February 2012 - moab-dev - mailman3.cels.anl.gov

Re: [MOAB-dev] iMesh_stepIter
by James Porter 05 Mar '12

05 Mar '12

On Wed, 2012-02-22 at 15:38 -0600, Tim Tautges wrote: > What's the difference between that and getNextEntIter? If you're likely to want to step over multiple items, wouldn't > you already be using an ArrIter? I'll let Steve chime in on this, but this is part of working with structured meshes. I think Steve wants to get one element at a time, but sometimes he needs to skip ahead to get to the elements he cares about. Since doing that with getNextEntIter is an O(N) operation, it's not as good as using stepIter, which is O(1), at least for list-type sets. - Jim

3 7

r5412 - MOAB/trunk/src/moab
by xbxu＠mcs.anl.gov 29 Feb '12

29 Feb '12

Author: xbxu Date: 2012-02-29 15:23:30 -0600 (Wed, 29 Feb 2012) New Revision: 5412 Modified: MOAB/trunk/src/moab/ScdInterface.hpp Log: set local ijk values to be global ijk values for sqij partition method when running in serial Modified: MOAB/trunk/src/moab/ScdInterface.hpp =================================================================== --- MOAB/trunk/src/moab/ScdInterface.hpp 2012-02-28 16:27:09 UTC (rev 5411) +++ MOAB/trunk/src/moab/ScdInterface.hpp 2012-02-29 21:23:30 UTC (rev 5412) @@ -771,40 +771,51 @@ { // square IxJ partition - std::vector<double> pfactors, ppfactors; - for (int i = 2; i <= np; i++) - if (!(np%i)) { - pfactors.push_back(i); - ppfactors.push_back(((double)(i*i))/np); + if (np == 1) { + if (lijk) { + lijk[0] = gijk[0]; + lijk[3] = gijk[3]; + lijk[1] = gijk[1]; + lijk[4] = gijk[4]; + lijk[2] = gijk[2]; + lijk[5] = gijk[5]; + } } - + else { + std::vector<double> pfactors, ppfactors; + for (int i = 2; i <= np; i++) + if (!(np%i)) { + pfactors.push_back(i); + ppfactors.push_back(((double)(i*i))/np); + } + // ideally, Px/Py = I/J - double ijratio = ((double)(gijk[3]-gijk[0]))/((double)(gijk[4]-gijk[1])); - - unsigned int ind = std::lower_bound(ppfactors.begin(), ppfactors.end(), ijratio) - ppfactors.begin(); - if (ind && fabs(ppfactors[ind-1]-ijratio) < fabs(ppfactors[ind]-ijratio)) ind--; - - int pi = pfactors[ind]; - int pj = np / pi; - - int I = (gijk[3] - gijk[0]), J = (gijk[4] - gijk[1]); - int iextra = I%pi, jextra = J%pj, i = I/pi, j = J/pj; - int nri = nr % pi, nrj = nr / pi; -// -// pi, pj = # blocks in i, j directions -// nri, nrj = column, row of this processor's block - if (lijk) { - lijk[0] = i*nri + std::min(iextra, nri); - lijk[3] = lijk[0] + i + (nri < iextra ? 1 : 0); - lijk[1] = j*nrj + std::min(jextra, nrj);

1 0

iMesh & iGeom
by Christopher Mueller 26 Feb '12

26 Feb '12

Hi Tim/all, I may be missing something obvious or I may be trying something that's not even possible, but how does one use the iGeom functions against a mesh created with iMesh? Thanks, Chris

2 4

iMesh/MOAB serialization
by Christopher Mueller 25 Feb '12

25 Feb '12

Hi Tim/all, We will be moving mesh and Tag data via messaging between processes from memory to memory. We see this as a need to marshall the imesh/moab structures in whole and in part over a message stream. I was wondering if you could help me understand how we would go about implementing our own serialization/deserialization strategy. The src/io directory seems to be the repository of existing I/o strategies; where are the interfaces for integrating a new I/O mechanism? The messaging protocol supports the typical set container classes, and handle based referencing. Also are there utilities for slicing (writing/reading portions of) the mesh? Thanks in advance, Chris

2 1

r5410 - in MOAB/trunk: itaps itaps/imesh src src/io src/moab test tools/vtkMOABReader
by tautges＠mcs.anl.gov 23 Feb '12

23 Feb '12

Author: tautges Date: 2012-02-23 13:26:20 -0600 (Thu, 23 Feb 2012) New Revision: 5410 Modified: MOAB/trunk/itaps/iBase_f.h.in MOAB/trunk/itaps/imesh/MBIter.hpp MOAB/trunk/itaps/imesh/iMesh_MOAB.cpp MOAB/trunk/itaps/imesh/iMesh_extensions.h MOAB/trunk/itaps/imesh/iMesh_f.h MOAB/trunk/src/Core.cpp MOAB/trunk/src/io/WriteDamsel.cpp MOAB/trunk/src/moab/Core.hpp MOAB/trunk/src/moab/Interface.hpp MOAB/trunk/test/coords_connect_iterate.cpp MOAB/trunk/tools/vtkMOABReader/vtkMOABReader.cxx Log: Mostly some imesh/ibase changes: - moved some arguments around in connect_iterate and coords_iterate functions - moved iterator #define's to iBase_f.h.in from iMesh_f.h - implemented new imesh extension functions iMesh_connectIterate and iMesh_coordsIterate, which provide direct access to memory used to store connectivity/coordinates, resp. - added an argument to get the # vertices/entity to connect_iterate - implemented new iMesh extension function iMesh_initEntArrIterRec, which is an iterator that recurses into contained entity sets; all other iterator-related functions operate as before, but with recursive access to sets (this also required changes to implementation in MBIter, but since MOAB's underlying functions already had a recursive flag, this just involved storing/sending the flag to the MOAB function - in vtkMOABReader, implemented more careful determination of vtk cell type, and support quadratic cell types in some cases Passes make check. Modified: MOAB/trunk/itaps/iBase_f.h.in =================================================================== --- MOAB/trunk/itaps/iBase_f.h.in 2012-02-20 22:59:28 UTC (rev 5409) +++ MOAB/trunk/itaps/iBase_f.h.in 2012-02-23 19:26:20 UTC (rev 5410) @@ -33,6 +33,8 @@ #define iBase_EntityHandle IBASE_HANDLE_T #define iBase_EntitySetHandle IBASE_HANDLE_T #define iBase_TagHandle IBASE_HANDLE_T +#define iBase_EntityIterator IBASE_HANDLE_T +#define iBase_EntityArrIterator IBASE_HANDLE_T #endif Modified: MOAB/trunk/itaps/imesh/MBIter.hpp =================================================================== --- MOAB/trunk/itaps/imesh/MBIter.hpp 2012-02-20 22:59:28 UTC (rev 5409) +++ MOAB/trunk/itaps/imesh/MBIter.hpp 2012-02-23 19:26:20 UTC (rev 5410) @@ -15,13 +15,15 @@ iMesh_EntityTopology entTopo; EntityHandle entSet; int arrSize; + bool isRecursive; public: iBase_EntityArrIterator_Private( iBase_EntityType type, iMesh_EntityTopology topology, EntityHandle set, - int array_size ) - : entType(type), entTopo(topology), entSet(set), arrSize(array_size) + int array_size, + bool recursive = false) + : entType(type), entTopo(topology), entSet(set), arrSize(array_size), isRecursive(recursive) {} virtual ~iBase_EntityArrIterator_Private() {} @@ -115,8 +117,9 @@ MBIter( iBase_EntityType type, iMesh_EntityTopology topology, EntityHandle set, - int array_size ) - : iBase_EntityArrIterator_Private( type, topology, set, array_size ), + int array_size, + bool recursive = false) + : iBase_EntityArrIterator_Private( type, topology, set, array_size, recursive ), iterPos(iterData.end()) {} ~MBIter() {} @@ -144,15 +147,15 @@ if (entTopo == iMesh_SEPTAHEDRON) result = MB_SUCCESS;

1 0

iMesh_stepIter
by James Porter 22 Feb '12

22 Feb '12

I notice that iMesh_stepIter (an iMesh extension function) applies only to array iterators. Should we make one for non-array iterators too? I think that would be more consistent, and it would help out Steve, who wants to use this function from Python. - Jim

1 0

exchange entity sets
by Lorenzo Alessio Botti 22 Feb '12

22 Feb '12

Thanks, and yes, what you propose is an effective workaround! I'll rely on this for now, but I'd like to know if you need and/or plan to introduce something like exchange_ghost_sets(...) at some point. Such a thing would allow me to save a lot of ghost cells. Best regards. Lorenzo > On Feb 22, 2012, at 6:53 PM, Hong-Jun Kim wrote: > >> "exchange_owned_meshs" exchanges shared entities and sets between processors. >> You should give exchange processor list and entity (including set) list for each processor in the exchange processor list to the function. >> The tag information is also exchanged by the function automatically but exchanged entities have shared status not ghost status. >> >> As an alternative, after exchanging ghost cells, I think, you can exchange "PARALLEL_PARTITION" tag, then make proper entity set by the partition information. >> Thanks. >> >> ----------------------------- >> Hong-Jun Kim >> Post-doc researcher >> MCS, Argonne National Laboratory >> 9700 S. Cass Ave. B240/R2147 >> Argonne, IL 60439 >> 630-252-4791 >> hongjun(a)mcs.anl.gov >> ----------------------------- >> >> ----- Original Message ----- >> From: "Lorenzo Alessio Botti" <ihabiamx(a)yahoo.it> >> To: MOAB-dev(a)mcs.anl.gov >> Sent: Wednesday, February 22, 2012 11:30:52 AM >> Subject: [MOAB-dev] exchange entity sets >> >> Hi all, >> I'd like to know if there is a way to communicate a whole entity set across processes. >> My actual goal is to create a layer of ghost "PARALLEL_PARTITION" entity sets instead of a layer of ghost cells. >> >> I know that it is possible to create an arbitrary extended layer of ghost cells but then I don't know how to recover >> the PARALLEL_PARTITION informations. >> >> I guess that exchange_owned_mesh/s(...) functions might help but I might be wrong. What's the purpose of these functions? >> >> Thanks for help. >> Lorenzo >> >

1 0

exchange entity sets
by Lorenzo Alessio Botti 22 Feb '12

22 Feb '12

Hi all, I'd like to know if there is a way to communicate a whole entity set across processes. My actual goal is to create a layer of ghost "PARALLEL_PARTITION" entity sets instead of a layer of ghost cells. I know that it is possible to create an arbitrary extended layer of ghost cells but then I don't know how to recover the PARALLEL_PARTITION informations. I guess that exchange_owned_mesh/s(...) functions might help but I might be wrong. What's the purpose of these functions? Thanks for help. Lorenzo

2 1

r5408 - MOAB/trunk/itaps/imesh
by jvporter＠wisc.edu 20 Feb '12

20 Feb '12

Author: jvporter Date: 2012-02-20 11:50:58 -0600 (Mon, 20 Feb 2012) New Revision: 5408 Modified: MOAB/trunk/itaps/imesh/MBIter.hpp Log: Inline the step_iterator functions to fix builds (and they should only get called from one place anyway) Modified: MOAB/trunk/itaps/imesh/MBIter.hpp =================================================================== --- MOAB/trunk/itaps/imesh/MBIter.hpp 2012-02-17 23:50:17 UTC (rev 5407) +++ MOAB/trunk/itaps/imesh/MBIter.hpp 2012-02-20 17:50:58 UTC (rev 5408) @@ -58,6 +58,7 @@ // for random-access iterators (vectors and Ranges) so that they perform better. template <typename T> +inline ErrorCode step_iterator(T &curr, const T &end, int num_steps, bool &at_end) { if (0 > num_steps) return MB_FAILURE; @@ -71,6 +72,7 @@ } template <typename T> +inline ErrorCode step_iterator(typename std::vector<T>::const_iterator &curr, const typename std::vector<T>::const_iterator &end, int num_steps, bool &at_end) @@ -87,6 +89,7 @@ return MB_SUCCESS; } +inline ErrorCode step_iterator(Range::const_iterator &curr, const Range::const_iterator &end, int num_steps, bool &at_end)

1 0

buildbot failure in moab-par: failed compile test
by moab-dev＠mcs.anl.gov 20 Feb '12

20 Feb '12

1 0