Abstract: The present disclosure relates generally to techniques for efficiently performing operations associated with artificial intelligence (AI), machine learning (ML), and/or deep learning (DL) applications, such as training and/or interference calculations, using an integrated circuit device. More specifically, the present disclosure relates to an integrated circuit design implemented to perform these operations with low latency and/or a high bandwidth of data. For example, embodiments of a computationally dense digital signal processing (DSP) circuitry, implemented to efficiently perform one or more arithmetic operations (e.g., a dot-product) on an input are disclosed. Moreover, embodiments described herein may relate to layout, design, and data scheduling of a processing element array implemented to compute matrix multiplications (e.g., systolic array multiplication).

Abstract: Systems and methods for dynamically sizing inter-kernel communication channels implemented on an integrated circuit (IC) are provided. Implementation characteristics of the channels, predication, and kernel scheduling imbalances may factor into properly sizing the channels for self-synchronization, resulting in optimized steady-state throughput.

Abstract: An integrated circuit may include elastic datapaths or pipelines, through which software threads or iterations of loops, may be executed. Throttling circuitry may be coupled along an elastic pipeline in the integrated circuit. The throttling circuitry may include dependency detection circuitry that dynamically detect memory dependency issues that may arise during runtime. To mitigate these dependency issues, the throttling circuitry may assert stall signals to upstream stages in the pipeline. Additionally, the throttling circuitry may control the pipeline to resolve a store operation prior to a corresponding load operation in order to avoid store/load conflicts. In an embodiment, the throttling circuitry may include a validator circuit, a rewind block, a revert block, and a flush block. The throttling circuitry may pass speculative iterations through the rewind block, and later validate the speculative iterations using the validator block.

Abstract: The present disclosure relates generally to techniques for efficiently performing operations associated with artificial intelligence (AI), machine learning (ML), and/or deep learning (DL) applications, such as training and/or interference calculations, using an integrated circuit device. More specifically, the present disclosure relates to an integrated circuit design implemented to perform these operations with low latency and/or a high bandwidth of data. For example, embodiments of a computationally dense digital signal processing (DSP) circuitry, implemented to efficiently perform one or more arithmetic operations (e.g., a dot-product) on an input are disclosed. Moreover, embodiments described herein may relate to layout, design, and data scheduling of a processing element array implemented to compute matrix multiplications (e.g., systolic array multiplication).

Abstract: The present disclosure relates generally to techniques for efficiently performing operations associated with artificial intelligence (AI), machine learning (ML), and/or deep learning (DL) applications, such as training and/or interference calculations, using an integrated circuit device. More specifically, the present disclosure relates to an integrated circuit design implemented to perform these operations with low latency and/or a high bandwidth of data. For example, embodiments of a computationally dense digital signal processing (DSP) circuitry, implemented to efficiently perform one or more arithmetic operations (e.g., a dot-product) on an input are disclosed. Moreover, embodiments described herein may relate to layout, design, and data scheduling of a processing element array implemented to compute matrix multiplications (e.g., systolic array multiplication).

Abstract: Systems and methods for dynamically sizing inter-kernel communication channels implemented on an integrated circuit (IC) are provided. Implementation characteristics of the channels, predication, and kernel scheduling imbalances may factor into properly sizing the channels for self-synchronization, resulting in optimized steady-state throughput.

Abstract: Systems and methods for dynamically sizing inter-kernel communication channels implemented on an integrated circuit (IC) are provided. Implementation characteristics of the channels, predication, and kernel scheduling imbalances may factor into properly sizing the channels for self-synchronization, resulting in optimized steady-state throughput.

Abstract: The present embodiments relate to interface circuitry between a serial interface circuit and an array of processing elements in an integrated circuit. The interface circuitry may include a daisy chain of feeder circuits and a daisy chain of drain circuits. If desired, the interface circuitry may include multiple daisy chains of feeder circuits and/or multiple daisy chains of drain circuits. These multiple daisy chains of feeder circuits and drains circuits may be coupled in parallel, respectively. In some embodiments, the interface circuitry may include synchronization circuitry that is coupled between the daisy chains of drain circuits and the serial interface circuit. Pipeline register stages between feeder circuits and/or between drain circuits may enable the placement of the feeder circuits and/or the drain circuits spatially close to the processing elements of the array of processing elements.

Abstract: The present disclosure relates generally to techniques for efficiently performing operations associated with artificial intelligence (AI), machine learning (ML), and/or deep learning (DL) applications, such as training and/or interference calculations, using an integrated circuit device. More specifically, the present disclosure relates to an integrated circuit design implemented to perform these operations with low latency and/or a high bandwidth of data. For example, embodiments of a computationally dense digital signal processing (DSP) circuitry, implemented to efficiently perform one or more arithmetic operations (e.g., a dot-product) on an input are disclosed. Moreover, embodiments described herein may relate to layout, design, and data scheduling of a processing element array implemented to compute matrix multiplications (e.g., systolic array multiplication).

Abstract: Systems and methods for calculating a dot product using digital signal processing units that are organized into a dot product processing unit for dot product processing using multipliers and adders of the digital signal processing units.

Abstract: A method for designing a system on a target device includes generating a scheduled netlist and a hardware description language (HDL) of the system from a computer language description of the system. An area report is generated prior to HDL compilation, based on estimates from the scheduled netlist, that identifies resources from the target device required to implement portions of the computer language description of the system.

Abstract: An integrated circuit may include elastic datapaths or pipelines, through which software threads or iterations of loops, may be executed. Throttling circuitry may be coupled along an elastic pipeline in the integrated circuit. The throttling circuitry may include dependency detection circuitry that dynamically detect memory dependency issues that may arise during runtime. To mitigate these dependency issues, the throttling circuitry may assert stall signals to upstream stages in the pipeline. Additionally, the throttling circuitry may control the pipeline to resolve a store operation prior to a corresponding load operation in order to avoid store/load conflicts. In an embodiment, the throttling circuitry may include a validator circuit, a rewind block, a revert block, and a flush block. The throttling circuitry may pass speculative iterations through the rewind block, and later validate the speculative iterations using the validator block.

Abstract: Systems and methods for calculating a dot product using digital signal processing units that are organized into a dot product processing unit for dot product processing using multipliers and adders of the digital signal processing units.

Abstract: Systems and methods for calculating a dot product using digital signal processing units that are organized into a dot product processing unit for dot product processing using multipliers and adders of the digital signal processing units.

Abstract: Systems and methods for dynamically sizing inter-kernel communication channels implemented on an integrated circuit (IC) are provided. Implementation characteristics of the channels, predication, and kernel scheduling imbalances may factor into properly sizing the channels for self-synchronization, resulting in optimized steady-state throughput.

Abstract: Integrated circuits may be programmed using configuration data to implement desired custom logic circuits. The configuration data may be generated using a logic design system. The logic design system may include first and second compilers and an emulation engine. The first compiler may compile a computer program language description of the logic circuit to generate a hardware description language (HDL) description. The emulation engine may emulate performance of the logic circuit when loaded on a target device and may monitor the emulated performance to generate emulated profile data characterizing the emulated performance of the logic circuit. The first compiler may process the emulated profile data to identify optimizations to perform on the logic circuit and may compile an optimized HDL description. The second compiler may compile optimized configuration data using the optimized HDL.