Abstract: Implementing data flows of an application across a memory hierarchy of a data processing array includes receiving a data flow graph specifying an application for execution on the data processing array. A plurality of buffer objects corresponding to a plurality of different levels of the memory hierarchy of the data processing array and an external memory are identified. The plurality of buffer objects specify data flows. Buffer object parameters are determined. The buffer object parameters define properties of the data flows. Data that configures the data processing array to implement the data flows among the plurality of different levels of the memory hierarchy and the external memory is generated based on the plurality of buffer objects and the buffer object parameters.

Abstract: Modeling and compiling tensor processing applications using multi-layer adaptive data flow (ML-ADF) graphs, including folding the ML-ADF graph for temporal sharing of platform resources, computing schedules for runtime orchestration of kernel execution, memory reuse, tensor and sub-volume movement, and dataflow synchronization, and generating binary code for processors of the target computing platform and re-targetable controller code. The ML-ADF graph may represent: tensor processing of a layer of a neural network as data flow through the data nodes and distribution to compute tiles across memory hierarchy; data flow amongst layers of the neural network using connections amongst data nodes of the respective layers; and multi-dimension data partitioning and distribution using tiling parameters associated with ports of the data nodes.

Abstract: Modeling and compiling tensor processing applications using multi-layer adaptive data flow (ML-ADF) graphs, including folding the ML-ADF graph for temporal sharing of platform resources, computing schedules for runtime orchestration of kernel execution, memory reuse, tensor and sub-volume movement, and dataflow synchronization, and generating binary code for processors of the target computing platform and re-targetable controller code. The ML-ADF graph may represent: tensor processing of a layer of a neural network as data flow through the data nodes and distribution to compute tiles across memory hierarchy; data flow amongst layers of the neural network using connections amongst data nodes of the respective layers; and multi-dimension data partitioning and distribution using tiling parameters associated with ports of the data nodes.

Abstract: Implementing data flows of an application across a memory hierarchy of a data processing array includes receiving a data flow graph specifying an application for execution on the data processing array. A plurality of buffer objects corresponding to a plurality of different levels of the memory hierarchy of the data processing array and an external memory are identified. The plurality of buffer objects specify data flows. Buffer object parameters are determined. The buffer object parameters define properties of the data flows. Data that configures the data processing array to implement the data flows among the plurality of different levels of the memory hierarchy and the external memory is generated based on the plurality of buffer objects and the buffer object parameters.

Abstract: An example method of implementing an application for a system-on-chip (SOC) having a data processing engine (DPE) array includes determining a graph representation of the application, the graph representation including nodes representing kernels of the application and edges representing communication between the kernels, mapping, based on the graph, the kernels onto DPEs of the DPE array and data structures of the kernels onto memory in the DPE array, routing communication channels between DPEs and circuitry of the application configured in programmable logic of the SOC, and generating implementation data for programming the SOC to implement the application based on results of the mapping and the routing.

Abstract: An example method of implementing an application for a system-on-chip (SOC) having a data processing engine (DPE) array includes determining a graph representation of the application, the graph representation including nodes representing kernels of the application and edges representing communication between the kernels, mapping, based on the graph, the kernels onto DPEs of the DPE array and data structures of the kernels onto memory in the DPE array, routing communication channels between DPEs and circuitry of the application configured in programmable logic of the SOC, and generating implementation data for programming the SOC to implement the application based on results of the mapping and the routing.

Abstract: An example method of implementing an application for a system-on-chip (SOC) having a data processing engine (DPE) array includes obtaining a graph representation of the application, the graph representation including nodes representing kernels of the application and edges representing communication between the kernels, mapping, based on the graph, the kernels onto DPEs of the DPE array and data structures of the kernels onto memory in the DPE array, routing communication channels between DPEs and circuitry of the application configured in programmable logic of the SOC, adding at least one first-in-first-out (FIFO) buffer to at least one of the communication channels, and generating implementation data for programming the SOC to implement the application based on results of the mapping and the routing.

Abstract: Systems and techniques are described for producing a synthesized IC design that includes design-for-testability (DFT) circuitry. A register-transfer-level (RTL) representation of an IC design can be received, wherein the RTL representation includes functional logic. Next, DFT logic can be added to the RTL representation, and DFT placement guidance for placing the DFT logic can be generated. Synthesis can be performed on the RTL representation to obtain the synthesized IC design, wherein during synthesis, (1) the functional logic and the DFT logic can be placed, wherein the DFT logic is placed based on the DFT placement guidance, (2) scan chains can be inserted and placed, and (3) the DFT logic can be electrically connected with the scan chains.

Abstract: Systems and techniques are described for producing a synthesized IC design that includes design-for-testability (DFT) circuitry. A register-transfer-level (RTL) representation of an IC design can be received, wherein the RTL representation includes functional logic. Next, DFT logic can be added to the RTL representation, and DFT placement guidance for placing the DFT logic can be generated. Synthesis can be performed on the RTL representation to obtain the synthesized IC design, wherein during synthesis, (1) the functional logic and the DFT logic can be placed, wherein the DFT logic is placed based on the DFT placement guidance, (2) scan chains can be inserted and placed, and (3) the DFT logic can be electrically connected with the scan chains.

Abstract: An embodiment of the present invention includes a pipeline comprising a plurality of stages and a pipeline timing controller controlling a plurality of predetermined delays, wherein, when one of the predetermined delays has expired, the pipeline timing controller sends a control signal to initiate at least one process within associated ones of the plurality of stages.

Abstract: A method for determining validity of a proposed loop iteration schedule comprising the steps of receiving a dependence graph including operations and edges between said operations; receiving a performance specification; receiving an assignment of latencies to operations of said dependence graph; and determining existence of a placement of clock cycle-boundaries in said dependence graph such that all dependence and timing constraints are satisfied for the performance specification.

Abstract: A method for performing a global timing analysis of a proposed digital circuit comprising receiving timing models and the proposed digital circuit; determining at least one mode of circuit operation of the proposed digital circuit; deriving a sub-circuit corresponding to each of at least one mode of circuit operation; performing timing analysis on each of the sub-circuits derived corresponding to each of the modes; and combining the timing analysis results for all of the modes to determine an overall maximum circuit delay.

Abstract: A method of considering circuit timing requirements during the circuit design process, comprising receiving a clock cycle-time constraint; receiving delay characteristics of hardware resources from a macrocell library; receiving an operation, an alternative clock cycle associated with said operation and an alternative hardware resource associated with said operation; and determining validity of the received alternative with respect to timing constraints using a hardware structural representation of the program graph.

Abstract: A technique for synthesizing digital circuit designs by incorporating timing convergence and routability considerations. In one aspect, the invention provides a system and programmatic method for generating a circuit design from a functional specification according to at least one design objective. An intermediate representation of the functional specification is formed. The intermediate representation is analyzed for identifying a physical instantiation that will possibly result in unacceptable interconnect delay or congestion. Functional units are allocated from among a plurality of candidate functional units for performing operations of the intermediate representation. Operations are scheduled to occur at specified times on said selected functional units. An architectural representation of the circuit design is formed according to results of scheduling.

Abstract: An embodiment of the present invention includes a pipeline comprising a plurality of stages and a pipeline timing controller controlling a plurality of predetermined delays, wherein, when one of the predetermined delays has expired, the pipeline timing controller sends a control signal to initiate at least one process within associated ones of the plurality of stages.

Abstract: A technique for synthesizing digital circuit designs by incorporating timing convergence and routability considerations. In one aspect, the invention provides a system and programmatic method for generating a circuit design from a functional specification according to at least one design objective. An intermediate representation of the functional specification is formed. The intermediate representation is analyzed for identifying a physical instantiation that will possibly result in unacceptable interconnect delay or congestion. Functional units are allocated from among a plurality of candidate functional units for performing operations of the intermediate representation. Operations are scheduled to occur at specified times on said selected functional units. An architectural representation of the circuit design is formed according to results of scheduling.

Abstract: A method of considering circuit timing requirements during the circuit design process, comprising receiving a clock cycle-time constraint; receiving delay characteristics of hardware resources from a macrocell library; receiving an operation, an alternative clock cycle associated with said operation and an alternative hardware resource associated with said operation; and determining validity of the received alternative with respect to timing constraints using a hardware structural representation of the program graph.

Abstract: A method for determining validity of a proposed loop iteration schedule comprising the steps of receiving a dependence graph including operations and edges between said operations; receiving a performance specification; receiving an assignment of latencies to operations of said dependence graph; and determining existence of a placement of clock cycle-boundaries in said dependence graph such that all dependence and timing constraints are satisfied for the performance specification.

Abstract: A method for performing a global timing analysis of a proposed digital circuit comprising receiving timing models and the proposed digital circuit; determining at least one mode of circuit operation of the proposed digital circuit; deriving a sub-circuit corresponding to each of at least one mode of circuit operation; performing timing analysis on each of the sub-circuits derived corresponding to each of the modes; and combining the timing analysis results for all of the modes to determine an overall maximum circuit delay.