Merge branch 'development'

.travis.yml

@@ -27,7 +27,7 @@ addons:
 install:
   - pip install sphinx sphinx_rtd_theme breathe
 
-script: bash ./build_and_deploy.sh
+script: travis_wait bash ./build_and_deploy.sh
 
 env:
   global:

Docs/sphinx_documentation/source/AmrLevel_Chapter.rst

@@ -18,14 +18,23 @@ than the classes created for the Advection_AmrCore tutorial.
    and attributes.
 
 Many of our mature, public application codes contain derived classes that
-inherit directly from :cpp:`AmrLevel`. These include our compressible
-astrophysics code, CASTRO, (available in the AMReX-Astro/Castro github
-repository) and our computational cosmology code, Nyx (available in the
-AMReX-Astro/Nyx github repository) .  Our incompressible Navier-Stokes code,
-IAMR (available in the AMReX-codes/IAMR github repository) has a pure virtual
-class called :cpp:`NavierStokesBase` that inherits from :cpp:`AmrLevel`, and an
-additional derived class :cpp:`NavierStokes`.  Our low Mach number combustion
-code PeleLM (not yet public) also inherits from :cpp:`NavierStokesBase`.
+inherit directly from :cpp:`AmrLevel`. These include:
+
+-  The :cpp:`Castro` class in our compressible astrophysics code, CASTRO,
+   (available in the AMReX-Astro/Castro github repository)
+
+-  The :cpp:`Nyx` class in our computational cosmology code, Nyx
+   (available in the AMReX-Astro/Nyx github repository).
+
+-  Our incompressible Navier-Stokes code, IAMR
+   (available in the AMReX-codes/IAMR github repository) has a pure virtual
+   class called :cpp:`NavierStokesBase` that inherits from :cpp:`AmrLevel`, and an
+   additional derived class :cpp:`NavierStokes`.
+
+-  Our low Mach number combustion code PeleLM
+   (available in the AMReX-Combustion/PeleLM github repository)
+   contains a derived class :cpp:`PeleLM` that also inherits from
+   :cpp:`NavierStokesBase` (but does not use :cpp:`NavierStokes`).
 
 The tutorial code in ``amrex/Tutorials/Amr/Advection_AmrLevel`` gives a simple
 example of a class derived from :cpp:`AmrLevel` that can be used to solve the

Docs/sphinx_documentation/source/BuildingAMReX.rst

@@ -242,7 +242,6 @@ use homebrew's mpich. Normally is it fine to simply install its binaries:
 ``brew install mpich``. But if you are experiencing problems, we suggest
 building mpich using homebrew's gcc: ``brew install mpich --cc=gcc-8``.
 
-
 .. _sec:build:lib:
 
 Building libamrex
@@ -634,3 +633,26 @@ You can tell CMake to look for the AMReX library in non-standard paths by settin
 More details on ``find_package`` can be found 
 `here <https://cmake.org/cmake/help/v3.14/command/find_package.html>`_.
 
+.. _sec:build:windows:
+
+AMReX on Windows
+================
+
+The AMReX team does development on Linux machines, from desktop workstations to supercomputers. Many people also use AMReX on Macs without issues.  
+
+We do not officially support AMReX on Windows.  However, we believe there are no fundamental issues for making it work on Windows.  
+AMReX mostly uses standard C++11,  and there are only a few places that are UNIX/Linux specific. These are:
+
+(1) File system:  We use some of the POSIX standard functions for operations like making a new directory, detecting if a file exists, etc.  
+C++17 now has a filesystem library that should work on any platform.  AMReX does not require C++17, but we are happy to provide a C++17 support for the file system part.
+
+(2) Signal handling:  We use POSIX handling when floating point exceptions, segmentation faults, etc. happen.  
+This capability allows us to print a backtrace of what leads to the error and is very useful for debugging but not required for using AMReX.  
+Some of the POSIX handling is platform-dependent, and Windows does seem to have this capability.  If you need it, it should not be hard for you to make it work on Windows.
+
+(3) Memory profiling:  This is an optional feature in AMReX that is not enabled by default.  
+It reads memory system information from the OS to give us a summary of our memory usage.
+
+One other caveat is regarding the size of ``long``, which is 4 on Windows and 8 on other 64-bit systems.  
+This might cause integer overflow for really big runs (unlikely on Windows desktops as opposd to clusters). 
+If this becomes an issue, please let us know and we could define amrex::Long that would be guaranteed to be 64 bits.

Docs/sphinx_documentation/source/Introduction.rst

@@ -22,7 +22,7 @@ Key features of AMReX include:
 
 -  Support for embedded boundary (cut cell) representations of complex geometries
 
--  Parallelization via flat MPI, OpenMP, hybrid MPI/OpenMP, or MPI/MPI
+-  Parallelization via flat MPI, OpenMP, hybrid MPI/OpenMP, hybrid MPI/CUDA, or MPI/MPI
 
 -  Parallel I/O