July 2013 - petsc-dev - mailman3.cels.anl.gov

Re: [petsc-dev] Adding DMLocalToLocalBegin/End: What about PetscSF?
by Matthew Knepley 08 Jul '13

08 Jul '13

On Mon, Jul 8, 2013 at 4:16 PM, Richard Tran Mills <rtm(a)eecs.utk.edu> wrote: > Hi Jed, > > Revisiting this now; thanks for your help so far. I understand that I can > use PetscSFBcast() for this, but it is not clear to me whether there is a > good way that I can make use of the "default" SF that exists (or the > "default" PetscSection used to construct this). Should I just construct > some other "default" for local-to-local, or is there a good way to remap > the "global" vector layout onto a local one? > The section stuff will not help you much, because you would still need to create a mesh point SF. Its useful if you want to manage all data that way. Matt > --Richard > > On 6/12/13 11:25 AM, Jed Brown wrote: > >> Richard Tran Mills <rtm(a)eecs.utk.edu> writes: >> >> PETSc developers, >>> >>> I am finally getting around to adding the DMLocalToLocalBegin/End >>> routines >>> that I had mentioned several months ago. Looking at >>> DMGlobalToLocalBegin() in src/interface/dm/dm.c, I see that a little over >>> a year ago, Matt added the use of a default 'PetscSF' ("star forest") if >>> present to handle the global-to-local and local-to-global operations. I >>> am not familiar with PetscSF, but I assume that it should be possible to >>> use these somehow (maybe needing to add a little bit of code) for >>> local-to-local operations, since the communication pattern is the same as >>> for a global-to-local. Could someone who is familiar with PetscSF >>> comment >>> on this? >>> >> You can use PetscSFBcast for this. For each point (which will be a leaf >> in star-forest terminology), set the (rank, offset) of the owner >> ("root"). Use PetscSFSetGraph rather than PetscSFSetGraphLayout, >> because you don't have a global ordering/PetscLayout to use the latter. >> > > > -- > Richard Tran Mills, Ph.D. > Computational Earth Scientist | Joint Assistant Professor > Hydrogeochemical Dynamics Team | EECS and Earth & Planetary Sciences > Oak Ridge National Laboratory | University of Tennessee, Knoxville > E-mail: rmills(a)ornl.gov V: 865-241-3198 http://climate.ornl.gov/~**rmills<http://climate.ornl.gov/~rmills> > > -- 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] Adding DMLocalToLocalBegin/End: What about PetscSF?
by Richard Tran Mills 08 Jul '13

08 Jul '13

Hi Jed, Revisiting this now; thanks for your help so far. I understand that I can use PetscSFBcast() for this, but it is not clear to me whether there is a good way that I can make use of the "default" SF that exists (or the "default" PetscSection used to construct this). Should I just construct some other "default" for local-to-local, or is there a good way to remap the "global" vector layout onto a local one? --Richard On 6/12/13 11:25 AM, Jed Brown wrote: > Richard Tran Mills <rtm(a)eecs.utk.edu> writes: > >> PETSc developers, >> >> I am finally getting around to adding the DMLocalToLocalBegin/End routines >> that I had mentioned several months ago. Looking at >> DMGlobalToLocalBegin() in src/interface/dm/dm.c, I see that a little over >> a year ago, Matt added the use of a default 'PetscSF' ("star forest") if >> present to handle the global-to-local and local-to-global operations. I >> am not familiar with PetscSF, but I assume that it should be possible to >> use these somehow (maybe needing to add a little bit of code) for >> local-to-local operations, since the communication pattern is the same as >> for a global-to-local. Could someone who is familiar with PetscSF comment >> on this? > You can use PetscSFBcast for this. For each point (which will be a leaf > in star-forest terminology), set the (rank, offset) of the owner > ("root"). Use PetscSFSetGraph rather than PetscSFSetGraphLayout, > because you don't have a global ordering/PetscLayout to use the latter. -- Richard Tran Mills, Ph.D. Computational Earth Scientist | Joint Assistant Professor Hydrogeochemical Dynamics Team | EECS and Earth & Planetary Sciences Oak Ridge National Laboratory | University of Tennessee, Knoxville E-mail: rmills(a)ornl.gov V: 865-241-3198 http://climate.ornl.gov/~rmills

1 0

Migrate SLEPc to bitbucket
by Jose E. Roman 08 Jul '13

08 Jul '13

Now that slepc-3.4 is out, I would like to migrate the SLEPc repository to bitbucket. I think it is good to have a user experience and workflow as close to PETSc as possible. To avoid sending lots of emails to the list, I would appreciate if anyone can volunteer to help me in the process (converting from svn to git, creating users and branches, and so on). Jose

1 0

Re: [petsc-dev] nightly builds
by Jed Brown 08 Jul '13

08 Jul '13

Satish Balay <balay(a)mcs.anl.gov> writes: > We have a nightlybuild with the above CFLAGS - it gives a bunch more > warnings apart from -Wdeclaration-after-statement - so these warnings > get lost in the noise. Hmm, I get warnings about C90 not supporting complex, strings exceeding the 509 character length that ISO C compilers are required to support, and conversion of object pointer to function pointer (most of these can be fixed by eradicating the old-style Fortran function pointers).

1 0

Re: [petsc-dev] Cross-compiling flag in PETSc
by Jed Brown 08 Jul '13

08 Jul '13

Chen Shapira <chen2600(a)gmail.com> writes: > Hi, > > Could anyone tell me what's the flag that's responsible to > Cross-Compilation in petsc's configure file ? > I'm building petsc under Linux CentOS, and I would like the library > binaries to be built > for a different Linux architecture ( called Intel's uOS Linux for Xeon Phi > ), Sounds like you want to run unhosted on MIC. I don't know if anyone has tested this yet. > > I just need a way to tell PETSc that the test programs it runs during > configure, might not work naturally, since I'm cross-compiling for a > different Linux-based OS.. Use --with-batch and execute the test program on the target architecture.

1 0

Cross-compiling flag in PETSc
by Chen Shapira 08 Jul '13

08 Jul '13

Hi, Could anyone tell me what's the flag that's responsible to Cross-Compilation in petsc's configure file ? I'm building petsc under Linux CentOS, and I would like the library binaries to be built for a different Linux architecture ( called Intel's uOS Linux for Xeon Phi ), I just need a way to tell PETSc that the test programs it runs during configure, might not work naturally, since I'm cross-compiling for a different Linux-based OS.. Anyone? Thank you!

1 0

Re: [petsc-dev] Building PETSc for Intel's Xeon Phi
by Karl Rupp 07 Jul '13

07 Jul '13

Hi Chen, > Did anyone try to build PETSc for Intel's Xeon Phi (MIC architecture) ? > I'm having configure errors , if you had a chance to do it , please > write here > how you configured petsc in this case, You may use the Xeon Phi via OpenCL and ViennaCL, but the performance is rather bad because Intel's OpenCL SDK for MIC isn't mature yet. If you want to give it a try nonetheless, I can send you instructions. I'm not aware of a PETSc build for 'native' execution on the Xeon Phi. 'Offloaded' execution using OpenMP-style pragmas is not supported, as this is too vendor-specific. Best regards, Karli

1 0

Building PETSc for Intel's Xeon Phi
by Chen Shapira 07 Jul '13

07 Jul '13

Hi Everyone, Did anyone try to build PETSc for Intel's Xeon Phi (MIC architecture) ? I'm having configure errors , if you had a chance to do it , please write here how you configured petsc in this case, Thank you, Chen

1 0

Re: [petsc-dev] nightly builds
by Satish Balay 06 Jul '13

06 Jul '13

On Sat, 6 Jul 2013, Jed Brown wrote: > Barry Smith <bsmith(a)mcs.anl.gov> writes: > > > --with-c89 ? Since that is essentially what we are trying to achieve? > > Yeah, some of us have PETSC_ARCHes with > > CFLAGS='-std=c89 -pedantic -Wno-long-long' -pedantic does '-Wdeclaration-after-statement' - but not '-Werror=declaration-after-statement'. And using '-pedantic -Werror=declaration-after-statement' doesn't work as -Werror appears to get ignored..noise [but get caught in the windows build] We have a nightlybuild with the above CFLAGS - it gives a bunch more warnings apart from -Wdeclaration-after-statement - so these warnings get lost in the noise. Satish

1 0

Re: [petsc-dev] nightly builds
by Barry Smith 06 Jul '13

06 Jul '13

On Jul 6, 2013, at 8:14 PM, Jed Brown <jedbrown(a)mcs.anl.gov> wrote: > Barry Smith <bsmith(a)mcs.anl.gov> writes: > >> I doubt these options are universal for all C compilers? Since >> BuildSystem knows about all compilers it would always use the >> appropriate options when --with-c89 are passed. > > The above works for GCC, Clang, Intel, Open64, and PathScale. Some > compilers have different executable names for C89 (IBM and Sun). PGI > uses -c89. I didn't find a C89 mode for Cray. Nice to see you were able to retrieve your laptop as you got off the plane.

1 0