I may read as a 1 bit or a 0 bit. In the 1 bit situation this means that the machine will not copy the ALU output to the memory reference register 40; however, if the bit is "0", then the ALU output will be copied on to the memory reference register 40. The C bit can be set to 0 on all instructions except GET's and RDB, RDBW, RBWR, SWAC and WRTB. These operators are discussed subsequently hereinafter.
Flag Registers:
Two flag registers are provided so that at least one flag register is fully available for each of the two operating modes. Thus, flag register 35 is the "foreground mode" flag used under normal operating conditions while register 36 is the "background mode" flag which is used during interrupt or other unusual conditions. As seen in FIG. 2, flag registers 35, 36 receive output from the arithmetic logic unit 32 and provide an output line to feed input to the instruction decoder-controller 23. The flag register is updated at the end of all ALU operators (which include arithmetic OP's, logical OP's and load/store/move OP's). On OP's which "byte swap" the ALU 32 output before storage into the destination location, the flag register is updated on the basis of the full two bytes of information from the ALU 32 after being swapped. In the specific cases of REBW OP (read before write byte) and the RBWR OP (read before write byte reversed), a full memory word is read into the UIO state machine, but only one of the two bytes read is written into accumulator register 30 (the low order byte of the accumulator destination register remains unchanged). The flag registers (35,36) are updated on the basis of the full word read from RAM memory 90.
As indicated in FIG. 2, the accumulator registers 30 actually constitute two different sets of eight registers. One set 30.sub.f of accumulator registers is for normal operating conditions (foreground mode) while the other set 30.sub.b of eight accumulator registers is for abnormal conditions (background mode) as during interrupts.
Instruction Repetition:
The use of the repetition counter 42 provides an instruction repetition feature where a PUT or a GET OP may be repeated up to 256 times by the use of a PUT OP to load the repetition counter 42 which must be followed by the PUT or the GET OP to be repeated. Any value N from 0 to 255 (from the software) may be loaded into the repetition counter 42 and N+1 operations will be performed (1 to 256). The address source for the repeated PUT data or the address destination for the repeated GET data is found in MRR 40 and thus these addresses will originate from the memory reference register 40 and not from the program counter 41 (PC+1). In other words, the M-bit is set to "1" in the repeated PUT or GET. After each PUT or get OPs, the hardware will increment the MRR 40 in preparation for the next PUT or GET. When repeated, each individual PUT or GET will execute in one clock period, unless the PUT or the GET addresses a "slow memory", in which case there will be one extra clock period at the beginning of the repeated PUT's or at the end of the repeated GET's.
The use of the repeated PUT's or GET's serves to greatly increase code compaction while also greatly increasing the speed of the clock transfers to or from RAM memory 90.
Any logical OP (AND, CMA, CMM, OR,XOR, ONES, ZERO, ADDC, SUBC, etc. as described hereinafter) may be repeated up to 256 times by the use of a PUT OP to load the repetition counter 42 which must be followed by the logical OP to be repeated. Any value of N from 0 to 255 may be loaded into the repetition counter 42 and N+1 operations will be performed (1 to 256). For the repeated OP to work as expected, the source accumulator 30 should be equal to the destination accumulator 30. The address source for the operand data should originate from the memory reference register 40 (MRR) just as for the repeated PUT or GET. After each OP of the repeat sequence is executed, the hardware will increment MRR 40 in preparation for the next OP. When repeated, each individual logical OP will execute in one clock period, unless the operands are located in a "slow memory".
One particular case, where a repeated logical OP can increase code compaction and execution speed, is the use of the repeated "Exclusive OR" for the purpose of generating longitudinal parity on a data block located in external memory 90.
The PUT OP which loads the repetition counter 42 must have the N value in the "B" byte (low order 8-bits) of the PUT data. The "A" byte (high order 8-bits) is a "don't care".
The application dependent logic 60.sub.L has the capability of stopping instruction repetition even though the full number of operations has not occurred, as will be later discussed in connection with FIG. 2F.
The capability exists to do repeat operations with PC (program counter 41) used for the memory operand source address (M-bit equal to 0). In this case, PC 41 will increment during the repeat operation instead of MRR 40 (provided that the operator does use a memory operand). Using PC 41 during the repeated operations will allow the use of a multiple word literal following the single repeated operator. The size of the "literal" field must be exactly the number of words specified by the repeat count value plus 1 word. The HLT REP/interface line (halt repetition) must remain "high" during all repeat operations using PC 41 as an address source for operand data.
In regard to the instruction execution speed, the bacic clock rate fo the UIO-SM is 8 MHZ (125 nanosecond clock period). All OP's , with the exception of the program stack operators, execute in two clock periods except such OP's that write from the UIO-SM accumulators 30 into RAM memory 90 (3 clocks required) or have operands to be fetched from or stored into "slow memory" (three or more clocks required). The program stack operators require 3 clock periods to execute if the condition is true (Branch or Call "executed") and only require 2 clock periods if the condition is false (Branch or Call "not executed"). The NO-OP executes in 2 clock periods. The OP IReturn executes in 3 clock periods, while Load and Store Accumulator executes in 4 clock periods.
The execution of any given instruction requires an "instruction fetch" along with the above-noted execution cycles; however, in the case of the UIO-SM, the instruction fetch for the next instruction is performed during the last execute cycle of the current instruction. Therefore, the "instruction fetch" is masked from the timing.
The NO-OP:
The NO-OP consists of any program stack operator (such as Branch) with the condition test field set to "UNCONDITIONAL" and the destination specifier set to "EXECUTE ON CONDITION TEST NOT TRUE" (Branch never). Whether or not the NO-OP occupies one or occupies two memory locations, is a function of whether or not the operand (Branch never address) is located next to the instruction (PC+1) or is addressed by the memory reference register 40 or comes from an accumulator 30. The flag registers 35, 36 will not be updated on the NO-OP.
Clear Sequence:
The clear sequence is a "zeroing" operation. When a "Clear" is issued to the UIO-SM, the program counter 41, the repetition counter 42, and the stack pointer 46 are reset to zero and an instruction is fetched. After a "Power-up Clear" has been issued, the accumulators 30 cannot be predicted. However, if a "Clear" is issued while power remains up, the accumulator registers 30 will retain their states just previous to the "Clear". While "Clear" is true, at least two clocks must be issued; after "Clear" has gone false, one clock is required before execution of the first instruction begins.
Output Control Registers:
There are two output control registers 37 and 38 in the UIO-SM as seen in FIG. 2. These can be used by the application dependent logic 60.sub.L to hold "PUT Data" for a longer period of time than can normally occur directly off of the I/O bus 10. Both output control registers 37 and 38 are 8-bits wide and have their data inputs connected to the low order 8 bits of the I/O bus 10 ("B" byte). Both of these output control registers must be clocked by the application dependent logic 60.sub.L. FIG. 5 illustrates a preferred method for clocking the output control registers 37 and 38. A triple input NAND gate 74 in FIG. 5 has three input line signals: PUT STROBE/signal; address signals for an external register (60 of FIG. 2); and the clock signal. These signals are correlated in FIG. 5.
Once clocked by a leading edge, the outputs of these registers will stabilize in 17 nanoseconds or less. A Schottky TTL NAND device 74 is used for the control register clock generator since the "PUT Data" can be guaranteed for only 10 nanoseconds past the leading edge of CLOCK as received on the UIO-SM, and the output control registers require a data hold time of 2 nanoseconds minimum past the leading edge of their clock.
Memory Addressing and Expansion:
In the preferred embodiment and without expansion of memory, the UIO-SM has the capability of directly addressing 0 to 64K words of program memory 50, FIG. 2, and from 0 to 32K words of data memory in RAM 90. To facilitate any further expansion of memory, such as by using paging techniques or otherwise, the enabling of the program PROM 50 in the UIO-SM can be controlled by an expanded memory in the application dependent logic 60.sub.L.
FIG. 6 illustrates how such an expanded memory can control the enabling of the UIO-SM program PROM 50 through the use of a front-plane external signal. If the expanded memory drives this signal to a "high" level, the UIO-SM program PROM 50 will be disabled. If the application dependent logic 60.sub.L does not have a requirement for expanded memory, then it will simply tie this line to a logic "low". In FIG. 6, a 5-input NOR gate 75 has inputs from program counter 41 in order to provide a parity disable signal (high) to disable PROM 50.
The UIO-SM program PROM 50 may be expanded from 8K up to 64K words by using different size PROMs. The chip select and the addressing functions of the PROM 50 are automatically adjusted without circuit changes. A jumper arrangement is used to enable the parity bit output for the size of the program memory being used. With reference to FIG. 6, jumpers can be installed according to the following Table VI for each word size of PROM shown below.
The UIO-SM receives its clock signal from the backplane of the UIO base module. To receive clock signals, the UIO-SM preferably uses Schottky type flip-flops followed by Schottky type inverters in order to generate the internal UIO-SM clock (CLOCK). All the timings are relative to this internal clock which preferably has a minimum delay of four nanoseconds and a maximum delay of 13 nanoseconds from the clock on the backplane. In order to insure minimum clock skew between the application logic 60.sub.L and the UIO-SM, the application logic 60.sub.L is preferably made to use the same type of devices to receive and distribute the clock.
The following paragraphs will illustrate detailed timing and explanation for the interface between the UIO-SM and the application dependent logic 60.sub.L.
Referring to FIG. 7, which is entitled "Clear Line Timing", it will be seen that in order to initialize the UIO-SM, the CLEAR line must be "true" (active low) for at least two full clock periods.
Another feature implemented into the UIO-SM is the WAIT line (WAIT/). This is used such that when "slow memory" is addressed, the slow memory can force the UIO-SM to wait however long required for the "read" or the "write" to be valid. This line can also be used to halt the machine for any length of time required. The timing requirements for the WAIT/line are illustrated in FIG. 8. Since this line has multiple sources, including one line on the UIO-SM itself, it must be driven by an open-collector (TTL device).
A CLOCK enable (CLOCKENB) line is implemented on the UIO-SM for use in single pulse or for direct memory access (DMA) functions. The timing here is the same as the WAIT/ of FIG. 8.
GET/PUT Timings:
The timing relationships for the non-repeated GET's and PUT's are illustrated in FIG. 9. The WAIT line can be used on the GET to increase the memory write time in the case of a "slow" write memory. Another application for use of the WAIT line on GET's is to increase the amount of time available for the application dependent logic 60.sub.L to drive the "GET DATA" and meet set-up times after having received the GET enable signal.
In the case of the PUT OP, the effect of the WAIT line is to increase the period of time that "PUT DATA" is on the I/O bus 10, although the UIO-SM would be halted during this time. It should also be noted that the timing of the PUT Strobe can only be guaranteed relative to the CLOCK signal and not relative to the PUT DATA. As such, neither edge of the PUT Strobe is used to directly clock the PUT Data into an application dependent register 60.
The first transfer of data in a "repeated" GET to an addressed register in accumulator 30 will be invalid. All data transfers (after the first) will be valid. The same accumulator register 30 is addressed as a "destination" for each repeat GET operation.
If M=1 (refer to Table I), the memory reference register 40 will be incremented during each repeat GET operation, even though external RAM memory 90 is not used by the GET operator.
The timing relationships for repeated GET's and repeated PUT's are shown in FIG. 10. Specifically this timing chart further illustrates when repeated "GET Data" can change and when repeated "PUT Data" will also change.
The timing relationships for memory "Reads" in the UIO-SM are shown in FIG. 11. The WAIT line has been utilized on one of the memory reads to illustrate how one (or more) extra clock periods can be inserted after address stabilization to allow "slow access" memory to meet the setup time requirements of the UIO-SM.
In regard to the memory "Write" timings for the UIO-SM, these memory Writes are shown in FIG. 12. UIO-SM instructions which can do memory Writes are: WRTW (Write Word), WRTB (Write Byte), and also GET. The WAIT line 11 has been utilized on one of the memory Writes to illustrate how one (or more) extra clock periods can be inserted after address stabilization in order to allow "slow" Write memory an adequate Write time.
Halting "Repeated" Instructions:
Although the total number of operations on a repeated instruction may not be completed, the repeated instruction itself can be halted by the application dependent logic 60.sub.L. The application dependent logic produces the signal HLTREP/ (active low), which is the signal to the UIO-SM that halts the repetition. In FIG. 13 there is shown the timing of this signal. If the application dependent logic 60.sub.L has no requirement for halting repeated instructions, then the application dependent logic can simply leave the HLTREP/ signal unconnected.
When the HLTREP signal goes false, the current processing operation will be completed and a fetch of the next instruction is started at the "high" clock transition. The HLTREP/ line must remain "high" during repeat operations using the program counter (PC) 41 to address the memory operands (M bit=0).
Parity Error Line:
On every memory fetch the UIO-SM checks for odd parity and signals the results to the application dependent logic 60.sub.L. If incorrect parity is detected, the UIO-SM takes no action other than to signal the error to the application dependent logic 60.sub.L. This timing of the parity error line PARERR/ is shown in FIG. 14.
Referring to FIG. 3 and to the memory reference register 40 (MRR) there is a line which is used by external memory mapping logic to control address bit 15 of the memory reference register 40. This line is designated as MRR 15 and the state of this line is loaded into MRR 40 at bit 15 whenever a loading of memory reference register 40 is called for by a UIO-SM instruction. The state of this line should be stable for at least 30 nanoseconds prior to the high transition of the clock to ensure that it can be properly loaded. This line is tied to ground (low side) if memory mapping is not used.
There is a line designated IRQ/ on bus 11 which line is the "interrupt request line". This line is driven "low" preferably by an open-collector device in order to request an interrupt of the UIO-SM. The UIO-SM will only recognize an interrupt request if the interrupt mask is reset (EI instruction executed, Table XII-F, for resetting the interrupt mask bit) and an instruction fetch has just been completed when not in repeat mode (high clock transition at end of T0 and the start of T1).
The UIO-SM responds to an interrupt by forcing execution of an ICALL instruction and switching the UIO-SM "background" mode "on". The ICALL instruction generates an acknowledge signal (ACKI/) in order to allow the application logic 60.sub.L to place an interrupt call address on the I/O bus 10. The ICALL instruction saves the memory reference register 40 data in the MRR save register 47 and it saves the old PC (program counter) value on the stack before loading PC with the call address (placed on the I/O bus 10 by the application logic 60.sub.L). Further interrupts are inhibited in the "background" mode and the duplicate set of flags (35) and accumulators (30.sub.f) are used so that the "foreground" operating environment is preserved therein.
A return to the "foreground" mode is accomplished by executing an IRET instruction. The top value from the stack memory 45 is placed in the program counter PC 41 for the return address and the contents of the MRR save register 47 is placed into the memory reference register MRR 40. The original foreground flag 35 and foreground accumulator 30.sub.f are again made active and contain the same values as before the interrupt. The clear line will place the UIO-SM in the "foreground" mode and set the interrupt mask to disable interrupts. The IRQ/ line must be stable at least 60 nanoseconds before the leading edge (high transition) of the clock signal.
The DMA request line 11 of FIG. 3 is a line which is driven "low" by the application dependent logic 60.sub.L in order to isolate the UIO-SM address outputs on memory address bus line 16 so that direct memory access (DMA) logic can drive this bus. This line is left unconnected when not using the DMA.
The bus enable line (BUSEN/LINE) from Control Logic 66 is driven "low" by the application logic 60.sub.L in order to enable the I/O bus drivers on I/O bus 10, including data, GETENB, PUTSTB and REGADR signals. The GETENB is a signal used to indicate that GET data is driven on to the I/O bus 10; the PUTSTB is the signal used to indicate that PUT data is on the I/O bus 10; the REGADR signal is the signal used to indicate the enabling of the external register address lines for GET's and PUT's. When "high", this line disables all I/O signals. This line is tied to "low" if it is not used.
Universal Input/Output State Machine Front Plane Interface:
The front plane connectors (FIG. 4A, FIGS. 4B-1, 4B-2) which provide the interface for the UIO-SM are preferably four in number wherein each of the four front plane connectors provides 25 pin connections on each of two sides of each front plane connector. The following Tables VII-X show the signal name for each pin number for each side of the four front plane connectors.
Table XI is a glossary of the front plane signal names for the Universal I/O State Machine. This glossary shows the signal name, the number of the front plane connector involved, when the signal is a Fan-in or a Fan-out signal and, in addition, a functional description of the signal involved.
The following Tables XII and XII-A through G show the operators of the UIO state Machine
Operational Factors:
The external memory storage RAM 90 is used to hold programs which have to do with the transferring of data between the external memory 90, the microprocessor unit and the application dependent registers 60 where data can be transferred to selected peripheral terminals or receive from selected peripheral terminals. While the application dependent register 60 can buffer and generally holds several words, it is also necessary that large batches of character transfers, such as a 1,000 character batch, will be temporarily stored or buffered in the external RAM memory 90. This is necessary since data transfers to and from the peripheral devices are very slow in comparison with the fast speed of operation of the microprocessor and the host computer.
It will be noted that the main host system 100 is connected by means of a message level interface bus 105 which, through a host adapter, not shown, connects to the application dependent registers 60. Further, as seen in FIG. 1 individual peripherals can be shown as items 109.sub.a and 109.sub.b which are connected to the external registers. Each of these peripherals also has an adapter unit 109.sub.c which controls the peripheral terminal unit and can also be instructed by the microprocessor system through instruction words passed through application dependent registers 60.
As seen in FIGS. 2A and 2B the general control and coordination is handled by the instruction decoder-controller 23 which has connections to all of the units in the system; however, the main interconnections have been shown on the drawings which indicate control lines to the major units involved.
Since the basic function of the overall system is to control and effectuate the transfer of data between selected peripherals and the main host system, it is helpful to consult the data provided in the previously cited patents which were incorporated by reference. The system operation is initiated when the main host system 100 of FIG. 1 conveys an I/O descriptor command word on the message level interface lines 105, this command being placed into a selected application dependent register 60. As mentioned in the previously cited patent references, the main host system also provides a "descriptor link" word signal which idenfities a task to be performed so that when that task is either completed or incompleted, the peripheral-controller can formulate a Result Descriptor word which can be conveyed to the main host system to inform it of the condition status as either completion or non-completion of the task to be performed and also to identify the task so that the result word data will be channeled to the proper memory location in the main host system where data concerning a given task has been located.
The microprocessor system is placed into operation by means of the "clear" signal shown in FIG. 1 at line 11, at which time it starts operations so that the first address of the program counter is used to address the program memory 50 in order to initiate the fetching and execution of instruction words.
The receipt of the I/O command descriptor word into the application dependent registers will cause the Interrupt line signal (on lines of bus 11) to be sent to instruction decoder-controller 23 whereby the microprocessor will fetch, analyze and execute operations as a result of the I/O command descriptor.
With the I/O descriptor in the application dependent register and with the Interrupt signal activated, the microprocessor will use a GET operator to get this command data from the application dependent register 60 and put it into a selected accumulator register 30 where it can then be read, analyzed and interpreted by the arithmetic logic unit (ALU 32) and its associated data manipulation circuitry so that the required operators and data will be sent again to the application dependent register 60, or the external memory 90 and such that the control logic 66 will permit the transfer of data to occur.
Assuming a Write operation (transfer from host computer system to peripheral), the I/O command descriptor from the main host system will have the address of a selected peripheral in its instructions. The address of this peripheral will also be taken on into a GET operation by the microprocessor and translated into an instruction which can then be put (PUT operation) into the peripheral (via an adapter) by using control logic 66 and selected application dependent registers 60 in order to select the desired peripheral device 109. As seen in FIG. 1 each of the peripheral devices will have an adapter 109.sub.c which is associated with the peripheral device in order to control its operation and data transfers. Then the microprocessor can send appropriate instruction commands to the adapter unit of the selected peripheral such that the input line to the peripheral can be enabled and instructions can be provided by the microprocessor, for example, to take incoming data from the main host system 100 (which may be temporarily stored in the application dependent register 60 or temporarily stored in the external memory 90) and then transfer data to the selected peripheral terminal device.
For example, if the selected peripheral terminal device is a "card reader", then this peripheral unit is oriented to handle only one byte at a time so that it is then necessary for the microprocessor instructions to select only one byte at a time and to transfer the data one byte at a time from its storage in external memory into the peripheral terminal unit. On the other hand, if data, for example, is being transferred to a disk-type peripheral, which generally handles a two byte word (16 bits) then the data can be transferred in terms of "words" (symbolically designated AB) directly to the peripheral adapter and thence the peripheral disk.
Thus, it is seen that in certain situations it is necessary to select only one byte, at a time, out of a two byte word in order that one byte, at a time, can be transferred out into the peripheral which can only accept one byte at a time. In this case, if there is an accessed word AB composed of two bytes, it is then necessary for the microprocessor to select out one byte at a time for transfer to the peripheral. This is done generally by using the "low order" byte as the byte to be selected for transfer. Thus, in the word AB, the byte B would be selected for transfer to the peripheral, after which the word-byte positions AB would be byte-swapped into the word positions BA, and then the "low order" byte A could then be transferred into the peripheral terminal unit.
The I/O command descriptor may eventually wind up in a selected one of the accumulator registers 30 or in a selected external register 60 or in external memory 90. In any case it must be available from the host computer 100 so that the microprocessor can access it, analyze it and make the appropriate signals as to what further instructions must be generated.
The program counter 41 holds addresses which are used to select instruction data from the program memory 50. This program counter is reset to zero with the initiation of the "power on" signal and the clear signal.
With the initiation of the fetching of instruction words from the program memory and with the receipt of an I/O command descriptor, the command descriptor data will be transferred to an internal register of the microprocessor system, such as the accumulator register 30 where it can then be inspected and analyzed by the data manipulation circuitry (ALU 32) and the associated shift logic and byte swap circuitry 33, 34. Thus, the meaning and intent of the descriptor command can be translated into instructions for the selection of a peripheral device, through its adapter, and for a subsequent transfer of data to or from the particular peripheral terminal device.
For example, the PUT operator is generated by the microprocessor and is applied to the application dependent logic module 60.sub.L. An address (of a peripheral adapter and a peripheral terminal unit) will then be provided and the application dependent logic module will operate on this address to select the appropriate adapter and peripheral unit for data to be transferred. The software of the microprocessor system generates the PUT commands to the peripheral adapter.
When the microprocessor system generates a PUT operator to the application dependent logic module 60.sub.L, then this enables the system to operate such that the host computer will send data on message level interface 105 to selected application dependent registers 60 where it can then be transferred into and stored in the external memory 90. As the microprocessor system provides a multiple series of repeated GET operators, then each word of data (or each byte of data if the peripheral can only accept bytes) will be transferred from the external register 60 to the selected peripheral terminal.
Since the microprocessor operates much faster than the peripheral terminal, then what would generally be done is that if there is a multiple word transfer of data, is that the words would be stored in the external memory 90 and then fetched and transferred out to a selected peripheral at the slower speed with which the peripheral terminal device can accept the data. In this respect the external memory 90 acts as a storage buffer which permits the microprocessor to operate at its high speed while data can reside in the external memory 90 and then be transferred at a slower speed to the selected peripheral.
Data will be sent from external memory 90 to the peripheral terminal by the use of a PUT operator which is used to control the peripheral adapter and terminal unit. That is to say, the PUT operator will command a word to be transferred from external memory 90 into a selected external register 60 and thence the peripheral. The PUT operator always deals with full "words" (AB). For byte-oriented usage, the microprocessor generates intervening instructions to isolate a "byte" from a memory word into a selected accumulator from which a PUT operation is then executed.
The use of the memory reference register 40 is prominent here in that this register provides a source of addresses for the selection of operands and for the selection of programs which reside in the external memory 90.
In the situation of a Read operation where it is desired that a selected peripheral terminal unit will "send" data to the main memory of the host system, again the I/O descriptor command will be placed by the host computer into the application dependent register 60 which will then interrupt the microprocessor and ask it to "GET" the I/O command and to analyze and execute it. Since this is a Read operation from a selected peripheral device, the microprocessor will then generate the addresses and the transfer commands so that a particularly selected peripheral and its control adapter will be selected and the data from that particularly selected peripheral will be transferred to application dependent register 60 and thence stored temporarily in external memory 90. Then a series of GET operators will cause the transfer of the words stored in external memory 90 to the external register 60 and thence through the message level interface 105 over to the host system 100. Transfers to the host system will be done on a full word (two byte) basis, AB, in repetitive cycles since the host system can accept the complete 16 bit word AB.
In summary, a multiple number of GET operators will cause the transfer of data from the peripheral terminal unit into the external memory 90 which acts as a buffer. Then a multiple number of PUT operators will be used to transfer this data from external memory 90 to the application register 60 and thence through the message level interface to the main host system 100. The adapters 109.sub.c of FIG. 1 which are supplied for each of the peripheral devices are useful in accepting commands and for controlling operations of the particular peripheral. For example, at certain times it might be necessary to rewind the tape mechanism of a tape peripheral terminal unit or to eject or slew the paper on a printing device type terminal unit, or to select a certain address sector in the disk terminal unit.
In regard to the "C" bit previously discussed at page 28 under the C field, this feature permits certain economies in the number of instructions.
Because there is only one memory reference register (MRR), loops of code which deal with multiple address ranges, such as a loop to move `n` words from address `s` to address `d`, involve alternately loading MRR with source (`s`) and destination (`d`) addresses. Such loops also involve incrementing the `s` and `d` addresses for each iteration. Without the copy (C) bit, such a loop looks like:
Using the copy bit it is possible to combine the incrementing of the address accumulator with the setting of MRR in one step. To do so, adopting the trick of starting with `s-1` and `d-1` in the accumulators, this is done only once:
Thus, it is possible to save ((8-6)*n)-2 instructions by using move `n` words, or 2n-2 instructions. For 100 words, this is 198; at current speeds that equals about 50 microseconds saved for 100 words moved.
Byte Oriented Read/Write Operators:
As was previously discussed, the UIO-SM has been implemented with several operators to simplify the processing of byte-oriented data even though the memory is organized in words which are two bytes wide. These particular operators work in conjunction with the memory reference register 40 of FIG. 2 which addresses external RAM data memory 90.
The byte-oriented read/write operators use the "byte-swap" feature of the UIO/SM such that the "desired" byte is always read from external memory (90) into the "low order" byte of a selected register of the destination accumulator A(D) even though the desired byte might have been in the "high" order byte of the memory location in 90. The desired byte, after being placed into the destination accumulator A(D), can then be processed. After the low order byte is processed, the high order byte of the accumulator A(D) must be updated with the other byte from the current data memory location, and then the two bytes are written into memory 90 as one word, and each byte will go into the proper position. This would be needed, for example, when a card reader peripheral is supplying only one byte of a word for each transfer operation.
In the case of byte-oriented operators, the decision whether or not to "byte-swap" is determined by the low order bit of the memory reference register 40 which, in essence, is a byte-address. The following Table XIII is an example which illustrates the use of these operators in handling sequential byte data which is stored in memory 90.
The next following Table XIV shows a sequence of instructions and the resultant signals in the memory reference register 40, the accumulator register 30 and the data memory 90. Thus, the sequence shown in Table XIV starts with loading the memory reference register 40; reading the byte (RDB); translation; read before write byte (RDBW); write byte (WRTB); incrementing the memory reference register (MRR 40); again read byte (RDB); again translate; again read before write byte (RDBW); write byte (WRTB); incrementing memory reference register 40; read byte (RDB); translate; read before write byte (RDBW); write byte (WRTB); and so on.
A further "byte-oriented" operator has been implemented into the UIO-SM and is illustrated in the following Table XV. This operator aids in the formation of a "two-byte" data word where the two bytes are sequentially located in data memory 90. But unless several OP's are used, it is not known whether or not the two desired bytes are in a single memory location or whether they straddle two memory locations.
The upper portion of Table XV shows the arrangement of data in a first Memory 1 and a second Memory 2 corresponding to true and false signals of the data memory address. Then the lower part of Table XV shows an exemplary sequence for the sequential instructions with reference to data in the memory reference register 40 and in the accumulator 30.
Linear Micro-Sequencing
FIG. 2C is a schematic block drawing of the Instruction Decoder 23 with its linear micro-sequencing circuit. A State Counter 23.sub.s is initiated into a Fetch operation upon receipt of a Clear signal to its input. The State Counter 23.sub.s feeds address bits to the PROM 23.sub.p which contains a set of internal micro-instructions. The PROM 23.sub.p provides one output line designated Count Enable which is used to sequence the State Counter 23.sub.s.
One set of instruction signal outputs, selected from PROM 23.sub.p, are a set of control terms which are conveyed to a pipeline register 23.sub.r which provides control signals to the units of the microprocessor and the application dependent logic 60.sub.L. The pipeline register 23.sub.r is used (a) to synchronize the output control term signals and also (b) to delay (by one clock) an output control term so that the signal from PROM 23.sub.p will be adequately stabilized. Another set of control signal terms are provided directly from the output lines of PROM 23.sub.p.
As seen by the dotted lines above PROM 23.sub.p, an alternate plug-compatible PROM 23.sub.x can replace 23.sub.p and be utilized by up to two extra address lines from State Counter 23.sub.s. Each line (1 bit) will permit the addressing of an alternate or added set of instructions from PROm 23.sub.x (in addition to using the original 2 bit lines).
In FIG. 2C the Instruction Register 22 provides input instruction to PROM 23.sub.p. An input of 15 bits (instruction fields) is fed to Instruction Register 22 from either Program Memory 50 or external memory 90 depending on which memory has been selected as a source of data.
One register of Instruction Register 22 is dedicated to receipt of the one bit of data called the "M" bit which is placed in each instruction word (from Program Memory 50). As discussed earlier, this field (M) is the one which selects either the Program Counter 41 or the Memory Reference Register 40 as the source of addresses to be next used for the selection of data.
In FIG. 2C the Condition Select Multiplexer 23.sub.m has inputs from internal flag registers 35, 36, and from external flags 23.sub.e in the Application Dependent Logic 60.sub.L. The output of Condition Select Multiplexer 23.sub.m is used to provide an input to PROM 23.sub.p of the linear micro-sequencer. Operationally, it should be noted that, at the same time that an instruction is being decoded, there is simultaneously a Fetch of the operand, since the M bit, when set, will initiate selection of an address from Memory Reference Register 40 (or when M not set, selects Program Counter) to access data from a location in external memory 90 (or Program Memory 50).
FIG. 2C-1 is a schematic illustration indicating modes of the State Counter 23.sub.s. At state number 0 (upon activation of the Clear signal) a Fetch operation is initiated. Thence the "1" state is achieved where the instructions may start execution of an instruction.
After completion of the "1" state, the sequencer may return to "0" for a Fetch and proceed to the "2" state. Here an operation step can occur.
After completion of this cycle, the State Counter returns to "0" for another Fetch operation, or else sequence to state "3" which permits an additional operation step.
After this, the State Counter goes to "0" (or "4") for Fetch.
The remaining portions of FIG. 2C-1 showing the states "4", "5", "638 and "7" refer to the situation in FIG. 2C where an alternate larger instruction PROM 23.sub.x has been used to replace 23.sub.p with the extra two address lines of 1 bit each for addressing the micro-instruction PROM. Here, an alternate instruction set can be accessed and sequenced, similar to the sequencing of the PROM 23.sub.p using the states 0 through 3. The State Counter 23.sub.s can thus be used to address and select instructions from either PROM 23.sub.p or from a larger PROM 23.sub.x.
Referring again to FIG. 2C, the Instruction Register 22 is seen to provide a 5 bit address line to select an external register 60. There is further provided a 3 bit "source address" line and a 3 bit "destination address" line to Multiplexer 30.sub.x (which is shown again in FIG. 2D). Referring to FIG. 2D, the accumulator registers 30 are seen clocked by a 8 megahertz clock, and also have a Write Enable (WE) input line coming from decoder-controller 23.
Multiplexer 30.sub.x (FIG. 2D) receives input addresses (from instruction register 22) for the source of data and for the destination of data. The SELECT line to MUX 30.sub.x enables the source address to be read on the first half of the clock as seen in FIG. 2E at "Read", and enables the destination address to be written on the second half of the clock at "Write". Thus, the source location of data to be processed, and the destination location, where the data is to be placed, are all accomplished in one clock (125 nanoseconds).
Referring to FIG. 2E, a latch 30.sub.t is provided to the accumulator registers 30 to hold and convey data to ALU 32. The other data input to ALU 32 is from Memory Operand Register 31 which receives data and instructions along instruction bus 15 (from Program Memory 50), and/or memory data bus 12 (from external memory).
The I/O bus 10 provides a path for processed data to be fed back to accumulator registers 30 or to the external memory 90 or to the external registers 60 (of application dependent logic unit 60.sub.L).
Referring to FIG. 2F, there is seen a schematic drawing showing the intercooperation of the Repetition Counter 42 and Repeat Mode Flip-Flop 23.sub.f and decoder-controller 23.
When a LOAD/PUT OP occurs which loads a number N into the Repetition Counter 42, this operation also sets the Repeat Mode Flip-Flop 23.sub.f which then continues to signal the decoder-controller 23 to continue iterative operations of the next instruction OP. On each iteration, the Repetition Counter is decremented. When the count therein reads "0", then a reset signal is conveyed to the Repeat Mode Flip-Flop 23.sub.f to turn it off which will signal the decoder-controller 23 that the iteration is finished (or to be halted), at which time the microprocessor will then select the next instruction from Program Memory 50 (via program counter 41).
In FIG. 2F, there will be seen another line for resetting (shutting off) the Repeat Mode Flip-Flop 23.sub.f. This is done by using a HALT/REP signal from Control Logic 66 in the Application Dependent Logic Unit 60.sub.L. This could occur because Control Logic 66 has sensed an early ending in the data transfer operation.
The linear micro-sequencer enables functions to be partitioned among the various PROMs so that each PROM will handle certain functions. There is no need for a stack memory or a next address register in the micro-sequencing circuitry. The micro-sequencer lowers the need for multiple IC's and permits expansion of PROM usage with no increase in hardware.
The microprocessor system permits simplicity and time-economy with which instructions can be fetched and executed, and data fetched and manipulated for data transfer purposes can be accomplished by what may be called (1) instruction pipelining and (2) operand pipelining.
Certain overlapping operations are made to occur which provide for an exceptionally short time for instruction and operand fetch and for execution of instructions.
Thus, while the processor is fetching an instruction in one clock, on the next clock, it is storing the result data in an I/O (peripheral) device by means of the I/O bus 10, or it may be storing the result data in the external memory 90, for example. At the same time that the result data is being stored, another fetch cycle occurs which is a fetch cycle using a memory address on the address bus 16 which will then provide an output of memory data on the memory data bus 12.
Note that the external memory read path 12 is set separate from the memory write path 10 and the system can store the result data in a peripheral I/O device while at the same time reading instructions from memory.
There are basically two different lines or separate paths, one for instructions an one for operands. For example, the instruction bus 15 feeds instructions to the instruction register 22. The memory operand register 31 also can receive operands from the memory data bus 12. Thus, on one clock the instruction bus can feed instructions to the instruction register (and then to instruction decoder-controller) while on the next clock the operand can be fed to the memory operand register 31 to facilitate the execution of the instructions.
Since the address of the I/O device (peripheral) is contained in the instruction word, the system can immediately store data in the I/O device. One level of the instruction pipelining includes (a) storing result data into the internal register files (as the accumulators 30 or the external registers 60) and (b) storage of results into an I/O port adapter which connects to peripheral devices.
The operand pipeline uses the M bit. Thus, in one clock while the processor is storing result data by using the I/O bus and using an address that has already been fetched, then at the same time an instruction fetch cycle can be occurring from program memory 15 on the next clock a decoding operation takes place simultaneously with the fetching of the operand after which on the next clock the storage of result data takes place in an addressed location.
The M bit is used as part of the operand pipeline in order to determine which source of addressing will be used--that is whether the memory reference register 40 will be used (data pointer) or whether the program counter 41 will be used (instruction pointer). This was previously discussed in the specification under the heading of Byte Oriented Read/Write operators. It may be noted that the arithmetic logic unit 32 has the one input from the memory operand register 31 while the other input is from the accumulator registers 30.
While a preferred embodiment of the disclosed invention has been described and explained, it is, of course, understood that modifications and changes may be made without departing from the concept of the invention as defined by the following claims.
Claims
What is claimed is:
1. A universal-type microprocessor system which cooperates with an application-dependent logic module to form a peripheral-controller capable of handling data transfers between the main memory of a host computer and a plurality of peripheral terminal units which can be either "word-oriented" or "byte-oriented", and wherein said application-dependent logic module provides external register means having bus connections to said plurality of peripheral terminal units and to said host computer, and wherein said application-dependent logic module further includes an external memory for storage of programs related to control of said peripheral terminal units and for temporary storage of data undergoing transfer, and wherein said application-dependent logic module further includes control logic means for communicating with and controlling said external register means, said external memory, and communicating with a decoder-controller in said microprocessor system, the said microprocessor system comprising: (a) data processing means including: (a1) an arithmetic logic unit providing an output to a shift logic circuit and to a byte-swap circuit; (a2) said shift logic circuit providing output to an I/O bus, said shift logic circuit functioning to rotate one or more bits of a word being processed to the right or to the left; (a3) said byte-swap circuit providing output to said I/O bus, said byte-swap circuit functioning to exchange the sequential positions of the higher order byte and the lower order byte of a received two-byte word; (b) said I/O bus providing connecting lines from said data processing means to said external registers, to said external memory, to an accumulator register means, and to an addressing means; (c) said addressing means receiving input data from said I/O bus and storing addresses useful for accessing data from internal program or external memory storage, said addressing means including: (c1) a program counter storing consecutive addresses of data located in said internal program memory; (c2) a memory reference register for storing addresses of data located in said external memory, and including: (c2-1) a memory address bus which connects to said external memory; (c2-2) and wherein a dedicated non-address bit in said memory reference register is set to signal said decoder-controller to initiate a byte-swap operation in said byte-swap circuit, or when not-set, to pass data without a byte-swap operation; (c3) an address register means in said decoder-controller connecting by means of said I/O bus to said accumulator register means which include a plurality of accumulator registers for selection of data from an addressed accumulator register; (d) said internal program memory storage for storing program instructions and data words, said internal memory being addressed from said program counter and providing instruction words and data, via an instruction bus, to an instruction register and to a memory operand register; (e) register means for temporary storage of data, said register means including: (e1) said plurality of addressable accumulator registers providing an output to the input of said arithmetic logic unit, and said accumulator registers receiving input data words from said I/O bus, said plurality of accumulator registers including: (e1-1) address bus input means from said decoder-controller for addressing selected registers in said plurality of registers; (e2) said memory operand register for receiving operand data from said internal program memory via said instruction bus or from said external memory via a memory data bus and providing an output to the input of said arithmetic logic unit; (e3) said instruction register for receiving instruction words from said internal program memory via said instruction bus and providing instruction words to said decoder-controller; (f) said instruction decoder-controller receiving instruction signals from said instruction register and including incoming and outgoing control signal lines connected to said data processing means, to said addressing means, to said register means, to said internal and external memory means, to said external registers and said control logic means; (f1) and wherein said decoder-controller further includes: (f1o) linear sequencing means for selecting microcode instructions for execution in a predetermined sequence, said sequencing means including a microcode instruction memory for storing microcode instructions, said linear sequencing means including: (f1o-1) a state counter connected to said microcode instruction memory for receiving count control signals from said microcode instruction memory and for providing a first portion of address signals to said microcode instruction memory; (f1o-2) input control lines from said instruction register to provide a second portion of address signals to said microcode instruction memory; (f1o-3) and wherein said microcode instruction memory provides said microcode instructions for the output lines of said decoder-controller in response to said first and second address portions; (f1a) connection means for receiving control signals from said control logic in said application-dependent logic module for performing clocking, interrupt and halt functions; (f1b) address lines to said external registers for selecting a specific register; (f1c) control lines to said external registers for strobing address signals; (g) said memory data bus for carrying data from said external memory to said instruction register and to said memory operand register; (h) a repetition counter having an input from said I/O bus and an output to said memory reference register; (i) instruction word means in said internal program memory for signalling said decoder-controller, said instruction word means including: (i1) a GET operator which functions to place the contents of said repetition counter into a selected accumulator register as its destination; (i2) a PUT operator which functions to take the contents of said destination accumulator and place it via said data processing means into a selected external register; (j) means for controlling the number of repetitive operations of a selected instruction word means, said means for controlling including flip-flop means, in said decoder-controller, which receives information signals from said repetition counter.
2. The system of claim 1, wherein said means for controlling the number of repetitive operations includes: (j1) means to load, via said I/O bus, a number N into said repetition counter for establishing a series of N+1 repeat operations of the same instruction, said repetition counter acting to decrement the number N for each repeat operation of said instruction; (j2) a repeat mode flip-flop, being set by said decoder-controller, for causing repetition of the current instruction until the repetition counter is equal to zero.
3. The system of claim 2, wherein said repeat-mode flip-flop includes: (k) means for setting said flip-flop upon receipt of a PUT operation which loads said number N into said repetition counter.
4. The system of claim 3, wherein said repeat-mode flip-flop may be reset and shut off by a signal from said repetition counter or by a signal from said external control logic in said application dependent logic module.
5. The microprocessor system of claim 2, wherein said instruction word means further includes: a dedicated bit field (M) for signalling said instruction decoder-controller to select either said memory reference register or said program counter as the source of memory addresses for the fetching of data or instructions.
6. The microprocessor system of claim 5, wherein a value of M equal to "one" for said dedicated bit field, will select said memory reference register as the source of memory addresses used to fetch data from said external data memory.
7. The microprocessor system of claim 6, wherein said memory reference register provides the source of addresses for the placement of repeated PUT data and further provides the destination addresses for repeated GET data.
8. The microprocessor system of claim 5, wherein a value of M equal to zero for said dedicated bit field, will select said program counter as the source of memory addresses to be used to fetch instructions or data.
9. The microprocessor of claim 2, wherein said internal memory, said external memory, and said register means are organized into locations for holding words of two-bytes, symbolically designated as AB where A is the higher order byte and B is the lower order byte of the word; and wherein said means for loading said repetition counter with a number N includes: means for selecting the lower order byte B of a selected word AB during a PUT operation, said number N being derived from the lower order byte B of the PUT data.
10. The microprocessor system of claim 4, wherein said control logic, in said application dependent logic module includes: a halt-repetition control line, connected, via said repeat-mode flip-flop, to said decoder-controller, for signalling said decoder-controller to stop the repetition sequence of an instruction.
11. The microprocessor system of claim 1, wherein said PUT instruction operator controls the placement of data from a selected accumulator register, acting as source, or from a selected address in said external memory, into a selected external register; and wherein said PUT operator is being used to load a number N from the said internal program memory into said repetition counter in order to control the execution of N+1 operations of a selected instruction to be repeated.