Applied-GeoSolutions / multitemporal

Efficient, chainable time series processing of raster stacks.
GNU General Public License v2.0
4 stars 3 forks source link

scope lapack-based module #31

Closed ircwaves closed 6 years ago

ircwaves commented 6 years ago

@naterubin scope out calling lapack from module (maybe from cython module)

bziniti commented 6 years ago

it's relatively easy to deal with: https://docs.scipy.org/doc/numpy-1.14.0/reference/generated/numpy.isfortran.html it has to do with the way that fortran arrays need to be aligned you can make sure the numpy arrays are correct by creating them like this: np.array([1,2,3], order='FORTRAN') and you can check that arrays passed into a function are aligned correctly by checking ary.flags.f_contiguous

bziniti commented 6 years ago

recommendation is to use the functions in scipy.linalg as these are using blas and lapack under the hood

bziniti commented 6 years ago

@naterubin anything else I missed in the documentation of this scoping? If not, feel free to close when appropriate.

naterubin commented 6 years ago

Yeah, I think that covers it. I'll close this and we can create new issues for more specific questions/problems.