Abstract: A voltage droop mitigation system, that includes a first processor core that executes computer executable components stored in a memory. A time-based sensor component generates digital data representing voltage values associated with a power supply. A filtering component digitally conditions the generated digital data, and an analysis component analyzes the conditioned data and determines slope of the power supply voltage and employs counters to determine rate of data change over time; and if the slope is negative and exceeds a first pre-determined value for a pre-determined time period. The system implements one or more voltage droop-reduction techniques at the first processor core; and the first processor core transmits at least one of the following types of information: its voltage value, slope information or decision to apply droop reduction to one or more other cores.

Abstract: Disclosed herein includes a system, a method, and a device including shift circuitry and add circuitry for performing multiplication of a first value and a second value for a neural network. The first value has a predetermined format including a first bit, and two or more second bits to represent a value of zero or 2n where n is an integer greater than or equal to 0. The device shifts, when the two or more second bits represent the value of 2n, the second value by (n+1) bits via the shift circuitry to provide a first result, selectively outputs zero or the second value, based on a value of the first bit of the first value, to provide a second result, and adds the first result and the second results via the add circuitry to provide a result of the multiplication of the first and second values.

Abstract: Disclosed herein includes a system, a method, and a device including shift circuitry and add circuitry for performing multiplication of a first value and a second value for a neural network. The first value has a predetermined format including a first bit, and two or more second bits to represent a value of zero or 2n where n is an integer greater than or equal to 0. The device shifts, when the two or more second bits represent the value of 2n, the second value by (n+1) bits via the shift circuitry to provide a first result, selectively outputs zero or the second value, based on a value of the first bit of the first value, to provide a second result, and adds the first result and the second results via the add circuitry to provide a result of the multiplication of the first and second values.

Abstract: Disclosed herein includes a system, a method, and a device for receiving input data to generate a plurality of outputs for a layer of a neural network. The plurality of outputs are arranged in a first array. Dimensions of the first array may be compared with dimensions of a processing unit (PE) array including a plurality of PEs. According to a result of the comparing, the first array is partitioned into subarrays by the processor. Each of the subarrays has dimensions less than or equal to the dimensions of the PE array. A first group of PEs in the PE array is assigned to a first one of the subarrays. A corresponding output of the plurality of outputs is generated using a portion of the input data by each PE of the first group of PEs assigned to the first one of the subarrays.

Abstract: Disclosed herein includes a system, a method, and a device for reading and writing sparse data in a neural network accelerator. A plurality of slices can be established to access a memory having an access size of a data word. A first slice can be configured to access a first side of the data word in memory. Circuitry can access a mask identifying byte positions within the data word having non-zero values. The circuitry can modify the data word to have non-zero byte values stored starting at an end of the first side, and any zero byte values stored in a remainder of the data word. A determination can be made whether a number of non-zero byte values is less than or equal to a first access size of the first slice. The circuitry can write the modified data word to the memory via at least the first slice.

Abstract: Disclosed herein includes a system, a method, and a device for early-exit from convolution. In some embodiments, at least one processing element (PE) circuit is configured to perform, for a node of a neural network corresponding to a dot-product operation with a set of operands, computation using a subset of the set of operands to generate a dot-product value of the subset of the set of operands. The at least one PE circuit can compare the dot-product value of the subset of the set of operands, to a threshold value. The at least one PE circuit can determine whether to activate the node of the neural network, based at least on a result of the comparing.

Abstract: Disclosed herein includes a system, a method, and a device for asymmetrical scaling factor support for negative and positive values. A device can include a circuit having a shift circuitry and multiply circuitry. The circuit can be configured to perform computation for a neural network, including multiplying, via the multiply circuitry, a first value and a second value. The circuit can be configured to perform computation for a neural network, including shifting, via the shift circuitry, a result of the multiplying by a determined number of bits. The circuit can be configured to perform computation for a neural network, including outputting the result of the multiplying when a sign bit of the first value is negative, and a result of the shifting when the sign bit of the first value is positive.

Abstract: Disclosed herein is a method for using a neural network across multiple devices. The method can include receiving, by a first device configured with a first one or more layers of a neural network, input data for processing via the neural network implemented across the first device and a second device. The method can include outputting, by the first one or more layers of the neural network implemented on the first device, a data set that is reduced in size relative to the input data while identifying one or more features of the input data for processing by a second one or more layers of the neural network. The method can include communicating, by the first device, the data set to the second device for processing via the second one or more layers of the neural network implemented on the second device.

Abstract: Disclosed herein includes a system, a method, and a device for improving computation efficiency of a neural network. In one aspect, adder circuitry is configured to add input data from processing of the neural network and a first number of bits of accumulated data for the neural network to generate summation data. In one aspect, according to a carry value of the adding from the adder circuitry, a multiplexer is configured to select between i) a second number of bits of the accumulated data and ii) incremented data comprising the second number of bits of the accumulated data incremented by a predetermined value. The summation data appended with the selected one of the second number of bits of the accumulated data or the incremented data may form appended data.

Abstract: A method of forming a multi-state nanosheet transistor device is provided. The method includes forming an alternating sequence of sacrificial layer segments and differentially doped nanosheet layer segments on a substrate, wherein each of the differentially doped nanosheet layer segments has a different dopant concentration from the other differentially doped nanosheet layer segments. The method further includes forming a source/drain on each of opposite ends of the sacrificial layer segments and differentially doped nanosheet layer segments, and removing the sacrificial layer segments. The method further includes depositing a gate dielectric layer on the differentially doped nanosheet layer segments, and forming a gate electrode on the gate dielectric layer to form a common gate-all-around structure, where each of the differentially doped nanosheet layer segments conducts current at a different threshold voltage.

Abstract: Methods and systems for performing calculations with a neural network include determining a conductance drift coefficient for resistive processing unit (RPU) weights in a neural network. A correction factor is applied to neuron inputs in the neural network in accordance with the drift coefficient and a time that has elapsed since the RPU weights were programmed. A calculation is performed with the neural network. The correction factor compensates for conductance drift.

Abstract: A method of forming a multi-state nanosheet transistor device is provided. The method includes forming an alternating sequence of sacrificial layer segments and differentially doped nanosheet layer segments on a substrate, wherein each of the differentially doped nanosheet layer segments has a different dopant concentration from the other differentially doped nanosheet layer segments. The method further includes forming a source/drain on each of opposite ends of the sacrificial layer segments and differentially doped nanosheet layer segments, and removing the sacrificial layer segments. The method further includes depositing a gate dielectric layer on the differentially doped nanosheet layer segments, and forming a gate electrode on the gate dielectric layer to form a common gate-all-around structure, where each of the differentially doped nanosheet layer segments conducts current at a different threshold voltage.

Abstract: A processor receives input data and provides the input data to a first neural network including a first neural network model. The first neural network model has a first numerical precision level. A first feature vector is generated from the input data using the first neural network. The input data is provided to a second neural network including a second neural network model. The second neural network model has a second numerical precision level different from the first numerical precession level. A second feature vector is generated from the input data using the second neural network. A difference metric is computed between the first feature vector and the second feature vector. The difference metric is indicative of whether the input data includes adversarial data.

Abstract: Techniques facilitating voltage droop reduction and/or mitigation in a processor core are provided. In one example, a system can comprise a memory that stores, and a processor that executes, computer executable components. The computer executable components can comprise an observation component that detects one or more events at a first stage of a processor pipeline. An event of the one or more events can be a defined event determined to increase a level of power consumed during a second stage of the processor pipeline. The computer executable components can also comprise an instruction component that applies a voltage droop mitigation countermeasure prior to the increase of the level of power consumed during the second stage of the processor pipeline and a feedback component that provides a notification to the instruction component that indicates a success or a failure of a result of the voltage droop mitigation countermeasure.

Abstract: A random number signal generator used for performing dropout or weight initialization for a node in a neural network. The random number signal generator includes a transistor which generates a random noise signal. The transistor includes a substrate, source and drain regions formed in the substrate, a first insulating layer formed over a channel of the transistor, a first trapping layer formed over the first insulating layer, a second insulating layer formed over the first trapping layer, and a second trapping layer formed over the second insulating layer. One or more traps in the first and second trapping layers are configured to capture or release one or more carriers flowing through the channel region. The random noise signal is generated as a function of one or more carrier being captured or released by the one or more traps.

Abstract: Techniques facilitating voltage droop reduction and/or mitigation in a processor core are provided. In one example, a system can comprise a memory that stores, and a processor that executes, computer executable components. The computer executable components can comprise an observation component that detects one or more events at a first stage of a processor pipeline. An event of the one or more events can be a defined event determined to increase a level of power consumed during a second stage of the processor pipeline. The computer executable components can also comprise an instruction component that applies a voltage droop mitigation countermeasure prior to the increase of the level of power consumed during the second stage of the processor pipeline and a feedback component that provides a notification to the instruction component that indicates a success or a failure of a result of the voltage droop mitigation countermeasure.

Abstract: A voltage droop mitigation system, that includes a first processor core that executes computer executable components stored in a memory. A time-based sensor component generates digital data representing voltage values associated with a power supply. A filtering component digitally conditions the generated digital data, and an analysis component analyzes the conditioned data and determines slope of the power supply voltage and employs counters to determine rate of data change over time; and if the slope is negative and exceeds a first pre-determined value for a pre-determined time period. The system implements one or more voltage droop-reduction techniques at the first processor core; and the first processor core transmits at least one of the following types of information: its voltage value, slope information or decision to apply droop reduction to one or more other cores.

Abstract: Techniques for statistics-aware weight quantization are presented. To facilitate reducing the bit precision of weights, for a set of weights, a quantizer management component can estimate a quantization scale value to apply to a weight as a linear or non-linear function of the mean of a square of a weight value of the weight and the mean of an absolute value of the weight value, wherein the quantization scale value is determined to have a smaller quantization error than all, or at least almost all, other quantization errors associated with other quantization scale values. A quantizer component applies the quantization scale value to symmetrically and/or uniformly quantize weights of a layer of the set of weights to generate quantized weights, the weights being quantized using rounding. The respective quantized weights can be used to facilitate training and inference of a deep learning system.

Abstract: Techniques facilitating on-chip supply noise voltage reduction and/or mitigation using local detection loops in a processor core are provided. In one example, a computer-implemented method can comprise detecting, by a processor core, a voltage droop at a first area of the processor core. The computer-implemented method can also comprise transmitting, by the processor core, voltage droop information to a local controller located in the first area and to a global controller located in the processor core. Further, the computer-implemented method can comprise applying, by the processor core, a first mitigation countermeasure at the first area of the processor core in response to a local instruction received from the local controller. The local instruction can comprise an indication of the first mitigation countermeasure.

Abstract: Techniques facilitating voltage management via on-chip sensors are provided. In one example, a computer-implemented method can comprise measuring, by a first processor core, power supply information. The computer-implemented method can also comprise measuring, by the first processor core, a value of an electrical current generated by the first processor core. Further, the computer-implemented method can comprise applying, by the first processor core, a mitigation technique at the first processor core in response to a determination that a combination of the power supply noise information and the value of the electrical current indicates a presence of a voltage noise at the first processor core.