search

haasad / PyPardiso

Python interface to the Intel MKL Pardiso library to solve large sparse linear systems of equations
BSD 3-Clause "New" or "Revised" License
133 stars 20 forks source link

Speed decrease with zero-based indexing #6

Closed haasad closed 7 years ago

haasad commented 7 years ago

Passing A.indices and A.indptr with zero-based indexing to pardiso and setting iparm 35 to 1 makes the code look cleaner, but slows down the factorization phase a lot (~factor 5 for brightway2 technosphere matrix). Performance is similar to scipy for the first call to spsolve.

Converting from 0 to 1-based indexing should be handled in python.

haasad commented 7 years ago

fixed in 6884c5a5fb2bbb778773499c2c3ff2e1dd2e53ad. new version v0.2.1 uploaded to anaconda cloud