dolfin team mailing list archive

[Anders Logg <logg@xxxxxxxxx>]

Garth N. Wells wrote:
> Anders Logg wrote:
> > Jake Ostien wrote:
> > > Garth N. Wells wrote:
> > > > Alessio Quaglino wrote:
> > > >  
> > > > > > > Alessio Quaglino wrote:
> > > > > > >       
> > > > > > > > I'd like to turn off the progress bars for the assembly, I'm trying
> > > > > > > > something like this:
> > > > > > > >
> > > > > > > > set("OutputDestination", "silent");
> > > > > > > > assembler.assemble(load, L, mesh);
> > > > > > > > set("OutputDestination", "terminal");
> > > > > > > >
> > > > > > > > but it doesn't seem to work. Do I forget something?
> > > > > > > >
> > > > > > > > Thanks,
> > > > > > > > Alessio Quaglino
> > > > > > > >           
> > > > > > > It should be "output destination" (not "OutputDestination").
> > > > > > >
> > > > > > > It should also work setting "debug level" to -1.
> > > > > > >
> > > > > > > Perhaps we should increase the debug level for progress bars to 
> > > > > > > 1. They
> > > > > > > are currently at 0 and will thus be printed on the default debug 
> > > > > > > level
> > > > > > > which is 0. Thoughts?
> > > > > > >
> > > > > > > /Anders
> > > > > > >         
> > > > > > I've tried:
> > > > > >
> > > > > > set("output destination", "silent");
> > > > > > assembler.assemble(load, L, mesh);
> > > > > > set("output destination", "terminal");
> > > > > >
> > > > > > but the progress bars are still printed out. I think it's a good 
> > > > > > idea to
> > > > > > move them to level 1.
> > > > > >
> > > > > > Alessio
> > > > > >       
> > > > > Apart from this, I've got another problem with the output. It seems 
> > > > > that
> > > > > for both uBlas matrices and vectors disp() prints only a short 
> > > > > block of
> > > > > each row of the matrix or a short block of the whole vector. The 
> > > > > output
> > > > > for a matrix looks like this (first and second rows):
> > > > >
> > > > > | (0, 0, 3.66e-01) (0, 1, -8.33e-03) (0, 11, -4.17e-03) (0, 12, 
> > > > > -5.69e-06)
> > > > > (0, 121, -2.67e-01) (0, 122, 5.18e-07) (0, 132, 6.70e-07) (0, 133,
> > > > > 4.98e-06) (0, 242, -5.99e-07) (0, 243, 1.02e-07) (0, 253, 2.08e-03) 
> > > > > (0,
> > > > > 254, -2.08e-03) (0, 363, 6.00e-07) (0, 36
> > > > >
> > > > > | (1, 0, -8.33e-03) (1, 1, 5.38e-01) (1, 2, -8.33e-03) (1, 12, 
> > > > > -6.25e-03)
> > > > > (1, 13, -2.21e-05) (1, 122, -4.00e-01) (1, 123, 6.08e-06) (1, 133,
> > > > > 8.36e-06) (1, 134, 5.75e-06) (1, 242, 2.08e-03) (1, 243, -2.08e-03) 
> > > > > (1,
> > > > > 244, -2.72e-07) (1, 254, 2.09e-03) (1, 255
> > > > >
> > > > >     
> > > > This problem was introduced by the new log system. It cuts things 
> > > > off after 256 character (which is set by DOFLIN_LINELENGTH). Anyone 
> > > > know how to fix this?
> > > >
> > > > Garth
> > > >
> > > >   
> > > I couldn't tell if this ever got discussed again, but if there is a 
> > > fix or a work around, I'd like to hear it, too.
> > >
> > > Jake
> >
> > There's no simple fix. The log system needs to have a buffer of some 
> > size when calling vsnprintf(). (Used for parsing variable number of 
> > arguments for dolfin::message().)
>
> We could print matrix and vector components one-by-one. Then it should 
> be ok.

Yes, that sounds good.

/Anders

> Garth
>
>
> > If the matrix is big, just print it to file and then look at it:
> >
> > File file("matrix.m");
> > file << A;
> >
> > or
> >
> > File file("matrix.xml");
> > file << A;
> >
> > /Anders