Abstract: A novel semiconductor device is provided. The semiconductor device includes a driver circuit including a plurality of transistors using a silicon substrate for channels, and a first transistor layer and a second transistor layer including a plurality of transistors using a metal oxide for channels. The first transistor layer and the second transistor layer are provided over the silicon substrate layer. The first transistor layer includes a first memory cell including a first transistor and a first capacitor. The first transistor is electrically connected to a first local bit line. The second transistor layer includes a second transistor whose gate is electrically connected to the first local bit line and a first correction circuit electrically connected to the second transistor. The first correction circuit is electrically connected to a first global bit line.

Abstract: An optimization device includes: a memory; and a processor and configured to: store a coefficient indicating magnitude of an interaction between bits in a bit string representing a state of an Ising model; output, when any bit in the bit string is inverted, a signal indicating inversion availability of an own bit according to calculation of energy change in the Ising model using the coefficient corresponding to the inverted bit and the own bit read from the memory as bit operations; output a signal indicating a bit to be inverted in the bit string selected on the basis of the signal indicating inversion availability output from bit operations of a first number of bits of the bit string, of the bit operations; and change the first number of bits and change a second number of bits of the coefficient for each bit operations of the first number of bits.

Abstract: A quantum phased array comprising one or more arrays of emitter elements each emitting one or more particles having one or more quantum wavefunctions; one or more a phase shifting elements coupled to the emitter elements, each of the phase shifting elements comprising a source of a vector potential applying one or more phase shifts to the one or more quantum wavefunctions; and a control circuit coupled to the one or more phase shifting elements, the control circuit configuring the one or more vector potentials to control an interference of the quantum wavefunctions forming a distribution of the one or more particles at a target, and wherein the distribution is described by a wavefunction interference pattern resulting from the interference controlled by the vector potentials.

Abstract: Provided is a multiplier-accumulator (MAC) system, circuit, and method. The MAC system includes a MAC circuit, including a plurality of resistors, having respective resistances, a capacitor connected to the plurality of resistors to charge, in response to a plurality of input signals, the capacitor with electric charge, and a time-to-digital converter (TDC) configured to convert information of a charge time of the capacitor, due to the electric charge, into a digital value, wherein the digital value is an accumulation result of the MAC circuit.

Abstract: A sum-of-products operator including: a first circuit configured to generate a plurality of signals, each of which corresponds to each of a plurality of data; a second circuit including a first operation circuit configured to multiply each of the signals generated by the first circuit by a weight using a plurality of variable resistive elements having variable resistance values, and to calculate a sum of a plurality of results of multiplications; a third circuit configured to calculate a result of summing values corresponding to the data or a result of the summing value after being adjusted; and a fourth circuit including a differential circuit configured to output a difference between a calculated result in the first operation circuit of the second circuit and a calculated result in the third circuit.

Abstract: An arithmetic operation method for a convolutional layer in a neural network includes: generating a coefficient matrix by converting a kernel used in the convolutional layer such that the coefficient matrix is associated with an input vector obtained by expanding, into one column, a feature map input to the convolutional layer; searching for non-zero elements included in the coefficient matrix; assigning multiplications of the non-zero elements included in the coefficient matrix and corresponding elements of the input vector to a plurality of calculators with each of the multiplications being handled as a unit of process, so as to level out the numbers of units of process among the calculators, each of the calculators being capable of performing a process in parallel with one another; and sequentially performing, by the calculators, the assigned multiplications, and sequentially adding, by the calculators, results of the multiplications to corresponding elements of an output vector.

Abstract: The present disclosure relates generally to techniques for improving the implementation of certain operations on an integrated circuit. In particular, deep learning techniques, which may use a deep neural network (DNN) topology, may be implemented more efficiently using low-precision weights and activation values by efficiently performing down conversion of data to a lower precision and by preventing data overflow during suitable computations. Further, by more efficiently mapping multipliers to programmable logic on the integrated circuit device, the resources used by the DNN topology to perform, for example, inference tasks may be reduced, resulting in improved integrated circuit operating speeds.

Abstract: An arithmetic apparatus according to an embodiment outputs a multiplicative value obtained by multiplying a weight value and an input value. The arithmetic apparatus includes a memristor, a logarithmic transform circuit, and a current-voltage converter circuit. The memristor is a device capable of changing voltage-current characteristic, and the memristor is preset to voltage-current characteristic according to the weight value. The logarithmic transform circuit applies an intermediate voltage, to the memristor, that is obtained by logarithmically transforming an input voltage according to the input value in accordance with a logarithmic transform function obtained by multiplying a natural logarithm function by a preset coefficient. The current-voltage converter circuit outputs an output voltage obtained by performing current-voltage conversion of current flowing through the memristor according to a preset linear function, as a multiplicative value.

Abstract: A general matrix-matrix multiplication (GEMM) dataflow accelerator circuit is disclosed that includes a smart 3D stacking DRAM architecture. The accelerator circuit includes a memory bank, a peripheral lookup table stored in the memory bank, and a first vector buffer to store a first vector that is used as a row address into the lookup table. The circuit includes a second vector buffer to store a second vector that is used as a column address into the lookup table, and lookup table buffers to receive and store lookup table entries from the lookup table. The circuit further includes adders to sum the first product and a second product, and an output buffer to store the sum. The lookup table buffers determine a product of the first vector and the second vector without performing a multiply operation. The embodiments include a hierarchical lookup architecture to reduce latency. Accumulation results are propagated in a systolic manner.

Abstract: Examples of the present disclosure provide apparatuses and methods for performing multiplication operations in a memory. An example method comprises performing a multiplication operation on a first element stored in a group of memory cells coupled to a first access line and a number of sense lines of a memory array and a second element stored in a group of memory cells coupled to a second access line and the number of sense lines of the memory array. The method can include a number operations performed without transferring data via an input/output (I/O) line.

Abstract: Examples of the present disclosure provide apparatuses and methods for performing multi-variable bit-length multiplication operations in a memory. An example method comprises performing a multiplication operation on a first vector and a second vector. The first vector includes a number of first elements stored in a group of memory cells coupled to a first access line and a number of sense lines of a memory array. The second vector includes a number of second elements stored in a group of memory cells coupled to a second access line and the number of sense lines of the memory array. The example multiplication operation can include performing a number of AND operations, OR operations and SHIFT operations without transferring data via an input/output (I/O) line.

Abstract: Examples of the present disclosure provide apparatuses and methods for performing signed division operations. An apparatus can include a first group of memory cells coupled to a first access line and a number of sense lines. The apparatus can include a second group of memory cells coupled to a second access line and the number of sense lines. The apparatus can include a controller configured to cause sensing circuitry to divide a signed dividend element stored in the first group of memory cells by a signed divisor element stored in the second group of memory cells by performing a number of operations. At least one of the number of operations can be performed without transferring data via an input/output (I/O) line.

Abstract: Examples of the present disclosure provide apparatuses and methods for performing multiplication operations in a memory. An example method comprises performing a multiplication operation on a first element stored in a group of memory cells coupled to a first access line and a number of sense lines of a memory array and a second element stored in a group of memory cells coupled to a second access line and the number of sense lines of the memory array. The method can include a number operations performed without transferring data via an input/output (I/O) line.

Abstract: Examples of the present disclosure provide apparatuses and methods for performing multi-variable bit-length multiplication operations in a memory. An example method comprises performing a multiplication operation on a first vector and a second vector. The first vector includes a number of first elements stored in a group of memory cells coupled to a first access line and a number of sense lines of a memory array. The second vector includes a number of second elements stored in a group of memory cells coupled to a second access line and the number of sense lines of the memory array. The example multiplication operation can include performing a number of AND operations, OR operations and SHIFT operations without transferring data via an input/output (I/O) line.

Abstract: Provided are a four-dimensional non-equilibrium hyperchaotic system and analog circuit, based on the five simplest three-dimensional chaotic systems; an operational amplifier (U1), an operational amplifier (U2), and resistor and capacitor are used to constitute an inverting adder and an inverting integrator; multipliers (U3) and (U4) are used to perform multiplication operations; an 8V DC power supply is used for constant input; the operational amplifier (U1) and operational amplifier (U2) use LF347N, and the multipliers (U3) and (U4) use AD633JN; the operational amplifier (U1) is connected to the operational amplifier (U2) and the multiplier (U3); the operational amplifier (U2) is connected to the multiplier (U4), the DC power supply, and the operational amplifier (U1); the multiplier (U3) is connected to the operational amplifier (U1); the multiplier (U4) is connected to the operational amplifier (U2); the DC power supply is connected to the operational amplifier (U2); on the basis of the five simplest thre

Abstract: A random number processing device according to an aspect of the present disclosure is a random number processing device generating random number data by using data read from memory cells, the memory cells having a property such that, in a variable state, in response to application of different electrical signals, a resistance value of each of the memory cells reversibly transitions between resistance value ranges and, when the resistance value falls within at least one resistance value range among the resistance value ranges, the resistance value changes as time passes, the random number processing device including a random number processing circuit that, in operation, generates first random number data from a combination of first resistance value information and second resistance value information about the resistance values of first and second memory cells among the memory cells which fall within the at least one resistance value range.

Abstract: A motor control apparatus includes a by-wire control circuit for sequentially switching an energized phase of a motor. The by-wire control circuit pre-stores a first table defining an energized phase address corresponding to each address and a second table defining an energized phase corresponding to each energized phase address. When receiving a drive permission code from a second control circuit, the by-wire control circuit switches the energized phase in a correct order of driving the motor, by calculating an address for access to the first table based on the drive permission code, calculating an energized phase address corresponding to the address by referring to the first table, and determining the energized phase corresponding to the energized phase address by referring to the second table.

Abstract: Disclosed herein are systems and methods for a signed-magnitude adder based on one's complement logic, where the adder offers enhancements in both speed and chip area consumption. The one's complement based adder includes circuitry for converting operands from their signed-magnitude representations to their one's complement representations, circuitry for adding operands in their one's complement representations, and circuitry for converting the resulting sum into a signed-magnitude format.

Abstract: A display apparatus includes a display panel including a plurality of pixel columns to display an image, wherein each of the pixel columns includes a plurality of pixels arranged in a first direction and sequentially turned-on in the first direction; a gate driver disposed on the display panel and including a plurality of stages connected to the pixels to sequentially apply a gate signal to the pixels, where at least two stages of the stages are disposed adjacent to each other in a second direction different from the first direction; and a data driver which applies a data voltage to the pixels.

Abstract: The invention provides a logic device, e.g. a Boolean logic gate, comprising: a memristor having: an input for receiving a sequence of voltage states which represent a sequence of logical inputs for a logical operation, and an output for an output current response triggered by the sequence of voltage states; and a determining unit arranged to determine an output of the logical operation from the output current response.

Abstract: Methods for reconfiguring an ASIC at runtime without using voltage over scaling. A functional criticality of a set of logic in the ASIC is identified. Then, the set of logic are classified into a set of regions based on the functional criticality, each region of the set of regions having a target error threshold. Further, each region is power gated at runtime based on the functional criticality such that the target error threshold is achieved without using voltage over scaling.

Abstract: An integrated circuit die stack comprises a first die coupled with a second die. The first die has a first memory volume. The second die has a second memory volume different from the first memory volume. Each of the first and second dies comprises a functional circuitry and a programmable array coupled with the functional circuitry. The programmable arrays in the first and second dies are programmed to bypass one of the first die or the second die having the smaller of the first memory volume or the second memory volume at a first time period.

Abstract: Techniques are generally described for generating an identification number for an integrated circuit (IC). In some examples, methods for generating an identification for an IC may comprise selecting circuit elements of the IC, evaluating measurements of an attribute of the IC for the selected circuit elements, wherein individual measurements are associated with corresponding input vectors previously applied to the IC, solving a plurality of equations formulated based at least in part on the measurements taken of the attribute of the IC for the selected circuit elements to determine scaling factors for the selected circuit elements, and transforming the determined scaling factors for the selected circuit elements to generate an identification number of the IC. Additional variants and embodiments may also be disclosed.

Abstract: A method is provided for a synthesising In RTL, a logic circuit and for manufacturing an integrated circuit for performing a sum of addends with faithful rounding. In this, optimisation constraints for a value of bits which may be discarded and a constant to include in a sum of addends are determined (20). Next, the maximum number of whole columns that can be removed from the sum of addends array is derived (22) and those columns are discarded (24). Next, a number of bits which can be removed from the least significant column is derived (26) and these bits are discarded (28). The constant is included in the sum of addends and a logic array synthesised in RTL (31) before manufacturing an integrated circuit.

Abstract: A system for a conjugate gradient iterative linear solver that calculates the solution to a matrix equation comprises a plurality of gamma processing elements, a plurality of direction vector processing elements, a plurality of x-vector processing elements, an alpha processing element, and a beta processing element. The gamma processing elements may receive an A-matrix and a direction vector, and may calculate a q-vector and a gamma scalar. The direction vector processing elements may receive a beta scalar and a residual vector, and may calculate the direction vector. The x-vector processing elements may receive an alpha scalar, the direction vector, and the q-vector, and may calculate an x-vector and the residual vector. The alpha processing element may receive the gamma scalar and a delta scalar, and may calculate the alpha scalar. The beta processing element may receive the residual vector, and may calculate the delta scalar and the beta scalar.

Abstract: A circuit that provides a rotating coefficient FIR filter with all necessary coefficient sets present at the same time, without the need for delay elements or devices providing for adjustable impedances is described. An input signal is sampled in round robin fashion by a plurality of sample and hold devices. The outputs of the sample and hold devices are connected to sets of impedance devices. Each set of impedance devices implements the coefficients of the desired frequency response of the filter. The impedance devices in each set are connected to the sample and hold devices in a different order from each other set, so that each set of impedance devices will produce the desired frequency response when a different one of the sampling circuits contains a new sample of the input signal. Switches connect the sets of impedance devices to an output, only one switch being closed at a time to provide the output signal.

Abstract: A direct digital synthesis is provided with added circuitry to reduce jitter in an IC so that a programmable frequency output can be provided near the limits of the IC system clock with minimal jitter. The system derives the quotient Q as a remainder R in an accumulator at the instant of an overflow, divided by a programmable input N. The quotient Q is subjected to conversion logic that can be provided by a fast parallel to serial converter such as, for example a multi-gigabit transceiver (MGT) of an FPGA. As an alternative to an MGT, a series of delay devices such as found in a carry chain can be used if calibration is performed to assure the accuracy of delays.

Abstract: There is provided a semiconductor integrated circuit including: a plurality of first logic blocks which are reconfigurable, the plurality of first logic blocks inputting data of a first bit width and performing computation; a first network connecting the plurality of first logic blocks in a dynamically reconfigurable manner; a plurality of second logic blocks inputting data of a second bit width different from the first bit width and performing computation; a second network connected to outputs of the plurality of second logic blocks; and a third network connecting a carry bit output of a computing unit included in the first logic block to an input of a computing unit included in the second logic block in a dynamically reconfigurable manner.

Abstract: The invention reduces unnecessary electromagnetic radiation noise due to an operation clock signal generated by an oscillator circuit. Random number data outputted by a random number generation circuit is stored in a frequency variable data register. The data stored in the frequency variable data register is replaced by random number data sequentially generated by the random number generation circuit. An oscillator circuit is a circuit generating a clock signal, and the clock signal is supplied as an operation clock signal to an internal circuit through an operation clock signal generation circuit. The frequency of the clock signal from the oscillator circuit is variably controlled in response to the random number data stored in the frequency variable data register. A frequency variable range control register which stores control data for controlling the range of the frequency variably controlled in response to the random number data stored in the frequency variable data register is further provided.

Abstract: In one embodiment, the invention includes a method for compressing video data using redundant binary mathematics. Other embodiments are described and claimed.

Abstract: A semiconductor integrated circuit device, having a plurality of processing elements accommodated on a single semiconductor chip, has a latch circuit and a selecting circuit. The latch circuit is provided at an output of each of the processing elements. The selecting circuit selects an input source from a group consisting of upper, lower, left, and right processing elements and a zero signal.

Abstract: An output buffer is provided, to which first and second input signals are applied and that delivers an output signal. The output buffer includes a second offset switching stage installed in cascade downstream from a first switching stage. The second offset switching stage generates control points shifted in time with respect to memory points.

Abstract: A filter circuit includes: an adder/subtractor that performs at least addition; and a shifter that performs multiplication/division by a power of two through a shift operation. The adder/subtractors and the shifter are configured to obtain a first calculation result representing a pixel value of a target pixel included in image data multiplied by a first filter coefficient. At least the adder/subtractors and the shifter is configured to obtain a second calculation result representing pixel values of a plurality of peripheral pixels adjacent to the target pixel, with each of the pixel values being multiplied by a second filter coefficient. The adder/subtractor is configured obtain a third calculation result by adding the first and second calculation results. The shifter configured to divide the third calculation result by a power of two which is equivalent to a sum of the first and second filter coefficients, so as to output the division result.

Abstract: An encoder encodes data into a series of parallel codewords. Each codeword is expressed two sets of logic values (e.g., a set of logic 0s and a set of logic 1s) on output nodes. The encoder selects a current codeword from a group of codewords in a codespace which does not overlap the other group of codewords, i.e., codewords in a given group of codewords are not included in any other group of codewords in the codespace. This property allows a receiver of the codewords to be simplified. In particular, a mathematical operation performed on symbols in the current codeword uniquely specifies the corresponding group of codewords. This allows a decoder to decode the current codeword using comparisons of symbols received on a subset of all possible combinations of node pairs.

Abstract: There is provided a semiconductor integrated circuit including: a plurality of first logic blocks which are reconfigurable, the plurality of first logic blocks inputting data of a first bit width and performing computation; a first network connecting the plurality of first logic blocks in a dynamically reconfigurable manner; a plurality of second logic blocks inputting data of a second bit width different from the first bit width and performing computation; a second network connected to outputs of the plurality of second logic blocks; and a third network connecting a carry bit output of a computing unit included in the first logic block to an input of a computing unit included in the second logic block in a dynamically reconfigurable manner.

Abstract: The invention reduces unnecessary electromagnetic radiation noise due to an operation clock signal generated by an oscillator circuit. Random number data outputted by a random number generation circuit is stored in a frequency variable data register. The data stored in the frequency variable data register is replaced by random number data sequentially generated by the random number generation circuit. An oscillator circuit is a circuit generating a clock signal, and the clock signal is supplied as an operation clock signal to an internal circuit through an operation clock signal generation circuit. The frequency of the clock signal from the oscillator circuit is variably controlled in response to the random number data stored in the frequency variable data register. A frequency variable range control register which stores control data for controlling the range of the frequency variably controlled in response to the random number data stored in the frequency variable data register is further provided.

Abstract: A cryptography processor includes a central processing unit and a co-processor, the co-processor comprising a plurality of calculating subunits as well as a single control unit which is coupled to each of the plurality of calculating subunits. A cryptographic operation is distributed among the individual calculating subunits in the form of sub-operations by the control unit. The central processing unit, the plurality of calculating subunits and the control unit are integrated on a single chip, the chip comprising a common supply current access for supplying the plurality of calculating subunits and the control unit with current. Due to the arrangement of the calculating subunit in parallel, on the hand, the throughput of the cryptography processor is increased. On the other hand, however, the current profile that may be detected at the supply current access is randomised to such an extent that an attacker can no longer infer numbers processed in the individual calculating subunits.

Abstract: A method for reading out the state of a mesoscopic phase device. In the method the mesoscopic phase device is coherently coupled to a mesoscopic charge device using a phase shift device and the quantum state of the mesoscopic charge device is measured. A method for reading out the quantum state of a qubit in a heterogeneous quantum register. The heterogeneous quantum register includes a first plurality of phase qubits and a second plurality of charge qubits. In the method a first phase qubit or a first charge qubit in the heterogeneous quantum register is selected. The first phase qubit or the first charge qubit is coherently connected to a mesoscopic charge device for a duration tc. The quantum state of the mesoscopic charge device is read out after the duration tc has elapsed.

Abstract: A mass-production transfer support system has a mass-production transfer source managing computer for managing information generated in a trial-production process of a semiconductor device and a mass-production transfer destination managing computer for managing a mass-production process of the semiconductor device.

Abstract: A variable function calculator utilizes a fixed program memory array such as a programmed read only memory in which a number of programs are stored depending upon the desired functions of the calculator. The calculator also includes a program counter, an instruction register, control decoders, jump-condition circuits, a clock generator, a timing generator, digit and FLAG mask decoders, key input logic, a register and FLAG data storage array, a decimal and FLAG arithmetic logic unit, an output decoder, and a digit scanner which scans both the keyboard and display outputs. Aside from providing basic desk top calculator functions, the read only memory may be programmed so that the system provides metering functions, arithmetic teaching functions, control functions, etc.. A preferred embodiment of the invention is capable of being fabricated as a monolithic integrated semiconductor system utilizing contemporary metal-insulator-semiconductor techniques.

Abstract: A configurable code generator system (CGS) for spread spectrum applications is disclosed herein. The CGS includes a composite code generator unit (CGU), a global code generator, and an interface that is coupled to the composite code generator and the global code generator. The CGU has multiple independent code generators, each capable of generating an independent code sequence. The global code generator provides a global code sequence for synchronization. The interface has memory that stores at least one bit of the global sequence and at least one bit from at least one of the independent code sequences of the CGU from which an output conditioning circuit can selectively choose based on a desired communication protocol.

Abstract: A data processor has a multiple length arithmetic circuit set with control data by a CPU that decodes and executes instructions through a bus, which performs processing for multiple length data based on the set control data. The multiple length arithmetic circuit performs a multiple length data operation by repeating processing where it makes read access to multiple length data at every unit of processing for multiple bits, partially operates read data, makes write access to the partially operated result, and delivers arithmetic information needed for the next partial operation to the next partial operation. The multiple length arithmetic circuit is a bus master module performing addressing operations by itself. It only operates by receiving control data setting from the CPU. The CPU does not need to repeatedly execute data transfer and add-subtract instructions. Multiple length data operations required for the Elliptic Curve Cryptosystem can be executed faster.

Abstract: An inverse discrete cosine transforming (IDCT) system for a digital television receiver performing a variable length decoding and an inverse quantization of an applied bit stream and an IDCT with respect to inversely quantized DCT coefficients, including an integrated IDCT part which performs an inverse discrete cosine transformation of standards DCT coefficients inversely quantized or high definition DCT coefficients inversely quantized.

Abstract: To encode data Y after separating the data Y into a scale factor SF, a word length, and a data value X. For this purpose, X=Y·{2(WL−1)−1}/SF is calculated. Here, a value of {2(WL−1)−1}/SFV with respect to a number of SFV and a number of SL, which are obtained by separating into SF=SF·2SFF, is stored in a ROM in advance. Then, input data Y is separated into a mantissa part Yr and an index part Ye and Ye is added to SFF. Then, the shifter 16 shifts Yr according to an additional result to obtain Ye·2Ye·2SFF. On the other hand, based on SF and WL, which are determined with respect to Y, a value of corresponding {2(WL−1)−1}/SFV is read from the ROM. Then, the shifted result is multiplied by the output from the ROM to obtain a data value X. Also, Y=SF·X/{2(WL−1)−1} is calculated and decoded. Inputted SF is divided into SFV and SFF.

Abstract: A circuit arrangement with a data sequence generator for the generation of an oscillator signal as a sequence of digital data, and having the following circuit parts: a first multiplication unit driven by a first auxiliary signal and a first control signal, where the first control signal corresponds to the sine of a manipulated variable that determines the frequency of the oscillator signal; a second multiplication unit driven by a second auxiliary signal, where the second control signal corresponds to the cosine of the manipulated variable; a third multiplication unit driven by a third auxiliary signal and a third control signal, where the third control signal corresponds to the sum of the first and second control signal; a first adding unit driven by the first multiplication unit and third multiplication unit; a second adding unit driven by the first multiplication unit and the second multiplication unit; a first time-delay unit driven by the first adding unit, where the oscillator signal is provided as a t

Abstract: A method of manufacturing an integration circuit device that includes generating logic library data with respect to a macro that includes a predetermined macro core and boundary cells positioned near input and output terminals of the macro. The logic library data includes delay characteristic data of the boundary cells given as attribute data to the input and output terminals. A logic circuit, that includes a plurality of cells and the macro, is designed, with the cells being connected to the macro core through the boundary cells connected to the input and output terminals. A delay time of the macro is calculated based on the delay characteristic data with respect to the designed logic circuit. A logic simulation on the designed logic circuit is effected based on the calculated delay time.

Abstract: The present invention provides a data processing circuit and method for performing arithmetic processing on data signals input to the circuit, comprising: a plurality of input terminals for receiving a plurality of data signals to be processed; a plurality of interconnected arithmetic processing units, one corresponding to each input terminal, for processing the data signals received at the corresponding input terminal; and a selector for routing the data signals at said input terminals to the corresponding arithmetic processing units in a first mode of operation, or for routing a selected one of said data signals to said plurality of arithmetic processing units in a second mode of operation; whereby, in said first mode of operation, data signals arriving at said input terminals are processed in parallel by said corresponding arithmetic processing units, and, in said second mode of operation, at any point in time, one of said data signals is processed by said plurality of arithmetic processing units.