The assignee for this patent, patent number 8627041, is
Reporters obtained the following quote from the background information supplied by the inventors: "Embodiments of the invention relate generally to memory systems and more specifically to an efficient line and page organization for compression status bit caching.
"Performance requirements are continually increasing in data processing systems, conventionally comprising one or more processor chips and attached memory devices organized as independently operating partitions. System performance is generally determined by on-chip data processing performance as well as effective bandwidth to the attached memory devices. One technique for increasing effective memory bandwidth, and therefore overall performance, is to store certain blocks of data within the attached memory in a compressed format. A plurality of both loss-less and lossy compression formats, as well as blocks not subject to any compression may coexist within attached memory. A compression status is associated with each block to specify whether the block of original data is stored uncompressed or using one of the plurality of compression formats. Each compression format advantageously reduces the number of bits needed to represent a block of original data stored in attached memory. A compression status bit set is associated with each block to encode the compression status for the block. Compression status bit sets for all compressible blocks may be stored as a compression status structure within attached memory. Sequential compression status bit sets within the compression status structure correspond to sequential blocks in the attached memory devices. Dedicated circuits coupled to a memory interface module typically perform compression and decompression operations based on compression status for a block being accessed. The memory interface module maps each physical address to one of one or more memory partitions to access a specified block of memory within the partition.
"Another technique for increasing effective memory bandwidth is caching, whereby bandwidth demand is shifted from the attached memory devices to on-chip cache storage that provides low latency and high bandwidth access to data. Cache storage is typically organized as cache lines, with each complete cache line being filled or flushed in response to a respective read or write. A cache line is conventionally sized as an integral multiple of an access quantum to attached memory devices. A compression status bit cache is configured to store a plurality of compression status bit sets per cache line, thereby facilitating access to compression status bit sets for sequential physically addressed blocks in attached memory devices. Compression status stored by a given compression status bit set is used to determine compression format and therefore access size for a block of data prior to an access being initiated to the block of data.
"Memory management of modern data processing systems typically implements a virtual memory access model for memory clients. Regions of contiguous virtual memory may be allocated and used by the memory clients, with access locality associated with virtual addresses. However, a contiguous range of virtual addresses may map arbitrarily to physical addresses. As such, an arbitrarily large number of compression status bit cache lines may be needed to store compression status bits associated with a given contiguous virtual address range. This can lead to cache fragmentation in the compression status bit cache and relatively inefficient use of associated cache storage, reducing overall efficiency for the processing system.
"Accordingly, what is needed in the art is a technique for improving access efficiency for compression status bits in a virtual memory system."
In addition to obtaining background information on this patent, VerticalNews editors also obtained the inventors' summary information for this patent: "One embodiment of the present invention sets forth a method for determining and updating compression status for a virtually addressed unit of data residing within a frame buffer. The method includes selecting a page table entry (PTE) based on a virtual address, the PTE comprising a physical address for the unit of data within the frame buffer and a compression cache tag line, computing a compression cache tag based on the compression cache tag line and a partition count for the frame buffer, and querying a compression cache to determine whether the compression cache tag represents a cache hit. If the compression cache tag represents a cache hit, then one or more compression bits are accessed from a cache line, wherein the one or more compression bits represent a compression status for the unit of data to be used when accessing the unit of data from the frame buffer. If the compression cache tag represents a cache miss, then the cache line is retrieved from memory prior to accessing the designated compression bits within the cache line.
"One advantage of embodiments of the present invention is that a processing unit may efficiently access virtually mapped data that is compressed and distributed over an arbitrary number of partitions. Thus, embodiments of the present invention overcome prior art design deficiencies that limit the application of virtual memory mapping in systems having an arbitrary number of memory partitions."
For more information, see this patent: Glasco, David B.; Holmqvist, Peter B.; Lynch, George R.; Marchand, Patrick R.; Mehra, Karan;
Keywords for this news article include:
Our reports deliver fact-based news of research and discoveries from around the world. Copyright 2014, NewsRx LLC
Most Popular Stories
- Obama Administration Releases Proposal to Regulate For-Profit Colleges
- Apple, HP, Intel May Take a Hit from Slowdown in Smartphone Sales Growth
- Elizabeth Vargas' Husband Marc Cohn Addresses Rumors
- Keurig Adds Peet's coffee, Alters Starbucks deal
- U.S. to Relinquish Gov't Control Over Internet
- Motley Crue's Nikki Sixx Marries Model Courtney Bingham
- FDIC Files Lawsuit on Behalf of Banks Allegedly Hurt by Libor Scandal
- Chinese e-Commerce Giant Alibaba Gears for IPO in U.S.
- Quiznos Files for Chapter 11
- Some California Cities Seeking Water Independence