hipexpm - Matrix Exponential Approximation using HIP

Version: 2.0.0

Copyright: Maximilian Behr

License: The software is licensed under under MIT. See LICENSE for details.

hipexpm is a HIP library for the numerical approximation of the matrix exponential $e^A$.

hipexpm support single and double precision as well as real and complex matrices.

Functions	Data
hipexpms_bufferSize, hipexpms	real, single precision matrix
hipexpmd_bufferSize, hipexpmd	real, double precision matrix
hipexpmc_bufferSize, hipexpmc	complex, single precision matrix
hipexpmz_bufferSize, hipexpmz	complex, double precision matrix

Available functions:

int hipexpms_bufferSize(const int n, size_t *d_bufferSize, size_t *h_bufferSize);
int hipexpms(const int n, const float *d_A, const int ldA, void *d_buffer, void *h_buffer, float *d_expmA, const int ldexpmA);

int hipexpmd_bufferSize(const int n, size_t *d_bufferSize, size_t *h_bufferSize);
int hipexpmd(const int n, const double *d_A, const int ldA, void *d_buffer, void *h_buffer, double *d_expmA, const int ldexpmA);

int hipexpmc_bufferSize(const int n, size_t *d_bufferSize, size_t *h_bufferSize);
int hipexpmc(const int n, const hipComplex *d_A, const int ldA, void *d_buffer, void *h_buffer, hipComplex *d_expmA, const int ldexpmA);

int hipexpmz_bufferSize(const int n, size_t *d_bufferSize, size_t *h_bufferSize);
int hipexpmz(const int n, const hipDoubleComplex *d_A, const int ldA, void *d_buffer, void *h_buffer, hipDoubleComplex *d_expmA, const int ldexpmA);

Algorithm

hipexpm implements the scaling and squaring method for the matrix exponential approximation.

The Scaling and Squaring Method for the Matrix Exponential Revisited Nicholas J. Higham SIAM Journal on Matrix Analysis and Applications 2005 26:4, 1179-1193

Installation

Prerequisites:

• CMake >= 3.23
• ROCm >= 6.0.2

  mkdir build && cd build
  cmake ..
  make
  make install

Usage and Examples

We provide examples for all supported matrix formats:

File	Data
example_hipexpms.hip	real, single precision matrix
example_hipexpmd.hip	real, double precision matrix
example_hipexpmc.hip	complex, single precision matrix
example_hipexpmz.hip	complex, double precision matrix