Back to results
US 2026/0267641 A1Published application
Accelerating linear algebra kernels for any processor architecture
AssigneeNVIDIA Corporation
InventorsVenmugil Elango, Norm Rubin, Mahesh Ravishankar, Vinod K. Grover
Life of this patent
Estimating the term…
Abstract
Systems and methods for obtaining a set of instructions for executing a computer program and generating executable code for the computer program based, at least in part, on scheduling operations associated with the executable code according to a polyhedral representation of a directed acyclic graph. The set of instructions may be represented as a domain-specific language. The executable code may be executable code for a specific processor architecture.
Claims2
-20. (canceled)
A computer-implemented method, comprising:
- performing an application programming interface (API) to:
- construct a polyhedral representation of a computer program based, at least in part, on an input file and a configuration file for the computer program, wherein the input file includes one or more matrix declaration statements specifying one or more input matrix dimension sizes for one or more subsequently specified matrix operations, and the configuration file specifies one or more parameters for one or more optimizations for generating executable code for the computer program;
- determine, based, at least in part, on the polyhedral representation and the one or more parameters for the one or more optimizations, an affine schedule for the computer program;
- and
- generate executable code for the computer program based, at least in part, on the affine schedule.
2 more elements
Dependencies are inferred from the claim text.