ROBUST, LOW POWER, RECONFIGURABLE THRESHOLD LOGIC ARRAY

Abstract

A field programmable threshold-logic array (FPTLA) includes a number of threshold logic gates and a number of programmable interconnect elements. Each one of the programmable interconnect elements are connected between two or more of the threshold logic gates, such that the programmable interconnect elements route signals between the threshold logic gates. By using threshold logic gates for the FPTLA, the size of the FPTLA may be significantly smaller than conventional solutions. Further, using threshold logic gates results in significant improvements in the computation speed of the FPTLA when compared to conventional solutions.

Description

FIELD OF THE DISCLOSURE

The present disclosure relates to threshold logic circuitry and in particular to programmable threshold logic circuitry.

BACKGROUND

Logic circuitry continues to be the platform on which modern electronic devices are built. In recent years, advancements in electronic devices and their associated software have crystallized the demand for logic circuitry that is faster, smaller, and more efficient. Currently, logic circuitry (e.g., application specific integrated circuits (ASICs), field programmable gate arrays (FPGAs), and the like) is dominated by the use of design paradigms that have been in place since the dawn of computing technology. Specifically, current logic circuits utilize a large number of individual AND, OR, NOT, etc. logic gates connected to one another in a particular fashion to perform a particular operation. Although generally accepted as standard practice, the use of these conventional design paradigms is quickly resulting in a plateau in the size, speed, and efficiency achievable by the resulting logic circuitry.

In an effort to continue improving the capabilities logic circuitry, many designers and manufacturers have started to explore alternative design strategies. Threshold logic gates have been studied and determined to result in more compact and faster circuits when compared to conventional logic gate equivalents. As defined herein, threshold logic gates are circuits that evaluate a threshold function. A threshold function (denoted as f=[w₁, w₂, . . . , w_n|T]) is a Boolean function that receives multiple input values and a weight associated with each input value and computes a weighted sum of the input values and their corresponding weights. If the value of the weighted sum is above a predetermined threshold, the output of the threshold function is 1, whereas if the value of the weighted sum is below the predetermined threshold, the output of the threshold function is 0, as illustrated in Equation 1:

$\begin{array}{cc}f\left(x_1,x_2,\dots ,x_n\right)=\{\begin{array}{cc}1& \mathrm{if}{\sum }_{i=1}^nw_ix_i\ge T\\ 0& \mathrm{otherwise}\end{array}.& \left(1\right)\end{array}$

where f (x₁, x₂, . . . , x_n) is the Boolean function including input values x₁, x₂, . . . , x_n, w₁, w₂, . . . , w_i are the weights associated with each variable in the Boolean function, and T is the predetermined threshold. Each operation in a Boolean function can be expressed as a threshold function. Further, multiple Boolean operations in a Boolean function can often be expressed as a single threshold function. Since threshold logic gates evaluate a threshold function, a single threshold logic gate may replace a large number of conventional logic gates that would otherwise be required to perform a particular operation, thereby saving space and resulting in improvements in computation speed and efficiency.

Despite their advantages in size, speed, and efficiency, threshold logic gates have not entered the mainstream due to a lack of efficient and reliable gate implementations and a lack of necessary infrastructure for the design and programming of logic circuits including threshold logic gates. Specifically, threshold logic gates are known to be highly sensitive to manufacturing tolerances, which often results in an undesirable or unpredictable output. Further, the widespread adoption of the logic circuitry design paradigms discussed above has resulted in a large obstacle in the implementation of any logic gates that deviate from current design principles. As a result, logic circuitry including threshold logic gates, and in particular programmable logic circuitry including threshold logic gates, is nearly non-existent.

Accordingly, there is a present need for a threshold logic gate that is reliable, fast, small, and efficient. Further, there is a need for programmable logic circuitry incorporating one or more threshold logic gates that is compatible with current logic circuitry design paradigms.

SUMMARY

The present disclosure relates to threshold logic circuitry and in particular to programmable threshold logic circuitry. In one embodiment, a field programmable threshold-logic array (FPTLA) includes a number of threshold logic gates and a number of programmable interconnect elements. Each one of the programmable interconnect elements are connected between two or more of the threshold logic gates, such that the programmable interconnect elements route signals between the threshold logic gates. By using threshold logic gates for the FPTLA, the size of the FPTLA may be significantly smaller than conventional solutions. Further, using threshold logic gates for the FPTLA results in significant improvements in the computation speed and efficiency of the FPTLA when compared to conventional solutions.

In one embodiment, each one of the threshold logic gates is a differential threshold logic gate. Further, each one of the threshold logic gates may be configured to evaluate a majority function.

In one embodiment, each one of the threshold logic gates includes a first input network, a second input network, a differential amplifier, and a latch. The first input network receives a first set of input signals and generates a first activation signal at a speed proportional to a combined value of the first set of input signals. The second input network receives a second set of input signals and generates a second activation signal at a speed proportional to combined value of the second set of input signals. The differential amplifier is connected between the first input network and the second input network. Further, the differential amplifier receives the first activation signal and the second activation signal. If the first activation signal is received before the second activation signal, the differential amplifier generates a differential output signal having a first state. If the first activation signal is received after the second activation signal, the differential amplifier generates a differential output signal having a second state. The latch receives the differential output signal and generates a latched output signal based on the state of the differential output signal.

Those skilled in the art will appreciate the scope of the disclosure and realize additional aspects thereof after reading the following detailed description in association with the accompanying drawings.

BRIEF DESCRIPTION OF THE DRAWINGS

The accompanying drawings incorporated in and forming a part of this specification illustrate several aspects of the disclosure, and together with the description serve to explain the principles of the disclosure.

FIG. 1 is a schematic view of a differential threshold logic gate (DTG) according to one embodiment of the present disclosure.

FIG. 2 is a schematic view of a DTG according to an additional embodiment of the present disclosure.

FIG. 3 is a block diagram illustrating a field programmable threshold logic array (FPTLA) according to one embodiment of the present disclosure.

FIG. 4 is a flow diagram illustrating a method for programming the FPTLA shown in FIG. 3 according to one embodiment of the present disclosure.

DETAILED DESCRIPTION

The embodiments set forth below represent the necessary information to enable those skilled in the art to practice the disclosure and illustrate the best mode of practicing the disclosure. Upon reading the following description in light of the accompanying drawings, those skilled in the art will understand the concepts of the disclosure and will recognize applications of these concepts not particularly addressed herein. It should be understood that these concepts and applications fall within the scope of the disclosure and the accompanying claims.

It will be understood that, although the terms first, second, etc. may be used herein to describe various elements, these elements should not be limited by these terms. These terms are only used to distinguish one element from another. For example, a first element could be termed a second element, and, similarly, a second element could be termed a first element, without departing from the scope of the present disclosure. As used herein, the term “and/or” includes any and all combinations of one or more of the associated listed items.

Relative terms such as “below” or “above” or “upper” or “lower” or “horizontal” or “vertical” may be used herein to describe a relationship of one element, layer, or region to another element, layer, or region as illustrated in the Figures. It will be understood that these terms and those discussed above are intended to encompass different orientations of the device in addition to the orientation depicted in the Figures.

The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the disclosure. As used herein, the singular forms “a,” “an,” and “the” are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms “comprises,” “comprising,” “includes,” and/or “including” when used herein specify the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof.

Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure belongs. It will be further understood that terms used herein should be interpreted as having a meaning that is consistent with their meaning in the context of this specification and the relevant art and will not be interpreted in an idealized or overly formal sense unless expressly so defined herein.

Turning now to FIG. 1, a differential threshold logic gate (DTG) 10 is shown according to one embodiment of the present disclosure. The DTG 10 includes a first input network 12, a second input network 14, a differential amplifier 16, and a latch 18. The first input network 12 includes a first plurality of input transistors M_i11-M_i1k coupled in parallel between a first input control node 20 and a first activation signal node 22. Specifically, each one of the first plurality of input transistors M_i11-M_i1k includes a source contact coupled to the first input control node 20, a drain contact coupled to the first activation signal node 22, and a gate contact, which is configured to receive one of a first plurality of input signals IN₁₁-IN_1k. Further, the first input network 12 includes a first input control transistor M_c1 and a second input control transistor M_c2. The first input control transistor M_c1 includes a source contact coupled to a supply voltage (VDD), a drain contact coupled to the first input control node 20, and a gate contact, which is configured to receive the complement of a clock signal (C_CLK). The second input control transistor M_c2 includes a source contact coupled to ground, a drain contact coupled to the first activation signal node 22, and a gate contact, which is configured to receive the complement of the clock signal (C_CLK).

The second input network 14 is similar to the first input network 12, and includes a second plurality of input transistors M_i21-M_i2k coupled in parallel between a second input control node 24 and a second activation signal node 26. Specifically, each one of the second plurality of input transistors M_i21-M_i2k includes a source contact coupled to the second input control node 24, a drain contact coupled to the second activation signal node 26, and a gate contact, which is configured to receive one of a second plurality of input signals IN₂₁-IN_2k. Further, the second input network 14 includes a third input control transistor M_c3 and a fourth input control transistor M_c4. The third input control transistor M_c3 includes a source contact coupled to a supply voltage (VDD), a drain contact coupled to the second input control node 24, and a gate contact, which is configured to receive the complement of the clock signal (C_CLK). The fourth input control transistor M_c4 includes a source contact coupled to ground, a drain contact coupled to the second activation signal node 26, and a gate contact, which is configured to receive the complement of the clock signal (C_CLK).

The differential amplifier 16 includes a first differential input node 28, a second differential input node 30, a first differential output node 32, a second differential output node 34, and a plurality of amplifier transistors M_a1-M_a8. A first amplifier transistor M_a1 includes a source contact coupled to ground, a drain contact, and a gate contact coupled to the first differential input node 28. A second amplifier transistor M_a2 includes a source contact coupled to the drain contact of the first amplifier transistor M_a1, a drain contact coupled to the first differential output node 32, and a gate contact coupled to the second differential output node 34. A third amplifier transistor M_a3 includes a source contact coupled to the power supply voltage (VDD), a drain contact coupled to the first differential output node 32, and a gate contact coupled to the second differential output node 34. A fourth amplifier transistor M_a4 includes a source contact coupled to the power supply voltage (VDD), a drain contact coupled to the first differential output node 32, and a gate contact coupled to the first differential input node 28.

A fifth amplifier transistor M_a5 includes a source contact coupled to ground, a drain contact, and a gate contact coupled to the second differential input node 30. A sixth amplifier transistor M_a6 includes a source contact coupled to the drain contact of the fifth amplifier transistor M_a5, a drain contact couple to the second differential output node 34, and a gate contact coupled to the first differential output node 32. A seventh amplifier transistor M_a7 includes a source contact coupled to the power supply voltage (VDD), a drain contact coupled to the second differential output node 34, and a gate contact coupled to the first differential output node 32. Finally, an eighth amplifier transistor M_a8 includes a source contact coupled to the power supply voltage (VDD), a drain contact coupled to the second differential output node 34, and a gate contact coupled to the second differential input node 30.

In one embodiment, the latch 18 is a set-reset (SR) NAND latch. Accordingly, the latch 18 may include a set input (S) a reset input (R), a latching output (Q), and a complementary latching output (QN). The set input (S) is coupled to the first differential output node 32. The reset input (R) is coupled to the second differential output node 34. One of the latching output (Q) and the complementary latching output (QN) is used as the ultimate output (OUT, C_OUT) of the DTG 10. In other embodiments, any suitable component may be used for the latch 18.

In one embodiment, the first plurality of input transistors M_i11-M_i1k, the second set of input transistors M_i21-M_i2k, the input control transistors M_c1-M_c4, and the amplifier transistors M_a1-M_a8 are metal-oxide-semiconductor field-effect transistors (MOSFETs). Specifically, the first plurality of input transistors M_i11-M_i1k, the second set of input transistors M_i21-M_i2k, the input control transistors M_c1-M_c4, and the amplifier transistors M_a1-M_a8 may be a mix of NMOS and PMOS MOSFETs, such that the PMOS MOSFETs are indicated by the circle at the gate contact of the transistor. In other embodiments, the first plurality of input transistors M_i11-M_i1k, the second set of input transistors M_a1M_i2k, the input control transistors M_c1-M_c4, and the amplifier transistors M_a1-M_a8 may be any suitable transistors such as field-effect-transistors (FETs), high electron mobility transistors (HEMTs), bipolar junction transistors (BJTs) or the like.

As shown in FIG. 1, each one of the first input network 12 and the second input network 14 includes k input transistors. Accordingly, the DTG 10 is referred to as a DTG-K, and includes 2k inputs. DTGs 10 according to different embodiments of the present disclosure may be made with any number of input transistors (e.g., DTG-5, DTG-7, DTG-9), and therefore may provide a variety of input signal configurations.

In operation, the DTG 10 has two states of operation: a reset state of operation and a evaluation state of operation. In the reset state of operation, the clock signal (CLK) is 0, and the complement of the clock signal (C_CLK) provided to the gate contact of each one of the input control transistors M_c1-M_c4 is 1. As discussed herein, signals with a value of “1” refer to signals including a voltage, a current, or both a voltage and a current above a particular threshold, while signals with a value of “0” refer to signals with a voltage, a current, or both a voltage and a current below a particular threshold (i.e., digital signals). Using the complement of the clock signal (C_CLK) to operate the DTG 10 allows the reset state of operation to occur when the clock signal (CLK) is 0, and the evaluation state of operation of the DTG 10 to occur when the clock signal (CLK) is 1, thereby conforming to standard industry practices. When the clock signal (CLK) is 0 and the complement of the clock signal (C_CLK) is 1, the first control transistor M_c1 and the third control transistor M_c3 turn OFF (i.e., become an open circuit) and the second control transistor M_c2 and the fourth control transistor M_c4 to turn ON (i.e., become a closed circuit). Turning the second control transistor M_c2 and the fourth control transistor M_c4 ON connects the first activation signal node 22 and the second activation signal node 26 to ground, which in turn causes the first amplifier transistor M_a1 and the fifth amplifier transistor M_a5 to turn OFF and the fourth amplifier transistor M_a4 and the eighth amplifier transistor M_a8 to turn ON. This disconnects all discharge paths for the first differential output node 32 and the second differential output node 34, and pulls the first differential output node 32 and the second differential output node 34 high to the supply voltage (VDD) through the fourth amplifier transistor M_a4 and the eighth amplifier transistor M_a8, such that the first differential output node 32 and the second differential output node 34 have a value of 1. Note that the second amplifier transistor M_a2 and the sixth amplifier transistor M_a6 are turned ON, while the third amplifier transistor M_a3 and the seventh amplifier transistor M_a7 are OFF however, this does not affect the functionality of the DTG 10. As discussed above, the latch 18 may be an SR NAND latch, and therefore will not change the state of the latching output (Q) when both the set input (S) and the reset input (R) are 1. Accordingly, the latch 18 stores the previous output value of the first differential output node 32 and the second differential output node 34 in the reset state of operation.

In the evaluation state of operation, the clock signal (CLK) is 1, and the complement of the clock signal (C_CLK) provided to the gate contact of each one of the input control transistors M_c1-M_c4 is 0. This causes the first control transistor M_c1 and the third control transistor M_c3 to turn ON and the second control transistor M_c2 and the fourth control transistor M_c4 to turn OFF. Turning the first control transistor M_c1 and the third control transistor M_c3 ON causes the supply voltage (VDD) to be presented to the first input control node 20 and the second input control node 24. The supply voltage (VDD) at the first input control node 20 and the second input control node 24 will be delivered through one or more of the first plurality of input transistors M_i11-M_i1k and one or more of the second plurality of input transistors M_i21-M_i2k to the first activation signal node 22 and the second activation signal node 26, respectively. Depending on the input signals (referred to generally as “IN”) presented to the first input network 12 and the second input network 14, a different number of the first plurality of input transistors M_i11-M_i1k and the second plurality of input transistors M_i21-M_i2k will be turned ON in each one of the first input network 12 and the second input network 14. In general, the DTG 10 will always be operated in such a way that the number of active input transistors in the first input network 12 and the second input network 14 are different.

The number of the first plurality of input transistors M_i11-M_i1k that are ON in the first input network 12 determines how quickly the supply voltage (VDD) is passed from the first input control node 20 to the first activation signal node 22, and thus to the first differential input node 28. Similarly, the number of the second plurality of input transistors M_i21-M_i2k in the second input network 14 determines how quickly the supply voltage (VDD) is passed from the second input control node 24 to the second activation signal node 26 and thus the second differential input node 30. As an example, assume that the first plurality of input transistors M_i11-M_i1k includes a larger number of input transistors in the ON state than the second plurality of input transistors M_i21-M_i2k. Accordingly, the supply voltage (VDD) from the first input network 12 will reach the first differential input node 28 before the supply voltage (VDD) from the second input network 14 will reach the second differential input node 30. The supply voltage (VDD) as it is delivered from the first input network 12 and the second input network 14 is referred to herein as an “activation signal.”

As the supply voltage (VDD) from the first input network 12 reaches the first differential input node 28, the first amplifier transistor M_a1 is turned ON, the second amplifier transistor M_a2 is turned ON, the third amplifier transistor M_a3 is turned OFF, and the fourth amplifier transistor M_a4 is turned OFF. As a result, the first differential output node 32 is discharged through the first amplifier transistor M_a1 and the second amplifier transistor M_a2, which in turn causes the sixth amplifier transistor M_a6 to turn OFF and the seventh amplifier transistor M_a7 to turn ON. Accordingly, even if the supply voltage (VDD) from the second input network 14 reaches the second differential input node 30 at a later time, causing the second differential output node to begin to discharge, its further discharge is quickly impeded as the sixth amplifier transistor M_a6 turns OFF and the seventh amplifier transistor M_a7 turns ON and the second differential output node 34 is pulled back to the supply voltage (VDD). Accordingly, the differential amplifier 16 produces a 0 (ground) at the first differential output node 32 and a 1 (supply voltage VDD) at the second differential output node 34. As discussed above, the latch 18 may be an SR NAND latch, and therefore produces 1 at the latching output (Q).

In the event that the second plurality of input transistors M_i21-M_i2k includes a larger number of input transistors in the ON state than the first plurality of input transistors M_i11-M_i1k, the result is the opposite, wherein the differential amplifier 16 produces a 1 (supply voltage VDD) at the first differential output node 32 and a 0 (ground) at the second differential output node 34. Accordingly, the latch 18 produces a 0 at the latching output (Q).

As discussed above, a threshold function is one that computes a weighted sum for a number of variables x₁, x₂, . . . , x_n and their associated weights w₁, w₂, . . . , w_i, and compares the weighted sum to a predetermined threshold T. The variables x₁, x₂, . . . , x_n of a threshold function are represented in the DTG 10 by the first plurality of input signals IN₁₁-IN_1k provided to the first input network 12, while their associated weights w₁, w₂, . . . , w_i are represented by the number of different input transistors to which a particular input signal is provided. For example, if a given variable has an associated weight of two, the input signal that is representative of the variable will be provided to two different input transistors in the first plurality of input transistors M_i11M_i1k. Similarly, the predetermined threshold T is determined by the second plurality of input signals IN₂₁-IN_2k provided to the second plurality of input transistors M_i21-M_i2k in the second input network 14. By providing various input signals IN to the input transistors in the first input network 12 and the second input network 14, any number of threshold functions may be evaluated by the DTG 10.

There are many different ways to map signals to the DTG 10 while satisfying the constraint that the first input network 12 and the second input network 14 do not include the same number of active input transistors. The different signal assignments will result in different delay, power, and robustness to process variations in the various transistors in the DTG 10. In some embodiments, 2k inputs may be too many, as it may result in unnecessary congestion in larger circuitry in which the DTG 10 is integrated. Accordingly, the first input network 12 and the second input network 14 may be connected as shown in FIG. 2 such that the DTG 10 is configured to evaluate a majority function. The DTG 10 shown in FIG. 2 is substantially similar to that shown in FIG. 1, but further includes connections between the first plurality of input transistors M_i11-M_i1k in the first input network 12 and the second plurality of input transistors M_i21-M_i2k in the second input network 14. Further, an inverter 36A-36K is coupled between each one of the input transistors in the first input network 12 and the second input network 14. The result is a number of unified input nodes 38A-38K, which are connected to both the first input network 12 and the second input network 14. Connecting the first input network 12 and the second input network 14 as shown reduces the number of inputs to the DTG 10 by a factor of 2, resulting in a DTG 10 that is configured to evaluate a majority function.

As defined herein, a m out of n majority function is a Boolean function that produces a 1 if at least m of the input signals are high signals. The DTG shown in FIG. 2 implements a (k+1)/2 out of k majority function. Specifically, whenever n of the first plurality of input transistors M_i11-M_i1k are turned ON, k−n of the second plurality of input transistors M_i21-M_i2k are turned ON. To ensure that the number of active input transistors in the first input network 12 and the second input network 14 are not equal, k must be kept odd. Whenever at least (k+1)/2 of the input signals provided to the unified input nodes 38A-38K are high signals, the first input network 12 will have a larger number of active input transistors than the second input network 14, resulting in the latching output (Q) of the DTG 10 evaluating to a 1. In contrast, when less than (k+1)/2 of the input signals provided to the unified input nodes 38A-38K are high signals, the first input network 12 will have a smaller number of active input transistors than the second input network 14, resulting in the latching output (Q) of the DTG 10 evaluating to a 0.

Any threshold function f=[w₁, w₂, . . . , w_n|T] can be implemented using a T out of W majority function, where W=Σw_i. For example consider a threshold function f (a, b, c)=aVbc=[2,1,1; 2], where w_a=2, w_b=w_c=1 and T=2. Here T=2 and W=4. Given a 2 out of 4 majority function q (p, q, r, s )=[1,1,1,1; 2], f can be realized by simply connecting signal a to the inputs p and and q, signal b to r, and signal c to the input s of function g. The 2 out of 4 majority is also realizable using a 3 out of 5 majority by simply setting one of its inputs to 1. Accordingly, any majority function has a fixed subset of threshold functions it can implement under all possible assignments of signals to its inputs.

Consider implementing a 4-input AND gate f (a, b, c, d) using a 4 out of 7 majority q (p, q, r, s, t, u, v). The assignment p=a, q=b, r=c, s=d, t=u=v=0 realizes this function. Only when all of a, b, c and d are high signals is the 4 out of 7 majority reached, resulting in a high signal output. As illustrated by the examples above, a given DTG 10 is configured as shown in FIG. 2 and is capable of evaluating a number of Boolean operations by simply routing input signals (including constants) to the various unified input nodes 38A-38K. For example, a DTG 10 including seven input transistors in each one of the first input network 12 and the second input network 14 (and therefore seven unified input nodes 38A-40F) is capable of evaluating at least 30 Boolean expressions, some with up to 35 terms.

One advantage of the DTG 10 configuration shown in FIG. 2 is that it simplifies the design and testing of the DTG 10. In particular, the DTG 10 shown in FIG. 2 includes a worst-case performance scenario for both delay and robustness, which results in significant improvements in the ease of the design and testing of the DTG 10. Specifically, the maximum delay of the DTG 10 occurs for |m−n|=1, which is also the most likely input configuration for which the DTG 10 will fail due to intolerance to manufacturing variations. Accordingly, the design of the DTG 10 must only be done to optimize performance and robustness for a single input configuration at all process corners while accounting for process variations. For example, the delay and robustness of a DTG-7 need only be optimized for an input configuration of 4/3. If such a DTG evaluates correctly for this configuration under all conditions, then it will evaluate correctly for all other input configurations.

In addition to the advantages discussed above, the DTG 10 shown in FIG. 2 further includes the advantage that the probability of a correct computation performed by the DTG 10 increases with temperature. This property is herein referred to as temperature monotonicity. Temperature monotonicity ensures that if a particular DTG properly functions at a given temperature, it will continue to function properly at temperatures above the given temperature. This results in significant time and cost savings in the testing of simulated and manufactured DTG cells. The explanation for this behavior is simple. Suppose there are n input transistors ON in the first input network 12 and n−1 input transistors ON in the second input network 14. Let

$Z_L\approx \frac{1}{n}\mathrm{and}Z_R\approx \frac{1}{\left(n-1\right)}.$

The greater the difference

$Z_L-Z_R=\frac{1}{n\left(n-1\right)},$

the more likely that the DTG 10 will compute the correct value. As n increases, each individual impedance decreases, as does their difference (quadratically). The increase in temperature counteracts this effect by increasing the impedance of the first input network 12 and the second input network 14. Additionally, the temperature effect is nonlinear. Thus, the increase of the impedance of the first input network 12 and the second input network 14 need not be the same.

As discussed above, the DTG 10 shown in FIG. 2 is capable of evaluating one or more Boolean expressions simply by routing particular input signals to the unified input nodes 38A-38K of the DTG 10. Accordingly, the DTG 10 shown in FIG. 2 is particularly well suited for use in programmable logic circuitry. FIG. 3 shows a field programmable threshold logic array (FPTLA) 42 according to one embodiment of the present disclosure. The FPTLA 40 includes a plurality of DTG blocks (DTGBs) 42 and a plurality of programmable interconnect elements 44, each coupled between at least two of the plurality of DTGBs 42. As shown in the exploded view of one of the DTGBs 42, each DTGB 42 includes a DTG 10 (such as that shown in FIG. 2), a plurality of XOR gates 46, and a plurality of static random access memory (SRAM) cells 48. The XOR gates 46 together with the SRAM cells 48 allow the DTG 10 to receive a particular input signal or a complement thereof.

The programmable interconnect elements 44 include connection boxes (C) and switch boxes (S). The connection boxes (C) connect the channel wires to the input and output pins of the DTGBs 42. The switch boxes (S) allow wires to switch between vertical and horizontal wires. The purpose of the resulting routing grid is to connect appropriate signals to the inputs of each DTGB 42 including the required constants.

As discussed above, each one of the DTGs 10 is capable of computing a large set of Boolean expressions based on the input signals provided to the DTG 10. Accordingly, by appropriately routing signals between the DTGBs 42 in the FPTLA 40, the FPTLA 40 can compute complex Boolean expressions. Each DTGB 42 in the FPTLA 40 includes a significantly smaller number of components such as transistors and SRAM cells when compared to conventional configurable logic blocks (CLBs). Accordingly, each one of the DTGBs 42 is smaller, faster, and more efficient than its conventional counterpart. Additionally, due to the lack of intervening logic between DTGBs 42 in the FPTLA 40, the circuit is what is herein referred to as “nano-pipelined”. In other words, each and every one of the DTGBs 42 is clocked such that there are no un-clocked elements between the various DTGBs 42. This nano-pipelining results in further performance improvements when compared to conventional logic circuitry equivalents.

While the DTGBs 42 in the FPTLA 40 are a vast departure from conventional logic circuitry technology, the programmable interconnect elements 44 remain similar to existing logic circuitry such as field programmable gate arrays (FPGAs). Accordingly, the FPTLA 40 may be easily programmed according to current design paradigms. FIG. 4 is a flow diagram illustrating a method of programming the FPTLA 40 according to one embodiment of the present disclosure. First, a Boolean function for which evaluation is desired is received (step 100). Notably, the Boolean expression is developed according to current design paradigms, and thus each operation in the Boolean function includes only two inputs and a single output (each operation in the Boolean function represents an AND, OR, NOT, etc. logic gate). As discussed above, a Boolean function may be re-characterized as one or more threshold functions. Accordingly, the Boolean function is converted to a number of threshold functions, one for each operation within the Boolean function (step 102). Since each operation in the Boolean function has only two inputs and a single output, the resulting threshold functions will be highly inefficient if individually mapped onto the FPTLA 40. In other words, since the DTGBs 42 in the FPTLA 40 include more than two inputs, using only two of the inputs for each DTGB 42 will result in the use of an unnecessary number of the DTGBs 42. Thus, an area minimization method is performed in order to combine as many of the threshold functions as possible into a minimized threshold function (step 104). Well known techniques such as node elimination may be used to combine the threshold expressions. In particular, depending on the number of inputs available for each DTGB 42, a set of threshold functions whose combination is capable of evaluation by one of the DTGBs 42 may be combined in order to reduce the number of DTGBs 42 used for evaluation of the Boolean expression.

Since the FPTLA 40 is a nano-pipelined circuit, as discussed above, each path from a primary input to any primary output must have the same number of clocked elements. The minimized threshold expression may include computation paths of different lengths, thereby necessitating the insertion of timing buffers in order to ensure that each path from a primary input to a primary output has the same number of clocked elements (step 106). In one embodiment, a single DTGB 44 is used as a timing buffer. It should be noted that the number of timing buffers should be as small as possible since timing buffers do not compute to useful computation. A variety of techniques can be used to determine the minimum possible number of timing buffers necessary to ensure proper functionality of the FPTLA 40, all of which are contemplated herein. Finally, the minimized threshold expression including timing buffers can be mapped onto the FPTLA 40 by appropriately routing connections between each one of the DTGBs 42 (step 108). Using the method described above, conventional design paradigms may be used to program the FPTLA 40 without significant effort on behalf of the designer or programmer. Accordingly, the performance enhancements in size, speed, and efficiency may be obtained with little to no cost or effort.

As discussed above, DTGs 10 may be sensitive to manufacturing tolerances, which are often experienced by the transistor elements that make up the DTG 10. Accordingly, one or more of the DTGs 10 in the DTGBs 42 may fail to properly evaluate the intended majority function. However, due to the nature of threshold functions, if a particular DTG 10 does not implement an intended majority function, it can still be shown that the DTG 10 implements some threshold function. Consider the inequality Σ_iW_ix_i≧T. The manufacturing variations discussed above simply perturb the value of the weights represented by the conductance of each of the transistors in the first input network 12 and the second input network 14. Regardless of the perturbation in weights from normal, however, the resultant expression is still an arithmetic inequality and therefore represents a threshold function. Accordingly, in a field programmable environment such as the FPTLA 40, a set of functions that each DTGB 42 can be determined after the FPTLA 40 has been fabricated, and a particular DTGB 42 can be noted as capable only of evaluating certain threshold functions. When programming the FPTLA 40, such DTGBs 42 will be assigned only those threshold functions that they are capable of evaluating, thereby allowing DTGBs 42 that may otherwise be bypassed to be used and reducing the required redundancy measures of the FPTLA 40.

Alternatively, if a particular DTGB 42 is determined to improperly evaluate the intended majority function, the DTGB 42 may be noted and not used by the FPTLA 40. Instead, redundant DTGBs 42 may be used in place of the defective DTGB 42. The amount of DTGBs 42 that should be included in a particular FPTLA 40 is denoted in Equation 2:

$\begin{array}{cc}Y=\sum _{k=0}^m\left(\begin{array}{c}N+m\\ k\end{array}\right){P^{N+m-k}\left(1-P\right)}^k& \left(2\right)\end{array}$

Where N is the number of DTGBs 42 in the FPTLA 40, m is the number of additional cells required to ensure a probability P that a particular DTGB 42 functions correctly for a given probability of success (yield) Y. As an example, the number of additional DTGBs 42 to ensure a yield Y≧0.999 for a 128-bit comparator including DTGBs 42 configured to evaluate a 4/7 majority function is only 60 cells (with the base size of the total grid being 961). In other words, only 6% more DTGBs 42 are required to ensure a 99.9% yield for a 128-bit comparator using DTGBs 42 configured to evaluate a 4/7 majority function. Accordingly, the cost of redundancy measures on FPTLAs 40 may be reduced when compared to conventional logic circuitry.

Those skilled in the art will recognize improvements and modifications to the embodiments of the present disclosure. All such improvements and modifications are considered within the scope of the concepts disclosed herein and the claims that follow.

Claims

1. A field programmable threshold-logic array (FPTLA) comprising: a plurality of threshold logic gates;a plurality of programmable interconnect elements each coupled between at least two of the plurality of threshold logic gates and configured to route signals between the plurality of threshold logic gates.

2. The FPTLA of claim 1 wherein each one of the plurality of threshold logic gates is a differential threshold logic gate.

3. The FPTLA of claim 2 wherein each one of the plurality of threshold logic gates is configured to evaluate a majority function.

4. The FPTLA of claim 1 wherein each one of the plurality of threshold logic gates comprises: a first input network configured to receive a first plurality of input signals and generate a first activation signal at a speed proportional to a combined value of the first plurality of input signals;a second input network configured to receive a second plurality of input signals and generate a second activation signal at a speed proportional to a combined value of the second plurality of input signals;a differential amplifier coupled between the first input network and the second input network and configured to: receive the first activation signal and the second activation signal;generate a differential output signal having a first state if the first activation signal is received before the second activation signal; andgenerate a differential output signal having a second state if the second activation signal is received before the first activation signal; anda latch configured to receive the differential output signal from the differential amplifier and generate a latched output signal based on the state of the differential output signal.

5. The FPTLA of claim 4 wherein each one of the first plurality of input signals is equal to a complement of a corresponding one of the second plurality of input signals.

6. The FPTLA of claim 4 wherein: the first input network comprises a first plurality of transistors coupled in parallel between a first input control node and a first activation signal node; andthe second input network comprises a second plurality of transistors coupled in parallel between a second input control node and a second activation signal node.

7. The FPTLA of claim 6 wherein the first plurality of transistors and the second plurality of transistors are metal-oxide-semiconductor field-effect-transistors (MOSFETs).

8. The FPTLA of claim 6 wherein: in a reset state of operation of the first input network and the second input network, the first input control node and the second input control node are left floating and the first activation signal node and the second activation signal node are coupled to ground; andin a set state of operation of the first input network and the second input network, the first input control node and the second input control node are connected to a power supply voltage and the first activation signal node and the second activation signal node are left floating.

9. The FPTLA of claim 6 wherein the differential amplifier comprises: a first differential input node, a second differential input node, a first differential output node, and a second differential output node;a first transistor including a source contact coupled to ground, a drain contact, and a gate contact coupled to the first differential input node;a second transistor including a source contact coupled to the drain contact of the first transistor, a drain contact coupled to the first differential output node, and a gate contact coupled to the second output node;a third transistor including a source contact coupled to a power supply voltage, a drain contact coupled to the first differential output node, and a gate contact coupled to the second differential output node;a fourth transistor including a source contact coupled to the power supply voltage, a drain contact coupled to the first differential output node, and a gate contact coupled to the first input node;a fifth transistor including a source contact coupled to ground, a drain contact, and a gate contact coupled to the second differential input node;a sixth transistor including a source contact coupled to the drain contact of the fifth transistor, a drain contact coupled to the second differential output node, and a gate contact coupled to the first differential output node;a seventh transistor including a source contact coupled to the power supply voltage, a drain contact coupled to the first differential output node, and a gate contact coupled to the first differential output node; andan eighth transistor including a source contact coupled to the power supply voltage, a drain contact coupled to the second differential output node, and a gate contact coupled to the second differential input node.

10. The FPTLA of claim 9 wherein the first transistor, the second transistor, the third transistor, the fourth transistor, the fifth transistor, the sixth transistor, the seventh transistor, and the eighth transistor are metal-oxide-semiconductor field-effect-transistors (MOSFETs).

11. The FPTLA of claim 9 wherein the latch is a set-reset (SR) NOR latch.

12. The FPTLA of claim 9 wherein: the first input network comprises a first plurality of transistors coupled in parallel between a first input control node and a first activation signal node; andthe second input network comprises a second plurality of transistors coupled in parallel between a second input control node and a second activation signal node.

13. The FPTLA of claim 12 wherein the first activation signal node is coupled to the first differential input node and the second activation signal node is coupled to the second differential input node.

14. The FPTLA of claim 13 wherein: in a reset state of operation of the first input network and the second input network, the first input control node and the second input control node are left floating and the first activation signal node and the second activation signal node are coupled to ground; andin a set state of operation of the first input network and the second input network, the first input control node and the second input control node are connected to a power supply voltage and the first activation signal node and the second activation signal node are left floating.

15. The FPTLA of claim 14 wherein the latch comprises: a set input node coupled to the first differential output node;a reset input node coupled to the second differential output node; andone or more output nodes.

16. The FPTLA of claim 15 wherein the node is a set-reset (SR) NOR latch.

17. A differential threshold logic gate (DTG) comprising: a first input network configured to receive a first plurality of input signals and generate a first activation signal at a speed proportional to a combined value of the first plurality of input signals;a second input network configured to receive a second plurality of input signals and generate a second activation signal at a speed proportional to a combined value of the second plurality of input signals;a differential amplifier coupled between the first input network and the second input network and configured to: receive the first activation signal and the second activation signal;generate a differential output signal having a first state if the first activation signal is received before the second activation signal; andgenerate a differential output signal having a second state if the second activation signal is received before the first activation signal; anda latch configured to receive the differential output signal from the differential amplifier and generate a latched output signal based on the state of the differential output signal.

18. The DTG of claim 17 wherein each one of the first plurality of input signals is equal to a complement of a corresponding one of the second plurality of input signals.

19. The DTG of claim 17 wherein: the first input network comprises a first plurality of input transistors coupled in parallel between a first input control node and a first activation signal node; andthe second input network comprises a second plurality of input transistors coupled in parallel between a second input control node and a second activation signal node.

20. The DTG of claim 19 wherein the first plurality of transistors and the second plurality of transistors are metal-oxide-semiconductor field-effect-transistors (MOSFETs).

21. The DTG of claim 19 wherein: in a reset state of operation of the first input network and the second input network, the first input control node and the second input control node are left floating and the first activation signal node and the second activation signal node are coupled to ground; andin a set state of operation of the first input network and the second input network, the first input control node and the second input control node are connected to a power supply voltage and the first activation signal node and the second activation signal node are left floating.

22. The DTG of claim 17 wherein the differential amplifier comprises: a first differential input node, a second differential input node, a first differential output node, and a second differential output node;a first transistor including a source contact coupled to ground, a drain contact, and a gate contact coupled to the first differential input node;a second transistor including a source contact coupled to the drain contact of the first transistor, a drain contact coupled to the first differential output node, and a gate contact coupled to the second differential output node;a third transistor including a source contact coupled to a power supply voltage, a drain contact coupled to the first differential output node, and a gate contact coupled to the second differential output node;a fourth transistor including a source contact coupled to the power supply voltage, a drain contact coupled to the first differential output node, and a gate contact coupled to the first differential input node;a fifth transistor including a source contact coupled to ground, a drain contact, and a gate contact coupled to the second differential input node;a sixth transistor including a source contact coupled to the drain contact of the fifth transistor, a drain contact coupled to the second differential output node, and a gate contact coupled to the first differential output node;a seventh transistor including a source contact coupled to the power supply voltage, a drain contact coupled to the second differential output node, and a gate contact coupled to the first differential output node; andan eighth transistor including a source contact coupled to the power supply voltage, a drain contact coupled to the second differential output node, and a gate contact coupled to the second differential input node.

23. The DTG of claim 22 wherein the first transistor, the second transistor, the third transistor, the fourth transistor, the fifth transistor, the sixth transistor, the seventh transistor, and the eighth transistor are metal-oxide-semiconductor field-effect-transistors (MOSFETs).

24. The DTG of claim 22 wherein the latch is a set-reset (SR) NOR latch.

25. The DTG of claim 22 wherein: the first input network comprises a first plurality of transistors coupled in parallel between a first input control node and a first activation signal node; andthe second input network comprises a second plurality of transistors coupled in parallel between a second input control node and a second activation signal node.

26. The DTG of claim 25 wherein the first activation signal node is coupled to the first differential input node and the second activation signal node is coupled to the second differential input node.

27. The DTG of claim 26 wherein: in a reset state of operation of the first input network and the second input network, the first input control node and the second input control node are left floating and the first activation signal node and the second activation signal node are coupled to ground; andin a set state of operation of the first input network and the second input network, the first input control node and the second input control node are connected to a power supply voltage and the first activation signal node and the second activation signal node are left floating.

28. The DTG of claim 27 wherein the latch comprises: a set input node coupled to the first differential output node;a reset input node coupled to the second differential output node; andone or more output nodes.

29. The DTG of claim 28 wherein the latch is a set-reset (SR) NOR latch.