Slow speed after compile Opensees 64 Bit

[xmp222]

Hi, I successfully compiled Opensees 2.5.0 64-Bit with VS 2010 (and also 2012). However analysis are significantly slower with my Opensees.exe compared to one I downloaded (OpenSees2.5.0-x64.exe). My opensees solves a model in 1:30 mins, original one solves in 10 seconds!

I tried both debug and release mods. Maximize Speed (/O2) option SEE2 and the other combinations...

Is there something I missed ? File sizes are also different (10 mb vs 17 mb). I tried two different pcs result is same. Should I build on another platform or use some special optimization features ?

Thank you.

[fmk]

the released version is actually built using th eintel compilers and the intel mkl math libraries . though 13 times faster seems excessive .. what solver are you using? .. if umfpack, try another one to test

[xmp222]

Dear fmk, thank you for your quick answer.

The results are taken about 36000 dof model and i tried all the solvers (best one is Band SPD), the released one always 4-5 times faster.

So if i would like to build as you do, what should I do ? There is no mkl library in Other folder (I think its not free), and also do i need a additional project or the source code is sufficient for this ?

Thank you very much again.

[fmk]

you need to obtain the intel compilers (C,C++ and Fortran) .. they are not free .. then there is a bunch of work to do in building the fortran libs for all the fortran code that is in the repository ... then you need to replace the libs with the new fotran ones and with the ones intel provides in the linker sections of the OpenSees project ..
it takes time and it requires you know what you are doing.

[TPdct]

If you must use the Opensees compiled by youself, you can try building it on a faster computer. I'm using Ubuntu 16.10, 8 cores Intel CPU, 16 GB memory, the OpenSees was built by myself. It's still faster than the downloaded version of OpenseesSP installed on another computer using Mumps solver...