GitHub actions compiler

.github/workflows/unix_cpu_build.yml

@@ -9,9 +9,58 @@ on:
 name: ci
 
 jobs:
+    clang-format:
+        name: Clang Format Lint
+        runs-on: ubuntu-latest
+        steps:
+            - name: Checkout Respository
+              uses: actions/checkout@master
+
+            - name: Check Sources
+              uses: DoozyX/clang-format-lint-action@v0.14
+              with:
+                source: './src ./test ./examples'
+                extensions: 'h,cpp,hpp'
+                clangFormatVersion: 14
+
+    documentation:
+        name: Documentation
+        runs-on: ubuntu-18.04
+        env:
+          DOXYGEN_VER: 1.8.18
+        steps:
+            - name: Checkout Repository
+              uses: actions/checkout@master
+
+            - name: Install Doxygen
+              run: |
+                  wget --quiet https://sourceforge.net/projects/doxygen/files/rel-${DOXYGEN_VER}/doxygen-${DOXYGEN_VER}.linux.bin.tar.gz
+                  mkdir doxygen
+                  tar -xf doxygen-${DOXYGEN_VER}.linux.bin.tar.gz -C doxygen --strip 1
+
+            - name: Install Boost
+              run: |
+                  sudo add-apt-repository ppa:mhier/libboost-latest
+                  sudo apt-get -qq update
+                  sudo apt-get install -y libboost1.74-dev
+
+            - name: Configure
+              run: |
+                  mkdir build && cd build && unset VCPKG_ROOT
+                  cmake -DAF_BUILD_CPU:BOOL=OFF -DAF_BUILD_CUDA:BOOL=OFF \
+                        -DAF_BUILD_OPENCL:BOOL=OFF -DAF_BUILD_UNIFIED:BOOL=OFF \
+                        -DAF_BUILD_EXAMPLES:BOOL=OFF -DBUILD_TESTING:BOOL=OFF \
+                        -DDOXYGEN_EXECUTABLE:FILEPATH=${GITHUB_WORKSPACE}/doxygen/bin/doxygen ..
+
+            - name: Build
+              run: |
+                  cd ${GITHUB_WORKSPACE}/build
+                  cmake --build . --target docs
+
     build_cpu:
         name: CPU
         runs-on: ${{ matrix.os }}
+        needs: [clang-format, documentation]
         env:
           NINJA_VER: 1.10.2
           CMAKE_VER: 3.10.2
@@ -20,11 +69,16 @@ jobs:
             matrix:
                 blas_backend: [Atlas, MKL, OpenBLAS]
                 os: [ubuntu-18.04, ubuntu-20.04, macos-latest]
+                compiler: [gcc, clang, icx]
                 exclude:
                     - os: macos-latest
                       blas_backend: Atlas
                     - os: macos-latest
                       blas_backend: MKL
+                    - blas_backend: Atlas
+                      compiler: icx
+                    - blas_backend: OpenBLAS
+                      compiler: icx
         steps:
             - name: Checkout Repository
               uses: actions/checkout@master
@@ -43,6 +97,7 @@ jobs:
               if: matrix.os != 'macos-latest'
               env:
                   OS_NAME: ${{ matrix.os }}
+                  CC: ${{ matrix.compiler }}
               run: |
                   cmake_suffix=$(if [ $OS_NAME == 'macos-latest' ]; then echo "Darwin-x86_64"; else echo "Linux-x86_64"; fi)
                   cmake_url=$(echo "https://github.com/Kitware/CMake/releases/download/v${CMAKE_VER}/cmake-${CMAKE_VER}-${cmake_suffix}.tar.gz")
@@ -54,6 +109,17 @@ jobs:
                   cmake_osx_dir=$(echo "${cmake_install_dir}/CMake.app/Contents/bin")
                   cmake_dir=$(if [ $OS_NAME == 'macos-latest' ]; then echo "${cmake_osx_dir}"; else echo "${cmake_lnx_dir}"; fi)
                   echo "CMAKE_PROGRAM=$(pwd)/${cmake_dir}/cmake" >> $GITHUB_ENV
+                  case "$CC" in
+                    'gcc')
+                        echo "CXX=g++" >> $GITHUB_ENV
+                        ;;
+                    'clang')
+                        echo "CXX=clang++" >> $GITHUB_ENV
+                        ;;
+                    'icx')
+                        echo "CXX=icpx" >> $GITHUB_ENV
+                        ;;
+                  esac
 
             - name: Install Dependencies for Macos
               if: matrix.os == 'macos-latest'
@@ -62,7 +128,7 @@ jobs:
                   echo "CMAKE_PROGRAM=cmake" >> $GITHUB_ENV
 
             - name: Install Common Dependencies for Ubuntu
-              if: matrix.os == 'ubuntu-20.04' || matrix.os == 'ubuntu-18.04'
+              if: matrix.os == 'ubuntu-18.04' || matrix.os == 'ubuntu-20.04' || matrix.os == 'ubuntu-22.04'
               run: |
                   sudo add-apt-repository ppa:mhier/libboost-latest
                   sudo apt-get -qq update
@@ -78,12 +144,15 @@ jobs:
 
             - name: Install MKL for Ubuntu
               if: matrix.os != 'macos-latest' && matrix.blas_backend == 'MKL'
+              env:
+                  CC: ${{ matrix.compiler }}
               run: |
                   wget https://apt.repos.intel.com/intel-gpg-keys/GPG-PUB-KEY-INTEL-SW-PRODUCTS.PUB
                   sudo apt-key add GPG-PUB-KEY-INTEL-SW-PRODUCTS.PUB
                   sudo sh -c 'echo deb https://apt.repos.intel.com/oneapi all main > /etc/apt/sources.list.d/oneAPI.list'
                   sudo apt-get -qq update
                   sudo apt-get install -y intel-oneapi-mkl-devel
+                  if [ "$CC" == 'icx' ]; then sudo apt-get install -y intel-oneapi-compiler-dpcpp-cpp; fi
                   echo "MKLROOT=/opt/intel/oneapi/mkl/latest" >> ${GITHUB_ENV}
 
             - name: Install OpenBLAS for Ubuntu
@@ -94,6 +163,8 @@ jobs:
               env:
                   USE_MKL: ${{ matrix.blas_backend == 'MKL' }}
                   BLAS_BACKEND: ${{ matrix.blas_backend }}
+                  CC: ${{ matrix.compiler }}
+                  OS_NAME: ${{ matrix.os }}
               run: |
                   ref=$(echo ${GITHUB_REF} | awk '/refs\/pull\/[0-9]+\/merge/{print $0}')
                   prnum=$(echo $ref | awk '{split($0, a, "/"); print a[3]}')
@@ -103,6 +174,7 @@ jobs:
                   backend=$(if [ "$USE_MKL" == 1 ]; then echo "Intel-MKL"; else echo "FFTW/LAPACK/BLAS"; fi)
                   buildname="$buildname-cpu-$BLAS_BACKEND"
                   cmake_rpath=$(if [ $OS_NAME == 'macos-latest' ]; then echo "-DCMAKE_INSTALL_RPATH=/opt/arrayfire/lib"; fi)
+                  if [ "$CC" == 'icx' ]; then source /opt/intel/oneapi/setvars.sh intel64; fi
                   mkdir build && cd build && unset VCPKG_ROOT
                   ${CMAKE_PROGRAM} -G Ninja \
                       -DCMAKE_MAKE_PROGRAM:FILEPATH=${GITHUB_WORKSPACE}/ninja \
@@ -115,6 +187,9 @@ jobs:
                   echo "CTEST_DASHBOARD=${dashboard}" >> $GITHUB_ENV
 
             - name: Build and Test
+              env:
+                  CC: ${{ matrix.compiler }}
               run: |
                   cd ${GITHUB_WORKSPACE}/build
+                  if [ "$CC" == 'icx' ]; then source /opt/intel/oneapi/setvars.sh intel64; fi
                   ctest -D Experimental --track ${CTEST_DASHBOARD} -T Test -T Submit -R cpu -j2

CMakeLists.txt

@@ -92,6 +92,7 @@ option(AF_WITH_STATIC_MKL "Link against static Intel MKL libraries" OFF)
 option(AF_WITH_STATIC_CUDA_NUMERIC_LIBS "Link libafcuda with static numeric libraries(cublas, cufft, etc.)" OFF)
 option(AF_WITH_SPDLOG_HEADER_ONLY "Build ArrayFire with header only version of spdlog" OFF)
 option(AF_WITH_FMT_HEADER_ONLY "Build ArrayFire with header only version of fmt" OFF)
+option(AF_WITH_FAST_MATH "Use lower precision but high performance numeric optimizations" OFF)
 
 if(AF_WITH_STATIC_CUDA_NUMERIC_LIBS)
   option(AF_WITH_PRUNE_STATIC_CUDA_NUMERIC_LIBS "Prune CUDA static libraries to reduce binary size.(WARNING: May break some libs on older CUDA toolkits for some compute arch)" OFF)
@@ -317,14 +318,15 @@ if(CMAKE_CROSSCOMPILING)
                        "directory and build the bin2cpp target.")
   endif()
 else()
-  add_executable(bin2cpp ${ArrayFire_SOURCE_DIR}/CMakeModules/bin2cpp.cpp
-                         ${ArrayFire_SOURCE_DIR}/src/backend/common/util.cpp)
+  add_executable(bin2cpp CMakeModules/bin2cpp.cpp
+                         src/backend/common/deterministicHash.cpp
+                         src/backend/common/deterministicHash.hpp
+                         src/backend/common/Source.hpp)
   set_target_properties(bin2cpp
     PROPERTIES
       CXX_STANDARD 17)
+  target_link_libraries(bin2cpp PRIVATE nonstd::span-lite)
 
-  # NOSPDLOG is used to remove the spdlog dependency from bin2cpp
-  target_compile_definitions(bin2cpp PRIVATE NOSPDLOG)
   if(WIN32)
     target_compile_definitions(bin2cpp PRIVATE OS_WIN)
   elseif(APPLE)

CMakeModules/AFconfigure_forge_dep.cmake

@@ -75,7 +75,8 @@ else(AF_BUILD_FORGE)
 
     if(TARGET Forge::forge)
         get_target_property(fg_lib_type Forge::forge TYPE)
-        if(NOT ${fg_lib_type} STREQUAL "STATIC_LIBRARY")
+        if(NOT ${fg_lib_type} STREQUAL "STATIC_LIBRARY" AND
+           AF_INSTALL_STANDALONE)
             install(FILES
                     $<TARGET_FILE:Forge::forge>
                     $<$<PLATFORM_ID:Linux>:$<TARGET_SONAME_FILE:Forge::forge>>

CMakeModules/InternalUtils.cmake

@@ -25,6 +25,15 @@ if(WIN32)
   check_cxx_compiler_flag(/permissive- cxx_compliance)
 endif()
 
+check_cxx_compiler_flag(-ffast-math has_cxx_fast_math)
+check_cxx_compiler_flag("-fp-model fast" has_cxx_fp_model)
+check_cxx_compiler_flag(-fno-errno-math has_cxx_no_errno_math)
+check_cxx_compiler_flag(-fno-trapping-math  has_cxx_no_trapping_math)
+check_cxx_compiler_flag(-fno-signed-zeros  has_cxx_no_signed_zeros)
+check_cxx_compiler_flag(-mno-ieee-fp has_cxx_no_ieee_fp)
+check_cxx_compiler_flag(-Wno-unqualified-std-cast-call has_cxx_unqualified_std_cast_call)
+check_cxx_compiler_flag(-Werror=reorder-ctor has_cxx_error_reorder_ctor)
+
 function(arrayfire_set_default_cxx_flags target)
   target_compile_options(${target}
     PRIVATE
@@ -39,6 +48,7 @@ function(arrayfire_set_default_cxx_flags target)
                                           /wd4668
                                           /wd4710
                                           /wd4505
+                                          /we5038
                                           /bigobj
                                           /EHsc
                                           # MSVC incorrectly sets the cplusplus to 199711L even if the compiler supports
@@ -51,7 +61,21 @@ function(arrayfire_set_default_cxx_flags target)
               # ignored attribute warnings in the OpenCL
               # headers
               $<$<BOOL:${has_ignored_attributes_flag}>:-Wno-ignored-attributes>
-              $<$<BOOL:${has_all_warnings_flag}>:-Wall>>
+              $<$<BOOL:${has_all_warnings_flag}>:-Wall>
+              $<$<BOOL:${has_cxx_unqualified_std_cast_call}>:-Wno-unqualified-std-cast-call>
+              $<$<BOOL:${has_cxx_error_reorder_ctor}>:-Werror=reorder-ctor>
+
+              $<$<BOOL:${AF_WITH_FAST_MATH}>:
+                  $<$<BOOL:${has_cxx_fast_math}>:-ffast-math>
+                  $<$<BOOL:${has_cxx_no_errno_math}>:-fno-errno-math>
+                  $<$<BOOL:${has_cxx_no_trapping_math}>:-fno-trapping-math>
+                  $<$<BOOL:${has_cxx_no_signed_zeros}>:-fno-signed-zeros>
+                  $<$<BOOL:${has_cxx_no_ieee_fp}>:-mno-ieee-fp>
+                  >
+
+              $<$<NOT:$<BOOL:${AF_WITH_FAST_MATH}>>:
+                    $<$<BOOL:${has_cxx_fp_model}>:-fp-model precise>>
+                      >
     )
 
   target_compile_definitions(${target}
@@ -65,6 +89,7 @@ function(arrayfire_set_default_cxx_flags target)
 
       $<$<BOOL:${AF_WITH_LOGGING}>:           AF_WITH_LOGGING>
       $<$<BOOL:${AF_CACHE_KERNELS_TO_DISK}>:  AF_CACHE_KERNELS_TO_DISK>
+      $<$<BOOL:${AF_WITH_FAST_MATH}>:         AF_WITH_FAST_MATH>
   )
 endfunction()
 

CMakeModules/bin2cpp.cpp

@@ -28,7 +28,7 @@
 #include <utility>
 #include <vector>
 
-#include <common/util.hpp>
+#include <common/deterministicHash.hpp>
 
 using namespace std;
 using std::cout;
@@ -275,7 +275,7 @@ int main(int argc, const char *const *const argv) {
 
     cout << "#pragma once\n";
     cout << "#include <cstddef>\n";          // defines size_t
-    cout << "#include <common/util.hpp>\n";  // defines common::Source
+    cout << "#include <common/Source.hpp>\n";  // defines common::Source
 
     int ns_cnt = 0;
     int level  = 0;

docs/details/arith.dox

@@ -190,14 +190,19 @@ Bitwise xor operation of two inputs
 Minimum of two inputs.
 
 
-
 \defgroup arith_func_max max
 
 \ingroup numeric_mat
 
 Maximum of two inputs.
 
 
+\defgroup arith_func_clamp clamp
+
+\ingroup numeric_mat
+
+Limits the range of the in array to the values between lo and hi
+
 
 \defgroup arith_func_rem rem
 

examples/benchmarks/pi.cpp

@@ -35,8 +35,8 @@ static double pi_device() {
 static double pi_host() {
     int count = 0;
     for (int i = 0; i < samples; ++i) {
-        float x = float(rand()) / RAND_MAX;
-        float y = float(rand()) / RAND_MAX;
+        float x = float(rand()) / float(RAND_MAX);
+        float y = float(rand()) / float(RAND_MAX);
         if (sqrt(x * x + y * y) < 1) count++;
     }
     return 4.0 * count / samples;