[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[Help-glpk] performance of exact LP solver
From: |
Axel Simon |
Subject: |
[Help-glpk] performance of exact LP solver |
Date: |
Thu, 20 Mar 2008 11:19:03 +0100 |
Hi,
I need to use the lpx_exact function and find that the performance on
repeatedly running a 100x10 variable problem is about 70x slower than
using doubles. It turns out that the program spends 40% of it's time
calculating the gcd when canonicalizing fractions. Is there a way to use
integers (rather than rationals) to represent the coefficients of a row
rather than rationals? I would assume that this is possible since a row
can always be scaled by the smallest denominator. Normalizing a row
could be much cheaper since the gcd calculation can be aborted early
whenever the gcd drops to 1.
Am I missing something?
Cheers,
Axel.
- [Help-glpk] performance of exact LP solver,
Axel Simon <=