Abstract: In various examples, a single camera is used to capture two images of a scene from different locations. A trained neural network, taking the two images as inputs, outputs a scene structure map that indicates a ratio of height and depth values for pixel locations associated with the images. This ratio may indicate the presence of an object above a surface (e.g., road surface) within the scene. Object detection then can be performed on non-zero values or regions within the scene structure map.

Abstract: In various examples, firewalls may include machine learning models that are automatically trained and applied to analyze service inputs submitted to input processing services and to identify whether service inputs are desirable (e.g., will result in an undesirable status code if processed by a service). When a service input is determined by a firewall to be desirable, the firewall may push the service input through to the input processing service for normal processing. When a service input is determined by the firewall to be undesirable, the firewall may block or drop the service input before it reaches the input processing service and/or server. This may be used to prevent the service input, which is likely to be undesirable, from touching a server that hosts the input processing service (e.g., preventing a crash).

Abstract: Systems and methods for datacenter are disclosed. In at least one embodiment, a system or method herein causes a process for a structure that includes a component tracking system, where such a system includes individual components within individual servers or individual racks, the process being based in part on at least location information and configuration information from one or more of an optical sensor or a radio sensor associated with a motile-support that is adapted for at least three dimensional (3D) movement in a space having the individual servers or the individual racks.

Abstract: Systems and methods are disclosed related to low-power dynamic offset calibration of an error amplifier. An analog linear voltage regulator circuit tracks changes between a reference voltage and a regulated voltage to keep the regulated voltage as close as possible to the reference voltage. The analog linear voltage regulator includes an error amplifier that measures the error between the reference and regulated voltages and feedback circuitry. The error amplifier and feedback circuitry should be calibrated to correct for any offset within the circuits. The described offset calibration technique not only compensates for the offset in the error amplifier but also cancels any mismatch in the feedback network. During operation, conditions such as temperature and supply voltage may vary causing the offset to change. The technique is low power and dynamically cancels the offset even when the linear regulator is operating to supply the desired voltage.

Abstract: A manifold voxel mesh or surface mesh is manufacturable by carving a single block of material and a non-manifold mesh is not manufacturable. Conventional techniques for constructing or extracting a surface mesh from an input point cloud often produce a non-manifold voxel mesh. Similarly, extracting a surface mesh from a voxel mesh that includes non-manifold geometry produces a surface mesh that includes non-manifold geometry. To ensure that the surface mesh includes only manifold geometry, locations of the non-manifold geometry in the voxel mesh are detected and converted into manifold geometry. The result is a manifold voxel mesh from which a manifold surface mesh of the object may be extracted.

Abstract: State information can be determined for a subject that is robust to different inputs or conditions. For drowsiness, facial landmarks can be determined from captured image data and used to determine a set of blink parameters. These parameters can be used, such as with a temporal network, to estimate a state (e.g., drowsiness) of the subject. To improve robustness, an eye state determination network can determine eye state from the image data, without reliance on intermediate landmarks, that can be used, such as with another temporal network, to estimate the state of the subject. A weighted combination of these values can be used to determine an overall state of the subject. To improve accuracy, individual behavior patterns and context information can be utilized to account for variations in the data due to subject variation or current context rather than changes in state.

Abstract: Raytracing can be used to generate high quality, physics-based water caustics patterns in real time. A caustics map is generate to represent locations and normals of points across a water surface. Rays from a light source that are reflected and refracted from these points, as determined by the locations and normals, and can generate hit points on a surface. Neighboring points can be used to help determine the resulting caustics pattern. In one embodiment, information for neighboring points in the caustics map can be used to generate scale factors for geometric regions to be projected onto the surface for each hit point. In another embodiment, these points serve as vertices of a caustic mesh that can be projected onto the surface, where the brightness at a primitive is determined by the size of the primitive area defined by the vertices of the caustics mesh.

Abstract: An electronic device includes an integrated circuit and a heat exchanger. The heat exchanger includes a heat pipe and a first plurality of cooling fins and a second plurality of cooling fins. The heat pipe is thermally coupled to the integrated circuit and has an evaporator portion and a condenser portion, where the condenser portion extends away from the evaporator portion. The first plurality of cooling fins are attached to the condenser portion and proximate to the evaporation portion and form a plenum having a first associated pressure drop when a cooling fluid flows across the first plurality of cooling fins at a first velocity. The second plurality of cooling fins are attached to the condenser portion and distal from the evaporation portion and form a flow path having a second associated pressure drop when the cooling fluid flows across the second plurality of cooling fins at the first velocity.

Abstract: Systems and methods for cooling a datacenter are disclosed. In at least one embodiment, an alternate cooling loop with its own fluid source and a liquid-to-liquid heat exchanger is used to provide cooling for the at least one computing component alternatively from a secondary cooling loop that is associated with a primary cooling loop and a chilling facility.

Abstract: A cooling system for a datacenter is disclosed. A multiple mode cooling subsystem of the cooling system has a form factor for one or more racks of the datacenter, has two or more different cooling systems, and is adjustable to different cooling requirements of the datacenter within different cooling capacities offered by the two or more different cooling systems.

Abstract: One embodiment of a computer-implemented method for processing data within a fixed-function pipeline included in an execution unit includes receiving a first input from a first processing unit, wherein the first input corresponds to a first fixed-function; executing the first fixed-function on the first input to generate a first output, wherein the first fixed-function is executed on the first input prior to executing the first fixed-function on one or more inputs received from a plurality of processing cores that are processing a plurality of rays, and wherein each ray represents a path from a light source to at least one point within a three-dimensional (3D) environment; and transmitting the first output to the first processing unit for further processing.

Abstract: In various examples, systems and methods are disclosed relating to multi-domain generative adversarial networks with learned warp fields. Input data can be generated according to a noise function and provided as input to a generative machine-learning model. The generative machine-learning model can determine a plurality of output images each corresponding to one of a respective plurality of image domains. The generative machine-learning model can include at least one layer to generate a plurality of morph maps each corresponding to one of the respective plurality of image domains. The output images can be presented using a display device.

Abstract: In examples, a list of elements may be divided into spans and each span may be allocated a respective memory range for output based on a worst-case compression ratio of a compression algorithm that will be used to compress the span. Worker threads may output compressed versions of the spans to the memory ranges. To ensure placement constraints of a data structure will be satisfied, boundaries of the spans may be adjusted prior to compression. The size allocated to a span (e.g., each span) may be increased (or decreasing) to avoid padding blocks while allowing for the span's compressed data to use a block allocated to an adjacent span. Further aspects of the disclosure provide for compaction of the portions of compressed data in memory in order to free up space which may have been allocated to account for the memory gaps which may result from variable compression ratios.

Abstract: Recurrent blurring may be used to render frames of a virtual environment, where the radius of a filter for a pixel is based on a number of successfully accumulated frames that correspond to that pixel. To account for rejections of accumulated samples for the pixel, ray-traced samples from a lower resolution version of a ray-traced render may be used to increase the effective sample count for the pixel. Parallax may be used to control the accumulation speed along with an angle between a view vector that corresponds to the pixel. A magnitude of one or more dimensions of a filter applied to the pixel may be based on an angle of a view vector that corresponds to the pixel to cause reflections to elongate along an axis under glancing angles. The dimension(s) may be based on a direction of a reflected specular lobe associated with the pixel.

Abstract: Apparatuses, systems, and techniques to optimize device communications disclosed. In at least one embodiment, one or more neural networks are used to determine optimal power and frequency states for communication links between processing devices.

Abstract: One embodiment of a method for rendering one or more graphics images includes tracing one or more rays through a graphics scene; computing one or more surface normals associated with intersections of the one or more rays with one or more surfaces, where computing each surface normal includes: computing a plurality of intermediate surface normals associated with a plurality of adjacent voxels of a grid, and interpolating the plurality of intermediate surface normals; and rendering one or more graphics images based on the one or more surface normals.

Abstract: Pixel depth information is used to determine a weight to apply to neighboring pixels when using a sharpening filter. A difference between neighboring pixel depths is evaluated and pixels with pixel depths that exceed a threshold are given less weight than other pixels. A sharpening mask may be generated using adjusted pixel colors.

Abstract: Techniques are disclosed herein for interleaving textures. In the disclosed techniques, multiple textures that would otherwise be accessed separately are interleaved into a single, interleaved texture that can be used to access the multiple textures together. The interleaved texture can include alternating blocks from the multiple textures. The interleaved texture can be generated when the multiple textures are being loaded into memory. Further, the interleaved texture can be accessed using multiple texture headers that are associated with different textures in the interleaved texture. Each of texture headers includes a stride indicating the distance between two blocks from a same texture in the interleaved texture.

Abstract: Fabric Attached Memory (FAM) provides a pool of memory that can be accessed by one or more processors, such as a graphics processing unit(s) (GPU)(s), over a network fabric. In one instance, a technique is disclosed for using imperfect processors as memory controllers to allow memory, which is local to the imperfect processors, to be accessed by other processors as fabric attached memory. In another instance, memory address compaction is used within the fabric elements to fully utilize the available memory space.

Abstract: Systems and methods for cooling a datacenter are disclosed. In at least one embodiment, an alternate cooling loop with its own fluid source and a liquid-to-air heat exchanger is used to provide cooling for the at least one computing component alternatively from a secondary cooling loop that is associated with a primary cooling loop and a chilling facility.