摘要:
A method for sequentially coupling successive processor requests for a cache line before the data is received in the cache of a first coupled processor. Both homogenous and non-homogenous operations are chained to each other, and the coherency protocol includes several new intermediate coherency responses associated with the chained states. Chained coherency states are assigned to track the chain of processor requests and the grant of access permission prior to receipt of the data at the first processor. The chained coherency states also identify the address of the receiving processor. When data is received at the cache of the first processor within the chain, the processor completes its operation on (or with) the data and then forwards the data to the next processor in the chain. The chained coherency protocol frees up address bus bandwidth by reducing the number of retries.
摘要:
A system and method for cache management in a data processing system. The data processing system includes a processor and a memory hierarchy. The memory hierarchy includes at least an upper memory cache, at least a lower memory cache, and a write-back data structure. In response to replacing data from the upper memory cache, the upper memory cache examines the write-back data structure to determine whether or not the data is present in the lower memory cache. If the data is present in the lower memory cache, the data is replaced in the upper memory cache without casting out the data to the lower memory cache.
摘要:
A system and method of managing cache hierarchies with adaptive mechanisms. A preferred embodiment of the present invention includes, in response to selecting a data block for eviction from a memory cache (the source cache) out of a collection of memory caches, examining a data structure to determine whether an entry exists that indicates that the data block has been evicted from the source memory cache, or another peer cache, to a slower cache or memory and subsequently retrieved from the slower cache or memory into the source memory cache or other peer cache. Also, a preferred embodiment of the present invention includes, in response to determining the entry exists in the data structure, selecting a peer memory cache out of the collection of memory caches at the same level in the hierarchy to receive the data block from the source memory cache upon eviction.
摘要:
A system and method for cache management in a data processing system. The data processing system includes a processor and a memory hierarchy. The memory hierarchy includes at least an upper memory cache, at least a lower memory cache, and a write-back data structure. In response to replacing data from the upper memory cache, the upper memory cache examines the write-back data structure to determine whether or not the data is present in the lower memory cache. If the data is present in the lower memory cache, the data is replaced in the upper memory cache without casting out the data to the lower memory cache.
摘要:
A method and data processing system for sequentially coupling successive, homogenous processor requests for a cache line in a chain before the data is received in the cache of a first processor within the chain. Chained intermediate coherency states are assigned to track the chain of processor requests and subsequent access permission provided, prior to receipt of the data at the first processor starting the chain. The chained intermediate coherency state assigned identifies the processor operation and a directional identifier identifies the processor to which the cache line is to be forwarded. When the data is received at the cache of the first processor within the chain, the first processor completes its operation on (or with) the data and then forwards the data to the next processor in the chain. The chain is immediately stopped when a non-homogenous operation is snooped by the last-in-chain processor.
摘要:
A method and an apparatus for performing just-in-time data prefetching within a data processing system comprising a processor, a cache or prefetch buffer, and at least one memory storage device. The apparatus comprises a prefetch engine having means for issuing a data prefetch request for prefetching a data cache line from the memory storage device for utilization by the processor. The apparatus further comprises logic/utility for dynamically adjusting a prefetch distance between issuance by the prefetch engine of the data prefetch request and issuance by the processor of a demand (load request) targeting the data/cache line being returned by the data prefetch request, so that a next data prefetch request for a subsequent cache line completes the return of the data/cache line at effectively the same time that a demand for that subsequent data/cache line is issued by the processor.
摘要:
A system and method of managing cache hierarchies with adaptive mechanisms. A preferred embodiment of the present invention includes, in response to selecting a data block for eviction from a memory cache (the source cache) out of a collection of memory caches, examining a data structure to determine whether an entry exists that indicates that the data block has been evicted from the source memory cache, or another peer cache, to a slower cache or memory and subsequently retrieved from the slower cache or memory into the source memory cache or other peer cache. Also, a preferred embodiment of the present invention includes, in response to determining the entry exists in the data structure, selecting a peer memory cache out of the collection of memory caches at the same level in the hierarchy to receive the data block from the source memory cache upon eviction.
摘要:
A method of assigning virtual memory to physical memory in a data processing system allocates a set of contiguous physical memory pages for a new page mapping, instructs the memory controller to move the virtual memory pages according to the new page mapping, and then allows access to the virtual memory pages using the new page mapping while the memory controller is still copying the virtual memory pages to the set of physical memory pages. The memory controller can use a mapping table which temporarily stores entries of the old and new page addresses, and releases the entries as copying for each entry is completed. The translation lookaside buffer (TLB) entries in the processor cores are updated for the new page addresses prior to completion of copying of the memory pages by the memory controller. The invention can be extended to non-uniform memory array (NUMA) systems. For systems with cache memory, any cache entry which is affected by the page move can be updated by modifying its address tag according to the new page mapping. This tag modification may be limited to cache entries in a dirty coherency state. The cache can further relocate a cache entry based on a changed congruence class for any modified address tag.
摘要:
A multiprocessor system includes a plurality of data processing nodes. Each node has a processor coupled to a system memory, a cache memory, and a cache directory. The cache directory contains cache coherency information for a predetermined range of system memory addresses. An interconnection enables the nodes to exchange messages. A node initiating a function shipping request identifies an intermediate destination directory based on a list of the function's operands and sends a message indicating the function and its corresponding operands to the identified destination directory. The destination cache directory determines a target node based, at least in part, on its cache coherency status information to reduce memory access latency by selecting a target node where all or some of the operands are valid in the local cache memory. The destination directory then ships the function to the target node over the interconnection.
摘要:
A method and memory controller for adaptive row management within a memory subsystem provides metrics for evaluating row access behavior and dynamically adjusting the row management policy of the memory subsystem in conformity with measured metrics to reduce the average latency of the memory subsystem. Counters provided within the memory controller track the number of consecutive row accesses and optionally the number of total accesses over a measurement interval. The number of counted consecutive row accesses can be used to control the closing of rows for subsequent accesses, reducing memory latency. The count may be validated using a second counter or storage for improved accuracy and alternatively the row close count may be set via program or logic control in conformity with a count of consecutive row hits in ratio with a total access count. The control of row closure may be performed by a mode selection between always closing a row (non-page mode) or always holding a row open (page mode) or by intelligently closing rows after a count interval (row hold count) determined from the consecutive row access measurements. The logic and counters may be incorporated within the memory controller or within the memory devices and the controller/memory devices may provide I/O ports or memory locations for reading the count values and/or setting a row management mode or row hold count.