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

Re: PETSc/ridgeSim 3D scaling on BG/L
by Matthew Knepley 02 Apr '06

02 Apr '06

Thats awesome. 80% true efficiency for 1K nodes on under 1 million unknowns. Thats only 1,000 unknowns per proc. If we were cranking the problems size, I am sure you'd be up in the 90s. Hopefully this summer we can deliver you much improved serial performance as well. BTW, I want to talk sometime about the subduction benchmark. Van Keken talked about it at Purdue, so I got a diametrically opposed viewpoint from a sympathetic person. It was eye opening. Matt On 4/2/06, Richard Katz <katz(a)ldeo.columbia.edu> wrote: > > Hi All, > > Thought you might be interested to see this. > > The simulation is finite-volume, steady-state, non-Newtonian CFD in 3D. > And, of course, it is done using PETSc (version 2.3.1). > > How does this compare to other PETSc-based simulations? > > Cheers > Rich > > Barry -- I had a problem with the the jobs that we discussed but I'm > hoping to run them properly tonight. > > > > -- "Failure has a thousand explanations. Success doesn't need one" -- Sir Alec Guiness

1 0

PETSc/ridgeSim 3D scaling on BG/L
by Richard Katz 02 Apr '06

02 Apr '06

Hi All, Thought you might be interested to see this. The simulation is finite-volume, steady-state, non-Newtonian CFD in 3D. And, of course, it is done using PETSc (version 2.3.1). How does this compare to other PETSc-based simulations? Cheers Rich Barry -- I had a problem with the the jobs that we discussed but I'm hoping to run them properly tonight.

1 0

Re: Native Windows port of PETSc library - not cygwin based.
by Satish Balay 28 Mar '06

28 Mar '06

Petsc can also be used from gnu compilers. You need to specify the correct compilers you wish to use to PETSc configure I presume you would like to use PETSc from fortran [using intel fortran] on windows? If so: - install cygwin [for build tools per installation instructons] - install MLK for blas/lapack [if available or use configure option --download-f-blas-lapack=1] - install MPICH for windows [or if you wish to use PETSc sequentially, use --with-mpi=0] Now counfigure PETSc with the required compilers [check config/cygwin-ms.py] ./config/configure.py --with-cc='win32fe cl' --with-fc='win32fe ifort' [autodetect mpich & mlk from default locations] BTW: We test with compaq fortran - and haven't tested with intel fortran on windows for a while. If you encounter problems - let us know. Satish On Tue, 28 Mar 2006, Girish Pillai wrote: > Hi Satish, > > Somehow I thought it was the Gnu compilers that were > being used for the Windows port done with the help of > cygwin. So I think I was wrong in making that > assumption. > > So what I really want is to build it with the > MS-VisualC++ compiler and Intel Fortran Compiler. So > cygwin is just used for the build purpose and the GNU > compilers never get used. Is that correct? > > Thanks much, > Regards, > Girish > > --- Satish Balay <balay(a)mcs.anl.gov> wrote: > > > yes - PETSc can be installed with > > MS-C/Compaq-Fortran [or intel > > compilers] on windows. > > > > However the build system still requires cygwin to > > build the PETSc > > libraries. > > > > Check the installation instructions for additional > > info. If you > > encounter problems - send us the erros with > > appropriate logfiles to > > petsc-maint(a)mcs.anl.gov > > > > [If configure doesn't work - send us configure.log] > > > > Satish > > > > On Tue, 28 Mar 2006, Girish Pillai wrote: > > > > > Folks, > > > > > > Is there a port of PETSc library on Windows > > platform. > > > I am looking for native Windows port, not the > > cygwin > > > port. > > > > > > Can one of you please respond? > > > > > > The config script does not work properly for > > Windows > > > when I tried to run it. > > > > > > Thanks, > > > Girish > > > > > > > > > __________________________________________________ > > > Do You Yahoo!? > > > Tired of spam? Yahoo! Mail has the best spam > > protection around > > > http://mail.yahoo.com > > > > > > > > > > > > > __________________________________________________ > Do You Yahoo!? > Tired of spam? Yahoo! Mail has the best spam protection around > http://mail.yahoo.com > >

1 0

Re: Native Windows port of PETSc library - not cygwin based.
by Girish Pillai 28 Mar '06

28 Mar '06

Hi Satish, Somehow I thought it was the Gnu compilers that were being used for the Windows port done with the help of cygwin. So I think I was wrong in making that assumption. So what I really want is to build it with the MS-VisualC++ compiler and Intel Fortran Compiler. So cygwin is just used for the build purpose and the GNU compilers never get used. Is that correct? Thanks much, Regards, Girish --- Satish Balay <balay(a)mcs.anl.gov> wrote: > yes - PETSc can be installed with > MS-C/Compaq-Fortran [or intel > compilers] on windows. > > However the build system still requires cygwin to > build the PETSc > libraries. > > Check the installation instructions for additional > info. If you > encounter problems - send us the erros with > appropriate logfiles to > petsc-maint(a)mcs.anl.gov > > [If configure doesn't work - send us configure.log] > > Satish > > On Tue, 28 Mar 2006, Girish Pillai wrote: > > > Folks, > > > > Is there a port of PETSc library on Windows > platform. > > I am looking for native Windows port, not the > cygwin > > port. > > > > Can one of you please respond? > > > > The config script does not work properly for > Windows > > when I tried to run it. > > > > Thanks, > > Girish > > > > > > __________________________________________________ > > Do You Yahoo!? > > Tired of spam? Yahoo! Mail has the best spam > protection around > > http://mail.yahoo.com > > > > > > __________________________________________________ Do You Yahoo!? Tired of spam? Yahoo! Mail has the best spam protection around http://mail.yahoo.com

1 0

Re: Native Windows port of PETSc library - not cygwin based.
by Satish Balay 28 Mar '06

28 Mar '06

yes - PETSc can be installed with MS-C/Compaq-Fortran [or intel compilers] on windows. However the build system still requires cygwin to build the PETSc libraries. Check the installation instructions for additional info. If you encounter problems - send us the erros with appropriate logfiles to petsc-maint(a)mcs.anl.gov [If configure doesn't work - send us configure.log] Satish On Tue, 28 Mar 2006, Girish Pillai wrote: > Folks, > > Is there a port of PETSc library on Windows platform. > I am looking for native Windows port, not the cygwin > port. > > Can one of you please respond? > > The config script does not work properly for Windows > when I tried to run it. > > Thanks, > Girish > > > __________________________________________________ > Do You Yahoo!? > Tired of spam? Yahoo! Mail has the best spam protection around > http://mail.yahoo.com > >

1 0

Native Windows port of PETSc library - not cygwin based.
by Girish Pillai 28 Mar '06

28 Mar '06

Folks, Is there a port of PETSc library on Windows platform. I am looking for native Windows port, not the cygwin port. Can one of you please respond? The config script does not work properly for Windows when I tried to run it. Thanks, Girish __________________________________________________ Do You Yahoo!? Tired of spam? Yahoo! Mail has the best spam protection around http://mail.yahoo.com

1 0

Fwd: New Additions
by Matthew Knepley 17 Mar '06

17 Mar '06

---------- Forwarded message ---------- From: Matthew Knepley <knepley(a)gmail.com> Date: Mar 17, 2006 10:42 AM Subject: New Additions To: petsc-dev(a)mc.anl.gov Cc: Dmitry Karpeev <karpeev(a)mcs.anl.gov> I just pushed: 1) A new changes file for dev. We can add things we do to this file, and then move them to whatever version we decide to release. That way people can see what is happening in dev realtime. 2) I added a nice Matlab DA viewer to viewers/util that Ethan and Rich wrote. There is a script to use it in bin/matlab. Ethan now believes all this can be converted seamlessly to Matplotlib in Python (which has Matlab syntax). 3) I added the method of charcteristics solver to src/contrib/semiLagrange. Check out the examples. They are really cool. Matt -- "Failure has a thousand explanations. Success doesn't need one" -- Sir Alec Guiness -- "Failure has a thousand explanations. Success doesn't need one" -- Sir Alec Guiness

1 0

more updates to VecScatter code
by Barry Smith 10 Mar '06

10 Mar '06

I have pushed another round of improvements to the parallel VecScatter code. There is now a universal parallel VecScatterBegin/End templated over block size. Also further optimizations in generation and use of the scatters. Be on the look out for subtle bugs in scatters. Barry

1 0

testing petsc-dev for upcoming release
by Satish Balay 02 Feb '06

02 Feb '06

PETSc Dev users, We hope to make a release tomorrow [friday]. This is a call for help to test configure. This is regression testing to make sure petsc-dev configure works for all working 2.3.0 installs So, try building petsc-dev with the _equivalent_ configure options that worked before for petsc-2.3.0. If this doesn't work - let us know [at petsc-maint(a)mcs.anl.gov] Previously used configure command can be obtained from petsc-2.3.0/bmake/$PETSC_ARCH/configure.py petsc-dev tar file is located at: ftp://info.mcs.anl.gov/pub/petsc/petsc-dev.tar.gz [Configure] changes are listed at: http://www-unix.mcs.anl.gov/petsc/petsc-as/documentation/changes/231.html Thanks, Satish

1 0

Re: efficient small matrix solver
by Barry Smith 26 Jan '06

26 Jan '06

Matt, You can the routines in src/mat/impls/baij/seq/dgefa*.c see also src/inline/ilu.c See also src/mat/impls/baij/seq/baijfact12.c where there are some factorizations specifically for the Intel SSE. Barry On Thu, 26 Jan 2006, Matt wrote: > Hi, > > We use PETSc and noticed on the web site that you built (from linpack) > some routines to factor/solve small dense matrices efficiently. What > are those routines? I'd like to get the bottom level routines as I want > to modify them just a bit more. (I want to set the matrix size as a > constant to help the compiler, and possibly add some pragmas to help > the compiler.) > > > > Thanks, > > Matt > >

1 0