Convert run_test.yml to run_test.sh

script/compiler_base.yml

@@ -1,6 +1,3 @@
-include:
-  - local: '/script/run_test.yml'
-
 .base_gcc:
   image: registry.gitlab.com/hzdr/crp/alpaka-group-container/alpaka-ci:gcc
   variables:
@@ -9,7 +6,8 @@ include:
                   -DALPAKA_ACC_CPU_B_SEQ_T_OMP2_ENABLE=ON
                   -DALPAKA_ACC_CPU_B_OMP2_T_SEQ_ENABLE=ON"
                   # -DALPAKA_ACC_CPU_B_SEQ_T_THREADS_ENABLE=ON
-  extends: .test_job
+  script:
+    - script/run_test.sh
   # x86_64 tag is used to get a multi-core CPU for the tests
   tags:
     - x86_64
@@ -22,7 +20,8 @@ include:
                   -DALPAKA_ACC_CPU_B_OMP2_T_SEQ_ENABLE=ON"
                   # -DALPAKA_ACC_CPU_B_SEQ_T_OMP2_ENABLE=ON
                   # -DALPAKA_ACC_CPU_B_SEQ_T_THREADS_ENABLE=ON
-  extends: .test_job
+  script:
+      - script/run_test.sh
   # x86_64 tag is used to get a multi-core CPU for the tests
   tags:
     - x86_64
@@ -36,7 +35,8 @@ include:
   before_script:
     - nvidia-smi
     - nvcc --version
-  extends: .test_job
+  script:
+      - script/run_test.sh
   tags:
     - cuda
     - intel

script/run_test.sh

@@ -0,0 +1,76 @@
+#!/bin/bash
+
+# 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 CUPLA_BUILD_TYPE
+
+if [[ ! -v CUPLA_BUILD_TYPE ]] ; then
+    CUPLA_BUILD_TYPE=Release ;
+fi
+
+echo "number of processor threads $(nproc)"
+$CXX --version
+cmake --version
+echo "Boost version-> $BOOST_VERSION"
+export cupla_DIR=$CI_PROJECT_DIR
+
+# use one build directory for all build configurations
+mkdir build
+cd build
+echo "Build type-> $CUPLA_BUILD_TYPE"
+
+# ALPAKA_ACCS contains the backends, which are used for each build
+# the backends are set in the sepcialized base jobs .base_gcc,.base_clang and.base_cuda
+for CMAKE_FLAGS in $ALPAKA_ACCS ; do
+    echo "###################################################"
+    echo "# Example Matrix Multiplication (adapted original)"
+    echo "###################################################"
+    echo "can not run with CPU_B_SEQ_T_SEQ due to missing elements layer in original SDK example"
+    echo "CPU_B_SEQ_T_OMP2/THREADS too many threads necessary (256)"
+    if [[ $CMAKE_FLAGS =~ -*DALPAKA_ACC_GPU_CUDA_ENABLE=ON.* ]]; then
+        cmake $cupla_DIR/example/CUDASamples/matrixMul/ \
+	      -DBOOST_ROOT=/opt/boost/$BOOST_VERSION \
+	      $CMAKE_FLAGS \
+	      -DCMAKE_BUILD_TYPE=$CUPLA_BUILD_TYPE
+        make -j
+        time ./matrixMul -wA=64 -wB=64 -hA=64 -hB=64
+        rm -r * ;
+    fi
+
+    echo "###################################################"
+    echo "# Example Async API (adapted original)"
+    echo "###################################################"
+    echo "can not run with CPU_B_SEQ_T_SEQ due to missing elements layer in original SDK example"
+    echo "CPU_B_SEQ_T_OMP2/THREADS too many threads necessary (512)"
+    if [[ $CMAKE_FLAGS =~ -*DALPAKA_ACC_GPU_CUDA_ENABLE=ON.* ]]; then
+        cmake $cupla_DIR/example/CUDASamples/asyncAPI/ \
+	      -DBOOST_ROOT=/opt/boost/$BOOST_VERSION \
+	      $CMAKE_FLAGS \
+	      -DCMAKE_BUILD_TYPE=$CUPLA_BUILD_TYPE
+        make -j
+        time ./asyncAPI
+        rm -r * ;
+    fi
+
+    echo "###################################################"
+    echo "# Example Async API (added elements layer)"
+    echo "###################################################"
+    cmake $cupla_DIR/example/CUDASamples/asyncAPI_tuned/ \
+	  -DBOOST_ROOT=/opt/boost/$BOOST_VERSION \
+	  $CMAKE_FLAGS \
+	  -DCMAKE_BUILD_TYPE=$CUPLA_BUILD_TYPE
+    make -j
+    time ./asyncAPI_tuned
+    rm -r *
+
+    echo "###################################################"
+    echo "Example vectorAdd (added elements layer)"
+    echo "###################################################"
+    cmake $cupla_DIR/example/CUDASamples/vectorAdd/ \
+	  -DBOOST_ROOT=/opt/boost/$BOOST_VERSION \
+	  $CMAKE_FLAGS \
+	  -DCMAKE_BUILD_TYPE=$CUPLA_BUILD_TYPE
+    make -j
+    time ./vectorAdd 100000
+    rm -r * ;
+done