Abstract: A chiplet-based architecture may quantize, or reduce, the number of bits at various stages of the data path in an artificial-intelligence processor. This architecture may leverage the synergy between quantizing multiple dimensions together to greatly decrease the memory usage and data path bandwidth. Internal weights may be quantized statically after a training procedure. Accumulator bits and activation bits may be quantized dynamically during an inference operation. New hardware logic may be configured to quantize the outputs of each operation directly from the core or other processing node before the tensor is stored in memory. Quantization may use a statistic from a previous tensor for a current output tensor, while also calculating a statistic to be used on a subsequent output tensor. In addition to quantizing based on a statistic, bits can be further quantized using a Kth percentile clamping operation.

Abstract: A smart network interface controller (NIC) implemented using a stacked die configuration is provided. The NIC may include user-customizable networking circuits formed in a top programmable die and primitive network function blocks formed in a bottom application-specific integrated circuit (ASIC) die. The top programmable die may provide a flexible packet processing pipeline to facilitate efficient control and data communication between the user-customizable networking circuits and the primitive network function blocks. The bottom ASIC die may also include an array of memory blocks operable as lookup tables and intermediate buffers for other network processing circuitry in the NIC. A NIC configured in this way provides both performance, power, and area benefits and superior customer configurability.

Abstract: A router may include input buffers that receive a packet being transmitted from a source to a destination, a state generator that determines a state for the packet, and a memory representing weights for actions corresponding to possible states. The memory may be configured to return an action corresponding to the state of the packet, where the action may indicate a next hop in the route between the source and the destination. The router may also include reward logic configured to generate the weights for the plurality of actions in the memory. The reward logic may receive a global reward corresponding to the route between the source and the destination, calculate a local reward corresponding to next hops available to the router; and combine the global reward and the local reward to generate the weights for the plurality of actions in the memory.

Abstract: A smart network interface controller (NIC) implemented using a stacked die configuration is provided. The NIC may include user-customizable networking circuits formed in a top programmable die and primitive network function blocks formed in a bottom application-specific integrated circuit (ASIC) die. The top programmable die may provide a flexible packet processing pipeline to facilitate efficient control and data communication between the user-customizable networking circuits and the primitive network function blocks. The bottom ASIC die may also include an array of memory blocks operable as lookup tables and intermediate buffers for other network processing circuitry in the NIC. A NIC configured in this way provides both performance, power, and area benefits and superior customer configurability.

Abstract: A router may include input buffers that receive a packet being transmitted from a source to a destination, a state generator that determines a state for the packet, and a memory representing weights for actions corresponding to possible states. The memory may be configured to return an action corresponding to the state of the packet, where the action may indicate a next hop in the route between the source and the destination. The router may also include reward logic configured to generate the weights for the plurality of actions in the memory. The reward logic may receive a global reward corresponding to the route between the source and the destination, calculate a local reward corresponding to next hops available to the router; and combine the global reward and the local reward to generate the weights for the plurality of actions in the memory.

Abstract: A method of inducing sparsity for outputs of neural network layer may include receiving outputs from a layer of a neural network; partitioning the outputs into a plurality of partitions; identifying first partitions in the plurality of partitions that can be treated as having zero values; generating an encoding that identifies locations of the first partitions among remaining second partitions in the plurality of partitions; and sending the encoding and the second partitions to a subsequent layer in the neural network.

Abstract: A network-on-package (NoPK) for connecting a plurality of chiplets may include a plurality of interface bridges configured to convert a plurality of protocols used by the plurality of chiplets into a common protocol, a routing network configured to route traffic between the plurality of interface bridges using the common protocol, and a controller configured to program the plurality of interface bridges and the routing network based on types of the plurality of chiplets connected to the NoPK. The NoPK may provide a scalable connection for any number of chiplets from different ecosystems using different communication protocols.

Abstract: A network-on-package (NoPK) for connecting a plurality of chiplets may include a plurality of interface bridges configured to convert a plurality of protocols used by the plurality of chiplets into a common protocol, a routing network configured to route traffic between the plurality of interface bridges using the common protocol, and a controller configured to program the plurality of interface bridges and the routing network based on types of the plurality of chiplets connected to the NoPK. The NoPK may provide a scalable connection for any number of chiplets from different ecosystems using different communication protocols.

Abstract: A smart network interface controller (NIC) implemented using a stacked die configuration is provided. The NIC may include user-customizable networking circuits formed in a top programmable die and primitive network function blocks formed in a bottom application-specific integrated circuit (ASIC) die. The top programmable die may provide a flexible packet processing pipeline to facilitate efficient control and data communication between the user-customizable networking circuits and the primitive network function blocks. The bottom ASIC die may also include an array of memory blocks operable as lookup tables and intermediate buffers for other network processing circuitry in the NIC. A NIC configured in this way provides both performance, power, and area benefits and superior customer configurability.

Abstract: A circuit and algorithm are disclosed for a step2 search of a three step search of synchronization channels in a W-CDMA system. A mobile terminal of the CDMA system includes an RF downconverter for receiving I and Q signals. A searcher, responsive to the I and Q signals, includes a first correlator for correlating the I and Q signals with a primary synchronization code on a primary synchronization channel, and a second correlator for correlating I and Q signals with a secondary synchronization code on a secondary synchronization channel. The correlated I and Q signals are added for each of the secondary synchronization codes. An energy calculator and a maximum energy detector use the correlated I and Q signals of both the primary and secondary synchronization channels to detect the most likely scrambling code group of secondary synchronization codes.

Abstract: A circuit and algorithm are disclosed for a step2 search of a three step search of synchronization channels in a W-CDMA system. A mobile terminal of the CDMA system includes an RF downconverter for receiving I and Q signals. A searcher, responsive to the I and Q signals, includes a first correlator for correlating the I and Q signals with a primary synchronization code on a primary synchronization channel, and a second correlator for correlating I and Q signals with a secondary synchronization code on a secondary synchronization channel. The correlated I and Q signals are added for each of the secondary synchronization codes. An energy calculator and a maximum energy detector use the correlated I and Q signals of both the primary and secondary synchronization channels to detect the most likely scrambling code group of secondary synchronization codes.

Abstract: A circuit and algorithm are disclosed for a step2 search of a three step search of synchronization channels in a W-CDMA system. A mobile terminal of the CDMA system includes an RF downconverter for receiving I and Q signals. A searcher, responsive to the I and Q signals, includes a first correlator for correlating the I and Q signals with a primary synchronization code on a primary synchronization channel, and a second correlator for correlating I and Q signals with a secondary synchronization code on a secondary synchronization channel. The correlated I and Q signals are added for each of the secondary synchronization codes. An energy calculator and a maximum energy detector use the correlated I and Q signals of both the primary and secondary synchronization channels to detect the most likely scrambling code group of secondary synchronization codes.

Abstract: A circuit and algorithm are disclosed for a step2 search of a three step search of synchronization channels in a W-CDMA system. A mobile terminal of the CDMA system includes an RF downconverter for receiving I and Q signals. A searcher, responsive to the I and Q signals, includes a first correlator for correlating the I and Q signals with a primary synchronization code on a primary synchronization channel, and a second correlator for correlating I and Q signals with a secondary synchronization code on a secondary synchronization channel. The correlated I and Q signals are added for each of the secondary synchronization codes. An energy calculator and a maximum energy detector use the correlated I and Q signals of both the primary and secondary synchronization channels to detect the most likely scrambling code group of secondary synchronization codes.

Abstract: Method and apparatus for circuit testing with ring-connected test instrument modules. A system for controlling one or more test instruments to test one or more integrated circuits includes a master clock and a controller. The test instruments are connected to form a communication ring. The master clock is connected to each test instrument and provides a clock signal to the one or more test instruments. The controller is connected to the communication ring and is configured to align counters of test instruments to derive a common clock time value from the clock signal. The controller is further configured to generate and send data words into the communication ring to carry the data words to each test instrument. The data words includes at least one data word specifying a test event to be performed, a common clock time value, and at least one of the test instruments.

Abstract: Method and apparatus for circuit testing with ring-connected test instrument modules. A system for controlling one or more test instruments to test one or more integrated circuits includes a master clock and a controller. The test instruments are connected to form a communication ring. The master clock is connected to each test instrument and provides a clock signal to the one or more test instruments. The controller is connected to the communication ring and is configured to align counters of test instruments to derive a common clock time value from the clock signal. The controller is further configured to generate and send data words into the communication ring to carry the data words to each test instrument. The data words includes at least one data word specifying a test event to be performed, a common clock time value, and at least one of the test instruments.