Abstract: An example computing system includes: a processing system, a hardware accelerator coupled to the processing system, and a software platform executing on the processing system. The hardware accelerator includes: a programmable integrated circuit (IC) configured with an acceleration circuit having a static region and a programmable region; a memory in the programmable IC configured to store metadata describing interface circuitry in at least one of the static region and the programmable region of the acceleration circuit. The software platform includes program code executable by the processing system to read the metadata from the memory of the hardware accelerator.

Abstract: Embodiments herein describe techniques for interfacing a neural network application with a neural network accelerator that operate on two heterogeneous computing systems. For example, the neural network application may execute on a central processing unit (CPU) in a computing system while the neural network accelerator executes on a FPGA. As a result, when moving a software-hardware boundary between the two heterogeneous systems, changes may be made to both the neural network application (using software code) and to the accelerator (using RTL). The embodiments herein describe a software defined approach where shared interface code is used to express both sides of the interface between the two heterogeneous systems in a single abstraction (e.g., a software class).

Abstract: Methods and apparatus are described for simultaneously buffering and reformatting (e.g., transposing) a matrix for high-speed data streaming in general matrix multiplication (GEMM), which may be implemented by a programmable integrated circuit (IC). Examples of the present disclosure increase the effective double data rate (DDR) memory throughput for streaming data into GEMM digital signal processing (DSP) engine multifold, as well as eliminate slow data reformatting on a host central processing unit (CPU). This may be accomplished through software-defined (e.g., C++) data structures and access patterns that result in hardware logic that simultaneously buffers and reorganizes the data to achieve linear DDR addressing.

Abstract: The disclosed approaches process a circuit design having first attributes associated with two or more signals or with sources of the two or more signals. The first attributes specify identifier values. The elements of the circuit design are placed on a target integrated circuit (IC), and timing analysis of the circuit design is performed after placing the elements of the circuit design. In response to the first attributes of the two or more signals or sources specifying equivalent identifier values and a path of at least one of the two or more signals or sources being timing-critical, equal numbers of one or more pipeline registers are inserted on paths of the two or more signals or sources.

Abstract: System and method generally relate to protection of a bussed network. In such a system, an access controller is configured for bussed communication via a communication bus to obtain a current transaction. An interface firewall is coupled for bussed communication with the access controller and configured to check for a fault associated with a transfer. A data processing device is coupled for communication with the interface firewall and configured to execute the current transaction to provide the transfer for the interface firewall. The interface firewall is configured to detect the fault associated with the transfer, to block access to the data processing device associated with the fault, and to communicate a blocked status for the data processing device.

Abstract: A system may include a host processor coupled to a communication bus, a first hardware accelerator communicatively linked to the host processor through the communication bus, and a second hardware accelerator communicatively linked to the host processor through the communication bus. The first hardware accelerator and the second hardware accelerator are directly coupled through an accelerator link independent of the communication bus. The host processor is configured to initiate a data transfer between the first hardware accelerator and the second hardware accelerator directly through the accelerator link.

Abstract: A system may include a host processor coupled to a communication bus, a first hardware accelerator communicatively linked to the host processor through the communication bus, and a second hardware accelerator communicatively linked to the host processor through the communication bus. The first hardware accelerator and the second hardware accelerator are directly coupled through an accelerator link independent of the communication bus. The host processor is configured to initiate a data transfer between the first hardware accelerator and the second hardware accelerator directly through the accelerator link.

Abstract: A method of circuit design can include detecting, using a processor, a transactional inefficiency within trace data including transactions involving a first circuit block of a circuit design and, in response to the detecting, generating a modified version of the circuit design by including a transaction converter circuit block within the circuit design. The transaction converter circuit block can be coupled to the first circuit block and can be adapted to correct the transactional inefficiency.

Abstract: The disclosed approaches compile a hierarchical representation of a circuit design into a flattened netlist and store the flattened netlist a memory circuit. The circuit design instantiates a plurality of memory blocks of a target device and specifies logic circuits that access the plurality of memory blocks, respectively. The flattened netlist is modified by determining a subset of the plurality of memory blocks. The quantity of memory reserved in each memory block of the subset is less than a capacity of said each memory block. One memory block is instantiated, for a pair of the memory blocks of the subset, in place of each memory block of the pair in the flattened netlist in the memory circuit. A portion of the flattened netlist that specifies the logic circuits that access each memory block of the pair is modified to access the one memory block instead of each memory block of the pair.

Abstract: Approaches for verifying connectivity of signals in a circuit design include generating a configured version of the circuit design based on input parameter values. The configured version specifies connections from source pins of ports of circuit blocks of the configured version to destination pins of ports of circuit blocks. Expected source-destination connections between source pins and destination pins of the ports of the circuit blocks of the configured version are determined from the input parameter values. A connectivity checker that includes HDL code is generated based on the expected source-destination connections. For each of the expected source-destination connections, the HDL code forces a first signal value on a source pin of the expected source-destination connection in the configured version of the circuit design and determines whether or not a second signal value at a destination pin of the expected source-destination connection matches the first signal value.

Abstract: Approaches for testing a module of a circuit design include tagging flip-flops in a netlist of the module with respective path names of the flip-flops from a hardware description language specification of the module. In simulating with the netlist, event data are captured to a first file. A process determines whether or not event data in the first file matches event data in a second file of event data. In response to a difference determined between the first file and the second file, an earliest occurrence of an event in the first file having an associated signal value of a first signal that does not match an associated signal value of a corresponding event in the second file is determined. The one of the plurality of flip-flops that output the first signal is determined, and the respective path name of the one flip-flop is output.

Abstract: Methods for generating a circuit design are disclosed. A plurality of cells is instantiated in the circuit design in response to user input. The set of interface parameters of each cell is arranged into a hierarchy of interface levels as indicated by an interface model corresponding to the cell. For each of the interface levels, values of the sets of interface parameters of cells included in the interface level are respectively propagated to other cells directly connected to the cell. In response to propagating a value of an interface parameter from another cell of the plurality of cells to the cell and the cell having a value of the corresponding interface parameter that is different from the propagated value, a value for the corresponding interface parameter of the cell is determined using a respective propagation function associated with the corresponding interface level.

Abstract: Method and apparatus for controlling access to memory circuitry is described. In one example, access to the memory circuitry is controlled among a plurality of bus interfaces of a data processing system. A plurality of ports is respectively coupled to said plurality of bus interfaces. Arbitration logic is configured for communication with the plurality of ports. The arbitration logic arbitrates access to the memory circuitry among the plurality of bus interfaces on a time shared basis.