Abstract: A system and method are presented. Some embodiments include a processing unit, at least one memory coupled to the processing unit, and at least one cache coupled to the processing unit and divided into a series of blocks, wherein at least one of the series of cache blocks includes data identified as being in a modified state. The modified state data is flushed by writing the data to the at least one memory based on a write back policy and the aggressiveness of the policy is based on at least one factor including the number of idle cores, the proximity of the last cache flush, the activity of the thread associated with the data, and which cores are idle and if the idle core is associated with the data.

Abstract: A method and apparatus for idle phase prediction in integrated circuits is disclosed. In one embodiment, an integrated circuit (IC) includes a functional unit configured to cycle between intervals of an active state and an idle state. The IC further includes a prediction unit configured to record a history of idle state durations for a plurality of intervals of the idle state. Based on the history of idle state durations, the prediction unit is configured to generate a prediction of the duration of the next interval of the idle state. The prediction may be used by a power management unit to, among other uses, determine whether to place the functional unit in a low power (e.g., sleep) state.

Abstract: A method and apparatus for exiting a low power state based on a prior prediction is disclosed. An integrated circuit (IC) includes a functional unit configured to, during operation, cycle between intervals of an active state and intervals of an idle state. The IC also include a power management unit configured to place the functional unit in a low power state responsive to the functional unit entering the idle state. The power management unit is further configured to preemptively cause the functional unit to exit the low power state at a predetermined time after entering the low power. The predetermined time is based on a prediction of idle state duration made prior to entering the low power state. The prediction may be generated by a prediction unit, based on a history of durations of intervals in which the functional unit was in the idle state.

Abstract: A multi-core data processor includes multiple data processor cores and a power controller. Each data processor core has a first input for receiving a clock signal, a second input for receiving a power supply voltage, and an output for providing an idle signal. The power controller is coupled to each of the data processor cores for providing the clock signal and the power supply voltage to each of the data processor cores. The power controller provides at least one of the clock signal and the power supply voltage to an active one of the data processor cores in dependence on a number of idle signals received from the data processor cores.

Abstract: A processor discards spill data from a memory hierarchy in response to the final access to the spill data has been performed by a compiled program executing at the processor. In some embodiments, the final access determined based on a special-purpose load instruction configured for this purpose. In some embodiments the determination is made based on the location of a stack pointer indicating that a method of the executing program has returned, so that data of the returned method that remains in the stack frame is no longer to be accessed. Because the spill data is discarded after the final access, it is not transferred through the memory hierarchy.

Abstract: The described embodiments include a computing device with one or more entities (processor cores, processors, etc.). In some embodiments, during operation, a thermal power management unit in the computing device uses a linear prediction to compute a predicted duration of a next idle period for an entity based on the duration of one or more previous idle periods for the entity. Based on the predicted duration of the next idle period, the thermal power management unit configures the entity to operate in a corresponding idle state.

Abstract: A method of partitioning a data cache comprising a plurality of sets, the plurality of sets comprising a plurality of ways, is provided. Responsive to a stack data request, the method stores a cache line associated with the stack data in one of a plurality of designated ways of the data cache, wherein the plurality of designated ways is configured to store all requested stack data.

Abstract: The described embodiments include a computing device with a first entity and a second entity. In the computing device, a management controller dynamically sets a power-state limit for the first entity based on a performance coupling and a thermal coupling between the first entity and the second entity.

Abstract: The described embodiments include a core that uses predictions for store-to-load forwarding. In the described embodiments, the core comprises a load-store unit, a store buffer, and a prediction mechanism. During operation, the prediction mechanism generates a prediction that a load will be satisfied using data forwarded from the store buffer because the load loads data from a memory location in a stack. Based on the prediction, the load-store unit first sends a request for the data to the store buffer in an attempt to satisfy the load using data forwarded from the store buffer. If data is returned from the store buffer, the load is satisfied using the data. However, if the attempt to satisfy the load using data forwarded from the store buffer is unsuccessful, the load-store unit then separately sends a request for the data to a cache to satisfy the load.

Abstract: A method of way prediction for a data cache having a plurality of ways is provided. Responsive to an instruction to access a stack data block, the method accesses identifying information associated with a plurality of most recently accessed ways of a data cache to determine whether the stack data block resides in one of the plurality of most recently accessed ways of the data cache, wherein the identifying information is accessed from a subset of an array of identifying information corresponding to the plurality of most recently accessed ways; and when the stack data block resides in one of the plurality of most recently accessed ways of the data cache, the method accesses the stack data block from the data cache.

Abstract: A method of storing stack data in a cache hierarchy is provided. The cache hierarchy comprises a data cache and a stack filter cache. Responsive to a request to access a stack data block, the method stores the stack data block in the stack filter cache, wherein the stack filter cache is configured to store any requested stack data block.

Abstract: A device receives an indication that a memory bank is to be powered down, and determines, based on receiving the indication, shutdown scores corresponding to powered up memory banks. Each shutdown score is based on a shutdown metric associated with powering down a powered up memory bank. The device may power down a selected memory bank based on the shutdown scores.

Abstract: A device receives an indication that a memory bank is to be powered up, and determines, based on receiving the indication, power scores corresponding to powered down memory banks. Each power score corresponds to a power metric associated with powering up a powered down memory bank. The device powers up a selected memory bank based on the plurality of power scores.

Abstract: The present disclosure relates to a method and apparatus for dynamically controlling power consumption by at least one processor. A power management method includes monitoring, by power control logic of the at least one processor, performance data associated with each of a plurality of executions of a repetitive workload by the at least one processor. The method includes adjusting, by the power control logic following an execution of the repetitive workload, an operating frequency of at least one of a compute unit and a memory controller upon a determination that the at least one processor is at least one of compute-bound and memory-bound based on monitored performance data associated with the execution of the repetitive workload.

Abstract: An apparatus and method to enable a fast cache shutdown is disclosed. In one embodiment, a cache subsystem includes a cache memory and a cache controller coupled to the cache memory. The cache controller is configured to, upon restoring power to the cache subsystem, inhibit writing of modified data exclusively into the cache memory.

Abstract: One or more lines of a cache are prefetched according to a first prefetch routine while training a prefetcher to prefetch one or more lines of the cache according to a second prefetch routine. In response to determining that the prefetcher has been trained, one or more lines of the cache may be prefetched according to the second prefetch routine.

Abstract: A processor core includes a fetch control unit for fetching instructions and placing the instructions into an instruction queue and includes a branch predictor for controlling the fetch control unit to speculatively fetch at least one instruction subsequent to an unresolved branch instruction. The processor further includes a controller configured to dispatch instructions from the instruction queue and, in response to a branch misprediction of an unresolved control instruction, to apply a selected one of a checkpointing-based recovery mode and a commit-time-based recovery mode.

Abstract: A data prefetcher includes a controller to control operation of the data prefetcher. The controller receives data associated with cache misses and data associated with events that do not rely on a prefetching function of the data prefetcher. The data prefetcher also includes a counter to maintain a count associated with the data prefetcher. The count is adjusted in a first direction in response to detection of a cache miss, and in a second direction in response to detection of an event that does not rely on the prefetching function. The controller disables the prefetching function when the count reaches a threshold value.

Abstract: Methods and apparatus to efficiently detect faults are described. In an embodiment, an encoded value may be generated based on a portion of an instruction address and a portion of a corresponding result value. The encoded value may be used to determine whether an entry corresponding to the encoded value is absent from a screening storage unit. Other embodiments are also described.

Abstract: A processor is disclosed. The processor includes an execution core, a cache memory, and a prefetcher coupled to the cache memory. The prefetcher is configured to fetch a first cache line from a lower level memory and to load the cache line into the cache. The cache is further configured to designate the cache line as a most recently used (MRU) cache line responsive to the execution core asserting N demand requests for the cache line, wherein N is an integer greater than 1. The cache is configured to inhibit the cache line from being promoted to the MRU position if it receives fewer than N demand requests.