News

FMDB 1.4.0 Released

Added by Seegyoung Seol about 5 years ago

Announcing FMDB v1.4.0 release with focus on enrichment of multi-part related functions & static partitioning performance improvement

  • Click here for full release note
[NEW/IMPROVED FEATURES]
  • FMDB_Mesh_LoadFromFile - mesh loading time reduced to one third (1/3) of the original
      - prev: the same migration algorithm was used for both static and dynamic partitioning
      - new: migration algorithm developed for static partitioning
  • FMDB_Mesh_Merge
      - merging N part mesh (migrating all parts to part 0 on master process)
  • (#405) FMDB_Vtx_GetLoadPartOrderPair
      - For a vertex, getting loading part id and read-in order from mesh file
  • weight control for dynamic partitioning through Zoltan/ParMETIS
      - FMDB_Ent_Set/Get/DelWeight - set/get/delete weight of individual entity.
      - FMDB_Topo_Set/Get/DelWeight - set/get/delete weight of all entities of topology.
  • FMDB_Mesh_DspStat - mesh load imbalance ratio computation improved to be consisitent with Zoltan/ParMETIS
       - prev: mesh load imbalance ratio computed based on # partition object entities (e.g. regions in 3D)
       - new:  mesh load imbalance ratio computed based on weights per process/part
[ERROR FIX]
  • (#416) FMDB_Mesh_Verify collective status report
  • (#415) FMDB_Mesh_Verify for optimized calls
  • (#414) FMDB does not compile on BG/P
[FASTMath INTERFACE COMPLIANCE]
  • iMesh.h pass rate: 100%
  • iMeshP.h passrate: 99%
  • iMeshP_ext.h pass rate: not applicable

FMDB 1.3.9 Released

Added by Seegyoung Seol about 5 years ago

Announcing FMDB v1.3.9 release with focus on multiple-part & p-set migration/partitioning support

  • Please be advised that the new features are supported only through FMDB.h.
  • Please obtain the source (SCUtil, FMDB) from FMDB repo or tarballs":http://www.scorec.rpi.edu/FMDB/source.html
  • Click here for full release note
[NEW FEATURES]
  • (#332) setting ZOLTAN parameters (LB_Method, LB_Approach, imbalance tolerance) for static/dynamic mesh partitioning
    - the user can control partitioning parameters for mesh partitioning
    - static mesh partitioning is to load and distribute serial mesh in parallel
    - dynamic mesh partitioning is local/global mesh re-distribution during parallel mesh modification
  • (#279) parallel mesh file i/o with p-set
    - p-set is one kind of ordered entity set which is assigned to a single part so can be designated as "partition object (a basic unit to assign destination part in migration)"
  • (#303) setting/getting #parts per process
    - The application can change # parts per process dynamically
  • (#328) getting neighboring process/parts
    - A process i neighbors process j if they share mesh entities on part boundary between parts located in process i and process j.
    - A part x neighbors part y if they share mesh entities on part boundary
  • displaying mesh statistics (click here for sample)
    - global statistics with respect to the count (#p-set, #entity) per process. E.g. mesh load imbalance ratio, avg/max/min #p-set, #entity
    - global statistics with respect to the count (#p-set, #entity) per part
    - local (per process) statistics
[ERROR FIX]
  • (#392) multi-part partitioning support with P-set
  • (#387) hang during migration after partitionedmesh loading

FMDB 1.3.8 Released

Added by Seegyoung Seol over 5 years ago

Announcing FMDB v1.3.8 release with focus on

  • New Feature (#303) process grouping
    - each process group loads individual mesh onto multi-part and modify/migrate/load_balance/export the mesh independently
  • Error fix (#373) failures with iMesh fortran tests
  • Error fix (#320) FMDB_FMDB_LoadFromFile hangs with wrong input
  • Error fix (#323) multi-part file i/o
  • Error fix (#339) multi-part partitioning/migration

Please obtain

For help or more information, please refer to User's Guide and send feedback to .

FMDB 1.3.7 Released

Added by Seegyoung Seol over 5 years ago

Announcing FMDB v1.3.7 release with focus on

  • API for manual part boundary update
  • iMeshP compliance (current pass rate: 76%)
  • Error fix [valgrind error with iMesh perfTest example run]
  • Error fix (#323) [Incorrect mesh size/entities' owning part info for multi-part distributed mesh]

Please obtain

  • Tarballs (GMI, SCUtil, FMDB, IPComMan) and build instructions from FMDB web site
  • Click here for full release note

For help, please refer to User's Guide and send feedback to .

FMDB 1.3.6 Released

Added by Seegyoung Seol over 5 years ago

Announcing FMDB v1.3.6 release with focus on

  • Valgrind memory errors with iMesh compliance test
  • Tagging import/export functionality for VTK files
  • "Multiple parts per process" support in global load balancing through FMDB API
  • Error fixes (#301) [Writing VTK files from FMDB API]

Please obtain

For help, please refer to User's Guide and send feedback to

FMDB 1.3.5 and GMI 1.0.1 Released

Added by Seegyoung Seol almost 6 years ago

Announcing FMDB v1.3.5 and GMI v1.0.1 release with focus on

  • (#305) failure with iZoltan-parallel
  • (#299) entity creation/search
  • (#294) sample mesh folder enriched with the mesh which mimics what MOAB iMeshP test constructs
  • (#203) exporting iGeom-Defs.inc

Please obtain

For help, please refer to User's Guide and send feedback to

FMDB v1.3.4 Released

Added by Seegyoung Seol about 6 years ago

Announcing FMDB v1.3.4 release with focus on

  • tagging (api, performance, code reusability)
  • memory leak error (#295)
  • make check error (#302)

Please obtain

For help, please refer to User's Guide and send feedback to

FMDB Weekly #17 (Apr 25 - Apr 29)

Added by Seegyoung Seol over 6 years ago

[ Week 17 Activities ]

FMDB v1.3

  • constraint: single part per mesh and single partition
  • (on-going) entity set and entity group integration
    - mesh migration
    - distributed entity set file I/O
  • (on-going) FMDB.h development
    - entity creation/search
    - higher-order node, etc.
  • (on-going) parallel code improvement (cleanup, performance, etc.)

iMesh/iMeshP

  • (on-going) swapping issue f/u (crash with iterator)
  • (on-going) iMeshP_FMDB development (itaps spec compliance)
  • (on-going) build issues f/u (fortran support, name mangling, linking etc.)

SCUtil/iBase

  • employing SCOREC error management scheme (fault tolerance with error code)

GRM/iRel

  • NONE

[ Planned Week 18 Activities ]

FMDB v1.3

  • MeshAdapt support centric release to SCOREC
  • FMDB.h development
  • entity set/entity group integration
  • parallel code improvement (cleanup, performance, etc.)

iMesh/iMeshP

  • swapping issue f/u
  • iMeshP_FMDB development
  • build issue f/u

GMI/iGeom

  • NONE

SCUtil/iBase

  • employing SCOREC error management scheme

GRM/iRel

  • NONE

[ Independent Study ]

  • Template Meta Programming
  • Design Pattern

[ Remarks ]

  • FMDB team meeting: Wednesday 11am-12pm
  • ITAPS telecon: Thursday 12pm-1pm
  • iMeshP compliance test verification with FMDB pended due to incomplete implementation such as ghosting & multiple part per process
  • entity group migration with "splitted" entity data stream won't be considered to put into the repository
    Instead, better algorithm for migration procedure will be investigated.

FMDB Weekly #15 (Apr 11 - April 15)

Added by Seegyoung Seol over 6 years ago

[ Week 15 Activities ]

FMDB v1.3

  • constraint: single part per mesh and single partition
  • (on-going) entity set and entity group integration
    - mesh migration
    - distributed entity set file I/O
  • (on-going) FMDB.h development
    - entity creation/search
    - higher-order node, etc.
  • (on-going) parallel code improvement (cleanup, performance, etc.)

iMesh/iMeshP

  • (on-going) iMeshP_FMDB development
  • (on-going) build issues (fortran support, name mangling, linking etc.)

SCUtil/iBase

  • NONE

GRM/iRel

  • NONE

[ Planned Week 16 Activities ]

FMDB v1.3

  • FMDB.h development
  • entity set/entity group integration
  • parallel code improvement (cleanup, performance, etc.)

iMesh/iMeshP

  • iMeshP_FMDB development

GMI/iGeom

  • MeshModel review and improvement

SCUtil/iBase

  • NONE

GRM/iRel

  • NONE

[ Independent Study ]

  • C++ Templates
  • Design Pattern

[ Remarks ]

  • FMDB team meeting: Wednesday 11am-12pm
  • ITAPS telecon: Thursday 12pm-1pm
  • iMeshP compliance test verification with FMDB pended due to incomplete implementation such as ghosting & multiple part per process
  • entity group migration with "splitted" entity data stream won't be considered to put into the repository
    Instead, better algorithm for migration procedure will be investigated.

FMDB Weekly #14 (Apr 4 - April 8)

Added by Seegyoung Seol over 6 years ago

[ Week 14 Activities ]

FMDB

  • constraint: single part per mesh
  • (on-going) entity set and entity group integration
    - mesh migration
    - distributed entity set file I/O
  • (on-going) ghosting
    - N-layer creation
    - layer accumulation
    - mesh verification
    - thesis writeup
  • (on-going) FMDB.h development
    - missing api functions (region creation, entity search, etc)
  • (on-going) User's Guide update

iMesh/iMeshP

  • (completed) iMesh_FMDB development
    - iMesh compliance test passed
  • (on-going) memory leak w/ valgrind
  • (on-going) iMeshP_FMDB development
  • (on-going) build issues (fortran support, name mangling, etc.)

SCUtil/iBase

  • NONE

GRM/iRel

  • NONE

[ Planned Week 15 Activities ]

FMDB v1.2

  • FMDB.h development
  • entity set/entity group integration
  • ghosting development
  • parallel code improvement (cleanup, performance, etc.)

iMesh/iMeshP

  • memory leak debugging w/ valgrind (serial)
  • iMeshP_FMDB development

GMI/iGeom

  • autotool support (iGeom-Defs.inc, fortran, etc.)
  • iGeom status (implementation/testing) review

SCUtil/iBase

  • NONE

GRM/iRel

  • NONE

[ Independent Study ]

  • C++ Templates
  • Design Pattern

[ Remarks ]

  • FMDB team meeting: Wednesday 11am-12pm
  • ITAPS telecon: Thursday 12pm-1pm
  • iMeshP compliance test verification with FMDB pended due to incomplete implementation such as ghosting & multiple part per process
  • entity group migration with "splitted" entity data stream won't be considered to put into the repository
    Instead, better algorithm for migration procedure will be investigated.

1 2 3 4 Next »

Also available in: Atom