potok/opencv
1
0
Fork 0
mirror of https://github.com/opencv/opencv.git synced 2026-07-29 23:33:05 +04:00
Code Packages Releases Wiki Activity

Merge remote-tracking branch 'upstream/3.4' into merge-3.4

Browse Source
This commit is contained in:
Alexander Alekhin
2019-07-02 21:17:45 +00:00
parent 2bf8793632 2370079220
commit 097d81363b
35 changed files with 467 additions and 540 deletions
Download Patch File Download Diff File Expand all files Collapse all files
modules/core/include/opencv2/core/base.hpp
+2 -2
View File
@@ -301,8 +301,8 @@ CV_EXPORTS CV_NORETURN void error(int _code, const String& _err, const char* _fu
// In practice, some macro are not processed correctly (noreturn is not detected).
// We need to use simplified definition for them.
#define CV_Error(...) do { abort(); } while (0)
#define CV_Error_( code, args ) do { cv::format args; abort(); } while (0)
#define CV_Error(code, msg) do { (void)(code); (void)(msg); abort(); } while (0)
#define CV_Error_(code, args) do { (void)(code); (void)(cv::format args); abort(); } while (0)
#define CV_Assert( expr ) do { if (!(expr)) abort(); } while (0)
#else // CV_STATIC_ANALYSIS
modules/core/include/opencv2/core/hal/intrin.hpp
+28 -26
View File
@@ -55,6 +55,34 @@
#define OPENCV_HAL_NOP(a) (a)
#define OPENCV_HAL_1ST(a, b) (a)
namespace {
inline unsigned int trailingZeros32(unsigned int value) {
#if defined(_MSC_VER)
#if (_MSC_VER < 1700) || defined(_M_ARM)
unsigned long index = 0;
_BitScanForward(&index, value);
return (unsigned int)index;
#elif defined(__clang__)
// clang-cl doesn't export _tzcnt_u32 for non BMI systems
return value ? __builtin_ctz(value) : 32;
#else
return _tzcnt_u32(value);
#endif
#elif defined(__GNUC__) || defined(__GNUG__)
return __builtin_ctz(value);
#elif defined(__ICC) || defined(__INTEL_COMPILER)
return _bit_scan_forward(value);
#elif defined(__clang__)
return llvm.cttz.i32(value, true);
#else
static const int MultiplyDeBruijnBitPosition[32] = {
0, 1, 28, 2, 29, 14, 24, 3, 30, 22, 20, 15, 25, 17, 4, 8,
31, 27, 13, 23, 21, 19, 16, 7, 26, 12, 18, 6, 11, 5, 10, 9 };
return MultiplyDeBruijnBitPosition[((uint32_t)((value & -value) * 0x077CB531U)) >> 27];
#endif
}
}
// unlike HAL API, which is in cv::hal,
// we put intrinsics into cv namespace to make its
// access from within opencv code more accessible
@@ -419,32 +447,6 @@ namespace CV__SIMD_NAMESPACE {
using namespace CV__SIMD_NAMESPACE;
#endif
inline unsigned int trailingZeros32(unsigned int value) {
#if defined(_MSC_VER)
#if (_MSC_VER < 1700) || defined(_M_ARM)
unsigned long index = 0;
_BitScanForward(&index, value);
return (unsigned int)index;
#elif defined(__clang__)
// clang-cl doesn't export _tzcnt_u32 for non BMI systems
return value ? __builtin_ctz(value) : 32;
#else
return _tzcnt_u32(value);
#endif
#elif defined(__GNUC__) || defined(__GNUG__)
return __builtin_ctz(value);
#elif defined(__ICC) || defined(__INTEL_COMPILER)
return _bit_scan_forward(value);
#elif defined(__clang__)
return llvm.cttz.i32(value, true);
#else
static const int MultiplyDeBruijnBitPosition[32] = {
0, 1, 28, 2, 29, 14, 24, 3, 30, 22, 20, 15, 25, 17, 4, 8,
31, 27, 13, 23, 21, 19, 16, 7, 26, 12, 18, 6, 11, 5, 10, 9 };
return MultiplyDeBruijnBitPosition[((uint32_t)((value & -value) * 0x077CB531U)) >> 27];
#endif
}
#ifndef CV_DOXYGEN
CV_CPU_OPTIMIZATION_HAL_NAMESPACE_END
#endif
modules/core/include/opencv2/core/hal/intrin_avx.hpp
+11
View File
@@ -1244,6 +1244,17 @@ inline int v_signmask(const v_float32x8& a)
inline int v_signmask(const v_float64x4& a)
{ return _mm256_movemask_pd(a.val); }
inline int v_scan_forward(const v_int8x32& a) { return trailingZeros32(v_signmask(v_reinterpret_as_s8(a))); }
inline int v_scan_forward(const v_uint8x32& a) { return trailingZeros32(v_signmask(v_reinterpret_as_s8(a))); }
inline int v_scan_forward(const v_int16x16& a) { return trailingZeros32(v_signmask(v_reinterpret_as_s8(a))) / 2; }
inline int v_scan_forward(const v_uint16x16& a) { return trailingZeros32(v_signmask(v_reinterpret_as_s8(a))) / 2; }
inline int v_scan_forward(const v_int32x8& a) { return trailingZeros32(v_signmask(v_reinterpret_as_s8(a))) / 4; }
inline int v_scan_forward(const v_uint32x8& a) { return trailingZeros32(v_signmask(v_reinterpret_as_s8(a))) / 4; }
inline int v_scan_forward(const v_float32x8& a) { return trailingZeros32(v_signmask(v_reinterpret_as_s8(a))) / 4; }
inline int v_scan_forward(const v_int64x4& a) { return trailingZeros32(v_signmask(v_reinterpret_as_s8(a))) / 8; }
inline int v_scan_forward(const v_uint64x4& a) { return trailingZeros32(v_signmask(v_reinterpret_as_s8(a))) / 8; }
inline int v_scan_forward(const v_float64x4& a) { return trailingZeros32(v_signmask(v_reinterpret_as_s8(a))) / 8; }
/** Checks **/
#define OPENCV_HAL_IMPL_AVX_CHECK(_Tpvec, and_op, allmask) \
inline bool v_check_all(const _Tpvec& a) \
modules/core/include/opencv2/core/hal/intrin_avx512.hpp
+18 -2
View File
@@ -2719,7 +2719,7 @@ OPENCV_HAL_IMPL_AVX512_LOADSTORE_INTERLEAVE(v_float64x8, double, f64, v_uint64x8
////////// Mask and checks /////////
/** Mask **/
inline int64 v_signmask(const v_int8x64& a) { return (int64)_mm512_cmp_epi8_mask(a.val, _mm512_setzero_si512(), _MM_CMPINT_LT); }
inline int64 v_signmask(const v_int8x64& a) { return (int64)_mm512_movepi8_mask(a.val); }
inline int v_signmask(const v_int16x32& a) { return (int)_mm512_cmp_epi16_mask(a.val, _mm512_setzero_si512(), _MM_CMPINT_LT); }
inline int v_signmask(const v_int32x16& a) { return (int)_mm512_cmp_epi32_mask(a.val, _mm512_setzero_si512(), _MM_CMPINT_LT); }
inline int v_signmask(const v_int64x8& a) { return (int)_mm512_cmp_epi64_mask(a.val, _mm512_setzero_si512(), _MM_CMPINT_LT); }
@@ -2733,7 +2733,7 @@ inline int v_signmask(const v_float64x8& a) { return v_signmask(v_reinterpret_as
/** Checks **/
inline bool v_check_all(const v_int8x64& a) { return !(bool)_mm512_cmp_epi8_mask(a.val, _mm512_setzero_si512(), _MM_CMPINT_NLT); }
inline bool v_check_any(const v_int8x64& a) { return (bool)_mm512_cmp_epi8_mask(a.val, _mm512_setzero_si512(), _MM_CMPINT_LT); }
inline bool v_check_any(const v_int8x64& a) { return (bool)_mm512_movepi8_mask(a.val); }
inline bool v_check_all(const v_int16x32& a) { return !(bool)_mm512_cmp_epi16_mask(a.val, _mm512_setzero_si512(), _MM_CMPINT_NLT); }
inline bool v_check_any(const v_int16x32& a) { return (bool)_mm512_cmp_epi16_mask(a.val, _mm512_setzero_si512(), _MM_CMPINT_LT); }
inline bool v_check_all(const v_int32x16& a) { return !(bool)_mm512_cmp_epi32_mask(a.val, _mm512_setzero_si512(), _MM_CMPINT_NLT); }
@@ -2754,6 +2754,22 @@ inline bool v_check_any(const v_uint16x32& a) { return v_check_any(v_reinterpret
inline bool v_check_any(const v_uint32x16& a) { return v_check_any(v_reinterpret_as_s32(a)); }
inline bool v_check_any(const v_uint64x8& a) { return v_check_any(v_reinterpret_as_s64(a)); }
inline int v_scan_forward(const v_int8x64& a)
{
int64 mask = _mm512_movepi8_mask(a.val);
int mask32 = (int)mask;
return mask != 0 ? mask32 != 0 ? trailingZeros32(mask32) : 32 + trailingZeros32((int)(mask >> 32)) : 0;
}
inline int v_scan_forward(const v_uint8x64& a) { return v_scan_forward(v_reinterpret_as_s8(a)); }
inline int v_scan_forward(const v_int16x32& a) { return trailingZeros32(v_signmask(v_reinterpret_as_s16(a))); }
inline int v_scan_forward(const v_uint16x32& a) { return trailingZeros32(v_signmask(v_reinterpret_as_s16(a))); }
inline int v_scan_forward(const v_int32x16& a) { return trailingZeros32(v_signmask(v_reinterpret_as_s16(a))) / 2; }
inline int v_scan_forward(const v_uint32x16& a) { return trailingZeros32(v_signmask(v_reinterpret_as_s16(a))) / 2; }
inline int v_scan_forward(const v_float32x16& a) { return trailingZeros32(v_signmask(v_reinterpret_as_s16(a))) / 2; }
inline int v_scan_forward(const v_int64x8& a) { return trailingZeros32(v_signmask(v_reinterpret_as_s16(a))) / 4; }
inline int v_scan_forward(const v_uint64x8& a) { return trailingZeros32(v_signmask(v_reinterpret_as_s16(a))) / 4; }
inline int v_scan_forward(const v_float64x8& a) { return trailingZeros32(v_signmask(v_reinterpret_as_s16(a))) / 4; }
inline void v512_cleanup() { _mm256_zeroall(); }
CV_CPU_OPTIMIZATION_HAL_NAMESPACE_END
modules/core/include/opencv2/core/hal/intrin_cpp.hpp
+18
View File
@@ -1072,6 +1072,7 @@ template<typename _Tp, int n> inline typename V_TypeTraits< typename V_TypeTrait
}
/** @brief Get negative values mask
@deprecated v_signmask depends on a lane count heavily and therefore isn't universal enough
Returned value is a bit mask with bits set to 1 on places corresponding to negative packed values indexes.
Example:
@@ -1088,6 +1089,23 @@ template<typename _Tp, int n> inline int v_signmask(const v_reg<_Tp, n>& a)
return mask;
}
/** @brief Get first negative lane index
Returned value is an index of first negative lane (undefined for input of all positive values)
Example:
@code{.cpp}
v_int32x4 r; // set to {0, 0, -1, -1}
int idx = v_heading_zeros(r); // idx = 2
@endcode
*/
template <typename _Tp, int n> inline int v_scan_forward(const v_reg<_Tp, n>& a)
{
for (int i = 0; i < n; i++)
if(V_TypeTraits<_Tp>::reinterpret_int(a.s[i]) < 0)
return i;
return 0;
}
/** @brief Check if all packed values are less than zero
Unsigned values will be casted to signed: `uchar 254 => char -2`.
modules/core/include/opencv2/core/hal/intrin_neon.hpp
+16 -1
View File
@@ -1096,17 +1096,32 @@ inline int v_signmask(const v_int32x4& a)
{ return v_signmask(v_reinterpret_as_u32(a)); }
inline int v_signmask(const v_float32x4& a)
{ return v_signmask(v_reinterpret_as_u32(a)); }
#if CV_SIMD128_64F
inline int v_signmask(const v_uint64x2& a)
{
int64x1_t m0 = vdup_n_s64(0);
uint64x2_t v0 = vshlq_u64(vshrq_n_u64(a.val, 63), vcombine_s64(m0, m0));
return (int)vgetq_lane_u64(v0, 0) + ((int)vgetq_lane_u64(v0, 1) << 1);
}
inline int v_signmask(const v_int64x2& a)
{ return v_signmask(v_reinterpret_as_u64(a)); }
#if CV_SIMD128_64F
inline int v_signmask(const v_float64x2& a)
{ return v_signmask(v_reinterpret_as_u64(a)); }
#endif
inline int v_scan_forward(const v_int8x16& a) { return trailingZeros32(v_signmask(a)); }
inline int v_scan_forward(const v_uint8x16& a) { return trailingZeros32(v_signmask(a)); }
inline int v_scan_forward(const v_int16x8& a) { return trailingZeros32(v_signmask(a)); }
inline int v_scan_forward(const v_uint16x8& a) { return trailingZeros32(v_signmask(a)); }
inline int v_scan_forward(const v_int32x4& a) { return trailingZeros32(v_signmask(a)); }
inline int v_scan_forward(const v_uint32x4& a) { return trailingZeros32(v_signmask(a)); }
inline int v_scan_forward(const v_float32x4& a) { return trailingZeros32(v_signmask(a)); }
inline int v_scan_forward(const v_int64x2& a) { return trailingZeros32(v_signmask(a)); }
inline int v_scan_forward(const v_uint64x2& a) { return trailingZeros32(v_signmask(a)); }
#if CV_SIMD128_64F
inline int v_scan_forward(const v_float64x2& a) { return trailingZeros32(v_signmask(a)); }
#endif
#define OPENCV_HAL_IMPL_NEON_CHECK_ALLANY(_Tpvec, suffix, shift) \
inline bool v_check_all(const v_##_Tpvec& a) \
{ \
modules/core/include/opencv2/core/hal/intrin_sse.hpp
+11
View File
@@ -1617,6 +1617,17 @@ OPENCV_HAL_IMPL_SSE_CHECK_SIGNS(v_int32x4, epi8, v_packq_epi32, OPENCV_HAL_AND,
OPENCV_HAL_IMPL_SSE_CHECK_SIGNS(v_float32x4, ps, OPENCV_HAL_NOP, OPENCV_HAL_1ST, 15, 15)
OPENCV_HAL_IMPL_SSE_CHECK_SIGNS(v_float64x2, pd, OPENCV_HAL_NOP, OPENCV_HAL_1ST, 3, 3)
inline int v_scan_forward(const v_int8x16& a) { return trailingZeros32(v_signmask(v_reinterpret_as_s8(a))); }
inline int v_scan_forward(const v_uint8x16& a) { return trailingZeros32(v_signmask(v_reinterpret_as_s8(a))); }
inline int v_scan_forward(const v_int16x8& a) { return trailingZeros32(v_signmask(v_reinterpret_as_s8(a))) / 2; }
inline int v_scan_forward(const v_uint16x8& a) { return trailingZeros32(v_signmask(v_reinterpret_as_s8(a))) / 2; }
inline int v_scan_forward(const v_int32x4& a) { return trailingZeros32(v_signmask(v_reinterpret_as_s8(a))) / 4; }
inline int v_scan_forward(const v_uint32x4& a) { return trailingZeros32(v_signmask(v_reinterpret_as_s8(a))) / 4; }
inline int v_scan_forward(const v_float32x4& a) { return trailingZeros32(v_signmask(v_reinterpret_as_s8(a))) / 4; }
inline int v_scan_forward(const v_int64x2& a) { return trailingZeros32(v_signmask(v_reinterpret_as_s8(a))) / 8; }
inline int v_scan_forward(const v_uint64x2& a) { return trailingZeros32(v_signmask(v_reinterpret_as_s8(a))) / 8; }
inline int v_scan_forward(const v_float64x2& a) { return trailingZeros32(v_signmask(v_reinterpret_as_s8(a))) / 8; }
#if CV_SSE4_1
#define OPENCV_HAL_IMPL_SSE_SELECT(_Tpvec, cast_ret, cast, suffix) \
inline _Tpvec v_select(const _Tpvec& mask, const _Tpvec& a, const _Tpvec& b) \
modules/core/include/opencv2/core/hal/intrin_vsx.hpp
+11
View File
@@ -891,6 +891,17 @@ inline int v_signmask(const v_uint64x2& a)
inline int v_signmask(const v_float64x2& a)
{ return v_signmask(v_reinterpret_as_s64(a)); }
inline int v_scan_forward(const v_int8x16& a) { return trailingZeros32(v_signmask(a)); }
inline int v_scan_forward(const v_uint8x16& a) { return trailingZeros32(v_signmask(a)); }
inline int v_scan_forward(const v_int16x8& a) { return trailingZeros32(v_signmask(a)); }
inline int v_scan_forward(const v_uint16x8& a) { return trailingZeros32(v_signmask(a)); }
inline int v_scan_forward(const v_int32x4& a) { return trailingZeros32(v_signmask(a)); }
inline int v_scan_forward(const v_uint32x4& a) { return trailingZeros32(v_signmask(a)); }
inline int v_scan_forward(const v_float32x4& a) { return trailingZeros32(v_signmask(a)); }
inline int v_scan_forward(const v_int64x2& a) { return trailingZeros32(v_signmask(a)); }
inline int v_scan_forward(const v_uint64x2& a) { return trailingZeros32(v_signmask(a)); }
inline int v_scan_forward(const v_float64x2& a) { return trailingZeros32(v_signmask(a)); }
template<typename _Tpvec>
inline bool v_check_all(const _Tpvec& a)
{ return vec_all_lt(a.val, _Tpvec().val); }
modules/core/include/opencv2/core/matx.hpp
+4
View File
@@ -385,6 +385,10 @@ public:
const _Tp& operator ()(int i) const;
_Tp& operator ()(int i);
#ifdef CV_CXX11
Vec<_Tp, cn>& operator=(const Vec<_Tp, cn>& rhs) = default;
#endif
Vec(const Matx<_Tp, cn, 1>& a, const Matx<_Tp, cn, 1>& b, Matx_AddOp);
Vec(const Matx<_Tp, cn, 1>& a, const Matx<_Tp, cn, 1>& b, Matx_SubOp);
template<typename _T2> Vec(const Matx<_Tp, cn, 1>& a, _T2 alpha, Matx_ScaleOp);