MPI error compiling 5.2.10

[Anonymous]

Hello everybody,

the new versopn of elk is out and that means a new fight with compiler variable to get it running on all my machines :(
Does anybdy got a similar error while compiling the new elk with MPI support?

modmpi.f90(11): error #6401: The attributes of this name conflict with those made accessible by a USE statement. [MPI_COMM]
integer mpi_comm
--------^
compilation aborted for modmpi.f90 (code 1)
make[1]: [modmpi.o] Error 1
make[1]: Leaving directory `/home/schwier/elk-5.2.10/src'
make: [all] Error 2

The compilation works without a problem on my linux box (using only openmp) and the previous version compiled without a hitch on the cluster...
I tried to use my optimized make.inc as well as the standard make.inc (only exchanging the name of my fortran compiler mpif90 > mpiifort)

google is not very supporting in identifying the error as it only points to an error within the code.

If anbody has an idea how to solve the problem it would be highly apreciated.

best regards,
Eike

[J. K. Dewhurst]

Hi Eike,

It appears that 'mpi_comm' is a restricted name on your compiler.

I've renamed it to 'mpicom'. Download a pre-release from: http://elk.sourceforge.net/elk-5.2.11.tgz and try and compile it.

Also, please don't post anonymously. All anonymous posts require moderator approval.

Let me know how you get on.

Regards,
Kay.

[Eike Schwier]

Hi Kay,

sorry for the inconvenience, I thought I was still logged in to sourceforge.

As for the pre-release: Thank you for the link. It compiles now on the cluster and I could test-run calculations. (I just need to update the libxc now at some point)

best regards,
Eike

[Eike Schwier]

Dear Kay,

while libxc works nicely now I found that elk 5 crashes on all my linux machines if I try to create species.in files from the elk.in file in the species folder with:

forrtl: severe (174): SIGSEGV, segmentation fault occurred
Image              PC                Routine            Line        Source             
elk                0000000000D73B09  Unknown               Unknown  Unknown
elk                0000000000D723DE  Unknown               Unknown  Unknown
elk                0000000000D269B2  Unknown               Unknown  Unknown
elk                0000000000CC68C8  Unknown               Unknown  Unknown
elk                0000000000CCB2EB  Unknown               Unknown  Unknown
libpthread.so.0    00007FBCD2C7C1B0  Unknown               Unknown  Unknown
elk                0000000000414CF9  Unknown               Unknown  Unknown
elk                0000000000485598  Unknown               Unknown  Unknown
elk                00000000004EC8C0  Unknown               Unknown  Unknown
elk                0000000000458A87  Unknown               Unknown  Unknown
elk                0000000000415470  Unknown               Unknown  Unknown
elk                0000000000404B66  Unknown               Unknown  Unknown
libc.so.6          00007FBCD2589F2A  Unknown               Unknown  Unknown
elk                0000000000404A6A  Unknown               Unknown  Unknown

with elk 4 it works fine. Am I doing something wrong with my libraries during compilation? I tried to use gfortran via the ./setup routine and recompile elk. In this case I get

Backtrace for this error:
#0  0x7f351bb02f3a in ???
#1  0x7f351bb02163 in ???
#2  0x7f351ab776df in ???
#3  0x4085ee in ???
#4  0x47e6a2 in ???
#5  0x4eeb2b in ???
#6  0x4545e3 in ???
#7  0x4089dd in ???
#8  0x40226c in ???
#9  0x7f351ab63f29 in ???
#10  0x4022b9 in ???
#11  0xffffffffffffffff in ???
Segmentation fault (core dumped)

while if I use ifort via setup created make.inc I have to change -qopenmp to -openmp due to my 2015 ifort compiler, but after compilation the run give an error again

Elk code version 5.2.11 started
Info(genspecies): running Z =    1, (hydrogen)
forrtl: severe (174): SIGSEGV, segmentation fault occurred
Image              PC                Routine            Line        Source             
elk                0000000000822299  Unknown               Unknown  Unknown
elk                0000000000820B6E  Unknown               Unknown  Unknown
elk                00000000007C79B2  Unknown               Unknown  Unknown
elk                00000000007678C8  Unknown               Unknown  Unknown
elk                000000000076C2EB  Unknown               Unknown  Unknown
libpthread.so.0    00007FE7107841B0  Unknown               Unknown  Unknown
elk                00000000004095F9  Unknown               Unknown  Unknown
elk                0000000000479285  Unknown               Unknown  Unknown
elk                00000000004DFD20  Unknown               Unknown  Unknown
elk                000000000044C967  Unknown               Unknown  Unknown
elk                0000000000409B70  Unknown               Unknown  Unknown
elk                00000000004037C6  Unknown               Unknown  Unknown
libc.so.6          00007FE7103DCF2A  Unknown               Unknown  Unknown
elk                00000000004036CA  Unknown               Unknown  Unknown

For now I simply use elk 4 to create modified .in files so its not so serious, but I would like to retire elk 4 soon.

best,
Eike

[martin_frbg]

Could you recompile with -g to get a more meaningful backtrace ?

[Eike Schwier]

Hi Martin,

I recompiled using the setup provided make.inc for gfortran adding the -g option:

In an unrelated matter the gfortran seems to find errors of conversion I missed to recognize previously (errors are not there with ifort)

Warning: Type mismatch in argument ‘cc’ at (1); passed COMPLEX(8) to REAL(8) [-Wargument-mismatch]
cfftnd.f90:472:72:

Warning: Type mismatch in argument ‘zv’ at (1); passed COMPLEX(8) to REAL(8) [-Wargument-mismatch]
oepresk.f90:92:65:

Now for the output of the execution in the speciels folder:

Elk code version 5.2.11 started
Info(genspecies): running Z =    1, (hydrogen)

Program received signal SIGSEGV: Segmentation fault - invalid memory reference.

Backtrace for this error:
#0  0x7f96869cef3a in ???
#1  0x7f96869ce163 in ???
#2  0x7f9685a436df in ???
#3  0x4085ee in __modomp_MOD_omp_hold
    at /home/calc/data/xChange/elk-5.2.11/src/modomp.f90:125
#4  0x47e6a2 in atom_
    at /home/calc/data/xChange/elk-5.2.11/src/atom.f90:107
#5  0x4eeb2b in genspecies_
    at /home/calc/data/xChange/elk-5.2.11/src/genspecies.f90:79
#6  0x4545e3 in readinput_
    at /home/calc/data/xChange/elk-5.2.11/src/readinput.f90:348
#7  0x4089dd in elk
    at /home/calc/data/xChange/elk-5.2.11/src/elk.f90:37
#8  0x40226c in main
    at /home/calc/data/xChange/elk-5.2.11/src/elk.f90:8
Segmentation fault (core dumped)

best,
Eike

[J. K. Dewhurst]

Hi Eike and Martin,

The problem has been fixed in version 5.2.14. It was due to an uninitialised OpenMP variable.

Regards,
Kay.