Abstract: A programmable hardware architecture for machine learning (ML) is proposed, which includes at least a host, a memory, a core, a data streaming engine, a instruction-streaming engine, and an interference engine. The core interprets a plurality of ML commands for a ML operation and/or data received from the host and coordinate activities of the engines based on the data in the received ML commands. The instruction-streaming engine translates the ML commands received from the core and provide a set of programming instructions to the data streaming engine and the inference engines based on the translated parameters. The data steaming engine sends one or more data streams to the inference engine in response to the received programming instructions. The inference engine then processes the data streams received from the data stream engine according to the programming instructions received from the instruction-streaming engine.

Abstract: A method for encapsulating metadata in a contextual data stream is described. The method may include receiving sensor data from a sensor. The method may include receiving metadata associated with the sensor data, wherein the metadata is defined as a probability vector based on a context set associated with the sensor data. The method may also include combining the sensor data and the metadata into a contextual data stream.

Abstract: A programmable hardware architecture for machine learning (ML) is proposed, which includes at least a host, a memory, a core, a data streaming engine, a instruction-streaming engine, and an interference engine. The core interprets a plurality of ML commands for a ML operation and/or data received from the host and coordinate activities of the engines based on the data in the received ML commands. The instruction-streaming engine translates the ML commands received from the core and provide a set of programming instructions to the data streaming engine and the inference engines based on the translated parameters. The data steaming engine sends one or more data streams to the inference engine in response to the received programming instructions. The inference engine then processes the data streams received from the data stream engine according to the programming instructions received from the instruction-streaming engine.

Abstract: Technologies for aggregation-based message processing include multiple computing nodes in communication over a network. A computing node receives a message from a remote computing node, increments an event counter in response to receiving the message, determines whether an event trigger is satisfied in response to incrementing the counter, and writes a completion event to an event queue if the event trigger is satisfied. An application of the computing node monitors the event queue for the completion event. The application may be executed by a processor core of the computing node, and the other operations may be performed by a host fabric interface of the computing node. The computing node may be a target node and count one-sided messages received from an initiator node, or the computing node may be an initiator node and count acknowledgement messages received from a target node. Other embodiments are described and claimed.

Abstract: Computing devices, computer-readable storage media, and methods associated with providing an operating system (OS)-absent firmware sensor layer to support a boot process are disclosed. A computing device may include a processor and firmware to be operated on the processor. The firmware may include one or more modules and a sensor layer. The sensor layer may be configured to receive, in the OS-absent environment, sensor data produced by a plurality of sensors. The sensor layer may be further configured to selectively provide the sensor data to the one or more modules via an interface of the sensor layer that abstracts the plurality of sensors.

Abstract: Techniques for adaptive demand/response power management. Power consumption and battery charge level of a platform having a battery with a smart power module are monitored. Information indicating the power consumption and battery charge level for the platform is provided to a remote demand/response management device. The remote demand/response management device and the smart power module receive a command to modify one or more power consumption characteristics of the platform. The one or more power consumption characteristics of the platform are to be changed in response to the command.

Abstract: Technologies for dynamic work queue management include a producer computing device communicatively coupled to a consumer computing device. The consumer computing device is configured to transmit a pop request (e.g., a one-sided pull request) that includes consumption constraints indicating an amount of work (e.g., a range of acceptable fraction of work elements to return from a work queue of the producer computing device) to pull from the producer computing device. The producer computing device is configured to determine whether the pop request can be satisfied and generate a response that includes an indication of the result of the determination and one or more producer metrics usable by the consumer computing device to determine a subsequent action to be performed by the consumer computing device upon receipt of the response message. Other embodiments are described and claimed herein.

Abstract: Technologies for aggregation-based message processing include multiple computing nodes in communication over a network. A computing node receives a message from a remote computing node, increments an event counter in response to receiving the message, determines whether an event trigger is satisfied in response to incrementing the counter, and writes a completion event to an event queue if the event trigger is satisfied. An application of the computing node monitors the event queue for the completion event. The application may be executed by a processor core of the computing node, and the other operations may be performed by a host fabric interface of the computing node. The computing node may be a target node and count one-sided messages received from an initiator node, or the computing node may be an initiator node and count acknowledgement messages received from a target node. Other embodiments are described and claimed.

Abstract: Computing devices, computer-readable storage media, and methods associated with providing an operating system (OS)-absent firmware sensor layer to support a boot process are disclosed herein. In embodiments, a computing device may include a processor and firmware to be operated on the processor. The firmware may include one or more modules and a sensor layer. The sensor layer may be configured to receive, in the OS-absent environment, sensor data produced by a plurality of sensors. The sensor layer may be further configured to selectively provide the sensor data to the one or more modules via an interface of the sensor layer that abstracts the plurality of sensors. Other embodiments may be described and/or claimed.

Abstract: A method for determining a location change based on contextual data is disclosed. The method may include gathering contextual data from ambient sensors of a device. The method may include determining a probability value that the device has changed location. The method may include comparing the probability value with a predetermined value. The method may include determining whether the location has changed based on the comparison of the probability value with the predetermined value.

Abstract: Techniques for adaptive demand/response power management. Power consumption and battery charge level of a platform having a battery with a smart power module are monitored. Information indicating the power consumption and battery charge level for the platform is provided to a remote demand/response management device. The remote demand/response management device and the smart power module receive a command to modify one or more power consumption characteristics of the platform. The one or more power consumption characteristics of the platform are to be changed in response to the command.

Abstract: In one embodiment, the present invention includes a power manager to receive a memory power usage value, to determine an available power based at least in part on a power budget and the memory power usage value, and to change a memory power state based at least in part on the available power, wherein the memory power state comprises a memory frequency and a memory voltage. Other embodiments are described and claimed.

Abstract: Described herein are techniques for dynamic memory frequency/voltage scaling to augment existing memory power management techniques and further improve memory power efficiency. Each operating point is defined as an operational state for the memory.

Abstract: Memory power estimation by means of calibrated weights and activity counters are generally presented. In this regard, in one embodiment, a memory power is introduced to read a value from a memory activity counter, to determine a memory power estimation based at least in part on the value and a calibration, and to store the memory power estimation to a register. Other embodiments are also described and claimed.

Abstract: Methods and apparatuses for adaptive memory operational state management. A memory performance parameter is determined for at least a portion of a memory system. The memory performance parameter is compared to one or more threshold values. An operating frequency of the memory system can be modified based on results of the comparison of the memory performance parameter and the one or more threshold values.

Abstract: In one embodiment, the present invention includes a power manager to receive a memory power usage value, to determine an available power based at least in part on a power budget and the memory power usage value, and to change a memory power state based at least in part on the available power, wherein the memory power state comprises a memory frequency and a memory voltage. Other embodiments are described and claimed.

Abstract: Methods and apparatuses for adaptive memory operational state management. A memory performance parameter is determined for at least a portion of a memory system. The memory performance parameter is compared to one or more threshold values. An operating frequency of the memory system can be modified based on results of the comparison of the memory performance parameter and the one or more threshold values.

Abstract: Memory power estimation by means of calibrated weights and activity counters are generally presented. In this regard, in one embodiment, a memory power is introduced to read a value from a memory activity counter, to determine a memory power estimation based at least in part on the value and a calibration, and to store the memory power estimation to a register. Other embodiments are also described and claimed.

Abstract: Described herein are techniques for dynamic memory frequency/voltage scaling to augment existing memory power management techniques and further improve memory power efficiency. Each operating point is defined as an operational state for the memory.

Abstract: Techniques for providing holistic views of energy consumption. Energy consumption of one or more energy consuming devices corresponding to a user is monitored. The energy consumption for the one or more energy consuming devices is aggregated. A graphical representation of the energy consumption is provided to the user for the one or more energy consuming devices and aggregate energy consumption. The graphical representation comprises at least one visual metaphor for energy consumption.