PIConGPU unit tests

.gitlab-ci.yml

@@ -126,6 +126,20 @@ pmacc-compile-reduced-matrix:
         job: pmacc-generate-reduced-matrix
     strategy: depend
 
+picongpu-unittest-generate-reduced-matrix:
+  variables:
+    PIC_INPUTS: "unit"
+    TEST_TUPLE_NUM_ELEM: 1
+  extends: ".base_generate-reduced-matrix"
+
+picongpu-unittest-compile-reduced-matrix:
+  stage: test
+  trigger:
+    include:
+      - artifact: compile.yml
+        job: picongpu-unittest-generate-reduced-matrix
+    strategy: depend
+
 pypicongpu-generate-full-matrix:
   stage: generate
   image: ubuntu:22.04

include/picongpu/param/precision.param

@@ -30,6 +30,7 @@
 
 #pragma once
 
+#include <picongpu/simulation_types.hpp>
 
 namespace picongpu
 {

include/picongpu/particles/Particles.hpp

@@ -249,14 +249,6 @@ namespace picongpu
         }
     };
 
-    namespace traits
-    {
-        template<typename T_Name, typename T_Attributes, typename T_Flags>
-        struct GetDataBoxType<picongpu::Particles<T_Name, T_Attributes, T_Flags>>
-        {
-            using type = typename picongpu::Particles<T_Name, T_Attributes, T_Flags>::ParticlesBoxType;
-        };
-    } // namespace traits
 } // namespace picongpu
 
 namespace pmacc

include/picongpu/particles/shapes/CIC.hpp

@@ -19,7 +19,8 @@
 
 #pragma once
 
-#include "picongpu/simulation_defines.hpp"
+#include <pmacc/attribute/FunctionSpecifier.hpp>
+#include <pmacc/memory/Array.hpp>
 
 #include <cstdint>
 

include/picongpu/particles/shapes/Counter.hpp

@@ -19,7 +19,8 @@
 
 #pragma once
 
-#include "picongpu/simulation_defines.hpp"
+#include <pmacc/attribute/FunctionSpecifier.hpp>
+#include <pmacc/memory/Array.hpp>
 
 #include <cstdint>
 

include/picongpu/particles/shapes/NGP.hpp

@@ -19,7 +19,8 @@
 
 #pragma once
 
-#include "picongpu/simulation_defines.hpp"
+#include <pmacc/attribute/FunctionSpecifier.hpp>
+#include <pmacc/memory/Array.hpp>
 
 #include <cstdint>
 

include/picongpu/particles/shapes/PCS.hpp

@@ -19,7 +19,8 @@
 
 #pragma once
 
-#include "picongpu/simulation_defines.hpp"
+#include <pmacc/attribute/FunctionSpecifier.hpp>
+#include <pmacc/memory/Array.hpp>
 
 #include <cstdint>
 

include/picongpu/particles/shapes/PQS.hpp

@@ -19,7 +19,8 @@
 
 #pragma once
 
-#include "picongpu/simulation_defines.hpp"
+#include <pmacc/attribute/FunctionSpecifier.hpp>
+#include <pmacc/memory/Array.hpp>
 
 #include <cstdint>
 

include/picongpu/particles/shapes/TSC.hpp

@@ -20,7 +20,8 @@
 
 #pragma once
 
-#include "picongpu/simulation_defines.hpp"
+#include <pmacc/attribute/FunctionSpecifier.hpp>
+#include <pmacc/memory/Array.hpp>
 
 #include <cstdint>
 

include/picongpu/simulation_defines.hpp

@@ -26,11 +26,13 @@
 #include <pmacc/types.hpp>
 #include <picongpu/simulation_types.hpp>
 #include "pmacc_renamings.hpp"
+#include "picongpu/traits/GetMargin.hpp"
 
 
 namespace picongpu
 {
     using namespace pmacc;
+    using namespace picongpu::traits;
 }
 
 /* IMPORTANT we need to use #include <...> for local files

include/picongpu/simulation_types.hpp

@@ -20,8 +20,6 @@
 
 #pragma once
 
-#include "picongpu/traits/GetDataBoxType.hpp"
-#include "picongpu/traits/GetMargin.hpp"
 #include "version.hpp"
 
 #include <pmacc/algorithms/PromoteType.hpp>
@@ -56,6 +54,5 @@ namespace picongpu
 
     using namespace pmacc::algorithms::promoteType;
     using namespace pmacc::traits;
-    using namespace picongpu::traits;
 
 } // namespace picongpu

include/picongpu/unitless/precision.unitless

@@ -19,8 +19,9 @@
 
 #pragma once
 
-#include <pmacc/math/Vector.hpp>
+#include "picongpu/param/dimension.param"
 
+#include <pmacc/math/Vector.hpp>
 
 namespace picongpu
 {

share/ci/backendFlags.sh

@@ -0,0 +1,57 @@
+#!/bin/bash
+
+###################################################
+# translate PIConGPU backend names into CMake Flags
+###################################################
+
+get_backend_flags()
+{
+    backend_cfg=(${1//:/ })
+    num_options="${#backend_cfg[@]}"
+    if [ $num_options -gt 2 ] ; then
+        echo "-b|--backend must be contain 'backend:arch' or 'backend'" >&2
+        exit 1
+    fi
+    if [ "${backend_cfg[0]}" == "cuda" ] ; then
+        result+=" -Dalpaka_ACC_GPU_CUDA_ENABLE=ON -Dalpaka_ACC_GPU_CUDA_ONLY_MODE=ON"
+        if [ $num_options -eq 2 ] ; then
+            result+=" -DCMAKE_CUDA_ARCHITECTURES=\"${backend_cfg[1]}\""
+        else
+            result+=" -DCMAKE_CUDA_ARCHITECTURES=52"
+        fi
+    elif [ "${backend_cfg[0]}" == "omp2b" ] ; then
+        result+=" -Dalpaka_ACC_CPU_B_OMP2_T_SEQ_ENABLE=ON"
+        if [ $num_options -eq 2 ] ; then
+            result+=" -DPMACC_CPU_ARCH=\"${backend_cfg[1]}\""
+        fi
+    elif [ "${backend_cfg[0]}" == "serial" ] ; then
+        result+=" -Dalpaka_ACC_CPU_B_SEQ_T_SEQ_ENABLE=ON"
+        if [ $num_options -eq 2 ] ; then
+            result+=" -DPMACC_CPU_ARCH=\"${backend_cfg[1]}\""
+        fi
+    elif [ "${backend_cfg[0]}" == "tbb" ] ; then
+        result+=" -Dalpaka_ACC_CPU_B_TBB_T_SEQ_ENABLE=ON"
+        if [ $num_options -eq 2 ] ; then
+            result+=" -DPMACC_CPU_ARCH=\"${backend_cfg[1]}\""
+        fi
+    elif [ "${backend_cfg[0]}" == "threads" ] ; then
+        result+=" -Dalpaka_ACC_CPU_B_SEQ_T_THREADS_ENABLE=ON"
+        if [ $num_options -eq 2 ] ; then
+            result+=" -DPMACC_CPU_ARCH=\"${backend_cfg[1]}\""
+        fi
+    elif [ "${backend_cfg[0]}" == "hip" ] ; then
+        result+=" -Dalpaka_ACC_GPU_HIP_ENABLE=ON -Dalpaka_ACC_GPU_HIP_ONLY_MODE=ON"
+        if [ $num_options -eq 2 ] ; then
+            result+=" -DGPU_TARGETS=\"${backend_cfg[1]}\""
+        else
+            # If no architecture is given build for Radeon VII or MI50/60.
+            result+=" -DGPU_TARGETS=gfx906"
+        fi
+    else
+        echo "unsupported backend given '$1'" >&2
+        exit 1
+    fi
+
+    echo "$result"
+    exit 0
+}

share/ci/generate_reduced_matrix.sh

@@ -33,6 +33,9 @@ folders=()
 if [ "$PIC_INPUTS" == "pmacc" ] ; then
   # create test cases for PMacc
   echo "pmacc" | tr " " "\n" | n_wise_generator.py $@ --limit_boost_version
+elif [ "$PIC_INPUTS" == "unit" ] ; then
+   # create test cases for PMacc
+   echo "unit" | tr " " "\n" | n_wise_generator.py $@ --limit_boost_version
 else
   # create test cases for PIConGPU
   for CASE in ${PIC_INPUTS}; do

share/ci/n_wise_generator.py

@@ -296,7 +296,8 @@ def is_valid_combination(row):
             v_cuda_hip_str = "" if v_cuda_hip == 0 else str(v_cuda_hip)
             os_name = pairs[2][0]
             os_version = get_version(pairs[2])
-            image_prefix = "_run" if folder == "pmacc" else "_compile"
+            image_prefix = "_run" if folder == "pmacc" or folder == "unit"\
+                else "_compile"
             job_name = compiler + "_" + backend + v_cuda_hip_str + \
                 "_boost" + boost_version + "_" + folder.replace("/", ".")
             print(job_name + ":")

share/ci/run_picongpu_tests.sh

@@ -4,7 +4,6 @@ set -e
 set -o pipefail
 
 # the default build type is Release
-# if neccesary, you can rerun the pipeline with another build type-> https://docs.gitlab.com/ee/ci/pipelines.html#manually-executing-pipelines
 # to change the build type, you must set the environment variable PIC_BUILD_TYPE
 if [[ ! -v PIC_BUILD_TYPE ]] ; then
     PIC_BUILD_TYPE=Release ;

share/ci/run_picongpu_unit_tests.sh

@@ -0,0 +1,98 @@
+#!/bin/bash
+# Execute PIConGPU's unit tests
+
+set -e
+set -o pipefail
+
+export code_DIR=$CI_PROJECT_DIR
+source $code_DIR/share/ci/backendFlags.sh
+
+# the default build type is Release
+# to change the build type, you must set the environment variable PIC_BUILD_TYPE
+if [[ ! -v PIC_BUILD_TYPE ]] ; then
+    PIC_BUILD_TYPE=Release ;
+fi
+
+if [[ "$CI_RUNNER_TAGS" =~ .*cpuonly.* ]] ; then
+    # In cases where the compile-only job is executed on a GPU runner but with different kinds of accelerators
+    # we need to reset the variables to avoid compiling for the wrong architecture and accelerator.
+    unset CI_GPUS
+    unset CI_GPU_ARCH
+fi
+
+if [ -n "$CI_GPUS" ] ; then
+    # select randomly a device if multiple exists
+    # CI_GPUS is provided by the gitlab CI runner
+    SELECTED_DEVICE_ID=$((RANDOM%CI_GPUS))
+    export HIP_VISIBLE_DEVICES=$SELECTED_DEVICE_ID
+    export CUDA_VISIBLE_DEVICES=$SELECTED_DEVICE_ID
+    echo "selected device '$SELECTED_DEVICE_ID' of '$CI_GPUS'"
+else
+    echo "No GPU device selected because environment variable CI_GPUS is not set."
+fi
+
+if [[ "$PIC_BACKEND" =~ hip.* ]] || [[ "$PIC_BACKEND" =~ cuda.* ]] ; then
+    if [ -n "$CI_GPU_ARCH" ] ; then
+        export PIC_BACKEND="${PIC_BACKEND}:${CI_GPU_ARCH}"
+    fi
+fi
+
+###################################################
+# cmake config builder
+###################################################
+
+PIC_CONST_ARGS=""
+PIC_CONST_ARGS="${PIC_CONST_ARGS} -DCMAKE_BUILD_TYPE=${PIC_BUILD_TYPE}"
+CMAKE_ARGS="${PIC_CONST_ARGS} ${PIC_CMAKE_ARGS} -DCMAKE_CXX_COMPILER=${CXX_VERSION} -DBOOST_ROOT=/opt/boost/${BOOST_VERSION}"
+CMAKE_ARGS="$CMAKE_ARGS -DUSE_MPI_AS_ROOT_USER=ON"
+
+# check and activate if clang should be used as CUDA device compiler
+if [ -n "$CI_CLANG_AS_CUDA_COMPILER" ] ; then
+  export PATH="$(agc-manager -b cuda)/bin:$PATH"
+  CMAKE_ARGS="$CMAKE_ARGS -DCMAKE_CUDA_COMPILER=${CXX_VERSION}"
+fi
+
+alpaka_backend=$(get_backend_flags ${PIC_BACKEND})
+CMAKE_ARGS="$CMAKE_ARGS $alpaka_backend"
+
+###################################################
+# build and run unit tests
+###################################################
+
+# adjust number of parallel builds to avoid out of memory errors
+# PIC_BUILD_REQUIRED_MEM_BYTES is a configured variable in the CI web interface
+PIC_PARALLEL_BUILDS=$(($CI_RAM_BYTES_TOTAL/$PIC_BUILD_REQUIRED_MEM_BYTES))
+
+# limit to number of available cores
+if [ $PIC_PARALLEL_BUILDS -gt $CI_CPUS ] ; then
+    PIC_PARALLEL_BUILDS=$CI_CPUS
+fi
+
+# CI_MAX_PARALLELISM is a configured variable in the CI web interface
+if [ $PIC_PARALLEL_BUILDS -gt $CI_MAX_PARALLELISM ] ; then
+    PIC_PARALLEL_BUILDS=$CI_MAX_PARALLELISM
+fi
+
+## run unit tests
+export unitTest_folder=$HOME/buildPICUnitTest
+mkdir -p $unitTest_folder
+cd $unitTest_folder
+
+echo -e "\033[0;32m///////////////////////////////////////////////////"
+echo "PIC_BUILD_REQUIRED_MEM_BYTES-> ${PIC_BUILD_REQUIRED_MEM_BYTES}"
+echo "CI_RAM_BYTES_TOTAL          -> ${CI_RAM_BYTES_TOTAL}"
+echo "CI_CPUS                     -> ${CI_CPUS}"
+echo "CI_MAX_PARALLELISM          -> ${CI_MAX_PARALLELISM}"
+echo "number of processor threads -> $(nproc)"
+echo "number of parallel builds   -> $PIC_PARALLEL_BUILDS"
+echo "cmake version               -> $(cmake --version | head -n 1)"
+echo "build directory             -> $(pwd)"
+echo "CMAKE_ARGS                  -> ${CMAKE_ARGS}"
+echo "accelerator                 -> ${PIC_BACKEND}"
+echo "input set                   -> ${PIC_TEST_CASE_FOLDER}"
+echo -e "/////////////////////////////////////////////////// \033[0m \n\n"
+
+cmake $CMAKE_ARGS $code_DIR/share/picongpu/unit
+make -j $PIC_PARALLEL_BUILDS
+# execute on one device
+ctest -V