A storage device is configured to manage a plurality of nonvolatile memories with a plurality of physical streams. An operation method of the storage device includes receiving an input/output request from an external host device, determining a 0-th virtual stream identifier, extracting a 0-th representative value from a 0-th virtual stream feature, extracting a first and second representative values corresponding to first and second physical streams, calculating distance information including first and second similarities between the 0-th virtual stream and each of the first and second physical streams, based on the extracted representative values, assigning one of the plurality of physical streams to the 0-th virtual stream, based on the distance information, and performing an operation corresponding to the input/output request, at the assigned physical stream, and the extracting and the calculating are performed by using machine learning model.

A method and apparatus for determining the sector size and concomitant host metadata size to determine the difference between total size of the data block to be stored, and using the difference for parity data. This allows an increase in parity bits available for smaller sector sizes and/or data with smaller host metadata sizes. Because the amount of space available for additional parity bits is known, data with lower numbers of parity bits may be assigned to higher quality portions a memory array written with longer programming trim times, and/or written to memory dies with good redundant columns, further increasing performance and reliability.

An apparatus includes a processor core and a memory hierarchy. The memory hierarchy includes main memory and one or more caches between the main memory and the processor core. A plurality of hardware pre-fetchers are coupled to the memory hierarchy and a pre-fetch control circuit is coupled to the plurality of hardware pre-fetchers. The pre-fetch control circuit is configured to compare changes in one or more cache performance metrics over two or more sampling intervals and control operation of the plurality of hardware pre-fetchers in response to a change in one or more performance metrics between at least a first sampling interval and a second sampling interval.

A streaming engine employed in a digital data processor specifies a fixed read only data stream defined by plural nested loops. An address generator produces address of data elements. A steam head register stores data elements next to be supplied to functional units for use as operands. The streaming engine stores an early address of next to be fetched data elements and a late address of a data element in the stream head register for each of the nested loops. The streaming engine stores an early loop counts of next to be fetched data elements and a late loop counts of a data element in the stream head register for each of the nested loops.

Disclosed are systems and methods that determine whether instances of data (e.g., forward activations, backward derivatives of activations) that are used to train deep neural networks are to be stored on-chip or off-chip. The disclosed systems and methods are also used to prune the data (discard or delete selected instances of data). A system includes a hierarchical arrangement of on-chip and off-chip memories, and also includes a hierarchical arrangement of data selector devices that are used to decide whether to discard data and where in the system the data is to be discarded.

A memory module includes a device memory configured to store data and including a first memory area and a second memory area, and a controller including an accelerator circuit. The controller is configured to control the device memory, transmit a command to exclude the first memory area from the system memory map to a host processor in response to a mode change request, and modify a memory configuration register to exclude the first memory area from the memory configuration register. The accelerator circuit is configured to use the first memory area to perform an acceleration operation.

According to an embodiment of the present disclosure, a storage device may include a memory device, and a memory controller configured to receive a read command from an external host and control the memory device according to the read command, wherein the read command may include a basic header segment commonly included in commands transferred between the external host and the memory controller and including information indicating that the read command is a command for requesting data stored in the memory device, a transaction specific field including information indicating that the read command is a read command for at least two or more logical addresses, and an extra header segment including information on the at least two or more logical addresses.

A method is described, which includes receiving, by a memory subsystem controller from a host system, a host read memory command that references a set of logical block addresses associated with a set of transfer units of a memory device. The controller converts the set of logical block addresses to a set of physical block addresses for the set of transfer units; generates a set of device read memory commands based on the physical block addresses, wherein each device read memory command references at least one physical block address; and generates a first aggregated device read memory command based on a first device read memory command and a second read memory command in response to determining that the first device read memory command is associated with the second device read memory command. The controller thereafter transmits the first aggregated device read memory command to the memory device.

A technical solution to the technical problem of how to support memory-centric operations on cached data uses a novel memory-centric memory operation that invokes write back functionality on cache controllers and memory controllers. The write back functionality enforces selective flushing of dirty, i.e., modified, cached data that is needed for memory-centric memory operations from caches to the completion level of the memory-centric memory operations, and updates the coherence state appropriately at each cache level. The technical solution ensures that commands to implement the selective cache flushing are ordered before the memory-centric memory operation at the completion level of the memory-centric memory operation.

An electronic device includes a cache memory and a controller. The cache memory includes a set of cache blocks, each cache block having a number of locations usable for storing cache lines. The cache memory also includes a separate set of error correction code (ECC) bits for each of the locations. The controller stores a victim cache line, evicted from a first location in the cache block, in a second location in the cache block. The controller next stores victim reference information in a portion of the plurality of ECC bits for the first location, the victim reference information indicating that the victim cache line is stored in the second location.