dolfin team mailing list archive

[Anders Logg <logg@xxxxxxxxx>]

On Wed, Oct 15, 2008 at 02:09:45PM -0400, Shawn Walker wrote:
> One of the other ODE demos does not work (Ilmar pointed this out).  It is 
> the 'Harmonic" demo.  I get the following output:
> 
> walker@box227 ~/FENICS/dolfin/demo/ode/harmonic/cpp
> $ ./demo
> cG(1): e = 1.047e-04
> cG(2): e = 1.745e-10
> cG(3): e = 4.663e-15
> cG(4): e = 4.630e-15
> cG(5): e = 3.553e-15
> 
> dG(0): e = 6.088e-02
> dG(1): e = 1.745e-07
> Segmentation fault

This should be fixed now.

> I looked at main.cpp and it seems straightforward.  I thought maybe there 
> was a problem with ODE.h/ODE.cpp, and I noticed that the 'real' keyword 
> was still used there.  Could this be causing a problem?

Yes, "real" is still used in the ODE solvers and that's the only place
it can/should be used. The ODE solvers now work with either "double"
or GMP high precision types (mpf_class). The purpose is to solve ODEs
with very high precision (like hundreds of digits).

-- 
Anders

Attachment: signature.asc
Description: Digital signature