81 results about "Dram cache" patented technology

A system and method for efficient cache data access in a large row-based memory of a computing system. A computing system includes a processing unit and an integrated three-dimensional (3D) dynamic random access memory (DRAM). The processing unit uses the 3D DRAM as a cache. Each row of the multiple rows in the memory array banks of the 3D DRAM stores at least multiple cache tags and multiple corresponding cache lines indicated by the multiple cache tags. In response to receiving a memory request from the processing unit, the 3D DRAM performs a memory access according to the received memory request on a given cache line indicated by a cache tag within the received memory request. Rather than utilizing multiple DRAM transactions, a single, complex DRAM transaction may be used to reduce latency and power consumption.

A method and apparatus for refreshing data in a dynamic random access memory (DRAM) cache memory in a computer system are provided to perform a data refresh operation without refresh penalty (e.g., delay in a processor). A data refresh operation is performed with respect to a DRAM cache memory by detecting a request address from a processor, stopping a normal refresh operation when the request address is detected, comparing the request address with TAG addresses stored in a TAG memory, generating refresh addresses to refresh data stored in the cache memory, each of which is generated based on an age of data corresponding to the refresh address, and performing a read/write operation on a wordline accessed by the request addresses and refreshing data on wordlines accessed by the refresh addresses, wherein the read/write operation and the refreshing of data are performed simultaneously.

An instruction cache and data cache used to virtualize the storage of global data and instructions used by graphics shaders. Present day hardware design stores the global data and instructions used by the shaders in a fixed amount of registers or writable control store (WCS). However, this traditional approach limits the size and the complexity of the shaders that can be supported. By virtualizing the storage of the global data and instructions, the amount of global or state memory available to the shader and the length of the shading programs are no longer constrained by the physical on-chip memory.

The present invention provides a DRAM/NVM hierarchical heterogeneous memory system with software-hardware cooperative management schemes. In the system, NVM is used as large-capacity main memory, and DRAM is used as a cache to the NVM. Some reserved bits in the data structure of TLB and last-level page table are employed effectively to eliminate hardware costs in the conventional hardware-managed hierarchical memory architecture. The cache management in such a heterogeneous memory system is pushed to the software level. Moreover, the invention is able to reduce memory access latency in case of last-level cache misses. Considering that many applications have relatively poor data locality in big data application environments, the conventional demand-based data fetching policy for DRAM cache can aggravates cache pollution. In the present invention, an utility-based data fetching mechanism is adopted in the DRAM/NVM hierarchical memory system, and it determines whether data in the NVM should be cached in the DRAM according to current DRAM memory utilization and application memory access patterns. It improves the efficiency of the DRAM cache and bandwidth usage between the NVM main memory and the DRAM cache.

Embodiments of both a non-volatile main memory (NVMM) single node and a multi-node computing system are disclosed. One embodiment of the NVMM single node system has a cache subsystem composed of all DRAM, a large main memory subsystem of all NAND flash, and provides different address-mapping policies for each software application. The NVMM memory controller provides high, sustained bandwidths for client processor requests, by managing the DRAM cache as a large, highly banked system with multiple ranks and multiple DRAM channels, and large cache blocks to accommodate large NAND flash pages. Multi-node systems organize the NVMM single nodes in a large inter-connected cache/flash main memory low-latency network. The entire interconnected flash system exports a single address space to the client processors and, like a unified cache, the flash system is shared in a way that can be divided unevenly among its client processors: client processors that need more memory resources receive it at the expense of processors that need less storage. Multi-node systems have numerous configurations, from board-area networks, to multi-board networks, and all nodes are connected in various Moore graph topologies. Overall, the disclosed memory architecture dissipates less power per GB than traditional DRAM architectures, uses an extremely large solid-state capacity of a terabyte or more of main memory per CPU socket, with a cost-per-bit approaching that of NAND flash memory, and performance approaching that of an all DRAM system.

Techniques are disclosed relating to prefetching data from memory. In one embodiment, an integrated circuit may include a processor containing an execution core and a data cache. The execution core may be configured to receive an instance of a prefetch instruction that specifies a memory address from which to retrieve data. In response to the instance of the instruction, the execution core retrieves data from the memory address and stores it in the data in the data cache, regardless of whether the data corresponding to that particular memory address is already stored in the data cache. In this manner, the data cache may be used as a prefetch buffer for data in memory buffers where coherence has not been maintained.

The invention relates to a memory system cache mechanism based on flash memory. It is provided that two cache areas are constructed in the DRAM of the memory system based on the flash memory, namely a page cache area and a cache line cache area. On one hand, hot cache line in each data page is identified and stored through a history perception hotspot identification mechanism, and the hit rate of DRAM cache is increased. On the other hand, a delay refresh mechanism is adopted, and a clean data block is preferably removed when the cache area is full so as to reduce writing for the flash memory. Meanwhile, a weak variable coefficient algorithm is adopted, modes of history visit records, dirty mark bits and the like are added, and the expenses of the cache mechanism on time and space are not high. The memory system cache mechanism based on the flash memory utilizes the nature of the flash memory, writing delay of the memory system can be effectively improved, and service life of the memory system can be effectively prolonged.

The invention discloses an error corrosion encoding method for a dynamic random access memory (DRAM) buffer, which comprises the following steps of: (1) prolonging a refresh cycle of the DRAM buffer and reducing refresh frequency; and (2) encoding and decoding data in the buffer. In the method, refresh time is set according to the most error-prone bits, some error data is compensated by error correction technology which is powerful enough to correct a plurality of bits, which means the refresh time can be prolonged, thereby reducing power consumption.

The invention discloses a data packet low-delay caching device and method for switch equipment. The device comprises a data packet aggregation module, a scheduler module, a data receiving module, a caching region module, a dynamic memory distribution module, a data sending module, a memory block state table module, and an idle memory block first-in first-out queue module. The method comprises thefollowing steps: the aggregation module aggregates a data packet as an aggregated data block with fixed size; the scheduler module distributes a caching address for the aggregated data block; the dynamic memory management module maintains memory block use information; the data receiving module moves the aggregated data block to the corresponding DRAM caching unit; an output port sends a read application to the scheduler module, and the data sending module outputs the data packet to the external from the caching region after the conflict detection. The delay of the cache is lowered, and the processing speed of the exchange equipment is improved.

A host adapter, which interfaces two I/O buses, caches data transferred from one I/O bus to another I/O bus in a data first-in-first-out (FIFO)/caching memory. In addition, when a target device on the another I/O bus is ready to receive the data, data is transferred from the data FIFO/caching memory even though not all of the data may be cached in that memory. Hence, data is concurrently transferred to and transferred from the data FIFO/caching memory. The data transfer to the target device is throttled if cached data is unavailable in the data FIFO/caching memory for transfer, e.g., the data cache is empty for the current context.

A high speed DRAM cache architecture. One disclosed embodiment includes a multiplexed bus interface to interface with a multiplexed bus. A cache control circuit drives a row address portion of an address on the multiplexed bus interface and a command to open a memory page containing data for a plurality of ways. The cache control circuit subsequently drives a column address including at least a way indicator to the multiplexed bus interface.