Abstract: Disclosed herein includes a system, a method, and a device for reading and writing sparse data in a neural network accelerator. A mask identifying byte positions within a data word having non-zero values in memory can be accessed. Each bit of the mask can have a first value or a second value, the first value indicating that a byte of the data word corresponds to a non-zero byte value, the second value indicating that the byte of the data word corresponds to a zero byte value. The data word can be modified to have non-zero byte values stored at an end of a first side of the data word in the memory, and any zero byte values stored in a remainder of the data word. The modified data word can be written to the memory via at least a first slice of a plurality of slices that is configured to access the first side of the data word in the memory.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for controlling, by an on-chip memory controller, a plurality of hardware components that are configured to perform computations to access a shared memory. One of the on-chip memory controller includes at least one backside arbitration controller communicatively coupled with a memory bank group and a first hardware component, wherein the at least one backside arbitration controller is configured to perform bus arbitrations to determine whether the first hardware component can access the memory bank group using a first memory access protocol; and a frontside arbitration controller communicatively coupled with the memory bank group and a second hardware component, wherein the frontside arbitration controller is configured to perform bus arbitrations to determine whether the second hardware component can access the memory bank group using a second memory access protocol different from the first memory access protocol.

Abstract: An electronic device includes a first memory controller, a second memory controller, and a memory access controller. The first memory controller stores setting information of a predetermined memory, wherein the predetermined memory is defined as an execute-only-memory. The second memory controller provides and sets an enabling register according to the setting information of the predetermined memory, and generates an enabling signal. The memory access controller accesses the first memory controller and the second memory controller to move the data of the predetermined memory to a predetermined memory space corresponding to the enabling register according to the enabling signal and the setting information of the predetermined memory.

Abstract: A system includes a memory device and a processing device, operatively coupled to the memory device. The processing device is to perform operations, including initializing a block family associated with the memory device; aggregating temperature values received from one or more temperature sensors of the memory device over time to determine an aggregate temperature; responsive to beginning to program a block residing on the memory device, associating the block with the block family; and in response to the aggregate temperature being greater than or equal to a specified threshold temperature value: performing a soft closure of the block family; initializing an extension timer; continuing to program data to the block; and performing a hard closure of the block family in response to one of the extension timer reaching an extension time value or the block family satisfying a hard closure criteria.

Abstract: Methods, systems, and devices for usage level identification for memory device addresses are described. Systems, techniques, and devices are described herein in which a memory device may determine where to store data according to a level of usage of the data. The memory device may receive a write command indicating data to be written, a type of the data, and a logical address of a memory array for writing the data. The memory device may identify an entry associated with the logical address in a table that maps the logical address to a physical address of the memory array. The entry may include a field configured to maintain a level of usage for the logical address. The memory device may update the level of usage value according to a process and write the data to a physical address of the memory array based on the level of usage value.

Abstract: In certain aspects, dynamic remapping of memory addresses is provided and includes initiating a remapping of a logical block from a “mapped block” to a “remapped block.” Logical address locations for the logical block are mapped to physical address locations in the mapped block. The mapped and remapped blocks include non-volatile memory. A read command is received and determined to be for reading from a logical address location of the logical block, and the logical address location is determined to be mapped to a physical address location. Data is read from the physical address location of the mapped block. A write command is received and determined to be for writing data to the logical address location. Data is written to the physical address location of the remapped block. The read command is received after the initiation of the remapping and before the writing of the data to the remapped block.

Abstract: This application provides a data backup method. The method includes: obtaining, by a first node, an identifier of a backup execution node from a storage device; and backing up data of the first node in the storage device responsive to determining that an ID of the first node is the same as the ID of the backup execution node. The embodiments of this application can improve reliability of the data stored by the first node, and prevent a plurality of nodes from redundantly backing up duplicate data in the storage device. The techniques disclosed herein reduce resource consumption.

Abstract: A SD Card including, in one implementation, a memory array, a controller coupled to the memory array, and a bus for transferring data between the memory array and a host device in communication with the SD Card. The controller is configured to perform background maintenance operations on the memory array during execution of a read command received from the host device.

Abstract: A storage system has a memory that is organized in wordlines. Each wordline has a number of strings. A controller in the storage system changes, in each of the wordlines, which of the strings is a last string programmed. Doing so can unmask a program disturb error when triple-level cells in the memory are used as pseudo-multi-level cells. By unmasking the program disturb error, the controller can detect and correct the error.

Abstract: A caching system including a first sub-cache, and a second sub-cache, coupled in parallel with the first cache, for storing cache data evicted from the first sub-cache and write-memory commands that are not cached in the first sub-cache, and wherein the second sub-cache includes: color tag bits configured to store an indication that a corresponding cache line of the second sub-cache storing write miss data is associated with a color tag, and an eviction controller configured to evict cache lines of the second sub-cache storing write-miss data based on the color tag associated with the cache line.

Abstract: There is provided a data processing apparatus and method of data processing. The data processing apparatus comprises storage circuitry to store a hierarchy of page tables comprising an intermediate level page table. Each entry of the intermediate level page table comprises base address information of a next level page table and control information indicating whether an addressing function has been applied to reorder physical storage locations of entries of the next level page table. Address translation circuitry is provided to perform address translations in response to receipt of a virtual address by performing a lookup in a next level page table dependent on the base address information and a page table index from the virtual address. When the control information indicates that the addressing function has been applied, the lookup is performed at a modified storage location generated by applying the addressing function to the page table index.

Abstract: According to one embodiment, a nonvolatile memory includes a memory chip and a command processing unit. The command processing unit stores data read from a first position of the memory chip in a memory when a first command for compaction is received from a controller, transmits validity determination information used for determining whether or not the data read from the first position is valid to the controller, and writes valid data of the data stored in the memory to a second position of the memory chip when a second command for the compaction and validity identification information that identifies the valid data are received from the controller.

Abstract: A method includes accessing a first memory component of a memory sub-system via a first interface, accessing a second memory component of the memory sub-system via a second interface, and transferring data between the first memory component and the second memory component via the first interface. The method further includes initially writing data in the first memory component via a first address window and accessing data in the second memory component via a second address window in response to caching the data in first memory component to the second memory component, wherein caching the data in the first memory component to the second component includes changing an address for the data from the first address window to the second address window.

Abstract: Techniques are provided for implementing intelligent defragmentation in a storage system. A storage control system manages a logical address space of a storage volume. The logical address space is partitioned into a plurality of extents, wherein each extent comprises a contiguous block of logical addresses of the logical address space. The storage control system monitors input/output (I/O) operations for logical addresses associated with the extents, and estimates fragmentation levels of the extents based on metadata associated with the monitored I/O operations. The storage control system identifies one or more extents as candidates for defragmentation based at least on the estimated fragmentation levels of the extents.

Abstract: Provided are techniques for providing a global unique identifier for a storage volume. Under control of a storage initiator, a Global Universally Unique Identifier (GUUID) is identified for a storage volume of a storage device in a cloud system storing a plurality of storage devices, wherein the GUUID is generated for use with an ATA over Ethernet (AoE) protocol. The GUUID is stored in bytes of a packet header structure. Metadata is stored in remaining portions of the packet header structure. A request with the packet header structure is sent to a storage target.

Abstract: Aspects of a storage device including a master chip controller and a slave chip processor and memory including a plurality of memory locations are provided which allow for simplified processing of descriptors associated with host commands in the slave chip based on an adaptive context metadata message from the master chip. When the controller receives a host command, the controller in the master chip provides to the processor in the slave chip a descriptor associated with a host command, an instruction to store the descriptor in the one of the memory locations, and the adaptive context metadata message mapping a type of the descriptor to the one of the memory locations. The processor may then process the descriptor stored in the one of the memory locations based on the message, for example, by refraining from identifying certain information indicated in the descriptor. Reduced latency in command execution may thereby result.

Abstract: A data storage system in which different copies of a data object (e.g., a file) can be compressed using different compression processes (e.g. different compression algorithms/processes and/or compression parameters), with some favoring faster decompression, while others favoring storage space savings. When a data object needs to be accessed, the copy of the data object that can be decompressed using minimal resource (computing and/or time) can be located and retrieved.

Abstract: Each of a plurality of memory blocks of a nonvolatile memory device is divided into two or more wordline groups having different characteristics. A write command for at least two memory blocks among the plurality of memory blocks is received. During a first partial time interval included in an entire write time interval for two or more memory blocks, a data write operation is performed on a wordline group included in one memory block among the two or more memory blocks in response to a reception of an address for the one memory block. During a second other partial time interval included in the entire write time interval, a data write operation is performed on wordline groups included in the two or more memory blocks in response to a reception of an address for the two or more memory blocks.

Abstract: A ternary content addressable memory is provided comprising; a memory device that includes a plurality of memory address locations; hash logic operative to determine a hash value, based upon a ternary key, wherein the determined hash value corresponds to a memory address location of the memory device; an encoder operable to convert the ternary key to a binary bit representation; wherein converting includes determining binary mapping bits based upon number and positions of ternary non-X (don't care) value bits of the ternary key; wherein converting further includes determining a different binary data bit to correspond to each different ternary non-X value bit of the ternary key; and memory controller logic to cause the memory device to store the binary bit representation at the memory address location that corresponds to the determined hash value.

Abstract: In an embodiment, a system may support programmable hashing of address bits at a plurality of levels of granularity to map memory addresses to memory controllers and ultimately at least to memory devices. The hashing may be programmed to distribute pages of memory across the memory controllers, and consecutive blocks of the page may be mapped to physically distant memory controllers. In an embodiment, address bits may be dropped from each level of granularity, forming a compacted pipe address to save power within the memory controller. In an embodiment, a memory folding scheme may be employed to reduce the number of active memory devices and/or memory controllers in the system when the full complement of memory is not needed.