Multiwfn forum

i.s.ger

Member

Registered: 2020-12-01

Posts: 55

Defensive compilation of O1.f90

Dear Tian Lu,

As I can see, O1.f90 has defensive compilation due to a possible compiler bug. How can I repeat your observation of zero or NaN?

Best regards,
Igor

sobereva

Tian Lu (Multiwfn developer)

From: Beijing

Registered: 2017-09-11

Posts: 1,768

Website

Re: Defensive compilation of O1.f90

Dear Igor,

In Makefile, if "OPT1 = -O1" is replaced with "OPT1 = -O2", then after loading a wavefunction file and then run
200
11
1
The result will be mostly zero or NaN, for example: (The ifort version is 19.0.1.144)

Multiwfn -- A Multifunctional Wavefunction Analyzer Version 3.8(dev), release date: 2022-May-9 Developer: Tian Lu (Beijing Kein Research Center for Natural Sciences) Below paper ***MUST BE CITED*** if Multiwfn is utilized in your work: Tian Lu, Feiwu Chen, J. Comput. Chem., 33, 580-592 (2012) See "How to cite Multiwfn.pdf" in Multiwfn binary package for more information Multiwfn official website: //www.umsyar.com/multiwfn Multiwfn English forum: //www.umsyar.com/wfnbbs Multiwfn Chinese forum: http://bbs.keinsci.com/wfn ( Number of parallel threads: 4 Current date: 2022-05-13 Time: 00:35:51 ) Please wait... Total energy: -2611.994301127532 Hartree, Virial ratio: 1.99994135 Total/Alpha/Beta electrons: 44.0000 22.0000 22.0000 Net charge: 0.00000 Expected multiplicity: 1 The number of orbitals: 22, Atoms: 5, GTFs: 145 This is a restricted closed-shell single-determinant wavefunction Title line of this file: opt hf/6-311+g(d) 10f 6d out=wfn Loaded CH3Br.wfn successfully! Formula: H3 C1 Br1 Molecule weight: 94.93809 Da Point group: C3v "q": Exit program gracefully "r": Load a new file ************ Main function menu ************ 0 Show molecular structure and view orbitals 1 Output all properties at a point 2 Topology analysis 3 Output and plot specific property in a line 4 Output and plot specific property in a plane 5 Output and plot specific property within a spatial region (calc. grid data) 6 Check & modify wavefunction 7 Population analysis and calculation of atomic charges 8 Orbital composition analysis 9 Bond order analysis 10 Plot total DOS, partial DOS, OPDOS, local DOS and photoelectron spectrum 11 Plot IR/Raman/UV-Vis/ECD/VCD/ROA/NMR spectrum 12 Quantitative analysis of molecular surface 13 Process grid data (No grid data is presented currently) 14 Adaptive natural density partitioning (AdNDP) analysis 15 Fuzzy atomic space analysis 16 Charge decomposition analysis (CDA) and plot orbital interaction diagram 17 Basin analysis 18 Electron excitation analysis 19 Orbital localization analysis 20 Visual study of weak interaction 21 Energy decomposition analysis 22 Conceptual DFT (CDFT) analysis 23 ETS-NOCV analysis 24 (Hyper)polarizability analysis 25 Electron delocalization and aromaticity analyses 100 Other functions (Part 1) 200 Other functions (Part 2) 300 Other functions (Part 3) 200 ============ Other functions (Part 2) ============ 0 Return 1 Calculate core-valence bifurcation (CVB) index and related quantities 2 Calculate atomic and bond dipole moments in Hilbert space 3 Generate cube file for multiple orbital wavefunctions 4 Generate iso-chemical shielding surfaces (ICSS) and related quantities 5 Plot radial distribution function for a real space function 6 Analyze correspondence between orbitals in two wavefunctions 7 Parse output of (hyper)polarizability task of Gaussian 8 Study (hyper)polarizability by sum-over-states (SOS) method 9 Calculate average bond length and average coordinate number 10 Output various kinds of integral between orbitals 11 Calculate center, first/second moments and radius of gyration of a function 12 Calculate energy index (EI) or bond polarity index (BPI) 13 Evaluate orbital contributions to density difference or other grid data 14 Domain analysis (Obtaining properties within isosurfaces of a function) 15 Calculate electron correlation index (PCCP, 18, 24015) 16 Generate natural orbitals based on the density matrix in .fch/.fchk file 17 Calculate Coulomb and exchange integrals between two orbitals 18 Calculate bond length/order alternation (BLA/BOA) 20 Bond order density (BOD) and natural adaptive orbital (NAdO) analyses 11 0 Return 1 Calculate various quantities of the selected function 2 Calculate center and integral of the selected function 3 Select the function to be studied, current: 1 4 Set the center for option 1, current: 0.00000 0.00000 0.00000 Ang 5 Calculate center and integral of the absolute of the selected function 1 Radial points: 75 Angular points: 434 Total: 32550 per center Processing center 1(C ) / 5 Processing center 2(H ) / 5 Processing center 3(H ) / 5 Processing center 4(H ) / 5 Processing center 5(Br) / 5 Calculation took up wall clock time 0 s Note: All data shown below are in a.u. Integral over whole space: 0.00000000E+00 The first moment: X= 0.00000000E+00 Y= 0.00000000E+00 Z= 0.00000000E+00 Norm= 0.00000000E+00 The second moment: XX= 0.00000000E+00 XY= 0.00000000E+00 XZ= 0.00000000E+00 YX= 0.00000000E+00 YY= 0.00000000E+00 YZ= 0.00000000E+00 ZX= 0.00000000E+00 ZY= 0.00000000E+00 ZZ= 0.00000000E+00 Eigenvalues: 0.00000000E+00 0.00000000E+00 0.00000000E+00 Sum of eigenvalues (trace of the second moment tensor): 0.00000000E+00 Anisotropy: 0.00000000E+00 Radius of gyration: NaN Spatial extent of the function : 0.000000 The quadrupole moment of nuclear charges: XX= 5.62036577 XY= 0.00000000 XZ= 0.00000000 YX= 0.00000000 YY= 5.62036573 YZ= 0.00000000 ZX= 0.00000000 ZY= 0.00000000 ZZ= 108.86830814 The quadrupole moment of the system: XX= 5.62036577 XY= 0.00000000 XZ= 0.00000000 YX= 0.00000000 YY= 5.62036573 YZ= 0.00000000 ZX= 0.00000000 ZY= 0.00000000 ZZ= 108.86830814

The result is normal if using -O1 in Linux or use -O2 in Windows for compiling O1.f90:

Integral over whole space: 4.39999897E+01 The first moment: X= -8.88178420E-15 Y= -6.86588534E-06 Z= 8.96717848E-01 Norm= 8.04102898E-01 The second moment: XX= 2.50666385E+01 XY= 7.32920669E-15 XZ= -1.50990331E-14 YX= 7.32920669E-15 YY= 2.50666353E+01 YZ= 1.39406886E-05 ZX= -1.50990331E-14 ZY= 1.39406886E-05 ZZ= 1.25287900E+02 Eigenvalues: 2.50666353E+01 2.50666385E+01 1.25287900E+02 Sum of eigenvalues (trace of the second moment tensor): 1.75421173E+02 Anisotropy: 1.00221263E+02 Radius of gyration: 1.99670874E+00 Spatial extent of the function : 175.421173 The quadrupole moment of nuclear charges: XX= 5.62036577 XY= 0.00000000 XZ= 0.00000000 YX= 0.00000000 YY= 5.62036573 YZ= 0.00000000 ZX= 0.00000000 ZY= 0.00000000 ZZ= 108.86830814 The quadrupole moment of the system: XX= -19.44627278 XY= -0.00000000 XZ= 0.00000000 YX= -0.00000000 YY= -19.44626957 YZ= -0.00000000 ZX= 0.00000000 ZY= -0.00000000 ZZ= -16.41959138

The reason is unclear to me.

Best regards,

Tian

i.s.ger

Member

Registered: 2020-12-01

Posts: 55

Re: Defensive compilation of O1.f90

Dear Tian Lu,

I found the problem that tmpval is over-optimized. Here, these is a patch that fixes this problem:0001-Remove-defensive-compilation.patch.txt

Sorry for my a little explicit code-style :-)
I OMP-parallelized the problem cycle and the problem is gone away :-)

Best regards,
Igor

sobereva

Tian Lu (Multiwfn developer)

From: Beijing

Registered: 2017-09-11

Posts: 1,768

Website

Re: Defensive compilation of O1.f90

Dear Igor,

Thank you very much for your modification, now this subroutine runs normally under Linux with -O2. I have applied the patch to official source code in the new version updated today. Although O1.f90 is empty now, considering that in the future it may still be used, I didn't delete it.

Best regards,

Tian