add the CI job to check bfloat16

Author: yhmtsai

.github/workflows/intel.yml

@@ -24,8 +24,9 @@ jobs:
       fail-fast: false
       matrix:
         config:
-          - {compiler: "dpcpp", build_type: "Release", name: "intel/dpcpp/release/shared", mixed: "ON"}
-          - {compiler: "icpx", build_type: "Release", name: "intel/icpx/release/shared", mixed: "OFF"}
+          - {compiler: "dpcpp", build_type: "Release", name: "intel/dpcpp/release/shared", mixed: "ON", half: "ON", bfloat16: "OFF"}
+          - {compiler: "icpx", build_type: "Release", name: "intel/icpx/release/shared", mixed: "OFF", half: "ON", bfloat16: "OFF"}
+          - {compiler: "icpx", build_type: "Release", name: "intel/icpx/release/shared-bfloat16", mixed: "ON", half: "OFF", bfloat16: "ON"}
     name: ${{ matrix.config.name }}
     runs-on: [gpu_intel]
 
@@ -39,7 +40,7 @@ jobs:
           module load intel-oneapi-compilers intel-oneapi-dpl intel-oneapi-mkl cmake
           mkdir build
           cd build
-          cmake .. -DCMAKE_INSTALL_PREFIX=install_ginkgo -DCMAKE_CXX_FLAGS="-Wpedantic -ffp-model=precise" -DCMAKE_CXX_COMPILER=${{ matrix.config.compiler }} -DCMAKE_BUILD_TYPE=${{ matrix.config.build_type }} -DGINKGO_MIXED_PRECISION=${{ matrix.config.mixed }} -DGINKGO_BUILD_CUDA=OFF -DGINKGO_BUILD_HIP=OFF -DGINKGO_BUILD_MPI=OFF -DGINKGO_DPCPP_SINGLE_MODE=ON
+          cmake .. -DCMAKE_INSTALL_PREFIX=install_ginkgo -DCMAKE_CXX_FLAGS="-Wpedantic -ffp-model=precise" -DCMAKE_CXX_COMPILER=${{ matrix.config.compiler }} -DCMAKE_BUILD_TYPE=${{ matrix.config.build_type }} -DGINKGO_MIXED_PRECISION=${{ matrix.config.mixed }} -DGINKGO_BUILD_CUDA=OFF -DGINKGO_BUILD_HIP=OFF -DGINKGO_BUILD_MPI=OFF -DGINKGO_DPCPP_SINGLE_MODE=ON -DGINKGO_ENABLE_HALF=${{ matrix.config.half }} -DGINKGO_ENABLE_BFLOAT16=${{ matrix.config.bfloat16 }}
           make -j8
           ONEAPI_DEVICE_SELECTOR=level_zero:gpu ctest -j10 --output-on-failure
   

.github/workflows/osx.yml

@@ -30,8 +30,9 @@ jobs:
       fail-fast: false
       matrix:
         config:
-          - {shared: "ON", build_type: "Debug", name: "omp/debug/shared", "mixed": "OFF"}
-          - {shared: "OFF", build_type: "Release", name: "omp/release/static", "mixed": "ON"}
+          - {shared: "ON", build_type: "Debug", name: "omp/debug/shared", mixed: "OFF", half: "ON", , bfloat16: "OFF"}
+          - {shared: "OFF", build_type: "Release", name: "omp/release/static", mixed: "ON", half: "ON", bfloat16: "OFF"}
+          - {shared: "ON", build_type: "Release", name: "omp/release/shared", mixed: "ON", half: "OFF", bfloat16: "ON"}
     name: ${{ matrix.config.name }}
     runs-on: [macos-latest]
 
@@ -58,7 +59,7 @@ jobs:
           mkdir install
           export INSTALL_PREFIX=`pwd`/install
           cd build
-          cmake .. -DCMAKE_PREFIX_PATH=/opt/homebrew/opt/libomp/ -DCMAKE_CXX_FLAGS=-Wpedantic -DBUILD_SHARED_LIBS=${{ matrix.config.shared }} -DCMAKE_INSTALL_PREFIX=${INSTALL_PREFIX} -DCMAKE_BUILD_TYPE=${{ matrix.config.build_type }} -DGINKGO_MIXED_PRECISION=${{ matrix.config.mixed }}
+          cmake .. -DCMAKE_PREFIX_PATH=/opt/homebrew/opt/libomp/ -DCMAKE_CXX_FLAGS=-Wpedantic -DBUILD_SHARED_LIBS=${{ matrix.config.shared }} -DCMAKE_INSTALL_PREFIX=${INSTALL_PREFIX} -DCMAKE_BUILD_TYPE=${{ matrix.config.build_type }} -DGINKGO_MIXED_PRECISION=${{ matrix.config.mixed }} -DGINKGO_ENABLE_HALF=${{ matrix.config.half }} -DGINKGO_ENABLE_BFLOAT16=${{ matrix.config.bfloat16 }}
           make -j8
           ctest -j10 --output-on-failure
 

.gitlab-ci.yml

@@ -175,6 +175,8 @@ build/amd/nompi/clang/rocm45/release/static:
     BUILD_HIP: "ON"
     BUILD_TYPE: "Release"
     BUILD_SHARED_LIBS: "OFF"
+    ENABLE_HALF: "OFF"
+    ENABLE_BFLOAT16: "ON"
 
 build/amd/nompi/clang/rocm45/debug/shared:
   extends:
@@ -212,6 +214,8 @@ build/amd/nompi/clang/rocm514/release/shared:
     BUILD_OMP: "ON"
     BUILD_HIP: "ON"
     BUILD_TYPE: "Release"
+    ENABLE_HALF: "OFF"
+    ENABLE_BFLOAT16: "ON"
 
 # without omp
 build/amd/nompi/gcc/rocm514_wo_omp/release/shared:
@@ -248,6 +252,8 @@ build/nocuda/nompi/clang/core/release/shared:
   variables:
     CXX_COMPILER: "clang++"
     BUILD_TYPE: "Release"
+    ENABLE_HALF: "OFF"
+    ENABLE_BFLOAT16: "ON"
 
 build/nocuda/nompi/gcc/omp/release/shared:
   extends:
@@ -258,6 +264,8 @@ build/nocuda/nompi/gcc/omp/release/shared:
   variables:
     BUILD_OMP: "ON"
     BUILD_TYPE: "Release"
+    ENABLE_HALF: "OFF"
+    ENABLE_BFLOAT16: "ON"
 
 build/nocuda/openmpi/clang/omp/debug/static:
   extends:
@@ -300,6 +308,8 @@ build/nocuda/nompi/gcc/omp/release/static:
     BUILD_OMP: "ON"
     BUILD_TYPE: "Release"
     BUILD_SHARED_LIBS: "OFF"
+    ENABLE_HALF: "OFF"
+    ENABLE_BFLOAT16: "ON"
 
 build/nocuda-nomixed/nompi/clang/omp/release/static:
   extends:

.gitlab/scripts.yml

@@ -49,6 +49,8 @@
         -DGINKGO_BUILD_TESTS=ON -DGINKGO_BUILD_EXAMPLES=ON
         -DGINKGO_FAST_TESTS=${FAST_TESTS}
         -DGINKGO_TEST_NONDEFAULT_STREAM=${NONDEFAULT_STREAM}
+        -DGINKGO_ENABLE_HALF=${ENABLE_HALF}
+        -DGINKGO_ENABLE_BFLOAT16=${ENABLE_BFLOAT16}
         -DGINKGO_MIXED_PRECISION=${MIXED_PRECISION}
         -DGINKGO_CONFIG_LOG_DETAILED=${CONFIG_LOG}
         -DGINKGO_DPCPP_SINGLE_MODE=${DPCPP_SINGLE_MODE}
@@ -91,6 +93,8 @@
         -DGINKGO_BUILD_PAPI_SDE=${BUILD_PAPI_SDE}
         -DGINKGO_BUILD_TESTS=ON -DGINKGO_BUILD_EXAMPLES=ON
         -DGINKGO_FAST_TESTS=${FAST_TESTS}
+        -DGINKGO_ENABLE_HALF=${ENABLE_HALF}
+        -DGINKGO_ENABLE_BFLOAT16=${ENABLE_BFLOAT16}
         -DGINKGO_MIXED_PRECISION=${MIXED_PRECISION}
         -DGINKGO_CONFIG_LOG_DETAILED=${CONFIG_LOG}
         -DGINKGO_DPCPP_SINGLE_MODE=${DPCPP_SINGLE_MODE}

.gitlab/variables.yml

@@ -19,6 +19,8 @@
     NONDEFAULT_STREAM: "OFF"
     DPCPP_SINGLE_MODE: "OFF"
     MIXED_PRECISION: "ON"
+    ENABLE_HALF: "ON"
+    ENABLE_BFLOAT16: "OFF"
     CONFIG_LOG: "ON"
     CXX_FLAGS: "-Wpedantic"
     EXTRA_CMAKE_FLAGS: ""