In response to an instruction decoder decoding a range prefetch instruction specifying first and second address-range-specifying parameters and a stride parameter, prefetch circuitry controls, depending on the first and second address-range-specifying parameters and the stride parameter, prefetching of data from a plurality of specified ranges of addresses into the at least one cache. A start address and size of each specified range is dependent on the first and second address-range-specifying parameters. The stride parameter specifies an offset between start addresses of successive specified ranges. Use of the range prefetch instruction helps to improve programmability and improve the balance between prefetch coverage and circuit area of the prefetch circuitry.

Methods, systems, apparatuses, and computer-readable storage mediums described herein are directed to techniques for efficient data encoding for neural network training. In particular, the embodiments described herein train a DNN based on a selective encoding (e.g., compressing) of data structures that are generated during training. For example, multiple training sessions may be performed where, in each training session, a different set of data structures performed by various operators of the DNN are encoded. Memory allocation information generated based on each training session is analyzed to determine which combination of encoded data structures results in a reduction of memory required to train the DNN.

The present application provides a data communication method, a communication system and a computer-readable storage medium. The method comprises: acquiring, by a data production module, target data to be sent to a data consumption module; determining in a preset GPU shared memory, by the data production module, a target memory block into which the target data is to be written, wherein the GPU shared memory is a predetermined GPU memory for data communication between the data production module and the data consumption module; writing, by the data production module, the target data into the target memory block to obtain memory address information corresponding to the target data; and sending, by the data production module, the memory address information to the data consumption module so that the data consumption module is operable to access the target data based on the memory address information.

A memory system includes a memory device suitable for storing data and a controller suitable for determining an operation state of the memory device and carrying out garbage collection to the memory device in response to the operation state. The controller can ignore a first command entered from a host while performing the garbage collection.

A memory module includes cache of relatively fast and durable dynamic, random-access memory (DRAM) in service of a larger amount of relatively slow and wear-sensitive nonvolatile memory. Local controller manages communication between the DRAM cache and nonvolatile memory to accommodate disparate access granularities, reduce the requisite number of memory transactions, and minimize the flow of data external to nonvolatile memory components.

A data storage device includes a memory device and a controller coupled to the memory device. The data storage device supports zoned namespace. The controller is configured to maintain a zone timestamp table that includes a corresponding timestamp for each zone and add a timestamp to each garbage collection block of the memory device. The controller is further configured to scan a garbage collection block from a last physical block address (PBA) entry to a first PBA entry, determine a zone timestamp for the scanned PBA entry, and compare the zone timestamp to a timestamp of the garbage collection block. The controller is further configured to create and maintain a zone timestamp table and create and maintain a zone based defragmentation table.

A storage device includes a nonvolatile memory, a volatile memory, and a controller accesses the nonvolatile memory using an address conversion table including regions, each region including entries, each entry storing a physical address of the nonvolatile memory in association with a logical address, and reads and writes data of the address conversion table from and to the nonvolatile memory and the volatile memory in a unit of a frame. The controller writes, to the nonvolatile memory, data of a first region in a first format in which a head address of data of a region aligns with a head address of a frame, and writes, to the volatile memory, data of a second region in either the first format or a second format in which a head address of data of a region does not align with a head address of a frame.

Virtual memory space may be saved in a clone environment by leveraging the similarity of the data signatures in swap files when a chain of virtual machines (VMs) includes clones spawned from a common parent and executing common applications. Deduplication is performed across the chain, rather than merely within each VM. Examples include generating a common deduplication identifier (ID) for the chain; generating a logical addressing table linked to the deduplication ID, for each of the VMs in the chain; and generating a hash table for the chain. Examples further include, based at least on a swap out request, generating a hash value for a block of memory to be written to a storage medium; and based at least on finding the hash value within the hash table, updating the logical addressing table to indicate a location of a prior-existing duplicate of the block on the storage medium.

A data storage device and method for auto-peeling of surveillance video content to increase archival storage is provided. In one embodiment, a data storage device is provided comprising a memory and a controller. The controller is configured to determine that available storage space in the memory is less than a threshold; in response to determining that the available storage space in the memory is less than the threshold: read a video file from the memory; and re-encode the video file to decrease a size of the video file, wherein re-encoding the video file increases available storage space in the memory without deleting the video file. Other embodiments are provided.

A storage system and method for automatic data phasing are disclosed. In one embodiment, a storage system is configured to receive, from a host, data to be written in the memory and an indication of an expected lifespan of the data; and determine whether to perform a garbage collection operation on the data based on the expected lifespan of the data. Other embodiments are provided.