Merge pull request opencv#9763 from seiko2plus:addVsxCore

vpisarev · vpisarev · commit 7d55c09a9fff · 2017-10-10T10:00:31.000Z
diff --git a/CMakeLists.txt b/CMakeLists.txt
@@ -298,6 +298,7 @@ OCV_OPTION(ENABLE_PROFILING           "Enable profiling in the GCC compiler (Add
 OCV_OPTION(ENABLE_COVERAGE            "Enable coverage collection with  GCov"                    OFF  IF CMAKE_COMPILER_IS_GNUCXX )
 OCV_OPTION(ENABLE_OMIT_FRAME_POINTER  "Enable -fomit-frame-pointer for GCC"                      ON   IF CMAKE_COMPILER_IS_GNUCXX AND NOT (APPLE AND CMAKE_COMPILER_IS_CLANGCXX) )
 OCV_OPTION(ENABLE_POWERPC             "Enable PowerPC for GCC"                                   ON   IF (CMAKE_COMPILER_IS_GNUCXX AND CMAKE_SYSTEM_PROCESSOR MATCHES powerpc.*) )
+OCV_OPTION(ENABLE_VSX                 "Enable POWER8 and above VSX (64-bit little-endian)"       ON   IF (CMAKE_COMPILER_IS_GNUCXX AND PPC64LE) )
 OCV_OPTION(ENABLE_FAST_MATH           "Enable -ffast-math (not recommended for GCC 4.6.x)"       OFF  IF (CMAKE_COMPILER_IS_GNUCXX AND (X86 OR X86_64)) )
 OCV_OPTION(ENABLE_NEON                "Enable NEON instructions"                                 (NEON OR ANDROID_ARM_NEON OR AARCH64) IF CMAKE_COMPILER_IS_GNUCXX AND (ARM OR AARCH64 OR IOS) )
 OCV_OPTION(ENABLE_VFPV3               "Enable VFPv3-D32 instructions"                            OFF  IF CMAKE_COMPILER_IS_GNUCXX AND (ARM OR AARCH64 OR IOS) )
diff --git a/cmake/OpenCVCompilerOptimizations.cmake b/cmake/OpenCVCompilerOptimizations.cmake
@@ -28,6 +28,7 @@
 
 set(CPU_ALL_OPTIMIZATIONS "SSE;SSE2;SSE3;SSSE3;SSE4_1;SSE4_2;POPCNT;AVX;FP16;AVX2;FMA3") # without AVX512
 list(APPEND CPU_ALL_OPTIMIZATIONS NEON VFPV3 FP16)
+list(APPEND CPU_ALL_OPTIMIZATIONS VSX)
 list(REMOVE_DUPLICATES CPU_ALL_OPTIMIZATIONS)
 
 ocv_update(CPU_VFPV3_FEATURE_ALIAS "")
@@ -79,6 +80,7 @@ ocv_optimization_process_obsolete_option(ENABLE_FMA3 FMA3 ON)
 ocv_optimization_process_obsolete_option(ENABLE_VFPV3 VFPV3 OFF)
 ocv_optimization_process_obsolete_option(ENABLE_NEON NEON OFF)
 
+ocv_optimization_process_obsolete_option(ENABLE_VSX VSX OFF)
 
 macro(ocv_is_optimization_in_list resultvar check_opt)
   set(__checked "")
@@ -266,6 +268,15 @@ elseif(ARM OR AARCH64)
     ocv_update(CPU_FP16_IMPLIES "NEON")
     set(CPU_BASELINE "NEON;FP16" CACHE STRING "${HELP_CPU_BASELINE}")
   endif()
+elseif(PPC64LE)
+  ocv_update(CPU_KNOWN_OPTIMIZATIONS "VSX")
+  ocv_update(CPU_VSX_TEST_FILE "${OpenCV_SOURCE_DIR}/cmake/checks/cpu_vsx.cpp")
+
+  if(CMAKE_COMPILER_IS_CLANGCXX AND (NOT ${CMAKE_CXX_COMPILER} MATCHES "xlc"))
+    ocv_update(CPU_VSX_FLAGS_ON "-mvsx -maltivec")
+  else()
+    ocv_update(CPU_VSX_FLAGS_ON "-mcpu=power8")
+  endif()
 endif()
 
 # Helper values for cmake-gui
diff --git a/cmake/OpenCVDetectCXXCompiler.cmake b/cmake/OpenCVDetectCXXCompiler.cmake
@@ -72,6 +72,8 @@ elseif(CMAKE_SYSTEM_PROCESSOR MATCHES "^(arm.*|ARM.*)")
   set(ARM 1)
 elseif(CMAKE_SYSTEM_PROCESSOR MATCHES "^(aarch64.*|AARCH64.*)")
   set(AARCH64 1)
+elseif(CMAKE_SYSTEM_PROCESSOR MATCHES "^ppc64le.*|PPC64LE.*")
+  set(PPC64LE 1)
 endif()
 
 # Workaround for 32-bit operating systems on 64-bit x86_64 processor
diff --git a/cmake/OpenCVPackaging.cmake b/cmake/OpenCVPackaging.cmake
@@ -31,6 +31,9 @@ elseif(ARM)
 elseif(AARCH64)
   set(CPACK_DEBIAN_PACKAGE_ARCHITECTURE "arm64")
   set(CPACK_RPM_PACKAGE_ARCHITECTURE "aarch64")
+elseif(PPC64LE)
+  set(CPACK_DEBIAN_PACKAGE_ARCHITECTURE "ppc64el")
+  set(CPACK_RPM_PACKAGE_ARCHITECTURE "ppc64le")
 else()
   set(CPACK_DEBIAN_PACKAGE_ARCHITECTURE ${CMAKE_SYSTEM_PROCESSOR})
   set(CPACK_RPM_PACKAGE_ARCHITECTURE ${CMAKE_SYSTEM_PROCESSOR})
@@ -164,4 +167,4 @@ endif(NOT OPENCV_CUSTOM_PACKAGE_INFO)
 
 include(CPack)
 
-ENDif(EXISTS "${CMAKE_ROOT}/Modules/CPack.cmake")
+ENDif(EXISTS "${CMAKE_ROOT}/Modules/CPack.cmake")
diff --git a/cmake/checks/cpu_vsx.cpp b/cmake/checks/cpu_vsx.cpp
@@ -0,0 +1,12 @@
+# if defined(__VSX__)
+#   include <altivec.h>
+# else
+#   error "VSX is not supported"
+# endif
+
+int main()
+{
+    __vector float testF = vec_splats(0.f);
+    testF = vec_madd(testF, testF, testF);
+    return 0;
+}
diff --git a/modules/core/include/opencv2/core/base.hpp b/modules/core/include/opencv2/core/base.hpp
@@ -740,5 +740,6 @@ CV_EXPORTS_W void setUseIPP_NE(bool flag);
 } // cv
 
 #include "opencv2/core/neon_utils.hpp"
+#include "opencv2/core/vsx_utils.hpp"
 
 #endif //OPENCV_CORE_BASE_HPP
diff --git a/modules/core/include/opencv2/core/cv_cpu_dispatch.h b/modules/core/include/opencv2/core/cv_cpu_dispatch.h
@@ -99,6 +99,14 @@
 #  include <arm_neon.h>
 #endif
 
+#if defined(__VSX__) && defined(__PPC64__) && defined(__LITTLE_ENDIAN__)
+#  include <altivec.h>
+#  undef vector
+#  undef pixel
+#  undef bool
+#  define CV_VSX 1
+#endif
+
 #endif // CV_ENABLE_INTRINSICS && !CV_DISABLE_OPTIMIZATION && !__CUDACC__
 
 #if defined CV_CPU_COMPILE_AVX && !defined CV_CPU_BASELINE_COMPILE_AVX
@@ -135,6 +143,12 @@ struct VZeroUpperGuard {
 #elif defined(__ARM_NEON__) || (defined (__ARM_NEON) && defined(__aarch64__))
 #  include <arm_neon.h>
 #  define CV_NEON 1
+#elif defined(__VSX__) && defined(__PPC64__) && defined(__LITTLE_ENDIAN__)
+#  include <altivec.h>
+#  undef vector
+#  undef pixel
+#  undef bool
+#  define CV_VSX 1
 #endif
 
 #endif // !__OPENCV_BUILD && !__CUDACC (Compatibility code)
@@ -208,3 +222,7 @@ struct VZeroUpperGuard {
 #ifndef CV_NEON
 #  define CV_NEON 0
 #endif
+
+#ifndef CV_VSX
+#  define CV_VSX 0
+#endif
diff --git a/modules/core/include/opencv2/core/cv_cpu_helper.h b/modules/core/include/opencv2/core/cv_cpu_helper.h
@@ -180,5 +180,20 @@
 #endif
 #define __CV_CPU_DISPATCH_CHAIN_NEON(fn, args, mode, ...)  CV_CPU_CALL_NEON(fn, args); __CV_EXPAND(__CV_CPU_DISPATCH_CHAIN_ ## mode(fn, args, __VA_ARGS__))
 
+#if !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_COMPILE_VSX
+#  define CV_TRY_VSX 1
+#  define CV_CPU_HAS_SUPPORT_VSX 1
+#  define CV_CPU_CALL_VSX(fn, args) return (opt_VSX::fn args)
+#elif !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_DISPATCH_COMPILE_VSX
+#  define CV_TRY_VSX 1
+#  define CV_CPU_HAS_SUPPORT_VSX (cv::checkHardwareSupport(CV_CPU_VSX))
+#  define CV_CPU_CALL_VSX(fn, args) if (CV_CPU_HAS_SUPPORT_VSX) return (opt_VSX::fn args)
+#else
+#  define CV_TRY_VSX 0
+#  define CV_CPU_HAS_SUPPORT_VSX 0
+#  define CV_CPU_CALL_VSX(fn, args)
+#endif
+#define __CV_CPU_DISPATCH_CHAIN_VSX(fn, args, mode, ...)  CV_CPU_CALL_VSX(fn, args); __CV_EXPAND(__CV_CPU_DISPATCH_CHAIN_ ## mode(fn, args, __VA_ARGS__))
+
 #define CV_CPU_CALL_BASELINE(fn, args) return (cpu_baseline::fn args)
 #define __CV_CPU_DISPATCH_CHAIN_BASELINE(fn, args, mode, ...)  CV_CPU_CALL_BASELINE(fn, args) /* last in sequence */
diff --git a/modules/core/include/opencv2/core/cvdef.h b/modules/core/include/opencv2/core/cvdef.h
@@ -153,6 +153,8 @@ namespace cv { namespace debug_build_guard { } using namespace debug_build_guard
 
 #define CV_CPU_NEON   100
 
+#define CV_CPU_VSX 200
+
 // when adding to this list remember to update the following enum
 #define CV_HARDWARE_MAX_FEATURE 255
 
@@ -182,7 +184,9 @@ enum CpuFeatures {
     CPU_AVX_512VBMI     = 20,
     CPU_AVX_512VL       = 21,
 
-    CPU_NEON            = 100
+    CPU_NEON            = 100,
+
+    CPU_VSX             = 200
 };
 
 
diff --git a/modules/core/include/opencv2/core/hal/intrin.hpp b/modules/core/include/opencv2/core/hal/intrin.hpp
@@ -308,6 +308,7 @@ CV_CPU_OPTIMIZATION_HAL_NAMESPACE_END
 #ifdef CV_DOXYGEN
 #   undef CV_SSE2
 #   undef CV_NEON
+#   undef CV_VSX
 #endif
 
 #if CV_SSE2
@@ -318,6 +319,10 @@ CV_CPU_OPTIMIZATION_HAL_NAMESPACE_END
 
 #include "opencv2/core/hal/intrin_neon.hpp"
 
+#elif CV_VSX
+
+#include "opencv2/core/hal/intrin_vsx.hpp"
+
 #else
 
 #include "opencv2/core/hal/intrin_cpp.hpp"
diff --git a/modules/core/include/opencv2/core/hal/intrin_vsx.hpp b/modules/core/include/opencv2/core/hal/intrin_vsx.hpp
diff --git a/modules/core/include/opencv2/core/vsx_utils.hpp b/modules/core/include/opencv2/core/vsx_utils.hpp
diff --git a/modules/core/src/precomp.hpp b/modules/core/src/precomp.hpp
diff --git a/modules/core/src/system.cpp b/modules/core/src/system.cpp
diff --git a/modules/ts/src/ts_func.cpp b/modules/ts/src/ts_func.cpp
