Linear system solver in D?
ao at pathlink.com
Tue Feb 19 18:05:52 PST 2008
Reply to Bill,
> BCS wrote:
>> thanks, both or you, I'll look at those. Performance isn't that big
>> an issue as I'm only looking at about 15-30 equations and a few
>> minutes run time would be ok, but I'm going to have to run ~1500
>> passes through it.
> Are the equations the same each time? I.e. do the a_m1 coeffs stay
> the same for all 1500 passes? or do they change each time?
> If they stay the same then you can factorize A once and do the much
> faster back-substitution 1500 times.
no such luck. This will be part of a newton Newton–Raphson non-linear solver
for doing a numerical approximation of a system of non-linear differential
equations. (a.k.a. bad, worse, nasty)
> If you don't care about performance just search the web for some
> C/C++/C#/Java code for Gaussian Elimination with Partial (or better,
> Full) pivoting. It should be pretty straightforward to port.
performance takes a back seat to complexity but is still an issue. OTOH generating
bindings for canned C code is less complex than porting it.
> Then you won't have to worry about dependencies and building
> BLAS/LAPACK etc.
building and what not isn't a big issue.
> Actually you didn't say, but are m and n not the same?
On the way out the door after posting that I relized that I put in NxM rather
than NxN. oops.
More information about the Digitalmars-d-learn