An accelerator is disclosed. A tier storage may store data. A circuit may process the data to produce a processed data. The accelerator may load the data from a device using a cache-coherent interconnect protocol.

A novel and useful neural network (NN) processing core incorporating inter-device connectivity and adapted to implement artificial neural networks (ANNs). A chip-to-chip interface spreads a given ANN model across multiple devices in a seamless manner. The NN processor is constructed from self-contained computational units organized in a hierarchical architecture. The homogeneity enables simpler management and control of similar computational units, aggregated in multiple levels of hierarchy. Computational units are designed with minimal overhead as possible, where additional features and capabilities are aggregated at higher levels in the hierarchy. On-chip memory provides storage for content inherently required for basic operation at a particular hierarchy and is coupled with the computational resources in an optimal ratio. Lean control provides just enough signaling to manage only the operations required at a particular hierarchical level. Dynamic resource assignment agility is provided which can be adjusted as required depending on resource availability and capacity of the device.

A semiconductor device includes a first memory controller configured to output a first control signal to first and second external memories through a first memory interface, a second memory controller configured to output a second control signal to the second external memory through a second memory interface, an inter-device interface for communicating with another semiconductor device, terminals configured to output the second control signal that has passed through the second memory interface, and a first selector configured to select between the second memory interface and the inter-device interface in accordance with an operation mode of the semiconductor device and to couple the selected interface to the terminals.

A method and system for permitting a guest to program a message-signaled interrupt-based device is disclosed. A hypervisor of a host detects a request by a guest to map an address range of memory of the guest to a message signaled-interrupt capability table associated with a device. The hypervisor maps the message signaled-interrupt capability table from a message signaled-interrupt capability register of a programmable interrupt controller associated with the host to the address range of memory of the guest. The hypervisor detects an attempt by the guest to program the device with the message-signaled interrupt configuration located in the address range of memory of the guest. The hypervisor programs the device with the message-signaled interrupt configuration specified by the guest in the address range of memory of the guest.

Aspects of the embodiments are directed to systems and methods for providing and using hints in data packets to perform memory transaction optimization processes prior to receiving one or more data packets that rely on memory transactions. The systems and methods can include receiving, from a device connected to the root complex across a PCIe-compliant link, a data packet; identifying from the received device a memory transaction hint bit; determining a memory transaction from the memory transaction hint bit; and performing an optimization process based, at least in part, on the determined memory transaction.

A method of accessing a server address space of a shared PCIe end point system includes programming a primary address translation table with a server address of a server address space, setting up a direct memory access (DMA) to access a primary port memory map, the primary port memory map correlating with addresses in the primary address translation table, and re-directing the direct memory accesses to the primary port memory map to the server address space according to the primary address translation table.

A method comprising: receiving, at a vector processor, a request to store data; performing, by the vector processor, one or more transforms on the data; and directly instructing, by the vector processor, one or more storage device to store the data; wherein performing one or more transforms on the data comprises: erasure encoding the data to generate n data fragments configured such that any k of the data fragments are usable to regenerate the data, where k is less than n; and wherein directly instructing one or more storage device to store the data comprises: directly instructing the one or more storage devices to store the plurality of data fragments.

Method and system embodying the method for a general address transformation for an access to a shared memory comprising at least one tile and each tile comprising at least one memory bank, comprising selecting a mode of a general address transformation; providing a general address comprising a plurality of bits by at least one of a plurality of devices; and transforming the general address onto a transformed address according to the selected mode; wherein in a first selected mode the transforming comprises determining each of a plurality of bits of a transformed address as an exclusive or of at least two bits of the plurality of bits of the general address provided that the shared memory comprises a plurality of tiles, and/or each tile comprises a plurality of banks is disclosed.

A multiple processor architecture with flexible external input/output interface is provided. In one embodiment, an open flexible processor architecture avionics device comprises: a multiple processor architecture having a primary processor, a secondary processor, a random access memory (RAM) coupled to at least the secondary processor, and a shared memory coupled to the primary and secondary processor; and a flexible input/output (I/O) interface coupled to the multiple processor architecture, wherein the flexible I/O interface provides I/O access to the primary processor using a fixed I/O protocol, and provides I/O access to the secondary processor using at least one re-configurable I/O protocol; wherein the primary processor is dedicated to executing embedded software for implementing a primary base functionality, the primary processor has read and write access to the shared memory, and the primary processor is not reprogrammable; and wherein the secondary processor has read-only access to the shared memory and is programmable.

An adaptive interface high availability storage device. In some embodiments, the adaptive interface high availability storage device includes: a rear storage interface connector; a rear multiplexer, connected to the rear storage interface connector; an adaptable circuit connected to the rear multiplexer; a front multiplexer, connected to the adaptable circuit; and a front storage interface connector, connected to the front multiplexer. The adaptive interface high availability storage device may be configured to operate in a single-port state or in a dual-port state. The adaptive interface high availability storage device may be configured: in the single-port state, to present a single-port host side storage interface according to a first storage protocol at the rear storage interface connector, and in the dual-port state, to present a dual-port host side storage interface according to the first storage protocol at the rear storage interface connector.