Merge 8757910 into c06bb71

LLNL · Jun 25, 2018 · b50898a · b50898a
2 parents c06bb71 + 8757910
commit b50898a
Show file tree

Hide file tree

Showing 152 changed files with 3,004 additions and 1,213 deletions.
diff --git a/CMakeLists.txt b/CMakeLists.txt
@@ -14,8 +14,8 @@
 ###############################################################################
 # Set version number
 set(RAJA_VERSION_MAJOR 0)
-set(RAJA_VERSION_MINOR 5)
-set(RAJA_VERSION_PATCHLEVEL 3)
+set(RAJA_VERSION_MINOR 6)
+set(RAJA_VERSION_PATCHLEVEL 0)
 
 if (RAJA_LOADED AND (NOT RAJA_LOADED STREQUAL "${RAJA_VERSION_MAJOR}.${RAJA_VERSION_MINOR}.${RAJA_VERSION_PATCHLEVEL}"))
   message(FATAL_ERROR "You are mixing RAJA versions. Loaded is ${RAJA_LOADED}, expected ${RAJA_VERSION_MAJOR}.${RAJA_VERSION_MINOR}.${RAJA_VERSION_PATCHLEVEL}")
@@ -83,8 +83,7 @@ if (NOT RAJA_LOADED)
     src/AlignedRangeIndexSetBuilders.cpp
     src/DepGraphNode.cpp
     src/LockFreeIndexSetBuilders.cpp
-    src/MemUtils_CUDA.cpp
-    src/ThreadUtils_CPU.cpp)
+    src/MemUtils_CUDA.cpp)
 
   set (raja_depends)
 

diff --git a/README.md b/README.md
@@ -1,4 +1,4 @@
-RAJA v0.5.3
+RAJA v0.6.0
 ============
 
 [![Build Status](https://travis-ci.org/LLNL/RAJA.svg?branch=develop)](https://travis-ci.org/LLNL/RAJA)

diff --git a/RELEASE b/RELEASE
@@ -1,6 +1,6 @@
 *******************************************************************************
 
-RAJA: ................................, version 0.5.3
+RAJA: ................................, version 0.6.0
 
 Copyright (c) 2016-17, Lawrence Livermore National Security, LLC. 
 Produced at the Lawrence Livermore National Laboratory.

diff --git a/docs/sphinx/user_guide/app_considerations.rst b/docs/sphinx/user_guide/app_considerations.rst
@@ -0,0 +1,21 @@
+.. ##
+.. ## Copyright (c) 2016-18, Lawrence Livermore National Security, LLC.
+.. ##
+.. ## Produced at the Lawrence Livermore National Laboratory
+.. ##
+.. ## LLNL-CODE-689114
+.. ##
+.. ## All rights reserved.
+.. ##
+.. ## This file is part of RAJA.
+.. ##
+.. ## For details about use and distribution, please read RAJA/LICENSE.
+.. ##
+
+.. _app-considerations-label:
+
+***************************
+Application Considerations
+***************************
+
+.. warning:: **This section is a work-in-progress!!**
diff --git a/docs/sphinx/user_guide/conf.py b/docs/sphinx/user_guide/conf.py
@@ -58,17 +58,17 @@
 
 # General information about the project.
 project = u'RAJA'
-copyright = u'2017'
+copyright = u'2016-2018'
 author = u'LLNS'
 
 # The version info for the project you're documenting, acts as replacement for
 # |version| and |release|, also used in various other places throughout the
 # built documents.
 #
 # The short X.Y version.
-version = u'0.5'
+version = u'0.6'
 # The full version, including alpha/beta/rc tags.
-release = u'0.5.3'
+release = u'0.6.0rc3'
 
 # The language for content autogenerated by Sphinx. Refer to documentation
 # for a list of supported languages.
@@ -138,7 +138,7 @@
 # Theme options are theme-specific and customize the look and feel of a theme
 # further.  For a list of options available for each theme, see the
 # documentation.
-#html_theme_options = {}
+#html_theme_options = {'codebgcolor': 'lightgrey'}
 
 # Add any paths that contain custom themes here, relative to this directory.
 #html_theme_path = []

diff --git a/docs/sphinx/user_guide/config_options.rst b/docs/sphinx/user_guide/config_options.rst
@@ -14,9 +14,9 @@
 
 .. _configopt-label:
 
-***********************
-Configuration Options
-***********************
+****************************
+Build Configuration Options
+****************************
 
 RAJA uses `BLT <https://github.com/LLNL/blt>`_, a CMake-based build system .
 In :ref:`getting_started-label`, we described how to run CMake to configure

diff --git a/docs/sphinx/user_guide/contributing.rst b/docs/sphinx/user_guide/contributing.rst
@@ -115,5 +115,5 @@ is used on a wide variety of systems with a number of configurations, and adding
 new tests helps ensure that all features work as expected across these
 environments.
 
-RAJA's tests are all in the ``test`` directory and are split up by backend and
-feature.
+RAJA's tests are all in the ``test`` directory and are split up by programming
+model back-end and feature.
diff --git a/docs/sphinx/user_guide/feature/atomic.rst b/docs/sphinx/user_guide/feature/atomic.rst
@@ -18,16 +18,22 @@
 Atomics
 ========
 
-To avoid race conditions at specific memory locations, RAJA provides 
-portable atomic operations, which are described in this section.
+RAJA provides portable atomic operations that can be used to update values
+at arbitrary memory locations while avoiding data races. They are described
+in this section.
+
+A complete working example code that shows RAJA atomic usage can be found in 
+:ref:`atomichist-label`.
+
+.. note:: * All RAJA atomic operations are in the namespace ``RAJA::atomic``.
 
 -----------------
 Atomic Operations
 -----------------
 
-.. note:: * All RAJA atomic operations are in the namespace ``RAJA::atomic``.
-          * Each RAJA atomic operation is templated on an *atomic policy*.
-            Atomic policies are described in the next section.
+RAJA atomic support includes a range of the most common atomic operations.
+
+.. note:: * Each RAJA atomic operation is templated on an *atomic policy*.
           * Each of methods below returns the value of the potentially modified
             argument (i.e., \*acc) immediately before the atomic operation is 
             applied.
@@ -60,8 +66,6 @@ Increment/decrement
 
 * ``atomicDec< atomic_policy >(T* acc, T compare)`` - Subtract 1 from \*acc if \*acc != 0 and \*acc <= compare, else set \*acc to compare.
 
-* ``atomicDec< atomic_policy >(T* acc, T compare)`` - Subtract 1 from \*acc if \*acc != 0 and \*acc <= compare, else set \*acc to compare.
-
 ^^^^^^^^^^^^^^^^^^^^
 Bitwise atomics
 ^^^^^^^^^^^^^^^^^^^^
@@ -97,7 +101,7 @@ a integral sum on a CUDA GPU device::
   RAJA::forall< RAJA::cuda_exec >(RAJA::RangeSegment(0, N), 
     [=] RAJA_DEVICE (RAJA::Index_type i) {
 
-    RAJA::atomic::atomicAdd< RAJA::cuda_atomic >(sum, 1);
+    RAJA::atomic::atomicAdd< RAJA::cuda_atomic >(&sum, 1);
 
   });
 
@@ -107,32 +111,33 @@ After this operation, 'sum' will be equal to 'N'.
 AtomicRef
 ^^^^^^^^^^^^^^^^^^^^
 
-RAJA also provides an atomic interface similar to C++ 'std::atomic', but for 
-arbitrary memory locations. The class ``RAJA::atomic::AtomicRef`` provides
-an object-oriented interface to the atomic methods described above. For 
-example, after the following operations:: 
+RAJA also provides an atomic interface similar to C++ 'std::atomic', but which
+works for arbitrary memory locations. The class ``RAJA::atomic::AtomicRef`` 
+provides an object-oriented interface to the atomic methods described above. 
+For example, after the following operations:: 
 
   double val = 2.0;
-  RAJA::atomic::AtomicRef<double,  RAJA::seq_atomic > sum(&val);
+  RAJA::atomic::AtomicRef<double,  RAJA::omp_atomic > sum(&val);
 
   sum++;
   ++sum;
   sum += 1.0; 
 
 the value of 'val' will be 5.
 
-Note that the operators that the 'AtomicRef' class provide return the object
-that holds the address of the data given passes the constructor. If you need 
-to keep the original value of the data before the atomic call, you need to 
-use the atomic methods listed above.
+Note that the operations provided by the 'AtomicRef' class return the object
+that holds the address of the data given to the constructor. It will likely 
+change with each atomic update call. If you need to keep the original value 
+of the data before an atomic call, you need to use the atomic methods described 
+earlier and not the ``RAJA::atomic::AtomicRef`` interface.
 
 ---------------
 Atomic Policies
 ---------------
 
 .. note:: * All RAJA atomic policies are in the namespace ``RAJA::atomic``.
           * There are no RAJA atomic policies for TBB (Intel Threading Building 
-            Blocks) execution contexts.
+            Blocks) execution contexts currently.
 
 * ``seq_atomic``     - Policy for use in sequential execution contexts, primarily for consistency with parallel policies. Note that sequential atomic operations are not protected and will likely produce incorrect results when used in a parallel execution context.
 
@@ -149,14 +154,11 @@ For example, we could use the 'auto_atomic' policy in the example above::
   RAJA::forall< RAJA::cuda_exec >(RAJA::RangeSegment seg(0, N), 
     [=] RAJA_DEVICE (RAJA::Index_type i) {
 
-    RAJA::atomic::atomicAdd< RAJA::auto_atomic >(sum, 1);
+    RAJA::atomic::atomicAdd< RAJA::auto_atomic >(&sum, 1);
 
   });
 
 Here, the atomic operation knows that it is used within a CUDA execution 
 context and does the right thing. Similarly, if the 'forall' method used 
 an OpenMP execution policy, the OpenMP version of the atomic operation 
 would be used.
-
-A complete working code that shows RAJA atomic usage can be found in 
-``<build-dir>/examples/ex8-pi-reduce_vs_atomic.cpp``. 
diff --git a/docs/sphinx/user_guide/feature/index.rst b/docs/sphinx/user_guide/feature/index.rst