Abstract:
Aspects include computing devices, systems, and methods for implementing a cache memory access requests for data smaller than a cache line and eliminating overfetching from a main memory by combining the data with padding data of a size of a difference between a size of a cache line and the data. A processor may determine whether the data, uncompressed or compressed, is smaller than a cache line using a size of the data or a compression ratio of the data. The processor may generate the padding data using constant data values or a pattern of data values. The processor may send a write cache memory access request for the combined data to a cache memory controller, which may write the combined data to a cache memory. The cache memory controller may send a write memory access request to a memory controller, which may write the combined data to a memory.
Abstract:
Devices and methods for monitoring one or more central processing units in real time are disclosed. The method may include monitoring state data associated with the one or more CPUs in real-time, filtering the state data, and at least partially based on filtered state data, selectively altering one or more system settings. A device may include means for monitoring state data associated with the one or more CPUs in real-time, means for filtering the state data, and means for selectively altering one or more system settings at least partially based on filtered state data. A device may also include a sub-sampling circuit configured to receive a hardware core signal from the central processing unit and output a central processing unit state indication, and an infinite impulse response filter connected to the sub-sampling circuit and configured to receive the central processing unit state indication from the sub-sampling circuit.
Abstract:
A method of controlling power within a multicore central processing unit (CPU) is disclosed. The method may include monitoring a die temperature, determining a degree of parallelism within a workload of the CPU, and powering one or more cores of the CPU up or down based on the degree of parallelism, the die temperature, or a combination thereof.
Abstract:
Methods, systems and devices that include a dynamic clock and voltage scaling (DCVS) solution configured to compute and enforce performance guarantees for a group of processors to ensure that the processors does not remain in a busy state (e.g., due to transient workloads) for a combined period that is more than a predetermined amount of time above that which is required for one of the processors to complete its pre-computed steady state workload. The DCVS may adjust the frequency and/or voltage of one or more of the processors based on a variable delay to ensure that the multiprocessor system only falls behind its steady state workload by, at most, a predefined maximum amount of work, irrespective of the operating frequency or voltage of the processors.
Abstract:
Devices and methods for monitoring one or more central processing units in real time is disclosed. The method may include monitoring state data associated with the one or more CPUs in real-time, filtering the state data, and at least partially based on filtered state data, selectively altering one or more system settings. A device may include means for monitoring state data associated with the one or more CPUs in real-time, means for filtering the state data, and means for selectively altering one or more system settings at least partially based on filtered state data. A device may also include a sub-sampling circuit configured to receive a hardware core signal from the central processing unit and output a central processing unit state indication, and an infinite impulse response filter connected to the sub-sampling circuit and configured to receive the central processing unit state indication from the sub-sampling circuit.
Abstract:
Aspects include computing devices, systems, and methods for implementing a cache memory access requests for data smaller than a cache line and eliminating overfetching from a main memory by combining the data with padding data of a size of a difference between a size of a cache line and the data. A processor may determine whether the data, uncompressed or compressed, is smaller than a cache line using a size of the data or a compression ratio of the data. The processor may generate the padding data using constant data values or a pattern of data values. The processor may send a write cache memory access request for the combined data to a cache memory controller, which may write the combined data to a cache memory. The cache memory controller may send a write memory access request to a memory controller, which may write the combined data to a memory.
Abstract:
A method of controlling power within a multicore central processing unit (CPU) is disclosed. The method may include monitoring a die temperature, determining a degree of parallelism within a workload of the CPU, and powering one or more cores of the CPU up or down based on the degree of parallelism, the die temperature, or a combination thereof.
Abstract:
Methods, systems and devices that include a dynamic clock and voltage scaling (DCVS) solution configured to compute and enforce performance guarantees for a group of processors to ensure that the processors does not remain in a busy state (e.g., due to transient workloads) for a combined period that is more than a predetermined amount of time above that which is required for one of the processors to complete its pre-computed steady state workload. The DCVS may adjust the frequency and/or voltage of one or more of the processors based on a variable delay to ensure that the multiprocessor system only falls behind its steady state workload by, at most, a predefined maximum amount of work, irrespective of the operating frequency or voltage of the processors.
Abstract:
Methods, systems and devices that include a dynamic clock and voltage scaling (DCVS) solution configured to compute and enforce performance guarantees to ensure that a processor does not remain in a busy state (e.g., due to transient workloads) for more than a predetermined amount of time above that which is required for that processor to complete its pre-computed steady state workload. The DCVS may adjust the frequency and/or voltage of a processor based on a variable delay to ensure that the processing core only falls behind its steady state workload by, at most, a predefined maximum amount of work, irrespective of the operating frequency or voltage of the processor.
Abstract:
Methods, systems and devices that include a dynamic clock and voltage scaling (DCVS) solution configured to compute and enforce performance guarantees to ensure that a processor does not remain in a busy state (e.g., due to transient workloads) for more than a predetermined amount of time above that which is required for that processor to complete its pre-computed steady state workload. The DCVS may adjust the frequency and/or voltage of a processor based on a variable delay to ensure that the processing core only falls behind its steady state workload by, at most, a predefined maximum amount of work, irrespective of the operating frequency or voltage of the processor.