Abstract: A system can include a plurality of processors. Each processor of the plurality of processors can be configured to execute program code. The system can include a direct memory access system configured for multi-processor operation. The direct memory access system can include a plurality of data engines coupled to a plurality of interfaces via a plurality of switches. The plurality of switches can be programmable to couple different ones of the plurality of data engines to different ones of the plurality of processors for performing direct memory access operations based on a plurality of host profiles corresponding to the plurality of processors.

Abstract: A system can include a plurality of processors. Each processor of the plurality of processors can be configured to execute program code. The system can include a direct memory access system configured for multi-processor operation. The direct memory access system can include a plurality of data engines coupled to a plurality of interfaces via a plurality of switches. The plurality of switches can be programmable to couple different ones of the plurality of data engines to different ones of the plurality of processors for performing direct memory access operations based on a plurality of host profiles corresponding to the plurality of processors.

Abstract: A direct memory access (DMA) system can include a memory configured to store a plurality of host profiles, a plurality of interfaces, wherein two or more of the plurality of interfaces correspond to different ones of a plurality of host processors, and a plurality of data engines coupled to the plurality of interfaces. The plurality of data engines are independently configurable to access different ones of the plurality of interfaces for different flows of a DMA operation based on the plurality of host profiles.

Abstract: A circuit arrangement includes a memory circuit, data upload circuitry, data formatting circuitry, and a systolic array (SA). The data upload circuitry inputs a multi-dimensional data set and stores the multi-dimensional data set in the memory circuit. The data formatting circuitry reads subsets of the multi-dimensional data set from the memory circuit. The data formatting circuitry arranges data elements of the subsets into data streams, and outputs data elements in the data streams in parallel. The SA includes rows and columns of multiply-and-accumulate (MAC) circuits. The SA inputs data elements of the data streams to columns of MAC circuits in parallel, inputs filter values to rows of MAC circuits in parallel, and computes an output feature map from the data streams and the filter values.

Abstract: A device includes a platform implemented, at least in part, in a static region of programmable circuitry and a dynamic region of programmable circuitry configured to implement user-specified circuitry in communication with the platform. The platform is configured to establish and maintain a first communication link with a host data processing system and a second communication link with a network while at least a portion of the dynamic region of programmable circuitry is dynamically reconfigured.

Abstract: Examples herein describe techniques for providing a customizable direct memory access (DMA) interface which can permit user logic to change or control how DMA read and writes are performed. In one example, a DMA engine may be hardened (e.g., include circuitry formed from a semiconductor material) which prevents the DMA engine from being reconfigured like programmable logic. Instead of changing the DMA engine, the user logic can change or customize the DMA interface between the user logic and the DMA engine. In this way, the manner in which the DMA engine performs DMA write and reads can be changed by the user logic. In one example, the DMA engine includes a bypass mode of operation where descriptors associated with DMA queues are passed through the DMA engine and to the user logic.

Abstract: A device includes a platform implemented, at least in part, in a static region of programmable circuitry and a dynamic region of programmable circuitry configured to implement user-specified circuitry in communication with the platform. The platform is configured to establish and maintain a first communication link with a host data processing system and a second communication link with a network while at least a portion of the dynamic region of programmable circuitry is dynamically reconfigured.

Abstract: An integrated circuit (IC) includes a first kernel circuit implemented in programmable circuitry, a second kernel circuit implemented in programmable circuitry, and a stream traffic manager circuit coupled to the first kernel circuit and the second kernel circuit. The stream traffic manager circuit is configured to control data streams exchanged between the first kernel circuit and the second kernel circuit.

Abstract: A circuit arrangement includes an encryption circuit and a decryption circuit. A cryptographic shell circuit has a transmit channel and a receive channel in parallel with the transmit channel. The transmit channel includes an encryption interface circuit coupled to the encryption circuit. The encryption interface circuit determines first cryptographic parameters based on data in a plaintext input packet and inputs the first cryptographic parameters and plaintext input packet to the encryption circuit. The receive channel includes a decryption interface circuit coupled to the decryption circuit. The decryption interface circuit determines second cryptographic parameters based on data in a ciphertext input packet and inputs the second cryptographic parameters and ciphertext input packet to the decryption circuit.

Abstract: An integrated circuit (IC) includes a first kernel circuit implemented in programmable circuitry, a second kernel circuit implemented in programmable circuitry, and a stream traffic manager circuit coupled to the first kernel circuit and the second kernel circuit. The stream traffic manager circuit is configured to control data streams exchanged between the first kernel circuit and the second kernel circuit.

Abstract: A circuit arrangement includes an encryption circuit and a decryption circuit. A cryptographic shell circuit has a transmit channel and a receive channel in parallel with the transmit channel. The transmit channel includes an encryption interface circuit coupled to the encryption circuit. The encryption interface circuit determines first cryptographic parameters based on data in a plaintext input packet and inputs the first cryptographic parameters and plaintext input packet to the encryption circuit. The receive channel includes a decryption interface circuit coupled to the decryption circuit. The decryption interface circuit determines second cryptographic parameters based on data in a ciphertext input packet and inputs the second cryptographic parameters and ciphertext input packet to the decryption circuit.

Abstract: Examples herein describe techniques for providing a customizable direct memory access (DMA) interface which can permit user logic to change or control how DMA read and writes are performed. In one example, a DMA engine may be hardened (e.g., include circuitry formed from a semiconductor material) which prevents the DMA engine from being reconfigured like programmable logic. Instead of changing the DMA engine, the user logic can change or customize the DMA interface between the user logic and the DMA engine. In this way, the manner in which the DMA engine performs DMA write and reads can be changed by the user logic. In one example, the DMA engine includes a bypass mode of operation where descriptors associated with DMA queues are passed through the DMA engine and to the user logic.

Abstract: A system can include a first memory section comprising a plurality of volatile memory cells; a second memory section comprising a plurality of nonvolatile memory cells; a first data path configured to transfer data between the first and second memory sections; an interface circuit coupled to receive access commands and address values, the interface circuit configured to determine if a data transfer operation is occurring in the device, and if the data transfer operation is occurring, accessing the address in the first memory section or accessing a location in the second memory section based on a select value, and if the data transfer operation is not occurring, accessing the address in the first memory section; and a compare circuit configured to compare a received address to a predetermined value to generate the select value.

Abstract: Structures and methods for improving logging in network structures are disclosed herein. In one embodiment, an apparatus can include: (i) a network interface card (NIC) configured to receive data, to transmit data, and to send data for logging; (ii) a memory log coupled to the NIC, where the memory log comprises non-volatile memory (NVM) configured to write the data sent for logging from the NIC; and (iii) where the data being sent for logging by the memory log occurs substantially simultaneously with the data being received by the NIC, and the data being transmitted from the NIC.

Abstract: A memory device can include at least two ports for transferring data to and from the memory device; and plurality of memory cells, each memory cell including at least one element programmable between different impedance states, and a plurality of access devices, each access device providing a current path between the element and a different one of the ports.

Abstract: A system can include a first memory section comprising a plurality of volatile memory cells accessible via a first data path having a first bit width; a second memory section comprising a plurality of programmable impedance memory cells, each having at least one solid electrolyte layer; and a second data path configured to transfer data between the first and second memory sections independent of the first data path, the second data path having a greater bit width than the first data path.

Abstract: Structures and methods for encoding data to reduce write cycles in a semiconductor memory device are disclosed herein. In one embodiment, a method of writing data to a semiconductor memory device can include: (i) determining a number of significant bits for data to be written in the semiconductor memory device; (ii) determining a tag associated with the data to be written in the semiconductor memory device, where the tag is determined based on the determined number of significant bits; (iii) encoding the data when the tag has a first state, where the tag is configured to indicate data encoding that comprises using N bits of the encoded data to store M bits of the data, where M and N are both positive integers and N is greater than M; and (iv) writing the encoded data and the tag in the semiconductor memory device.

Abstract: A memory device can include a plurality of memory elements programmable between different impedance states; and circuits configured to apply first electrical conditions to one group of memory elements and second electrical conditions, different from the first electrical conditions, to another group of memory elements to vary a speed of an access operation to the different groups of memory elements.

Abstract: Methods, systems, and circuits for forming and operating a global hierarchical clock tree are described. The global hierarchical clock tree may comprise a clock circuit that operates to provide clock signals to a core circuit surrounded by the clock circuit. The clock circuit may include two or more first and second clock generator modules to generate a first and a second set of clock signals, respectively. The first and second clock modules may be located so that the first set of clock signals experience approximately equal first latencies and the second set of clock signals experience approximately equal second latencies. Additional methods, systems, and circuits are disclosed.

Abstract: A memory device can include a plurality of physical blocks that each include a number of memory elements programmable between at least two different impedance states, the memory elements being subject to degradation in performance; and bias circuits configured to applying healing electrical conditions to at least one spare physical block that does not contain valid data; wherein the healing electrical conditions are different from write operation electrical conditions, and reverse degradation of the memory elements of the at least one spare physical block.