Provided are a computer program product, system, and method for staging data from storage to a fast cache tier of a multi-tier cache in a non-adaptive sector caching mode in which data staged in response to a read request is limited to track sectors required to satisfy the read request. Data is also staged from storage to a slow cache tier of the multi-tier cache in a selected adaptive caching mode of a plurality of adaptive caching modes available for staging data of tracks. Adaptive caching modes are selected for the slow cache tier as a function of historical access ratios. Prestage requests for the slow cache tier are enqueued in one of a plurality of prestage request queues of various priority levels as a function of the selected adaptive caching mode and historical access ratios. Other aspects and advantages are provided, depending upon the particular application.

A cache management system includes a sequentiality determination process configured to determine sequentiality profiles of a workload of IO traces as the workload dynamically changes over time. A learning process is trained to learn a correlation between workload sequentiality and cache pollution, and the trained learning process is used to predict cache pollution before the cache starts to experience symptoms of excessive pollution. The predicted pollution value is used by a cache policy adjustment process to change the prefetch policy applied to the cache, to proactively control the manner in which prefetching is used to write data to the cache. Selection of the cache policy is implemented on a per-LUN basis, so that cache performance for each LUN is individually managed by the cache management system.

Limitations on memory access decrease the computing capability of related-art semiconductor devices during convolution processing in a convolutional neural network. A semiconductor device according to an aspect of the present invention includes an accelerator section that performs computation on a plurality of intermediate layers included in a convolutional neural network by using a memory having a plurality of banks capable of changing the read/write status on an individual bank basis. The accelerator section includes a network layer control section that controls a memory control section in such a manner as to change the read/write status assigned to the banks storing input data or output data of the intermediate layers in accordance with the transfer amounts and transfer rates of the input data and output data of the intermediate layers included in the convolutional neural network.

Techniques provide cache service in a storage system. Such techniques involve a storage cell pool, a cache and an underlying storage system. The storage cell pool includes multiple storage cells, a storage cell among the multiple storage cells being mapped to a physical address in the underlying storage system via an address mapping of the storage system. Specifically, an access request for target data at a virtual address in the storage cell pool is received, and the type of the access request is determined. The access request is served with the cache on the basis of the determined type, where the cache is used to cache data according to a format of a storage cell in the storage cell pool. The cache directly stores data in various storage cells in the pool that is visible to users, so that response speed for the access request may be increased.

A method for managing a cache memory, including executing first and second processes, when the second process modifies the state of the cache memory, updating the value of an indicator associated with this second process, and comparing the value of this indicator to a predefined threshold and, when this predefined threshold is exceeded, detecting an abnormal use of the cache memory by the second process, in response to this detection, modifying pre-recorded relationships in order to associate with the identifier of the second process a value of a parameter q different from the value of the parameter q associated with the first process so that, after this modification, when the received address of a word to be read is the same for the first and second processes, then the set addresses used to read this word from the cache memory are different.

Techniques described herein generally include methods and systems related to cache partitioning in a chip multiprocessor. Cache-partitioning for a single thread or application between multiple data sources improves energy or latency efficiency of a chip multiprocessor by exploiting variations in energy cost and latency cost of the multiple data sources. Partition sizes for each data source may be selected using an optimization algorithm that minimizes or otherwise reduces latencies or energy consumption associated with cache misses.

A memory system includes: a memory device; a host interface suitable for receiving write commands and queueing the received write commands in an interface queue; a workload manager suitable for detecting, in a cache program mode, a mixed workload when a read count is greater than a first threshold value, the read count representing a number of read commands queued in the interface queue and the mixed workload representing receipt of a mix of read and write commands; a mode manager suitable for switching from the cache program mode to a normal program mode when the mixed workload is detected; and a processor suitable for processing write commands queued in a command queue in the cache program mode and processing write commands queued in the interface queue in the normal program mode when the mixed workload is detected.

Embodiments of the present invention disclose a method, computer program product, and system for utilizing a block storage device as Dynamic Random-Access Memory (DRAM) space, wherein a computer includes at least one DRAM module and at least one block storage device interfaced to the computer using a double data rate (DDR) interface. During boot up, the computer configures DRAM and block storage devices of the computer for utilization as DRAM or block storage. Then the computer determines that more DRAM space is required. Responsive to determining that more DRAM space is required, the computer transforms a block storage device into DRAM space. Once the computer determines that the transformed block storage device that is being used for DRAM space is no longer needed to be used as DRAM space, the computer transforms the block storage device back to block storage space.

Embodiments of the invention provide systems and methods for managing processing, memory, storage, network, and cloud computing to significantly improve the efficiency and performance of processing nodes. Embodiments can implement an object memory fabric including object memory modules storing memory objects created natively within the object memory module and may be a managed at a memory layer. The memory module object directory may index all memory objects within the object memory module. A hierarchy of object routers communicatively coupling the object memory modules may each include a router object directory that indexes all memory objects and portions contained in object memory modules below the object router in the hierarchy. The hierarchy of object routers may behave in aggregate as a single object directory communicatively coupled to all object memory modules and to process requests based on the router object directories.

Embodiments of the invention provide systems and methods for managing processing, memory, storage, network, and cloud computing to significantly improve the efficiency and performance of processing nodes. Embodiments can implement an object memory fabric including object memory modules storing memory objects created natively within the object memory module and may be a managed at a memory layer. The memory module object directory may index all memory objects within the object memory module. A hierarchy of object routers communicatively coupling the object memory modules may each include a router object directory that indexes all memory objects and portions contained in object memory modules below the object router in the hierarchy. The hierarchy of object routers may behave in aggregate as a single object directory communicatively coupled to all object memory modules and to process requests based on the router object directories.