Merge pull request #2811 from ComputationalRadiationPhysics/release-0…

….4.2 Release 0.4.2
ComputationalRadiationPhysics · Nov 19, 2018 · 067d2d3 · 067d2d3
2 parents f6f8ebd + 3351a70
commit 067d2d3
Show file tree

Hide file tree

Showing 18 changed files with 369 additions and 148 deletions.
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -1,6 +1,38 @@
 Changelog
 =========
 
+0.4.2
+-----
+**Date:** 2018-11-TBA
+
+CPU Plugin Performance
+
+This release fixes a performance regression for energy histograms and phase
+space plugins on CPU with our OpenMP backend on CPU. At least OpenMP 3.1 is
+needed to benefit from this. Additionally, several small documentation issues
+have been fixed and the energy histogram python tool forgot to return the first
+iteration.
+
+Thanks to Axel Huebl, René Widera, Sebastian Starke, and Marco Garten for
+contributions to this release!
+
+### Changes to "0.4.1"
+
+**Bug Fixes:**
+- Plugin performance regression:
+  - Speed of plugins `EnergyHistogram` and `PhaseSpace` on CPU (`omp2b`) #2802
+- Tools:
+  - Python `EnergyHistogramData`: skip of first iteration #2799
+
+**Misc:**
+- update Alpaka to 0.3.5 to fix #2802
+- Docs:
+  - CFL Static Assert: new grid.param #2804
+  - missing exponent in fieldIonization.rst #2790
+  - remove grep file redirect #2788
+  - Calorimeter Plugin: Document File Suffix #2800
+
+
 0.4.1
 -----
 **Date:** 2018-11-06
@@ -16,7 +48,7 @@ moving window simulations.
 Thanks to Axel Huebl, René Widera, Richard Pausch and Andrei Berceanu for
 contributions to this release!
 
-### Changes to "0.4.1"
+### Changes to "0.4.0"
 
 **Bug Fixes:**
 - PIConGPU:

diff --git a/INSTALL.rst b/INSTALL.rst
@@ -119,7 +119,7 @@ rsync
 - *Arch Linux:* ``sudo pacman --sync rsync``
 - *Spack:* ``spack install rsync``
 
-alpaka 0.3.4
+alpaka 0.3.5
 """"""""""""
 - `alpaka <https://github.com/ComputationalRadiationPhysics/alpaka>`_ is included in the PIConGPU source code
 

diff --git a/docs/source/conf.py b/docs/source/conf.py
@@ -105,9 +105,9 @@
 # built documents.
 #
 # The short X.Y version.
-version = u'0.4.1'
+version = u'0.4.2'
 # The full version, including alpha/beta/rc tags.
-release = u'0.4.1'
+release = u'0.4.2'
 
 # The language for content autogenerated by Sphinx. Refer to documentation
 # for a list of supported languages.

diff --git a/docs/source/install/instructions/spack.rst b/docs/source/install/instructions/spack.rst
@@ -31,7 +31,7 @@ First `install spack <http://spack.readthedocs.io/en/latest/getting_started.html
    source $HOME/src/spack/share/spack/setup-env.sh
 
    # install a supported compiler
-   spack compiler list | grep [email protected] 1>/dev/null && spack install [email protected] && spack load [email protected] && spack compiler add
+   spack compiler list | grep -q [email protected] && spack install [email protected] && spack load [email protected] && spack compiler add
 
    # add the PIConGPU repository
    git clone https://github.com/ComputationalRadiationPhysics/spack-repo.git $HOME/src/spack-repo

diff --git a/docs/source/models/field_ionization.rst b/docs/source/models/field_ionization.rst
@@ -98,7 +98,7 @@ Ammosov-Delone-Krainov (ADK)
    :nowrap:
 
     \begin{align}
-        \Gamma_\mathrm{ADK} &= \underbrace{\sqrt{\frac{3 n^{*3} F}{\pi Z^3}}}_\text{lin. pol.} \frac{F D^2}{8 \pi Z} \exp\left(-\frac{2Z}{3n^{*3}F}\right) \\
+        \Gamma_\mathrm{ADK} &= \underbrace{\sqrt{\frac{3 n^{*3} F}{\pi Z^3}}}_\text{lin. pol.} \frac{F D^2}{8 \pi Z} \exp\left(-\frac{2Z^3}{3n^{*3}F}\right) \\
         D &\equiv \left( \frac{4 \mathrm{e} Z^3}{F n^{*4}} \right)^{n^*} \hspace{2cm} n^* \equiv \frac{Z}{\sqrt{2 E_\mathrm{ip}}}
     \end{align}
 

diff --git a/docs/source/usage/plugins/particleCalorimeter.rst b/docs/source/usage/plugins/particleCalorimeter.rst
@@ -23,15 +23,15 @@ Therein, a species can be also be marked for detecting particles leaving the sim
 .cfg file
 ^^^^^^^^^
 
-All options are denoted for the photon (``ph``) particle species here.
+All options are denoted exemplarily for the photon (``ph``) particle species here.
 
 
 ================================== =========================================================================================
 PIConGPU command line option       Description
 ================================== =========================================================================================
 ``--ph_calorimeter.period``        The ouput periodicity of the plugin.
                                    A value of ``100`` would mean an output at simulation time step *0, 100, 200, ...*.
-``--ph_calorimeter.file``          Output file prefix. Files will be stored in the folder ``ph_calorimeter``
+``--ph_calorimeter.file``          Output file suffix. Put unique name if same ``species + filter`` is used multiple times.
 ``--ph_calorimeter.filter``        Use filtered particles. All available filters will be shown with ``picongpu --help``
 ``--ph_calorimeter.numBinsYaw``    Specifies the number of bins used for the yaw axis of the calorimeter.
                                    Defaults to ``64``.
@@ -93,6 +93,8 @@ Output
 ^^^^^^
 
 The calorimeters are stored in hdf5-files in the ``simOutput/<species>_calorimeter/<filter>/`` directory.
+The file names are ``<species>_calorimeter_<file>_<sfilter>_<timestep>_0_0_0.h5``.
+
 The dataset within the hdf5-file is located at ``/data/<timestep>/calorimeter``.
 Depending on whether energy binning is enabled the dataset is two or three dimensional.
 The dataset has the following attributes:
@@ -129,6 +131,13 @@ Attribute          Description
    #. calorimeter for species ph each 128th time step **with** logarithmic energy binning.
    #. calorimeter for species ph each 1000th time step **without** (this is the default) logarithmic energy binning.
 
+.. attention::
+
+   When using the plugin multiple times for the same combination of ``species`` and ``filter``, you *must* provide a unique ``file`` suffix.
+   Otherwise output files will overwrite each other, since only ``species``, ``filter`` and ``file`` suffix are encoded in it.
+
+   An example use case would be two (or more) calorimeters for the same species and filter but with differing position in space or different binning, range, linear and log scaling, etc.
+
 Analysis Tools
 ^^^^^^^^^^^^^^
 

diff --git a/include/picongpu/fields/MaxwellSolver/DirSplitting/DirSplitting.hpp b/include/picongpu/fields/MaxwellSolver/DirSplitting/DirSplitting.hpp
@@ -69,14 +69,14 @@ namespace dirSplitting
          * A workaround is to add a template dependency to the expression.
          * `sizeof(ANY_TYPE) != 0` is always true and defers the evaluation.
          */
-        PMACC_CASSERT_MSG(DirectionSplitting_Set_dX_equal_dt_times_c____check_your_gridConfig_param_file,
+        PMACC_CASSERT_MSG(DirectionSplitting_Set_dX_equal_dt_times_c____check_your_grid_param_file,
                           (SI::SPEED_OF_LIGHT_SI * SI::DELTA_T_SI) == SI::CELL_WIDTH_SI &&
                           (sizeof(T_Dummy) != 0));
-        PMACC_CASSERT_MSG(DirectionSplitting_use_cubic_cells____check_your_gridConfig_param_file,
+        PMACC_CASSERT_MSG(DirectionSplitting_use_cubic_cells____check_your_grid_param_file,
                           SI::CELL_HEIGHT_SI == SI::CELL_WIDTH_SI &&
                           (sizeof(T_Dummy) != 0));
 #if (SIMDIM == DIM3)
-        PMACC_CASSERT_MSG(DirectionSplitting_use_cubic_cells____check_your_gridConfig_param_file,
+        PMACC_CASSERT_MSG(DirectionSplitting_use_cubic_cells____check_your_grid_param_file,
                           SI::CELL_DEPTH_SI == SI::CELL_WIDTH_SI &&
                           (sizeof(T_Dummy) != 0));
 #endif

diff --git a/include/picongpu/fields/MaxwellSolver/None/None.hpp b/include/picongpu/fields/MaxwellSolver/None/None.hpp
@@ -51,7 +51,7 @@ namespace none
     >
     {
         /* Courant-Friedrichs-Levy-Condition for Yee Field Solver: */
-        PMACC_CASSERT_MSG(Courant_Friedrichs_Levy_condition_failure____check_your_gridConfig_param_file,
+        PMACC_CASSERT_MSG(Courant_Friedrichs_Levy_condition_failure____check_your_grid_param_file,
             (SPEED_OF_LIGHT*SPEED_OF_LIGHT*DELTA_T*DELTA_T*INV_CELL2_SUM)<=1.0);
     };
 } // namespace none

diff --git a/include/picongpu/fields/MaxwellSolver/Yee/Yee.hpp b/include/picongpu/fields/MaxwellSolver/Yee/Yee.hpp
@@ -62,7 +62,7 @@ namespace maxwellSolver
         void updateE()
         {
             /* Courant-Friedrichs-Levy-Condition for Yee Field Solver: */
-            PMACC_CASSERT_MSG(Courant_Friedrichs_Levy_condition_failure____check_your_gridConfig_param_file,
+            PMACC_CASSERT_MSG(Courant_Friedrichs_Levy_condition_failure____check_your_grid_param_file,
                 (SPEED_OF_LIGHT*SPEED_OF_LIGHT*DELTA_T*DELTA_T*INV_CELL2_SUM)<=1.0);
 
             typedef SuperCellDescription<

diff --git a/include/picongpu/version.hpp b/include/picongpu/version.hpp
@@ -21,5 +21,5 @@
 
 #define PICONGPU_VERSION_MAJOR 0
 #define PICONGPU_VERSION_MINOR 4
-#define PICONGPU_VERSION_PATCH 1
+#define PICONGPU_VERSION_PATCH 2
 #define PICONGPU_VERSION_LABEL ""
diff --git a/lib/python/picongpu/plugins/data/energy_histogram.py b/lib/python/picongpu/plugins/data/energy_histogram.py
@@ -94,7 +94,6 @@ def get_iterations(self, species, species_filter="all"):
 
         # the first column contains the iterations
         return pd.read_csv(data_file_path,
-                           skiprows=1,
                            usecols=(0,),
                            delimiter=" ",
                            dtype=np.uint64).values[:, 0]
@@ -137,7 +136,6 @@ def get(self, species, species_filter="all", iteration=None,
         # read whole file as pandas.DataFrame
         data = pd.read_csv(
             data_file_path,
-            skiprows=1,
             delimiter=" "
         )
         # upper range of each bin in keV
@@ -170,7 +168,7 @@ def get(self, species, species_filter="all", iteration=None,
         if not set(iteration).issubset(data.index.values):
             raise IndexError('Iteration {} is not available!\n'
                              'List of available iterations: \n'
-                             '{}'.format(iteration, data['iteration']))
+                             '{}'.format(iteration, data.index.values))
 
         # remove unused columns
         del data['sum']

diff --git a/thirdParty/alpaka/.travis.yml b/thirdParty/alpaka/.travis.yml
@@ -1,5 +1,5 @@
 #
-# Copyright 2015-2017 Benjamin Worpitz, Erik Zenker
+# Copyright 2015-2018 Benjamin Worpitz, Erik Zenker
 #
 # This file is part of alpaka.
 #
@@ -20,7 +20,6 @@
 
 language: generic
 os: linux
-sudo: required
 dist: trusty
 services:
   - docker
@@ -34,8 +33,8 @@ services:
 #   [clang++] ALPAKA_CI_CLANG_VER               : {3.5.2, 3.6.2, 3.7.1, 3.8.1, 3.9.1, 4.0.0, 5.0.0}
 # CMAKE_BUILD_TYPE                              : {Debug, Release}
 # ALPAKA_CI                                     : {TRAVIS}
-# ALPAKA_CI_BOOST_BRANCH                        : {boost-1.62.0, boost-1.63.0, boost-1.64.0, boost-1.65.1, boost-1.66.0, boost-1.67.0}
-# ALPAKA_CI_CMAKE_VER                           : {3.7.2, 3.8.2, 3.9.6, 3.10.3, 3.11.1}
+# ALPAKA_CI_BOOST_BRANCH                        : {boost-1.62.0, boost-1.63.0, boost-1.64.0, boost-1.65.1, boost-1.66.0, boost-1.67.0, boost-1.68.0}
+# ALPAKA_CI_CMAKE_VER                           : {3.7.2, 3.8.2, 3.9.6, 3.10.3, 3.11.4, 3.12.3}
 # ALPAKA_CI_SANITIZERS                          : {ASan, UBsan, TSan, ESan}
 #    TSan is not currently used because it produces many unexpected errors
 # ALPAKA_CI_ANALYSIS                            : {ON, OFF}

diff --git a/thirdParty/alpaka/include/alpaka/acc/AccCpuOmp2Blocks.hpp b/thirdParty/alpaka/include/alpaka/acc/AccCpuOmp2Blocks.hpp
@@ -33,7 +33,7 @@
 #include <alpaka/idx/bt/IdxBtZero.hpp>
 #include <alpaka/atomic/AtomicNoOp.hpp>
 #include <alpaka/atomic/AtomicStlLock.hpp>
-#include <alpaka/atomic/AtomicOmpCritSec.hpp>
+#include <alpaka/atomic/AtomicOmpBuiltIn.hpp>
 #include <alpaka/atomic/AtomicHierarchy.hpp>
 #include <alpaka/math/MathStl.hpp>
 #include <alpaka/block/shared/dyn/BlockSharedMemDynBoostAlignedAlloc.hpp>
@@ -85,7 +85,7 @@ namespace alpaka
             public idx::bt::IdxBtZero<TDim, TSize>,
             public atomic::AtomicHierarchy<
                 atomic::AtomicStlLock<16>,   // grid atomics
-                atomic::AtomicOmpCritSec,    // block atomics
+                atomic::AtomicOmpBuiltIn,    // block atomics
                 atomic::AtomicNoOp           // thread atomics
             >,
             public math::MathStl,
@@ -116,7 +116,7 @@ namespace alpaka
                     idx::bt::IdxBtZero<TDim, TSize>(),
                     atomic::AtomicHierarchy<
                         atomic::AtomicStlLock<16>,// atomics between grids
-                        atomic::AtomicOmpCritSec, // atomics between blocks
+                        atomic::AtomicOmpBuiltIn, // atomics between blocks
                         atomic::AtomicNoOp        // atomics between threads
                     >(),
                     math::MathStl(),

diff --git a/thirdParty/alpaka/include/alpaka/acc/AccCpuOmp2Threads.hpp b/thirdParty/alpaka/include/alpaka/acc/AccCpuOmp2Threads.hpp
@@ -32,7 +32,7 @@
 #include <alpaka/idx/gb/IdxGbRef.hpp>
 #include <alpaka/idx/bt/IdxBtOmp.hpp>
 #include <alpaka/atomic/AtomicStlLock.hpp>
-#include <alpaka/atomic/AtomicOmpCritSec.hpp>
+#include <alpaka/atomic/AtomicOmpBuiltIn.hpp>
 #include <alpaka/atomic/AtomicHierarchy.hpp>
 #include <alpaka/math/MathStl.hpp>
 #include <alpaka/block/shared/dyn/BlockSharedMemDynBoostAlignedAlloc.hpp>
@@ -87,8 +87,8 @@ namespace alpaka
             public idx::bt::IdxBtOmp<TDim, TSize>,
             public atomic::AtomicHierarchy<
                 atomic::AtomicStlLock<16>,   // grid atomics
-                atomic::AtomicOmpCritSec,    // block atomics
-                atomic::AtomicOmpCritSec     // thread atomics
+                atomic::AtomicOmpBuiltIn,    // block atomics
+                atomic::AtomicOmpBuiltIn     // thread atomics
             >,
             public math::MathStl,
             public block::shared::dyn::BlockSharedMemDynBoostAlignedAlloc,
@@ -118,8 +118,8 @@ namespace alpaka
                     idx::bt::IdxBtOmp<TDim, TSize>(),
                     atomic::AtomicHierarchy<
                         atomic::AtomicStlLock<16>,// atomics between grids
-                        atomic::AtomicOmpCritSec, // atomics between blocks
-                        atomic::AtomicOmpCritSec  // atomics between threads
+                        atomic::AtomicOmpBuiltIn, // atomics between blocks
+                        atomic::AtomicOmpBuiltIn  // atomics between threads
                     >(),
                     math::MathStl(),
                     block::shared::dyn::BlockSharedMemDynBoostAlignedAlloc(static_cast<std::size_t>(blockSharedMemDynSizeBytes)),

diff --git a/thirdParty/alpaka/include/alpaka/acc/AccCpuOmp4.hpp b/thirdParty/alpaka/include/alpaka/acc/AccCpuOmp4.hpp
@@ -32,7 +32,7 @@
 #include <alpaka/idx/gb/IdxGbRef.hpp>
 #include <alpaka/idx/bt/IdxBtOmp.hpp>
 #include <alpaka/atomic/AtomicStlLock.hpp>
-#include <alpaka/atomic/AtomicOmpCritSec.hpp>
+#include <alpaka/atomic/AtomicOmpBuiltIn.hpp>
 #include <alpaka/atomic/AtomicHierarchy.hpp>
 #include <alpaka/math/MathStl.hpp>
 #include <alpaka/block/shared/dyn/BlockSharedMemDynBoostAlignedAlloc.hpp>
@@ -87,8 +87,8 @@ namespace alpaka
             public idx::bt::IdxBtOmp<TDim, TSize>,
             public atomic::AtomicHierarchy<
                 atomic::AtomicStlLock<16>,   // grid atomics
-                atomic::AtomicOmpCritSec,    // block atomics
-                atomic::AtomicOmpCritSec     // thread atomics
+                atomic::AtomicOmpBuiltIn,    // block atomics
+                atomic::AtomicOmpBuiltIn     // thread atomics
             >,
             public math::MathStl,
             public block::shared::dyn::BlockSharedMemDynBoostAlignedAlloc,
@@ -118,8 +118,8 @@ namespace alpaka
                     idx::bt::IdxBtOmp<TDim, TSize>(),
                     atomic::AtomicHierarchy<
                         atomic::AtomicStlLock<16>,// atomics between grids
-                        atomic::AtomicOmpCritSec, // atomics between blocks
-                        atomic::AtomicOmpCritSec  // atomics between threads
+                        atomic::AtomicOmpBuiltIn, // atomics between blocks
+                        atomic::AtomicOmpBuiltIn  // atomics between threads
                     >(),
                     math::MathStl(),
                     block::shared::dyn::BlockSharedMemDynBoostAlignedAlloc(static_cast<std::size_t>(blockSharedMemDynSizeBytes)),

diff --git a/thirdParty/alpaka/include/alpaka/alpaka.hpp b/thirdParty/alpaka/include/alpaka/alpaka.hpp
@@ -44,7 +44,7 @@
 // atomic
 #include <alpaka/atomic/AtomicCudaBuiltIn.hpp>
 #include <alpaka/atomic/AtomicNoOp.hpp>
-#include <alpaka/atomic/AtomicOmpCritSec.hpp>
+#include <alpaka/atomic/AtomicOmpBuiltIn.hpp>
 #include <alpaka/atomic/AtomicStlLock.hpp>
 #include <alpaka/atomic/Op.hpp>
 #include <alpaka/atomic/Traits.hpp>