Merged in knepley/feature-plex-pullback (pull request #1632)

Plex: Implemented pullback/pushforward

Approved-by: BarryFSmith <bsmith@mcs.anl.gov>

Fix http -> https for Hass diagram URL and remove stray a inside (

Commit-type: documentation
Reported-by: Junchao Zhang <jczhang@mcs.anl.gov>

plexgenerate.c edited online with Bitbucket

Fixed out-dated URLS using make checkbadURLS REPLACE=1

One manual fix of https://en.wikipedia.org/wiki/Buckingham_π_theorem since the automatic tool dropped the first _ for some reason
Manual fix of

Fixed out-dated URLS using make checkbadURLS REPLACE=1

One manual fix of https://en.wikipedia.org/wiki/Buckingham_π_theorem since the automatic tool dropped the first _ for some reason
Manual fix of //faq.html since the automatic tool incorrectly replaced /faq.html to //faq.html in some locations

These are largely updates of http to https but have a variety of other website location changes

Once this branch is in master individual developers can run make checkbadURLS and fix URLs related to their projects manually

Commit-type: bug-fix, documentation

show more ...

DMPlexPartitionLabelInvert: loop only over the neighbors

Update output files since the previous code was accessing the label values in an ordered way

DMPlexPartitionLabelInvert: added info on algorithm used

DMPLEX: fix improper usage of STDOUT_WORLD

DMPLEX: fix coding standards

PetscSF: added logging support for functions used by DMPlex

DMPLEX: speed up DMPlexPartitionLabelClosure_Private

instead of computing the closure for each point, recursively add the cone
with this optimization, using a 128^3 simplex mesh an 16 processes on m

DMPLEX: speed up DMPlexPartitionLabelClosure_Private

instead of computing the closure for each point, recursively add the cone
with this optimization, using a 128^3 simplex mesh an 16 processes on my workstation -> from 7 seconds to 1.2

Change the name of the function and its prototype since the label and ranks arguments are unused

show more ...

DMPLEX: add few missing events to help optimizing the code

Merged in dalcinl/fix-io-gmsh (pull request #1660)

Binary I/O And Gmsh DMPlex Reader

Approved-by: Matthew Knepley <knepley@gmail.com>
Approved-by: BarryFSmith <bsmith@mcs.anl.gov>

DMPlex test ex11: Remove redundant test and use simple partitioner

DMPlex: Update Gmsh reader to set hybrid bounds by default

If the reader is asked to return an interpolated mesh and the mesh
has tri+quad or tetra+wedge cells, you most likely want the output
DMPle

DMPlex: Update Gmsh reader to set hybrid bounds by default

If the reader is asked to return an interpolated mesh and the mesh
has tri+quad or tetra+wedge cells, you most likely want the output
DMPlex to have the hybrid bounds set, so make it default behaviour.

If for whatever reason the users do not want the hybrid bounds set,
they can still use `-dm_plex_gmsh_hybrid 0`.

show more ...

DMPLEX: remove debug view from options

fix comment

DMPlex: Update Gmsh reader to read periodic section by default if present

Binary I/O: Various fixes and enhancements

* Change PetscBinary[Synchronized]Read() to allow for reading
items up to EOF without erroring
* Fix PetscViewerBinaryRead() to properly return the numbe

Binary I/O: Various fixes and enhancements

* Change PetscBinary[Synchronized]Read() to allow for reading
items up to EOF without erroring
* Fix PetscViewerBinaryRead() to properly return the number of items read
* Fix PetscViewerRead() to prevent infinite loops at EOF
* Various fixes for PetscDataType PETSC_COMPLEX

show more ...

Partitioner: Prefer default type PARMETIS, then PTSCOTCH, then CHACO

DMPlex test ex11: Use simple partitioner and test gmsh/exodusii meshes

DMPLEX tests ex1: activate test for overlap distribution of non-conforming mesh

This got fixed by Lisandro in 31bc6364794885a9add836c67d7739665f4e4535

PetscPartitioner: Move option for randomSeed to where it belongs

Move PtScotch internal data structure to dmpleximpl.h
Update output files

DMPlex test ex1: add tests for load balance and redistribution

PetscPartitioner: initial vertex weights

In the case of overlapped meshes from p4est,
the assumption of cells at the end is wrong.
This at least silence the uninitialized values warning from valgrind

DMPLEX: add support to compute some statistics for parallel meshes

compute edgeCut, node-aware edgeCut and owned cell balance

DMPLEX: add support to compute CSR via matrix operations