shuren/fdfd_tools
1
0
Fork 0
forked from jan/fdfd_tools
Code Issues Pull Requests Releases Wiki Activity

add possibility to use csr opengl solver

Browse Source
This commit is contained in:
jan 2016-07-04 23:53:54 -07:00
parent eb4d9be6cf
commit 56df805e24
1 changed files with 17 additions and 11 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

22
examples/test.py
View File

@ -10,7 +10,7 @@ import gridlock
from matplotlib import pyplot from matplotlib import pyplot
from opencl_fdfd import cg_solver from opencl_fdfd import cg_solver, csr
__author__ = 'Jan Petykiewicz' __author__ = 'Jan Petykiewicz'
@ -180,7 +180,7 @@ def test1():
H_overlap = waveguide_mode.compute_overlap_e(**wg_args, **wg_results) H_overlap = waveguide_mode.compute_overlap_e(**wg_args, **wg_results)
pecg = gridlock.Grid(edge_coords, initial=0.0, num_grids=3) pecg = gridlock.Grid(edge_coords, initial=0.0, num_grids=3)
pecg.draw_cuboid(center=[700, 0, 0], dimensions=[80, 1e8, 1e8], eps=1) # pecg.draw_cuboid(center=[700, 0, 0], dimensions=[80, 1e8, 1e8], eps=1)
# pecg.visualize_isosurface() # pecg.visualize_isosurface()
pmcg = gridlock.Grid(edge_coords, initial=0.0, num_grids=3) pmcg = gridlock.Grid(edge_coords, initial=0.0, num_grids=3)
@ -190,15 +190,21 @@ def test1():
''' '''
Solve! Solve!
''' '''
A = fdfd_tools.operators.e_full(omega, dxes, sim_args = {
epsilon=vec(grid.grids), 'omega': omega,
pec=vec(pecg.grids), 'dxes': dxes,
pmc=vec(pmcg.grids)).tocsr() 'epsilon': vec(grid.grids),
'pec': vec(pecg.grids),
'pmc': vec(pmcg.grids),
}
b = -1j * omega * vec(J) b = -1j * omega * vec(J)
A = fdfd_tools.operators.e_full(**sim_args).tocsr()
# x = solve_A(A, b) # x = solve_A(A, b)
x = cg_solver(omega, dxes, J=vec(J), epsilon=vec(grid.grids), # x = csr.cg_solver(J=vec(J), **sim_args)
pec=vec(pecg.grids), pmc=vec(pmcg.grids)) x = cg_solver(J=vec(J), **sim_args)
E = unvec(x, grid.shape) E = unvec(x, grid.shape)
print('Norm of the residual is ', numpy.linalg.norm(A @ x - b)) print('Norm of the residual is ', numpy.linalg.norm(A @ x - b))