Hallo Andreas,
thanks for your fast respond. First of all I still receive
unsatisfactionary results, no matter if I watch them with paraview or
visit(axial velocity doesn't become null at the noslip boundary). What
about the warnings I get(I sent to you last time)? Are they supposed to
appear by running the square.py or naca.py example?
Secondly, trying to run hedge with cuda by writing out silo, I get an
error:
Traceback (most recent call last):
File "square-neu.py", line 304, in <module>
main()
File "square-neu.py", line 166, in main
tune_for=op.op_template())
File "/home/elsalawi/src/hedge/hedge/backends/__init__.py", line 120, in
make_discretization
return self.discr_class(mesh_data, *args, **kwargs)
File "/home/elsalawi/src/hedge/hedge/backends/cuda/__init__.py", line
489, in __init__
overall_plans = list(generate_overall_plans())
File "/home/elsalawi/src/hedge/hedge/backends/cuda/__init__.py", line
456, in generate_overall_plans
flux_plan, flux_time = fluxgather.make_plan(self, given,
my_tune_for)
File "/home/elsalawi/src/hedge/hedge/backends/cuda/fluxgather.py", line
311, in make_plan
debug_flags=discr.debug)
File "/home/elsalawi/src/hedge/hedge/backends/cuda/plan.py", line 58, in
optimize_plan
plans = [p for p in plan_generator() if p.invalid_reason() is None]
File "/home/elsalawi/src/hedge/hedge/backends/cuda/fluxgather.py", line
287, in generate_valid_plans
mbs_per_block*given.microblock.elements))
File "/home/elsalawi/src/hedge/hedge/backends/cuda/__init__.py", line
297, in _get_partition_data
max_block_size)
File "/home/elsalawi/src/hedge/hedge/backends/cuda/__init__.py", line
211, in make_gpu_partition_metis
adjgraph, vweights=[1000]*len(adjgraph))
File
"/home/elsalawi/src/env/lib/python2.6/site-packages/PyMetis-0.91.1-py2.6-linux-x86_64.egg/pymetis/__init__.py",
line 30, in part_graph
from pymetis._internal import part_graph
ImportError:
/home/elsalawi/src/env/lib/python2.6/site-packages/PyMetis-0.91.1-py2.6-linux-x86_64.egg/pymetis/_internal.so:
undefined symbol: regerrorA
-------------------------------------------------------------------
PyCUDA ERROR: The context stack was not empty upon module cleanup.
-------------------------------------------------------------------
A context was still active when the context stack was being
cleaned up. At this point in our execution, CUDA may already
have been deinitialized, so there is no way we can finish
cleanly. The program will be aborted now.
Use Context.pop() to avoid this problem.
-------------------------------------------------------------------
Aborted
And my .aksetup-default.py looks like:
#BOOST_INC_DIR = ['${HOME}/pool/include/']
BOOST_INC_DIR = ['/usr/include']
#BOOST_LIB_DIR = ['${HOME}/pool/lib']
BOOST_LIB_DIR = ['/usr/lib']
BOOST_BINDINGS_INC_DIR = ['${HOME}/pool/include/boost-numeric-bindings/']
BOOST_COMPILER = 'gcc44'
# WITH_SPARSE_WRAPPERS = True
BOOST_PYTHON_LIBNAME=["boost_python"]
BOOST_THREAD_LIBNAME=["boost_thread"]
SILO_INC_DIR = ['${HOME}/pool/include']
SILO_LIB_DIR = ['${HOME}/pool/lib']
If I change BOOST_INC_DIR and type in /usr/include/boost instead, I receive
the same error message. This problem is already posted in the pycuda
mailing list, though it couldn't help me out solve my problem. Maybe you
have an idea? Is there any trick to compile silo? I followed the
instructions introduced on scicomp. If you need any more information, just
give me a sign.
Regards,
Islam
>
> -------- Original-Nachricht --------
> Datum: Wed, 23 Feb 2011 10:08:07 -0500
> Von: Andreas Kloeckner <lists(a)informa.tiker.net>
> An: Islam El Salawi <islamelsalawi(a)gmx.de>, hedge(a)tiker.net
> Betreff: Re: [Hedge] square, naca results
>
> Hi Islam,
>
> I suspect a visualization issue. Can you check the normal of the
> vertical component of velocity? Also, try visualizing using Silo instead
> of Vtk.
>
> On Wed, 23 Feb 2011 15:20:52 +0100, "Islam El Salawi"
> <islamelsalawi(a)gmx.de> wrote:
> > Receiving results from the square and naca examples I am wondering if I
> > missed something to install. Cuda works fine and there are no errors,
> but
> > the results are far away from being realistic. I posted this problem
> > already once and I was asking if I took the right initialization
> settings.
> > But I am pretty sure that the error source is somewhere else. I've
> attached
> > a screenshot of the streamlines surrounding the square. As you can see
> the
> > vectors close to the noslip boundary condation on the wall of the
> square
> > are not null.
>
> Two answers to that:
>
> - No-slip is only genuine no-slip if you are using a viscosity mu
> greater than 0. Otherwise, there's physically no mechanism that would
> slow/stop tangential flow near the boundary. Not sure what problem you
> are doing, though. If you have mu>0, there might be an issue.
>
> - Even then, you won't see the boundary velocity being *exactly*
> zero. DG enforces boundary conditions weakly, i.e. as you refine your
> discretization, your BC will asymptotically be satisfied--that's the
> best you can ask for. (but BC values should still be "small")
>
> Andreas
>
>
--
Schon gehört? GMX hat einen genialen Phishing-Filter in die
Toolbar eingebaut! http://www.gmx.net/de/go/toolbar
--
NEU: FreePhone - kostenlos mobil telefonieren und surfen!
Jetzt informieren: http://www.gmx.net/de/go/freephone
Hi,
I'm encountering a strange error on the staging branch of hedge. I created a
new virtualenv with distribute and no-site-packages and made a new clean
install according to the wiki instructions in case it was something strange
on my side, and I got the same error again. Checking out my local copy of
the master branch and doing setup.py install makes the error go away.
Traceback (most recent call last):
File "shearflow.py", line 198, in <module>
main()
File "shearflow.py", line 116, in main
rhs(0, fields)
File "shearflow.py", line 111, in rhs
ode_rhs, speed = navierstokes_ex(t, q)
File
> "/home/daniel/cookie/src/env/lib/python2.6/site-packages/hedge-0.91-py2.6-linux-i686.egg/hedge/models/gas_dynamics/__init__.py",
> line 841, in rhs
**extra_kwargs
File
> "/home/daniel/cookie/src/env/lib/python2.6/site-packages/hedge-0.91-py2.6-linux-i686.egg/hedge/backends/jit/__init__.py",
> line 485, in __call__
self.discr.exec_mapper_class(context, self))
File
> "/home/daniel/cookie/src/env/lib/python2.6/site-packages/hedge-0.91-py2.6-linux-i686.egg/hedge/compiler.py",
> line 492, in execute
return self.execute_dynamic(exec_mapper, pre_assign_check)
File
> "/home/daniel/cookie/src/env/lib/python2.6/site-packages/hedge-0.91-py2.6-linux-i686.egg/hedge/compiler.py",
> line 447, in execute_dynamic
insn.get_executor_method(exec_mapper)(insn)
File
> "/home/daniel/cookie/src/env/lib/python2.6/site-packages/hedge-0.91-py2.6-linux-i686.egg/hedge/backends/jit/__init__.py",
> line 54, in exec_vector_expr_assign
compiled(self, stats_callback)), []
File
> "/home/daniel/cookie/src/env/lib/python2.6/site-packages/hedge-0.91-py2.6-linux-i686.egg/hedge/backends/jit/vector_expr.py",
> line 53, in __call__
shape = single_valued(vec.shape for vec in vectors)
File "/home/daniel/cookie/src/pytools/pytools/__init__.py", line 292, in
> single_valued
assert others_same()
File "/home/daniel/cookie/src/pytools/pytools/__init__.py", line 288, in
> others_same
for other_item in it:
File
> "/home/daniel/cookie/src/env/lib/python2.6/site-packages/hedge-0.91-py2.6-linux-i686.egg/hedge/backends/jit/vector_expr.py",
> line 53, in <genexpr>
shape = single_valued(vec.shape for vec in vectors)
*AttributeError: 'int' object has no attribute 'shape'*
Is this error message familiar to anyone? I guess if no one recognizes the
error I'll try to checkout older commits of staging until I find one that
works.
Thanks!
Daniel Gempesaw