Abstract: A circular queue implementing a scheme for prioritized reads is disclosed. In one embodiment, a circular queue (or buffer) includes a number of storage locations each configured to store a data value. A multiplexer tree is coupled between the storage locations and a read port. A priority circuit is configured to generate and provide selection signals to each multiplexer of the multiplexer tree, based on a priority scheme. Based on the states of the selection signals, one of the storage locations is coupled to the read port via the multiplexers of the multiplexer tree.

Abstract: A method and apparatus for selectively powering down a portion of a cache memory includes determining a power down condition dependent upon a number of accesses to the cache memory. In response to the detection of the power down condition, selecting a group of cache ways included in the cache memory dependent upon a number of cache lines in each cache way that are also included in another cache memory. The method further includes locking and flushing the selected group of cache ways, and then activating a low power mode for the selected group of cache ways.

Abstract: A mechanism for evicting a cache line from a cache memory includes first selecting for eviction a least recently used cache line of a group of invalid cache lines. If all cache lines are valid, selecting for eviction a least recently used cache line of a group of cache lines in which no cache line of the group of cache lines is also stored within a higher level cache memory such as the L1 cache, for example. Lastly, if all cache lines are valid and there are no non-inclusive cache lines, selecting for eviction the least recently used cache line stored in the cache memory.

Abstract: Techniques are disclosed relating to completion of load and store instructions in a weakly-ordered memory model. In one embodiment, a processor includes a load queue and a store queue and is configured to associate queue information with a load instruction in an instruction stream. In this embodiment, the queue information indicates a location of the load instruction in the load queue and one or more locations in the store queue that are associated with one or more store instructions that are older than the load instruction. The processor may determine, using the queue information, that the load instruction does not conflict with a store instruction in the store queue that is older than the load instruction. The processor may remove the load instruction from the load queue while the store instruction remains in the store queue. The queue information may include a wrap value for the load queue.

Abstract: A method and apparatus for evicting cache lines from a cache memory includes receiving a request from one of a plurality of processors. The cache memory is configured to store a plurality of cache lines, and a given cache line includes an identifier indicating a processor that performed a most recent access of the given cache line. The method further includes selecting a cache line for eviction from a group of least recently used cache lines, where each cache line of the group of least recently used cache lines occupy a priority position less that a predetermined value, and then evicting the selected cache line.

Abstract: Systems, methods, and apparatuses for implementing selective victimization to reduce power and utilized bandwidth in a multi-level cache hierarchy. Each set of an upper-level cache includes a counter that keeps track of the number of times the set was accessed. These counters are periodically decremented by another counter that tracks the total number of accesses to the cache. If a given set counter is below a certain threshold value, clean victims are dropped from this given set instead of being sent to a lower-level cache. Also, a separate counter is used to track the total number of outstanding requests for the cache as a proxy for bus-bandwidth in order to gauge the total amount of traffic in the system. The cache will implement selective victimization whenever there is a large amount of traffic in the system.

Abstract: A mechanism for evicting a cache line from a cache memory includes first selecting for eviction a least recently used cache line of a group of invalid cache lines. If all cache lines are valid, selecting for eviction a least recently used cache line of a group of cache lines in which no cache line of the group of cache lines is also stored within a higher level cache memory such as the L1 cache, for example. Lastly, if all cache lines are valid and there are no non-inclusive cache lines, selecting for eviction the least recently used cache line stored in the cache memory.

Abstract: Systems, methods, and apparatuses for reducing writes to the data array of a cache. A cache hierarchy includes one or more L1 caches and a L2 cache inclusive of the L2 cache(s). When a request from the L1 cache misses in the L2 cache, the L2 cache sends a fill request to memory. When the fill data returns from memory, the L2 cache delays writing the fill data to its data array. Instead, this cache line is written to the L1 cache and a clean-evict bit corresponding to the cache line is set in the L1 cache. When the L1 cache evicts this cache line, the L1 cache will write back the cache line to the L2 cache even if the cache line has not been modified.

Abstract: A mechanism for evicting a cache line from a cache memory includes first selecting for eviction a least recently used cache line of a group of invalid cache lines. If all cache lines are valid, selecting for eviction a least recently used cache line of a group of cache lines in which no cache line of the group of cache lines is also stored within a higher level cache memory such as the L1 cache, for example. Lastly, if all cache lines are valid and there are no non-inclusive cache lines, selecting for eviction the least recently used cache line stored in the cache memory.

Abstract: A method and apparatus for selectively powering down a portion of a cache memory includes determining a power down condition dependent upon a number of accesses to the cache memory. In response to the detection of the power down condition, selecting a group of cache ways included in the cache memory dependent upon a number of cache lines in each cache way that are also included in another cache memory. The method further includes locking and flushing the selected group of cache ways, and then activating a low power mode for the selected group of cache ways.

Abstract: A method and apparatus for evicting cache lines from a cache memory includes receiving a request from one of a plurality of processors. The cache memory is configured to store a plurality of cache lines, and a given cache line includes an identifier indicating a processor that performed a most recent access of the given cache line. The method further includes selecting a cache line for eviction from a group of least recently used cache lines, where each cache line of the group of least recently used cache lines occupy a priority position less that a predetermined value, and then evicting the selected cache line.

Abstract: A system and method for efficiently handling misaligned memory accesses within a processor. A processor comprises a load-store unit (LSU) with a banked data cache (d-cache) and a banked store queue. The processor generates a first address corresponding to a memory access instruction identifying a first cache line. The processor determines the memory access is misaligned which crosses over a cache line boundary. The processor generates a second address identifying a second cache line logically adjacent to the first cache line. If the instruction is a load instruction, the LSU simultaneously accesses the d-cache and store queue with the first and the second addresses. If there are two hits, the data from the two cache lines are simultaneously read out. If the access is a store instruction, the LSU separates associated write data into two subsets and simultaneously stores these subsets in separate cache lines in separate banks of the store queue.

Abstract: Processors and methods for coordinating prefetch units at multiple cache levels. A single, unified training mechanism is utilized for training on streams generated by a processor core. Prefetch requests are sent from the core to lower level caches, and a packet is sent with each prefetch request. The packet identifies the stream ID of the prefetch request and includes relevant training information for the particular stream ID. The lower level caches generate prefetch requests based on the received training information.

Abstract: A circular queue implementing a scheme for prioritized reads is disclosed. In one embodiment, a circular queue (or buffer) includes a number of storage locations each configured to store a data value. A multiplexer tree is coupled between the storage locations and a read port. A priority circuit is configured to generate and provide selection signals to each multiplexer of the multiplexer tree, based on a priority scheme. Based on the states of the selection signals, one of the storage locations is coupled to the read port via the multiplexers of the multiplexer tree.

Abstract: Processors and methods for preventing lower level prefetch units from stalling at page boundaries. An upper level prefetch unit closest to the processor core issues a preemptive request for a translation of the next page in a given prefetch stream. The upper level prefetch unit sends the translation to the lower level prefetch units prior to the lower level prefetch units reaching the end of the current page for the given prefetch stream. When the lower level prefetch units reach the boundary of the current page, instead of stopping, these prefetch units can continue to prefetch by jumping to the next physical page number provided in the translation.

Abstract: Systems, methods, and apparatuses for reducing writes to the data array of a cache. A cache hierarchy includes one or more L1 caches and a L2 cache inclusive of the L2 cache(s). When a request from the L1 cache misses in the L2 cache, the L2 cache sends a fill request to memory. When the fill data returns from memory, the L2 cache delays writing the fill data to its data array. Instead, this cache line is written to the L1 cache and a clean-evict bit corresponding to the cache line is set in the L1 cache. When the L1 cache evicts this cache line, the L1 cache will write back the cache line to the L2 cache even if the cache line has not been modified.

Abstract: Systems, methods, and apparatuses for implementing selective victimization to reduce power and utilized bandwidth in a multi-level cache hierarchy. Each set of an upper-level cache includes a counter that keeps track of the number of times the set was accessed. These counters are periodically decremented by another counter that tracks the total number of accesses to the cache. If a given set counter is below a certain threshold value, clean victims are dropped from this given set instead of being sent to a lower-level cache. Also, a separate counter is used to track the total number of outstanding requests for the cache as a proxy for bus-bandwidth in order to gauge the total amount of traffic in the system. The cache will implement selective victimization whenever there is a large amount of traffic in the system.

Abstract: In an embodiment, a processor may implement an access map-pattern match (AMPM)-based prefetcher in which patterns may include wild cards for some cache blocks. The wild card may match any access for the corresponding cache block (e.g. no access, demand access, prefetch, successful prefetch, etc.). Furthermore, patterns with irregular strides and/or irregular access patterns may be included in the matching patterns and may be detected for prefetch generation. In an embodiment, the AMPM prefetcher may implement a chained access map for large streaming prefetches. If a stream is detected, the AMPM prefetcher may allocate a pair of map entries for the stream and may reuse the pair for subsequent access map regions within the stream. In some embodiments, a quality factor may be associated with each access map and may control the rate of prefetch generation.

Abstract: A circular queue implementing a scheme for prioritized reads is disclosed. In one embodiment, a circular queue (or buffer) includes a number of storage locations each configured to store a data value. A multiplexer tree is coupled between the storage locations and a read port. A priority circuit is configured to generate and provide selection signals to each multiplexer of the multiplexer tree, based on a priority scheme. Based on the states of the selection signals, one of the storage locations is coupled to the read port via the multiplexers of the multiplexer tree.

Abstract: A mechanism for evicting a cache line from a cache memory includes first selecting for eviction a least recently used cache line of a group of invalid cache lines. If all cache lines are valid, selecting for eviction a least recently used cache line of a group of cache lines in which no cache line of the group of cache lines is also stored within a higher level cache memory such as the L1 cache, for example. Lastly, if all cache lines are valid and there are no non-inclusive cache lines, selecting for eviction the least recently used cache line stored in the cache memory.