solv_CUDA_sparse#

MCEq.solvers.solv_CUDA_sparse(nsteps, dX, rho_inv, context, phi, grid_idcs)[source]#

NVIDIA CUDA cuSPARSE implementation of forward-euler integration.

Function requires a working accelerate installation.

Parameters:

nsteps (int) – number of integration steps
dX (numpy.array() [nsteps]) – vector of step-sizes \(\Delta X_i\) in g/cm**2
rho_inv (numpy.array() [nsteps]) – vector of density values \(\frac{1}{\rho(X_i)}\)
int_m (numpy.array()) – interaction matrix (1) in dense or sparse representation
dec_m (numpy.array()) – decay matrix (2) in dense or sparse representation
phi (numpy.array()) – initial state vector \(\Phi(X_0)\)
mu_loss_handler (object) – object of type SemiLagrangianEnergyLosses

Returns:

state vector \(\Phi(X_{nsteps})\) after integration

Return type:

numpy.array()