Hello,<br><br>Thank you for all your reply.<br>I have tested my code using the debug version and there is the same problem. The error message is attached.<br><br>The problem I am solving is a nonlinear system. I use inexact Newton method with preconditioned restarted GMRES (-ksp_type gnres) for the linear system.<br>
The preconditioner is ASM (-pc_type asm) and the sub problem is solved by LU (-sub_pc_type lu). Since there are zeros on the diagonal, I use the command line "-sub_pc_factor_shift_amount 1.e-10 -sub_pc_factor_shift_type nonzero" to add a small number to the zero diagonals.<br>
<br>I found some error message in the configure.log(see bellow). Could this be a problem?<br>*********************************************************<br># Directories<br> files, <br># %s (device %ld, inode %ld): <br> impossibilities<br>
so far.<br> impossibilities in %lu directories.<br># %s: could not be stat'd.<br># %s (device %ld, inode %ld): could not be opened.<br>Recursive variable `%s' references itself (eventually)<br>warning: undefined variable `%.*s'<br>
unterminated variable reference<br>%s:%s<br> | %s<br># Implicit rule search has been done.<br># Implicit rule search has not been done.<br># Modification time never checked.<br># File does not exist.<br># Last modified %s<br>
# File has been updated.<br># File has not been updated.<br># Invalid value in `update_status' member!<br># Not a target:<br># Also makes:<br># File is an intermediate prerequisite.<br># Implicit/static pattern stem: `%s'<br>
# A default or MAKEFILES makefile.<br># Command-line target.<br># Phony target (prerequisite of .PHONY).<br># Precious file (prerequisite of .PRECIOUS).<br># Failed to be updated.<br># Successfully updated.<br># Needs to be updated (-q is set).<br>
question_flag<br>file.c<br>print_file<br># Dependencies commands running (THIS IS A BUG).<br># File is very old.<br># Commands currently running (THIS IS A BUG).<br># Invalid value in `command_state' member!<br># Files<br>
# files hash-table stats:<br>%04d-%02d-%02d %02d:%02d:%02d<br>.%09d<br>%s: Timestamp out of range; substituting %s<br>Current time<br>*** Deleting intermediate file `%s'<br>Removing intermediate files...<br>can't rename single-colon `%s' to double-colon `%s'<br>
Commands were specified for file `%s' at %s:%lu,<br>but `%s' is now considered the same file as `%s'.<br>Commands for `%s' will be ignored in favor of those for `%s'.<br>Commands for file `%s' were found by implicit rule search,<br>
can't rename double-colon `%s' to single-colon `%s'<br>.........<br><br>TEST checkDynamicLinker from config.setCompilers(/home/rchen/soft/petsc-3.1-p7-nodebug/config/BuildSystem/config/setCompilers.py:1257)<br>
TESTING: checkDynamicLinker from config.setCompilers(config/BuildSystem/config/setCompilers.py:1257)<br> Check that the linker can produce dynamic libraries<br>Checking for header: dlfcn.h<br>sh: /contrib/bgl/bin/mpxlc -E conftest.c <br>
Executing: /contrib/bgl/bin/mpxlc -E conftest.c <br>sh: <br>Possible ERROR while running preprocessor: "conftest.c", line 3.10: 1506-296 (S) #include file <dlfcn.h> not found.<br>ret = 256<br>error message = {"conftest.c", line 3.10: 1506-296 (S) #include file <dlfcn.h> not found.<br>
}<br>Source:<br>#include "confdefs.h"<br>#include "conffix.h"<br>#include <dlfcn.h><br> Dynamic libraries disabled since dlfcn.h was missing<br>********************************************<br>
<br>Thank you!<br><br>Regards,<br><br>Rongliang<br><div class="gmail_quote"><br><blockquote class="gmail_quote" style="margin: 0pt 0pt 0pt 0.8ex; border-left: 1px solid rgb(204, 204, 204); padding-left: 1ex;">
--------------------------------------------------------------------------<br>
Message: 5<br>
Date: Tue, 25 Jan 2011 23:19:27 -0600 (CST)<br>
From: Satish Balay <<a href="mailto:balay@mcs.anl.gov" target="_blank">balay@mcs.anl.gov</a>><br>
Subject: Re: [petsc-users] Problem on LU factorization<br>
To: Rongliang Chen <<a href="mailto:rongliang.chan@gmail.com" target="_blank">rongliang.chan@gmail.com</a>><br>
Cc: PETSc Users mail list <<a href="mailto:petsc-users@mcs.anl.gov" target="_blank">petsc-users@mcs.anl.gov</a>><br>
Message-ID:<br>
<alpine.LFD.2.02.1101252316570.2510@localhost6.localdomain6><br>
Content-Type: TEXT/PLAIN; charset=US-ASCII<br>
<br>
I don't see anything obviously wrong with this build<br>
<br>
I guess the other thing to do is to build debug version on the machine<br>
and run in a debugger to determine the problem. [I believe there is a<br>
way to debug on bgl..]<br>
<br>
Satish<br>
<br>
<br>
------------------------------<br>
<br>
Message: 7<br>
Date: Wed, 26 Jan 2011 09:39:17 -0600<br>
From: Hong Zhang <<a href="mailto:hzhang@mcs.anl.gov" target="_blank">hzhang@mcs.anl.gov</a>><br>
Subject: Re: [petsc-users] Problem on LU factorization<br>
To: PETSc users list <<a href="mailto:petsc-users@mcs.anl.gov" target="_blank">petsc-users@mcs.anl.gov</a>><br>
Message-ID:<br>
<<a href="mailto:AANLkTikJJCFQnESxfTMrwgiTjHMmGSoYDNFG6GivQO84@mail.gmail.com" target="_blank">AANLkTikJJCFQnESxfTMrwgiTjHMmGSoYDNFG6GivQO84@mail.gmail.com</a>><br>
Content-Type: text/plain; charset=ISO-8859-1<br>
<br>
As Matt said, this is no enough information on what solver combination<br>
is being used.<br>
Since superlu_dist works while petsc/superlu's sequential lu fails,<br>
you might need<br>
parallel lu, or shift if zero pivot causes crash. As suggested by Satish,<br>
run code in debug version from which informative error display would be shown.<br>
<br>
Hong<br>
<br></blockquote></div>