[Wien] lapw2 SIGSEV (9.1)
static and/or dynamic linking ??? Laurence Marks schrieb: On a 32bit machine, I cannot get the latest lapw2 to run -- even dropping the compilation down to FOPT = -FR -mp1 -prec_div -C -g. I always get a SIGSEV on first entry into l2main, with nothing useful (to me anyway) showing up with idb. Dropping down to an older version works. Has anyone else seen this? ifort 10.1.018 , mkl 10.1.1.019
[Wien] lapw2 SIGSEV (9.1)
The answer of Intel is found in all release notes of the Fortran Compilers 8.0 and higher: ... On Linux also note that many 32bit Linux distributions ship with a pthread static library ibpthread.a) that at runtime will fix the stacksize to 2093056 bytes regardless of the ulimit setting. To avoid this problem do not link with the -static option or the {\it -fast} option. Instead of {\it -fast}, use options: {\it -ipo -O3 -no-prec-div -xP}. This only affects the 32bit {\scshape Linux} distributions and does not apply to the 64bit {\scshape Linux} distributions. Thats why ulimit something does not help. There are different ways, either use an older compiler or use 64 bit Linux. Ciao Gerhard Dr. Gerhard H. Fecher Institut of Inorganic and Analytical Chemistry Johannes Gutenberg - University 55099 Mainz Von: wien-bounces at zeus.theochem.tuwien.ac.at [wien-bounces at zeus.theochem.tuwien.ac.at] im Auftrag von Laurence Marks [L-marks at northwestern.edu] Gesendet: Freitag, 7. August 2009 00:35 An: A Mailing list for WIEN2k users Betreff: [Wien] lapw2 SIGSEV (9.1) On a 32bit machine, I cannot get the latest lapw2 to run -- even dropping the compilation down to FOPT = -FR -mp1 -prec_div -C -g. I always get a SIGSEV on first entry into l2main, with nothing useful (to me anyway) showing up with idb. Dropping down to an older version works. Has anyone else seen this? ifort 10.1.018 , mkl 10.1.1.019 -- Laurence Marks Department of Materials Science and Engineering MSE Rm 2036 Cook Hall 2220 N Campus Drive Northwestern University Evanston, IL 60208, USA Tel: (847) 491-3996 Fax: (847) 491-7820 email: L-marks at northwestern dot edu Web: www.numis.northwestern.edu Chair, Commission on Electron Crystallography of IUCR www.numis.northwestern.edu/ Electron crystallography is the branch of science that uses electron scattering to study the structure of matter. ___ Wien mailing list Wien at zeus.theochem.tuwien.ac.at http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
[Wien] lapw2 SIGSEV (9.1)
The fix was to remove a -static option in the linking, at least for the 32bit machines. (N.B., I would not recommend the options -no-prec-div and instead would use -mpi -prec_div -ipo1; the options you are using will be very slightly faster but at the expense of accuracy, and -ipo1 is faster for execution and safe.) I don't know what you meant by the NO! with respect to /etc/csh.cshrc and ulimit. Even with a 64bit OS, depending upon specific details of the OS (and these things are never standard) many parameters such as stack-size limits, LD_LIBRARY_PATHS and other environmental variables are not inherited from the calling shell. On Fri, Aug 7, 2009 at 6:39 AM, Gerhard Fecherfecher at uni-mainz.de wrote: The answer of Intel is found in all release notes of the Fortran Compilers 8.0 and higher: ... On Linux also note that many 32bit ?Linux distributions ship with a pthread static library ibpthread.a) that at runtime will fix the stacksize to 2093056 bytes regardless of the ulimit setting. To avoid this problem do not link with the -static option or the {\it -fast} option. Instead of {\it -fast}, use options: {\it -ipo -O3 -no-prec-div -xP}. This only affects the 32bit {\scshape Linux} distributions and does not apply to the 64bit {\scshape Linux} distributions. Thats why ulimit something does not help. There are different ways, either use an older compiler or use 64 bit Linux. Ciao Gerhard Dr. Gerhard H. Fecher Institut of Inorganic and Analytical Chemistry Johannes Gutenberg - University 55099 Mainz Von: wien-bounces at zeus.theochem.tuwien.ac.at [wien-bounces at zeus.theochem.tuwien.ac.at] im Auftrag von Laurence Marks [L-marks at northwestern.edu] Gesendet: Freitag, 7. August 2009 00:35 An: A Mailing list for WIEN2k users Betreff: [Wien] lapw2 SIGSEV (9.1) On a 32bit machine, I cannot get the latest lapw2 to run -- even dropping the compilation down to FOPT = -FR -mp1 -prec_div -C -g. I always get a SIGSEV on first entry into l2main, with nothing useful (to me anyway) showing up with idb. Dropping down to an older version works. Has anyone else seen this? ifort 10.1.018 , mkl 10.1.1.019 -- Laurence Marks Department of Materials Science and Engineering MSE Rm 2036 Cook Hall 2220 N Campus Drive Northwestern University Evanston, IL 60208, USA Tel: (847) 491-3996 Fax: (847) 491-7820 email: L-marks at northwestern dot edu Web: www.numis.northwestern.edu Chair, Commission on Electron Crystallography of IUCR www.numis.northwestern.edu/ Electron crystallography is the branch of science that uses electron scattering to study the structure of matter. ___ Wien mailing list Wien at zeus.theochem.tuwien.ac.at http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien ___ Wien mailing list Wien at zeus.theochem.tuwien.ac.at http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien -- Laurence Marks Department of Materials Science and Engineering MSE Rm 2036 Cook Hall 2220 N Campus Drive Northwestern University Evanston, IL 60208, USA Tel: (847) 491-3996 Fax: (847) 491-7820 email: L-marks at northwestern dot edu Web: www.numis.northwestern.edu Chair, Commission on Electron Crystallography of IUCR www.numis.northwestern.edu/ Electron crystallography is the branch of science that uses electron scattering to study the structure of matter.
[Wien] lapw2 SIGSEV (9.1)
sorry if there is a misunderstanding, the text below is nothing but an original part of the Intel Compiler Release Notes, my recommendation that I usually give is indeed ... -mp1 (not -mpi) -prec_div etc. ... (note that Version 11 of the Fortran Compiler has some new switches to control the numerical precision of the calculation, it usually depends also which type of optimization was selected, if the optimization does not touch the divisions, there is no need to tell the compiler another time it should not do so. The switches to preserve the numeric precision are mainly needed together with -O3, -fast or similar) NO just means that you do not have a chance to have the stacksize larger than 2 093 056 bytes if your libpthread fixes the stacksize to that or whatever else value, and it seems you were running just into that trouble. Ciao Gerhard Dr. Gerhard H. Fecher Institut of Inorganic and Analytical Chemistry Johannes Gutenberg - University 55099 Mainz Von: wien-bounces at zeus.theochem.tuwien.ac.at [wien-bounces at zeus.theochem.tuwien.ac.at] im Auftrag von Laurence Marks [L-marks at northwestern.edu] Gesendet: Freitag, 7. August 2009 14:51 An: A Mailing list for WIEN2k users Betreff: Re: [Wien] lapw2 SIGSEV (9.1) The fix was to remove a -static option in the linking, at least for the 32bit machines. (N.B., I would not recommend the options -no-prec-div and instead would use -mpi -prec_div -ipo1; the options you are using will be very slightly faster but at the expense of accuracy, and -ipo1 is faster for execution and safe.) I don't know what you meant by the NO! with respect to /etc/csh.cshrc and ulimit. Even with a 64bit OS, depending upon specific details of the OS (and these things are never standard) many parameters such as stack-size limits, LD_LIBRARY_PATHS and other environmental variables are not inherited from the calling shell. On Fri, Aug 7, 2009 at 6:39 AM, Gerhard Fecherfecher at uni-mainz.de wrote: The answer of Intel is found in all release notes of the Fortran Compilers 8.0 and higher: ... On Linux also note that many 32bit Linux distributions ship with a pthread static library ibpthread.a) that at runtime will fix the stacksize to 2093056 bytes regardless of the ulimit setting. To avoid this problem do not link with the -static option or the {\it -fast} option. Instead of {\it -fast}, use options: {\it -ipo -O3 -no-prec-div -xP}. This only affects the 32bit {\scshape Linux} distributions and does not apply to the 64bit {\scshape Linux} distributions. Thats why ulimit something does not help. There are different ways, either use an older compiler or use 64 bit Linux. Ciao Gerhard Dr. Gerhard H. Fecher Institut of Inorganic and Analytical Chemistry Johannes Gutenberg - University 55099 Mainz Von: wien-bounces at zeus.theochem.tuwien.ac.at [wien-bounces at zeus.theochem.tuwien.ac.at] im Auftrag von Laurence Marks [L-marks at northwestern.edu] Gesendet: Freitag, 7. August 2009 00:35 An: A Mailing list for WIEN2k users Betreff: [Wien] lapw2 SIGSEV (9.1) On a 32bit machine, I cannot get the latest lapw2 to run -- even dropping the compilation down to FOPT = -FR -mp1 -prec_div -C -g. I always get a SIGSEV on first entry into l2main, with nothing useful (to me anyway) showing up with idb. Dropping down to an older version works. Has anyone else seen this? ifort 10.1.018 , mkl 10.1.1.019 -- Laurence Marks Department of Materials Science and Engineering MSE Rm 2036 Cook Hall 2220 N Campus Drive Northwestern University Evanston, IL 60208, USA Tel: (847) 491-3996 Fax: (847) 491-7820 email: L-marks at northwestern dot edu Web: www.numis.northwestern.edu Chair, Commission on Electron Crystallography of IUCR www.numis.northwestern.edu/ Electron crystallography is the branch of science that uses electron scattering to study the structure of matter. ___ Wien mailing list Wien at zeus.theochem.tuwien.ac.at http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien ___ Wien mailing list Wien at zeus.theochem.tuwien.ac.at http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien -- Laurence Marks Department of Materials Science and Engineering MSE Rm 2036 Cook Hall 2220 N Campus Drive Northwestern University Evanston, IL 60208, USA Tel: (847) 491-3996 Fax: (847) 491-7820 email: L-marks at northwestern dot edu Web: www.numis.northwestern.edu Chair, Commission on Electron Crystallography of IUCR www.numis.northwestern.edu/ Electron crystallography is the branch of science that uses electron scattering to study the structure of matter. ___ Wien mailing list Wien at zeus.theochem.tuwien.ac.at http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien