potok/opencv
1
0
Fork 0
mirror of https://github.com/opencv/opencv.git synced 2026-07-31 00:03:03 +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
2018-09-28 12:48:51 +03:00
parent 29e88e50ff 4b895a4d1f
commit a8b0db4e5d
77 changed files with 586 additions and 443 deletions
Download Patch File Download Diff File Expand all files Collapse all files
modules/core/include/opencv2/core/ocl.hpp
+75 -75
View File
@@ -59,7 +59,7 @@ CV_EXPORTS_W void finish();
CV_EXPORTS bool haveSVM();
class CV_EXPORTS Context;
class CV_EXPORTS Device;
class CV_EXPORTS_W_SIMPLE Device;
class CV_EXPORTS Kernel;
class CV_EXPORTS Program;
class CV_EXPORTS ProgramSource;
@@ -67,14 +67,14 @@ class CV_EXPORTS Queue;
class CV_EXPORTS PlatformInfo;
class CV_EXPORTS Image2D;
class CV_EXPORTS Device
class CV_EXPORTS_W_SIMPLE Device
{
public:
Device();
CV_WRAP Device();
explicit Device(void* d);
Device(const Device& d);
Device& operator = (const Device& d);
~Device();
CV_WRAP ~Device();
void set(void* d);
@@ -89,24 +89,24 @@ public:
TYPE_ALL = 0xFFFFFFFF
};
String name() const;
String extensions() const;
bool isExtensionSupported(const String& extensionName) const;
String version() const;
String vendorName() const;
String OpenCL_C_Version() const;
String OpenCLVersion() const;
int deviceVersionMajor() const;
int deviceVersionMinor() const;
String driverVersion() const;
CV_WRAP String name() const;
CV_WRAP String extensions() const;
CV_WRAP bool isExtensionSupported(const String& extensionName) const;
CV_WRAP String version() const;
CV_WRAP String vendorName() const;
CV_WRAP String OpenCL_C_Version() const;
CV_WRAP String OpenCLVersion() const;
CV_WRAP int deviceVersionMajor() const;
CV_WRAP int deviceVersionMinor() const;
CV_WRAP String driverVersion() const;
void* ptr() const;
int type() const;
CV_WRAP int type() const;
int addressBits() const;
bool available() const;
bool compilerAvailable() const;
bool linkerAvailable() const;
CV_WRAP int addressBits() const;
CV_WRAP bool available() const;
CV_WRAP bool compilerAvailable() const;
CV_WRAP bool linkerAvailable() const;
enum
{
@@ -119,21 +119,21 @@ public:
FP_SOFT_FLOAT=(1 << 6),
FP_CORRECTLY_ROUNDED_DIVIDE_SQRT=(1 << 7)
};
int doubleFPConfig() const;
int singleFPConfig() const;
int halfFPConfig() const;
CV_WRAP int doubleFPConfig() const;
CV_WRAP int singleFPConfig() const;
CV_WRAP int halfFPConfig() const;
bool endianLittle() const;
bool errorCorrectionSupport() const;
CV_WRAP bool endianLittle() const;
CV_WRAP bool errorCorrectionSupport() const;
enum
{
EXEC_KERNEL=(1 << 0),
EXEC_NATIVE_KERNEL=(1 << 1)
};
int executionCapabilities() const;
CV_WRAP int executionCapabilities() const;
size_t globalMemCacheSize() const;
CV_WRAP size_t globalMemCacheSize() const;
enum
{
@@ -141,38 +141,38 @@ public:
READ_ONLY_CACHE=1,
READ_WRITE_CACHE=2
};
int globalMemCacheType() const;
int globalMemCacheLineSize() const;
size_t globalMemSize() const;
CV_WRAP int globalMemCacheType() const;
CV_WRAP int globalMemCacheLineSize() const;
CV_WRAP size_t globalMemSize() const;
size_t localMemSize() const;
CV_WRAP size_t localMemSize() const;
enum
{
NO_LOCAL_MEM=0,
LOCAL_IS_LOCAL=1,
LOCAL_IS_GLOBAL=2
};
int localMemType() const;
bool hostUnifiedMemory() const;
CV_WRAP int localMemType() const;
CV_WRAP bool hostUnifiedMemory() const;
bool imageSupport() const;
CV_WRAP bool imageSupport() const;
bool imageFromBufferSupport() const;
CV_WRAP bool imageFromBufferSupport() const;
uint imagePitchAlignment() const;
uint imageBaseAddressAlignment() const;
/// deprecated, use isExtensionSupported() method (probably with "cl_khr_subgroups" value)
bool intelSubgroupsSupport() const;
CV_WRAP bool intelSubgroupsSupport() const;
size_t image2DMaxWidth() const;
size_t image2DMaxHeight() const;
CV_WRAP size_t image2DMaxWidth() const;
CV_WRAP size_t image2DMaxHeight() const;
size_t image3DMaxWidth() const;
size_t image3DMaxHeight() const;
size_t image3DMaxDepth() const;
CV_WRAP size_t image3DMaxWidth() const;
CV_WRAP size_t image3DMaxHeight() const;
CV_WRAP size_t image3DMaxDepth() const;
size_t imageMaxBufferSize() const;
size_t imageMaxArraySize() const;
CV_WRAP size_t imageMaxBufferSize() const;
CV_WRAP size_t imageMaxArraySize() const;
enum
{
@@ -181,53 +181,53 @@ public:
VENDOR_INTEL=2,
VENDOR_NVIDIA=3
};
int vendorID() const;
CV_WRAP int vendorID() const;
// FIXIT
// dev.isAMD() doesn't work for OpenCL CPU devices from AMD OpenCL platform.
// This method should use platform name instead of vendor name.
// After fix restore code in arithm.cpp: ocl_compare()
inline bool isAMD() const { return vendorID() == VENDOR_AMD; }
inline bool isIntel() const { return vendorID() == VENDOR_INTEL; }
inline bool isNVidia() const { return vendorID() == VENDOR_NVIDIA; }
CV_WRAP inline bool isAMD() const { return vendorID() == VENDOR_AMD; }
CV_WRAP inline bool isIntel() const { return vendorID() == VENDOR_INTEL; }
CV_WRAP inline bool isNVidia() const { return vendorID() == VENDOR_NVIDIA; }
int maxClockFrequency() const;
int maxComputeUnits() const;
int maxConstantArgs() const;
size_t maxConstantBufferSize() const;
CV_WRAP int maxClockFrequency() const;
CV_WRAP int maxComputeUnits() const;
CV_WRAP int maxConstantArgs() const;
CV_WRAP size_t maxConstantBufferSize() const;
size_t maxMemAllocSize() const;
size_t maxParameterSize() const;
CV_WRAP size_t maxMemAllocSize() const;
CV_WRAP size_t maxParameterSize() const;
int maxReadImageArgs() const;
int maxWriteImageArgs() const;
int maxSamplers() const;
CV_WRAP int maxReadImageArgs() const;
CV_WRAP int maxWriteImageArgs() const;
CV_WRAP int maxSamplers() const;
size_t maxWorkGroupSize() const;
int maxWorkItemDims() const;
CV_WRAP size_t maxWorkGroupSize() const;
CV_WRAP int maxWorkItemDims() const;
void maxWorkItemSizes(size_t*) const;
int memBaseAddrAlign() const;
CV_WRAP int memBaseAddrAlign() const;
int nativeVectorWidthChar() const;
int nativeVectorWidthShort() const;
int nativeVectorWidthInt() const;
int nativeVectorWidthLong() const;
int nativeVectorWidthFloat() const;
int nativeVectorWidthDouble() const;
int nativeVectorWidthHalf() const;
CV_WRAP int nativeVectorWidthChar() const;
CV_WRAP int nativeVectorWidthShort() const;
CV_WRAP int nativeVectorWidthInt() const;
CV_WRAP int nativeVectorWidthLong() const;
CV_WRAP int nativeVectorWidthFloat() const;
CV_WRAP int nativeVectorWidthDouble() const;
CV_WRAP int nativeVectorWidthHalf() const;
int preferredVectorWidthChar() const;
int preferredVectorWidthShort() const;
int preferredVectorWidthInt() const;
int preferredVectorWidthLong() const;
int preferredVectorWidthFloat() const;
int preferredVectorWidthDouble() const;
int preferredVectorWidthHalf() const;
CV_WRAP int preferredVectorWidthChar() const;
CV_WRAP int preferredVectorWidthShort() const;
CV_WRAP int preferredVectorWidthInt() const;
CV_WRAP int preferredVectorWidthLong() const;
CV_WRAP int preferredVectorWidthFloat() const;
CV_WRAP int preferredVectorWidthDouble() const;
CV_WRAP int preferredVectorWidthHalf() const;
size_t printfBufferSize() const;
size_t profilingTimerResolution() const;
CV_WRAP size_t printfBufferSize() const;
CV_WRAP size_t profilingTimerResolution() const;
static const Device& getDefault();
CV_WRAP static const Device& getDefault();
protected:
struct Impl;
modules/core/include/opencv2/core/opengl.hpp
+1 -5
View File
@@ -558,13 +558,11 @@ by the call to mapGLBuffer() function.
*/
CV_EXPORTS void unmapGLBuffer(UMat& u);
//! @}
}} // namespace cv::ogl
namespace cv { namespace cuda {
//! @addtogroup cuda
//! @{
/** @brief Sets a CUDA device and initializes it for the current thread with OpenGL interoperability.
This function should be explicitly called after OpenGL context creation and before any CUDA calls.
@@ -573,8 +571,6 @@ This function should be explicitly called after OpenGL context creation and befo
*/
CV_EXPORTS void setGlDevice(int device = 0);
//! @}
}}
//! @cond IGNORED
modules/core/include/opencv2/core/private.hpp
+3 -3
View File
@@ -704,12 +704,12 @@ CV_EXPORTS InstrNode* getCurrentNode();
if(::cv::instr::useInstrumentation()){\
::cv::instr::IntrumentationRegion __instr__(#FUN, __FILE__, __LINE__, NULL, false, TYPE, IMPL);\
try{\
auto status = ((FUN)(__VA_ARGS__));\
auto instrStatus = ((FUN)(__VA_ARGS__));\
if(ERROR_COND){\
::cv::instr::getCurrentNode()->m_payload.m_funError = true;\
CV_INSTRUMENT_MARK_META(IMPL, #FUN " - BadExit");\
}\
return status;\
return instrStatus;\
}catch(...){\
::cv::instr::getCurrentNode()->m_payload.m_funError = true;\
CV_INSTRUMENT_MARK_META(IMPL, #FUN " - BadExit");\
@@ -750,7 +750,7 @@ CV_EXPORTS InstrNode* getCurrentNode();
// Wrapper region instrumentation macro
#define CV_INSTRUMENT_REGION_IPP(); CV_INSTRUMENT_REGION_META(__FUNCTION__, false, ::cv::instr::TYPE_WRAPPER, ::cv::instr::IMPL_IPP)
// Function instrumentation macro
#define CV_INSTRUMENT_FUN_IPP(FUN, ...) CV_INSTRUMENT_FUN_RT_META(::cv::instr::TYPE_FUN, ::cv::instr::IMPL_IPP, status < 0, FUN, __VA_ARGS__)
#define CV_INSTRUMENT_FUN_IPP(FUN, ...) CV_INSTRUMENT_FUN_RT_META(::cv::instr::TYPE_FUN, ::cv::instr::IMPL_IPP, instrStatus < 0, FUN, __VA_ARGS__)
// Diagnostic markers
#define CV_INSTRUMENT_MARK_IPP(NAME) CV_INSTRUMENT_MARK_META(::cv::instr::IMPL_IPP, NAME)
modules/core/src/ocl.cpp
+2 -2
View File
@@ -3073,7 +3073,7 @@ bool Kernel::run(int dims, size_t _globalsize[], size_t _localsize[],
dims == 1 ? 64 : dims == 2 ? (i == 0 ? 256 : 8) : dims == 3 ? (8>>(int)(i>0)) : 1;
CV_Assert( val > 0 );
total *= _globalsize[i];
if (_globalsize[i] == 1)
if (_globalsize[i] == 1 && !_localsize)
val = 1;
globalsize[i] = divUp(_globalsize[i], (unsigned int)val) * val;
}
@@ -3086,7 +3086,7 @@ bool Kernel::run(int dims, size_t _globalsize[], size_t _localsize[],
bool Kernel::Impl::run(int dims, size_t globalsize[], size_t localsize[],
bool sync, int64* timeNS, const Queue& q)
{
CV_INSTRUMENT_REGION_OPENCL_RUN(name.c_str(););
CV_INSTRUMENT_REGION_OPENCL_RUN(name.c_str());
if (!handle || isInProgress)
return false;
modules/core/src/opengl.cpp
+8 -8
View File
@@ -1434,14 +1434,14 @@ void cv::ogl::render(const ogl::Texture2D& tex, Rect_<double> wndRect, Rect_<dou
gl::TexParameteri(gl::TEXTURE_2D, gl::TEXTURE_MIN_FILTER, gl::LINEAR);
CV_CheckGlError();
const float vertex[] =
const double vertex[] =
{
wndRect.x, wndRect.y, 0.0f,
wndRect.x, (wndRect.y + wndRect.height), 0.0f,
wndRect.x + wndRect.width, (wndRect.y + wndRect.height), 0.0f,
wndRect.x + wndRect.width, wndRect.y, 0.0f
wndRect.x, wndRect.y, 0.0,
wndRect.x, (wndRect.y + wndRect.height), 0.0,
wndRect.x + wndRect.width, (wndRect.y + wndRect.height), 0.0,
wndRect.x + wndRect.width, wndRect.y, 0.0
};
const float texCoords[] =
const double texCoords[] =
{
texRect.x, texRect.y,
texRect.x, texRect.y + texRect.height,
@@ -1454,7 +1454,7 @@ void cv::ogl::render(const ogl::Texture2D& tex, Rect_<double> wndRect, Rect_<dou
gl::EnableClientState(gl::TEXTURE_COORD_ARRAY);
CV_CheckGlError();
gl::TexCoordPointer(2, gl::FLOAT, 0, texCoords);
gl::TexCoordPointer(2, gl::DOUBLE, 0, texCoords);
CV_CheckGlError();
gl::DisableClientState(gl::NORMAL_ARRAY);
@@ -1464,7 +1464,7 @@ void cv::ogl::render(const ogl::Texture2D& tex, Rect_<double> wndRect, Rect_<dou
gl::EnableClientState(gl::VERTEX_ARRAY);
CV_CheckGlError();
gl::VertexPointer(3, gl::FLOAT, 0, vertex);
gl::VertexPointer(3, gl::DOUBLE, 0, vertex);
CV_CheckGlError();
gl::DrawArrays(gl::QUADS, 0, 4);
modules/core/src/system.cpp
+1 -1
View File
@@ -1768,7 +1768,7 @@ FLAGS getFlags()
NodeData::NodeData(const char* funName, const char* fileName, int lineNum, void* retAddress, bool alwaysExpand, cv::instr::TYPE instrType, cv::instr::IMPL implType)
{
m_funName = funName;
m_funName = funName ? cv::String(funName) : cv::String(); // std::string doesn't accept NULL
m_instrType = instrType;
m_implType = implType;
m_fileName = fileName;
modules/core/test/ocl/test_gemm.cpp
+15
View File
@@ -145,6 +145,21 @@ OCL_INSTANTIATE_TEST_CASE_P(Core, Gemm, ::testing::Combine(
testing::Values(CV_32FC1, CV_32FC2, CV_64FC1, CV_64FC2),
Bool(), Bool(), Bool(), Bool()));
// Test for non-Intel GPUs to check CL_INVALID_WORK_GROUP_SIZE when localsize > globalsize
OCL_TEST(Gemm, small)
{
UMat A(2, 3, CV_32F), B(4, 3, CV_32F), uC(2, 4, CV_32F);
Mat C(2, 4, CV_32F);
randu(A, -1, 1);
randu(B, -1, 1);
OCL_OFF(cv::gemm(A, B, 1, noArray(), 0, C, GEMM_2_T));
OCL_ON(cv::gemm(A, B, 1, noArray(), 0, uC, GEMM_2_T));
EXPECT_LE(cvtest::norm(C, uC, cv::NORM_INF), 1e-5);
}
} } // namespace opencv_test::ocl
#endif // HAVE_OPENCL
modules/core/test/test_intrin.cpp
+3
View File
@@ -12,6 +12,9 @@
#include "test_intrin256.simd.hpp"
#include "test_intrin256.simd_declarations.hpp"
#ifdef _MSC_VER
# pragma warning(disable:4702) // unreachable code
#endif
namespace opencv_test { namespace hal {