Abstract: Systems, apparatuses, and methods include technology that determines, with a neural network, that a first eviction node stored in a cache will be evicted from the cache based on a cache policy. The first eviction node is part of a plurality of nodes associated with a graph. Further, a subset of nodes of the plurality of nodes remains in the cache after the eviction of the first eviction node from the cache. The technology further tracks a number of cache hits on the cache during an aggregation operation associated with a hardware accelerator, where the aggregation operation is executed on the subset of nodes that remain in the cache after the eviction of the eviction node from the cache. The technology executes a training process on the neural network to adjust the cache policy based on the number of the cache hits.

Abstract: Embodiments of the invention are generally directed to systems, methods, and apparatuses for linear to physical address translation with support for page attributes. In some embodiments, a system receives an instruction to translate a memory pointer to a physical memory address for a memory location. The system may return the physical memory address and one or more page attributes. Other embodiments are described and claimed.

Abstract: Methods and apparatus to implement multiple inference compute engines are disclosed herein. A disclosed example apparatus includes a first inference compute engine, a second inference compute engine, and an accelerator on coherent fabric to couple the first inference compute engine and the second inference compute engine to a converged coherency fabric of a system-on-chip, the accelerator on coherent fabric to arbitrate requests from the first inference compute engine and the second inference compute engine to utilize a single in-die interconnect port.

Abstract: Embodiments of the invention are generally directed to systems, methods, and apparatuses for linear to physical address translation with support for page attributes. In some embodiments, a system receives an instruction to translate a memory pointer to a physical memory address for a memory location. The system may return the physical memory address and one or more page attributes. Other embodiments are described and claimed.

Abstract: Systems, apparatus and methods are described including operations for a flexible neural network accelerator.

Abstract: Techniques to output a media stream, capture a media stream, or synchronize the output or capture of the media stream at a specified time are described. A media stream output or capture apparatus may include a media processor to receive a media stream to output or a request to capture a media stream and a start time. A buffer generator may be included to generate an input or an output buffer and a media mixer may be included to mix the media stream into the output buffer at the start time or capture the media stream from the input buffer at the start time.

Abstract: Systems, apparatus and methods are described including operations for a dual mode GMM (Gaussian Mixture Model) scoring accelerator for both speech and video data.

Abstract: Methods and apparatus to implement multiple inference compute engines are disclosed herein. A disclosed example apparatus includes a first inference compute engine, a second inference compute engine, and an accelerator on coherent fabric to couple the first inference compute engine and the second inference compute engine to a converged coherency fabric of a system-on-chip, the accelerator on coherent fabric to arbitrate requests from the first inference compute engine and the second inference compute engine to utilize a single in-die interconnect port.

Abstract: A device for positioning a tile, comprising: a beam reader for receiving a laser beam within a predetermined range of locations, and providing a height indication; a level for providing a tilt indication of a tilt of the device; a processor for receiving the height and tilt indications, and combining them for determining gradient and height of a part of the tile; a display component for displaying in at least two dimensions positioning information for the device, the positioning information relating to the gradient and height of the tile part, as obtained by combining the height and tilt indications, wherein the beam being received at a predetermined location within the range of locations, and the level indicating that the device is at a predetermined tilt, provides that the tile being positioned is at a required height, and wherein the device is adapted to be stably placed on the tile.

Abstract: Examples are disclosed for exchanging a key between an input/output device for network device and a first processing element operating on the network device. Data having a destination associated with the first processing element may be received by the input/output device. The exchanged key may be used to encrypt the received data. The encrypted data may then be sent to a buffer maintained at least in part in a memory for the network device. The memory may be arranged to enable sharing of the buffer with at least a second processing element operating on the network device. Examples are also disclosed for the processing element to receive an indication of the storing of the encrypted data in the buffer. The processing element may then obtain the encrypted data from the buffer and decrypt the data using the exchanged key.

Abstract: Systems, apparatus and methods are described including operations for a flexible neural network accelerator.

Abstract: Embodiments of the invention are generally directed to systems, methods, and apparatuses for linear to physical address translation with support for page attributes. In some embodiments, a system receives an instruction to translate a memory pointer to a physical memory address for a memory location. The system may return the physical memory address and one or more page attributes. Other embodiments are described and claimed.

Abstract: Embodiments of an apparatus for improving performance for events handling are presented. In one embodiment, the apparatus includes a number of processing elements and task routing logic. If at least one of the processing elements is in a turbo mode, the task routing logic selects a processing element for executing a task based at least on a comparison of performance losses.

Abstract: Embodiments of systems, apparatuses, and methods for posting interrupts to virtual processors are disclosed. In one embodiment, an apparatus includes look-up logic and posting logic. The look-up logic is to look-up an entry associated with an interrupt request to a virtual processor in a data structure. The posting logic is to post the interrupt request in a data structure specified by information in the first data structure.

Abstract: Systems, apparatus and methods are described including operations for a dual mode GMM (Gaussian Mixture Model) scoring accelerator for both speech and video data.

Abstract: A processor includes at least one core, a power control unit, and a first interconnect to couple with a peripheral controller. The first interconnect is to provide a first uni-directional communication path for communication of first power management data from the processor to the peripheral controller. Other embodiments are described and claimed.

Abstract: Systems, apparatus and methods are described including operations for memory access via direct memory access engines, of a Gaussian Mixture Model Accelerator, corresponding to individual data streams.

Abstract: A device for positioning a tile, comprising: a beam reader for receiving a laser beam within a predetermined range of locations, and providing a height indication; a level for providing a tilt indication of a tilt of the device; a processor for receiving the height and tilt indications, and combining them for determining gradient and height of a part of the tile; a display component for displaying in at least two dimensions positioning information for the device, the positioning information relating to the gradient and height of the tile part, as obtained by combining the height and tilt indications, wherein the beam being received at a predetermined location within the range of locations, and the level indicating that the device is at a predetermined tilt, provides that the tile being positioned is at a required height, and wherein the device is adapted to be stably placed on the tile.

Abstract: Methods and apparatus relating to controlling power consumption by a power management link are described. In one embodiment, the physical interface of a power management (PM) link is shut down when a processor is in a sleep state (e.g., to conserve power), while maintaining the availability of the processor for communication to a (e.g., embedded) controller over the PM link. Other embodiments are also disclosed and claimed.

Abstract: Methods and apparatus relating to measuring time offsets between devices with independent silicon clocks are described. In some embodiments, logic is provided to synchronize a first clock of a first agent with a second clock of a second agent based on one or more messages exchanged between the first agent and the second agent and a platform time. The first agent and the second agent are coupled via a link. Other embodiments are also disclosed and claimed.