use universal intrinsic and SSE4 popcount instruction in normHamming

- add v_popcount in universal intrinsic
 - add test for v_popcount
 - add wrapper of popcount for both MSVC and GCC

modules/core/include/opencv2/core/cvdef.h

@@ -188,8 +188,16 @@ enum CpuFeatures {
 #  if defined __POPCNT__ || (defined _MSC_VER && _MSC_VER >= 1500)
 #    ifdef _MSC_VER
 #      include <nmmintrin.h>
+#      if defined(_M_X64)
+#        define CV_POPCNT_U64 _mm_popcnt_u64
+#      endif
+#      define CV_POPCNT_U32 _mm_popcnt_u32
 #    else
 #      include <popcntintrin.h>
+#      if defined(__x86_64__)
+#        define CV_POPCNT_U64 __builtin_popcountll
+#      endif
+#      define CV_POPCNT_U32 __builtin_popcount
 #    endif
 #    define CV_POPCNT 1
 #  endif

modules/core/include/opencv2/core/hal/intrin_cpp.hpp

@@ -149,7 +149,7 @@ Element-wise binary and unary operations.
 
 Most of these operations return only one value.
 
-- Reduce: @ref v_reduce_min, @ref v_reduce_max, @ref v_reduce_sum
+- Reduce: @ref v_reduce_min, @ref v_reduce_max, @ref v_reduce_sum, @ref v_popcount
 - Mask: @ref v_signmask, @ref v_check_all, @ref v_check_any, @ref v_select
 
 ### Other math
@@ -574,6 +574,49 @@ Scheme:
 For 32-bit integer and 32-bit floating point types. */
 OPENCV_HAL_IMPL_REDUCE_MINMAX_FUNC(v_reduce_max, std::max)
 
+static const unsigned char popCountTable[] =
+{
+    0, 1, 1, 2, 1, 2, 2, 3, 1, 2, 2, 3, 2, 3, 3, 4,
+    1, 2, 2, 3, 2, 3, 3, 4, 2, 3, 3, 4, 3, 4, 4, 5,
+    1, 2, 2, 3, 2, 3, 3, 4, 2, 3, 3, 4, 3, 4, 4, 5,
+    2, 3, 3, 4, 3, 4, 4, 5, 3, 4, 4, 5, 4, 5, 5, 6,
+    1, 2, 2, 3, 2, 3, 3, 4, 2, 3, 3, 4, 3, 4, 4, 5,
+    2, 3, 3, 4, 3, 4, 4, 5, 3, 4, 4, 5, 4, 5, 5, 6,
+    2, 3, 3, 4, 3, 4, 4, 5, 3, 4, 4, 5, 4, 5, 5, 6,
+    3, 4, 4, 5, 4, 5, 5, 6, 4, 5, 5, 6, 5, 6, 6, 7,
+    1, 2, 2, 3, 2, 3, 3, 4, 2, 3, 3, 4, 3, 4, 4, 5,
+    2, 3, 3, 4, 3, 4, 4, 5, 3, 4, 4, 5, 4, 5, 5, 6,
+    2, 3, 3, 4, 3, 4, 4, 5, 3, 4, 4, 5, 4, 5, 5, 6,
+    3, 4, 4, 5, 4, 5, 5, 6, 4, 5, 5, 6, 5, 6, 6, 7,
+    2, 3, 3, 4, 3, 4, 4, 5, 3, 4, 4, 5, 4, 5, 5, 6,
+    3, 4, 4, 5, 4, 5, 5, 6, 4, 5, 5, 6, 5, 6, 6, 7,
+    3, 4, 4, 5, 4, 5, 5, 6, 4, 5, 5, 6, 5, 6, 6, 7,
+    4, 5, 5, 6, 5, 6, 6, 7, 5, 6, 6, 7, 6, 7, 7, 8,
+};
+/** @brief Count the 1 bits in the vector and return 4 values
+
+Scheme:
+@code
+{A1 A2 A3 ...} => popcount(A1)
+@endcode
+Any types but result will be in v_uint32x4*/
+template<typename _Tp, int n> inline v_uint32x4 v_popcount(const v_reg<_Tp, n>& a)
+{
+    v_uint8x16 b;
+    b = v_reinterpret_as_u8(a);
+    for( int i = 0; i < v_uint8x16::nlanes; i++ )
+    {
+        b.s[i] = popCountTable[b.s[i]];
+    }
+    v_uint32x4 c;
+    for( int i = 0; i < v_uint32x4::nlanes; i++ )
+    {
+        c.s[i] = b.s[i*4] + b.s[i*4+1] + b.s[i*4+2] + b.s[i*4+3];
+    }
+    return c;
+}
+
+
 //! @cond IGNORED
 template<typename _Tp, int n>
 inline void v_minmax( const v_reg<_Tp, n>& a, const v_reg<_Tp, n>& b,

modules/core/include/opencv2/core/hal/intrin_neon.hpp

@@ -813,6 +813,22 @@ OPENCV_HAL_IMPL_NEON_REDUCE_OP_4(v_float32x4, float32x2, float, sum, add, f32)
 OPENCV_HAL_IMPL_NEON_REDUCE_OP_4(v_float32x4, float32x2, float, max, max, f32)
 OPENCV_HAL_IMPL_NEON_REDUCE_OP_4(v_float32x4, float32x2, float, min, min, f32)
 
+#define OPENCV_HAL_IMPL_NEON_POPCOUNT(_Tpvec, cast) \
+inline v_uint32x4 v_popcount(const _Tpvec& a) \
+{ \
+    uint8x16_t t = vcntq_u8(cast(a.val)); \
+    uint16x8_t t0 = vpaddlq_u8(t);  /* 16 -> 8 */ \
+    uint32x4_t t1 = vpaddlq_u16(t0); /* 8 -> 4 */ \
+    return v_uint32x4(t1); \
+}
+
+OPENCV_HAL_IMPL_NEON_POPCOUNT(v_uint8x16, OPENCV_HAL_NOP)
+OPENCV_HAL_IMPL_NEON_POPCOUNT(v_uint16x8, vreinterpretq_u8_u16)
+OPENCV_HAL_IMPL_NEON_POPCOUNT(v_uint32x4, vreinterpretq_u8_u32)
+OPENCV_HAL_IMPL_NEON_POPCOUNT(v_int8x16, vreinterpretq_u8_s8)
+OPENCV_HAL_IMPL_NEON_POPCOUNT(v_int16x8, vreinterpretq_u8_s16)
+OPENCV_HAL_IMPL_NEON_POPCOUNT(v_int32x4, vreinterpretq_u8_s32)
+
 inline int v_signmask(const v_uint8x16& a)
 {
     int8x8_t m0 = vcreate_s8(CV_BIG_UINT(0x0706050403020100));

modules/core/include/opencv2/core/hal/intrin_sse.hpp

@@ -1121,6 +1121,28 @@ OPENCV_HAL_IMPL_SSE_REDUCE_OP_4(v_float32x4, float, sum, OPENCV_HAL_ADD)
 OPENCV_HAL_IMPL_SSE_REDUCE_OP_4(v_float32x4, float, max, std::max)
 OPENCV_HAL_IMPL_SSE_REDUCE_OP_4(v_float32x4, float, min, std::min)
 
+#define OPENCV_HAL_IMPL_SSE_POPCOUNT(_Tpvec) \
+inline v_uint32x4 v_popcount(const _Tpvec& a) \
+{ \
+    __m128i m1 = _mm_set1_epi32(0x55555555); \
+    __m128i m2 = _mm_set1_epi32(0x33333333); \
+    __m128i m4 = _mm_set1_epi32(0x0f0f0f0f); \
+    __m128i p = a.val; \
+    p = _mm_add_epi32(_mm_and_si128(_mm_srli_epi32(p, 1), m1), _mm_and_si128(p, m1)); \
+    p = _mm_add_epi32(_mm_and_si128(_mm_srli_epi32(p, 2), m2), _mm_and_si128(p, m2)); \
+    p = _mm_add_epi32(_mm_and_si128(_mm_srli_epi32(p, 4), m4), _mm_and_si128(p, m4)); \
+    p = _mm_adds_epi8(p, _mm_srli_si128(p, 1)); \
+    p = _mm_adds_epi8(p, _mm_srli_si128(p, 2)); \
+    return v_uint32x4(_mm_and_si128(p, _mm_set1_epi32(0x000000ff))); \
+}
+
+OPENCV_HAL_IMPL_SSE_POPCOUNT(v_uint8x16)
+OPENCV_HAL_IMPL_SSE_POPCOUNT(v_uint16x8)
+OPENCV_HAL_IMPL_SSE_POPCOUNT(v_uint32x4)
+OPENCV_HAL_IMPL_SSE_POPCOUNT(v_int8x16)
+OPENCV_HAL_IMPL_SSE_POPCOUNT(v_int16x8)
+OPENCV_HAL_IMPL_SSE_POPCOUNT(v_int32x4)
+
 #define OPENCV_HAL_IMPL_SSE_CHECK_SIGNS(_Tpvec, suffix, pack_op, and_op, signmask, allmask) \
 inline int v_signmask(const _Tpvec& a) \
 { \

modules/core/src/stat.cpp

@@ -44,6 +44,7 @@
 #include "precomp.hpp"
 #include <climits>
 #include <limits>
+#include "opencv2/core/hal/intrin.hpp"
 
 #include "opencl_kernels_core.hpp"
 
@@ -4238,21 +4239,7 @@ int normHamming(const uchar* a, int n)
 {
     int i = 0;
     int result = 0;
-#if CV_NEON
-    {
-        uint32x4_t bits = vmovq_n_u32(0);
-        for (; i <= n - 16; i += 16) {
-            uint8x16_t A_vec = vld1q_u8 (a + i);
-            uint8x16_t bitsSet = vcntq_u8 (A_vec);
-            uint16x8_t bitSet8 = vpaddlq_u8 (bitsSet);
-            uint32x4_t bitSet4 = vpaddlq_u16 (bitSet8);
-            bits = vaddq_u32(bits, bitSet4);
-        }
-        uint64x2_t bitSet2 = vpaddlq_u32 (bits);
-        result = vgetq_lane_s32 (vreinterpretq_s32_u64(bitSet2),0);
-        result += vgetq_lane_s32 (vreinterpretq_s32_u64(bitSet2),2);
-    }
-#elif CV_AVX2
+#if CV_AVX2
     if(USE_AVX2)
     {
         __m256i _r0 = _mm256_setzero_si256();
@@ -4274,12 +4261,40 @@ int normHamming(const uchar* a, int n)
         _r0 = _mm256_add_epi32(_r0, _mm256_shuffle_epi32(_r0, 2));
         result = _mm256_extract_epi32_(_mm256_add_epi32(_r0, _mm256_permute2x128_si256(_r0, _r0, 1)), 0);
     }
+#elif CV_POPCNT
+    if(checkHardwareSupport(CV_CPU_POPCNT))
+    {
+#  if defined CV_POPCNT_U64
+        for(; i <= n - 8; i += 8)
+        {
+            result += (int)CV_POPCNT_U64(*(uint64*)(a + i));
+        }
 #  endif
         for(; i <= n - 4; i += 4)
+        {
+            result += CV_POPCNT_U32(*(uint*)(a + i));
+        }
+    }
+#elif CV_SIMD128
+    if(hasSIMD128())
+    {
+        v_uint32x4 t = v_setzero_u32();
+        for(; i <= n - v_uint8x16::nlanes; i += v_uint8x16::nlanes)
+        {
+            t += v_popcount(v_load(a + i));
+        }
+        result = v_reduce_sum(t);
+    }
+#endif
+    for(; i <= n - 4; i += 4)
+    {
         result += popCountTable[a[i]] + popCountTable[a[i+1]] +
         popCountTable[a[i+2]] + popCountTable[a[i+3]];
+    }
     for(; i < n; i++)
+    {
         result += popCountTable[a[i]];
+    }
     return result;
 }
 
@@ -4287,23 +4302,7 @@ int normHamming(const uchar* a, const uchar* b, int n)
 {
     int i = 0;
     int result = 0;
-#if CV_NEON
-    {
-        uint32x4_t bits = vmovq_n_u32(0);
-        for (; i <= n - 16; i += 16) {
-            uint8x16_t A_vec = vld1q_u8 (a + i);
-            uint8x16_t B_vec = vld1q_u8 (b + i);
-            uint8x16_t AxorB = veorq_u8 (A_vec, B_vec);
-            uint8x16_t bitsSet = vcntq_u8 (AxorB);
-            uint16x8_t bitSet8 = vpaddlq_u8 (bitsSet);
-            uint32x4_t bitSet4 = vpaddlq_u16 (bitSet8);
-            bits = vaddq_u32(bits, bitSet4);
-        }
-        uint64x2_t bitSet2 = vpaddlq_u32 (bits);
-        result = vgetq_lane_s32 (vreinterpretq_s32_u64(bitSet2),0);
-        result += vgetq_lane_s32 (vreinterpretq_s32_u64(bitSet2),2);
-    }
-#elif CV_AVX2
+#if CV_AVX2
     if(USE_AVX2)
     {
         __m256i _r0 = _mm256_setzero_si256();
@@ -4328,12 +4327,40 @@ int normHamming(const uchar* a, const uchar* b, int n)
         _r0 = _mm256_add_epi32(_r0, _mm256_shuffle_epi32(_r0, 2));
         result = _mm256_extract_epi32_(_mm256_add_epi32(_r0, _mm256_permute2x128_si256(_r0, _r0, 1)), 0);
     }
+#elif CV_POPCNT
+    if(checkHardwareSupport(CV_CPU_POPCNT))
+    {
+#  if defined CV_POPCNT_U64
+        for(; i <= n - 8; i += 8)
+        {
+            result += (int)CV_POPCNT_U64(*(uint64*)(a + i) ^ *(uint64*)(b + i));
+        }
+#  endif
+        for(; i <= n - 4; i += 4)
+        {
+            result += CV_POPCNT_U32(*(uint*)(a + i) ^ *(uint*)(b + i));
+        }
+    }
+#elif CV_SIMD128
+    if(hasSIMD128())
+    {
+        v_uint32x4 t = v_setzero_u32();
+        for(; i <= n - v_uint8x16::nlanes; i += v_uint8x16::nlanes)
+        {
+            t += v_popcount(v_load(a + i) ^ v_load(b + i));
+        }
+        result = v_reduce_sum(t);
+    }
 #endif
     for(; i <= n - 4; i += 4)
+    {
         result += popCountTable[a[i] ^ b[i]] + popCountTable[a[i+1] ^ b[i+1]] +
                 popCountTable[a[i+2] ^ b[i+2]] + popCountTable[a[i+3] ^ b[i+3]];
+    }
     for(; i < n; i++)
+    {
         result += popCountTable[a[i] ^ b[i]];
+    }
     return result;
 }
 

modules/core/test/test_intrin.cpp

@@ -404,6 +404,18 @@ template<typename R> struct TheTest
         return *this;
     }
 
+    TheTest & test_popcount()
+    {
+        static unsigned popcountTable[] = {0, 1, 2, 4, 5, 7, 9, 12, 13, 15, 17, 20, 22, 25, 28, 32, 33};
+        Data<R> dataA;
+        R a = dataA;
+
+        unsigned resB = (unsigned)v_reduce_sum(v_popcount(a));
+        EXPECT_EQ(popcountTable[R::nlanes], resB);
+
+        return *this;
+    }
+
     TheTest & test_absdiff()
     {
         typedef typename V_RegTrait128<LaneType>::u_reg Ru;
@@ -798,6 +810,7 @@ TEST(hal_intrin, uint8x16) {
         .test_min_max()
         .test_absdiff()
         .test_mask()
+        .test_popcount()
         .test_pack<1>().test_pack<2>().test_pack<3>().test_pack<8>()
         .test_pack_u<1>().test_pack_u<2>().test_pack_u<3>().test_pack_u<8>()
         .test_unpack()
@@ -819,6 +832,7 @@ TEST(hal_intrin, int8x16) {
         .test_absdiff()
         .test_abs()
         .test_mask()
+        .test_popcount()
         .test_pack<1>().test_pack<2>().test_pack<3>().test_pack<8>()
         .test_unpack()
         .test_extract<0>().test_extract<1>().test_extract<8>().test_extract<15>()
@@ -844,6 +858,7 @@ TEST(hal_intrin, uint16x8) {
         .test_absdiff()
         .test_reduce()
         .test_mask()
+        .test_popcount()
         .test_pack<1>().test_pack<2>().test_pack<7>().test_pack<16>()
         .test_pack_u<1>().test_pack_u<2>().test_pack_u<7>().test_pack_u<16>()
         .test_unpack()
@@ -870,6 +885,7 @@ TEST(hal_intrin, int16x8) {
         .test_abs()
         .test_reduce()
         .test_mask()
+        .test_popcount()
         .test_pack<1>().test_pack<2>().test_pack<7>().test_pack<16>()
         .test_unpack()
         .test_extract<0>().test_extract<1>().test_extract<4>().test_extract<7>()
@@ -894,6 +910,7 @@ TEST(hal_intrin, uint32x4) {
         .test_absdiff()
         .test_reduce()
         .test_mask()
+        .test_popcount()
         .test_pack<1>().test_pack<2>().test_pack<15>().test_pack<32>()
         .test_unpack()
         .test_extract<0>().test_extract<1>().test_extract<2>().test_extract<3>()
@@ -910,6 +927,7 @@ TEST(hal_intrin, int32x4) {
         .test_mul()
         .test_abs()
         .test_cmp()
+        .test_popcount()
         .test_shift<1>().test_shift<8>()
         .test_logic()
         .test_min_max()