Re: [Wien] segfault in mixer
On 11/11/2015 03:40 PM, Laurence Marks wrote: > There is some code inside mixer.F which reduces the number of PW's to > only those which are non-zero. With your clmval this is zero, so the > array kzz in setn probably has a size of (3,0) which is zero. A zero > size array will lead to a SIGSEGV, I suspect that ifort has decided that > line 27 is where it is going to prefetch the first values of kzz (i.e. > kzz(1,2) at line 30). > > Something went wrong earlier either (or both) in lapw1 & lapw2. I think you are right, and I think I localized the problem. EMIN in case.in2 was set too large (0.53 to be precise). Thus in case.output2 the band energies were reasonable, but the occupations were all zero. I have to wait for my batch job to start, but this looks promising. The EMIN was probably a leftover from a charge distribution calculation. Thanks for the help! Elias -- Elias Assmann Institute of Theoretical and Computational Physics TU Graz ⟨https://itp.tugraz.at/⟩ signature.asc Description: OpenPGP digital signature ___ Wien mailing list Wien@zeus.theochem.tuwien.ac.at http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien SEARCH the MAILING-LIST at: http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html
Re: [Wien] segfault in mixer
-BEGIN PGP SIGNED MESSAGE- Hash: SHA1 On 11/11/2015 04:04 PM, Zhu, Jianxin wrote: > I am curios. How come the value emin becomes so big? It is > automatically set, no. I think I set it when I was calculating a charge density (to get only the “valence” density). When I picked the calculation up again, I forgot to change it back. Elias > The EMIN was probably a leftover from a charge distribution > calculation. -BEGIN PGP SIGNATURE- Version: GnuPG v1 Comment: Using GnuPG with Icedove - http://www.enigmail.net/ iQIcBAEBAgAGBQJWQ1s6AAoJEE/4gtQZfOqPKXUQAKF8yBlnXybrBcqOFkO5swxU YoVvRaSHXFjvqoKVvdRAQyX6XyCQq+croO0TeMllMms2jGQ75Orwt1hPj0mY5fgS pJ/8TnXDqWM7kI9fCbb/ImlXnVNazzLN4NfHKXXAK9Zh5SF/bg2JY1sWb1I3NQVD ifvTCr8wkSkPx4w4OgHAB11SbXYQUvx3l4U2KQ42DNVRUvNpI1+SQeLOg8iTmeuJ OGeH2XDHerfEtRL4inY67hK7y/MK8QXvo9mGRtFoxOg9CvLXkPt4kBY+E7yDRcLW tYF2n2BjhHd12vtaSwy+ZEppACbEspUMrkoM5lKRcu2V/fUtYYfXNrnpuqAd4Byj RGnfDSSZK+BqfvineuhZx1VxV7lCw9A4WFClV0u5UNkmqFHDpEA2uM45WAPyi9Fo KIamOaVKDtrOmKU/FhnbRZHQ3JbOWUVphP4QdHyxBkC/a+n95lb5iqKHP21YUtbp v4q4O5ZUQW4DBNJLeuzGwNMbnyWYap1E1oFnrdG5Mqvqtmk6J+nin1BXA00qmmty qtaXpuOrpk111OAKNYoWwfApatomD90f/gZ9tC901AdmUbbpjpjiVZYGkpQxdIPD jz4WoRhYd95y0/r/s/Xy0Vi7yaAFxonAeEgN98Ybea181PXshVAWgU2O8BEpG35u 0xTwhiZEaRzAEnXV2xKd =ZsIi -END PGP SIGNATURE- ___ Wien mailing list Wien@zeus.theochem.tuwien.ac.at http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien SEARCH the MAILING-LIST at: http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html
Re: [Wien] segfault in mixer
I am curios. How come the value emin becomes so big? It is automatically set, no. Jianxin Sent from my BlackBerry 10 smartphone on the Verizon Wireless 4G LTE network. Original Message From: Elias Assmann Sent: Wednesday, November 11, 2015 8:53 AM To: A Mailing list for WIEN2k users Reply To: A Mailing list for WIEN2k users Subject: Re: [Wien] segfault in mixer On 11/11/2015 03:40 PM, Laurence Marks wrote: > There is some code inside mixer.F which reduces the number of PW's to > only those which are non-zero. With your clmval this is zero, so the > array kzz in setn probably has a size of (3,0) which is zero. A zero > size array will lead to a SIGSEGV, I suspect that ifort has decided that > line 27 is where it is going to prefetch the first values of kzz (i.e. > kzz(1,2) at line 30). > > Something went wrong earlier either (or both) in lapw1 & lapw2. I think you are right, and I think I localized the problem. EMIN in case.in2 was set too large (0.53 to be precise). Thus in case.output2 the band energies were reasonable, but the occupations were all zero. I have to wait for my batch job to start, but this looks promising. The EMIN was probably a leftover from a charge distribution calculation. Thanks for the help! Elias -- Elias Assmann Institute of Theoretical and Computational Physics TU Graz ⟨https://itp.tugraz.at/⟩ ___ Wien mailing list Wien@zeus.theochem.tuwien.ac.at http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien SEARCH the MAILING-LIST at: http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html
Re: [Wien] segfault in mixer
Never had such a problem. Anything in case.outputm ? setn has not much to do with the actual clmsum/val file input, except when the number of PW is zero (check out clmval and clmsum and clmsum_old files, if the K-lists are ok.), as this fft array gets allocated with nkknew1 Otherwise more with the struct file. Is this still ok ? Can you run x nnwith this struct file ? On 11/11/2015 02:43 PM, Elias Assmann wrote: Hi List, In a fairly large calculation (200 atoms) I am running I get a segfault in mixer: $ x mixer forrtl: severe (174): SIGSEGV, segmentation fault occurred Image RoutineLineSource mixer Unknown Unknown Unknown mixer Unknown Unknown Unknown mixer Unknown Unknown Unknown mixer Unknown Unknown Unknown mixer Unknown Unknown Unknown libpthread.so.0 Unknown Unknown Unknown mixer setn_ 27 setn.f mixer MAIN__ 1232 mixer.F mixer Unknown Unknown Unknown libc.so.6 Unknown Unknown Unknown mixer Unknown Unknown Unknown Line 27 in setn.f is just FFT(1)=U0 the ‘FFT’ array corresponds to ‘cfft’ in mixer.F, and indeed $ tail STO6LVO2_ovac.outputm … Allocating CLM-arrays 678 MB Allocating nwav-arrays 57 MB Allocating cfft-array0 MB Any idea where this might come from? I strongly suspect some error in the input because another, similar, calculation is running without this problem; but I have no idea what I might have changed to cause the error. Here is my case.inm: MSR1 0.0 YES (BROYD/PRATT, extra charge (+1 for additional e), norm) 0.20mixing FACTOR for BROYD/PRATT scheme 1.00 1.00 PW and CLM-scaling factors 8 idum, HISTORY The error keeps happening even after I removed case.broyd*. Elias ___ Wien mailing list Wien@zeus.theochem.tuwien.ac.at http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien SEARCH the MAILING-LIST at: http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html -- P.Blaha -- Peter BLAHA, Inst.f. Materials Chemistry, TU Vienna, A-1060 Vienna Phone: +43-1-58801-165300 FAX: +43-1-58801-165982 Email: bl...@theochem.tuwien.ac.atWIEN2k: http://www.wien2k.at WWW: http://www.imc.tuwien.ac.at/staff/tc_group_e.php -- ___ Wien mailing list Wien@zeus.theochem.tuwien.ac.at http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien SEARCH the MAILING-LIST at: http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html
Re: [Wien] segfault in mixer
On 11/11/2015 02:53 PM, Peter Blaha wrote: > Anything in case.outputm ? No warnings or errors. Apart from the size information I quoted, the only “suspicious” thing is the first line filename of case.inc: case.incup where case.incup is empty. But I checked other calculations and case.incup seems to be empty always. > setn has not much to do with the actual clmsum/val file input, except > when the number of PW is zero (check out clmval and clmsum and > clmsum_old files, if the K-lists are ok.), as this fft array gets > allocated with nkknew1 The clmval files are zero (the whole files), but there are definitely plane waves: $ grep -e PW *.clmvalup -A10 594657 NUMBER OF PW 000 0.E+00 0.E+00 00 -1 0.E+00 0.E+00 001 0.E+00 0.E+00 00 -2 0.E+00 0.E+00 002 0.E+00 0.E+00 00 -3 0.E+00 0.E+00 003 0.E+00 0.E+00 00 -4 0.E+00 0.E+00 004 0.E+00 0.E+00 00 -5 0.E+00 0.E+00 $ grep -e PW *.clmsum -A10 594657 NUMBER OF PW Change Residue 000 5.630359813726E-02 0.E+00 00 -1-1.150778984163E-02 8.051431134897E-03 001-1.150778984163E-02-8.051431134897E-03 00 -2-9.470078634937E-04 1.160395119838E-02 002-9.470078634937E-04-1.160395119838E-02 00 -3 3.822712367509E-03 5.400713534390E-03 003 3.822712367509E-03-5.400713534390E-03 00 -4 2.813831811510E-03 4.780854323303E-04 004 2.813831811510E-03-4.780854323303E-04 00 -5-1.336690365246E-03 4.849945762786E-04 > Otherwise more with the struct file. > > Is this still ok ? Can you run x nnwith this struct file ? ‘x nn’ is fine, in fact, the whole SCF cycle is fine up to mixer. I checked all files in the directory for NaN's, there are none. Elias -- Elias Assmann Institute of Theoretical and Computational Physics TU Graz ⟨https://itp.tugraz.at/⟩ signature.asc Description: OpenPGP digital signature ___ Wien mailing list Wien@zeus.theochem.tuwien.ac.at http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien SEARCH the MAILING-LIST at: http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html
Re: [Wien] segfault in mixer
I agree with Peter. A problem with SIGSEGV is that the line it indicates is often not where the problem actually is. My guess would be that there is something wrong with the struct format, and/or you have no PW's. For the later try grep -e PW case.clmval(up/dn) -A10 (change case etc) and see if there are NAN's present. On Wed, Nov 11, 2015 at 7:53 AM, Peter Blahawrote: > Never had such a problem. > > Anything in case.outputm ? > > setn has not much to do with the actual clmsum/val file input, except > when the number of PW is zero (check out clmval and clmsum and > clmsum_old files, if the K-lists are ok.), as this fft array gets > allocated with nkknew1 > > Otherwise more with the struct file. > > Is this still ok ? Can you run x nnwith this struct file ? > > > > On 11/11/2015 02:43 PM, Elias Assmann wrote: > > Hi List, > > > > In a fairly large calculation (200 atoms) I am running I get a segfault > > in mixer: > > > > $ x mixer > > forrtl: severe (174): SIGSEGV, segmentation fault occurred > > Image RoutineLineSource > > mixer Unknown Unknown Unknown > > mixer Unknown Unknown Unknown > > mixer Unknown Unknown Unknown > > mixer Unknown Unknown Unknown > > mixer Unknown Unknown Unknown > > libpthread.so.0 Unknown Unknown Unknown > > mixer setn_ 27 setn.f > > mixer MAIN__ 1232 mixer.F > > mixer Unknown Unknown Unknown > > libc.so.6 Unknown Unknown Unknown > > mixer Unknown Unknown Unknown > > > > Line 27 in setn.f is just > > > >FFT(1)=U0 > > > > the ‘FFT’ array corresponds to ‘cfft’ in mixer.F, and indeed > > > > $ tail STO6LVO2_ovac.outputm > > … > > Allocating CLM-arrays 678 MB > > Allocating nwav-arrays 57 MB > > Allocating cfft-array0 MB > > > > Any idea where this might come from? I strongly suspect some error in > > the input because another, similar, calculation is running without this > > problem; but I have no idea what I might have changed to cause the error. > > > > Here is my case.inm: > > > > MSR1 0.0 YES (BROYD/PRATT, extra charge (+1 for additional e), norm) > > 0.20mixing FACTOR for BROYD/PRATT scheme > > 1.00 1.00 PW and CLM-scaling factors > > 8 idum, HISTORY > > > > The error keeps happening even after I removed case.broyd*. > > > > > > Elias > > > > > > > > > > ___ > > Wien mailing list > > Wien@zeus.theochem.tuwien.ac.at > > http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien > > SEARCH the MAILING-LIST at: > http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html > > > > -- > >P.Blaha > -- > Peter BLAHA, Inst.f. Materials Chemistry, TU Vienna, A-1060 Vienna > Phone: +43-1-58801-165300 FAX: +43-1-58801-165982 > Email: bl...@theochem.tuwien.ac.atWIEN2k: http://www.wien2k.at > WWW: http://www.imc.tuwien.ac.at/staff/tc_group_e.php > -- > ___ > Wien mailing list > Wien@zeus.theochem.tuwien.ac.at > http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien > SEARCH the MAILING-LIST at: > http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html > -- Professor Laurence Marks Department of Materials Science and Engineering Northwestern University www.numis.northwestern.edu Corrosion in 4D: MURI4D.numis.northwestern.edu Co-Editor, Acta Cryst A "Research is to see what everybody else has seen, and to think what nobody else has thought" Albert Szent-Gyorgi ___ Wien mailing list Wien@zeus.theochem.tuwien.ac.at http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien SEARCH the MAILING-LIST at: http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html
[Wien] segfault in mixer
Hi List, In a fairly large calculation (200 atoms) I am running I get a segfault in mixer: $ x mixer forrtl: severe (174): SIGSEGV, segmentation fault occurred Image RoutineLineSource mixer Unknown Unknown Unknown mixer Unknown Unknown Unknown mixer Unknown Unknown Unknown mixer Unknown Unknown Unknown mixer Unknown Unknown Unknown libpthread.so.0 Unknown Unknown Unknown mixer setn_ 27 setn.f mixer MAIN__ 1232 mixer.F mixer Unknown Unknown Unknown libc.so.6 Unknown Unknown Unknown mixer Unknown Unknown Unknown Line 27 in setn.f is just FFT(1)=U0 the ‘FFT’ array corresponds to ‘cfft’ in mixer.F, and indeed $ tail STO6LVO2_ovac.outputm … Allocating CLM-arrays 678 MB Allocating nwav-arrays 57 MB Allocating cfft-array0 MB Any idea where this might come from? I strongly suspect some error in the input because another, similar, calculation is running without this problem; but I have no idea what I might have changed to cause the error. Here is my case.inm: MSR1 0.0 YES (BROYD/PRATT, extra charge (+1 for additional e), norm) 0.20mixing FACTOR for BROYD/PRATT scheme 1.00 1.00 PW and CLM-scaling factors 8 idum, HISTORY The error keeps happening even after I removed case.broyd*. Elias -- Elias Assmann Institute of Theoretical and Computational Physics TU Graz ⟨https://itp.tugraz.at/⟩ signature.asc Description: OpenPGP digital signature ___ Wien mailing list Wien@zeus.theochem.tuwien.ac.at http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien SEARCH the MAILING-LIST at: http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html
Re: [Wien] segfault in mixer
Aha! There is some code inside mixer.F which reduces the number of PW's to only those which are non-zero. With your clmval this is zero, so the array kzz in setn probably has a size of (3,0) which is zero. A zero size array will lead to a SIGSEGV, I suspect that ifort has decided that line 27 is where it is going to prefetch the first values of kzz (i.e. kzz(1,2) at line 30). Something went wrong earlier either (or both) in lapw1 & lapw2. N.B., I will try and remember to add a trap for this in mixer, but it is not really a fault in that code. On Wed, Nov 11, 2015 at 8:20 AM, Elias Assmannwrote: > On 11/11/2015 02:53 PM, Peter Blaha wrote: > > Anything in case.outputm ? > > No warnings or errors. Apart from the size information I quoted, the > only “suspicious” thing is the first line > > filename of case.inc: case.incup > > where case.incup is empty. But I checked other calculations and > case.incup seems to be empty always. > > > setn has not much to do with the actual clmsum/val file input, except > > when the number of PW is zero (check out clmval and clmsum and > > clmsum_old files, if the K-lists are ok.), as this fft array gets > > allocated with nkknew1 > > The clmval files are zero (the whole files), but there are definitely > plane waves: > > $ grep -e PW *.clmvalup -A10 > 594657 NUMBER OF PW >000 0.E+00 0.E+00 >00 -1 0.E+00 0.E+00 >001 0.E+00 0.E+00 >00 -2 0.E+00 0.E+00 >002 0.E+00 0.E+00 >00 -3 0.E+00 0.E+00 >003 0.E+00 0.E+00 >00 -4 0.E+00 0.E+00 >004 0.E+00 0.E+00 >00 -5 0.E+00 0.E+00 > > $ grep -e PW *.clmsum -A10 > 594657 NUMBER OF PW Change > Residue >000 5.630359813726E-02 0.E+00 >00 -1-1.150778984163E-02 8.051431134897E-03 >001-1.150778984163E-02-8.051431134897E-03 >00 -2-9.470078634937E-04 1.160395119838E-02 >002-9.470078634937E-04-1.160395119838E-02 >00 -3 3.822712367509E-03 5.400713534390E-03 >003 3.822712367509E-03-5.400713534390E-03 >00 -4 2.813831811510E-03 4.780854323303E-04 >004 2.813831811510E-03-4.780854323303E-04 >00 -5-1.336690365246E-03 4.849945762786E-04 > > > Otherwise more with the struct file. > > > > Is this still ok ? Can you run x nnwith this struct file ? > > ‘x nn’ is fine, in fact, the whole SCF cycle is fine up to mixer. > > I checked all files in the directory for NaN's, there are none. > > > Elias > > > -- > Elias Assmann > Institute of Theoretical and Computational Physics > TU Graz ⟨https://itp.tugraz.at/⟩ > > -- Professor Laurence Marks Department of Materials Science and Engineering Northwestern University www.numis.northwestern.edu Corrosion in 4D: MURI4D.numis.northwestern.edu Co-Editor, Acta Cryst A "Research is to see what everybody else has seen, and to think what nobody else has thought" Albert Szent-Gyorgi ___ Wien mailing list Wien@zeus.theochem.tuwien.ac.at http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien SEARCH the MAILING-LIST at: http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html