Multi-tile graphics processor rendering
AssigneeIntel Corporation
InventorsPrasoonkumar Surti, Arthur Hunter, Jr., Kamal Sinha, Scott Janus, Brent Insko, Vasanth Ranganathan, Lakshminarayanan Striramassarma
Life of this patent
Estimating the term…
Abstract
Embodiments are generally directed to multi-tile graphics processor rendering. An embodiment of an apparatus includes a memory for storage of data; and one or more processors including a graphics processing unit (GPU) to process data, wherein the GPU includes a plurality of GPU tiles, wherein, upon geometric data being assigned to each of a plurality of screen tiles, the apparatus is to transfer the geometric data to the plurality of GPU tiles.
Claims14
An apparatus comprising:
- a memory for storage of data, the data including geometric data for graphics processing, the memory including a distributed data structure;
- and
- one or more processors including a graphics processing unit (GPU) to process data, wherein the GPU includes a plurality of GPU tiles on a substrate, each GPU tile being a separate chiplet, and each GPU tile having a respective tile-based storage;
- wherein, upon a respective set of geometric data being assigned to each of a plurality of screen tiles, the apparatus is to transfer the set of geometric data assigned to each of the plurality of screen tiles from the distributed data structure to a tile-based storage of a respective GPU tile of the plurality of GPU tiles, including the apparatus to transfer a first set of geometric data assigned to a first screen tile from the distributed data structure to a tile-based storage of a first GPU tile, and
- wherein each set of geometric data is to be processed locally in the respective GPU tile of the plurality of GPU tiles to which the set of geometric data was transferred, including processing the set of first geometric data locally in the first GPU tile.
One or more non-transitory computer-readable storage mediums having stored thereon executable computer program instructions that, when executed by one or more processors, cause the one or more processors to perform operations comprising:
- assigning a set of geometric data for graphics processing by a graphics processing unit (GPU) of a computing system to each of a plurality of screen tiles, the GPU including a plurality of GPU tiles on a substrate, each GPU tile being a separate chiplet, and each GPU tile having a respective tile-based storage;
- transferring each set of geometric data that is assigned to each screen tile of the plurality of screen tiles from a distributed data structure of a memory of the computing system to a tile-based storage of a respective GPU tile of the plurality of GPU tiles of the GPU, including transferring a first set of geometric data assigned to a first screen tile from the distributed data structure to a tile-based storage of a first GPU tile;
- and
- processing each set of geometric data locally in the respective GPU tile of the plurality of GPU tiles to which the set of geometric data was transferred, including processing the set of first geometric data locally in the first GPU tile.
A method comprising:
- assigning a set of geometric data for graphics processing by a graphics processing unit (GPU) of a computing system to each of a plurality of screen tiles, the GPU including a plurality of GPU tiles on a substrate, each GPU tile being a separate chiplet, and each GPU tile having a respective tile-based storage;
- transferring each set of geometric data from the that is assigned to each screen tile of the plurality of screen tiles from a distributed data structure of a memory of the computing system to a tile-based storage of a respective GPU tile of the plurality of GPU tiles of the GPU including transferring a first set of geometric data assigned to a first screen tile from the distributed data structure to a tile-based storage of a first GPU tile;
- and
- processing each set of geometric data locally in the respective GPU tile of the plurality of GPU tiles to which the set of geometric data was transferred, including processing the set of first geometric data locally in the first GPU tile.
Dependencies are inferred from the claim text.