Fixed in r2788.
Thanks all!
** Changed in: yade
Status: Confirmed => Fix Released
--
You received this bug notification because you are a member of Yade
developers, which is the registrant for Yade.
https://bugs.launchpad.net/bugs/724396
Title:
crash after add particles in parallel mo
** Branch linked: lp:yade
--
You received this bug notification because you are a member of Yade
developers, which is the registrant for Yade.
https://bugs.launchpad.net/bugs/724396
Title:
crash after add particles in parallel mode
Status in Yet Another Dynamic Engine:
Fix Released
Bug des
I have checked the last patch, all is working well for me!
The performance is slightly higher now, than without patch, so I think we did
not break the performance.
I modified model.py a little bit (attached), now it reads forces of
every body on each step. I did not see any problem with that as
fixed
** Patch added: "fix_724396-v3.diff"
https://bugs.launchpad.net/yade/+bug/724396/+attachment/1908010/+files/fix_724396-v3.diff
--
You received this bug notification because you are a member of Yade
developers, which is the registrant for Yade.
https://bugs.launchpad.net/bugs/724396
Ti
Backtrace is attached
Anton
** Attachment added: "segfault_fix_724396-v2"
https://bugs.launchpad.net/bugs/724396/+attachment/1907394/+files/segfault_fix_724396-v2
--
You received this bug notification because you are a member of Yade
developers, which is the registrant for Yade.
https://bu
The last patch gives a segmentation fault in --test:
yade --test
Math: Matrix3 operations ... ok
Math: Quaternion operations ... ok
Math: Vector2 operations ... ok
Math: Vector3 operations ... ok
Core: correct types are instantiated ... ok
Core: dispatcher ctors with functors ... ok
Core: Attr::hi
Fixed path is attached.
Compilation and test model.py from this thread is ok.
Need test of reading forces from python and perfomance benchmark.
** Patch added: "fix_724396-v2.diff"
https://bugs.launchpad.net/yade/+bug/724396/+attachment/1905334/+files/fix_724396-v2.diff
--
You received this
I tried to implement an idea, indicated here:
https://bugs.launchpad.net/yade/+bug/724396/comments/7
Sergei, could you not, please, test it?
Vaclav, could you not, please, have a look at it?
Patch is attached.
Thanks
** Patch added: "fix_patch.diff"
https://bugs.launchpad.net/yade/+bug/7243
Yes, you are right. The proposed way breaks simulations (--check does
not path with j>1).
I have done locks in non-parallel implementation, and the performance
goes down, as you said:
-j Parallel Non-Parallel %
1 9971002-0.5
2 821842 -2.6
Do you mean that you used non-parallel ForceContainer with OpenMP-
enabled simulation run with multiple threads? If so... threads could
have been overwriting each other's data. https://www.yade-
dem.org/doc/prog.html#forcecontainer explains why threads must have
separate data (unless you want to us
** Attachment added: "Results of Parallel ForceContainer"
https://bugs.launchpad.net/yade/+bug/724396/+attachment/1895043/+files/Parallel.tgz
--
You received this bug notification because you are a member of Yade
developers, which is the registrant for Yade.
https://bugs.launchpad.net/bugs/72
** Attachment added: "Results of NonParallel ForceContainer"
https://bugs.launchpad.net/yade/+bug/724396/+attachment/1895042/+files/NoParallel.tgz
--
You received this bug notification because you are a member of Yade
developers, which is the registrant for Yade.
https://bugs.launchpad.net/bu
I have done some performance tests on 6-core machine.
The simulation have been carried out with different thread numbers and
execution time has been measured. In the first case it was the latest
trunk revision (2781) clean compile, in the second case the
ForceContainer used only "non-parallel flav
** Attachment added: "Scripts to measure performance"
https://bugs.launchpad.net/yade/+bug/724396/+attachment/1895041/+files/scripts.tgz
--
You received this bug notification because you are a member of Yade
developers, which is the registrant for Yade.
https://bugs.launchpad.net/bugs/724396
** Changed in: yade
Importance: Undecided => High
** Changed in: yade
Status: New => Confirmed
--
You received this bug notification because you are a member of Yade
developers, which is the registrant for Yade.
https://bugs.launchpad.net/bugs/724396
Title:
crash after add particles
I don't think so. I think sizeof(vector) is always
sizeof(size_t)+sizeof(void*), the array is allocated on the heap
dynamically. For the independent resizing: sounds fine, though sizes
should be re-synced in sync() before the summation loop, otherwise it
will run slower due to in-loop size checks.
Apparently, _forceData will not store vvector, but vvector*.
Because, if it will store vvector (as now) then _forceData[A].resize(N) can
affect _forceData[B]. Or no?
--
You received this bug notification because you are a member of Yade
developers, which is the registrant for Yade.
https://bugs
I want to rewrite ensureSize() in order to the each thread do resize only the
"own" vector, not all.
So, no need locking at all. Vaclav, do you agree?
--
You received this bug notification because you are a member of Yade
developers, which is the registrant for Yade.
https://bugs.launchpad.net/b
Yes, you're right, the locking logic in ForceContainer::resize is not
sufficient. What you propose does not solve the problem of adding bodies
inside the loop, though -- that is what a factory does, for instance.
--
You received this bug notification because you are a member of Yade
developers, w
Seems, cause is next:
After adding new particles, the first thread runs on the "old" particles, so
without resizing of _forceData in ensureSize() (ForceContainer.hpp:87)
And at time, when first thread adds new force to _forceData, second thread
starts to run on the "new" particles and do resize
I have tried it on Ubuntu 10.04
--
You received this bug notification because you are a member of Yade
developers, which is the registrant for Yade.
https://bugs.launchpad.net/bugs/724396
Title:
crash after add particles in parallel mode
Status in Yet Another Dynamic Engine:
New
Bug descri
Hi, Sergei.
I have tried your script, but I cannot reproduce the crash.
Will try on other systems.
--
You received this bug notification because you are a member of Yade
developers, which is the registrant for Yade.
https://bugs.launchpad.net/bugs/724396
Title:
crash after add particles in p
Seems, something wrong with ForceContainer, see attached log file.
** Attachment added: "crash.log"
https://bugs.launchpad.net/yade/+bug/724396/+attachment/1869104/+files/crash.log
--
You received this bug notification because you are a member of Yade
developers, which is the registrant for
** Attachment added: "model.py"
https://bugs.launchpad.net/bugs/724396/+attachment/1869079/+files/model.py
--
You received this bug notification because you are a member of Yade
developers, which is the registrant for Yade.
https://bugs.launchpad.net/bugs/724396
Title:
crash after add parti
24 matches
Mail list logo