Nvidia patents method to schedule tasks on AI chips for better efficiency
Nvidia describes a compiler technique that organizes how tasks are executed on its specialized processors. It creates a map of how operations depend on each other and then rearranges them to fit the hardware's limits, ensuring the chip works without bottlenecks.
In the filing’s own words
Processors, systems, and techniques to generate schedules of operations of tile-based programs in accordance with constants on usage of functional units of an accelerator. In at least one embodiment a compiler is to generate a dependency graph indicative of operations of a tile program and determine functional units of an accelerator that would be used to perform operations in a first order indicated by the dependency graph. The compiler is further to determine a second order in which to perform the operations of the tile program based, at least in part, on constraints on usage of the determined functional units, and generate code to perform the tile program based, at least in part, on the second order.
What exactly they claim (claim 1)
1. One or more processors, comprising: circuitry to cause a compiler to: generate a dependency graph indicative of operations of a tile program; determine functional units of an accelerator that would be used to perform operations in a first order indicated by the dependency graph; determine a second order in which to perform the operations of the tile program based, at least in part, on constraints on usage of the determined functional units; and generate code to perform the tile program based, at least in part, on the second order.
Facts
- Company
- Nvidia (NVIDIA Corporation)
- Official title
- SCHEDULING TILE PROGRAM OPERATIONS
- Publication number
- US20260277554A1
- Application number
- 19530267
- Filed
- Feb 4, 2026
- Published
- Sep 17, 2026
- Inventors
- 5
- Claims
- 20
Classified under
G06F 8/30 Physics & computingG06F 8/433 Physics & computingG06F 9/5038 Physics & computingG06F 9/5044 Physics & computingG06F 2209/506 Physics & computingMore from Nvidia in Chips & hardware
- Nvidia patents method to skip zero data in neural network math Oct 13, 2026
- Nvidia patents hardware to speed up motion blur in ray tracing Oct 13, 2026
- Nvidia patents method to sync data across split processor chips Sep 17, 2026
- Nvidia patents chip math trick to speed up AI calculations Oct 8, 2026
- Nvidia patents data center cooling that uses cryogenic fluids Oct 15, 2026
Summaries are simplified from the filing’s own abstract and first claim. A patent filing is not a product announcement.
Patent data and drawings: U.S. Patent and Trademark Office (USPTO). Nebula is not affiliated with or endorsed by the USPTO.