Vitis Solver Library - Vitis Solver Library - 2026.1 English

AMD Vitis™ Solver Library (SolverLib) provides a collection of matrix decomposition operations, linear solvers, and eigenvalue solvers on PL and AI Engine.

The SolverLib contains:

• PL Solver library.
• AI Engine Solver Library.

PL Solver library

Currently, the Vitis PL Solver library includes the following operations for dense matrix:

• Matrix decomposition

  • Cholesky decomposition for symmetric positive definite matrix
  • LU decomposition without pivoting and with partial pivoting
  • QR decomposition for general matrix
  • SVD decomposition (singular value decomposition) for symmetric matrix and non-symmetric matrix (Jacobi method)

• Linear solver

  • Tridiagonal linear solver (Parallel cyclic reduction method)
  • Linear solver for triangular matrix
  • Linear solver for symmetric and non-symmetric matrix
  • Matrix inverse for symmetric and non-symmetric matrix

• Eigenvalue solver

  • Jacobi eigenvalue solver for symmetric matrix

AI Engine Solver Library

Currently, the Vitis AIE Solver Library provides the following operations on AI Engine. The AMD Vitis AI Engine Solver library encapsulates several solver algorithms, optimized to take full advantage of the processing power of AMD Versal™ Adaptive SoC devices, which contain an array of AI Engines.

• Matrix decomposition

  • Cholesky decomposition
  • QR decomposition (Gram-Schmidt method)
  • QR decomposition (House-Holder method)
  • Singular Value Decomposition (one-sided Jacobi with parallel-pairs schedule)

• Linear system solving

  • Substitution (forward and backward)