[Users] Compiling ET_2023_11 on Sciama

[Roland Haas]

Hello Robyn, all,

Just responding here to "close" the conversation.

Robyn and I have had an offline discussion and I was able to help with
higher level issue that was what triggered to attempt at using the
newer release.

Yours,
Roland

> Hi everyone,
> I've been unsuccessfully trying to run the latest version of Einstein
> Toolkit on the Sciama HPC cluster where I have been using the 2020_05
> version so far. Could someone please help? I have tried three
> different approaches and put the terminal output and config.log files
> of each in the folder here (too heavy for
> email)Compiling_ET_2023_11_on_Sciama - Google Drive as well as the
> Sciama config files I used in the B approach.
> 
> A) Using the config files already included in Einstein Toolkit, using
> intel_comp/2019.2 it gives the error Cactus requires a C++11 compiler
> -- check your C++ compiler and C++ compiler flags
> 
> If using g++ at least version 6 is required, if using icpc use
> -gxx-name to use libstc++ from g++ at least version 6 I tried using
> intel_comp/2020.1 (the latest available on Sciama) and got the same
> error (and some module warnings). I found the path to g++ and so
> tried to add -gxx-name =
> /opt/apps/compilers/gnu_comp/cfg/9.1.0/intel64/bin/g++ to CXXFLAGS in
> sciama.cfg but this didn't change anything. B) Using gnu_comp/13.2.0
> instead, I changed modules and flags where you can see in the Google
> Drive folder the .ini and .cfg files I use. It compiles all the way
> to the part "Creating cactus_sim" where it lists all the thorns then
> I get lots of errors of the type patch_system.cc:(.text+0xbe):
> undefined reference to `H5open'
> 
> /usr/bin/ld: Dwarf Error: found dwarf version '1', this reader only
> handles version 2, 3 and 4 information. I have this message for lots
> of files .cc, .c, .f90 undefined reference to lots of different H5
> functions (and also dsytrf_ dgetrf_ dgesv_  dsysv_  dposv_  dgbsv_)
> complaining about dwarf version 'any number'. The only other somewhat
> similar example I found of this online is here
> https://urldefense.com/v3/__https://github.com/tinygo-org/tinygo/issues/1415__;!!DZ3fjg!9mjP-VFecdeSB1L_ZQPEPFbDRHvM6zAk-eawkY7KzjykhzljCNXf7RKt-MwIZIuXct1A_TaOwfU6QxsJ5eDxq8I$
>  but it's not that helpful. I'm guessing it's unhappy with hdf5,
> unfortunately, the latest version on Sciama is 1.10.5. Do I just need
> a later version of hdf5 or is this something else? C) I tried with
> the EinsteinToolkit hdf5 (that I see is version 1.12.0)  with
> HDF5_DIR = BUILD and commented out HDF5_LIB_DIRS and HDF5_INC_DIRS,
> as it compiles it doesn't get as far. In the output, I'm under the
> impression it manages to build it properly, but later it seems to not
> be unable to find it, but it should just be within Einstein Toolkit
> right? Thank you for your help,Robyn
> 
> 


-- 
My email is as private as my paper mail. I therefore support encrypting
and signing email messages. Get my PGP key from http://pgp.mit.edu .
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 833 bytes
Desc: OpenPGP digital signature
URL: <http://lists.einsteintoolkit.org/pipermail/users/attachments/20240222/11a7c794/attachment.sig>