dolfin team mailing list archive

[Johan Hake <johan.hake@xxxxxxxxx>]

On Monday August 30 2010 12:06:30 Anders Logg wrote:
> On Mon, Aug 30, 2010 at 12:02:14PM -0700, Johan Hake wrote:
> > On Monday August 30 2010 11:53:54 Anders Logg wrote:
> > > Is timing broken? It seems to be for me.
> > > 
> > > Can anyone try the following:
> > > 
> > > python -c "from dolfin import *; import time; tic(); time.sleep(1);
> > > print toc()"
> > > 
> > > It should print out something like 1.0, but I get 0.0.
> > 
> > I get 0.0.
> > 
> > > std::clock() is giving very strange values.
> > 
> > Do you suggest that it is broken in std::clock and not in our wrapped
> > version?
> 
> Yes, since we didn't change anything. (At least I don't remember
> changing anything.)
> 
> When I print the values I get from std::clock(), I get values like
> 
> 66400000
> 
> The precision seems to be very low.

Something is wrong, but I cannot figure out what. My DOLFIN installation is 
kindoff broken due to some CMake trouble* so I am not up to speed to debug 
this. But can it be broken just for the Python interface. I mean something 
happens to std::clock when Python is imported?

* My petsc and slepc is not found due to some broken tests (probably due to my 
empty PETSC_ARCH), and libpytrilinos.so is included in the linker to dolfin, 
and my Python library is not... Why are we linking to PyTrilinos for the 
DOLFIN library?

Johan