Full AARCH64 support with NEON and new CUDA toolkit.

diff --git a/3rdparty/libpng/CMakeLists.txt b/3rdparty/libpng/CMakeLists.txt
index 1b2dddd..88bf04d 100644 (file)
--- a/3rdparty/libpng/CMakeLists.txt
+++ b/3rdparty/libpng/CMakeLists.txt
@@ -14,7 +14,7 @@ ocv_include_directories("${CMAKE_CURRENT_SOURCE_DIR}" ${ZLIB_INCLUDE_DIR})
 file(GLOB lib_srcs *.c)
 file(GLOB lib_hdrs *.h)
 
-if(NEON AND CMAKE_SIZEOF_VOID_P EQUAL 4)
+if(ARM AND NEON)
   list(APPEND lib_srcs arm/filter_neon.S)
   add_definitions(-DPNG_ARM_NEON)
 endif()

diff --git a/3rdparty/tbb/CMakeLists.txt b/3rdparty/tbb/CMakeLists.txt
index f5c67fb..1fd31fa 100644 (file)
--- a/3rdparty/tbb/CMakeLists.txt
+++ b/3rdparty/tbb/CMakeLists.txt
@@ -204,7 +204,7 @@ endif()
 
 set(TBB_SOURCE_FILES ${lib_srcs} ${lib_hdrs})
 
-if (ARM AND NOT WIN32)
+if ((ARM OR AARCH64) AND NOT WIN32)
   if (NOT ANDROID)
     set(TBB_SOURCE_FILES ${TBB_SOURCE_FILES} "${CMAKE_CURRENT_SOURCE_DIR}/arm_linux_stub.cpp")
   endif()

diff --git a/CMakeLists.txt b/CMakeLists.txt
index 268c3cc..b10f2cc 100644 (file)
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -226,8 +226,8 @@ OCV_OPTION(ENABLE_SSE41               "Enable SSE4.1 instructions"
 OCV_OPTION(ENABLE_SSE42               "Enable SSE4.2 instructions"                               OFF  IF (CMAKE_COMPILER_IS_GNUCXX AND (X86 OR X86_64)) )
 OCV_OPTION(ENABLE_AVX                 "Enable AVX instructions"                                  OFF  IF ((MSVC OR CMAKE_COMPILER_IS_GNUCXX) AND (X86 OR X86_64)) )
 OCV_OPTION(ENABLE_AVX2                "Enable AVX2 instructions"                                 OFF  IF ((MSVC OR CMAKE_COMPILER_IS_GNUCXX) AND (X86 OR X86_64)) )
-OCV_OPTION(ENABLE_NEON                "Enable NEON instructions"                                 OFF  IF CMAKE_COMPILER_IS_GNUCXX AND ARM )
-OCV_OPTION(ENABLE_VFPV3               "Enable VFPv3-D32 instructions"                            OFF  IF CMAKE_COMPILER_IS_GNUCXX AND ARM )
+OCV_OPTION(ENABLE_NEON                "Enable NEON instructions"                                 AARCH64 IF CMAKE_COMPILER_IS_GNUCXX AND (ARM OR AARCH64))
+OCV_OPTION(ENABLE_VFPV3               "Enable VFPv3-D32 instructions"                            AARCH64 IF CMAKE_COMPILER_IS_GNUCXX AND (ARM OR AARCH64))
 OCV_OPTION(ENABLE_NOISY_WARNINGS      "Show all warnings even if they are too noisy"             OFF )
 OCV_OPTION(OPENCV_WARNINGS_ARE_ERRORS "Treat warnings as errors"                                 OFF )
 OCV_OPTION(ENABLE_WINRT_MODE          "Build with Windows Runtime support"                       OFF  IF WIN32 )

diff --git a/cmake/OpenCVDetectCUDA.cmake b/cmake/OpenCVDetectCUDA.cmake
index 24fbb03..99e4349 100644 (file)
--- a/cmake/OpenCVDetectCUDA.cmake
+++ b/cmake/OpenCVDetectCUDA.cmake
@@ -91,8 +91,13 @@ if(CUDA_FOUND)
 
   if(NOT DEFINED __cuda_arch_bin)
     if(ANDROID)
-      set(__cuda_arch_bin "3.2")
-      set(__cuda_arch_ptx "")
+      if (ARM)
+        set(__cuda_arch_bin "3.2")
+        set(__cuda_arch_ptx "")
+      elseif(AARCH64)
+        set(__cuda_arch_bin "5.2")
+        set(__cuda_arch_ptx "")
+      endif()
     else()
       if(${CUDA_VERSION} VERSION_LESS "5.0")
         set(__cuda_arch_bin "1.1 1.2 1.3 2.0 2.1(2.0) 3.0")
@@ -154,7 +159,10 @@ if(CUDA_FOUND)
 
   if(ANDROID)
     set(CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS} "-Xptxas;-dlcm=ca")
-    set(CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS} "-target-os-variant=Android")
+    if(${CUDA_VERSION} VERSION_LESS "7.0")
+      # since CUDA 7.0 OS variant is depricated
+      set(CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS} "-target-os-variant=Android")
+    endif()
   endif()
 
   message(STATUS "CUDA NVCC target flags: ${CUDA_NVCC_FLAGS}")

diff --git a/cmake/OpenCVDetectCXXCompiler.cmake b/cmake/OpenCVDetectCXXCompiler.cmake
index 819c2e0..674a6e4 100644 (file)
--- a/cmake/OpenCVDetectCXXCompiler.cmake
+++ b/cmake/OpenCVDetectCXXCompiler.cmake
@@ -109,6 +109,8 @@ elseif(CMAKE_SYSTEM_PROCESSOR MATCHES "i686.*|i386.*|x86.*|amd64.*|AMD64.*")
   set(X86 1)
 elseif (CMAKE_SYSTEM_PROCESSOR MATCHES "arm.*|ARM.*")
   set(ARM 1)
+elseif (CMAKE_SYSTEM_PROCESSOR MATCHES "aarch64.*|AARCH64.*")
+  set(AARCH64 1)
 endif()