Abstract: Systems and methods disclosed herein provide for an improved termination leg unit design and method of trimming impedance thereof, which provides for improved impedance matching for process variations, along with variations in temperature and voltage. Example implementation provide for a leg unit circuit design that includes a first circuit compensating for temperature and voltage variations and a second circuit, connected in series with the first circuit, compensating for process variations. Furthermore, disclosed herein is ZQ calibration method that provides for calibrating of the impedance of each of an on-die termination, a pull-up driver, and a pull-down driver using a single calibration circuit.

Abstract: Systems and methods are provided for generating a stable reference current that has low sensitivity to operating temperature and supply voltage variations and is stable across process corners. In an example implementation, an improved reference current generator circuit is provided that includes a first circuit generating a first current that is proportional to absolute temperature and a second circuit generating a second current that is complementary to absolute temperature based on first transistors operating in respective triode regions. The second current compensates for process, voltage, and temperature variations in the first current at a node. According to some examples, the second current is also generated based on second transistors operating in respective saturation regions. The first current may be generated using a forward biased PN junction diode.

Abstract: Systems and methods are provided for generating a stable DC reference voltage that has low sensitivity to operating temperature and supply voltage variations and is stable across process corners using complimentary metal-on-semiconductor field-effect transistors (MOSFETS). In an example implementation, a reference voltage generator circuit is provided that includes complimentary MOSFETs including a first complimentary MOSFET connected to a first node and having a first threshold voltage, and a second complimentary MOSFET connected to a second node and having a second threshold voltage that is greater than the first threshold voltage. The reference voltage generator circuit feeds the first node a first current based on mirroring a second current at the second node and outputs a stable DC reference voltage based on the first and second complimentary MOSFETs and configured operating in respective saturation regions.

Abstract: A mesh network-on-a-chip (NOC) with heterogenous routers for use with homogenous processing elements. Some of the routers are configured differently from other routers to interface more efficiently with particular physical resources that the processing elements require, such as particular input/output or memory devices. For example, one router is configured for use with the Peripheral Component Interconnect Express (PCIe) protocol, whereas another router is configured for use with the InterLaken communication protocol. Still further, the overall system is configured so that the various physical resources are physically adjacent to the particular router that is designed to access the resource to help ensure fair access by each processing element of the NOC to the particular resources that are required. The NOC may be part of a large manycore system on field programmable gate array (FPGA). The methods and apparatus described herein are generally applicable to all system-on-a-chip (SOC) designs.

Abstract: An innovative low-bit-width device may include a first digital-to-analog converter (DAC), a second DAC, a plurality of non-volatile memory (NVM) weight arrays, one or more analog-to-digital converters (ADCs), and a neural circuit. The first DAC is configured to convert a digital input signal into an analog input signal. The second DAC is configured to convert a digital previous hidden state (PHS) signal into an analog PHS signal. NVM weight arrays are configured to compute vector matrix multiplication (VMM) arrays based on the analog input signal and the analog PHS signal. The NVM weight arrays are coupled to the first DAC and the second DAC. The one or more ADCs are coupled to the plurality of NVM weight arrays and are configured to convert the VMM arrays into digital VMM values. The neural circuit is configured to process the digital VMM values into a new hidden state.

Abstract: Non-volatile memory structures for performing compute in memory inferencing for neural networks are presented. To improve performance, both in terms of speed and energy consumption, weight matrices are replaced with their singular value decomposition (SVD) and use of a low rank approximations (LRAs). The decomposition matrices can be stored in a single array, with the resultant LRA matrices requiring fewer weight values to be stored. The reduced sizes of the LRA matrices allow for inferencing to be performed more quickly and with less power. In a high performance and energy efficiency mode, a reduced rank for the SVD matrices stored on a memory die is determined and used to increase performance and reduce power needed for an inferencing operation.

Abstract: Methods and apparatus are disclosed for implementing principal component analysis (PCA) within a non-volatile memory (NVM) die of solid state drive (SSD) to reduce the dimensionality of machine learning data before the data is transferred to other components of the SSD, such as to a data storage controller equipped with a machine learning engine. The machine learning data may include, for example, training images for training an image recognition system in which the SSD is installed. In some examples, the on-chip PCA components of the NVM die are configured as under-the-array or next-to-the-array components. In other examples, one or more arrays of the NVM die are configured as multiplication cores for performing PCA matrix multiplication. In still other aspects, multiple NVM dies are arranged in parallel, each with on-chip PCA components to permit parallel concurrent on-chip processing of machine learning data.

Abstract: Techniques are presented for performing in-memory matrix multiplication operations for binary input, binary weight valued convolution neural network (CNN) inferencing. The weights of a filter are stored in pairs of memory cells of a storage class memory device, such as a ReRAM or phase change memory based devices. To reduce current consumption, the binary valued filters are transformed into ternary valued filters by taking sums and differences of binary valued filter pairs. The zero valued weights of the transformed filters are stored as a pair of high resistance state memory cells, reducing current consumption during convolution. The results of the in-memory multiplications are pair-wise combined to compensate for the filter transformations. To compensate for zero valued weights, a zero weight register stores the number of zero weights along each bit line and is used to initialize counter values for accumulating the multiplication operations.

Abstract: A non-volatile memory device is configured for in-memory computation of discrete Fourier transformations and their inverses. The real and imaginary components of the twiddle factors are stored as conductance values of memory cells in non-volatile memory arrays having a cross-point structure. The real and imaginary components of inputs are encoded as word line voltages applied to the arrays. Positive and negative valued components of the twiddle factors are stored separately and positive and negative of the inputs are separately applied to the arrays. Real and imaginary parts of the outputs for the discrete Fourier transformation are determined from combinations of the output currents from the arrays.

Abstract: Use of a NAND array architecture to realize a binary neural network (BNN) allows for matrix multiplication and accumulation to be performed within the memory array. A unit synapse for storing a weight of a BNN is stored in a pair of series connected memory cells. A binary input is applied on a pair of word lines connected to the unit synapse to perform the multiplication of the input with the weight. The results of such multiplications are determined by a sense amplifier, with the results accumulated by a counter. The arrangement extends to ternary inputs to realize a ternary-binary network (TBN) by adding a circuit to detect 0 input values and adjust the accumulated count accordingly. The arrangement further extends to a ternary-ternary network (TTN) by allowing 0 weight values in a unit synapse, maintaining the number of 0 weights in a register, and adjusting the count accordingly.

Abstract: A network device includes at least one control path port and data path ports configured to communicate on a network. A connection request is received from a host via a control path port, and a resource of the network device is allocated to the host. A data path port is determined from among the plurality of data path ports for communication between the host and the allocated resource. An indication of the determined data path port is sent to the host via the control path port for communication on a data path between the host and the allocated resource. In one aspect, a network interface includes at least one control path port and a first plurality of data path ports configured to communicate on a network. A connection request is received from a host via a control path port, and a locally connected device is allocated to the host.

Abstract: Techniques are presented for accelerating in-memory matrix multiplication operations for a convolution neural network (CNN) inference in which the weights of a filter are stored in the memory of a storage class memory device, such as a ReRAM or phase change memory based device. To improve performance for inference operations when filters exhibit sparsity, a zero column index and a zero row index are introduced to account for columns and rows having all zero weight values. These indices can be saved in a register on the memory device and when performing a column/row oriented matrix multiplication, if the zero row/column index indicates that the column/row contains all zero weights, the access of the corresponding bit/word line is skipped as the result will be zero regardless of the input.

Abstract: A non-volatile memory device includes arrays of non-volatile memory cells that are configured to the store weights for a recurrent neural network (RNN) inference engine with a gated recurrent unit (GRU) cell. A set three non-volatile memory arrays, such as formed of storage class memory, store a corresponding three sets of weights and are used to perform compute-in-memory inferencing. The hidden state of a previous iteration and an external input are applied to the weights of the first and the of second of the arrays, with the output of the first array used to generate an input to the third array, which also receives the external input. The hidden state of the current generation is generated from the outputs of the second and third arrays.

Abstract: Technology for reconfigurable input precision in-memory computing is disclosed herein. Reconfigurable input precision allows the bit resolution of input data to be changed to meet the requirements of in-memory computing operations. Voltage sources (that may include DACs) provide voltages that represent input data to memory cell nodes. The resolution of the voltage sources may be reconfigured to change the precision of the input data. In one parallel mode, the number of DACs in a DAC node is used to configure the resolution. In one serial mode, the number of cycles over which a DAC provides voltages is used to configure the resolution. The memory system may include relatively low resolution voltage sources, which avoids the need to have complex high resolution voltage sources (e.g., high resolution DACs). Lower resolution voltage sources can take up less area and/or use less power than higher resolution voltage sources.

Abstract: A non-volatile memory device includes an array of non-volatile memory cells that are configured to store weights of a neural network. Associated with the array is a data latch structure that includes a page buffer, which can store weights for a layer of the neural network that is read out of the array, and a transfer buffer, that can store inputs for the neural network. The memory device can perform multiply and accumulate operations between inputs and weight of the neural network within the latch structure, avoiding the need to transfer data out of the array and associated latch structure for portions of an inference operation. By using binary weights and inputs, multiplication can be performed by bit-wise XNOR operations. The results can then be summed and activation applied, all within the latch structure.

Abstract: An illustrative embodiment disclosed herein is an apparatus including a non-volatile memory cell and multi-bit input circuitry that simultaneously receives a plurality of bits, receives a supply voltage, converts the plurality of bits and the supply voltage into a multiply voltage, and applies the multiply voltage to the non-volatile memory cell. The non-volatile memory cell may pass a memory cell current in response to the multiply voltage. A magnitude of the multiply voltage may represent a multiplier. The memory cell current may represent a product of the multiplier and a multiplicand stored in the non-volatile memory cell.

Abstract: A non-volatile memory device is configured for in-memory computation of layers of a neural network by storing weight values as conductance values in memory cells formed of a series combination of a threshold switching selector, such as an ovonic threshold switch, and a programmable resistive element, such as a ReRAM element. By scaling the input voltages (representing inputs for the layer of the neural network) relative to the threshold values of the threshold switching selectors, dropout for inputs can be implemented to reduce overfitting by the neural network.

Abstract: Enhanced techniques and circuitry are presented herein for artificial neural networks. These artificial neural networks are formed from artificial synapses, which in the implementations herein comprise a memory arrays having non-volatile memory elements. In one implementation, an apparatus comprises a plurality of non-volatile memory arrays configured to store weight values for an artificial neural network. Each of the plurality of non-volatile memory arrays can be configured to receive data from a unified buffer shared among the plurality of non-volatile memory arrays, operate on the data, and shift at least portions of the data to another of the plurality of non-volatile memory arrays.

Abstract: To improve efficiencies for inferencing operations of neural networks, ensemble neural networks are used for compute-in-memory inferencing. In an ensemble neural network, the layers of a neural network are replaced by an ensemble of multiple smaller neural network generated from subsets of the same training data as would be used for the layers of the full neural network. Although the individual smaller network layers are “weak classifiers” that will be less accurate than the full neural network, by combining their outputs, such as in majority voting or averaging, the ensembles can have accuracies approaching that of the full neural network. Ensemble neural networks for compute-in-memory operations can have their efficiency further improved by implementations based binary memory cells, such as by binary neural networks using binary valued MRAM memory cells. The size of an ensemble can be increased or decreased to optimize the system according to error requirements.

Abstract: A network device includes at least one control path port and data path ports configured to communicate on a network. A connection request is received from a host via a control path port, and a resource of the network device is allocated to the host. A data path port is determined from among the plurality of data path ports for communication between the host and the allocated resource. An indication of the determined data path port is sent to the host via the control path port for communication on a data path between the host and the allocated resource. In one aspect, a network interface includes at least one control path port and a first plurality of data path ports configured to communicate on a network. A connection request is received from a host via a control path port, and a locally connected device is allocated to the host.