arrayfire · umar456 · Jan 21, 2023 · Nov 28, 2022 · Nov 29, 2022 · Nov 29, 2022
diff --git a/.github/workflows/win_cpu_build.yml b/.github/workflows/win_cpu_build.yml
@@ -13,7 +13,7 @@ jobs:
         name: CPU (fftw, OpenBLAS, windows-latest)
         runs-on: windows-latest
         env:
-          VCPKG_HASH: 6ca56aeb457f033d344a7106cb3f9f1abf8f4e98
+          VCPKG_HASH: f14984af3738e69f197bf0e647a8dca12de92996
           VCPKG_DEFAULT_TRIPLET: x64-windows
         steps:
             - name: Checkout Repository

diff --git a/CMakeModules/FindOpenCL.cmake b/CMakeModules/FindOpenCL.cmake
@@ -1,43 +1,51 @@
 # Distributed under the OSI-approved BSD 3-Clause License.  See accompanying
 # file Copyright.txt or https://cmake.org/licensing for details.
 
-#.rst:
-# FindOpenCL
-# ----------
-#
-# Try to find OpenCL
-#
-# IMPORTED Targets
-# ^^^^^^^^^^^^^^^^
-#
-# This module defines :prop_tgt:`IMPORTED` target ``OpenCL::OpenCL``, if
-# OpenCL has been found.
-#
-# Result Variables
-# ^^^^^^^^^^^^^^^^
-#
-# This module defines the following variables::
-#
-#   OpenCL_FOUND          - True if OpenCL was found
-#   OpenCL_INCLUDE_DIRS   - include directories for OpenCL
-#   OpenCL_LIBRARIES      - link against this library to use OpenCL
-#   OpenCL_VERSION_STRING - Highest supported OpenCL version (eg. 1.2)
-#   OpenCL_VERSION_MAJOR  - The major version of the OpenCL implementation
-#   OpenCL_VERSION_MINOR  - The minor version of the OpenCL implementation
-#
-# The module will also define two cache variables::
-#
-#   OpenCL_INCLUDE_DIR    - the OpenCL include directory
-#   OpenCL_LIBRARY        - the path to the OpenCL library
-#
+#[=======================================================================[.rst:
+FindOpenCL
+----------
+
+.. versionadded:: 3.1
+
+Finds Open Computing Language (OpenCL)
+
+.. versionadded:: 3.10
+  Detection of OpenCL 2.1 and 2.2.
+
+IMPORTED Targets
+^^^^^^^^^^^^^^^^
+
+.. versionadded:: 3.7
+
+This module defines :prop_tgt:`IMPORTED` target ``OpenCL::OpenCL``, if
+OpenCL has been found.
+
+Result Variables
+^^^^^^^^^^^^^^^^
+
+This module defines the following variables::
+
+  OpenCL_FOUND          - True if OpenCL was found
+  OpenCL_INCLUDE_DIRS   - include directories for OpenCL
+  OpenCL_LIBRARIES      - link against this library to use OpenCL
+  OpenCL_VERSION_STRING - Highest supported OpenCL version (eg. 1.2)
+  OpenCL_VERSION_MAJOR  - The major version of the OpenCL implementation
+  OpenCL_VERSION_MINOR  - The minor version of the OpenCL implementation
+
+The module will also define two cache variables::
+
+  OpenCL_INCLUDE_DIR    - the OpenCL include directory
+  OpenCL_LIBRARY        - the path to the OpenCL library
+
+#]=======================================================================]
 
 function(_FIND_OPENCL_VERSION)
   include(CheckSymbolExists)
   include(CMakePushCheckState)
   set(CMAKE_REQUIRED_QUIET ${OpenCL_FIND_QUIETLY})
 
   CMAKE_PUSH_CHECK_STATE()
-  foreach(VERSION "2_0" "1_2" "1_1" "1_0")
+  foreach(VERSION "3_0" "2_2" "2_1" "2_0" "1_2" "1_1" "1_0")
     set(CMAKE_REQUIRED_INCLUDES "${OpenCL_INCLUDE_DIR}")
 
     if(APPLE)
@@ -76,6 +84,9 @@ find_path(OpenCL_INCLUDE_DIR
     ENV NVSDKCOMPUTE_ROOT
     ENV CUDA_PATH
     ENV ATISTREAMSDKROOT
+    ENV OCL_ROOT
+    /usr/local/cuda
+    /opt/cuda
   PATH_SUFFIXES
     include
     OpenCL/common/inc
@@ -94,6 +105,7 @@ if(WIN32)
         ENV CUDA_PATH
         ENV NVSDKCOMPUTE_ROOT
         ENV ATISTREAMSDKROOT
+        ENV OCL_ROOT
       PATH_SUFFIXES
         "AMD APP/lib/x86"
         lib/x86
@@ -109,16 +121,39 @@ if(WIN32)
         ENV CUDA_PATH
         ENV NVSDKCOMPUTE_ROOT
         ENV ATISTREAMSDKROOT
+        ENV OCL_ROOT
       PATH_SUFFIXES
         "AMD APP/lib/x86_64"
         lib/x86_64
         lib/x64
         OpenCL/common/lib/x64)
   endif()
 else()
-  find_library(OpenCL_LIBRARY
-    NAMES OpenCL
-    PATH_SUFFIXES lib64/)
+  if(CMAKE_SIZEOF_VOID_P EQUAL 4)
+    find_library(OpenCL_LIBRARY
+      NAMES OpenCL
+      PATHS
+        ENV AMDAPPSDKROOT
+        ENV CUDA_PATH
+        /usr/local/cuda
+        /opt/cuda
+      PATH_SUFFIXES
+        lib/x86
+        lib)
+  elseif(CMAKE_SIZEOF_VOID_P EQUAL 8)
+    find_library(OpenCL_LIBRARY
+      NAMES OpenCL
+      PATHS
+        ENV AMDAPPSDKROOT
+        ENV CUDA_PATH
+        /usr/local/cuda
+        /opt/cuda
+      PATH_SUFFIXES
+        lib/x86_64
+        lib/x64
+        lib
+        lib64)
+  endif()
 endif()
 
 set(OpenCL_LIBRARIES ${OpenCL_LIBRARY})

diff --git a/CMakeModules/build_cl2hpp.cmake b/CMakeModules/build_cl2hpp.cmake
@@ -13,28 +13,30 @@
 
 find_package(OpenCL)
 
-find_path(cl2hpp_header_file_path
-  NAMES CL/cl2.hpp
-  PATHS ${OpenCL_INCLUDE_PATHS})
-
-if(cl2hpp_header_file_path)
-  add_library(cl2hpp IMPORTED INTERFACE GLOBAL)
-  add_library(OpenCL::cl2hpp IMPORTED INTERFACE GLOBAL)
-
-  set_target_properties(cl2hpp OpenCL::cl2hpp PROPERTIES
-    INTERFACE_INCLUDE_DIRECTORIES ${cl2hpp_header_file_path})
-elseif (NOT TARGET OpenCL::cl2hpp OR NOT TARGET cl2hpp)
-  af_dep_check_and_populate(${cl2hpp_prefix}
-    URI https://github.com/KhronosGroup/OpenCL-CLHPP.git
-    REF v2.0.12)
-
-  find_path(cl2hpp_var
+if(NOT TARGET OpenCL::cl2hpp)
+  find_path(cl2hpp_header_file_path
     NAMES CL/cl2.hpp
-    PATHS ${ArrayFire_BINARY_DIR}/extern/${cl2hpp_prefix}-src/include)
+    PATHS ${OpenCL_INCLUDE_PATHS})
 
-  add_library(cl2hpp IMPORTED INTERFACE GLOBAL)
-  add_library(OpenCL::cl2hpp IMPORTED INTERFACE GLOBAL)
+  if(cl2hpp_header_file_path)
+    add_library(cl2hpp IMPORTED INTERFACE GLOBAL)
+    add_library(OpenCL::cl2hpp IMPORTED INTERFACE GLOBAL)
 
-  set_target_properties(cl2hpp OpenCL::cl2hpp PROPERTIES
-    INTERFACE_INCLUDE_DIRECTORIES ${cl2hpp_var})
+    set_target_properties(cl2hpp OpenCL::cl2hpp PROPERTIES
+      INTERFACE_INCLUDE_DIRECTORIES ${cl2hpp_header_file_path})
+  elseif (NOT TARGET OpenCL::cl2hpp OR NOT TARGET cl2hpp)
+    af_dep_check_and_populate(${cl2hpp_prefix}
+      URI https://github.com/KhronosGroup/OpenCL-CLHPP.git
+      REF v2022.09.30)
+
+    find_path(cl2hpp_var
+      NAMES CL/cl2.hpp
+      PATHS ${ArrayFire_BINARY_DIR}/extern/${cl2hpp_prefix}-src/include)
+
+    add_library(cl2hpp IMPORTED INTERFACE GLOBAL)
+    add_library(OpenCL::cl2hpp IMPORTED INTERFACE GLOBAL)
+
+    set_target_properties(cl2hpp OpenCL::cl2hpp PROPERTIES
+      INTERFACE_INCLUDE_DIRECTORIES ${cl2hpp_var})
+  endif()
 endif()
diff --git a/CMakeModules/vcpkg/ports/lapack-reference/portfile.cmake b/CMakeModules/vcpkg/ports/lapack-reference/portfile.cmake
@@ -68,7 +68,7 @@ vcpkg_cmake_configure(
     OPTIONS
         "-DUSE_OPTIMIZED_BLAS=${USE_OPTIMIZED_BLAS}"
         "-DCBLAS=${CBLAS}"
-	"-DLAPACKE=ON"
+        "-DLAPACKE=ON"
         ${FORTRAN_CMAKE}
 )
 

diff --git a/include/af/opencl.h b/include/af/opencl.h
@@ -8,6 +8,9 @@
  ********************************************************/
 
 #pragma once
+#ifndef CL_TARGET_OPENCL_VERSION
+#define CL_TARGET_OPENCL_VERSION 120
+#endif
 #if defined(__APPLE__) || defined(__MACOSX)
 #include <OpenCL/cl.h>
 #else

diff --git a/src/api/c/fftconvolve.cpp b/src/api/c/fftconvolve.cpp
@@ -49,10 +49,11 @@ using std::vector;
 template<typename T>
 af_array fftconvolve_fallback(const af_array signal, const af_array filter,
                               const bool expand, const int baseDim) {
-    using convT =
-        typename conditional<is_integral<T>::value || is_same<T, float>::value,
-                             float, double>::type;
-    using cT = typename conditional<is_same<convT, float>::value, cfloat,
+    using convT = typename conditional<is_integral<T>::value ||
+                                           is_same<T, float>::value ||
+                                           is_same<T, cfloat>::value,
+                                       float, double>::type;
+    using cT    = typename conditional<is_same<convT, float>::value, cfloat,
                                     cdouble>::type;
 
     const Array<cT> S = castArray<cT>(signal);

diff --git a/src/api/c/imageio.cpp b/src/api/c/imageio.cpp
@@ -1091,5 +1091,4 @@ af_err af_delete_image_memory(void *ptr) {
     AF_RETURN_ERROR("ArrayFire compiled without Image IO (FreeImage) support",
                     AF_ERR_NOT_CONFIGURED);
 }
-}  // namespace arrayfire
 #endif  // WITH_FREEIMAGE
diff --git a/src/backend/common/ArrayFireTypesIO.hpp b/src/backend/common/ArrayFireTypesIO.hpp
@@ -14,13 +14,10 @@
 
 template<>
 struct fmt::formatter<af_seq> {
-    // Parses format specifications of the form ['f' | 'e'].
     constexpr auto parse(format_parse_context& ctx) -> decltype(ctx.begin()) {
         return ctx.begin();
     }
 
-    // Formats the point p using the parsed format specification (presentation)
-    // stored in this formatter.
     template<typename FormatContext>
     auto format(const af_seq& p, FormatContext& ctx) -> decltype(ctx.out()) {
         // ctx.out() is an output iterator to write to.
@@ -61,27 +58,24 @@ struct fmt::formatter<arrayfire::common::Version> {
             }
             ++it;
         } while (it != end && *it != '}');
-        return ctx.begin();
+        return it;
     }
 
-    // Formats the point p using the parsed format specification (presentation)
-    // stored in this formatter.
     template<typename FormatContext>
     auto format(const arrayfire::common::Version& ver, FormatContext& ctx)
         -> decltype(ctx.out()) {
-        // ctx.out() is an output iterator to write to.
-        // if (ver.major == -1) return format_to(ctx.out(), "N/A");
-        if (ver.minor == -1) show_minor = false;
-        if (ver.patch == -1) show_patch = false;
+        if (ver.major() == -1) return format_to(ctx.out(), "N/A");
+        if (ver.minor() == -1) show_minor = false;
+        if (ver.patch() == -1) show_patch = false;
         if (show_major && !show_minor && !show_patch) {
-            return format_to(ctx.out(), "{}", ver.major);
+            return format_to(ctx.out(), "{}", ver.major());
         }
         if (show_major && show_minor && !show_patch) {
-            return format_to(ctx.out(), "{}.{}", ver.major, ver.minor);
+            return format_to(ctx.out(), "{}.{}", ver.major(), ver.minor());
         }
         if (show_major && show_minor && show_patch) {
-            return format_to(ctx.out(), "{}.{}.{}", ver.major, ver.minor,
-                             ver.patch);
+            return format_to(ctx.out(), "{}.{}.{}", ver.major(), ver.minor(),
+                             ver.patch());
         }
         return ctx.out();
     }

diff --git a/src/backend/common/DependencyModule.cpp b/src/backend/common/DependencyModule.cpp
@@ -52,7 +52,7 @@ vector<string> libNames(const std::string& name, const string& suffix,
     UNUSED(suffix);
     const string noVerName = libraryPrefix + name + librarySuffix;
     if (ver != arrayfire::common::NullVersion) {
-        const string infix = "." + to_string(ver.major) + ".";
+        const string infix = "." + to_string(ver.major()) + ".";
         return {libraryPrefix + name + infix + librarySuffix, noVerName};
     } else {
         return {noVerName};
@@ -71,10 +71,11 @@ vector<string> libNames(const std::string& name, const string& suffix,
     UNUSED(suffix);
     const string noVerName = libraryPrefix + name + librarySuffix;
     if (ver != arrayfire::common::NullVersion) {
-        const string soname("." + to_string(ver.major));
+        const string soname("." + to_string(ver.major()));
 
-        const string vsfx = "." + to_string(ver.major) + "." +
-                            to_string(ver.minor) + "." + to_string(ver.patch);
+        const string vsfx = "." + to_string(ver.major()) + "." +
+                            to_string(ver.minor()) + "." +
+                            to_string(ver.patch());
         return {noVerName + vsfx, noVerName + soname, noVerName};
     } else {
         return {noVerName};

diff --git a/src/backend/common/Version.hpp b/src/backend/common/Version.hpp
@@ -21,11 +21,12 @@
 
 namespace arrayfire {
 namespace common {
-struct Version {
-    int major = -1;
-    int minor = -1;
-    int patch = -1;
+class Version {
+    int major_ = -1;
+    int minor_ = -1;
+    int patch_ = -1;
 
+   public:
     /// Checks if the major version is defined before minor and minor is defined
     /// before patch
     constexpr static bool validate(int major_, int minor_,
@@ -34,14 +35,18 @@ struct Version {
                !(minor_ < 0 && patch_ >= 0);
     }
 
+    constexpr int major() const { return major_; }
+    constexpr int minor() const { return minor_; }
+    constexpr int patch() const { return patch_; }
+
     constexpr Version(const int ver_major, const int ver_minor = -1,
                       const int ver_patch = -1) noexcept
-        : major(ver_major), minor(ver_minor), patch(ver_patch) {}
+        : major_(ver_major), minor_(ver_minor), patch_(ver_patch) {}
 };
 
 constexpr bool operator==(const Version& lhs, const Version& rhs) {
-    return lhs.major == rhs.major && lhs.minor == rhs.minor &&
-           lhs.patch == rhs.patch;
+    return lhs.major() == rhs.major() && lhs.minor() == rhs.minor() &&
+           lhs.patch() == rhs.patch();
 }
 
 constexpr bool operator!=(const Version& lhs, const Version& rhs) {
@@ -52,11 +57,11 @@ constexpr static Version NullVersion{-1, -1, -1};
 
 constexpr bool operator<(const Version& lhs, const Version& rhs) {
     if (lhs == NullVersion || rhs == NullVersion) return false;
-    if (lhs.major != -1 && rhs.major != -1 && lhs.major < rhs.major)
+    if (lhs.major() != -1 && rhs.major() != -1 && lhs.major() < rhs.major())
         return true;
-    if (lhs.minor != -1 && rhs.minor != -1 && lhs.minor < rhs.minor)
+    if (lhs.minor() != -1 && rhs.minor() != -1 && lhs.minor() < rhs.minor())
         return true;
-    if (lhs.patch != -1 && rhs.patch != -1 && lhs.patch < rhs.patch)
+    if (lhs.patch() != -1 && rhs.patch() != -1 && lhs.patch() < rhs.patch())
         return true;
     return false;
 }

diff --git a/src/backend/cuda/convolveNN.cpp b/src/backend/cuda/convolveNN.cpp
@@ -70,7 +70,7 @@ pair<cudnnConvolutionFwdAlgo_t, size_t> getForwardAlgorithm(
     size_t workspace_bytes = 0;
 
     auto version = getCudnnPlugin().getVersion();
-    if (version.major >= 8) {
+    if (version.major() >= 8) {
         int maxAlgoCount = 0;
         CUDNN_CHECK(cuda::cudnnGetConvolutionForwardAlgorithmMaxCount(
             cudnn, &maxAlgoCount));
@@ -419,7 +419,7 @@ pair<cudnnConvolutionBwdFilterAlgo_t, size_t> getBackwardFilterAlgorithm(
     size_t workspace_bytes = 0;
 
     auto version = getCudnnPlugin().getVersion();
-    if (version.major >= 8) {
+    if (version.major() >= 8) {
         int maxAlgoCount = 0;
         CUDNN_CHECK(cuda::cudnnGetConvolutionBackwardFilterAlgorithmMaxCount(
             cudnn, &maxAlgoCount));