A computing system using a bit counter may include a host device; a cache configured to temporarily store data of the host device, and including a plurality of sets; a cache controller configured to receive a multi-bit cache address from the host device, perform computation on the cache address using a plurality of bit counters, and determine a hash function of the cache; a semiconductor device; and a memory controller configured to receive the cache address from the cache controller, and map the cache address to a semiconductor device address.

Systems, apparatuses, and methods for implementing flexible dictionary sharing techniques for caches are disclosed. A set-associative cache includes a dictionary for each data array set. When a cache line is to be allocated in the cache, a cache controller determines to which set a base index of the cache line address maps. Then, a selector unit determines which dictionary of a group of dictionaries stored by those sets neighboring this set would achieve the most compression for the cache line. This dictionary is then selected to compress the cache line. An offset is added to the base index of the cache line to generate a full index in order to map the cache line to the set corresponding to this chosen dictionary. The compressed cache line is stored in this set with the chosen dictionary, and the offset is stored in the corresponding tag array entry.

In one embodiment, an apparatus includes a memory access circuit to receive memory access instructions and provide at least some of the memory access instructions to a memory subsystem for execution. The memory access circuit may have a conversion circuit to convert the first memory access instruction to a first subline memory access instruction, e.g., based at least in part on an access history for a first memory access instruction. Other embodiments are described and claimed.

Systems, methods, and apparatus including computer-readable mediums for managing memories by overlapping ranges of pages in nonvolatile memory systems are provided. An example memory system includes a memory controller coupled to a memory and configured to: determine a range of logical addresses associated with a command, search particular mapping tables including the range of logical addresses in mapping pages in the memory, determine whether a starting address of the range of logical addresses is in an overlapped range of first and second sequential mapping pages, the overlapped range including logical addresses of one or more mapping tables duplicated in the first and second mapping pages, determine which of the first and second mapping pages from which the particular mapping tables to be loaded based on a result of determining whether the starting address is in the overlapped range, and load the particular mapping tables from the determined mapping page.

Dynamic compression with dynamic multi-stage encryption for a data storage system in accordance with the present description includes, in one aspect of the present description, preserves end-to-end encryption between a host and a storage controller while compressing data which was received from the host in encrypted but uncompressed form, using MIPs and other processing resources of the storage controller instead of the host. In one embodiment, the storage controller decrypts encrypted but uncompressed data received from the host to unencrypted data and compresses the unencrypted data to compressed data. The storage controller then encrypts the compressed data to encrypted, compressed data and stores the encrypted, compressed data in a storage device controlled by the storage controller. Other aspects and advantages may be realized, depending upon the particular application.

A prefetch unit includes multiple memories; and a memory controller coupled to the multiple memories. The memory controller includes a prefetch stream filter and a prefetch buffer. The prefetch stream filter includes a first set of address slots and a set of direction prediction fields, each of which is associated with a respective one of the address slots of the first set of address slots. The prefetch buffer includes a set of buffer slots, each slot of the set of buffer slots including an address field, a direction prediction field, a data pending field, a data valid field, and a set of sub-slots configured to store data, wherein each address field of each slot of the set of buffer slots is configured to store at least a portion of an address associated with the corresponding slot.

Disclosed herein includes a system, a method, and a device for reading and writing sparse data in a neural network accelerator. A plurality of slices can be established to access a memory having an access size of a data word. A first slice can be configured to access a first side of the data word in memory. Circuitry can access a mask identifying byte positions within the data word having non-zero values. The circuitry can modify the data word to have non-zero byte values stored starting at an end of the first side, and any zero byte values stored in a remainder of the data word. A determination can be made whether a number of non-zero byte values is less than or equal to a first access size of the first slice. The circuitry can write the modified data word to the memory via at least the first slice.

According to an embodiment of the present invention, a probability to achieve a purpose is increased in a case where information is intermittently allocated to a plurality of elements. An information processing apparatus according to an embodiment of the present invention is an information processing apparatus that allocates information to respective elements included in a first set and includes an allocator. The allocator uses a first subset and a second subset, and allocates the information to at least one element included in the first subset. The first subset is constituted by elements which are included in the first set and in which information allocation may be performed at a present time. The second subset is constituted by elements which are included in the first set and in which the information allocation is not performed at the present time.

A method of operating a cache in a computing device includes, in response to receiving a memory access request at the cache, determining compressibility of data specified by the request, selecting in the cache a destination portion for storing the data based on the compressibility of the data and a persistent fault history of the destination portion, and storing a compressed copy of the data in a non-faulted subportion of the destination portion, wherein the persistent fault history indicates that the non-faulted subportion excludes any persistent faults.

Various embodiments are generally directed to techniques to store data for critical chunk operations, such as by utilizing a spare lane, for instance. Some embodiments are particularly directed to a memory controller that stores a portion of a critical chunk in a spare lane to enable the entire critical chunk to be stored in a half of the cache line.