Description of problem: RT-TDDFT fails under Fedora 28 with NWChem throwing an error message that suggests that it is GA's fault. Version-Release number of selected component (if applicable): ga-openmpi-5.6.5-1.fc28.x86_64 ga-mpich-5.6.5-1.fc28.x86_64 ga-common-5.6.5-1.fc28.noarch On (nwchem-6.8.1-4.fc28.x86_64 nwchem-openmpi-6.8.1-4.fc28.x86_64 nwchem-mpich-6.8.1-4.fc28.x86_64 nwchem-common-6.8.1-4.fc28.noarch) How reproducible: Every time it fails Steps to Reproduce: 1. Try to run this input: title "Water TD-PBE0 absorption spectrum" echo #scratch_dir ./scratch #permanent_dir ./perm start water ## ## aug-cc-pvtz / pbe0 optimized ## ## Note: you are required to explicitly name the geometry ## geometry "system" units angstroms nocenter noautoz noautosym O 0.00000000 -0.00001441 -0.34824012 H -0.00000000 0.76001092 -0.93285191 H 0.00000000 -0.75999650 -0.93290797 end ## Note: We need to explicitly set the "active" geometry even though there is only one geom. set geometry "system" ## All DFT and basis parameters are inherited by the RT-TDDFT code basis * library 6-31G end dft xc pbe0 end ## Compute ground state of the system task dft energy ## ## Now, we compute an x, y, and z kick simulation, which we give separate "tags" for post-processing. ## unset rt_tddft:* rt_tddft tmax 200.0 dt 0.2 tag "kick_x" field "kick" type delta polarization x max 0.0001 end excite "system" with "kick" end task dft rt_tddft unset rt_tddft:* rt_tddft tmax 200.0 dt 0.2 tag "kick_y" field "kick" type delta polarization y max 0.0001 end excite "system" with "kick" end task dft rt_tddft unset rt_tddft:* rt_tddft tmax 200.0 dt 0.2 tag "kick_z" field "kick" type delta polarization z max 0.0001 end excite "system" with "kick" end task dft rt_tddft 2. nwchem_openmpi example.nw Actual results: ************************* * Closed-shell RT-TDDFT * ************************* ****************************** WARNING ****************************** Schwartz screening tolerance is similar to the Magnus interpolation tolerance--this may cause problems with convergence. If so, try loosening tol_interpol. ********************************************************************* Geometry Atoms Basis func. Nuc. charge Nuc. dip. mom. ----------------------------------------------------------------------------------------- *system 3 (100% ) 13 (100% ) 10.00 0.00 -0.00 -8.79 0:0:ga_diag_std_seq: dsyev failed:: 0 (rank:0 hostname:localhost.localdomain pid:7252):ARMCI DASSERT fail. src/common/armci.c:ARMCI_Error():208 cond:0 Expected results: On Ubuntu (but with NWCHem 6.6 instead of 6.8) the calculation finishes and it also used to finish on Fedora 27/26 Additional info:
Edoardo may be able to help in debugging this.
Sure I will have a look Edo
Marcin Please propagate the following commits from my github repository to fix this (known) bug in ga_diag_std_seq https://github.com/edoapra/fedpkg/commit/b5e0c50258cd448b85b146a3752e616ef433946a https://github.com/edoapra/fedpkg/commit/88cd7a58f29555e2c1f04307b8a7112b7f6681be We might have to get rid of the patch once I will push the same fix to the GlobalArrays repository
ga-5.6.5-3.fc28 has been submitted as an update to Fedora 28. https://bodhi.fedoraproject.org/updates/FEDORA-2018-d0f6b9b482
ga-5.6.5-3.fc28 has been pushed to the Fedora 28 testing repository. If problems still persist, please make note of it in this bug report. See https://fedoraproject.org/wiki/QA:Updates_Testing for instructions on how to install test updates. You can provide feedback for this update here: https://bodhi.fedoraproject.org/updates/FEDORA-2018-d0f6b9b482
ga-5.6.5-3.fc28 has been pushed to the Fedora 28 stable repository. If problems still persist, please make note of it in this bug report.