Xilinx, Inc.

Xlnx, Inc.

Sundeep Ram Gopal Agarwal

Brian C Gaide

Santosh Yachareni

Prashant Dubey

Ramakrishna Kishore Tanikella

Vijay Kumar Reddy Enumula

Austin H Lesea

Michael J Hart

Jun Liu

Shidong Zhou

Praful Jain

Pierre Maillard

Anil Kumar Kandala

Narendra Kumar Pulipati

Sree Rkc Saraswatula

Vijay Kumar Koganti

Aashish Tripathi

An example method of implementing a quantized neural network (QNN) for a programmable device includes: identifying multiply-accumulate operations of neurons in the QNN; converting the multiply-accumulate operations to memory lookup operations; and implementing the memory lookup operations using a pre-compute circuit for the programmable device, the pre-compute circuit storing a pre-computed output of a neuron in the QNN for each of the memory lookup operations.

Partial sum pre-computation to implement quantized neural networks on programmable devices

Examples of the present disclosure provide power gating for stacked die structures. In some examples, a stacked die structure comprises a first die and a second die bonded to the first die. In some examples, a power gated power path is from a bonding interface between the dies through TSVs in the second die, a power gating device in the second die, and routing of metallization layers in the second die to the circuit region in the second die. In some examples, a power gated power path comprises a power gating device in a power gating region of the first die and is configured to interrupt a flow of current through the power gated power path to a circuit region in the second die.

Power gating in stacked die structures

Examples described herein provide a method for disabling a defective portion of a fabric die of a stacked IC device. The method includes receiving a signal indicating that a portion of a fabric die of a stacked IC device including at least two fabric dies is defective. The method further includes, in response to the signal, pulling a source voltage rail of the defective portion to ground, thereby disabling the portion, and operating the remainder of the fabric die without interference from or contention with the disabled portion. In one example, the stacked IC device is an active on active (AoA) device, and the portion of the fabric die includes a configuration memory cell. In one example, the signal is received after power-up of the stacked IC device.

Yield-centric power gated regulated supply design with programmable leakers

Examples described herein generally relate to multi-chip devices having stacked chips. In an example, a multi-chip device includes a chip stack that includes chips. Neighboring chips are connected to each other. Plural chips of the chips collectively include columns of broken via pillars and bridges. Each of the plural chips has a broken via pillar in each column. The broken via pillar has first and second continuous via pillar portions aligned in a direction normal to a side of a semiconductor substrate of the respective chip. The first continuous via pillar portion is not connected within the broken via pillar to the second continuous via pillar portion. Each of the plural chips has one or more of the bridges. Each bridge connects, within the respective chip, the first continuous via pillar portion in a column and the second continuous via pillar portion in another column.

Multi-chip stacked devices

Configurable termination circuits for use with programmable logic devices are disclosed. In one implementation, the termination circuit may include one or more components to couple unused inputs of one or more configurable logic blocks to a fixed voltage. In another implementation, the termination circuit may include one or more components to couple unused inputs of one or more configurable logic blocks to an output of the one or more configurable logic blocks. In some implementations, the programmable logic device may include a platform management controller to configure the termination circuits based on configuration data.

Programmable termination circuits for programmable devices

Embodiments herein describe an interface between PL fabric and a hardened block that includes a programmable pipeline. This pipeline includes at least a sequential element and a bypass path. For time critical nets in a netlist, the programmable IC routes a net through the sequential element. Doing so mitigates or eliminates the uncertainty associated with routing the net from the hardened block through PL fabric. Also, the sequential element can increase the available time for capturing the data. For less time critical nets, the net can route through the bypass path. This means the route from the hardened block to the PL fabric is determined on the fly by a routing algorithm rather than being fixed.

Programmable pipeline at interface of hardened blocks

Disclosed approaches for circuitry that implements a softmax function include difference calculation circuitry configured to calculate differences between combinations of elements, z&#x2212;z, of a vector. First lookup circuitry is configured to lookup and output representations of exponential values, eassociated with the differences in response to input of the differences. Each adder circuit of N adder circuits sums a subset of the exponential values output from the first lookup circuitry and a value of 1. The sum output by each adder circuit denotes a denominator of a plurality of denominators of the softmax function. Second lookup circuitry is configured with quotients and looks-up and outputs quotients associated with the plurality of denominators as results of the softmax function.

Softmax circuit

A circuit for transmitting signals in an integrated circuit device is described. The circuit comprises a first die; a second die stacked on the first die; and a buffer transmitting data between the first die and the second die; wherein a first inverter of the buffer is on the first die, and a second inverter of the buffer is on the second die. A method of transmitting signals in an integrated circuit device is also described.

Circuit for and method of transmitting a signal in an integrated circuit device

A master-slave flip-flop implemented in an integrated circuit comprises a master latch coupled to receive data at an input; and a slave latch coupled to an output of the master latch, wherein the slave latch comprises an SEU-enhanced latch, and the master latch is not enhanced for SEU protection. A method of implementing a master-slave flip-flop in an integrated circuit is also described.

Growing community of inventors

Hyderabad Telangana, India

Sundeep Ram Gopal Agarwal