Abstract: Apparatuses, systems, and techniques are presented to generate images. In at least one embodiment, at least a first optical flow network (OFN) and at least a first reconstruction network (RN) can be used to generate one or more images based, at least in part, upon the OFN and the RN using a shared loss function.

Abstract: Methods and systems may provide for identifying a plurality of subject commands that reference a common screen location and access a read/write resource, and serializing the plurality of subject commands according to a predefined order. Additionally, execution of the plurality of subject commands may be deferred until one or more additional commands referencing the common screen location are executed. In one example, the plurality of subject commands are serialized in response to a serialization command.

Abstract: In some embodiments, a given frame or picture may have different shading rates. In one embodiment in some areas of the frame or picture the shading rate may be less than once per pixel and in other places it may be once per pixel. An algorithm may be used to determine how the shading rate changes across the frame.

Abstract: Representing a transparent object as a summation of substantially zero step functions of a visibility curve for the object. An array may be used to store nodes to represent the visibility function. The size of the array may be limited to be storable within a memory of an on-chip graphics processing unit.

Abstract: Shadow map partitions may be automatically placed based on the location or concentration of sample data depth in eye space. An initial positioning for the partitions may be determined based on user specified budgets for number of partitions, computation time, or memory utilization, in some embodiments. The initial positioning may be refined using a clustering algorithm in some cases.

Abstract: In some embodiments, a given frame or picture may have different shading rates. In one embodiment in some areas of the frame or picture the shading rate may be less than once per pixel and in other places it may be once per pixel. An algorithm may be used to determine how the shading rate changes across the frame.

Abstract: Methods and systems may provide for obtaining a plurality of visibility samples for an image at a sample resolution, wherein a first subset of the plurality of visibility samples has corresponding color samples. One or more of the color samples may be replicated from the first subset to a second subset of the plurality of visibility samples, and the visibility samples and the color samples may be rendered at a native display resolution. Additionally, the sample resolution may be greater than a pixel resolution of the image and greater than or equal to the native display resolution. In one example, the image includes a plurality of pixels and the plurality of visibility samples are obtained in accordance with an axis-aligned grid within each pixel of the image.

Abstract: In some embodiments, tile lists may be avoided by storing the geometry of a scene in a bounding volume hierarchy (BVH). For each tile, the bounding volume hierarchy is traversed. The traversals continued only into children nodes that overlap with the frustum on the tile. By relaxing the ordering constraint of rendering primitives, the BVH is traversed such that nodes that are closer to the viewer are traversed first, increasing the occlusion culling efficiency in some embodiments. Rendering the full scene between the central processing cores and the graphics processor may be done through a shared memory in some embodiments.

Abstract: A dynamic volumetric medium, such as hair, fog, or smoke, may be represented, for purposes of shadow mapping, by transmittance versus depth data for that medium. In one embodiment, the representation may take the form of a plot of transmittance versus depth, with nodes where the transmittance changes non-live linearly with respect of depth into the medium. The number of nodes in the representation may be reduced to reduce memory footprint and to enable the storage of the representation on the same chip doing the shadow mapping. In some embodiments, the number of nodes may be reduced, one node at a time, by removing the node whose underlying trapezoid has the least area of all the remaining nodes.

Abstract: Methods and systems may provide for identifying a plurality of subject commands that reference a common screen location and access a read/write resource, and serializing the plurality of subject commands according to a predefined order. Additionally, execution of the plurality of subject commands may be deferred until one or more additional commands referencing the common screen location are executed. In one example, the plurality of subject commands are serialized in response to a serialization command.

Abstract: The problem of generating high quality images with a rendering pipeline based on decoupled sampling may be addressed by generating non-extrapolated shading locations and by determining improved texture filtering footprints. This may be accomplished by performing shading at the center of a bounding box that bounds mapped shading samples.

Abstract: In some embodiments, tile lists may be avoided by storing the geometry of a scene in a bounding volume hierarchy (BVH). For each tile, the bounding volume hierarchy is traversed. The traversals continued only into children nodes that overlap with the frustum on the tile. By relaxing the ordering constraint of rendering primitives, the BVH is traversed such that nodes that are closer to the viewer are traversed first, increasing the occlusion culling efficiency in some embodiments. Rendering the full scene between the central processing cores and the graphics processor may be done through a shared memory in some embodiments.

Abstract: Representing a transparent object as a summation of substantially zero step functions of a visibility curve for the object. An array may be used to store nodes to represent the visibility function. The size of the array may be limited to be storable within a memory of an on-chip graphics processing unit.

Abstract: A dynamic volumetric medium, such as hair, fog, or smoke, may be represented, for purposes of shadow mapping, by transmittance versus depth data for that medium. In one embodiment, the representation may take the form of a plot of transmittance versus depth, with nodes where the transmittance changes non-live linearly with respect of depth into the medium. The number of nodes in the representation may be reduced to reduce memory footprint and to enable the storage of the representation on the same chip doing the shadow mapping. In some embodiments, the number of nodes may be reduced, one node at a time, by removing the node whose underlying trapezoid has the least area of all the remaining nodes.

Abstract: Shadow map partitions may be automatically placed based on the location or concentration of sample data depth in eye space. An initial positioning for the partitions may be determined based on user specified budgets for number of partitions, computation time, or memory utilization, in some embodiments. The initial positioning may be refined using a clustering algorithm in some cases.

Abstract: Computer operations are partitioned for execution by a processor. A plurality of operations to be partitioned is received. The respective priorities for each of the plurality of operations is determined. Ready operations are identified, where a ready operation is an operation that does not have any predecessor operations that have not yet been added to a partition. A determination is made whether adding the ready operations with the highest priorities to the current partition would violate a hardware resource constraint. Based on this determination, the ready operations are added to the partition. The operations added to the partition are provided to a processor for execution.

Abstract: The described embodiments of the present invention include a method and system for partitioning and partitioning operations. The operations are first prioritized, then partitioned into one or more partitions.