The OpenSees Community

On linux you need to turn on optimizations, native architecture (-march option) support and also provide a native BLAS or intel MKL for fastest support.

You need to hand compile it for now. You can download the source code and then compile it with visual studio or on linux.
I believe the next version of OpenSees will have a python module download option. It's pretty cool, though I've been sticking with tcl lately.

In the end it was not compiling the AMD library. I included this in the DIRS variable and the compile and link worked.
Now I need to compile for parallel processing.

Cheers.

Thank you! I am using python (matplotlib) inspired in MatLab for plotting. After a few tests I found out that even though the output is binary (doubles) the recorder still outputs a newline '\n' character before starting a new time. This could be helpful to anyone out there not willing to read the s...

I am outputting huge ammounts of 'node disp recorder ' and I am using binary file output to minimize disk usage. I haven't been able to figure out the format of this output though to read it back into some other code for plotting.

What is the output format for the binary mode recorders?

Thanks!

Hi Frank, I can compile regular OpenSEES on any node of our Ubuntu cluster. We would now like to take it a step further and compile it for parallel processing first within one node and then across multiple nodes. How do I need to modify the Makefile.def file to achieve this assuming I have all the d...

Thank you frank, Now into the next error. I'm getting the following: /home/jose/lib/libOpenSees.a(AMDNumberer.o): In function `AMD::number(Graph&, int)': AMDNumberer.cpp:(.text+0x191): undefined reference to `amd_order' collect2: ld returned 1 exit status make[1]: *** [tcl] Error 1 make[1]: Leaving ...

BTW, I compiled the previous version of OpenSees using my current settings and the same makefile.

I'm having the same problem (Ubuntu 9.10 Karmic and Ubuntu 10.4 Lucid) using the g++4.4 compiler.

Thank you, this worked in Ubuntu Karmic.

On ubuntu 9.10 I complete the compiling but when I try to execute I get a Segmentation Fault. Are there additional settings?

The problem occurs even if I input a rectangular pulse. The solution looks nice for dt = 0.05 but diverges for dt = 0.02 or lower (faster for lower values).

Plus, I have damping so the solution, even if there is resonance, wouldnt explode undefinitely.

There is one freak symptom. If I make the integration step shorter, the instability gets worse!

The lower the step size the earlier it occurs.

The problem is still occurring, just not as fast as before so I missed it. It occurs further along the integration.

I'm out of ideas now.

I tried ProfileSPD, the model uses about 1GB ram. The program is not complaining, so I guess it's not a numbering problem. I've checked this a couple of times, but I'm not feeling too confident after all. The time it's taking to solve the problem with ProfileSPD is about 4 or 5 times the time UmfPac...