Abstract: Methods and systems include a neural network system that includes a neural network accelerator. The neural network accelerator includes multiple processing engines coupled together to perform arithmetic operations in support of an inference performed using the deep neural network system. The neural network accelerator also includes a schedule-aware tensor data distribution circuitry or software that is configured to load tensor data into the multiple processing engines in a load phase, extract output data from the multiple processing engines in an extraction phase, reorganize the extracted output data, and store the reorganized extracted output data to memory.

Abstract: An apparatus to facilitate accelerating neural networks with low precision-based multiplication and exploiting sparsity in higher order bits is disclosed. The apparatus includes a processor comprising a re-encoder to re-encode a first input number of signed input numbers represented in a first precision format as part of a machine learning model, the first input number re-encoded into two signed input numbers of a second precision format, wherein the first precision format is a higher precision format than the second precision format. The processor further includes a multiply-add circuit to perform operations in the first precision format using the two signed input numbers of the second precision format; and a sparsity hardware circuit to reduce computing on zero values at the multiply-add circuit, wherein the processor to execute the machine learning model using the re-encoder, the multiply-add circuit, and the sparsity hardware circuit.

Abstract: An apparatus to facilitate accelerating neural networks with low precision-based multiplication and exploiting sparsity in higher order bits is disclosed. The apparatus includes a processor comprising a re-encoder to re-encode a first input number of signed input numbers represented in a first precision format as part of a machine learning model, the first input number re-encoded into two signed input numbers of a second precision format, wherein the first precision format is a higher precision format than the second precision format. The processor further includes a multiply-add circuit to perform operations in the first precision format using the two signed input numbers of the second precision format; and a sparsity hardware circuit to reduce computing on zero values at the multiply-add circuit, wherein the processor to execute the machine learning model using the re-encoder, the multiply-add circuit, and the sparsity hardware circuit.

Abstract: Technology for estimating neural network (NN) power profiles includes obtaining a plurality of workloads for a compiled NN model, the plurality of workloads determined for a hardware execution device, determining a hardware efficiency factor for the compiled NN model, and generating, based on the hardware efficiency factor, a power profile for the compiled NN model on one or more of a per-layer basis or a per-workload basis. The hardware efficiency factor can be determined on based on a hardware efficiency measurement and a hardware utilization measurement, and can be determined on a per-workload basis. A configuration file can be provided for generating the power profile, and an output visualization of the power profile can be generated. Further, feedback information can be generated to perform one or more of selecting a hardware device, optimizing a breakdown of workloads, optimizing a scheduling of tasks, or confirming a hardware device design.

Abstract: A disclosed apparatus to multiply matrices includes a compute engine. The compute engine includes multipliers in a two dimensional array that has a plurality of array locations defined by columns and rows. The apparatus also includes a plurality of adders in columns. A broadcast interconnect between a cache and the multipliers broadcasts a first set of operand data elements to multipliers in the rows of the array. A unicast interconnect unicasts a second set of operands between a data buffer and the multipliers. The multipliers multiply the operands to generate a plurality of outputs, and the adders add the outputs generated by the multipliers.

Abstract: Methods and systems include a neural network system that includes a neural network accelerator comprising. The neural network accelerator includes multiple processing engines coupled together to perform arithmetic operations in support of an inference performed using the deep neural network system. The neural network accelerator also includes a schedule-aware tensor data distribution circuitry or software that is configured to load tensor data into the multiple processing engines in a load phase, extract output data from the multiple processing engines in an extraction phase, reorganize the extracted output data, and store the reorganized extracted output data to memory.

Abstract: An apparatus to facilitate accelerating neural networks with low precision-based multiplication and exploiting sparsity in higher order bits is disclosed. The apparatus includes a processor comprising a re-encoder to re-encode a first input number of signed input numbers represented in a first precision format as part of a machine learning model, the first input number re-encoded into two signed input numbers of a second precision format, wherein the first precision format is a higher precision format than the second precision format. The processor further includes a multiply-add circuit to perform operations in the first precision format using the two signed input numbers of the second precision format; and a sparsity hardware circuit to reduce computing on zero values at the multiply-add circuit, wherein the processor to execute the machine learning model using the re-encoder, the multiply-add circuit, and the sparsity hardware circuit.

Abstract: A disclosed apparatus to multiply matrices includes a compute engine. The compute engine includes multipliers in a two dimensional array that has a plurality of array locations defined by columns and rows. The apparatus also includes a plurality of adders in columns. A broadcast interconnect between a cache and the multipliers broadcasts a first set of operand data elements to multipliers in the rows of the array. A unicast interconnect unicasts a second set of operands between a data buffer and the multipliers. The multipliers multiply the operands to generate a plurality of outputs, and the adders add the outputs generated by the multipliers.

Abstract: Example apparatus disclosed herein include an array of processor elements, the array including rows each having a first number of processor elements and columns each having a second number of processor elements. Disclosed example apparatus also include configuration registers to store descriptors to configure the array to implement a layer of a convolutional neural network based on a dataflow schedule corresponding to one of multiple tensor processing templates, ones of the processor elements to be configured based on the descriptors to implement the one of the tensor processing templates to operate on input activation data and filter data associated with the layer of the convolutional neural network to produce output activation data associated with the layer of the convolutional neural network. Disclosed example apparatus further include memory to store the input activation data, the filter data and the output activation data associated with the layer of the convolutional neural network.

Abstract: Techniques of debugging a computing system are described herein. The techniques may include generating debug data at agents in the computing system. The techniques may include recording the debug data at a storage element, wherein the storage element is disposed in a non-core portion of the circuit interconnect accessible to the agents.

Abstract: A technique to monitor events within a computer system or integrated circuit. In one embodiment, a software-accessible event monitoring storage and hardware-specific monitoring logic are selectable and their corresponding outputs may be monitored by accessing a counter to count events corresponding to each of software-accessible storage and hardware-specific monitoring logic.

Abstract: Techniques of debugging a computing system are described herein. The techniques may include generating debug data at agents in the computing system. The techniques may include recording the debug data at a storage element, wherein the storage element is disposed in a non-core portion of the circuit interconnect accessible to the agents.

Abstract: A technique to monitor events within a computer system or integrated circuit. In one embodiment, a software-accessible event monitoring storage and hardware-specific monitoring logic are selectable and their corresponding outputs may be monitored by accessing a counter to count events corresponding to each of software-accessible storage and hardware-specific monitoring logic.

Abstract: An interconnect bandwidth throttler is disclosed. The interconnect bandwidth throttler turns off the interconnect, based on whether a maximum number of transactions has taken place within a predetermined throttle window. Both the maximum number of transactions and the throttle window are adjustable. Characteristics such as performance, thermal considerations, and average power are adjustable using the interconnect bandwidth throttler.

Abstract: A method to reduce idle leakage power in I/O pins of an integrated circuit using external circuitry. Initially, I/O pins on a package are subdivided into those that will also remain powered up and those that will power down during idle state. When a system enters a low power mode, a signal is sent to the external circuitry. The signal notifies the I/O pins that always remain powered up to notify the external circuitry to power down the other set of I/O pins.

Abstract: A technique to promote determinism among multiple clocking domains within a computer system or integrated circuit, In one embodiment, one or more execution units are placed in a deterministic state with respect to multiple clocks within a processor system having a number of different clocking domains.

Abstract: An interconnect bandwidth throttler is disclosed. The interconnect bandwidth throttler turns off the interconnect, based on whether a maximum number of transactions has taken place within a predetermined throttle window. Both the maximum number of transactions and the throttle window are adjustable. Characteristics such as performance, thermal considerations, and average power are adjustable using the interconnect bandwidth throttler.

Abstract: An interconnect bandwidth throttler is disclosed. The interconnect bandwidth throttler turns off the interconnect, based on whether a maximum number of transactions has taken place within a predetermined throttle window. Both the maximum number of transactions and the throttle window are adjustable. Characteristics such as performance, thermal considerations, and average power are adjustable using the interconnect bandwidth throttler.

Abstract: An interconnect bandwidth throttler is disclosed. The interconnect bandwidth throttler turns off the interconnect, based on whether a maximum number of transactions has take place within a predetermined throttle window. Both the maximum number of transactions and the throttle window are adjustable.

Abstract: A front side bus swizzle mechanism modifies the front side (address and data) bus on a chip so that, when the chip is positioned on one side of a printed circuit board, connection to a second chip located on the opposite side of the printed circuit board is simplified. The simplified connection may result in less complexity and minimize the consumption of additional printed circuit board real estate.