Various examples are directed to systems and methods for executing a loop in a reconfigurable compute fabric. A first flow controller may initiate a first thread at a first synchronous flow to execute a first portion of a first iteration of the loop. A second flow controller may receive a first asynchronous message instructing the second flow controller to initiate a first thread at a second synchronous flow to execute a second portion of the first iteration. The second flow controller may determine that the first iteration of the loop is the last iteration of the loop to be executed and initiate the first thread at the second synchronous flow with a last iteration flag set.

Embodiments are directed to a processor having a functional slice architecture. The processor is divided into tiles (or functional units) organized into a plurality of functional slices. The functional slices are configured to perform specific operations within the processor, which includes memory slices for storing operand data and arithmetic logic slices for performing operations on received operand data (e.g., vector processing, matrix manipulation). The processor includes a plurality of functional slices of a module type, each functional slice having a plurality of tiles. The processor further includes a plurality of data transport lanes for transporting data in a direction indicated in a corresponding instruction. The processor also includes a plurality of instruction queues, each instruction queue associated with a corresponding functional slice of the plurality of functional slices, wherein the instructions in the instruction queues comprise a functional slice specific operation code.

Techniques in advanced deep learning provide improvements in one or more of accuracy, performance, and energy efficiency. An array of processing elements performs flow-based computations on wavelets of data. Each processing element has a compute element and a routing element. Each compute element has memory. Each router enables communication via wavelets with at least nearest neighbors in a 2D mesh. Routing is controlled by respective virtual channel specifiers in each wavelet and routing configuration information in each router. A compute element conditionally selects for task initiation a previously received wavelet specifying a particular one of the virtual channels. The conditional selecting excludes the previously received wavelet for selection until at least block/unblock state maintained for the particular virtual channel is in an unblock state. The compute element executes block/unblock instructions to modify the block/unblock state.

Methods, apparatuses, and systems for implementing data flows in a processor are described herein. A data flow manager may be configured to generate a configuration packet for a compute operation based on status information regarding multiple processing elements of the processor. Accordingly, multiple processing elements of a processor may concurrently process data flows based on the configuration packet. For example, the multiple processing elements may implement a mapping of processing elements to memory, while also implementing identified paths, through the processor, for the data flows. After executing the compute operation at certain processing elements of the processor, the processing results may be provided. In speech signal processing operations, the processing results may be compared to phonemes to identify such components of human speech in the processing results. Once dynamically identified, the processing elements may continue comparing additional components of human speech to facilitate processing of an audio recording, for example.

A computer system according to the present invention includes N (N is an integer of 2 or more) data output devices, a transmission control device, and an arithmetic device, in which the arithmetic device executes predetermined arithmetic processing on data collected from the N data output devices via a communication network connecting the data output devices and the arithmetic device to each other, the transmission control device controls transmission timing of data output from the N data output devices according to a processing content of the predetermined arithmetic processing executed by the arithmetic device, and the N data storage devices are configured to output the data on the basis of the transmission timing notified by the transmission control device.

Methods and systems for executing an application data flow graph on a set of computational nodes are disclosed. The computational nodes can each include a programmable controller from a set of programmable controllers, a memory from a set of memories, a network interface unit from a set of network interface units, and an endpoint from a set of endpoints. A disclosed method comprises configuring the programmable controllers with instructions. The method also comprises independently and asynchronously executing the instructions using the set of programmable controllers in response to a set of events exchanged between the programmable controllers themselves, between the programmable controllers and the network interface units, and between the programmable controllers and the set of endpoints. The method also comprises transitioning data in the set of memories on the computational nodes in accordance with the application data flow graph and in response to the execution of the instructions.

Techniques in advanced deep learning provide improvements in one or more of accuracy, performance, and energy efficiency. An array of processing elements performs flow-based computations on wavelets of data. Each processing element has a respective compute element and a respective routing element. Instructions executed by the compute element include operand specifiers, some specifying a data structure register storing a data structure descriptor describing an operand as a fabric vector or a memory vector. The data structure descriptor further describes the memory vector as one of a one-dimensional vector, a four-dimensional vector, or a circular buffer vector. Optionally, the data structure descriptor specifies an extended data structure register storing an extended data structure descriptor. The extended data structure descriptor specifies parameters relating to a four-dimensional vector or a circular buffer vector.

An event-processing unit for processing tokens associated with a state or state transition, herein also referred to as an event, of an external device is disclosed. The EPU allows token-processing schemes, in which the processing of incoming tokens and the further handling of a processing result by the EPU are determined not only by the token identifier, but also by the payload data of the incoming token or by data in the data memory. A flag-processing capability of a processing-control stage allows applying flag-processing operations such as logical operations to data obtained as a processing result of an ALU-processing operation. The result of these operations determines a subsequent handling of ALU-result data by the EPU. Thus, whether or not the ALU-result data is written to the data memory also influences the processing of any subsequent incoming tokens for which that data is used in the ALU-processing operation.

Embodiments of the present disclosure are directed toward techniques and configurations enhancing the performance of hardware (HW) accelerators. The present disclosure provides a schedule-aware, dynamically reconfigurable, tree-based partial sum accumulator architecture for HW accelerators, wherein the depth of an adder tree in the HW accelerator is dynamically based on a dataflow schedule generated by a compiler. The adder tree depth is adjusted on a per-layer basis at runtime. Configuration registers, programmed via software, dynamically alter the adder tree depth for partial sum accumulation based on the dataflow schedule. By facilitating a variable depth adder tree during runtime, the compiler can choose a compute optimal dataflow schedule that minimizes the number of compute cycles needed to accumulate partial sums across multiple processing elements (PEs) within a PE array of a HW accelerator. Other embodiments may be described and/or claimed.

Techniques in advanced deep learning provide improvements in one or more of accuracy, performance, and energy efficiency. An array of processing elements performs flow-based computations on wavelets of data. Each processing element has a respective compute element and a respective routing element. Instructions executed by the compute element include operand specifiers, some specifying a data structure register storing a data structure descriptor describing an operand as a fabric vector or a memory vector. The data structure descriptor further describes the memory vector as one of a one-dimensional vector, a four-dimensional vector, or a circular buffer vector. Optionally, the data structure descriptor specifies an extended data structure register storing an extended data structure descriptor. The extended data structure descriptor specifies parameters relating to a four-dimensional vector or a circular buffer vector.