Abstract: Techniques are disclosed relating to cache thrash detection. In some embodiments, cache controller circuitry is configured to monitor and track performance metrics across multiple levels of a cache hierarchy, detect cache thrashing based on one or more performance metrics, and modify a cache insertion policy to mitigate cache thrashing. Disclosed techniques may advantageously detect and reduce or avoid cache thrashing, which may increase processor performance, decrease power consumption for a given workload, or both, relative to traditional techniques.

Abstract: A method of interleaving a memory by mapping address bits of the memory to a number N of memory channels iteratively in successive rounds, wherein in each round except the last round: selecting a unique subset of address bits, determining a maximum number (L) of unique combinations possible based on the selected subset of address bits, mapping combinations to the N memory channels a maximum number of times (F) possible where each of the N memory channels gets mapped to an equal number of combinations, and if and when a number of combinations remain (K, which is less than N) that cannot be mapped, one to each of the N memory channels, entering a next round. In the last round, mapping remaining most significant address bits, not used in the subsets in prior rounds, to each of the N memory channels.

Abstract: Ephemeral data stored in a cache is read when needed but is not written to system memory so as to save power and bandwidth. In an embodiment, a no-writeback bit associated with the ephemeral data is set in response to a read-no-writeback instruction. Data in a cache line for which its no-writeback bit has been set is not written back into system memory. Accordingly, when evicting cache lines, if a cache line has a no-writeback bit set, then the data in that cache line is discarded without being written back to system memory.

Abstract: A method of interleaving a memory by mapping address bits of the memory to a number N of memory channels iteratively in successive rounds, wherein in each round except the last round: selecting a unique subset of address bits, determining a maximum number (L) of unique combinations possible based on the selected subset of address bits, mapping combinations to the N memory channels a maximum number of times (F) possible where each of the N memory channels gets mapped to an equal number of combinations, and if and when a number of combinations remain (K, which is less than N) that cannot be mapped, one to each of the N memory channels, entering a next round. In the last round, mapping remaining most significant address bits, not used in the subsets in prior rounds, to each of the N memory channels.

Abstract: A system and method for a processor to determine a memory page management implementation used by a memory controller without necessarily having direct access to the circuits or registers of the memory controller is disclosed. In one embodiment, a matrix of counters correspond to potential page management implementations and numbers of pages per block. The counters may be incremented or decremented depending upon whether the corresponding page management implementations and numbers of pages predict a page boundary whenever a long access latency is observed. The counter with the largest value after a period of time may correspond to the actual page management implementation and number of pages per block.

Abstract: A method and apparatus to detect and filter out redundant cache line addresses in a prefetch input queue, and to adjust the detector window size dynamically according to the number of detector entries in the queue for the cache-to-memory controller bus. Detectors correspond to cache line addresses that may represent cache misses in various levels of cache memory.

Abstract: A method and apparatus to detect and filter out redundant cache line addresses in a prefetch input queue, and to adjust the detector window size dynamically according to the number of detector entries in the queue for the cache-to-memory controller bus. Detectors correspond to cache line addresses that may represent cache misses in various levels of cache memory.

Abstract: A system and method for a processor to determine a memory page management implementation used by a memory controller without necessarily having direct access to the circuits or registers of the memory controller is disclosed. In one embodiment, a matrix of counters correspond to potential page management implementations and numbers of pages per block. The counters may be incremented or decremented depending upon whether the corresponding page management implementations and numbers of pages predict a page boundary whenever a long access latency is observed. The counter with the largest value after a period of time may correspond to the actual page management implementation and number of pages per block.

Abstract: A system and method for a processor to determine a memory page management implementation used by a memory controller without necessarily having direct access to the circuits or registers of the memory controller is disclosed. In one embodiment, a matrix of counters correspond to potential page management implementations and numbers of pages per block. The counters may be incremented or decremented depending upon whether the corresponding page management implementations and numbers of pages predict a page boundary whenever a long access latency is observed. The counter with the largest value after a period of time may correspond to the actual page management implementation and number of pages per block.

Abstract: A method and apparatus to detect and filter out redundant cache line addresses in a prefetch input queue, and to adjust the detector window size dynamically according to the number of detector entries in the queue for the cache-to-memory controller bus. Detectors correspond to cache line addresses that may represent cache misses in various levels of cache memory.