libarpack_INTEL_i8.a

Probably a dumb question, but I have not been able to find an answer without your help so far.

I am trying to compile ccx for i8 under WSL/Ubuntu and cannot understand where to get / how to install libarpack_INTEL_i8.a or, more generally, arpack for i8.

Thank you

Before trying to build from source, try this executable:

https://www.feacluster.com/calculix.php#3

Otherwise, try this installation script which should automate everything ( though it will use spooles solver which is very slow)

https://www.feacluster.com/calculix.php#4

1 Like

Thank you very much indeed. However, I am not sure I will be able to use your advice for my purpose for the following reason. I don’t have enough RAM (I only have 64 GB), so I have to use Pardiso’s out-of-core capabilities. Are these capabilities included in your executable? And Spooles does not have out-of-core capabilities, as far as I know.

I don’t recall fiddling with any out-of-core functionality in my executable. If you have the Intel compiler, then you can try building the executable with the installation script mentioned above… Then you can modify the source code for the out of core stuff.

1 Like

Thank you. I have looked at the installation script. It looks like it installs arpack96*. That is what I have now, but my understanding is it is not i8. I believe I need arpack i8 mentioned in ccx installation guide, but I don’t know where to find it. My computation does not run properly with i4 (with the number of nodes that I need). I switched everything but arpack to i8 and am getting a segmentation fault. As for Pardiso, I have already set it up for out-of-core computation.

I think it does make some changes to arpack make file for i8.

1 Like

Maybe this post?

1 Like

Thank you, I’ve seen that. But I guess that is for Windows? Or can it be adjusted for linux as well?

Not sure if they are only for windows. I assume that since you have the source files, you could give it a try on Linux and report back?

Thank you I will look into it.

Thank you feacluster and jbr. Thank you very much indeed.

I think I have sorted it out with your help. So I used the same arpack96* as feacluster, added gfortran option -fdefault-integer-8 to FFLAGS in ARmake.inc and made a library libarpack_IA64.a (its name is defined by the value of PLAT in ARmake.inc : I set PLAT = IA64, but it does not matter) and referenced this library in my Make_i8 file. No segmentation fault anymore.

1 Like