WO1982002783A1

WO1982002783A1 - Stored-program control machine

Info

Publication number: WO1982002783A1
Application number: PCT/US1981/001745
Authority: WO
Inventors: Electric Co Inc Western; Donald Edgar Blahut; Marc Lawrence Harrison; Michael John Killian; Mark Ernest Thierbach
Original assignee: Western Electric Co
Priority date: 1981-02-10
Filing date: 1981-12-23
Publication date: 1982-08-19
Also published as: JPS58500043A; GB2092786B; CA1167169A; KR830009517A; EP0070863A1; JPH0666682B2; IT1149544B; EP0070863A4; GB2092786A; IE820286L; US4399516A; IL64937A0; IE53180B1; KR900005282B1; EP0070863B1; IT8219507A0; IL64937A

Abstract

A hierarchical organization of programmable logic arrays (505-507) permits the control of microprocessor functions to be achieved in a way which allows otherwise wasted clock time to be used. The mostly independent operations of the several PLA's is organized by 'handshake' signals from the latches (550) of one PLA to those of another via AND circuits (552) operative to selectively enable clock signals, in some instances, and data in other instances, to be applied to the latches. The use of the AND circuits enables requisite operations to be achieved with relatively small PLA's.

Description

STORED-PROGRAM CONTROL MACHINE

Field of the Invention

This invention relates to stored-program control machines and, more particularly to semiconductor integrated circuits such as microprocessors for implementing such machines. Background of the Invention

A microprocessor generally is accepted to include various elements of a computer on a single chip of semiconductor material, with the possible exception of memory (Program and Data Store).

The various processing functions of a microprocessor are carried out in an area of the chip which includes registers and an arithmetic logic unit and is referred to as the data path portion of the chip. The cooperation between the various elements of the data path portion as well as the sequences in which those elements cooperate is determined by sequences of outputs generated by a programmable logic array (PLA) and applied in a manner to control the data path portion of the chip. See "Introduction to VLSI Systems" by Carver Mead and Lynn Conway, Addison-Wesley, 1980 for a full explanation of PLA's. A PLA includes a decoder section and a READ ONLY

MEMORY (ROM) section with associated input and output registers respectively. The decoder section is known as an "AND" plane and includes a drive line for each input term and one for each complement term. Each of the drive and complement lines intersect electrically conducting output lines which extend into the ROM section where they become "word" lines. The intersecting lines are called decoder output lines. At selected intersections in the decoder section, pulldown transistors are formed. The transistors respond to various input codes to ground selected decoder output lines. The outputs of the AND plane thus are determined by the locations and gate connections of pulldown transistors connected to the decoder output lines.

The ROM section of a PLA also includes output lines which intersect the word lines. Again, pulldown transistors are formed at selected ones of those intersections with transistor gates connected to the word lines. If the gate of any transistor connected to a decoder output line is at a high voltage (a binary one), that output line is at a low voltage (a binary zero) as is the associated word line in the ROM section. On the other hand, if a decoder output line is high, the associated word line is grounded. By a selective placement of transistors in both the decoder section and the ROM section of a PLA, a particular output code appears on the ROM output lines for each input code applied to the inputs of the decoder section. In this manner, instructions of an input program are decoded into a sequence of cycle-by-cycle actions. A representation of the repertoire of actions is termed a state diagram. The action to be performed during a given cycle is defined at the output of a ROM section of a PLA at an output register clocked during phase U₂ of a two phase clock cycle. An input register to the PLA is operative to store, in a U₁ phase, input that were applied to it during the immediately proceeding U₂ phase.

As the number of operations performed by the elements, of the data path portion of a microprocessor chip increases, so does the requisite number of cycle-by-cycle actions. The number of distinct actions that a PLA can invoke is a function of the number of word lines.

Consequently, the PLA has to increase in size in order to invoke an increased number of actions. Since the available area on a semiconductor chip is limited, the space available for the PLA is also limited. Further, as the PLA increases in size it operates more slowly and thus limits the clock rate of the entire device. The problem thus is to implement a requisite state diagram with a PLA of relatively reduced area.

Brief Description of the Invention

The foregoing problem is solved in several ways in accordance with this invention. One solution involves the combination of logic circuitry with a PLA to implement the function of a state diagram implemented in the prior art only with a relatively large PLA. In a preferred solution two or more separate PLA's are cooperative in a manner to implement the function of a single prior art PLA. Particularly, the interconnection of two or more finite state machines via at least one electrical path which may or may not include logic is considered a significant departure from prior art thinking.

In one embodiment herein the problem is solved by applying the normal clock pulse to the input register of a PLA by way of an AND circuit which can be enabled selectively. In the absence of the AND circuit, whenever a phase U₁ clock pulse occurs in each cycle of operation, a PLA output (state) is required to orchestrate operation of the elements of the data path portion or to orchestrate a pause during the next consecutive U₂ phase operation. Additional ROM word lines are required to continue a state through consecutive clock cycles, and the power of a relatively large PLA results. The use, for example, of an AND circuit to inhibit the clock permits the number of output states to be reduced. Concomitants to this reduction are a reduction in the requisite number of word lines, a reduction in the requisite PLA area, and an increase in speed. With the AND circuit present, an enable signal, termed a "wait" signal, can be applied to provide for state continuance during a next consecutive U₁ phase. Pause states and associated word lines are obviated. The inclusion of means for enabling a logic element for applying a clock pulse to an input register to a PLA is also considered a significant departure from prior art thinking. The arrangement is particularly useful for a microprocessor which includes two or more PLA's each illustratively dedicated herein to associated elements in the data path portion of the chip. Each PLA and its associated data path elements are operative largely independently, being constrained primarily in that other largely independent operations are necessarily coordinated in order to achieve the usual kinds of processing results. The coordination typically is accomplished by a first PLA applying an enable "wait" signal to a second PLA. The signal operates to enable an AND circuit to provide an input clock signal to the second PLA. The coordination of multiple PLA's, particularly with each PLA having processing elements associated exclusively with it, is considered another departure from prior art thinking and leads to a powerful hierarchical PLA control arrangement. A contemporary arrangement discloses the use of logic elements between master-slave latches for performing processing during otherwise unused time which occurs between the U₁ and U₂ pulses of a standard clock cycle.

The present invention capitalizes on the use of such logic primarily to reduce PLA complexity by gating timing pulses but also to achieve such ends by gating data. In both cases, hierarchical PLA controls are achieved. Brief Description of the Drawing

FIG. 1 shows a schematic block diagram of a microprocessor organization;

FIGS. 2 and 3 show a prior art PLA arrangement in some detail and in schematic form, respectively; FIGS. 5 and 6 show, in some detail and schematically, respectively a PLA arrangement in accordance with one embodiment of this invention;

FIGS. 4 and 7 are state diagrams for the arrangements of FIGS. 2 and 5, respectively; and FIGS. 8, 9, and 10 are schematic diagrams of alternative embodiments of this invention. Detailed Description

FIG. 1 shows a semiconductor integrated circuit chip 10 including a PLA, a control, and a data path portion 11, 12, and 13, respectively. FIG. 2 shows a prior art arrangement for the PLA portion of FIG. 1. The arrangement includes the conventional components, namely the decoder portion 14 (the AND plane), the ROM portion 15 (the OR plane), and the associated input and output buffers 16 and 17, respectively. FIG. 2 shows the arrangement in some detail; FIG. 3 shows the arrangement schematically with the addition that the familiar input and output registers (latches) 18 and 19 are shown connected to the buffers. FIG. 3 also shows clock signals U₁ and U₂ being applied to input register 13 and to output register 19, respectively.

The circuit typically includes an array of transistors (not shown) in the encoder and in the ROM sections to define the specific pattern of bits generated by the output register in response to a particular input code as per the representative state diagram. The array of transistors is not being shown because a discussion of the generation of a particular pattern of bits is unnecessary to an understanding of the invention. All that is important here is that the arrangement of FIG. 2 shows four word lines 20, 21, 22, and 23 which permit a number of potentially different binary output words at output register 19.

A simple state diagram for the arrangement of FIGS. 2 and 3 is shown in FIG. 4. The possible input bits S₀, S₁, and wait are shown in FIG. 3. The possible ouput bits are N₁, N₀ and SIG1. Four states are shown, one represented by each of blocks 30, 31, 32, and 33 in FIG. 4. The operation represented in FIG. 3 typically requires four word lines in the (decoder AND) ROM portions of the PLA. FIGS. 5 and 6 show, in detail and schematically, respectively, a PLA arrangement in accordance with an embodiment of this invention. A comparison of FIGS. 5 and 2 shows that the arrangement of FIG. 5 has only three word lines, namely 120, 121, and 122. The arrangement of FIG. 5 does, on the other hand, include an AND circuit 125 to an input 126 of which clock signal U₁ is applied. An enable signal "wait" is applied selectively to a second input 127 of AND circuit 125. FIG. 7 shows the state diagram for the embodiment of FIGS. 5 and 6. It can be seen from a comparison of the state diagram of FIGS. 4 and 7 that the described operations are achieved with fewer word lines and one fewer input in the arrangement of FIGS. 5 and 6.

AND circuit 125 is shown in FIG. 6 also. The AND and OR planes in FIGS. 5 and 6 are designated 211 and 212, the corresponding input and output registers being designated 213 and 214, respectively. The wait signal is operative to gate the clock (U₁) pulse for the entire input code in this embodiment.

In a related embodiment of FIG. 8, a plurality of feedback loops are connected between the output register and the input register 323 and 324, respectively, of a ROM section and a decoder section 325 and 326 of a PLA 327. The feedback loops are designated 1_l ... 1_n. An AND circuit 328 is employed in a manner similar to that shown in FIG. 6. The presence of such loops also is implied by the state diagram of FIGS. 4 and 7. Consequently, each of the arrangements of FIGS. 3 and 6 may include feedback loops and FIG. 8 is intended to illustrate the presence of such loops. FIG. 9 shows an embodiment in which signals in feedback loops between the ROM section of a PLA and the encoder section of the same PLA as shown in FIG. 8 are selectively inhibited by a signal from a second PLA. The figure shows decoder and ROM sections 411 and 412 of a first PLA along with input and output registers 413 and 414. Decoder and ROM sections 415 and 416 of a second PLA along with input and output registers 417 and 418 are also shown. Representative feedback loops 420, 421, and 422 interconnect outputs of register 414 to input register 413. Each feedback loop includes an AND circuit, and an output of register 418 of PLA #2 is connected to an input of each of the AND circuits. The AND circuits are designated 130, 131, and 132 for loops 420, 421, and 422, respectively. We have seen that the provision of a control signal to a logic circuit such as an AND circuit to selectively enable clock pulses to be applied to an input register of a PLA results in the realization of a given set of operations with a PLA of reduced size. A like reduction in size is achieved with the embodiment of FIG. 9, again capitalizing on the use of a logic circuit but in this instance for processing data between U₁ and U₂ phases of a clock cycle. For the examples given, a reduction of the number of word lines of 25% is achieved. In a practical embodiment, a typical prior art PLA may contain 150 word lines which number is reduced to slightly over 100 by the use of a logic circuit in accordance with the foregoing discussion. Again size reductions of about 25% are achieved. We will now show that the judicious use of such AND circuits enables a hierarchical PLA architecture to be achieved, leading to improved microprocessor operation as well as to a reduction in size.

FIG. 10 shows a portion of a semiconductor microprocessor chip 500 and a memory 501 external to the chip. The microprocessor includes a Main PLA 505, a Fetch PLA 506, and an Arithmetic PLA 507. PLA 506 is associated with user, registers 510, 511,...517 and with associated output tri-state buffers 510A, 511A...517A. PLA 507 is associated with arithmetic logic unit 520. The microprocessor also includes two temporary registers 521 and 522, and data bus 525. Data and control inputs and outputs (I/O) are designated 526 and 527, respectively.

The first illustrative operation of the arrangement of FIG. 10 is directed at moving the contents of two selected ones of user registers 510...517 to the temporary registers 521 and 522 under the control of Fetch PLA 506 and, thereafter, to carry out an "add" operation in ALU 520 under the control of Arithmetic PLA 507. Both PLA's 506 and 507 are under the control of Main PLA 505 during the operation, and the PLA's are interconnected in a manner to enable clock pulses to be applied to input registers as described in connection with FIG. 6.

We will adopt the convention that action starts on a (U₁) clock cycle at which time the Main PLA is assumed to receive a valid command input from control I/O 527. On the next subsequent phase (U₂), Main PLA 505 applies its valid output to the input registers 561 and 559 of the Fetch and the Arithmetic PLA's 506 and 507. On the following phase U₁ , PLA's 506 and 507 have valid command inputs. On the next phase U₂, the Fetch PLA applies an output to a selected one of tri-state buffer circuits 510A, 511A, 512A,...517A and activates one user register. Outputs from the selected register are applied to bus 525. During this phase, PLA 507 enables temporary register 521 to receive data from bus 525.

During the next U₁ phase (cycle 3), the Fetch and Arithmetic PLA's receive a second valid command input. During the following U₂ phase, data from a second user register is applied to bus 525, and temporary register 522 is enabled. The operation to this point has resulted in data in first and second user registers to be stored in first and second temporary registers 521 and 522.

The fourth cycle of operation commences on a phase U₁ during which PLA 507 activates temporary registers 521 and 522. Registers 521 and 522 apply inputs to ALU 520 during this phase. During the following phase U₂, ALU 520 applies a valid output (data) to bus 525, and the Fetch PLA latches that data into a selected user register (510 517). The operation to this point results in the addition of two binary numbers stored in two user registers by moving those numbers to temporary registers and then driving the numbers through an ALU where addition occurs. The result is returned to a selected user register over the bus. The operation is summarized in TABLE I: TABLE I

Cycle Action (Single Cycle)

U₁ Main PLA receives first valid opcode from I/O

527 U₂ Main PLA applies valid output commands to Fetch and Arithmetic PLA's U₁ Fetch And Arithmetic PLA's have valid input command U₂ Fetch PLA enables first selected user register;

Arithmetic PLA enables first temporary register to receive data from bus; Main PLA outputs 2nd valid commands; assert data valid U₁ Fetch And Arithmetic PLA's receive second valid input command U₂ Fetch PLA enables second selected user register

Arithmetic PLA enables second temporary register to receive data from bus assert data valid U₁ temporary registers apply input data to ALU

U₂ ALU applies valid output data to bus,

Fetch PLA latches data from bus to user register

It is noted that tri-state buffer circuits 510A- 517A are enabled by outputs from PLA 506 applied from output register 562 through a slave latch designated 562S. Such a latch is employed because register 562 is operated in a U₂ phase and circuits 510A-517A are operative in U₁ and U₂ phases of a subsequent cycle and therefore isolation is required. A similar organization is required for PLA 507 in activating slave register 550. In each of these cases, a master-slave relationship exists and an opportunity arises for introducing logic in a manner to utilize unused time. No advantage of such an opportunity is taken in these instances. The use of slave latch 557S similarly provides isolation as required for latch 550 and permits operation of temporary registers 521 and 522. ALU 520 performs AND, OR, Add, Subtract, and

Complement functions. If the registers 521 and 522 have contents represented by TA and TB, respectively, the functions are symbolized by (TA OR TB), (TA AND TB), (TA + TB), (TA - TB), and (TA), respectively. Five bit register 550 determines the function performed and is itself enabled by an output from PLA 507 via line 558. The clock input to register 550 is connected to the output of AND circuit 552. One input to the AND circuit is connected to a clock source; the other to an output of output register 557 of PLA 507 via line 558. The output from

PLA 507 over line 551 is operative to program register 550 in a manner to enable the various ALU operations in sequence. The output from PLA 507 over line 558 enables a clock pulse to select the appropriate operation. Note that the clock to input registers 560, 561, and 559 of PLA's 505, 506, and 507 respectively is applied via inputs to AND circuits 563, 564, and 565, respectively. The second input to AND circuit 563 is connected to an output of register 565 of PLA 506. The second input to AND circuit 565 is connected to an output of register 562 also. The second input to AND circuit 564 is connected to an output of data I/O 526. The organization of the gated clock pulses is essentially as shown in FIGS. 5 and 6, being operative to selectively enable the Arithmetic PLA from stepping through its state diagram for selecting temporary registers and for determining the function of the ALU, etc.

Similarly, AND circuit 564 is operative responsive to a control signal from I/O 526 to enable selection of a next subsequent user register or memory address only when the prior fetch operation is completed. In the absence of such a signal, the clocks at the input registers 560 and 561 of PLA's 505 and 506 are inhibited. The signals may be understood as "assert data valid" signals or "handshake" signals and occur in the above example at cycle 2, phase U₂, and cycle 3, phase U₂. In cycle 2, phase U₂, the output from register 562 of PLA 506 selects the user register and enables output register 557 of PLA 507 to select the temporary register and AND circuit 552 to determine the function of ALU 520.

In cycle 3, phase U₂, the handshake is similar. The input register 560 of Main PLA 505 is connected to the output of AND circuit 563 in an arrangement similar to that of AND circuit 564 as noted hereinbefore. The handshake signals from PLA 506 are applied to selectively enable the clock at circuit 563 to permit outputs from PLA 505 to proceed to a next subsequent operation as would occur in cycle 3, phase U₁ in TABLE I.

When the output from the selected user from register (510-512) is an address to memory 501, an indeterminate number of cycles of operation may occur before the address is acquired as in the case, for example, when a disk file is searched. In such a multiple cycle of operation, cycle 2, phase U₂ of TABLE I is expanded as summarized in TABLE II.

TABLE II

Cycle Action (Multiple Cycle)

U₁ Fetch and Arithmetic PLA's have valid inputs commands U₂ Contents of selected user register applied to output latches of data I/O 526, U₁ No new command - address goes out on the pins U₂ Address goes from output latches into memory U₁ Memory responds to address, data ready? U₂ If data not ready, hold Fetch PLA (which in turn holds Main PLA)

(i.e., no handshake signal to AND circuit 564;)-Repeat If data ready, latch input of data I/O 526 to enable clock at AND circuit 564. data from memory applied to bus 525 for storage in temporary store TA or TB.

In the absence of a gated clock means at the input registers of the Main, the Arithmetic, and Fetch PLA's, each of those PLA's would have to be considerably larger as discussed hereinbefore. Moreover, reduction in PLA size (and thus increased speed) is achieved whenever gated clock means is used. The use of gated clock to an input register of a PLA (as shown in FIG. 6) or the use of a gate to gate data from or an output register of a PLA (as shown in FIG. 10) achieves like savings. The architectural strategy of the microprocessor determines which gating means is employed or whether both are used in any particular case.

The organization of the PLA's in a control hierarchy permits each PLA to apply successive commands to elements, such as on ALU and temporary registers, which are dedicated to it. In this manner, independent PLA's may proceed with successive operations independently where a Main PLA is operative to initiate those independent operations. The handshaking signals indicate that the various independent operations have been completed, and the next subsequent command is permitted. Concurrent PLA execution resulting in pipelining of data manipulation (Parallel Processing) thus is achieved. The gating of clock signals to the PLA's of a microprocessor with a plurality of PLA's organized in a hierarchy is a powerful arrangement leading not only to size and overall speed advantages, but also to throughput advantages. The lastmentioned advantage arises because the independent PLA's can utilize cycle time which would otherwise be unavailable for use if only a single (relatively large) PLA were to be used.

The implementation of a control hierarchy herein may include a direct interconnection between an output register of one PLA and an input register to another. Such an interconnection is represented by line 600 in FIG. 10 and may be used in reset operations.

The invention has been described herein to include programmable logic arrays which are state machines having input and output latches. But there are other potential elements for use with the invention. For example, it is possible to employ the gated clock means and the hierarchical organization in a multiple ROM or ROM/PLA arrangement directed to the same end. Further, the gating of data or the gating of clock pulses may be considered embodiments of the use of logic between master and slave latches. The embodiment of FIG. 9, for example, does not show a gated clock. It does show logic circuitry between a master and a slave latch, which in FIG. 9 are the output register (latch) of a first PLA and the input register (latch) of a second PLA, respectively. Gated clock means as shown in FIGS. 6, 8, and 10 also may be considered to constitute logic between a master latch and a slave latch. For example, in the embodiment of FIG. 10, the master latch and the slave latch are output and input registers, respectively. In any case, the master-slave relationship exists and logic circuitry is employed to utilize otherwise unused time therein. The inclusion of logic circuitry whether for gating clock signals or for manipulating data between input and output latches of a plurality of PLA's is considered particularly unique herein leading to the powerful PLA control hierarchy disclosed as was mentioned hereinbefore. What has been described is considered merely illustrative of the principles of this invention. Various modifications of the invention can be devised by those skilled in the art in accordance with those principles within the spirit and scope of the invention as encompassed by the following claims. Specifically, the invention can be implemented in NMOS , PMOS , pseudo NMOS , CMOS , etc . integrated circuit technology as is apparent to one skilled in the art. Moreover, although the invention has been described in terms of enabling a clock pulse or enabling data to be applied to input registers, an alternative mode in which normally-present clock pulses are disabled can be implemented to this same end. Also, other than AND circuits may be used to manipulate data or the clock. in otherwise unutilized time between consecutive clock pulses as described herein. Moreover, the invention need not be practiced in a single integrated circuit chip. Components may be defined in more than one chip (or discrete component) and still take advantage of unused tine as disclosed herein. In addition, as would be apparent to one skilled in the art, more than one input can be employed for gating a clock signal or data; more than one gate may be used also.

Claims

1. An integrated circuit structure including first and second logic arrays having input and output registers, respectively and being operative respectively in first and second phases of each of a succession of clock cycles, and means for applying inputs to said input register, said structure being CHARACTERIZED BY control means responsive to an output from said output register during a second phase and operative prior to the next subsequent first phase for selectively modifying said inputs during said next subsequent first phase.

2. An integrated circuit structure in accordance with claim 1 wherein said first logic array comprises a decoder section of a first PLA and said second logic array comprises a ROM section of a second PLA, and clock means for applying clock pulses in said first and second phases, said structure being further CHARACTERIZED IN THAT said control means is adapted to selectively apply clock pulses to said input register responsive to signals from said output register.

3. An integrated circuit structure in accordance with claim 1 wherein said first logic array comprises a decoder section of a first PLA and said second logic array comprises an input-output register, and clock means for applying clock pulses in said first and second phases, said structure being further CHARACTERIZED IN THAT said control means is adapted to selectively apply clock pulses to said input register responsive to signals from said input-output register.

4. An integrated circuit. structure in accordance with claim 2 wherein said control means comprises an AND circuit having at least first and second input terminals and at least one output terminal and said clock means and said output register are electrically connected to said first and second input terminals and said input register is connected to said one output terminal in a manner to clock input signals selectively thereto during a first phase.

5. An integrated circuit structure in accordance with claim 3 wherein said control means comprises an AND circuit having at least first and second input terminals and at least one output terminal and said clock means and said output register are electrically connected to said first and second input terminals and said input register is connected to said one output terminal in a manner to clock input signals selectively thereto during a first phase.

6. An integrated circuit structure in accordance with claim 4 also including a third logic array comprising a ROM section of said first PLA and having an associated output register and feedback loops between outputs of said associated output register and inputs to said decoder section, said structure being further CHARACTERIZED IN THAT said control means is adapted to selectively manipulate signals from said associated output register to said input register of said decoder section responsive to signals from said output register of said ROM section.

7. An integrated circuit structure in accordance with claim 6 wherein said control means comprises an AND circuit in at least one of said feedback loops, each of said AND circuits having at least first and second input and at least one output terminal, an output of the output register of each of said first and second PLA's being connected to said first and second input terminals, said output terminal being connected to an input of said input register.