Unfamiliar error during electron-phonon calculation

[mdogan]

Hello,

I am having trouble with an EPW run using the version that came with QE-6.6. I'm puzzled because I was able to complete this calculation with QE-6.4.1 earlier. Also, I ran the same calculation in a related system (the same system under a slightly different pressure) without a problem. The error message says:

Code: Select all

forrtl: severe (174): SIGSEGV, segmentation fault occurred
Image              PC                Routine            Line        Source
epw.x              000000000104942A  Unknown               Unknown  Unknown
libpthread-2.28.s  0000151475BC4DD0  Unknown               Unknown  Unknown
epw.x              00000000008C4B91  init_us_2_                157  init_us_2.f90
epw.x              0000000000600780  elphel2_shuffle_          430  elphel2_shuffle.f90
epw.x              00000000005DA8B4  elphon_shuffle_           149  elphon_shuffle.f90
epw.x              0000000000434438  elphon_shuffle_wr         722  elphon_shuffle_wrap.f90
epw.x              0000000000407CB9  MAIN__                    145  epw.f90
epw.x              0000000000407062  Unknown               Unknown  Unknown
libc-2.28.so       00001514754916A3  __libc_start_main     Unknown  Unknown
epw.x              0000000000406F6E  Unknown               Unknown  Unknown

I've put all the relevant files here: https://www.dropbox.com/sh/gdpmevzdd9ct ... wDUpa?dl=0

Thank you all for any suggestions in advance!

Best,
Mehmet

[hlee]

Dear Mehmet:

It seems that you encounter the out-of-bound error and it can be due to the problem in EPW or QE.
Unfortunately, it is not easy for me to solve your issue since it requires some debugging by printing relevant quantities to see exactly what happens.

Well, is it necessary to manually specify the number of grids like "nr1 = 20, nr2 = 20, nr3 = 20"?
If it is just due to the fractional translation in your system, I think that it is not needed in the recent version of QE.

Sincerely,

H. Lee

[mdogan]

Dear H. Lee,

Thank you for your response! I specified the nr variables in the scf calculation to fix an error I was receiving with EPW in QE-6.4.1, so you may be correct that it may not be needed in QE-6.6. I'd like to avoid changing the scf input at this stage, however, so I'm first going to try running this calculation with QE-6.7. If that doesn't work, I'll go back and remove the nr1, nr2, nr3 specifications and retry.

Best,
Mehmet

[mdogan]

Dear H. Lee,

I tried both of your suggestions (retrying with QE 6.7 and removing the nr1, nr2, nr3 variables) but the calculation fails at the exact same point. Because these are crucial calculations for me, I'm happy to work on debugging by modifying the code and printing extra things if you could instruct me on what to modify.

FYI, I have two other related systems (just different pressure); in one of them the calculation runs without a problem, and in two of them the calculation fails at different points. I put all of these calculation files in the original Dropbox folder [https://www.dropbox.com/sh/gdpmevzdd9ct ... wDUpa?dl=0].

Thank you very much for your help!

Best,
Mehmet

[hlee]

Dear Mehmet:

The first thing you can try is to switch to gcc programming environment, to configure QE with the debug mode enabled by issuing "configure --enable-debug", and to compile the code.

In this way, (I am not 100% sure) you might get the more information on your error (I want to confirm that you encountered the out-of-bound error).

Sincerly,

H. Lee

[mdogan]

Dear H. Lee,

I believe I managed to run the code with debugging turned on, and here is the error message I received:

Code: Select all

forrtl: severe (408): fort: (3): Subscript #2 of the array G has value 0 which is less than the lower bound of 1

Image              PC                Routine            Line        Source
epw.x              00000000015F8F5F  Unknown               Unknown  Unknown
epw.x              0000000000BA4D21  init_us_2_                 86  init_us_2.f90
epw.x              000000000079226C  elphel2_shuffle_          430  elphel2_shuffle.f90
epw.x              0000000000741BC0  elphon_shuffle_           149  elphon_shuffle.f90
epw.x              000000000044FA8E  elphon_shuffle_wr         717  elphon_shuffle_wrap.f90
epw.x              0000000000408E33  MAIN__                    145  epw.f90
epw.x              00000000004081D2  Unknown               Unknown  Unknown
libc-2.17.so       00002BA73B248555  __libc_start_main     Unknown  Unknown
epw.x              00000000004080E9  Unknown               Unknown  Unknown

So it looks like you were right, that it was the out-of-bound error. Please let me know what the next steps might be in fixing this error. As always, I greatly appreciate your help.

Best,
Mehmet

[hlee]

Dear Mehmet:

Could you show me (1) the lines around the 86th line of init_us_2.f90 and (2) the lines around the 430th line of elphel2_shuffle.f90?

Sincerely,

H. Lee

[mdogan]

Dear H. Lee,

I apologize for the late response. I was away from this project due to some unforeseen problems, but now I've fully returned to it.

I put all the f90 files involved in the error report into the Dropbox folder [https://www.dropbox.com/sh/gdpmevzdd9ct ... VZG6KwDUpa] in the H.I41amd2.P450.novdW.qe6.h1 subfolder. I also copy the lines you asked for below. Thank you very much for your help!

Best,
Mehmet

init_us_2.f90, line 80 - 90:

Code: Select all

  DO iblock = 1, numblock
     !
     realblocksize = MIN(npw_-(iblock-1)*blocksize,blocksize)
     !
     DO ig = 1, realblocksize
        ig_orig = (iblock-1)*blocksize+ig
        gk(1,ig) = q_(1) + g(1,igk_(ig_orig) )
        gk(2,ig) = q_(2) + g(2,igk_(ig_orig) )
        gk(3,ig) = q_(3) + g(3,igk_(ig_orig) )
        qg(ig) = gk(1, ig)**2 +  gk(2, ig)**2 + gk(3, ig)**2
     ENDDO

elphel2_shuffle.f90, line 425 - 435:

Code: Select all

    !
    ! now we generate vkb on the igkq() set because dvpsi is needed on that set                     
    ! we need S(k)+q_0 in the KB projector: total momentum transfer must be q_0                     
    !
    xkqtmp = sxk + xq0
    CALL init_us_2(npwq, igkq, xkqtmp, vkb)
    !
    ! --------------------------------------------------                                            
    !   Calculation of the matrix element                                                           
    ! --------------------------------------------------                                            
    !

[hlee]

Dear Mehmet:

Could you add to the subfolders in Dropbox the corresponding prefix.kgmap files?

Sincerely,

H. Lee

[mdogan]

Dear H. Lee,

I just put the files in their respective folders. Let me know what you think. Thank you!

Best,
Mehmet