Abstract: Various embodiments provide for determining a capacitance (or capacitor value) of a circuit, determining a resistance-capacitance time constant (or RC time constant) of a circuit, or both. The circuit can comprise an integrated circuit (IC), such as a circuit implemented on die. An IC of some embodiments generates a frequency of a dock wave signal (e.g., an output signal) such that the clock wave signal encodes an effective capacitance of the IC, a RC time constant of the IC, or both. A component external to the IC, such as a controller, can receive the clock wave signal and determine the effective capacitance of the IC, the RC time constant of the IC, or both based on the received clock wave signal.

Abstract: An approach includes a method, product, and apparatus for dynamically removing sparse data on a pixel by pixel basis. In some embodiments, a machine learning processing job is received. The machine learning processing job is then executed on a pixel by pixel basis by selecting non-zero data values for input into a systolic array, wherein sparse data is not selected for input into the systolic array. Subsequently, a message is generated that provides an indication of whether the execution completed successfully. In some embodiments, the machine learning processing job comprises at least a plurality of multiply and accumulate operations. In some embodiments, at least one data value equal to zero for the machine learning processing job is not input into a systolic array. In some embodiments, a plurality of weights are input into a plurality of columns for each cycle.

Abstract: Various embodiments provide for routing a net of a circuit design using rule-based routing blockage extension, which may be part of electronic design automation (EDA). In particular, some embodiments route a net of a circuit design by determining a dimension extension value based on a design rule of the circuit design and applying the dimension extension value to at least one existing routing blockage.

Abstract: A system includes a machine configured to perform operations including accessing an integrated circuit design including a buffer tree that interconnects a plurality of inputs and buffers. The buffer tree includes a baseline timing characteristic. The operations include identifying a set of candidate solutions for improving the baseline timing characteristic using an initial timing model and selecting a subset of candidate solutions that have a timing characteristic lower than the baseline timing characteristic. Then the subset of candidate solutions are evaluated using a detailed timing model and based on determining that at least one candidate solution in the subset has a timing characteristic that is better than the baseline timing characteristic, selecting a candidate solution from the set of candidate solutions, and updating the buffer tree based on the candidate solution.

Abstract: Using a buffer sized according to the size of the filters of a convolutional neural network (CNN), a processor may use a read pointer to generate a two-dimensional virtual matrix of inputs. The number of inputs in each row in the two-dimensional virtual matrix of inputs may match the one-dimensional filter size of the cubic filters. The processor may collapse each of the cubic filters to one-dimensional linear arrays and generate a two-dimensional filter matrix from the one-dimensional linear arrays. The convolution computations for a corresponding layer of the CNN therefore reduce to a single matrix multiplication without any memory movement operations. When the buffer is refreshed using a new input frame, the processor may increment the initial read address of each read pointer by one and increment the final read address by one, circling back to the corresponding initial read address.

Abstract: Embodiments include herein are directed towards a dynamic random access memory system. Embodiments may include a command queue that is configured to hold all commands that are currently selectable for bank operation and execution. Embodiments may further include bank logic operatively connected with the command queue. The bank logic may include a bank management module and a plurality of bank slices, wherein each of the plurality of bank slices is an independent, re-assignable bank tracking module.

Abstract: Techniques are disclosed relating to an apparatus, including a data storage circuit having a plurality of entries, and a load-store pipeline configured to allocate an entry in the data storage circuit in response to a determination that a first instruction includes an access to an external memory circuit. The apparatus further includes an execution pipeline configured to make a determination, while performing a second instruction and using the entry in the data storage circuit, that the second instruction uses a result of the first instruction, and cease performance of the second instruction in response to the determination.

Abstract: An approach is described for a method, product, and apparatus for a machine learning process using dynamic rearrangement of sparse data and corresponding weights. This approach includes a method, product, and apparatus for dynamically rearranging input data to move sparse data to a location such that computations on the sparse data might be avoided when executing a machine learning processing job. For example, sparse data within each row of the input matrix can be moved to the end of each corresponding row. When the input data is folded to fit the array, that sparse data might be at least partially contained within a fold that comprises only sparse data and possibly filler data. In such an event, computations on the fold are unnecessary and are avoided. In some embodiments, the approach includes dynamically rearranging a weight matrix to maintain a correspondence between the input data and the weights.

Abstract: Aspects of the present disclosure address systems and methods for performing a machine-learning based clustering of clock sinks during clock tree synthesis. An integrated circuit (IC) design comprising a clock net that includes a plurality of clock sinks is accessed. An initial number of clusters to generate from the set of clock sinks is determined using a machine-learning model. A first set of clusters is generated from the set of clocks sinks and includes the initial number of clusters. A timing analysis is performed to determine whether each cluster in the first set of clusters satisfies design rule constraints. The initial number of clusters is adjusted based on the timing analysis and a clustering solution is generated based on the adjusted number of clusters.

Abstract: Embodiments included herein are directed towards a fractional feedback divider circuit and associated method. The circuit may include a programmable feedback divider including a plurality of flip-flops arranged in series. The programmable feedback divider may be configured to receive an input clock signal and a reset signal comprising at least one pulse and to generate a divided clock. The circuit may include reset logic configured to receive an input from the programmable feedback divider and to generate a reset signal. The circuit may include a first D flip-flop configured to receive the reset signal and to generate an output and a second D flip-flop configured to receive the output from the first D flip-flop and to generate a second output. The circuit may further include a multiplexer configured to receive the second output and to generate an output clock signal.

Abstract: Aspects of the present disclosure address systems and methods for fixed-point quantization using a dynamic quantization level adjustment scheme. Consistent with some embodiments, a method comprises accessing a neural network comprising floating-point representations of filter weights corresponding to one or more convolution layers. The method further includes determining a peak value of interest from the filter weights and determining a quantization level for the filter weights based on a number of bits in a quantization scheme. The method further includes dynamically adjusting the quantization level based on one or more constraints. The method further includes determining a quantization scale of the filter weights based on the peak value of interest and the adjusted quantization level. The method further includes quantizing the floating-point representations of the filter weights using the quantization scale to generate fixed-point representations of the filter weights.

Abstract: Various embodiments provide for failure mode analysis of a circuit design, which can be used as part of electronic design automation (EDA). In particular, some embodiments provide for failure mode analysis of a circuit design by determining a set of functional primitives of a circuit design component (e.g., cell at gate level) that contribute to a root cause logic for a specific failure mode.

Abstract: The present disclosure relates to a high speed, differential input, single phase clock circuit. The circuit may include a cross-coupled PMOS connected with a cross-coupled NMOS via a pass gate. The circuit may further include a single-phase clock in communication with the cross-coupled PMOS and the cross-coupled NMOS. The circuit may also include a master and a slave each having an output node that charges and discharges to VDD or ground respectively, wherein there is no direct feedback from an output of the circuit to an input the circuit and there is no precharged state in the circuit.

Abstract: Various embodiments provide for clustering-based grouping of cells in a cell library, which can be used for pruning the cell library. In particular, various embodiments provide for a clustering-based grouping of cells in a cell library based on a criterion (or cell attribute), and for pruning of the cell library based on the grouping of cells, which can optimize the cell library for the criterion. For instance, some embodiments provide for a clustering-based grouping of cells based on leakage power and then applying cell library pruning to optimize for cell leakage power.

Abstract: Aspects of the present disclosure address systems, methods, and a user interface for providing interactive skew group visualizations for integrated circuit (IC) design. The method includes causing display of a user interface that includes a display of a grouped view of a clock-tree including a plurality of skew group indicators. The method further includes receiving a user selection of a skew group indicator and updating the user interface to display a detailed view of the skew group including a graphical representation of each clock sink in the skew group and corresponding timing information. The method further includes receiving a second user selection of a first clock sink and in response, the display is updated to display an indicator of a physical location of the first clock sink within the clock tree.

Abstract: The present disclosure relates to a computer-implemented method for electronic design is provided. Embodiments may include receiving, using at least one processor, an electronic design having an original schematic associated therewith and extracting one or more features for each net from the schematic. Embodiments may include storing one or more resistance or capacitance values for each net and applying the one or more resistance or capacitance values as labels for a machine learning model. Embodiments may also include training the machine learning model using one or more actual values to generate a trained model. Embodiments may further include receiving the trained model to predict parasitics for a stitching engine and generating a stitched schematic.

Abstract: Various embodiments provide a system for performing operations that comprise accessing an integrated circuit design that includes a clock tree interconnecting a clock source to a plurality of clock sinks. The operations include receiving a request to adjust a present timing offset of the clock tree to a target timing offset. In response, a group of clock sinks to be adjusted are identified to satisfy the request. The clock tree is then modified by moving a terminal of the group from a first location in the clock tree to a second location in the clock tree to update the clock tree. An indication is provided that the updated clock tree has been modified and complies with the target timing offset.

Abstract: Methods and systems are disclosed for an upstream facing port implementation for DisplayPort link-training tunable PHY repeaters (LTTPRs). The device includes an upstream facing port to interface with an external DisplayPort source device and a downstream facing port to interface with an external DisplayPort sink device and the upstream facing port. The upstream facing port is configured to perform operations including receiving a main link data stream from an external transmitting display device, generating an outbound main link data stream, and providing the outbound main link data stream for transmitting by the external device. The device is also configured for receiving an updated main link data stream corresponding to the outbound main link data stream and sending the updated main link data stream to the downstream facing port to be transmitted to a receiving display device.

Abstract: An approach includes identification of a machine learning model for processing and generating an ordered set of weights with varying precisions and metadata that specifies where those values can be found in order to allow the identification of weights needed during processing. In a first embodiment, the variable precision weights are separated into different memory segments where each segment has weights of only a single precision. In a second embodiment, the variable precision weights are provided in a memory where weights of different precisions are intermingled, and those weights are identified using a sequence of pairs of data representing a number of weights with the same precision and the precision of those weights. In some embodiments, both the first and second embodiments are combined, where some segments contain weights with only a single precision and at least one segment stores weights with different precisions within a respective segment.

Abstract: An approach includes a new power and ground structure description language (PSDL) will allow the user to describe the desired routing pattern for each layer and on a user defined region by region basis, including how the pattern will be laid out in the design with respect to other patterns from a different layer. The new PSDL also gives the complete picture of the entire power and ground structure, instead of just a layer-by-layer view from a single command. It also allowed flexibility in alignment especially when dealing with track misalignments, thus avoiding the extensive trial-and-error steps needed to calculate offsets and distances to maintain pattern alignment using previous approaches. Additionally, because PSDL is not tightly dependent on the design size and/or floorplan, transferring the desired power and ground structure from one design to another will be very easy with only few adjustments.