dolfin team mailing list archive

[Anders Logg <logg@xxxxxxxxx>]

On Mon, Nov 07, 2011 at 03:06:38PM -0800, Johan Hake wrote:
> Hello!
>  
> Are any of these parameters every used in JIT compilation?
> (From GlobalParameters.h)
>  
> p.add("optimize_form", false);
> p.add("optimize_use_dofmap_cache", false);
> p.add("optimize_use_tensor_cache", false);
> p.add("optimize", false);
>  
> Looks to me that these are legacy parameters. We now use:
>  
> parameters["form_compiler"].foo
>  
> to set the optimization.
>  
> Johan

Looks like only the last two are used and then in the same way.
>From site-packages/dolfin/fem/assembling.py:

  # Check if we should use the cache
  use_cache = cpp.parameters["optimize_use_tensor_cache"] or \
              cpp.parameters["optimize"]

That is a bit silly to say the least...

I suggest we remove those parameters and add a boolean parameter
use_tensor_cache to assemble() in Python.

--
Anders