April 2012 - petsc-dev - mailman3.cels.anl.gov

Re: [petsc-dev] [petsc-users] How to include PETSc in an Eclipse CDT project
by Satish Balay 18 Apr '12

18 Apr '12

On Wed, 18 Apr 2012, Matthew Knepley wrote: > On Wed, Apr 18, 2012 at 1:50 PM, Jed Brown <jedbrown(a)mcs.anl.gov> wrote: > > > On Wed, Apr 18, 2012 at 10:46, Christian Staudt < > > christian.staudt(a)ira.uka.de> wrote: > > > >> make getincludedirs > >> > >> failed for me with: > >> > >> cls /usr/local/petsc $ make getincludedirs > >> conf/petscvariables:3: > >> /usr/local/petsc-3.2-p6/arch-darwin-c-debug/conf/petscvariables: No such > >> file or directory > >> makefile:15: /usr/local/petsc-3.2-p6/conf/variables: No such file or > >> directory > >> makefile:16: /usr/local/petsc-3.2-p6/conf/rules: No such file or directory > >> makefile:17: /usr/local/petsc-3.2-p6/conf/test: No such file or directory > >> make: *** No rule to make target `/usr/local/petsc-3.2-p6/conf/test'. > >> Stop. > >> > > > > Matt, is this expected to work for a prefix install? > > > > No 'make getincludedirs' does work - provided one uses a petsc makefile [perhaps from build sources]. But no makefile is installed in the prefix location. Also this build is not a prefix build. With a prefix install both 'petsc.h' and 'petscconf.h' reside in prefix/include location. satish

1 0

Re: [petsc-dev] [petsc-users] How to include PETSc in an Eclipse CDT project
by Matthew Knepley 18 Apr '12

18 Apr '12

On Wed, Apr 18, 2012 at 1:50 PM, Jed Brown <jedbrown(a)mcs.anl.gov> wrote: > On Wed, Apr 18, 2012 at 10:46, Christian Staudt < > christian.staudt(a)ira.uka.de> wrote: > >> make getincludedirs >> >> failed for me with: >> >> cls /usr/local/petsc $ make getincludedirs >> conf/petscvariables:3: >> /usr/local/petsc-3.2-p6/arch-darwin-c-debug/conf/petscvariables: No such >> file or directory >> makefile:15: /usr/local/petsc-3.2-p6/conf/variables: No such file or >> directory >> makefile:16: /usr/local/petsc-3.2-p6/conf/rules: No such file or directory >> makefile:17: /usr/local/petsc-3.2-p6/conf/test: No such file or directory >> make: *** No rule to make target `/usr/local/petsc-3.2-p6/conf/test'. >> Stop. >> > > Matt, is this expected to work for a prefix install? > No Matt -- What most experimenters take for granted before they begin their experiments is infinitely more interesting than any results to which their experiments lead. -- Norbert Wiener

1 0

Re: [petsc-dev] [petsc-users] How to include PETSc in an Eclipse CDT project
by Jed Brown 18 Apr '12

18 Apr '12

On Wed, Apr 18, 2012 at 10:46, Christian Staudt <christian.staudt(a)ira.uka.de > wrote: > make getincludedirs > > failed for me with: > > cls /usr/local/petsc $ make getincludedirs > conf/petscvariables:3: > /usr/local/petsc-3.2-p6/arch-darwin-c-debug/conf/petscvariables: No such > file or directory > makefile:15: /usr/local/petsc-3.2-p6/conf/variables: No such file or > directory > makefile:16: /usr/local/petsc-3.2-p6/conf/rules: No such file or directory > makefile:17: /usr/local/petsc-3.2-p6/conf/test: No such file or directory > make: *** No rule to make target `/usr/local/petsc-3.2-p6/conf/test'. > Stop. > Matt, is this expected to work for a prefix install?

1 0

Re: [petsc-dev] PCApplyTranspose in BiCG
by Barry Smith 18 Apr '12

18 Apr '12

On Apr 18, 2012, at 9:51 AM, Alexander Grayver wrote: > Are there other methods affected by that? Possibly. > Now it seems to be not easy to say whether solver just doesn't converge or it's wrong because of that issue. The issue is Krylov solvers properly coded for complex numbers. We are not experts in that area and the Krylov solvers are a little iffy for that. Barry > > On 18.04.2012 15:26, Barry Smith wrote: >> On Apr 18, 2012, at 7:55 AM, Alexander Grayver wrote: >> >>>>> Jed, >>>>> >>>>> Thank you. >>>>> TFQMR works well now, however TCQMR doesn't. Although latter was working before I updated petsc-dev. Might be same issue? >>>> Please clarify. When was TCQMR working and when did it stop working? Was TCQMR working before you just updated in the last 24 hours? Did TCQMR work a long time ago (like TFQMR worked a long time ago?)? >>> Barry, that looks weird, but seems to be true: >>> >>> HG revision: a4dae06b4d05ef5b78a9a9aa2cbc31b2b158834f HG Date: Wed Feb 08 11:29:52 2012 -0600 --- TFQMR works, TCQMR does not >>> HG revision: 59ca71b85d784abbe22f12c34ffcb2ccc6dc68b0 HG Date: Tue Mar 20 01:05:13 2012 -0500 --- TFQMR does not work, TCQMR works >>> HG revision: 8f37c498b7b419ef0db5fb335a07c874e6dfe6d9 HG Date: Tue Apr 17 21:06:44 2012 -0500 --- TFQMR works, TCQMR does not >>> >> Not weird at all, makes perfect sense. Some inner products in TCQMR must be coded "backwards" (that is with the two vector arguments flipped) from what PETSc expects. When we accidently flipped the roles of the two arguments that triggered TCQMR to start working. There are three inner products in TCQMR >> >> ierr = VecDot(v0,y,&dp11);CHKERRQ(ierr); >> ierr = VecDot(v0,u,&dp2);CHKERRQ(ierr); >> alpha = dp11 / dp2; /* alpha = v0'*y/v0'*u */ >> deltmp = alpha; >> ierr = VecCopy(y,z);CHKERRQ(ierr); >> ierr = VecAXPY(z,-alpha,u);CHKERRQ(ierr); /* z = y - alpha u */ >> ierr = VecDot(v0,u,&rho);CHKERRQ(ierr); >> >> hence 8 possible combinations of orders for the vector arguments to the three calls. Now one could try to understand the algorithm for complex numbers to determine the correct orderings or one could simply try them :-) >> >> >> Barry >> >> >>> Alexander >>> >>>> Barry > > > -- > Regards, > Alexander >

1 0

Re: [petsc-dev] PCApplyTranspose in BiCG
by Alexander Grayver 18 Apr '12

18 Apr '12

Are there other methods affected by that? Now it seems to be not easy to say whether solver just doesn't converge or it's wrong because of that issue. On 18.04.2012 15:26, Barry Smith wrote: > On Apr 18, 2012, at 7:55 AM, Alexander Grayver wrote: > >>>> Jed, >>>> >>>> Thank you. >>>> TFQMR works well now, however TCQMR doesn't. Although latter was working before I updated petsc-dev. Might be same issue? >>> Please clarify. When was TCQMR working and when did it stop working? Was TCQMR working before you just updated in the last 24 hours? Did TCQMR work a long time ago (like TFQMR worked a long time ago?)? >> Barry, that looks weird, but seems to be true: >> >> HG revision: a4dae06b4d05ef5b78a9a9aa2cbc31b2b158834f HG Date: Wed Feb 08 11:29:52 2012 -0600 --- TFQMR works, TCQMR does not >> HG revision: 59ca71b85d784abbe22f12c34ffcb2ccc6dc68b0 HG Date: Tue Mar 20 01:05:13 2012 -0500 --- TFQMR does not work, TCQMR works >> HG revision: 8f37c498b7b419ef0db5fb335a07c874e6dfe6d9 HG Date: Tue Apr 17 21:06:44 2012 -0500 --- TFQMR works, TCQMR does not >> > Not weird at all, makes perfect sense. Some inner products in TCQMR must be coded "backwards" (that is with the two vector arguments flipped) from what PETSc expects. When we accidently flipped the roles of the two arguments that triggered TCQMR to start working. There are three inner products in TCQMR > > ierr = VecDot(v0,y,&dp11);CHKERRQ(ierr); > ierr = VecDot(v0,u,&dp2);CHKERRQ(ierr); > alpha = dp11 / dp2; /* alpha = v0'*y/v0'*u */ > deltmp = alpha; > ierr = VecCopy(y,z);CHKERRQ(ierr); > ierr = VecAXPY(z,-alpha,u);CHKERRQ(ierr); /* z = y - alpha u */ > ierr = VecDot(v0,u,&rho);CHKERRQ(ierr); > > hence 8 possible combinations of orders for the vector arguments to the three calls. Now one could try to understand the algorithm for complex numbers to determine the correct orderings or one could simply try them :-) > > > Barry > > >> Alexander >> >>> Barry -- Regards, Alexander

1 0

Re: [petsc-dev] PCApplyTranspose in BiCG
by Barry Smith 18 Apr '12

18 Apr '12

On Apr 18, 2012, at 7:55 AM, Alexander Grayver wrote: > >>> Jed, >>> >>> Thank you. >>> TFQMR works well now, however TCQMR doesn't. Although latter was working before I updated petsc-dev. Might be same issue? >> Please clarify. When was TCQMR working and when did it stop working? Was TCQMR working before you just updated in the last 24 hours? Did TCQMR work a long time ago (like TFQMR worked a long time ago?)? > > Barry, that looks weird, but seems to be true: > > HG revision: a4dae06b4d05ef5b78a9a9aa2cbc31b2b158834f HG Date: Wed Feb 08 11:29:52 2012 -0600 --- TFQMR works, TCQMR does not > HG revision: 59ca71b85d784abbe22f12c34ffcb2ccc6dc68b0 HG Date: Tue Mar 20 01:05:13 2012 -0500 --- TFQMR does not work, TCQMR works > HG revision: 8f37c498b7b419ef0db5fb335a07c874e6dfe6d9 HG Date: Tue Apr 17 21:06:44 2012 -0500 --- TFQMR works, TCQMR does not > Not weird at all, makes perfect sense. Some inner products in TCQMR must be coded "backwards" (that is with the two vector arguments flipped) from what PETSc expects. When we accidently flipped the roles of the two arguments that triggered TCQMR to start working. There are three inner products in TCQMR ierr = VecDot(v0,y,&dp11);CHKERRQ(ierr); ierr = VecDot(v0,u,&dp2);CHKERRQ(ierr); alpha = dp11 / dp2; /* alpha = v0'*y/v0'*u */ deltmp = alpha; ierr = VecCopy(y,z);CHKERRQ(ierr); ierr = VecAXPY(z,-alpha,u);CHKERRQ(ierr); /* z = y - alpha u */ ierr = VecDot(v0,u,&rho);CHKERRQ(ierr); hence 8 possible combinations of orders for the vector arguments to the three calls. Now one could try to understand the algorithm for complex numbers to determine the correct orderings or one could simply try them :-) Barry > Alexander > >> >> Barry

1 0

Re: [petsc-dev] PCApplyTranspose in BiCG
by Alexander Grayver 18 Apr '12

18 Apr '12

>> Jed, >> >> Thank you. >> TFQMR works well now, however TCQMR doesn't. Although latter was working before I updated petsc-dev. Might be same issue? > Please clarify. When was TCQMR working and when did it stop working? Was TCQMR working before you just updated in the last 24 hours? Did TCQMR work a long time ago (like TFQMR worked a long time ago?)? Barry, that looks weird, but seems to be true: HG revision: a4dae06b4d05ef5b78a9a9aa2cbc31b2b158834f HG Date: Wed Feb 08 11:29:52 2012 -0600 --- TFQMR works, TCQMR does not HG revision: 59ca71b85d784abbe22f12c34ffcb2ccc6dc68b0 HG Date: Tue Mar 20 01:05:13 2012 -0500 --- TFQMR does not work, TCQMR works HG revision: 8f37c498b7b419ef0db5fb335a07c874e6dfe6d9 HG Date: Tue Apr 17 21:06:44 2012 -0500 --- TFQMR works, TCQMR does not Alexander > > Barry

1 0

Re: [petsc-dev] pip cannot currently automatically download and install petsc on OS X
by Aron Ahmadia 18 Apr '12

18 Apr '12

The mpif90 on OS X.3-6 is a non-functional little script installed by the OpenMPI folks that mpi4py happily picks up and reports as installed. We need to patch mpi4py to correctly detect this little piece of disfunctionality as non-existent. sudo rm /usr/bin/mpif90 and a reinstall of mpi4py fixes that issue :) A On Wed, Apr 18, 2012 at 3:39 PM, Aron Ahmadia <aron.ahmadia(a)kaust.edu.sa>wrote: > Bogus! This is coming from a bad mpi4py install on my box! > > > On Wed, Apr 18, 2012 at 3:25 PM, Matthew Knepley <knepley(a)gmail.com>wrote: > >> On Wed, Apr 18, 2012 at 8:18 AM, Aron Ahmadia <aron.ahmadia(a)kaust.edu.sa>wrote: >> >>> Lisandro has fixed the first issue, but another has sprung up in its >>> place: >>> >>> pip seems to be installing the packages out of order and insists on >>> building petsc4py before petsc. We still have the issue with petsc wanting >>> mpif90 available on OS X... >>> >> >> Where are the configure arguments stored so I can look at them? >> >> Matt >> >> >>> A >>> >>> >>> On Wed, Apr 18, 2012 at 2:57 PM, Aron Ahmadia <aron(a)ahmadia.net> wrote: >>> >>>> This is on an OS X.6 box running Python 2.7 and pip 1.0.2 >>>> >>>> # at command line: >>>> pip install -vvv petsc4py >>>> >>>> # relevant bits: >>>> Downloading/unpacking petsc>=3.2,<3.3 (from petsc4py) >>>> Getting page http://pypi.python.org/simple/petsc >>>> URLs to search for versions for petsc>=3.2,<3.3 (from petsc4py): >>>> * http://pypi.python.org/simple/petsc/ >>>> Getting page http://www.mcs.anl.gov/petsc/ >>>> ... >>>> Skipping page >>>> http://ftp.mcs.anl.gov/pub/petsc/release-snapshots/petsc-lite-3.2-p6.tar.gz… >>>> http://pypi.python.org/simple/petsc/) because of Content-Type: >>>> application/x-gzip >>>> ... >>>> >>>> Found link >>>> http://ftp.mcs.anl.gov/pub/petsc/release-snapshots/petsc-lite-3.2-p6.tar.gz… >>>> http://pypi.python.org/simple/petsc/) version: lite-3.2-p6 >>>> >>>> The "lite-" prefix is being interpreted as part of the version, which >>>> pip doesn't quite know what to do with. It would probably just be easiest >>>> to host lite versions in a separate directory >>>> ../release-snapshots/lite/petsc-3.2-p6.tar.gz that petsc4py/PyPI can point >>>> to. Lisandro can also point to the non-lite versions until a solution is >>>> implemented. >>>> >>>> This is going to be an additional problem: >>>> >>>> >>>> ******************************************************************************* >>>> >>>> UNABLE to CONFIGURE with GIVEN OPTIONS (see configure.log >>>> for details): >>>> >>>> >>>> ------------------------------------------------------------------------------- >>>> >>>> Fortran compiler you provided with --with-fc=/usr/bin/mpif90 does not >>>> work >>>> >>>> ******************************************************************************* >>>> >>>> Obviously the Mac doesn't come with /usr/bin/mpif90, so a clean system >>>> won't have this either... I think you guys know your options for how to >>>> fix this, but let me know if you want my 2 cents :) >>>> >>>> =AA= >>>> >>> >>> >> >> >> -- >> What most experimenters take for granted before they begin their >> experiments is infinitely more interesting than any results to which their >> experiments lead. >> -- Norbert Wiener >> > >

1 0

Re: [petsc-dev] PCApplyTranspose in BiCG
by Barry Smith 18 Apr '12

18 Apr '12

On Apr 18, 2012, at 2:20 AM, Alexander Grayver wrote: > Jed, > > Thank you. > TFQMR works well now, however TCQMR doesn't. Although latter was working before I updated petsc-dev. Might be same issue? Please clarify. When was TCQMR working and when did it stop working? Was TCQMR working before you just updated in the last 24 hours? Did TCQMR work a long time ago (like TFQMR worked a long time ago?)? Barry > > On 18.04.2012 07:30, Jed Brown wrote: >> Thanks for the details, we should have identified the issue earlier. When VecDot changed from using custom code to BLAS xDOT, the conjugation definition was accidentally changed. If you pull after this patch, TFQMR will work for your problem again. >> >> http://petsc.cs.iit.edu/petsc/petsc-dev/rev/8f37c498b7b4 >> >> On Wed, Apr 11, 2012 at 00:38, Alexander Grayver <agrayver(a)gfz-potsdam.de> wrote: >> On 10.04.2012 14:19, Jed Brown wrote: >>> >>> There is no space allocated at (44,44); note that counting is 0-based. This consistency check has been in PETSc for many years, so I don't know why you didn't trip it in earlier runs. Do you have some options in PETSC_OPTIONS or .petscrc? >> >> Jed, >> >> This is not possible: >> >> >> A = PetscBinaryRead('A.dat','complex',true); >> >> find(diag(A)==0) >> ans = >> Empty matrix: 0-by-1 >> >> >> A(45,45) >> ans = (1,1) 7.999999999999997e+002 +6.316546778308559e-009i >> >> Sorry, I forgot to mention that this is double complex matrix and vector. Have you read it as real double? >> >> >>> >>> >>> If it is of any help files are here: >>> http://dl.dropbox.com/u/60982984/A.dat >>> http://dl.dropbox.com/u/60982984/RHS.dat >>> >>> Is this possible to look up changes in one file only? >>> >>> $ hg log src/ksp/ksp/impls/tfqmr/ >>> >>> But the problem was clearly not in that file. >> >> >> >> -- >> Regards, >> Alexander >> >> > > > -- > Regards, > Alexander >

1 0

Re: [petsc-dev] pip cannot currently automatically download and install petsc on OS X
by Aron Ahmadia 18 Apr '12

18 Apr '12

Bogus! This is coming from a bad mpi4py install on my box! On Wed, Apr 18, 2012 at 3:25 PM, Matthew Knepley <knepley(a)gmail.com> wrote: > On Wed, Apr 18, 2012 at 8:18 AM, Aron Ahmadia <aron.ahmadia(a)kaust.edu.sa>wrote: > >> Lisandro has fixed the first issue, but another has sprung up in its >> place: >> >> pip seems to be installing the packages out of order and insists on >> building petsc4py before petsc. We still have the issue with petsc wanting >> mpif90 available on OS X... >> > > Where are the configure arguments stored so I can look at them? > > Matt > > >> A >> >> >> On Wed, Apr 18, 2012 at 2:57 PM, Aron Ahmadia <aron(a)ahmadia.net> wrote: >> >>> This is on an OS X.6 box running Python 2.7 and pip 1.0.2 >>> >>> # at command line: >>> pip install -vvv petsc4py >>> >>> # relevant bits: >>> Downloading/unpacking petsc>=3.2,<3.3 (from petsc4py) >>> Getting page http://pypi.python.org/simple/petsc >>> URLs to search for versions for petsc>=3.2,<3.3 (from petsc4py): >>> * http://pypi.python.org/simple/petsc/ >>> Getting page http://www.mcs.anl.gov/petsc/ >>> ... >>> Skipping page >>> http://ftp.mcs.anl.gov/pub/petsc/release-snapshots/petsc-lite-3.2-p6.tar.gz… >>> http://pypi.python.org/simple/petsc/) because of Content-Type: >>> application/x-gzip >>> ... >>> >>> Found link >>> http://ftp.mcs.anl.gov/pub/petsc/release-snapshots/petsc-lite-3.2-p6.tar.gz… >>> http://pypi.python.org/simple/petsc/) version: lite-3.2-p6 >>> >>> The "lite-" prefix is being interpreted as part of the version, which >>> pip doesn't quite know what to do with. It would probably just be easiest >>> to host lite versions in a separate directory >>> ../release-snapshots/lite/petsc-3.2-p6.tar.gz that petsc4py/PyPI can point >>> to. Lisandro can also point to the non-lite versions until a solution is >>> implemented. >>> >>> This is going to be an additional problem: >>> >>> >>> ******************************************************************************* >>> >>> UNABLE to CONFIGURE with GIVEN OPTIONS (see configure.log >>> for details): >>> >>> >>> ------------------------------------------------------------------------------- >>> >>> Fortran compiler you provided with --with-fc=/usr/bin/mpif90 does not >>> work >>> >>> ******************************************************************************* >>> >>> Obviously the Mac doesn't come with /usr/bin/mpif90, so a clean system >>> won't have this either... I think you guys know your options for how to >>> fix this, but let me know if you want my 2 cents :) >>> >>> =AA= >>> >> >> > > > -- > What most experimenters take for granted before they begin their > experiments is infinitely more interesting than any results to which their > experiments lead. > -- Norbert Wiener >

1 0