New question #659557 on Yade:
https://answers.launchpad.net/yade/+question/659557
Hi folks,
I face a very strange behavior with the flow engine. With useSolver=3, the same
simulation can either work on not. More precisely, the flow can be computed or
not depending on... well, I don't know...
F
Question #659557 on Yade changed:
https://answers.launchpad.net/yade/+question/659557
Status: Open => Needs information
Bruno Chareyre requested more information:
Hi Luc,
I've heard the same story from different sides in the last months. It seems the
linear solver has issues.
Can you show he
Question #659557 on Yade changed:
https://answers.launchpad.net/yade/+question/659557
Status: Needs information => Open
Luc Scholtès gave more information on the question:
Hi Bruno,
Thank you for the prompt answer.
Here is what the ldd command gives:
linux-vdso.so.1 => (0x7ffe
Question #659557 on Yade changed:
https://answers.launchpad.net/yade/+question/659557
Status: Open => Answered
Robert Caulk proposed the following answer:
Hello Luc,
I ran your script 10 times with each of useSolver=0 and 3 (also 4 to
check the GPU consistency). I was unable to reproduce the
Question #659557 on Yade changed:
https://answers.launchpad.net/yade/+question/659557
Status: Answered => Open
Luc Scholtès is still having a problem:
Hi Robert,
Thank you for your advice but I already did that... twice... I even
removed yade and yadedaily and reinstalled them. Always the sa
Question #659557 on Yade changed:
https://answers.launchpad.net/yade/+question/659557
Robert Caulk posted a new comment:
H, when you mention that it fails 2 or 3 times to give the correct
result. What result is the failed result exactly? You mention that the
flow is simply not computed 20-30%
Question #659557 on Yade changed:
https://answers.launchpad.net/yade/+question/659557
Luc Scholtès gave more information on the question:
Typically, here is what I get:
Qin= -308.240063332 Qout= 0.0
Permeability [m2]= 0.0 || Hydraulic conductivity [m/s]= 0.0
and when I open the vtk file, I ca
Question #659557 on Yade changed:
https://answers.launchpad.net/yade/+question/659557
Bruno Chareyre posted a new comment:
>CMake Warning:
Manually-specified variables were not used by the project:
DFNFLOW
Syntax problem. DFNFLOW is a flag for gcc, not for cmake. Consequently you have
to
Question #659557 on Yade changed:
https://answers.launchpad.net/yade/+question/659557
Luc Scholtès gave more information on the question:
Before reinstalling ubuntu 14, I would liek to ask if this problem coul
dbe related to the libsuitesparse-metis-dev that needs to be install
(cf. yade installat
Question #659557 on Yade changed:
https://answers.launchpad.net/yade/+question/659557
Status: Open => Answered
Robert Caulk proposed the following answer:
Hello Luc,
Actually, you should be able to use these packages as a replacement:
sudo apt-get -y install libopenblas-dev libsuitesparse-d
Question #659557 on Yade changed:
https://answers.launchpad.net/yade/+question/659557
Status: Answered => Open
Luc Scholtès is still having a problem:
Alright...
I reinstalled ubuntu (16.04.3) and reinstalled yade (git version):
It seems to work correctly now.
Why?
I don't know... Probabl
Question #659557 on Yade changed:
https://answers.launchpad.net/yade/+question/659557
Status: Open => Answered
Robert Caulk proposed the following answer:
We might a swell sort it out here since FlowEngine's behavior doesn't
change.
Hmmm, I use the same method you describe for getting packin
Question #659557 on Yade changed:
https://answers.launchpad.net/yade/+question/659557
Status: Answered => Open
Luc Scholtès is still having a problem:
Here you go. It fails every time. Of course, you'll have to create your
own packing (cf PACK parameter in the first lines).
Also, I tried to
Question #659557 on Yade changed:
https://answers.launchpad.net/yade/+question/659557
Status: Open => Answered
Robert Caulk proposed the following answer:
The script runs with success and produces:
Qin= -5.06465867124e-20 Qout= 5.06465867124e-20 ARE THEY EQUAL? IF NOT-> NO
FLOW!
Permeabil
Question #659557 on Yade changed:
https://answers.launchpad.net/yade/+question/659557
Status: Answered => Open
Luc Scholtès is still having a problem:
Yes, same problem with either FlowEngine or DFNFlowEngine (which is kind
of reassuring and "logical"): the flow is not computed.
I guess the
Question #659557 on Yade changed:
https://answers.launchpad.net/yade/+question/659557
Luc Scholtès posted a new comment:
FYI, the O.reset() script works with yade installed from sudo-apt get
installed yade (yade 1.20.0).
--
You received this question notification because your team yade-users is
Question #659557 on Yade changed:
https://answers.launchpad.net/yade/+question/659557
Status: Open => Answered
Bruno Chareyre proposed the following answer:
>FlowEngine requires the walls to be appended in the simulation before
the spheres, right?
Not at all. But you have to specify their id
Question #659557 on Yade changed:
https://answers.launchpad.net/yade/+question/659557
Status: Answered => Open
Luc Scholtès is still having a problem:
Hi guys,
I reinstalled everything from scratch (ubuntu 16.04 LTS and yade from
git). The problem persists and I cannot figure out what is wro
Question #659557 on Yade changed:
https://answers.launchpad.net/yade/+question/659557
Status: Open => Answered
Bruno Chareyre proposed the following answer:
> boundaryUseMaxMin=[1,1,1,1,1,1]
It means that instead of walls positions it will use static positions
defined by max/min (the static
Question #659557 on Yade changed:
https://answers.launchpad.net/yade/+question/659557
Robert Caulk posted a new comment:
If you post the two scripts that you have created for testing
boundaryUseMaxMin() and wallIds(), I will gladly test them on my setup.
I have never used either of those functions
Question #659557 on Yade changed:
https://answers.launchpad.net/yade/+question/659557
Status: Answered => Open
Luc Scholtès is still having a problem:
- OK for the boundaryUseMaxMin=[1,1,1,1,1,1] . I misunderstood the doc.
- For the calibration of the coordination number, yes, we could do th
Question #659557 on Yade changed:
https://answers.launchpad.net/yade/+question/659557
Status: Open => Answered
Bruno Chareyre proposed the following answer:
> for the boundaryUseMaxMin=[1,1,1,1,1,1] . I misunderstood the doc.
Or the doc is not clear... feel free to improve when you spot miss
Question #659557 on Yade changed:
https://answers.launchpad.net/yade/+question/659557
lingran.zhang posted a new comment:
Dear all,
I confirm that I am facing exactly the same problem as Luc Scholtès,
that is, Flow engine works randomly! Not even to mention that strange
results were obtained duri
Question #659557 on Yade changed:
https://answers.launchpad.net/yade/+question/659557
Robert Caulk proposed the following answer:
- Regarding the problem with the solver, I guess I'll need to figure it
out by myself... Not an easy task...
I will gladly help you down that path if there really is a
Question #659557 on Yade changed:
https://answers.launchpad.net/yade/+question/659557
Status: Answered => Solved
Luc Scholtès confirmed that the question is solved:
Well, you are right, I made a mistake in my script when asking for the
Qin and Qout. It gives the correct result if, of course,
Question #659557 on Yade changed:
https://answers.launchpad.net/yade/+question/659557
Bruno Chareyre posted a new comment:
I'm confident in the fact that something random occurs. This is consistently
reported by you (Luc) and Lingran (thanks for detailed report).
This is most likely unrelated to
Question #659557 on Yade changed:
https://answers.launchpad.net/yade/+question/659557
Lingran Zhang posted a new comment:
Hi, Bruno,
Just let you know, I've installed openblas from git (compiled with 'make
FC=gfortran USE_OPENMP=0') as well as installed yade from git (with
cmake -DENABLE_OPENMP=0
Question #659557 on Yade changed:
https://answers.launchpad.net/yade/+question/659557
Bruno Chareyre posted a new comment:
@Lingran,
Thank you very much for reporting. Further, you could compile with a
replacement of openblas (atlas, or legacy blas) since I suspect openblas could
be the problem.
Question #659557 on Yade changed:
https://answers.launchpad.net/yade/+question/659557
Luc Scholtès posted a new comment:
Hi,
Just an additional comment:
Running my simulations with debug=1, I obtain the exact same output when
it works and when it does not work.
Even at the very last line of the
Question #659557 on Yade changed:
https://answers.launchpad.net/yade/+question/659557
Bruno Chareyre posted a new comment:
The total force is entirely defined by the fluid boundary conditions (not from
an algorithmic point of view, but mathematically it is the case).
For instance the total force
Question #659557 on Yade changed:
https://answers.launchpad.net/yade/+question/659557
Bruno Chareyre posted a new comment:
Hi Luc, Lingran,
Did you check if some older versions (that "were" working) would work for you?
I'd like to be sure that it is not a problem introduced recently in the code.
Y
Question #659557 on Yade changed:
https://answers.launchpad.net/yade/+question/659557
Lingran Zhang posted a new comment:
Hi Bruno,
I've just installed yade from git (Branch:2017.01, uncomment the '#define
DFNFLOW') and tested the 5 scripts provided above by Luc.
The random thing in Flow Engine
Question #659557 on Yade changed:
https://answers.launchpad.net/yade/+question/659557
Status: Solved => Open
Luc Scholtès is still having a problem:
Hi,
Just to confirm that I don't have the random behavior with yade
2017.10.a as mentioned by Lingran. Why? I have no idea but I'll stick to
th
Question #659557 on Yade changed:
https://answers.launchpad.net/yade/+question/659557
Status: Open => Answered
Robert Caulk proposed the following answer:
Hey Luc,
>However, this version does not allow to run simulations with
DFNFlowEngine as it produces a segfault
Yes, the first DFNFlow co
Question #659557 on Yade changed:
https://answers.launchpad.net/yade/+question/659557
Luc Scholtès posted a new comment:
Robert, you made my day!
There was another problem linked to the declaration of the
trickPermeability() function but, your bug report was what I was looking
for. It seems that
Question #659557 on Yade changed:
https://answers.launchpad.net/yade/+question/659557
Robert Caulk proposed the following answer:
Glad to help!
Correct me if I am wrong, but the randomness is caused by DFNFlow, you
observe it with the basic FlowEngine as well in your original post. So I
do not th
Question #659557 on Yade changed:
https://answers.launchpad.net/yade/+question/659557
Robert Caulk posted a new comment:
*randomness is [b]not[/b] caused
--
You received this question notification because your team yade-users is
an answer contact for Yade.
__
I spotted these:
https://github.com/yade/trunk/commit/d63293b5d6397c8d956e9d533f33ce4077667c0a
https://github.com/yade/trunk/commit/f6970362d9e6e866e8adbc8cbea18e54f677f785
B
___
Mailing list: https://launchpad.net/~yade-users
Post to : yade-users
Question #659557 on Yade changed:
https://answers.launchpad.net/yade/+question/659557
Bruno Chareyre posted a new comment:
I spotted these:
https://github.com/yade/trunk/commit/d63293b5d6397c8d956e9d533f33ce4077667c0a
https://github.com/yade/trunk/commit/f6970362d9e6e866e8adbc8cbea18e54f677f785
B
Question #659557 on Yade changed:
https://answers.launchpad.net/yade/+question/659557
Bruno Chareyre proposed the following answer:
Hi,
As far as I understand DFN is not the question, I agree with Robert.
Third party libraries may also not be to blame in the end, if a previous
revision is ok (I d
Question #659557 on Yade changed:
https://answers.launchpad.net/yade/+question/659557
Luc Scholtès posted a new comment:
Yes, DFNFlow is not the problem. The randomness appears with the "whole
FlowEngine". Sorry if my message made you think so.
Since it seems that my version built from 2017.01 do
Question #659557 on Yade changed:
https://answers.launchpad.net/yade/+question/659557
Robert Caulk posted a new comment:
Hey Bruno,
I am responsible for the second commit you posted [1]. I don't think it
is to blame, since Luc's original script here does not use
multithread=True, and therefore do
Question #659557 on Yade changed:
https://answers.launchpad.net/yade/+question/659557
Bruno Chareyre proposed the following answer:
@Luc
Yes, let's have a bug.
B
--
You received this question notification because your team yade-users is
an answer contact for Yade.
__
Question #659557 on Yade changed:
https://answers.launchpad.net/yade/+question/659557
Bruno Chareyre posted a new comment:
@Robert
I tried to identify the commits which touched relevant code, but I could not
spot anything wrong precisely.
Bruno
--
You received this question notification because
Bruno Chareyre said: (by the date of Wed, 08 Nov 2017 17:57:59 -)
> Question #659557 on Yade changed:
> https://answers.launchpad.net/yade/+question/659557
>
> Bruno Chareyre posted a new comment:
> @Robert
> I tried to identify the commits which touched relevant code, but I could not
>
Question #659557 on Yade changed:
https://answers.launchpad.net/yade/+question/659557
Janek Kozicki proposed the following answer:
Bruno Chareyre said: (by the date of Wed, 08 Nov 2017 17:57:59
-)
> Question #659557 on Yade changed:
> https://answers.launchpad.net/yade/+question/659557
>
Question #659557 on Yade changed:
https://answers.launchpad.net/yade/+question/659557
Robert Caulk proposed the following answer:
Looks like Bruno pointed us in the right direction at the end but I
disagreed :-/.
Luc, I just reread this thread and noticed you were installing i386
libraries? Is yo
Question #659557 on Yade changed:
https://answers.launchpad.net/yade/+question/659557
Status: Answered => Solved
Luc Scholtès confirmed that the question is solved:
No worries. I cannot really blame you for trying to improve the code.
That's part of the game I would say.
Regarding my compute
48 matches
Mail list logo