QCG: Conda CREST fails

[yanfeiguan]

Hi, just want to let you know CREST we got through CONDA fails for QCG computation with error:

`Note: The following floating-point exceptions are signalling: IEEE_INVALID_FLAG IEEE_UNDERFLOW_FLAG IEEE_DENORMAL ERROR STOP error while reading coord line

Error termination. Backtrace:

0 0x55555568f286 in ???

1 0x55555569213e in ???

2 0x55555557bb73 in ???

3 0x555555585e0f in ???

4 0x5555556c6c70 in ???

5 0x55555555f340 in ???

6 0x2aaaad2ce554 in ???

7 0x55555555f370 in ???"`

while the binary crest from github realse works.

[pprcht]

Thanks. The conda release hasn't been updated in some time, I will probably do that soon.

[jule-c]

Hi, for me conda crest fails on xTB optimised structures with the error that it fails in xTB optimisation (did not try binaries, yet):

xTB Geometry Optimization Initial geometry optimization failed! Please check your input. Error termination. Backtrace:

0 0x10d4384dc

1 0x10d439395

2 0x10d43ab04

3 0x1041bd959

4 0x1042c5008

5 0x1042c577e

As far as I can tell the input looks fine and is not broken (otherwise I guess xtb would have thrown an error while optimising it before). I attach an example structure here and would really appreciate your help! Thank you!

xtb_conformer_0.xyz.zip

[cplett]

Hi, This will probably be fixed with the next update of the conda release.

[pprcht]

I just checked it compared to the binary compiled from source where it works with the input you provided and the latest xtb release (6.5.1). So it's probably the faulty conda version.

[hpatel1567]

Hi, i compiled xtb and crest from source with gcc compiler using meson and ninja. Crest version 2.12. Im getting this error ONLY for positive charged solutes.

crest ../li.xyz --qcg ../h2o.xyz --nsolv 20 --alpb water --chrg 1 --gsolv --nopreopt

[cplett]

Hi, With my gcc-compiled versions, the example seems to work. Could you please tell me, which version you use for compilation so that I can check on this? Different versions might cause a problem. Moreover, it seems that the docking of the solvent to the solute does not work correctly. Could you please check the xtb_dock.out file in the qcg_tmp/tmp_grow directory if there is an error?

[hpatel1567]

Hi, thank you for your reply. I am using gcc 11.2.0. Here is thextb_dock.out file:

[cplett]

Hi, After several attempts to reproduce the error, I can tell that something is wrong when a single lithium atom is used in xtb, which causes a follow-up error in the aISS docking procedure. When using the intel icc compiler, this does not lead to a fatal error, but compiling xtb with the gcc compiler does so. I will work on a bug fix and inform you if I find a solution. For now, you might try to use the xtbiff binary (https://github.com/grimme-lab/xtbiff) by adding the --xtbiff flag to the qcg command line call or try the intel icc compiler instead of gcc.

[cplett]

As we fixed the continuous delivery for xtb recently, it is possible to use an xtb binary compiled with the intel icc compiler from the current source code. You can find it at https://github.com/grimme-lab/xtb/actions/runs/3913013270 at the bottom of the page.

[hpatel1567]

Hi, sadly my hpc system does not have intel icc compiler. I am limited to using gcc. Anyway to get it working with gcc? Thank you for all your help so far!

[cplett]

Hi, did you try the latest bleeding-edge build of xtb (xtb-bleed.tar.xz in the assets at https://github.com/grimme-lab/xtb/releases)? They are built statically with the intel compilers so that you do not have to have the intel compilers to run the program. I guess that this will work for your example.

[hpatel1567]

Hi, the bleeding edge build works for the cations, thank you very much for providing it!!

[github-actions[bot]]

This issue had no activity for 6 months. It will be closed in 1 week unless there is some new activity.