Installation

[jk262626]

Dear Nakib,
Thank you so much for your effort on a wonderful software, elphbolt!

I'm trying to install elphbolt in HPC system (Linux-rocky8-zen2) and this is the shared cluster system like Expanse.
Both the spglib and opencoarrays are installed on my HPC system by myself.
And then, I tried to do the next step '3. build' in your instruction with the following procedure.

1. Load opencoarrays and activate fpm.
2. I have all of the files and folders of your code in my working directory.
3. Type the command instead of executing fpm_config_caf.sh because I'm not sure if the command '. fpm_config_caf.sh' would work for this shared cluster system.
   export FPM_FC=caf
   export FPM_FFLAGS="-g -O2 -fPIC -fbounds-check -Wunused -Wconversion -Wunderflow -Wdo-subscript"
4. Type 'fpm build' and then I encountered error like the figure below. The permission for git@github.com was denied.
   
5. So I deleted the following dependencies line in the fpm.toml file, and then the error disappeared and new error came out like the figure below.
   [dependencies]
   testify = { git = "git@github.com:nakib/testify.git" } .
   

Can I get an advice on this? I want to stick to fpm to build elphbolt.
Thank you!

[nakib]

Dear jk262626,

Thanks for your kind words!

As for your questions:

3. . fpm_config_caf.sh or, equivalently, source fpm_config_caf.sh should work since it just defines a couple of environment variables for the current open shell. I am surprised that it didn't work for you. But manually exporting the flags also work. Additionally, can you confirm that echo $FPM_FC and echo $FPM_FLAGS are reproducing the contents of fpm_config_caf.sh? Also, could you confirm that which caf and which cafrun return something meaningful?

4. Hmm...this seems strange. Is it happening because your are within a conda environment? But let's revisit this since this part is about building the tests.

5. Have you included the header file and linked the shared library of spglib as explained here: https://spglib.readthedocs.io/en/latest/install.html ?

You can do the above once and for all by adding the following to your ~/.bashrc file:

export PATH="<path to spglib build directory>:$PATH"
export LD_LIBRARY_PATH="<path to spglib build directory>:$LD_LIBRARY_PATH"

Don't forget to source ~/.bashrc after the change.

Best,
Nakib

[jk262626]

I confirmed what you said.

I got the source code by
$ git clone https://github.com/spglib/spglib.git
and then tried to compile but encountered the error like this.

Should I need to do this? I skip this.
% tar xvfz spglib-2.0.0.tar.gz
% cd spglib-2.0.0

[nakib]

Ok, thanks for checking.

Regarding spglib, it looks like the build system is trying to update something but failing to connect to github. Could be a related issue to the error you got when fpm tried to clone the testify repo.

You mentioned that spglib is already available on your HPC system as a module. Could you check where libsymspg.so is located? That is, what is the output of locate libsymspg.so ?

[jk262626]

I checked!

I guess I already have libsymspg.so (FYI, I've installed it by using $ conda install -c conda-forge spglib).
If I correctly have it on my system, then should I add the following commands to the end of the ~/.bashrc file and then type source ~/.bashrc?
export PATH="path to spglib build directory:$PATH"
export LD_LIBRARY_PATH="path to spglib build directory:$LD_LIBRARY_PATH"

That "path to spglib build directory" would be one of the directories in the picture?

[nakib]

Since I am not a conda user, I am actually not sure if it sets up the environments such that the shared library (libsymspg.so) will be findable in your fpm environment. From the 2nd line in the screenshot, it looks like it is available. You can double check if echo $LD_LIBRARY_PATH includes this location.

It looks like you have managed to install spglib successfully. I think you can retry building elphbolt.

[nakib]

Could you please try to build the latest develop branch? I made 2 changes to the fpm.toml file:

1. I added the following in fpm.toml. You are using a new version of fpm which by default does not allow compilation of code with an implicit interface. This is why the file spglib_wrapper.f90 was not being built. With the following, this error should go away:

...

[fortran]
implicit-external = true #Needed for v>=0.7 fpm

...

2. With the following change, fpm will build testify from a local copy instead of trying to pull from the remote repository.

...

[dependencies]
#Pull from github
#testify = { git = "git@github.com:nakib/testify.git" }
#Use local copy from thirdparty directory
testify = {path = "thirdparty/testify"}

...

Would you please let me know if this solves you issues?

Best,
Nakib

[jk262626]

Thank you so much for the update!

Following your instruction, f90 files can be compiled but I encountered the error at the end like this.

[nakib]

This is because you don't have libsymspg, libblas, and liblapack available. Have you loaded the necessary modules? Typically openblas is available as a module. You also mentioned spglib is available as a module. Please try loading these first.

[jk262626]

Here is the directory of each libraries.

They are all included in /home/jkim3/miniconda3/envs/fpm/lib, so I added the path to the ~/.bashrc file and typed source ~/.bashrc.
export PATH="/home/jkim3/miniconda3/envs/fpm/lib:$PATH"
export LD_LIBRARY_PATH="/home/jkim3/miniconda3/envs/fpm/lib:$LD_LIBRARY_PATH"

And then I tried to build again, but have same error.

[nakib]

Hmm...this is quite surprising.

Can you do a module load spglib? I'm assuming that's what the module is called on your HPC stack. The thing is that spglib requires libblas and liblapack.

[jk262626]

Yea, I also loaded openblas module for both libblas and liblapack but even the errors for those two still appear.
I should check this with my cluster manager and then I will reach out to you again!

[jk262626]

Hi Nakib,
I have all of libsymspg, libblas, and liblapack in my conda environment and also have openblas, but I still have the same error like the figure below.

I already have changed the PATH and LD_LIBRARY_PATH to where library files are located like the below, but it is still finding the library files in /usr/bin/ld.
export PATH="/home/jkim3/miniconda3/envs/fpm/lib:$PATH"
export LD_LIBRARY_PATH="/home/jkim3/miniconda3/envs/fpm/lib:$LD_LIBRARY_PATH"

[jk262626]

What is the "path to spglib build directory"?

[jk262626]

I have also tried to build elphbolt by using cmake, but had same error.

[nakib]

The issue seems to be related to the compiler not finding the libraries to link to. Let's continue with your cmake build. Can you try the following replacement in your CMakeLists.txt file (I'm assuming your have openblas; change to whatever flavor you have):

...

#Set flags for caf
if(CMAKE_Fortran_COMPILER MATCHES "caf")
  set(CMAKE_Fortran_FLAGS "${CMAKE_Fortran_FLAGS} -g -O2 -cpp -fPIC -fbounds-check -Wunused -Wconversion -Wunderflow -Wdo-subscript -L<enter path to the openblas lib directory> -lopenblas")
  #set(CMAKE_Fortran_FLAGS "${CMAKE_Fortran_FLAGS} -g -O2 -fPIC -fbounds-check -Wunused -Wconversion -Wunderflow -Wdo-subscript")
endif()

#Set spg library
find_library(SPG_LIB NAMES symspg REQUIRED)
#find_library(BLAS_LIB NAMES blas REQUIRED)
#find_library(LAPACK_LIB NAMES lapack REQUIRED)

#Set library links
#set(LINKS "-lblas -llapack ${SPG_LIB}")
set(LINKS "${SPG_LIB}")

...

Basically, above I am explicitly setting the linker instead of asking cmake to find the libraries.

Could you also reach out to your cluster manager if the above does not work? But also let me know how it goes.

As for your other question, "path to spglib build directory" refers to where the libsymspg.a, libsymspg.so, etc. files are.

Best regards,
Nakib

[jk262626]

Hi Nakib,
Thanks to your comments, it seems to be linked to the library now.
I modified Cmake file like this.

#Set flags for caf
if(CMAKE_Fortran_COMPILER MATCHES "caf")
  set(CMAKE_Fortran_FLAGS "${CMAKE_Fortran_FLAGS} -g -O2 -cpp -fPIC -fbounds-check -Wunused -Wconversion -Wunderflow -Wdo-subscript -L/home/jkim3/miniconda3/pkgs/libopenblas-0.3.21-h043d6bf_0/lib/-lopenblas")
  #set(CMAKE_Fortran_FLAGS "${CMAKE_Fortran_FLAGS} -g -O2 -fPIC -fbounds-check -Wunused -Wconversion -Wunderflow -Wdo-subscript")
endif()

#Set spg library
find_library(SPG_LIB NAMES symspg REQUIRED)
#find_library(BLAS_LIB NAMES blas REQUIRED)
#find_library(LAPACK_LIB NAMES lapack REQUIRED)

#Set library links
#set(LINKS "-lblas -llapack ${SPG_LIB}")
set(LINKS "${SPG_LIB}")

Now, I'm getting a new error like this figure below!

[nakib]

Is there a space missing before -lopenblas in the linking command?

…

On Fri, Dec 1, 2023, 18:15 Jiyoung Kim ***@***.***> wrote: Hi Nakib, Thanks to your comments, it seems to be linked to the library now. I modified Cmake file like this. #Set flags for caf if(CMAKE_Fortran_COMPILER MATCHES "caf") set(CMAKE_Fortran_FLAGS "${CMAKE_Fortran_FLAGS} -g -O2 -cpp -fPIC -fbounds-check -Wunused -Wconversion -Wunderflow -Wdo-subscript -L/home/jkim3/miniconda3/pkgs/libopenblas-0.3.21-h043d6bf_0/lib/-lopenblas") #set(CMAKE_Fortran_FLAGS "${CMAKE_Fortran_FLAGS} -g -O2 -fPIC -fbounds-check -Wunused -Wconversion -Wunderflow -Wdo-subscript") endif() #Set spg library find_library(SPG_LIB NAMES symspg REQUIRED) #find_library(BLAS_LIB NAMES blas REQUIRED) #find_library(LAPACK_LIB NAMES lapack REQUIRED) #Set library links #set(LINKS "-lblas -llapack ${SPG_LIB}") set(LINKS "${SPG_LIB}") Now, I'm getting a new error like this figure below! image.png (view on web) <https://github.com/nakib/elphbolt/assets/138066037/82f24c1f-ab94-4c35-9b99-f169ef195525> — Reply to this email directly, view it on GitHub <#120 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAN7Y2K5F5FPXFNXLMJ7IO3YHIGEDAVCNFSM6AAAAAA7DFMDNGVHI2DSMVQWIX3LMV43SRDJONRXK43TNFXW4Q3PNVWWK3TUHM3TOMZRHAZDO> . You are receiving this because you commented.Message ID: ***@***.***>

[jk262626]

Yes I was missing some space

[jk262626]

Awesome!! I finally can compile it.


I'm running ctest now.

[nakib]

Great! I'll let you know when I figure out why the fpm build didn't work for you.

…

On Fri, Dec 1, 2023, 19:46 Jiyoung Kim ***@***.***> wrote: Awesome!! I finally can compile it. image.png (view on web) <https://github.com/nakib/elphbolt/assets/138066037/c0842662-0097-4041-b326-a6ed960672b4> image.png (view on web) <https://github.com/nakib/elphbolt/assets/138066037/59e214fb-55bf-46db-afbd-4f315a44ef4c> I'm running ctest now. — Reply to this email directly, view it on GitHub <#120 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAN7Y2L3EEHOVGXF4WKJKRTYHIQWVAVCNFSM6AAAAAA7DFMDNGVHI2DSMVQWIX3LMV43SRDJONRXK43TNFXW4Q3PNVWWK3TUHM3TOMZSGU3TE> . You are receiving this because you commented.Message ID: ***@***.***>

[jk262626]

Nice thanks!

[jk262626]

I encountered some errors after running ctest.
Should I address this or can I just move forward with Si example?

[nakib]

The tests should pass. Could you share the output of the test? I can take a look on Monday.

…

On Fri, Dec 1, 2023, 21:46 Jiyoung Kim ***@***.***> wrote: image.png (view on web) <https://github.com/nakib/elphbolt/assets/138066037/676d6d8f-6f79-4978-bb91-d12b6f5aa8e8> I encountered some errors after running ctest. Should I address this or can I just move forward with Si example? — Reply to this email directly, view it on GitHub <#120 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAN7Y2O6B6262AABRGUJNLTYHI6YPAVCNFSM6AAAAAA7DFMDNGVHI2DSMVQWIX3LMV43SRDJONRXK43TNFXW4Q3PNVWWK3TUHM3TOMZUGAZDQ> . You are receiving this because you commented.Message ID: ***@***.***>

[jk262626]

I understand now. Thank you for the clarification.
I tried running AlN calculation with full set of inputs and the calculation was terminated incorrectly.
Here is some part of the slurm file. Please take a look at it.

Number of coarray images = 20

---

______________________________________________________Setting up crystal...
At line 200 of file /home/jkim3/elphbolt/src/crystal.f90 (unit = 1, file = 'input.nml')
Fortran runtime error: Cannot match namelist object name autoisotopes

Error termination. Backtrace:
At line 200 of file /home/jkim3/elphbolt/src/crystal.f90 (unit = 1, file = 'input.nml')
Fortran runtime error: Cannot match namelist object name autoisotopes

Error termination. Backtrace:
At line 200 of file /home/jkim3/elphbolt/src/crystal.f90 (unit = 1, file = 'input.nml')
Fortran runtime error: Cannot match namelist object name autoisotopes

[nakib]

Since v1.1, the flag autoisotopes does not exist. Instead, set the flag VCA = .true. to tell the code to use the isotopic average of each basis atom. Alternatively, set the flag DIB = .true. to tell the code to use the dominant isotopic mass.

Best,
Nakib

[jk262626]

Thanks you so much! I resolved by using DIB = .true.
..............

---

______________________________________________________Setting up Wannier...
Wannier data from EPW will be read.
At line 312 of file /home/jkim3/elphbolt/src/wannier.f90 (unit = 1, file = 'epwdata.fmt')
Fortran runtime error: Bad real number in item 1 of list input

Error termination. Backtrace:
At line 312 of file /home/jkim3/elphbolt/src/wannier.f90 (unit = 1, file = 'epwdata.fmt')
Fortran runtime error: Bad real number in item 1 of list input

Error termination. Backtrace:
At line 312 of file /home/jkim3/elphbolt/src/wannier.f90 (unit = 1, file = 'epwdata.fmt')
Fortran runtime error: Bad real number in item 1 of list input

I encountered error like this for this time

[nakib]

What is the content of the file epwdata.fmt? Can you show the first few lines here?

[jk262626]

Oh I missed data from the epw dft calculation. I will let you know after doing that again. Thanks!