Expand description
Matrix functions: exponential, logarithm, and square root.
Provides GPU-accelerated matrix function computations via PTX kernel generation:
- Matrix Exponential (
expm): Scaling and squaring with Padé approximation. Computese^Afor a square matrix A. - Matrix Logarithm (
logm): Inverse scaling and squaring with Padé approximation. Computeslog(A)for a matrix with eigenvalues in the right half-plane. - Matrix Square Root (
sqrtm): Denman–Beavers iteration. ComputesA^{1/2}such thatsqrtm(A) * sqrtm(A) = A.
Each plan struct generates self-contained PTX kernels using
KernelBuilder/BodyBuilder from the oxicuda-ptx crate.
Structs§
- Matrix
ExpConfig - Configuration for matrix exponential computation.
- Matrix
ExpPlan - Execution plan for the matrix exponential.
- Matrix
LogConfig - Configuration for matrix logarithm computation.
- Matrix
LogPlan - Execution plan for the matrix logarithm.
- Matrix
Sqrt Config - Configuration for matrix square root computation.
- Matrix
Sqrt Plan - Execution plan for the matrix square root using Denman–Beavers iteration.