Programmable Application Specific Integrated Circuit and Logic Cell Therefor

1. Field of the Invention

The present invention relates to application specific integrated circuits, and more specifically to field programmable gate arrays and logic cells therefor.

2. Description of Related Art

Application specific integrated circuits ("ASIC") provide their users the ability to manufacture products having a proprietary design without having to begin the design at the device level. Many different ASIC technologies are available, including gate array, standard cell, full custom design, and programmable logic devices. The gate array, standard cell, and full custom technologies offer high performance and flexibility, although at the expense of a long development cycle and a high non-recurring engineering cost. Programmable logic devices, which include programmable logic array devices ("PLD") and field programmable logic array devices ("FPLA"), are useful in a variety of applications such as glue logic and simple integrated circuit designs. Their utility has been limited in general due to a relatively low number of available usable gates, poor flexibility in implementing the required logic functions, low speed of operation, and high power consumption.

The reason for the limited utility of FPLA and PLD devices is generally due to the exponential increase in the size of the array as the number of logic elements increase. A FPLA typically contains a programmable AND array and a programmable OR array, while a PLD typically contains a programmable AND array and a fixed OR array. The programmable elements in a programmable array are constructed of active devices such as diodes or transistors connected to a fusible link, ultraviolet erasable floating gate devices, and electrically erasable floating gate devices. Generally, such devices are resistive and high in parasitic capacitance, and consume a relatively large silicon area. Since generally each product term consumes a certain amount of power, as the array size increases, the power dissipation increases. Moreover, the speed decreases with increasing array size, due to the increasingly parasitic capacitance of the array. Hence, for a given technology, the size of the array is restricted by practical considerations.

The utility of PLDs and FPLAs has been further extended by the technique of a programmable interconnect network for modules consisting of small PLDs or FPLAs. While some improvement in maximum practical size is realized, the degree of improvement is limited by the same problems afflicting PLDs and FPLAs; specifically, speed deterioration and power dissipation.

One device of the programmable interconnect small array type is disclosed in U.S. Pat. No. 4,774,421, issued Sep. 22, 1988 to Hartmann et al. The device consists of a plurality of "macrocells," each including a programmable AND array, an OR/NOR array, a feedback row driver, and an I/O driver and input circuit. The macrocells are driven from two common buses, one carrying input signals and the other carrying feedback signals. This device achieves a logic complexity of 300 2-input NAND gates.

Another device of the programmable interconnect small array type is disclosed in U.S. Pat. No. 4,847,612, issued Jul. 11, 1989. A plurality of functional units are arranged in a matrix of rows and columns. Each functional unit has 12 inputs and 9 outputs, although the number may vary. Each functional unit performs one or more logical functions. Vertical lines are permanently connected to the outputs of the functional units, while horizontal lines are permanently connected to the inputs of the functional units. Each of the horizontal and vertical lines is a collection of conductive lines whose number matches the number of inputs and outputs of one of the functional units. The horizontal and vertical lines cross at various areas of the device to form programmable interconnection matrices. Programming may be by mask, conductive fuses, or EPROM/EEPROM switching transistors.

Each functional unit is a relatively small programmable logic array which is mask programmable, fuse programmable, or switch programmable. Other types of functional units such as random access memories and arithmetic logic units may be present. Moreover, additional specialized logic at the output of the sum terms may be present.

Electrically configurable gate arrays have been developed to overcome the gate density limitations of programmable array logic devices and field programmable logic arrays. One such device is disclosed in U.S. Pat. No. 4,758,745, issued Jul. 19, 1988 to El Gamal et al. In the El Gamal et al. patent, a universal logic module having 3 input terminals and 2 output terminals and functioning as a 2:1 multiplexer is shown. Different types of array modules as well as combinations of two or more types may be used, as well as RAM arrays, ROM arrays, multipliers, and ALUs optimized to implement certain functions. The five terminals of the logic cell are hardwired to five separate vertical wiring channels. Wire channels are segmented and offset.

A similar device to that disclosed in the aforementioned El Gamal et al. patent is disclosed in an article by K. A. El-Ayat et al., "A CMOS Electrically Configurable Gate Array," IEEE Journal of Solid State Circuits, Vol. 24, No. 3, June 1989, pp. 752-61. The chip has a channeled gate array architecture consisting of configurable logic modules organized in rows and columns and separated by wiring channels. The wiring channels contain predefined segmented metal tracks of different segment length to accommodate the routing requirements. Antifuse elements are located at the intersection of the horizontal and vertical wire segments, and also between wire segments in association with isolation transistors. Circuit connections and module configuration are established by programming the appropriate antifuse elements, which then form low impedance connections as required between metal segments. The logic module is configurable, as are the I/O buffers.

El-Ayat et al. discloses a configurable logic module having eight inputs and one output, which is claimed to efficiently implement both combinatorial and sequential circuits and to optimally utilize routing resources. The module implements a 4:1 multiplexer function with inputs A-D; select inputs SA, SB, S0 and S1; and output Y. To implement the required logic function, the module is configured as the desired macrocell by programming the appropriate antifuses at its input terminals to connect the inputs to the required nets and to VDD and VSS.

Despite many improvements having been made in increasing usable gate density while reducing power dissipation, a need exists for a field programmable gate array device having even higher speed, higher density, lower power dissipation, and a more flexible architecture.

These and other disadvantages of the prior art are overcome in accordance with the present invention, which in one embodiment involves a programmable routing network, a programmable configuration network integrated with the programmable routing network, and a logic cell (i.e., logic module) integrated with the programmable configuration network.

In accordance with one embodiment of the present invention, a logic module comprises a first, a second and a third multiplexer as well as a flip-flop. An output lead of the first multiplexer is coupled to a first data input lead of the third multiplexer and an output lead of the second multiplexer is coupled to a second data input lead of the third multiplexer. Select input leads of the first and second multiplexers are coupled to input terminals of the logic module and are not coupled within the logic module to any multi-input logic gate. Similarly, a select input lead of the third multiplexer is coupled to an input terminal of the logic module and is not coupled within the logic module to any multi-input logic gate. A data input lead of the flip-flop is coupled to an output lead of the third multiplexer and an output data lead of the flip-flop is coupled to an output terminal of the logic module.

In accordance with some embodiments, the logic module also includes a logic gate having multiple input leads so that each of the multiple input leads is coupled to a respective input terminal of the logic module and so that an output lead of the logic gate is coupled to an output terminal of the logic module. In some embodiments, this logic gate is a six input AND gate having both inverting and non-inverting input leads. In some embodiments the logic module includes a second logic gate, the input leads of the second logic gate being coupled to respective input terminals of the logic module and the output lead of the second logic gate being coupled to an output terminal of the logic module.

A logic module structure in accordance with some embodiments of the present invention achieves high speed and/or reduced silicon area in implementations of some types of user-specific circuits.

In the Figures, where like reference numerals indicate like parts,

FIG. 1 is a schematic block diagram showing a 2.times.2 field programmable gate array of logic cells, in accordance with the present invention;

FIG. 2 is a schematic diagram of a routing network and a configuration network associated with a logic cell and useful in the field programmable gate array of FIG. 1;

FIG. 3 is a schematic diagram of an input/output cell useful in the field programmable gate array of FIG. 1;

FIG. 4 is a schematic diagram of an input cell useful in the field programmable gate array of FIG. 1;

FIG. 5 is a schematic block diagram of a logic cell useful in the field programmable gate array of FIG. 1;

FIG. 6 is a gate level schematic diagram of a macrocell realizable with the logic cell of FIG. 5;

FIG. 7 is a schematic circuit level diagram of a two input AND gate useful in the logic cell of FIG. 5;

FIG. 8 is a schematic circuit level diagram of an inverter useful in the logic cell of FIG. 5;

FIG. 9 is a schematic circuit level diagram of a six input AND gate useful in the logic cell of FIG. 5;

FIG. 10 is a schematic gate level diagram of a multiplexer group useful in the logic cell of FIG. 5;

FIG. 11 is a schematic block level diagram of a register or delay flipflop useful in the logic cell of FIG. 5; and

FIG. 12 is a schematic circuit level diagram of a NAND gate useful in the register of FIG. 11.

FIGS. 13A-L are a hierarchical transistor level circuit diagram of a particular implementation in accordance with the present invention.

FIG. 14 is a circuit diagram of a logic cell in accordance with another embodiment of the present invention.

The routing and logic elements of an illustrative field programmable gate array having four logic cells is illustrated in FIG. 1. The field programmable gate array of FIG. 1 is configured as a 2.times.2 array of logic cells. A larger array may be achieved merely by extending the architectural features of the 2.times.2 array of FIG. 1 either horizontally or vertically, as desired. The array size of a practical field programmable gate array generally ranges from about 32 logic cells to about 1000 logic cells, depending on the application and the process technology used in fabrication.

A number of basic architectural features are illustrated, including logic cells 2, 4, 6, and 8; vertical wire segments 10, vertical express lines 11, and power lines 7 associated with a column of logic cells 2 and 6; vertical wire segments 12, vertical express lines 13, and power lines 9 associated with a column of logic cells 4 and 8; horizontal wire segments 14 and 16 and horizontal express lines 15 and 17 associated with a row of logic cells 2 and 4; horizontal wire segments 18 and 20 and horizontal express lines 19 and 21 associated with a row of logic cells 6 and 8; input/output ("IO") cells 22-29; input driver ("I") cells 30-33, power ("VCC") pins 34 and 35; and ground ("GND") pins 36 and 37.

Generally, pass links are provided between contiguous segments of horizontal and vertical wires, and cross links are provided where horizontal and vertical segments cross. As the architecture of FIG. 1 is a replicating architecture, one may conveniently view sections of cross links and pass links as associated with a particular logic cell. For example, each of the four quadrants of FIG. 1 may conveniently be represented by the arrangement of FIG. 2, which shows the illustrative logic cell 100 associated with a routing network and a cell configuration network. The cell configuration network generally comprises illustrative track and logic cell link section 80. The routing network generally comprises illustrative substantially identical switch sections 82 and 84, illustrative substantially identical vertical pass link sections 90 and 92, and illustrative substantially identical horizontal pass link sections 94 and 96.

In the track and logic cell link section 80, generally each crossing of a vertical wire such as 41 and a horizontal wire such as 125 from an input or output of the logic cell 100 is linked by cross links such as cross link 81. Note the cross links (not referenced) at the crossing of vertical wire groups 40, 45, 46, 47 and 50 (wires 51 and 52 only) with horizontal input wire groups 103, 105, 107, 109, 111, and 113; with horizontal input wires 122, 123 and 124; and with horizontal output wires 115, 117, 119, 121, and 125.

In the illustrative switch sections 82 and 84, generally each crossing of a vertical wire such as 41 and a horizontal wire such as 66 is linked by a cross link such as 83. Note the cross links (not referenced) at the crossing of vertical wire groups 40, 45, 46, 47 and 50 (wires 51 and 52 only) with horizontal wire groups 60, 67, 70 and 77.

Contrary to the general arrangement of cross links, the vertical wires 53 and 54 of wire group 50 and the vertical wires 56 and 57 of the wire group 55 are not provided with cross links to all of the horizontal wires. Vertical wire 56 is a power bus and vertical wire 57 is a ground bus. As ground and power are not applied to the output nodes of the logic cell, no cross links are provided in the track and logic cell link section 80 between the vertical wires 56 and 57 and the output wires 115, 117, 119, 121 and 125. Similarly, no cross links are provided in the switch sections 82 and 84 between the vertical wires 56 and 57 and the horizontal wire groups 60 and 70 and the horizontal express lines 67 and 77. Vertical wires 53 and 54 of wire group 50 are particularly suitable for certain widely distributed signals such as set, reset, and clock. Accordingly, cross links are provided between vertical wires 53 and 54 and the following horizontal wires: clock wire 122, set wire 123, reset wire 124, and express wires 68, 69, 78 and 79. In order to avoid unnecessarily loading these lines with the small but finite leakage and capacitance of additional cross links, no cross links are provided at the other wire crossings in track and logic cell link section 80 or in the switch sections 82 and 84.

As a general rule, all cross links such as 81 of track and logic cell link section 80, and 83 of switch section 82 are programmable. In the unprogrammed state, the cross links provide an open circuit. A particularly suitable type of cross link is the plasma-enhanced chemical vapor deposited ("PECVD") amorphous silicon antifuse disclosed in copending U.S. patent application Ser. No. 447,969 entitled "Method for Fabrication of Programmable Interconnect Structure," filed Dec. 8, 1989 by Whitten et al., which hereby is incorporated herein by reference. Other types of cross links may be suitable as well, however.

Many of the horizontal and vertical wires of the field programmable gate array of FIG. 1 are wires for global connections. Most of these are segmented, with pass links being provided at opposing ends of contiguous segments. As shown in FIG. 2, for example, the vertical wires of wire groups 40, 45, 46 and 47 are segmented with pass links being indicated generally at 90 and 92; horizontal wires 63, 64, 65 and 66 of wire group 60 are segmented with respective pass links indicated generally at 94; and horizontal wires 73, 74, 75 and 76 of wire group 70 are segmented with respective pass links indicated generally at 96. Some wires are not segmented. These are generally known as "express" wires and are particularly useful for routing such signals as clock, set, and reset, as well as other signals that must be widely distributed about the chip without unnecessarily loading the drivers with the small but finite leakage and capacitance of the pass links. Unsegmented wires include the vertical wires of wire groups 50 and 55; horizontal wires 61 and 62 of the wire group 60; horizontal wires 68 and 69 of wire group 67; horizontal wires 71 and 72 of wire group 70; and horizontal wires 78 and 79 of wire group 77.

Other wires are used for local purposes and are not segmented. Such wires include the horizontal wires of wire groups 103, 105, 107, 109, 111, and 113; and the horizontal wires 115, 117, 119, 121, 122, 123, 124 and 125.

A particularly suitable type of pass link is the PECVD amorphous silicon antifuse disclosed in the aforementioned copending United States Patent Application of Whitten et al., which is incorporated herein by reference. Other types of pass links may be suitable as well, however.

An illustrative IO cell 500 is shown in FIG. 3. The IO cell 500 is suitable for input only or output only, or for bidirectional signaling with a tristate enable control. IO cell 500 includes input/output pin 502. The input circuit comprises pin 502 connected to the input of a driver 504. The output of driver 504 is supplied to a vertical wire or a horizontal wire, as desired. The output circuit comprises OR gate 506 having one true input and one complement input. Output polarity is determined by whether the true or complement input to the OR gate 506 is used. The output of OR gate 506 is supplied to a tristate driver 508, the control terminal 509 of which is connected to a vertical or horizontal wire, as desired. The output of the tristate driver 508 is connected to pin 502.

An illustrative input or I cell 510 is shown in FIG. 4. I cell 510 includes an input pin 512 connected to a double driver buffer 514. Driver 514 is suitable for use in driving high fanout networks such as a clock distribution network, set networks, and reset networks. The driver 514 has a true output and a complement output, which are supplied to respective vertical or horizontal wires, as desired.

VCC pins 34 and 35 are connected to a power distribution bus which extends around the periphery of the field programmable gate array of FIG. 1 and which includes vertical power buses associated with respective columns of logic cells such as a first column comprising logic cells 2 and 6, and a second column comprising logic cells 4 and 8. Similarly, GND pins 36 and 37 are connected to a ground distribution bus which extends around the periphery of the field programmable gate array of FIG. 1 and which includes vertical ground buses associated with respective columns of logic cells such as a first column comprising logic cells 2 and 6, and a second column comprising logic cells 4 and 8.

An illustrative logic cell 100 is shown in FIG. 5. Logic cell 100 includes four two input AND gates 104, 106, 108 and 110, each of which has one true input and one complement input. Gate 104 receives a true input 105.1 and a complement input 105.2, and furnishes its output to a data input of 2:1 multiplexer 114. AND gate 106 receives a true input 107.1 and a complement input 107.2, and furnishes its output to the other data input of 2:1 multiplexer 114. Similarly, AND gate 108 receives a true input 109.1 and a complement input 109.2, and furnishes its output to a data input of 2:1 multiplexer 116. AND gate 110 receives a true input 111.1 and a complement input 111.2, and furnishes its output to the other data input of 2:1 multiplexer 116.

The select terminals of both multiplexer 114 and multiplexer 116 are driven by the output of six input AND gate 112, which receives true inputs on lines 113.1, 113.3 and 113.5, and complement inputs on lines 113.2, 113.4 and 113.6. The output of AND gate 112 is also furnished as an output 115 of the logic cell 100.

The output of multiplexer 114 is furnished to one data input of the 2:1 multiplexer 118, and the output of multiplexer 116 is furnished to the other data input of 2:1 multiplexer 118. The output of multiplexer 116 is also furnished as an output 117 of the logic cell 100. The select terminal of multiplexer 118 is driven by the output of AND gate 102, which receives true inputs on lines 103.1, 103.3 and 103.5, and complement inputs on lines 103.2, 103.4 and 103.6. The output of AND gate 102 is also furnished as an output 119 of the logic cell 100.

The output of multiplexer 118 is furnished as an output 121 of the logic cell 100 as well as to the input of a delay or "D" flipflop 120, the output of which is furnished as an output 125 of the logic cell 100. Inputs to the D flipflop 120 include set terminal S and reset terminal R, which are accessible from input lines 123 and 124 respectively, and clock terminal CLK, which is accessible from input line 122.

Other outputs of the logic cell 100 include the output 119 of AND gate 102, the output 115 of AND gate 112, the output 117 of multiplexer 116, and the output 121 of multiplexer 118.

Logic cell 100 is a powerful general purpose universal logic building block suitable for implementing most TTL and gate array macrolibrary functions. The complexity of logic cell 100 is moderate, varying from ten equivalent 2-input NAND gates to twenty-five equivalent 2-input NAND gates depending on user programming. Combinational logic functions as wide as thirteen inputs are realizable with one cell delay. All boolean transfer functions are realizable for up to three inputs. Sequential flipflop functions such as T, JK and count with carry-in are realized with the use of the high performance D flipflop 120, which is initializable with the asynchronous set and reset inputs.

Many of the functions realizable with the logic cell 100 are listed in Table 1. Various realizable AND gates, NAND gates, OR gates, NOR gates, XOR gates, XNOR gates, and MAJ ("majority") gates are listed in Table 1, in which the designation "XiY" following the gate name means a gate of the designated type with X inputs, Y of which are complement and X-Y of which are true. Various realizable buffers and inverters are listed in Table 1, in which BUFF means a buffer, the designation "INVa" means an inverter with its output inverted, and the designation "INVb" means an inverter with its input inverted. The designation SOP14i7 ("sum of products") means the gate shown in FIG. 6. Various realizable multiplexers are listed in Table 1, in which "MUX XxY" means a multiplexer having X inputs with the inputs at positions corresponding to a 1 in the binary equivalent of Y being complemented and the other inputs being true; and in which "MUX XDxY" means dual multiplexers, each having X inputs with the inputs at positions corresponding to a 1 in the binary equivalent of Y being complemented and the other inputs being true. Various realizable decoder gates are listed in Table 1, in which DEC2t4 means a two to four decoder. Various realizable adder gates are listed in Table 1, in which HADDA means a half adder with inputs A and B and outputs sum and carry; FADD1 means a full adder with inputs A, B and carry and outputs sum and carry; CLAADD3 means a three bit carry look ahead adder, and CLAGEN3 means a three bit carry look ahead generator with enable. Various realizable latches are listed in Table 1, in which DLA means a D latch, DLAC means a D latch with clear, DLAMUX means a D latch with 2:1 multiplexer, DLAD means a dual D latch, DLAE means a D latch with enable, DLAEC means a D latch with enable and clear, DLAEMUX means a D latch with enable and a 2:1 multiplexer, DLADE means a dual D latch with enable, DLAiE means a D latch with inverted enable, DLAiEC means a D latch with inverted enable and clear, DLAiEMUX means a D latch with inverted enable and a 2:1 multiplexer, and DLADiE means a dual D latch with inverted enable. Various realizable flipflops are listed in Table 1, in which DFF means a delay flipflop, TFF means a toggle flipflop, and JKFF means a JK flipflop, while the suffix E means "with enable," the suffix iE means "with inverted enable," the suffix P means "with preset," and the suffix C means "with clear." The counter COUNT1 realizable listed in Table 1 is a simple counter having a count-in input, Q and count-out outputs, and preset, clear, and clock inputs.

The logic blocks of the logic cell 100 may be implemented in a number of suitable ways. One suitable implementation of the two input AND gates 104, 106, 108 and 110 (FIG. 5) is shown in the gate 200 in FIG. 7. Input 202 corresponds to the operand A, and controls the gate of p-channel MOSFET 204. When A=0, transistor 204 is ON, thereby presenting a logical ONE to the input of inverter 222 and resulting in a logical ZERO at the output 208. When A=1 transistor 204 is OFF and n-channel transistor 218 is ON, so that the output 208 is controlled by the state of input 210 which corresponds to the operand B. When B=0, p-channel MOSFET 212 turns ON and n-channel MOSFET 214 turns OFF. The drains of transistors 212 and 214 are connected to a common circuit node, to which the gates of p-channel MOSFET 216 and n-channel MOSFET 220 are connected. Accordingly, transistor 216 is OFF and transistor 220 is ON. As transistor 218 is ON due to A=1, the input of inverter 222 is brought to about ground potential or logical ZERO through transistors 218 and 220, which results in a logical ONE at output 208. When B=1, p-channel MOSFET 212 turns OFF and n-channel MOSFET 214 turns ON. Accordingly, transistor 216 is turned ON and transistor 220 is turned OFF, thereby presenting a logical ONE at the input of inverter 222 and resulting in a logical ZERO at output 208.

The inverter 222 may be implemented in any suitable manner, an example of which is shown in FIG. 8. The inverter 230 of FIG. 8 includes a p-channel MOSFET 232 with its source connected to VCC, and a n-channel MOSFET 234 with its source connected to GND. The drains of transistors 232 and 234 are connected to a common output node 236, and the gates of transistors 232 and 234 are connected to a common input node 238. Transistor 232 conducts when input node 238 is LOW, driving the output node 236 HIGH. Transistor 234 conducts when input node 238 is HIGH, driving the output node 236 LOW.

One suitable implementation of the six input AND gates 102 and 112 (FIG. 5) is shown in the gate 250 of FIG. 9. Note the following general functional correspondence of components of circuit 250 with components of circuit 200 (FIG. 7): input terminals 252, 254 and 256 with terminal 202 (FIG. 7); transistors 258, 260 and 262 with transistor 204 (FIG. 7); output terminal 264 with terminal 208 (FIG. 7); input terminals 266, 268 and 270 with terminal 210 (FIG. 7); transistors 272, 274 and 276 with transistor 212 (FIG. 7); transistors 278, 280 and 282 with transistor 214 (FIG. 7); transistor 284 with transistor 216; transistor 286 with transistor 220 (FIG. 7); transistors 288, 290 and 292 with transistor 218 (FIG. 7) and inverter 263 with inverter 222 (FIG. 7). Because the operation of circuit 250 (FIG. 9) is apparent from the previous description of the operation of circuit 200 (FIG. 7), no further description of the operation of circuit 250 is necessary.

One suitable implementation of the multiplexers 114, 116 and 118 (FIG. 5) is shown as circuit 300 of FIG. 10. Multiplexer 114 (FIG. 5) includes n-channel MOSFET 314, the gate of which is connected to input 115; and n-channel MOSFET 316, the gate of which is connected to the output of inverter 322 and thereby controlled by the complement of the signal on input 115. Accordingly, only one of pass transistors 314 and 316 is selected at any given time in accordance with the logical state of the signal on input 115. Multiplexer 116 (FIG. 5) includes n-channel MOSFET 318, the gate of which is connected to input 115; and n-channel MOSFET 320, the gate of which is connected to the output of inverter 322 and thereby controlled by the complement of the signal on input 115. Accordingly, only one of pass transistors 318 and 320 is selected at any given time in accordance with the logical state of the signal on input 115.

Multiplexer 118 includes n-channel MOSFET 330, the gate of which is connected to input 102; and n-channel MOSFET 332, the gate of which is connected to the output of inverter 326 and thereby controlled by the complement of the signal on input 102. Accordingly, either pass transistor 330, which receives the output of multiplexer 114 from transistor 314 or 316, or pass transistor 332, which receives the output of multiplexer 116 from transistor 318 or 320, is selected at any given time in accordance with the logical state of the signal on input 102. The logical level passed by the selected one of the pass transistors 330 and 332 is furnished to an inverter 338, and supplied on output terminal 344. Inverter 338 is any suitable inverter, such as the inverter 230 of FIG. 8.

One suitable implementation of the flipflop 120 (FIG. 5) is shown as circuit 400 in FIG. 11. The master stage 430 of flipflop 400 comprises NAND gates 420 and 422 and transmission gates 412 and 414. The slave stage 432 of flipflop 400 comprises NAND gates 424 and 426 and transmission gates 416 and 418. Note that a typical transmission gate comprises a p-channel MOSFET and an n-channel MOSFET having their respective sources and drains connected to each other, and their respective gates controlled by the control signal and its complement. In FIG. 11, the control signal for the transmission gates 412, 414, 416 and 418 is signal CLK buffered by inverters 406 and 408, with the complement CLK being obtained from inverter 410. The flipflop 400 is reset when signal RT is HIGH (signal ST remains LOW), since the outputs of both NAND gates 422 and 426 are driven HIGH and furnished to output Q through buffer 428 through either transmission gate 416 or gate 418 depending on the state of signal CLK. Buffer 428 comprises two series-connected inverters. The flipflop 400 is set when signal ST is HIGH (signal RT remains LOW), since the outputs of both NAND gates 420 and 424 are driven HIGH, whereby the outputs of both NAND gates 422 and 426 are driven LOW and furnished to output Q through buffer 428 through either transmission gate 416 or gate 418 depending on the state of signal CLK.

When the signals on the direct inputs RT and ST are both LOW, the output of flipflop 120 is determined by the logical state of the clocked input D. Consider first the operation of the circuit 400 when the clock signal CLK is LOW, during which event the state of the flipflop 400 is latched in the slave stage 432 while the master stage 430 responds to the new state presented on input D. Transmission gate 412 is closed so that the master stage 430 is responsive to input D. Specifically, gate 412 passes the signal D to an input of NAND gate 420 (the other input being HIGH due to the inversion of signal ST by inverter 404). The output of NAND gate 420 is furnished to an input of NAND gate 422 (the other input being HIGH due to the inversion of signal RT by inverter 402), which causes D to appear at the output of NAND gate 422. Accordingly, the master stage 430 is driven into a particular state depending on the logical state of D. The output of NAND gate 422 and hence the master section 430 is isolated, however, because transmission gates 414 and 416 are open. Transmission gate 418 of the slave stage 432 passes the state of the output of NAND gate 426 to an input of NAND gate 424 (the other input being HIGH due to the inversion of signal ST by inverter 404), the output of which is furnished to an input of the NAND gate 426 (the other input being HIGH due to the inversion of signal RT by inverter 402). Accordingly, the state of the slave stage 432 is latched, and furnished as output Q through buffer 428. The state of the slave stage 432 is the previous state of the master stage 430.

Consider the operation of the circuit 400 when the clock signal CLK is HIGH, during which event the master stage 430 latches the prior state of input D and furnishes the state to output Q and the slave stage 432. Transmission gates 412 and 418 are open and transmission gates 414 and 416 are closed. Because gate 412 is open, input D is isolated. Transmission gate 414 of the master stage 430 passes the state of the output of NAND gate 422 to an input of NAND gate 420 (the other input being HIGH due to the inversion of signal ST by inverter 404), the output of which is furnished to an input of the NAND gate 422 (the other input being HIGH due to the inversion of signal RT by inverter 402). Accordingly, the state of the master stage 432 is latched. The latched state of the master stage 432 is furnished as output Q through transmission gate 416 and buffer 428, and also applied to an input of NAND gate 424 (the other input being HIGH due to the inversion of signal ST by inverter 404) through transmission gate 416. The output of NAND gate 424 is applied to an input of NAND gate 426, which drives the slave stage 432 into a particular state depending on the output of the master stage. The output of NAND gate 426 and hence the slave section 432 is isolated, however, because transmission gate 418 is open.

Note that the various gates of FIG. 11 may be implemented in any suitable manner. A suitable implementation of the inverter gates is the circuit of FIG. 8. A suitable implementation of the NAND gates is the circuit 440 of FIG. 12, in which the input to inverter 450 is raised to VCC through either p-channel MOSFET 442 or p-channel MOSFET 444 when either input A or input B respectively is LOW, and is lowered to GND through n-channel MOSFETs 446 and 448 only when both inputs A and B are HIGH. Inverters 450 and 452 are provided for purposes of driving the following device, and may be omitted if such drive is not required.

A variety of techniques may be used to program the pass links and cross links of the field programmable gate array 100. For example, in one suitable technique, programming transistors are provided in conjunction with each pass link, in conjunction with adjacent horizontal logic cell input wire segments (generally without pass links), and at edges of cell array, operatively connected to the horizontal and vertical wires. A programming voltage of, for example, 12 volts and a ground potential are selectively applied to horizontal and vertical wires in order to close certain antifuses, while a voltage of one-half the programming voltage is applied to other selected wires to prevent closing of other antifuses. The programming transistors are selectively driven ON in order to direct the programming voltages through the network of wires. Once the desired antifuses are closed, the programming transistors are no longer used. Antifuses may be programmed in accordance with techniques set forth in U.S. patent application Ser. Nos. 07/907,904, 07/937,331 and 07/937,071. The subject matter of U.S. patent application Ser. Nos. 07/907,904, 07/937,331 and 07/937,071 is incorporated herein by reference.

A variety of suitable technologies are available for fabricating the field programmable gate array circuit 100. One suitable process is a one micron, two layer metal CMOS process. In using such a process for the fabrication of field programmable gate array circuit 100, the configuration wires such as horizontal input wire groups 103, 105, 107, 109, 111 and 113, horizontal input wires 122, 123 and 124, and horizontal output wires 115, 117, 119, 121 and 125 preferably are fabricated in first metal; the vertical routing wires such as wire groups 40, 45, 46, 47, 50 and 55 (FIG. 2) are fabricated generally in second metal; and horizontal routing wires such as wire groups 60, 67, 70 and 77 (FIG. 2) are fabricated generally in first metal. Note that pass links such as the pass links indicated at 90, 92, 94 and 96 (FIG. 2) generally are fabricated using a combination of first and second metal, as more fully disclosed in the above-referenced application of Whitten et al., which is incorporated herein by reference.

FIGS. 13A-L is a hierarchical transistor level circuit diagram of a particular implementation of one portion of an integrated circuit chip having a plurality of such portions. FIG. 13A is high level diagram of such a portion. FIG. 13B is a diagram of the block labeled a1.times.8 in FIG. 13A; FIG. 13C is a diagram of the block labeled a4.times.1 in FIG. 13A; FIG. 13D is a diagram of the block labeled lrcon in FIG. 13A; FIG. 13E is a diagram of the block labeled a6.times.20 in FIG. 13A; FIG. 13F is a diagram of the block labeled a28.times.24 in FIG. 13A; and FIG. 13G is a diagram of the block labeled logblk in FIG. 13A. FIG. 13H is a diagram of the block labeled blkmux in FIG. 13G; and FIG. 13I is a diagram of the block labeled blkreg in FIG. 13G. FIG. 13J is a diagram of the gates labeled ptnand in FIG. 13H; FIG. 13K is a diagram of the gates labeled tccand in FIG. 13H; and FIG. 13L is diagram of the gate labeled a2oli.sub.-- 2 in FIG. 13I.

The logic cell 100 of FIG. 5 (sometimes referred to as a "logic module") comprises a gate 102, the output lead of which is coupled to the multiplexer select input lead of multiplexer 118. A gate 112 is also provided, the output lead of which is coupled to the multiplexer select input leads of multiplexers 114 and 116. It was believed that providing such gates in the signal path to the multiplexer select input leads would both minimize the size and also maximize the speed of user-specific designs realized in an integrated circuit having numerous such logic cells. It was believed that size would be minimized in part because the average user-specific design would utilize most of the multiplexer select input lead signals being driven by a signal which would be a logical combination of other signals. It was believed that the maximum operating speed of the average user-specific design would be maximized in part because the average user-specific design would entail the multiplexer select input leads being driven with a logical combination of the other signals. As a result, it was believed that providing the gates performing the logic functions inside the logic module would eliminate the delay associated with a long lead connecting the output of the gate performing the logic function and the multiplexer select input lead. Accordingly, the logic cell 100 of FIG. 5 comprises two logic gates 102 and 112 coupled inside the logic cell 100 to the multiplexer select input leads as illustrated. It has been discovered, however, that another logic cell results in at least some types of user-specific designs being realizable in a smaller silicon area and also results in an improvement in the maximum operating speed of such user-specific designs.

FIG. 14 is a circuit diagram of a logic cell 1400 in accordance with another embodiment of the present invention. Logic cell 1400 contains a first gate 1401, a second gate 1402, a third gate 1403, a fourth gate 1404, a fifth gate 1405, a first multiplexer 1406, a second multiplexer 1407, a third multiplexer 1408 and a D-type flip-flop 1409. Leads QFFS, A1-A6, OS, B1-B2, C1-C2, D1-D2, E1-E2, NS, QFFC and QFFR are input leads of the logic cell 1400. Leads AZ, OZ, QFFZ and NZ are output leads of the logic cell 1400. Whereas the multiplexer select input leads of multiplexers 114 and 116 of logic cell 100 of FIG. 5 are coupled inside the logic cell to an output lead of gate 112, the multiplexer select input leads of multiplexers 1406 and 1407 of FIG. 14 are both coupled to cell input lead NS. Whereas the multiplexer select input lead of multiplexer 118 of logic cell 100 of FIG. 5 is coupled inside the logic cell to an output lead of gate 102, the multiplexer select input lead of multiplexer 1408 of FIG. 14 is not coupled inside the logic cell 1400 to an output lead of gate 1401. Rather, the output lead of gate 1401 is coupled to cell output lead AZ and the multiplexer select input lead of multiplexer 1408 is coupled to cell input lead OS.

By eliminating gate 112 of the logic cell of FIG. 5 to achieve the logic cell 1400 of FIG. 14, silicon area in each logic cell is saved. First, eliminating the additional input leads supplying the inputs of gate 112 from outside the logic cell reduces the height of the logic cell. Moreover, eliminating transistors and interconnect forming the gate 112 results in a significant reduction in the silicon area associated with realizing each logic cell. Second, decoupling the gate 102 from the multiplexer select input lead of multiplexer 118 frees gate 102 for use in user-specific designs where the signal driving the multiplexer select input lead of multiplexer 118 need not be a logical function of multiple other signals received from outside the logic cell. This usable gate 1401 in each logic cell allows more of a user-specific design to be implemented in the same number of logic cells. Third, decoupling the output lead of gate 102 from multiplexer 118 breaks cell 100 into two smaller separately usable portions of logic. Accordingly, a gate 1401 of a logic cell located closer to a portion of the user-specific circuit requiring the gate can be used as opposed to a gate located in another logic cell located farther away. This reduces the amount of routing resources needed on-chip, thereby reducing the size of a chip containing a given number of logic cells.

Not only does the logic cell of FIG. 14 reduce the amount of silicon required to realize user-specific designs, but the logic cell of FIG. 14 also results in a speed improvement for several reasons. First, by extending the multiplexer select input leads of multiplexers 1406, 1407 and 1408 outside the logic cell, there is no propagation delay associated with driving the multiplexer select input leads indirectly through unnecessary gates. This has been found to more than compensate for the delay associated with situations in which a gate is required in the path driving a multiplexer select input lead and where connections outside the logic cell are required in order to couple such a gate to the multiplexer select input lead. Second, because the gates (such as gate 1401) of logic cells can be used independently of the use put to the remainder of the logic cells containing the gates, the length of interconnecting leads extending to and from such gates is reduced as described above. By shortening the leads associated with these gates, the capacitance and resistance of the leads are reduced, thereby increasing signal propagation speed over the leads. Where a shortened lead is a critical net in a user-specific design, the maximum operating speed of the user-specific design may be increased.

To verify that the logic cell 1400 results in smaller and faster implementations of at least some representative user-specific designs, hypothetical implementations of benchmark circuits using the logic cell 100 of FIG. 5 and using the logic cell 1400 of FIG. 14 were studied. The first benchmark circuit studied is a state machine known as "PREP3". The PREP3 state machine is commonly used as a benchmark in the field programmable gate array (FPGA) industry and this benchmark is believed to be representative of random logic in a typical user-specific design. The PREP3 state machine was estimated to be approximately 40 percent smaller and approximately 7 percent faster when realized using the logic cell 1400 of FIG. 14 versus using the logic cell 100 of FIG. 5. The second benchmark circuit studied is a four bit adder. This circuit is believed to be representative of circuits in a macro library of larger commonly used blocks of logic. The four bit adder was estimated to be approximately 60 percent smaller and about 7 percent faster when realized using the logic cell 1400 of FIG. 14 versus using the logic cell 100 of FIG. 5.

In accordance with some embodiments of the present invention, a portion of an integrated circuit chip may be implemented in similar fashion to the portion of an integrated circuit illustrated in FIGS. 13A-L. In one such embodiment, the logic cell of FIG. 14 would correspond with FIG. 13G and the corresponding lower level diagrams. In accordance with some embodiments of the present invention, an additional gate similar to input gate 1401 is provided in some or all logic cells, the output lead of the additional gate also extending from the logic cell containing the gate, the output lead of the gate not being coupled to a multiplexer select input lead, the input leads of the additional gate also extending outside the logic cell.

While our invention has been described with respect to certain embodiments, the scope of the present invention is defined by the appended claims. The embodiments described herein are merely illustrative. For example, the fabrication technology, the programming technique, the number and routing of the wires such as the various vertical wires and the various horizontal wires, and the precise composition of the logic cell are illustrative. Accordingly, other embodiments and variations not described herein may be within the scope of our invention, as defined by the appended claims.