Merge pull request #2836 from s98felix:2.4

modules/core/doc/utility_and_system_functions_and_macros.rst

@@ -317,6 +317,7 @@ Returns true if the specified feature is supported by the host hardware.
                         * ``CV_CPU_SSE4_2`` - SSE 4.2
                         * ``CV_CPU_POPCNT`` - POPCOUNT
                         * ``CV_CPU_AVX`` - AVX
+                        * ``CV_CPU_AVX2`` - AVX2
 
 The function returns true if the host hardware supports the specified feature. When user calls ``setUseOptimized(false)``, the subsequent calls to ``checkHardwareSupport()`` will return false until ``setUseOptimized(true)`` is called. This way user can dynamically switch on and off the optimized code in OpenCV.
 

modules/core/include/opencv2/core/core.hpp

@@ -284,6 +284,7 @@ CV_EXPORTS_W int64 getCPUTickCount();
   - CV_CPU_SSE4_2 - SSE 4.2
   - CV_CPU_POPCNT - POPCOUNT
   - CV_CPU_AVX - AVX
+  - CV_CPU_AVX2 - AVX2
 
   \note {Note that the function output is not static. Once you called cv::useOptimized(false),
   most of the hardware acceleration is disabled and thus the function will returns false,

modules/core/include/opencv2/core/core_c.h

@@ -1706,6 +1706,7 @@ CVAPI(double) cvGetTickFrequency( void );
 #define CV_CPU_SSE4_2  7
 #define CV_CPU_POPCNT  8
 #define CV_CPU_AVX    10
+#define CV_CPU_AVX2   11
 #define CV_HARDWARE_MAX_FEATURE 255
 
 CVAPI(int) cvCheckHardwareSupport(int feature);

modules/core/include/opencv2/core/internal.hpp

@@ -141,6 +141,10 @@ CV_INLINE IppiSize ippiSize(const cv::Size & _size)
 #      define __xgetbv() 0
 #    endif
 #  endif
+#  if defined __AVX2__
+#    include <immintrin.h>
+#    define CV_AVX2 1
+#  endif
 #endif
 
 
@@ -176,6 +180,9 @@ CV_INLINE IppiSize ippiSize(const cv::Size & _size)
 #ifndef CV_AVX
 #  define CV_AVX 0
 #endif
+#ifndef CV_AVX2
+#  define CV_AVX2 0
+#endif
 #ifndef CV_NEON
 #  define CV_NEON 0
 #endif

modules/core/src/system.cpp

@@ -253,6 +253,39 @@ struct HWFeatures
             f.have[CV_CPU_AVX]    = (((cpuid_data[2] & (1<<28)) != 0)&&((cpuid_data[2] & (1<<27)) != 0));//OS uses XSAVE_XRSTORE and CPU support AVX
         }
 
+    #if defined _MSC_VER && (defined _M_IX86 || defined _M_X64)
+        __cpuidex(cpuid_data, 7, 0);
+    #elif defined __GNUC__ && (defined __i386__ || defined __x86_64__)
+        #ifdef __x86_64__
+        asm __volatile__
+        (
+         "movl $7, %%eax\n\t"
+         "movl $0, %%ecx\n\t"
+         "cpuid\n\t"
+         :[eax]"=a"(cpuid_data[0]),[ebx]"=b"(cpuid_data[1]),[ecx]"=c"(cpuid_data[2]),[edx]"=d"(cpuid_data[3])
+         :
+         : "cc"
+        );
+        #else
+        asm volatile
+        (
+         "pushl %%ebx\n\t"
+         "movl $7,%%eax\n\t"
+         "movl $0,%%ecx\n\t"
+         "cpuid\n\t"
+         "popl %%ebx\n\t"
+         : "=a"(cpuid_data[0]), "=b"(cpuid_data[1]), "=c"(cpuid_data[2]), "=d"(cpuid_data[3])
+         :
+         : "cc"
+        );
+        #endif
+    #endif
+
+        if( f.x86_family >= 6 )
+        {
+            f.have[CV_CPU_AVX2] = (cpuid_data[1] & (1<<5)) != 0;
+        }
+
         return f;
     }