dolfin team mailing list archive

[Johan Hake <668862@xxxxxxxxxxxxxxxxxx>]

> For something else, I've just added a
> 
>    dolfin/graph/graph_types.h
> 
> which has typedefs for graph-like objects, We could use that.
> 
> Have you done any Set versus unordered_set performance testing for
> problems with 'Real'?

That was what I tried to say above ;)

I more or less removed the bottle neck wrt build sparsity pattern with Real 
spaces, when I used unordered_set. I am not sure how it affects performance 
for other situations. I am working on a code with a MixedSpace with lots of 
Real sub spaces. Building the sparsity pattern previously took ~18 s and after 
the change it was down to < 1s.

-- 
Sparsity pattern computation dead slow with 'real' spaces
https://bugs.launchpad.net/bugs/668862
You received this bug notification because you are a member of DOLFIN
Team, which is subscribed to DOLFIN.

Status in DOLFIN: Confirmed

Bug description:
The computation of the sparsity pattern is dead slow when real spaces are used, i.e.,

    R = FunctionSpace(mesh, 'R', 0)

To see the problem, compare the neumann-poisson and poisson demos with fine meshes (256x256)