[+] Pixel types via templates

[+] Color conversions stub via pixel types
[+] Pyramid calculation (required for mipmaps in CUDA 4.1)
[~] Changed C strings to C++ throughout NCV
[~] Fixed a couple of bugs in NCV

modules/gpu/src/cascadeclassifier.cpp

@@ -129,7 +129,7 @@ struct cv::gpu::CascadeClassifier_GPU::CascadeClassifierImpl
 private:
 
 
-    static void NCVDebugOutputHandler(const char* msg) { CV_Error(CV_GpuApiCallError, msg); }
+    static void NCVDebugOutputHandler(const std::string &msg) { CV_Error(CV_GpuApiCallError, msg.c_str()); }
 
 
     NCVStatus load(const string& classifierFile)

modules/gpu/src/nvidia/core/NCV.cu

@@ -40,10 +40,9 @@
 //M*/
 
 
-#include <ios>
-#include <stdarg.h>
+#include <iostream>
+#include <string>
 #include <vector>
-#include <cstdio>
 #include "NCV.hpp"
 
 using namespace std;
@@ -56,24 +55,18 @@ using namespace std;
 //==============================================================================
 
 
-static void stdioDebugOutput(const char *msg)
+static void stdDebugOutput(const string &msg)
 {
-    printf("%s", msg);
+    cout << msg;
 }
 
 
-static NCVDebugOutputHandler *debugOutputHandler = stdioDebugOutput;
+static NCVDebugOutputHandler *debugOutputHandler = stdDebugOutput;
 
 
-void ncvDebugOutput(const char *msg, ...)
+void ncvDebugOutput(const string &msg)
 {
-    const int K_DEBUG_STRING_MAXLEN = 1024;
-    char buffer[K_DEBUG_STRING_MAXLEN];
-    va_list args;
-    va_start(args, msg);
-    vsnprintf(buffer, K_DEBUG_STRING_MAXLEN, msg, args);
-    va_end (args);
-    debugOutputHandler(buffer);
+    debugOutputHandler(msg);
 }
 
 
@@ -288,7 +281,7 @@ NCVMemStackAllocator::NCVMemStackAllocator(NCVMemoryType memT, size_t capacity,
 
     allocBegin = NULL;
 
-    if (reusePtr == NULL)
+    if (reusePtr == NULL && capacity != 0)
     {
         bReusesMemory = false;
         switch (memT)
@@ -329,7 +322,7 @@ NCVMemStackAllocator::~NCVMemStackAllocator()
     {
         ncvAssertPrintCheck(currentSize == 0, "NCVMemStackAllocator dtor:: not all objects were deallocated properly, forcing destruction");
 
-        if (!bReusesMemory)
+        if (!bReusesMemory && (allocBegin != (Ncv8u *)(0x1)))
         {
             switch (_memType)
             {
@@ -355,7 +348,7 @@ NCVStatus NCVMemStackAllocator::alloc(NCVMemSegment &seg, size_t size)
     seg.clear();
     ncvAssertReturn(isInitialized(), NCV_ALLOCATOR_BAD_ALLOC);
 
-    size = alignUp(static_cast<Ncv32u>(size), this->_alignment);
+    size = alignUp(size, this->_alignment);
     this->currentSize += size;
     this->_maxSize = std::max(this->_maxSize, this->currentSize);
 
@@ -464,7 +457,7 @@ NCVStatus NCVMemNativeAllocator::alloc(NCVMemSegment &seg, size_t size)
         break;
     }
 
-    this->currentSize += alignUp(static_cast<Ncv32u>(size), this->_alignment);
+    this->currentSize += alignUp(size, this->_alignment);
     this->_maxSize = std::max(this->_maxSize, this->currentSize);
 
     seg.begin.memtype = this->_memType;
@@ -480,8 +473,8 @@ NCVStatus NCVMemNativeAllocator::dealloc(NCVMemSegment &seg)
     ncvAssertReturn(seg.begin.memtype == this->_memType, NCV_ALLOCATOR_BAD_DEALLOC);
     ncvAssertReturn(seg.begin.ptr != NULL, NCV_ALLOCATOR_BAD_DEALLOC);
 
-    ncvAssertReturn(currentSize >= alignUp(static_cast<Ncv32u>(seg.size), this->_alignment), NCV_ALLOCATOR_BAD_DEALLOC);
-    currentSize -= alignUp(static_cast<Ncv32u>(seg.size), this->_alignment);
+    ncvAssertReturn(currentSize >= alignUp(seg.size, this->_alignment), NCV_ALLOCATOR_BAD_DEALLOC);
+    currentSize -= alignUp(seg.size, this->_alignment);
 
     switch (this->_memType)
     {

modules/gpu/src/nvidia/core/NCV.hpp

@@ -42,7 +42,7 @@
 #ifndef _ncv_hpp_
 #define _ncv_hpp_
 
-#if (defined WIN32 || defined _WIN32 || defined WINCE) && defined CVAPI_EXPORTS //&& !defined(__CUDACC__) 
+#if (defined WIN32 || defined _WIN32 || defined WINCE) && defined CVAPI_EXPORTS
     #define NCV_EXPORTS __declspec(dllexport)
 #else
     #define NCV_EXPORTS
@@ -53,6 +53,8 @@
 #endif
 
 #include <cuda_runtime.h>
+#include <sstream>
+#include <iostream>
 
 
 //==============================================================================
@@ -78,7 +80,7 @@ namespace NcvCTprep
 }
 
 
-#define NCV_CT_PREP_PASTE_AUX(a,b)      a##b                           ///< Concatenation indirection macro
+#define NCV_CT_PREP_PASTE_AUX(a,b)      a##b                         ///< Concatenation indirection macro
 #define NCV_CT_PREP_PASTE(a,b)          NCV_CT_PREP_PASTE_AUX(a, b)  ///< Concatenation macro
 
 
@@ -181,6 +183,25 @@ struct NcvSize32u
     Ncv32u height; ///< Rectangle height.
     __host__ __device__ NcvSize32u() : width(0), height(0) {};
     __host__ __device__ NcvSize32u(Ncv32u width, Ncv32u height) : width(width), height(height) {}
+    __host__ __device__ bool operator == (const NcvSize32u &another) const {return this->width == another.width && this->height == another.height;}
+};
+
+
+struct NcvPoint2D32s
+{
+    Ncv32s x; ///< Point X.
+    Ncv32s y; ///< Point Y.
+    __host__ __device__ NcvPoint2D32s() : x(0), y(0) {};
+    __host__ __device__ NcvPoint2D32s(Ncv32s x, Ncv32s y) : x(x), y(y) {}
+};
+
+
+struct NcvPoint2D32u
+{
+    Ncv32u x; ///< Point X.
+    Ncv32u y; ///< Point Y.
+    __host__ __device__ NcvPoint2D32u() : x(0), y(0) {};
+    __host__ __device__ NcvPoint2D32u(Ncv32u x, Ncv32u y) : x(x), y(y) {}
 };
 
 
@@ -199,6 +220,7 @@ NCV_CT_ASSERT(sizeof(NcvRect8u) == sizeof(Ncv32u));
 NCV_CT_ASSERT(sizeof(NcvRect32s) == 4 * sizeof(Ncv32s));
 NCV_CT_ASSERT(sizeof(NcvRect32u) == 4 * sizeof(Ncv32u));
 NCV_CT_ASSERT(sizeof(NcvSize32u) == 2 * sizeof(Ncv32u));
+NCV_CT_ASSERT(sizeof(NcvPoint2D32u) == 2 * sizeof(Ncv32u));
 
 
 //==============================================================================
@@ -219,49 +241,44 @@ const Ncv32u K_LOG2_WARP_SIZE = 5;
 //==============================================================================
 
 
-#define NCV_CT_PREP_STRINGIZE_AUX(x)    #x
-#define NCV_CT_PREP_STRINGIZE(x)        NCV_CT_PREP_STRINGIZE_AUX(x)
-
+NCV_EXPORTS void ncvDebugOutput(const std::string &msg);
 
-NCV_EXPORTS void ncvDebugOutput(const char *msg, ...);
 
-
-typedef void NCVDebugOutputHandler(const char* msg);
+typedef void NCVDebugOutputHandler(const std::string &msg);
 
 
 NCV_EXPORTS void ncvSetDebugOutputHandler(NCVDebugOutputHandler* func);
 
 
 #define ncvAssertPrintCheck(pred, msg) \
-    ((pred) ? true : (ncvDebugOutput("\n%s\n", \
-    "NCV Assertion Failed: " msg ", file=" __FILE__ ", line=" NCV_CT_PREP_STRINGIZE(__LINE__) \
-    ), false))
-
-
-#define ncvAssertPrintReturn(pred, msg, err) \
-    if (ncvAssertPrintCheck(pred, msg)) ; else return err
-
-
-#define ncvAssertReturn(pred, err) \
     do \
     { \
         if (!(pred)) \
         { \
-            ncvDebugOutput("\n%s%d%s\n", "NCV Assertion Failed: retcode=", (int)err, ", file=" __FILE__ ", line=" NCV_CT_PREP_STRINGIZE(__LINE__)); \
-            return err; \
+            std::ostringstream oss; \
+            oss << "NCV Assertion Failed: " << msg << ", file=" << __FILE__ << ", line=" << __LINE__ << std::endl; \
+            ncvDebugOutput(oss.str()); \
         } \
     } while (0)
 
 
+#define ncvAssertPrintReturn(pred, msg, err) \
+    do \
+    { \
+        ncvAssertPrintCheck(pred, msg); \
+        if (!(pred)) return err; \
+    } while (0)
+
+
+#define ncvAssertReturn(pred, err) \
+    ncvAssertPrintReturn(pred, "retcode=" << (int)err, err)
+
+
 #define ncvAssertReturnNcvStat(ncvOp) \
     do \
     { \
         NCVStatus _ncvStat = ncvOp; \
-        if (NCV_SUCCESS != _ncvStat) \
-        { \
-            ncvDebugOutput("\n%s%d%s\n", "NCV Assertion Failed: NcvStat=", (int)_ncvStat, ", file=" __FILE__ ", line=" NCV_CT_PREP_STRINGIZE(__LINE__)); \
-            return _ncvStat; \
-        } \
+        ncvAssertPrintReturn(NCV_SUCCESS==_ncvStat, "NcvStat=" << (int)_ncvStat, _ncvStat); \
     } while (0)
 
 
@@ -270,18 +287,14 @@ NCV_EXPORTS void ncvSetDebugOutputHandler(NCVDebugOutputHandler* func);
     { \
         cudaError_t resCall = cudacall; \
         cudaError_t resGLE = cudaGetLastError(); \
-        if (cudaSuccess != resCall || cudaSuccess != resGLE) \
-        { \
-            ncvDebugOutput("\n%s%d%s\n", "NCV CUDA Assertion Failed: cudaError_t=", (int)(resCall | resGLE), ", file=" __FILE__ ", line=" NCV_CT_PREP_STRINGIZE(__LINE__)); \
-            return errCode; \
-        } \
+        ncvAssertPrintReturn(cudaSuccess==resCall && cudaSuccess==resGLE, "cudaError_t=" << (int)(resCall | resGLE), errCode); \
     } while (0)
 
 
 /**
 * Return-codes for status notification, errors and warnings
 */
-enum NCVStatus
+enum
 {
     //NCV statuses
     NCV_SUCCESS,
@@ -338,9 +351,14 @@ enum NCVStatus
     NPPST_MEM_INSUFFICIENT_BUFFER,            ///< Insufficient user-allocated buffer
     NPPST_MEM_RESIDENCE_ERROR,                ///< Memory residence error detected (check if pointers should be device or pinned)
     NPPST_MEM_INTERNAL_ERROR,                 ///< Internal memory management error
+
+    NCV_LAST_STATUS                           ///< Marker to continue error numeration in other files
 };
 
 
+typedef Ncv32u NCVStatus;
+
+
 #define NCV_SET_SKIP_COND(x) \
     bool __ncv_skip_cond = x
 
@@ -774,9 +792,20 @@ public:
         return ncvStat;
     }
 
+    T &at(Ncv32u x, Ncv32u y) const
+    {
+        if (x >= this->_width || y >= this->_height)
+        {
+            printf("Error addressing matrix at [%d, %d]\n", x, y);
+            return *this->_ptr;
+        }
+        return ((T *)((Ncv8u *)this->_ptr + y * this->_pitch))[x];
+    }
+
     T *ptr() const {return this->_ptr;}
     Ncv32u width() const {return this->_width;}
     Ncv32u height() const {return this->_height;}
+    NcvSize32u size() const {return NcvSize32u(this->_width, this->_height);}
     Ncv32u pitch() const {return this->_pitch;}
     NCVMemoryType memType() const {return this->_memtype;}
 
@@ -923,7 +952,7 @@ public:
         this->_width = roi.width;
         this->_height = roi.height;
         this->_pitch = mat.pitch();
-        this->_ptr = mat.ptr() + roi.y * mat.stride() + roi.x;
+        this->_ptr = &mat.at(roi.x, roi.y);
         this->_memtype = mat.memType();
 
         this->bReused = true;
@@ -962,4 +991,24 @@ NCV_EXPORTS NCVStatus ncvDrawRects_8u_device(Ncv8u *d_dst, Ncv32u dstStride, Ncv
 NCV_EXPORTS NCVStatus ncvDrawRects_32u_device(Ncv32u *d_dst, Ncv32u dstStride, Ncv32u dstWidth, Ncv32u dstHeight,
                                               NcvRect32u *d_rects, Ncv32u numRects, Ncv32u color, cudaStream_t cuStream);
 
+
+#define CLAMP(x,a,b)        ( (x) > (b) ? (b) : ( (x) < (a) ? (a) : (x) ) )
+#define CLAMP_TOP(x, a)     (((x) > (a)) ? (a) : (x))
+#define CLAMP_BOTTOM(x, a)  (((x) < (a)) ? (a) : (x))
+#define CLAMP_0_255(x)      CLAMP(x,0,255)
+
+
+#define SUB_BEGIN(type, name)    struct { __inline type name
+#define SUB_END(name)            } name;
+#define SUB_CALL(name)           name.name
+
+#define SQR(x)              ((x)*(x))
+
+
+#define ncvSafeMatAlloc(name, type, alloc, width, height, err) \
+    NCVMatrixAlloc<type> name(alloc, width, height); \
+    ncvAssertReturn(name.isMemAllocated(), err);
+
+
+
 #endif // _ncv_hpp_

modules/gpu/src/nvidia/core/NCVColorConversion.hpp

+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+// IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING. 
+// 
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2009-2010, NVIDIA Corporation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef _ncv_color_conversion_hpp_
+#define _ncv_color_conversion_hpp_
+
+#include "NCVPixelOperations.hpp"
+
+enum NCVColorSpace
+{
+    NCVColorSpaceGray,
+    NCVColorSpaceRGBA,
+};
+
+template<NCVColorSpace CSin, NCVColorSpace CSout, typename Tin, typename Tout> struct __pixColorConv {
+static void _pixColorConv(const Tin &pixIn, Tout &pixOut);
+};
+
+template<typename Tin, typename Tout> struct __pixColorConv<NCVColorSpaceRGBA, NCVColorSpaceGray, Tin, Tout> {
+static void _pixColorConv(const Tin &pixIn, Tout &pixOut)
+{
+    Ncv32f luma = 0.299f * pixIn.x + 0.587f * pixIn.y + 0.114f * pixIn.z;
+    _TDemoteClampNN(luma, pixOut.x);
+}};
+
+template<typename Tin, typename Tout> struct __pixColorConv<NCVColorSpaceGray, NCVColorSpaceRGBA, Tin, Tout> {
+static void _pixColorConv(const Tin &pixIn, Tout &pixOut)
+{
+    _TDemoteClampNN(pixIn.x, pixOut.x);
+    _TDemoteClampNN(pixIn.x, pixOut.y);
+    _TDemoteClampNN(pixIn.x, pixOut.z);
+    pixOut.w = 0;
+}};
+
+template<NCVColorSpace CSin, NCVColorSpace CSout, typename Tin, typename Tout>
+static
+NCVStatus _ncvColorConv_host(const NCVMatrix<Tin> &h_imgIn,
+                             const NCVMatrix<Tout> &h_imgOut)
+{
+    ncvAssertReturn(h_imgIn.size() == h_imgOut.size(), NCV_DIMENSIONS_INVALID);
+    ncvAssertReturn(h_imgIn.memType() == h_imgOut.memType() &&
+                    (h_imgIn.memType() == NCVMemoryTypeHostPinned || h_imgIn.memType() == NCVMemoryTypeNone), NCV_MEM_RESIDENCE_ERROR);
+    NCV_SET_SKIP_COND(h_imgIn.memType() == NCVMemoryTypeNone);
+    NCV_SKIP_COND_BEGIN
+
+    for (Ncv32u i=0; i<h_imgIn.height(); i++)
+    {
+        for (Ncv32u j=0; j<h_imgIn.width(); j++)
+        {
+            __pixColorConv<CSin, CSout, Tin, Tout>::_pixColorConv(h_imgIn.at(j,i), h_imgOut.at(j,i));
+        }
+    }
+
+    NCV_SKIP_COND_END
+    return NCV_SUCCESS;
+}
+
+#endif //_ncv_color_conversion_hpp_

modules/gpu/src/nvidia/core/NCVPyramid.hpp

+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+// IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING. 
+// 
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2009-2010, NVIDIA Corporation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+
+#ifndef _ncvpyramid_hpp_
+#define _ncvpyramid_hpp_
+
+#include <memory>
+#include <vector>
+#include "NCV.hpp"
+
+
+template <class T>
+class NCV_EXPORTS NCVMatrixStack
+{
+public:
+    NCVMatrixStack() {this->_arr.clear();}
+    ~NCVMatrixStack()
+    {
+        const Ncv32u nElem = this->_arr.size();
+        for (Ncv32u i=0; i<nElem; i++)
+        {
+            pop_back();
+        }
+    }
+    void push_back(NCVMatrix<T> *elem) {this->_arr.push_back(std::tr1::shared_ptr< NCVMatrix<T> >(elem));}
+    void pop_back() {this->_arr.pop_back();}
+    NCVMatrix<T> * operator [] (int i) const {return this->_arr[i].get();}
+private:
+    std::vector< std::tr1::shared_ptr< NCVMatrix<T> > > _arr;
+};
+
+
+template <class T>
+class NCV_EXPORTS NCVImagePyramid
+{
+public:
+
+    NCVImagePyramid(const NCVMatrix<T> &img,
+                    Ncv8u nLayers,
+                    INCVMemAllocator &alloc,
+                    cudaStream_t cuStream);
+    ~NCVImagePyramid();
+    NcvBool isInitialized() const;
+    NCVStatus getLayer(NCVMatrix<T> &outImg,
+                       NcvSize32u outRoi,
+                       NcvBool bTrilinear,
+                       cudaStream_t cuStream) const;
+
+private:
+
+    NcvBool _isInitialized;
+    const NCVMatrix<T> *layer0;
+    NCVMatrixStack<T> pyramid;
+    Ncv32u nLayers;
+};
+
+
+#endif //_ncvpyramid_hpp_

modules/gpu/src/optical_flow.cpp

@@ -68,10 +68,7 @@ namespace
 
 namespace
 {
-    void outputHandler(const char* msg) 
-    { 
-        CV_Error(CV_GpuApiCallError, msg); 
-    }
+    static void outputHandler(const std::string &msg) { CV_Error(CV_GpuApiCallError, msg.c_str()); }
 }
 
 void cv::gpu::BroxOpticalFlow::operator ()(const GpuMat& frame0, const GpuMat& frame1, GpuMat& u, GpuMat& v, Stream& s)