Abstract: A processing system includes an interconnect fabric coupleable to a local memory and at least one compute cluster coupled to the interconnect fabric. The compute cluster includes a processor core and a cache hierarchy. The cache hierarchy has a plurality of caches and a throttle controller configured to throttle a rate of memory requests issuable by the processor core based on at least one of an access latency metric and a prefetch accuracy metric. The access latency metric represents an average access latency for memory requests for the processor core and the prefetch accuracy metric represents an accuracy of a prefetcher of a cache of the cache hierarchy.

Abstract: A processing system employs an expandable memory buffer that supports enlarging the memory buffer when the processing system generates a large number of long latency memory transactions. The hybrid structure of the memory buffer allows a memory controller of the processing system to store a larger number of memory transactions while still maintaining adequate transaction throughput and also ensuring a relatively small buffer footprint and power consumption. Further, the hybrid structure allows different portions of the buffer to be placed on separate integrated circuit dies, which in turn allows the memory controller to be used in a wide variety of integrated circuit configurations, including configurations that use only one portion of the memory buffer.

Abstract: A method for steering data for an I/O write operation includes, in response to receiving the I/O write operation, identifying, at an interconnect fabric, a cache of one of a plurality of compute complexes as a target cache for steering the data based on at least one of: a software-provided steering indicator, a steering configuration implemented at boot initialization, and coherency information for a cacheline associated with the data. The method further includes directing, via the interconnect fabric, the identified target cache to cache the data from the I/O write operation. The data is temporarily buffered at the interconnect fabric, and if the target cache attempts to fetch the data while the data is still buffered at the interconnect fabric, the interconnect fabric provides a copy of the buffered data in response to the fetch operation instead of initiating a memory access operation to obtain the data from memory.

Abstract: A method includes monitoring, at a cache coherence directory, states of cachelines stored in a cache hierarchy of a data processing system using a plurality of entries of the cache coherence directory. Each entry of the cache coherence directory is associated with a corresponding cache page of a plurality of cache pages, and each cache page representing a corresponding set of contiguous cachelines. The method further includes selectively evicting cachelines from a first cache of the cache hierarchy based on cacheline utilization densities of cache pages represented by the corresponding entries of the plurality of entries of the cache coherence directory.

Abstract: A system for managing cache utilization includes a processor core, a lower-level cache, and a higher-level cache. In response to activating the higher-level cache, the system counts lower-level cache victims evicted from the lower-level cache. While a count of the lower-level cache victims is not greater than a threshold number, the system transfers each lower-level cache victim to a system memory without storing the lower-level cache victim to the higher-level cache. When the count of the lower-level cache victims is greater than the threshold number, the system writes each lower-level cache victim to the higher-level cache. In this manner, if the higher-level cache is deactivated before the threshold number of lower-level cache victims is reached, the higher-level cache is empty and thus may be deactivated without flushing.

Abstract: A method includes monitoring, at a cache coherence directory, states of cachelines stored in a cache hierarchy of a data processing system using a plurality of entries of the cache coherence directory. Each entry of the cache coherence directory is associated with a corresponding cache page of a plurality of cache pages, and each cache page representing a corresponding set of contiguous cachelines. The method further includes selectively evicting cachelines from a first cache of the cache hierarchy based on cacheline utilization densities of cache pages represented by the corresponding entries of the plurality of entries of the cache coherence directory.

Abstract: A processor core associated with a first cache initiates entry into a powered-down state. In response, information representing a set of entries of the first cache are stored in a retention region that receives a retention voltage while the processor core is in a powered-down state. Information indicating one or more invalidated entries of the set of entries is also stored in the retention region. In response to the processor core initiating exit from the powered-down state, entries of the first cache are restored using the stored information representing the entries and the stored information indicating the at least one invalidated entry.

Abstract: A processing system includes a plurality of processor cores and a plurality of private caches. Each private cache is associated with a corresponding processor core of the plurality of processor cores and includes a corresponding first set of cachelines. The processing system further includes a shared cache shared by the plurality of processor cores. The shared cache includes a second set of cachelines, and a shadow tag memory including a plurality of entries, each entry storing state information for a corresponding cacheline of the first set of cachelines of one of the private caches.

Abstract: A system for managing cache utilization includes a processor core, a lower-level cache, and a higher-level cache. In response to activating the higher-level cache, the system counts lower-level cache victims evicted from the lower-level cache. While a count of the lower-level cache victims is not greater than a threshold number, the system transfers each lower-level cache victim to a system memory without storing the lower-level cache victim to the higher-level cache. When the count of the lower-level cache victims is greater than the threshold number, the system writes each lower-level cache victim to the higher-level cache. In this manner, if the higher-level cache is deactivated before the threshold number of lower-level cache victims is reached, the higher-level cache is empty and thus may be deactivated without flushing.

Abstract: A system for managing cache utilization includes a processor core, a lower-level cache, and a higher-level cache. In response to activating the higher-level cache, the system counts lower-level cache victims evicted from the lower-level cache. While a count of the lower-level cache victims is not greater than a threshold number, the system transfers each lower-level cache victim to a system memory without storing the lower-level cache victim to the higher-level cache. When the count of the lower-level cache victims is greater than the threshold number, the system writes each lower-level cache victim to the higher-level cache. In this manner, if the higher-level cache is deactivated before the threshold number of lower-level cache victims is reached, the higher-level cache is empty and thus may be deactivated without flushing.

Abstract: The present application describes embodiments of a method and apparatus for concurrently accessing dirty bits in a cache. One embodiment of the apparatus includes a cache configurable to store a plurality of lines. The lines are grouped into a plurality of subsets the plurality of lines. This embodiment of the apparatus also includes a plurality of dirty bits associated with the plurality of lines and first circuitry configurable to concurrently access the plurality of dirty bits associated with at least one of the plurality of subsets of lines.

Abstract: A method and apparatus for performing a bus lock and a translation lookaside buffer invalidate transaction includes receiving, by a lock master, a lock request from a first processor in a system. The lock master sends a quiesce request to all processors in the system, and upon receipt of the quiesce request from the lock master, all processors cease issuing any new transactions and issue a quiesce granted transaction. Upon receipt of the quiesce granted transactions from all processors, the lock master issues a lock granted message that includes an identifier of the first processor. The first processor performs an atomic transaction sequence and sends a first lock release message to the lock master upon completion of the atomic transaction sequence. The lock master sends a second lock release message to all processors upon receiving the first lock release message from the first processor.

Abstract: A system for managing cache utilization includes a processor core, a lower-level cache, and a higher-level cache. In response to activating the higher-level cache, the system counts lower-level cache victims evicted from the lower-level cache. While a count of the lower-level cache victims is not greater than a threshold number, the system transfers each lower-level cache victim to a system memory without storing the lower-level cache victim to the higher-level cache. When the count of the lower-level cache victims is greater than the threshold number, the system writes each lower-level cache victim to the higher-level cache. In this manner, if the higher-level cache is deactivated before the threshold number of lower-level cache victims is reached, the higher-level cache is empty and thus may be deactivated without flushing.

Abstract: Apparatus and method embodiments for dynamically allocating cache space in a multi-threaded execution environment are disclosed. In some embodiments, a processor includes a cache shared by each of a plurality of processor cores and/or each of a plurality of threads executing on the processor. The processor further includes a cache allocation circuit configured to dynamically allocate space in the cache provided to each of the plurality of processor cores based on their respective usage patterns. The cache allocation unit may track cache usage by each of the processor cores/threads using subsets of usage bits and counters configured to update states of the usage bits. The cache allocation circuit may track the usage of cache space by the processor cores/threads and may allocate more space to those that exhibit more usage of the cache.

Abstract: A processing system includes a plurality of processor cores and a plurality of private caches. Each private cache is associated with a corresponding processor core of the plurality of processor cores and includes a corresponding first set of cachelines. The processing system further includes a shared cache shared by the plurality of processor cores. The shared cache includes a second set of cachelines, and a shadow tag memory including a plurality of entries, each entry storing state information for a corresponding cacheline of the first set of cachelines of one of the private caches.

Abstract: Apparatus and method embodiments for dynamically allocating cache space in a multi-threaded execution environment are disclosed. In some embodiments, a processor includes a cache shared by each of a plurality of processor cores and/or each of a plurality of threads executing on the processor. The processor further includes a cache allocation circuit configured to dynamically allocate space in the cache provided to each of the plurality of processor cores based on their respective usage patterns. The cache allocation unit may track cache usage by each of the processor cores/threads using subsets of usage bits and counters configured to update states of the usage bits. The cache allocation circuit may track the usage of cache space by the processor cores/threads and may allocate more space to those that exhibit more usage of the cache.

Abstract: Apparatus and method embodiments for dynamically allocating cache space in a multi-threaded execution environment are disclosed. In some embodiments, a processor includes a cache shared by each of a plurality of processor cores and/or each of a plurality of threads executing on the processor. The processor further includes a cache allocation circuit configured to dynamically allocate space in the cache provided to each of the plurality of processor cores based on their respective usage patterns. The cache allocation unit may track cache usage by each of the processor cores/threads using subsets of usage bits and counters configured to update states of the usage bits. The cache allocation circuit may track the usage of cache space by the processor cores/threads and may allocate more space to those that exhibit more usage of the cache.

Abstract: Methods, integrated circuit devices, and fabrication processes relating to power management transitions of a compute unit comprising a cache are presented. One method includes, responsive to an indication that the compute unit is attempting to enter a low power state, detecting at least one line of the cache differing from the corresponding line in memory, writing differing data from the at least one differing line to the memory, flushing at least one remaining differing line of the cache, and permitting the compute unit to enter the low power state, wherein the detecting and the writing are performed at a first frequency prior to the indication and at a second frequency subsequent the indication, and the second frequency is higher than the first frequency.

Abstract: A method includes storing architectural state data associated with a processing unit in a cache memory using an allocate without fill mode. A system includes a processing unit, a cache memory, and a cache controller. The cache controller is to receive architectural state data associated with the processing unit and store at least a first portion of the architectural state data in the cache memory using a first fill mode responsive to a first value of a fill mode flag and store at least a second portion of the architectural state data in the cache memory using a second fill mode responsive to a second value of a fill mode flag, wherein the first fill mode differs from the second fill mode with respect to whether previous values of the architectural state data are retrieved prior to storing the first or second portions in the cache memory.

Abstract: We report methods, integrated circuit devices, and fabrication processes relating to power management transitions of multiple compute units sharing a cache. One method includes indicating that a first compute unit of a plurality of compute units of an integrated circuit device is attempting to enter a low power state, determining if the first compute unit is the only compute unit of the plurality in a normal power state, and in response to determining the first compute unit is the only compute unit in the normal power state: saving a state of a shared cache unit of the integrated circuit device, flushing at least a portion of a cache of the shared cache unit, repeating the flushing until either a second compute unit exits the low power state or the cache is completely flushed, and permitting the first compute unit to enter the low power state.