Re: [Pw_forum] A problem of parallel excuting pwcond.x
Dear Paolo Thanks for your advise. In fact I have examed line444, however, there is only a summation, and if I use other operation instead, wirte(*,*) for example, the error appeared at the same position, too. I have send a email to the cluster manager. If he find the reason, I will post it. Sincerely Pang Rui -- 庞瑞(PANG Rui) South University of Science and Technology of China/Department of Physics No.1088,Xueyuan Road, Shenzhen,Guangdong -- Original -- From: "Paolo Giannozzi"; Date: Sat, Aug 15, 2015 05:16 PM To: "PWSCF Forum"; Subject: Re: [Pw_forum] A problem of parallel excuting pwcond.x in theory, well-designed parallel code should work (or cleanly stop) on any number of processors. In practice, if you use a disproportionate number of processors, the code may not be able to deal with it. IN any case: look at what happens at line 444 of PWCOND/src/scatter_forw.f90 aqnd maybe you sill understand what is happening and why Paolo Paolo On Mon, Aug 10, 2015 at 9:27 AM, 庞瑞(PANG Rui) wrote: However, I found I could run the example correctly by using only 8 cores. If using 16, 32 and 64 cores, the code will stop at " ngper, shell number = 804 82 ngper, n2d = 804 391 --- E-Ef =2.000 k =0.500 0.500 --- ie = 1 ik = 1" and showed "forrtl: severe (174): SIGSEGV, segmentation fault occurred Image PCRoutineLineSource pwcond.x 025AD7C9 Unknown Unknown Unknown pwcond.x 025AC140 Unknown Unknown Unknown pwcond.x 0255DBC2 Unknown Unknown Unknown pwcond.x 024F15F3 Unknown Unknown Unknown pwcond.x 024F753B Unknown Unknown Unknown libpthread.so.0003F2280F710 Unknown Unknown Unknown pwcond.x 004F9B35 scatter_forw_ 444 scatter_forw. f90 pwcond.x 004ADD66 do_cond_ 518 do_cond.f90 pwcond.x 004A8B42 MAIN__ 22 condmain.f90 pwcond.x 00498036 Unknown Unknown Unknown libc.so.6 003F2241ED1D Unknown Unknown Unknown pwcond.x 00497F29 Unknown Unknown Unknown" So could anyone tell me how can I fix this error so that I can using more cores to excute pwcond.x? I met this problem in both 5.1.2 and 5.2. Thanks very much for any help. Sincerely. Pang Rui -- 庞瑞(PANG Rui) South University of Science and Technology of China/Department of Physics No.1088,Xueyuan Road, Shenzhen,Guangdong ___ Pw_forum mailing list Pw_forum@pwscf.org http://pwscf.org/mailman/listinfo/pw_forum -- Paolo Giannozzi, Dept. Chemistry&Physics&Environment, Univ. Udine, via delle Scienze 208, 33100 Udine, Italy Phone +39-0432-558216, fax +39-0432-558222___ Pw_forum mailing list Pw_forum@pwscf.org http://pwscf.org/mailman/listinfo/pw_forum
Re: [Pw_forum] A problem of parallel excuting pwcond.x
in theory, well-designed parallel code should work (or cleanly stop) on any number of processors. In practice, if you use a disproportionate number of processors, the code may not be able to deal with it. IN any case: look at what happens at line 444 of PWCOND/src/scatter_forw.f90 aqnd maybe you sill understand what is happening and why Paolo Paolo On Mon, Aug 10, 2015 at 9:27 AM, 庞瑞(PANG Rui) wrote: However, I found I could run the example correctly by using only 8 cores. > If using 16, 32 and 64 cores, the code will stop at > " ngper, shell number = 804 82 > ngper, n2d = 804 391 > --- E-Ef =2.000 k =0.500 0.500 > --- ie = 1 ik = 1" > > and showed > > > "forrtl: severe (174): SIGSEGV, segmentation fault occurred > Image PCRoutineLine > Source > > pwcond.x 025AD7C9 Unknown Unknown Unknown > pwcond.x 025AC140 Unknown Unknown Unknown > pwcond.x 0255DBC2 Unknown Unknown Unknown > pwcond.x 024F15F3 Unknown Unknown Unknown > pwcond.x 024F753B Unknown Unknown Unknown > libpthread.so.0003F2280F710 Unknown Unknown Unknown > pwcond.x 004F9B35 scatter_forw_ 444 > scatter_forw. > f90 > pwcond.x 004ADD66 do_cond_ 518 > do_cond.f90 > pwcond.x 004A8B42 MAIN__ 22 > condmain.f90 > pwcond.x 00498036 Unknown Unknown Unknown > libc.so.6 003F2241ED1D Unknown Unknown Unknown > pwcond.x 00497F29 Unknown Unknown > Unknown" > > So could anyone tell me how can I fix this error so that I can using more > cores to excute pwcond.x? I met this problem in both 5.1.2 and 5.2. > > Thanks very much for any help. > > Sincerely. > > Pang Rui > > > > > > -- > 庞瑞(PANG Rui) > > South University of Science and Technology of China/Department of Physics > > No.1088,Xueyuan Road, Shenzhen,Guangdong > > > > ___ > Pw_forum mailing list > Pw_forum@pwscf.org > http://pwscf.org/mailman/listinfo/pw_forum > -- Paolo Giannozzi, Dept. Chemistry&Physics&Environment, Univ. Udine, via delle Scienze 208, 33100 Udine, Italy Phone +39-0432-558216, fax +39-0432-558222 ___ Pw_forum mailing list Pw_forum@pwscf.org http://pwscf.org/mailman/listinfo/pw_forum
[Pw_forum] A problem of parallel excuting pwcond.x
Dear all I want to follow the example of the following page titled with " Monatomic Ni wire with a spin reversal" http://iramis.cea.fr/Pisp/alexander.smogunov/PWCOND/examples.html However, I found I could run the example correctly by using only 8 cores. If using 16, 32 and 64 cores, the code will stop at " ngper, shell number = 804 82 ngper, n2d = 804 391 --- E-Ef =2.000 k =0.500 0.500 --- ie = 1 ik = 1" and showed "forrtl: severe (174): SIGSEGV, segmentation fault occurred Image PCRoutineLineSource pwcond.x 025AD7C9 Unknown Unknown Unknown pwcond.x 025AC140 Unknown Unknown Unknown pwcond.x 0255DBC2 Unknown Unknown Unknown pwcond.x 024F15F3 Unknown Unknown Unknown pwcond.x 024F753B Unknown Unknown Unknown libpthread.so.0003F2280F710 Unknown Unknown Unknown pwcond.x 004F9B35 scatter_forw_ 444 scatter_forw. f90 pwcond.x 004ADD66 do_cond_ 518 do_cond.f90 pwcond.x 004A8B42 MAIN__ 22 condmain.f90 pwcond.x 00498036 Unknown Unknown Unknown libc.so.6 003F2241ED1D Unknown Unknown Unknown pwcond.x 00497F29 Unknown Unknown Unknown" So could anyone tell me how can I fix this error so that I can using more cores to excute pwcond.x? I met this problem in both 5.1.2 and 5.2. Thanks very much for any help. Sincerely. Pang Rui -- 庞瑞(PANG Rui) South University of Science and Technology of China/Department of Physics No.1088,Xueyuan Road, Shenzhen,Guangdong___ Pw_forum mailing list Pw_forum@pwscf.org http://pwscf.org/mailman/listinfo/pw_forum
