A data processing system includes one or more nodes, each node including a memory sub-system. The sub-system includes a fine-grained, memory, and a less-fine-grained (e.g., page-based) memory. The fine-grained memory optionally serves as a cache and/or as a write buffer for the page-based memory. Software executing on the system uses a node address space which enables access to the page-based memories of all nodes. Each node optionally provides ACID memory properties for at least a portion of the space. In at least a portion of the space, memory elements are mapped to locations in the page-based memory. In various embodiments, some of the elements are compressed, the compressed elements are packed into pages, the pages are written into available locations in the page-based memory, and a map maintains an association between the some of the elements and the locations.

System including a fine-grained memory and a less-fine-grained memory

A system, method, and computer program product are provided for a memory system. The system includes a first semiconductor platform including at least one first circuit, and at least one additional semiconductor platform stacked with the first semiconductor platform and including at least one additional circuit.

System, method, and computer program product for improving memory systems

Several system-level operations trigger bulk data copy or initialization. Even though these bulk data operations do not require any computation, current systems transfer a large quantity of data back and forth on the memory channel to perform such operations. As a result, bulk data operations consume high latency, bandwidth, and energy — degrading both system performance and energy efficiency. In this work, we propose RowClone, a new and simple mechanism to perform bulk copy and initialization completely within DRAM — eliminating the need to transfer any data over the memory channel to perform such operations. Our key observation is that DRAM can internally and efficiently transfer a large quantity of data (multiple KBs) between a row of DRAM cells and the associated row buffer. Based on this, our primary mechanism can quickly copy an entire row of data from a source row to a destination row by first copying the data from the source row to the row buffer and then from the row buffer to the destination row, via two back-to-back activate commands. This mechanism, which we call the Fast Parallel Mode of RowClone, reduces the latency and energy consumption of a 4KB bulk copy operation by 11.6× and 74.4×, respectively, and a 4KB bulk zeroing operation by 6.0× and 41.5×, respectively. To efficiently copy data between rows that do not share a row buffer, we propose a second mode of RowClone, the Pipelined Serial Mode, which uses the shared internal bus of a DRAM chip to quickly copy data between two banks. RowClone requires only a 0.01% increase in DRAM chip area. We quantitatively evaluate the benefits of RowClone by focusing on fork, one of the frequently invoked system calls, and five other copy and initialization intensive applications. Our results show that RowClone can significantly improve both single-core and multi-core system performance, while also significantly reducing main memory bandwidth and energy consumption.

/pdf/rowclone-fast-and-energy-efficient-in-dram-bulk-data-copy-2wpwggalyr.pdf

RowClone: fast and energy-efficient in-DRAM bulk data copy and initialization

A computer implemented method of analyzing and graphically representing the correlation of a plurality of transaction items, the method comprising the steps of: retrieving data associated with groups of the transaction items, correlating a plurality of groups of transaction items in a dimensionally reduced manner, creating a tree hierarchy which classifies the groups of transaction items in a hierarchy according to a defined user understandable factor, wherein the tree hierarchy is linked to the groups of transaction items, and graphically representing the correlated groups of transaction items and tree hierarchy to enable interaction between the correlated groups of transaction items and the linked tree hierarchy.

Methods, apparatus and systems for data visualization and related applications

Modern DRAMs have multiple banks to serve multiple memory requests in parallel. However, when two requests go to the same bank, they have to be served serially, exacerbating the high latency of off-chip memory. Adding more banks to the system to mitigate this problem incurs high system cost. Our goal in this work is to achieve the benefits of increasing the number of banks with a low cost approach. To this end, we propose three new mechanisms that overlap the latencies of different requests that go to the same bank. The key observation exploited by our mechanisms is that a modern DRAM bank is implemented as a collection of subarrays that operate largely independently while sharing few global peripheral structures. Our proposed mechanisms (SALP-1, SALP-2, and MASA) mitigate the negative impact of bank serialization by overlapping different components of the bank access latencies of multiple requests that go to different subarrays within the same bank. SALP-1 requires no changes to the existing DRAM structure and only needs reinterpretation of some DRAM timing parameters. SALP-2 and MASA require only modest changes (&#60; 0.15% area overhead) to the DRAM peripheral structures, which are much less design constrained than the DRAM core. Evaluations show that all our schemes significantly improve performance for both single-core systems and multi-core systems. Our schemes also interact positively with application-aware memory request scheduling in multi-core systems.

/pdf/a-case-for-exploiting-subarray-level-parallelism-salp-in-371d40m7v0.pdf

A case for exploiting subarray-level parallelism (SALP) in DRAM

A memory controller includes at least one interface adapted to be coupled to one or more first memory devices of a first memory type having a first set of attributes, and to one or more second memory devices of a second memory type having a second set of attributes. The first and second sets of attributes have at least one differing attribute. The controller also includes interface logic configured to direct memory transactions having a predefined first characteristic to the first memory devices and to direct memory transactions having a predefined second characteristic to the second memory devices. Pages having a usage characteristic of large volumes of write operations may be mapped to the one or more first memory devices, while pages having a read-only or read-mostly usage characteristic may be mapped to the one or more second memory devices.

Memory controller for non-homogeneous memory system

Described are dynamic memory systems that perform overlapping refresh and data-access (read or write) transactions that minimize the impact of the refresh transaction on memory performance. The memory systems support independent and simultaneous activate and precharge operations directed to different banks. Two sets of address registers enable the system to simultaneously specify different banks for refresh and data-access transactions.

Dynamic memory supporting simultaneous refresh and data-access transactions

A signaling system is disclosed. The signaling system includes a first integrated circuit (IC) chip to receive a data signal and a strobe signal. The first IC includes circuitry to sample the data signal at times indicated by the strobe signal to generate phase error information and circuitry to output the phase error information from the first IC device. The system further includes a signaling link and a second IC chip coupled to the first IC chip via the signaling link to output the data signal and the strobe signal to the first IC chip. The second IC chip includes delay circuitry to generate the strobe signal by delaying an aperiodic timing signal for a first time interval and timing control circuitry to receive the phase error information from the first IC chip and adjust the first time interval in accordance with the phase error information.

Signaling system with adaptive timing calibration

Described is a memory system in which the memory core organization changes with device width. The number of physical memory banks accessed reduces with device width, resulting in reduced power usage for relatively narrow memory configurations. Increasing the number of logic memory banks for narrow memory widths reduces the likelihood of bank conflicts, and consequently improves speed performance.

Memory device supporting a dynamically configurable core organisation

A memory module having reduced access granularity. The memory module includes a substrate having signal lines thereon that form a control path and first and second data paths, and further includes first and second memory devices coupled in common to the control path and coupled respectively to the first and second data paths. The first and second memory devices include control circuitry to receive respective first and second memory access commands via the control path and to effect concurrent data transfer on the first and second data paths in response to the first and second memory access commands.

Frederick A. Ware

Papers

Memory controller for non-homogeneous memory system

Dynamic memory supporting simultaneous refresh and data-access transactions

Signaling system with adaptive timing calibration

Memory device supporting a dynamically configurable core organisation

Memory Module With Reduced Access Granularity