On Wed, 26 Feb, 2014 at 12:11 PM, Martin Sandve Alnæs
<[email protected]> wrote:
Fix is in next.
Seems to have done the trick - the tests are no longer hanging in
parallel for me.
Garth
Martin
On 25 February 2014 17:52, Martin Sandve Alnæs <[email protected]>
wrote:
Thanks! Helper me to quickly narrow it down, the Domain label
affects the signature (it shouldn't):
from ufl import *
for i in range(2):
j = i # This is not
#j = 0 # This is ok
domain = Domain(triangle, label="domain%d"%j)
V = FiniteElement("CG", domain, 1)
a = Coefficient(V)**2*dx
print a.compute_form_data().signature[:8]
I'll fix it soon but not today.
Martin
On 25 February 2014 18:35, Jan Blechta <[email protected]>
wrote:
On Tue, 25 Feb 2014 15:48:37 +0000
Martin Sandve Alnæs <[email protected]> wrote:
> Probably something with signature computation in ufl. I'll take a
> look.
This triggers JITing twice
from dolfin import *
for i in range(2):
mesh = UnitSquareMesh(3, 3)
elem = FiniteElement('CG', mesh, 1)
V = FunctionSpaceBase(mesh, elem)
Is it correct?
Jan
>
> Martin
> 25. feb. 2014 15:20 skrev "Garth N. Wells" <[email protected]>
følgende:
>
> > I'm finding that of late JIT compilation is flaky in parallel,
i.e.
> > it hangs. What recent changes could be causing this? I typically
> > see the problems in test/unit/function/python/, and problems are
> > are not consistently reproducible. Moreover, it seems that the
JIT
> > caching is not working properly in parallel. When running a
script
> > for a second time, JIT compilation is sometimes triggered.
> >
> > Garth
> >
> > _______________________________________________
> > fenics mailing list
> > [email protected]
> > http://fenicsproject.org/mailman/listinfo/fenics
> >
_______________________________________________
fenics mailing list
[email protected]
http://fenicsproject.org/mailman/listinfo/fenics
