Abstract: Methods and apparatuses relating to processing neural networks are described. In one embodiment, an apparatus to process a neural network includes a plurality of fully connected layer chips coupled by an interconnect; a plurality of convolutional layer chips each coupled by an interconnect to a respective fully connected layer chip of the plurality of fully connected layer chips and each of the plurality of fully connected layer chips and the plurality of convolutional layer chips including an interconnect to couple each of a forward propagation compute intensive tile, a back propagation compute intensive tile, and a weight gradient compute intensive tile of a column of compute intensive tiles between a first memory intensive tile and a second memory intensive tile.

Abstract: Methods and apparatuses relating to processing neural networks are described. In one embodiment, an apparatus to process a neural network includes a plurality of fully connected layer chips coupled by an interconnect; a plurality of convolutional layer chips each coupled by an interconnect to a respective fully connected layer chip of the plurality of fully connected layer chips and each of the plurality of fully connected layer chips and the plurality of convolutional layer chips including an interconnect to couple each of a forward propagation compute intensive tile, a back propagation compute intensive tile, and a weight gradient compute intensive tile of a column of compute intensive tiles between a first memory intensive tile and a second memory intensive tile.

Abstract: In an embodiment, a processor includes a plurality of cores to independently execute instructions, a shared cache coupled to the cores and including a plurality of lines to store data, and a power controller including a low power control logic to calculate a flush latency to flush the shared cache based on a state of the plurality of lines. Other embodiments are described and claimed.

Abstract: Methods and apparatus to permit a system low power consumption state when CPU (Central Processing Unit) or generically any compute element is active are described. In an embodiment, a fabric and a memory controller are caused to enter a low power consumption state at least partially in response to a determination that the fabric and the memory controller are idle. The entry into the low power consumption state occurs while a compute element, coupled to the fabric and the memory controller, is in an active state. Other embodiments are also disclosed and claimed.

Abstract: Methods and apparatuses relating to processing neural networks are described. In one embodiment, an apparatus to process a neural network includes a plurality of fully connected layer chips coupled by an interconnect; a plurality of convolutional layer chips each coupled by an interconnect to a respective fully connected layer chip of the plurality of fully connected layer chips and each of the plurality of fully connected layer chips and the plurality of convolutional layer chips including an interconnect to couple each of a forward propagation compute intensive tile, a back propagation compute intensive tile, and a weight gradient compute intensive tile of a column of compute intensive tiles between a first memory intensive tile and a second memory intensive tile.

Abstract: In an embodiment, a processor includes a plurality of cores to independently execute instructions, a shared cache coupled to the cores and including a plurality of lines to store data, and a power controller including a low power control logic to calculate a flush latency to flush the shared cache based on a state of the plurality of lines. Other embodiments are described and claimed.

Abstract: In one example a sensor module comprises at least one sensor and a controller communicatively coupled to the at least one sensor by a communication bus, the controller comprising logic, at least partially including hardware logic, configured to generate a signal to configure the at least one sensor in a notify power state mode and place the signal on a communication bus coupled to the at least one sensor. Other examples may be described.

Abstract: The present disclosure describes techniques related to media caching. A media hub device may include a media hub device configured to execute an operation on a current frame of media having a frame period. The media hub device may include a cache configured to provide, to a media accelerator of the media hub device, data associated with the frame period of the current frame.

Abstract: In an embodiment, a processor includes a plurality of cores to independently execute instructions, a shared cache coupled to the cores and including a plurality of lines to store data, and a power controller including a low power control logic to calculate a flush latency to flush the shared cache based on a state of the plurality of lines. Other embodiments are described and claimed.

Abstract: Methods and apparatus to permit a system low power consumption state when CPU (Central Processing Unit) or generically any compute element is active are described. In an embodiment, a fabric and a memory controller are caused to enter a low power consumption state at least partially in response to a determination that the fabric and the memory controller are idle. The entry into the low power consumption state occurs while a compute element, coupled to the fabric and the memory controller, is in an active state. Other embodiments are also disclosed and claimed.

Abstract: In an embodiment, a processor includes a plurality of cores to independently execute instructions, a shared cache coupled to the cores and including a plurality of lines to store data, and a power controller including a low power control logic to calculate a flush latency to flush the shared cache based on a state of the plurality of lines. Other embodiments are described and claimed.

Abstract: In an embodiment, a processor includes a plurality of cores to independently execute instructions, a shared cache coupled to the cores and including a plurality of lines to store data, and a power controller including a low power control logic to calculate a flush latency to flush the shared cache based on a state of the plurality of lines. Other embodiments are described and claimed.

Abstract: In one example a sensor module comprises at least one sensor and a controller communicatively coupled to the at least one sensor by a communication bus, the controller comprising logic, at least partially including hardware logic, configured to generate a signal to configure the at least one sensor in a notify power state mode and place the signal on a communication bus coupled to the at least one sensor. Other examples may be described.

Abstract: Technologies for low-power display refresh standby include a computing device with a display such as an LCD panel. The computing device may include a system-on-a-chip (SoC) with a processor, I/O subsystem, display controller, and memory. When the computing device determines that a display image is static, the computing device enters a low-power display refresh standby mode, powering down unneeded components of the SoC such as processor cores, peripheral devices, and memory other than a dedicated display buffer. The display controller may access the dedicated display buffer via the I/O subsystem and output the image to the display. The computing device may power down the I/O subsystem and the dedicated display buffer when a display controller FIFO is full of image data, and periodically power on the I/O subsystem and display buffer to fill the display controller FIFO. Other embodiments are described and claimed.

Abstract: The present disclosure describes techniques related to media caching. A media hub device may include a media hub device configured to execute an operation on a current frame of media having a frame period. The media hub device may include a cache configured to provide, to a media accelerator of the media hub device, data associated with the frame period of the current frame.

Abstract: Technologies for low-power display refresh standby include a computing device with a display such as an LCD panel. The computing device may include a system-on-a-chip (SoC) with a processor, I/O subsystem, display controller, and memory. When the computing device determines that a display image is static, the computing device enters a low-power display refresh standby mode, powering down unneeded components of the SoC such as processor cores, peripheral devices, and memory other than a dedicated display buffer. The display controller may access the dedicated display buffer via the I/O subsystem and output the image to the display. The computing device may power down the I/O subsystem and the dedicated display buffer when a display controller FIFO is full of image data, and periodically power on the I/O subsystem and display buffer to fill the display controller FIFO. Other embodiments are described and claimed.

Abstract: In an embodiment, a processor includes a plurality of cores to independently execute instructions, a shared cache coupled to the cores and including a plurality of lines to store data, and a power controller including a low power control logic to calculate a flush latency to flush the shared cache based on a state of the plurality of lines. Other embodiments are described and claimed.

Abstract: A method includes generating a noise frame of data that is representative of a dark current image. Video frames of data are generated that represent video images. The video frames include noise. Information from the noise frame is used to compensate for the noise.

Abstract: A method and apparatus for processing an image is described. A first area having a first value is selected. A second value for a second area surrounding the first area is determined. The first and second values are compared with a predetermined threshold. The first value is transformed according to the comparison.

Abstract: A method includes determining an operating frequency of a processor. Based on the operating frequency, a routine is selected from a group of at least two routines, and the selected routine is executed. The routine may also be selected based on an unused processing bandwidth that is available to the processor.