BUG scipy.linalg.lapack: potrf/ptroi interpret their 'lower' argument differently #2691
Labels
defect
A clear bug or issue that prevents SciPy from being installed or used as expected
scipy.linalg
Milestone
The second argument to these functions should define if the cholesky decomposition works on the lower/upper part of the matrix, however something doesn't work out.
This is a simple test to try it out:
The normal inverse of this would be:
Using dprotf/dproti:
Clearly, this is not the same as the answer given above. In fact, the answer is only exact when the 2nd argument to dpotrf and dpotri are different:
I tried this with SciPy 0.11 and 0.12, the underlying BLAS library was OpenBLAS.
The text was updated successfully, but these errors were encountered: