Embodiments may be generally direct to apparatuses, systems, method, and techniques to determine a configuration for a plurality of connectors, the configuration to associate a first interconnect protocol with a first subset of the plurality of connectors and a second interconnect protocol with a second subset of the plurality of connectors, the first interconnect protocol and the second interconnect protocol are different interconnect protocols and each comprising one of a serial link protocol, a coherent link protocol, and an accelerator link protocol, cause processing of data for communication via the first subset of the plurality of connectors in accordance with the first interconnect protocol, and cause processing of data for communication via the second subset of the plurality of connector in accordance with the second interconnect protocol.

An input output scheduler. The scheduler runs in user space and is associated with one core of a multi-core central processing unit. Applications submit input output commands to the scheduler, which queues the input output commands and submits them in batches to a mass storage device. The input output scheduler may include a plurality of command queues with different batching strategies configured to provide, e.g., different performance characteristics as measured, for example, by latency or input output throughput.

Apparatuses and methods for performing memory operations are described. An example apparatus includes a memory operation controller. The memory operation controller is configured to receive memory instructions and decode the same to provide internal signals for performing memory operations for the memory instructions. The memory operation controller is further configured to provide information indicative of a time for a variable latency period of a memory instruction during the variable latency period. In an example method, a write instruction and an address to which write data is to be written is received at a memory and an acknowledgement indicative of an end of a variable latency period for the write instruction is provided. After waiting a variable bus turnaround after the acknowledgement, write data for the write instruction is received.

Systems, apparatuses, and methods for performing efficient memory accesses for a computing system are disclosed. In various embodiments, a computing system includes a computing resource and a memory controller coupled to a memory device. The computing resource selectively generates a hint that includes a target address of a memory request generated by the processor. The hint is sent outside the primary communication fabric to the memory controller. The hint conditionally triggers a data access in the memory device. When no page in a bank targeted by the hint is open, the memory controller processes the hint by opening a target page of the hint without retrieving data. The memory controller drops the hint if there are other pending requests that target the same page or the target page is already open.

Managing input/output (‘I/O’) queues in a data storage system, including: receiving, by a host that is coupled to a plurality of storage devices via a storage network, a plurality of I/O operations to be serviced by a target storage device; determining, for each of a plurality of paths between the host and the target storage device, a data transfer maximum associated with the path; determining, for one or more of the plurality of paths, a cumulative amount of data to be transferred by I/O operations pending on the path; and selecting a target path for transmitting one or more of the plurality of I/O operations to the target storage device in dependence upon the cumulative amount of data to be transferred by I/O operations pending on the path and the data transfer maximum associated with the path.

A data storage device includes a nonvolatile memory device; a buffer memory for storing temporarily data to be transmitted from the nonvolatile memory device to a host device or data to be transmitted from the host device to the nonvolatile memory device; a memory control unit for performing a control operation for controlling the nonvolatile memory device; and a direct memory access (DMA) unit for performing a data transmission operation associated with the buffer memory, according to control of the memory control unit, wherein the DMA block transmits a first data from the nonvolatile memory device to the buffer memory, and wherein the DMA unit transmits a second data from the nonvolatile memory device to the buffer memory, while the first data stored in the buffer memory is transmitted from the buffer memory to the host device.

Methods, systems, and apparatus for implementing low latency interconnect switches between CPU's and associated protocols. CPU's are configured to be installed on a main board including multiple CPU sockets linked in communication via CPU socket-to-socket interconnect links forming a CPU socket-to-socket ring interconnect. The CPU's are also configured to transfer data between one another by sending data via the CPU socket-to-socket interconnects. Data may be transferred using a packetized protocol, such as QPI, and the CPU's may also be configured to support coherent memory transactions across CPU's.

A system having multiple devices that can host different versions of an artificial neural network (ANN) as well as different versions of a feature dictionary. In the system, encoded inputs for the ANN can be decoded by the feature dictionary, which allows for encoded input to be sent to a master version of the ANN over a network instead of an original version of the input which usually includes more data than the encoded input. Thus, by using the feature dictionary for training of a master ANN there can be reduction of data transmission.

Apparatuses and methods for variable latency memory operations are disclosed herein. An example apparatus may include a memory configured to receive an activate command indicative of a type of a command during a first addressing phase and to receive the command during a second addressing phase. The memory may further be configured to provide information indicating that the memory is not available to perform a command responsive, at least in part, to receiving the command during a variable latency period and to provide information indicating that the memory is available to perform a command responsive, at least in part, to receiving the command after the variable latency period.

A method of managing a memory bus includes identifying sub-operations required for execution of commands, maintaining a list of released sub-operations containing only released unexecuted sub-operations directed to individual dies that are identified as available, accessing the dies until the list is empty, subsequently, polling to identify dies that are available, and subsequently resuming accessing the dies by executing only sub-operations from the list until the list is empty.