Abstract: A method of performing automatic tuning on a deep learning model includes: utilizing an instruction-based learned cost model to estimate a first type of operational performance metrics based on a tuned configuration of layer fusion and tensor tiling; utilizing statistical data gathered during a compilation process of the deep learning model to determine a second type of operational performance metrics based on the tuned configuration of layer fusion and tensor tiling; performing an auto-tuning process to obtain a plurality of optimal configurations based on the first type of operational performance metrics and the second type of operational performance metrics; and configure the deep learning model according to one of the plurality of optimal configurations.

Abstract: A sense amplifier circuit includes a sense amplifier, a switch and a temperature compensation circuit. The temperature compensation circuit provides a control signal having a positive temperature coefficient, based on which the switch provides reference impedance for temperature compensation. The sense amplifier includes a first input end coupled to a target bit and a second input end coupled to the switch. The sense amplifier outputs a sense amplifier signal based on the reference impedance and the impedance of the target bit.

Abstract: A bottom-pinned spin-orbit torque magnetic random access memory (SOT-MRAM) is provided in the present invention, including a substrate, a bottom electrode layer on the substrate, a magnetic tunnel junction (MTJ) on the bottom electrode layer, a spin-orbit torque (SOT) layer on the MTJ, a capping layer on the SOT layer, and an injection layer on the capping layer, wherein the injection layer is divided into individual first part and second part, and the first part and the second part are connected respectively with two ends of the capping layer.

Abstract: Disclosed herein are system, method, and computer program product embodiments for reducing GPU load by programmatically controlling shading rates in computer graphics. GPU load may be reduced by applying different shading rates to different screen regions. By reading the depth buffer of previous frames and performing image processing, thresholds may be calculated that control the shading rates. The approach may be run on any platform that supports VRS hardware and primitive- or image-based VRS. The approach may be applied on a graphics driver installed on a client device, in a firmware layer between hardware and a driver, in a software layer between a driver and an application, or in hardware on the client device. The approach is flexible and adaptable and calculates and sets the variable rate shading based on the graphics generated by an application without requiring the application developer to manually set variable rate shading.

Abstract: The present invention provides a method for monitoring tracks and a track monitoring module. The method comprises: maintaining a target list, stored a target track, by a monitoring server; recording a first track related to a first mobile device by a first monitoring application; and comparing the similarity between the first track and the target track by the first monitoring application to generate a comparison result.

Abstract: Provided are compositions that may be referred to as photoreactive compositions or inks. The compositions may have a plurality of reactive components, which are silica nanocages with one or more photoreactive ligand(s). Also provided are methods of making an article of manufacture. Also provided are articles of manufacture and uses thereof. The article of manufacture may be formed from a composition of the present disclosure.

Abstract: Compositions including a plurality of reactive components. The reactive components each include an inorganic core with one or more photoresponsive ligand(s) covalently bonded to a surface of the inorganic core. A composition may also include a photoinitiator. Methods of making an article of manufacture includes photochemically reacting at least a portion of one or more layer(s) formed from one or more composition(s). The photochemical reaction may be carried using a laser as a source of electromagnetic radiation. An article of manufacture, which may be a three-dimensional article of manufacture, may be or is a part of a microfluidic device, HPLC column, fluidic channel, point of care device, or diagnostics device.

Abstract: Data management methods and systems for handheld devices are provided. First, a plurality of email messages are displayed in a first display area of a touch-sensitive display unit, and at least one tab icon is displayed in a second display area of the touch-sensitive display unit, wherein the tab icon records at least one condition, and the condition is used to determine whether the respective email message has a flag or has a meeting invitation format. A contact corresponding to the tab icon on the touch-sensitive display unit is detected, and in response to the contact, at least one candidate email message having the flag or having the meeting invitation format is filtered from the email messages according to the condition corresponding to the tab icon, and the candidate email message is displayed in the first display area of the touch-sensitive display unit.