Fixed several issues found by static analysis in core module

pull/8769/head
Maksim Shabunin 8 years ago
parent c5e9d1adae
commit b04ed5956e
  1. 4
      modules/core/include/opencv2/core/mat.inl.hpp
  2. 3
      modules/core/include/opencv2/core/types.hpp
  3. 2
      modules/core/src/arithm.cpp
  4. 3
      modules/core/src/array.cpp
  5. 1
      modules/core/src/copy.cpp
  6. 6
      modules/core/src/matrix.cpp
  7. 2
      modules/core/src/ocl.cpp
  8. 2
      modules/core/src/out.cpp
  9. 4
      modules/core/src/parallel_pthreads.cpp
  10. 94
      modules/core/src/persistence.cpp
  11. 20
      modules/core/src/rand.cpp
  12. 91
      modules/core/src/stat.cpp
  13. 1
      modules/core/src/umatrix.cpp

@ -872,7 +872,7 @@ size_t Mat::step1(int i) const
inline
bool Mat::empty() const
{
return data == 0 || total() == 0;
return data == 0 || total() == 0 || dims == 0;
}
inline
@ -3730,7 +3730,7 @@ size_t UMat::step1(int i) const
inline
bool UMat::empty() const
{
return u == 0 || total() == 0;
return u == 0 || total() == 0 || dims == 0;
}
inline

@ -1040,7 +1040,8 @@ Complex<_Tp> operator / (const Complex<_Tp>& a, const Complex<_Tp>& b)
template<typename _Tp> static inline
Complex<_Tp>& operator /= (Complex<_Tp>& a, const Complex<_Tp>& b)
{
return (a = a / b);
a = a / b;
return a;
}
template<typename _Tp> static inline

@ -523,7 +523,7 @@ static bool ocl_arithm_op(InputArray _src1, InputArray _src2, OutputArray _dst,
float usrdata_f[3];
int i, n = oclop == OCL_OP_MUL_SCALE || oclop == OCL_OP_DIV_SCALE ||
oclop == OCL_OP_RDIV_SCALE || oclop == OCL_OP_RECIP_SCALE ? 1 : oclop == OCL_OP_ADDW ? 3 : 0;
if( n > 0 && wdepth == CV_32F )
if( usrdata && n > 0 && wdepth == CV_32F )
{
for( i = 0; i < n; i++ )
usrdata_f[i] = (float)usrdata_d[i];

@ -545,7 +545,7 @@ cvCreateSparseMat( int dims, const int* sizes, int type )
if( pix_size == 0 )
CV_Error( CV_StsUnsupportedFormat, "invalid array data type" );
if( dims <= 0 || dims > CV_MAX_DIM_HEAP )
if( dims <= 0 || dims > CV_MAX_DIM )
CV_Error( CV_StsOutOfRange, "bad number of dimensions" );
if( !sizes )
@ -1839,6 +1839,7 @@ cvPtr2D( const CvArr* arr, int y, int x, int* _type )
}
else if( CV_IS_SPARSE_MAT( arr ))
{
CV_Assert(((CvSparseMat*)arr)->dims == 2);
int idx[] = { y, x };
ptr = icvGetNodePtr( (CvSparseMat*)arr, idx, _type, 1, 0 );
}

@ -448,6 +448,7 @@ Mat& Mat::operator = (const Scalar& s)
for( size_t j = 0; j < elsize; j += blockSize )
{
size_t sz = MIN(blockSize, elsize - j);
CV_Assert(sz <= sizeof(scalar));
memcpy( dptr + j, scalar, sz );
}
}

@ -5786,10 +5786,10 @@ void minMaxLoc( const SparseMat& src, double* _minval, double* _maxval, int* _mi
else
CV_Error( CV_StsUnsupportedFormat, "Only 32f and 64f are supported" );
if( _minidx )
if( _minidx && minidx )
for( i = 0; i < d; i++ )
_minidx[i] = minidx[i];
if( _maxidx )
if( _maxidx && maxidx )
for( i = 0; i < d; i++ )
_maxidx[i] = maxidx[i];
}
@ -5898,7 +5898,7 @@ _IplImage::_IplImage(const cv::Mat& m)
CvSparseMat* cvCreateSparseMat(const cv::SparseMat& sm)
{
if( !sm.hdr )
if( !sm.hdr || sm.hdr->dims > (int)cv::SparseMat::MAX_DIM)
return 0;
CvSparseMat* m = cvCreateSparseMat(sm.hdr->dims, sm.hdr->size, sm.type());

@ -1204,7 +1204,7 @@ not_found:
if (!configuration)
return NULL; // suppress messages on stderr
std::cerr << "ERROR: Requested OpenCL device not found, check configuration: " << (configuration == NULL ? "" : configuration) << std::endl
std::cerr << "ERROR: Requested OpenCL device not found, check configuration: " << configuration << std::endl
<< " Platform: " << (platform.length() == 0 ? "any" : platform) << std::endl
<< " Device types: ";
for (size_t t = 0; t < deviceTypes.size(); t++)

@ -103,7 +103,7 @@ namespace cv
}
else
{
sprintf(floatFormat, "%%.%dg", std::min(precision, 20));
snprintf(floatFormat, 8, "%%.%dg", std::min(precision, 20));
}
switch(mtx.depth())

@ -141,10 +141,10 @@ private:
pthread_t m_posix_thread;
pthread_mutex_t m_thread_mutex;
pthread_cond_t m_cond_thread_task;
bool m_task_start;
volatile bool m_task_start;
ThreadManager* m_parent;
ForThreadState m_state;
volatile ForThreadState m_state;
size_t m_id;
};

@ -366,6 +366,11 @@ namespace base64
private:
void check_dt(const char* dt);
private:
// disable copy and assignment
Base64Writer(const Base64Writer &);
Base64Writer & operator=(const Base64Writer &);
private:
Base64ContextEmitter * emitter;
@ -1376,7 +1381,8 @@ static char* icvYMLParseBase64(CvFileStorage* fs, char* ptr, int indent, CvFileN
beg = end;
icvYMLGetMultilineStringContent( fs, beg, indent, beg, end );
}
if ( !base64::base64_valid(base64_buffer.data(), 0U, base64_buffer.size()) )
if ( base64_buffer.empty() ||
!base64::base64_valid(base64_buffer.data(), 0U, base64_buffer.size()) )
CV_PARSE_ERROR( "Invalid Base64 data." );
/* buffer for decoded data(exclude header) */
@ -1451,7 +1457,7 @@ static char*
icvYMLParseValue( CvFileStorage* fs, char* ptr, CvFileNode* node,
int parent_flags, int min_indent )
{
char buf[CV_FS_MAX_LEN + 1024];
char buf[CV_FS_MAX_LEN + 1024] = {0};
char* endptr = 0;
char c = ptr[0], d = ptr[1];
int is_parent_flow = CV_NODE_IS_FLOW(parent_flags);
@ -1668,7 +1674,7 @@ force_int:
{
int new_min_indent = min_indent + !is_parent_flow;
int struct_flags = CV_NODE_FLOW + (c == '{' ? CV_NODE_MAP : CV_NODE_SEQ);
int is_simple = 1;
bool is_simple = true;
icvFSCreateCollection( fs, CV_NODE_TYPE(struct_flags) +
(node->info ? CV_NODE_USER : 0), node );
@ -1706,16 +1712,18 @@ force_int:
break;
elem = (CvFileNode*)cvSeqPush( node->data.seq, 0 );
}
CV_Assert(elem);
ptr = icvYMLParseValue( fs, ptr, elem, struct_flags, new_min_indent );
if( CV_NODE_IS_MAP(struct_flags) )
elem->tag |= CV_NODE_NAMED;
is_simple &= !CV_NODE_IS_COLLECTION(elem->tag);
is_simple = is_simple && !CV_NODE_IS_COLLECTION(elem->tag);
}
node->data.seq->flags |= is_simple ? CV_NODE_SEQ_SIMPLE : 0;
}
else
{
int indent, struct_flags, is_simple;
int indent, struct_flags;
bool is_simple;
if( is_parent_flow || c != '-' )
{
@ -1760,7 +1768,7 @@ force_string:
(node->info ? CV_NODE_USER : 0), node );
indent = (int)(ptr - fs->buffer_start);
is_simple = 1;
is_simple = true;
for(;;)
{
@ -1778,12 +1786,12 @@ force_string:
elem = (CvFileNode*)cvSeqPush( node->data.seq, 0 );
}
CV_Assert(elem);
ptr = icvYMLSkipSpaces( fs, ptr, indent + 1, INT_MAX );
ptr = icvYMLParseValue( fs, ptr, elem, struct_flags, indent + 1 );
if( CV_NODE_IS_MAP(struct_flags) )
elem->tag |= CV_NODE_NAMED;
is_simple &= !CV_NODE_IS_COLLECTION(elem->tag);
is_simple = is_simple && !CV_NODE_IS_COLLECTION(elem->tag);
ptr = icvYMLSkipSpaces( fs, ptr, 0, INT_MAX );
if( ptr - fs->buffer_start != indent )
@ -2355,7 +2363,8 @@ static char* icvXMLParseBase64(CvFileStorage* fs, char* ptr, CvFileNode * node)
beg = end;
icvXMLGetMultilineStringContent( fs, beg, beg, end );
}
if ( !base64::base64_valid(base64_buffer.data(), 0U, base64_buffer.size()) )
if ( base64_buffer.empty() ||
!base64::base64_valid(base64_buffer.data(), 0U, base64_buffer.size()) )
CV_PARSE_ERROR( "Invalid Base64 data." );
/* alloc buffer for all decoded data(include header) */
@ -2402,7 +2411,7 @@ icvXMLParseValue( CvFileStorage* fs, char* ptr, CvFileNode* node,
int value_type CV_DEFAULT(CV_NODE_NONE))
{
CvFileNode *elem = node;
int have_space = 1, is_simple = 1;
bool have_space = true, is_simple = true;
int is_user_type = CV_NODE_IS_USER(value_type);
memset( node, 0, sizeof(*node) );
@ -2416,7 +2425,7 @@ icvXMLParseValue( CvFileStorage* fs, char* ptr, CvFileNode* node,
if( cv_isspace(c) || c == '\0' || (c == '<' && ptr[1] == '!' && ptr[2] == '-') )
{
ptr = icvXMLSkipSpaces( fs, ptr, 0 );
have_space = 1;
have_space = true;
c = *ptr;
}
@ -2482,7 +2491,7 @@ icvXMLParseValue( CvFileStorage* fs, char* ptr, CvFileNode* node,
elem = (CvFileNode*)cvSeqPush( node->data.seq, 0 );
else
elem = cvGetFileNode( fs, node, key, 1 );
CV_Assert(elem);
if (!is_binary_string)
ptr = icvXMLParseValue( fs, ptr, elem, elem_type);
else {
@ -2493,12 +2502,12 @@ icvXMLParseValue( CvFileStorage* fs, char* ptr, CvFileNode* node,
if( !is_noname )
elem->tag |= CV_NODE_NAMED;
is_simple &= !CV_NODE_IS_COLLECTION(elem->tag);
is_simple = is_simple && !CV_NODE_IS_COLLECTION(elem->tag);
elem->info = info;
ptr = icvXMLParseTag( fs, ptr, &key2, &list, &tag_type );
if( tag_type != CV_XML_CLOSING_TAG || key2 != key )
CV_PARSE_ERROR( "Mismatched closing tag" );
have_space = 1;
have_space = true;
}
else
{
@ -2547,7 +2556,7 @@ icvXMLParseValue( CvFileStorage* fs, char* ptr, CvFileNode* node,
else
{
// string
char buf[CV_FS_MAX_LEN+16];
char buf[CV_FS_MAX_LEN+16] = {0};
int i = 0, len, is_quoted = 0;
elem->tag = CV_NODE_STRING;
if( c == '\"' )
@ -2630,7 +2639,7 @@ icvXMLParseValue( CvFileStorage* fs, char* ptr, CvFileNode* node,
if( !CV_NODE_IS_COLLECTION(value_type) && value_type != CV_NODE_NONE )
break;
have_space = 0;
have_space = false;
}
}
@ -2708,6 +2717,7 @@ icvXMLParseTag( CvFileStorage* fs, char* ptr, CvStringHashNode** _tag,
while( cv_isalnum(c) || c == '_' || c == '-' );
attrname = cvGetHashedKey( fs, ptr, (int)(endptr - ptr), 1 );
CV_Assert(attrname);
ptr = endptr;
if( !tagname )
@ -2840,6 +2850,7 @@ icvXMLParse( CvFileStorage* fs )
CvFileNode* root_node;
ptr = icvXMLParseTag( fs, ptr, &key, &list, &tag_type );
if( tag_type != CV_XML_OPENING_TAG ||
!key ||
strcmp(key->str.ptr,"opencv_storage") != 0 )
CV_PARSE_ERROR( "<opencv_storage> tag is missing" );
@ -3665,6 +3676,9 @@ static char* icvJSONParseMap( CvFileStorage* fs, char* ptr, CvFileNode* node );
static char* icvJSONParseSeq( CvFileStorage* fs, char* ptr, CvFileNode* node )
{
if (!ptr)
CV_PARSE_ERROR( "ptr is NULL" );
if ( *ptr != '[' )
CV_PARSE_ERROR( "'[' - left-brace of seq is missing" );
else
@ -3703,6 +3717,9 @@ static char* icvJSONParseSeq( CvFileStorage* fs, char* ptr, CvFileNode* node )
CV_PARSE_ERROR( "Unexpected character" );
}
if (!ptr)
CV_PARSE_ERROR("ptr is NULL");
if ( *ptr != ']' )
CV_PARSE_ERROR( "']' - right-brace of seq is missing" );
else
@ -3713,6 +3730,9 @@ static char* icvJSONParseSeq( CvFileStorage* fs, char* ptr, CvFileNode* node )
static char* icvJSONParseMap( CvFileStorage* fs, char* ptr, CvFileNode* node )
{
if (!ptr)
CV_PARSE_ERROR("ptr is NULL");
if ( *ptr != '{' )
CV_PARSE_ERROR( "'{' - left-brace of map is missing" );
else
@ -3731,6 +3751,8 @@ static char* icvJSONParseMap( CvFileStorage* fs, char* ptr, CvFileNode* node )
{
CvFileNode* child = 0;
ptr = icvJSONParseKey( fs, ptr, node, &child );
if ( ptr == 0 || fs->dummy_eof )
break;
ptr = icvJSONSkipSpaces( fs, ptr );
if ( ptr == 0 || fs->dummy_eof )
break;
@ -3774,6 +3796,9 @@ static char* icvJSONParseMap( CvFileStorage* fs, char* ptr, CvFileNode* node )
CV_PARSE_ERROR( "Unexpected character" );
}
if (!ptr)
CV_PARSE_ERROR("ptr is NULL");
if ( *ptr != '}' )
CV_PARSE_ERROR( "'}' - right-brace of map is missing" );
else
@ -4180,6 +4205,7 @@ cvOpenFileStorage( const char* query, CvMemStorage* dststorage, int flags, const
CV_Error( CV_StsBadFlag, "CV_STORAGE_APPEND and CV_STORAGE_MEMORY are not currently compatible" );
fs = (CvFileStorage*)cvAlloc( sizeof(*fs) );
CV_Assert(fs);
memset( fs, 0, sizeof(*fs));
fs->memstorage = cvCreateMemStorage( default_block_size );
@ -4344,6 +4370,7 @@ cvOpenFileStorage( const char* query, CvMemStorage* dststorage, int flags, const
}
icvCloseFile( fs );
fs->file = fopen( fs->filename, "r+t" );
CV_Assert(fs->file);
fseek( fs->file, last_occurence, SEEK_SET );
// replace the last "</opencv_storage>" with " <!-- resumed -->", which has the same length
icvPuts( fs, " <!-- resumed -->" );
@ -4397,6 +4424,7 @@ cvOpenFileStorage( const char* query, CvMemStorage* dststorage, int flags, const
{
icvCloseFile( fs );
fs->file = fopen( fs->filename, "r+t" );
CV_Assert(fs->file);
fseek( fs->file, roffset, SEEK_END );
fputs( ",", fs->file );
}
@ -5316,7 +5344,7 @@ icvReadMatND( CvFileStorage* fs, CvFileNode* node )
const char* dt;
CvFileNode* data;
CvFileNode* sizes_node;
int sizes[CV_MAX_DIM], dims, elem_type;
int sizes[CV_MAX_DIM] = {0}, dims, elem_type;
int i, total_size;
sizes_node = cvGetFileNodeByName( fs, node, "sizes" );
@ -5341,7 +5369,10 @@ icvReadMatND( CvFileStorage* fs, CvFileNode* node )
for( total_size = CV_MAT_CN(elem_type), i = 0; i < dims; i++ )
{
CV_Assert(sizes[i]);
total_size *= sizes[i];
}
int nelems = icvFileNodeSeqLen( data );
@ -5481,7 +5512,7 @@ icvReadSparseMat( CvFileStorage* fs, CvFileNode* node )
dims = CV_NODE_IS_SEQ(sizes_node->tag) ? sizes_node->data.seq->total :
CV_NODE_IS_INT(sizes_node->tag) ? 1 : -1;
if( dims <= 0 || dims > CV_MAX_DIM_HEAP )
if( dims <= 0 || dims > CV_MAX_DIM)
CV_Error( CV_StsParseError, "Could not determine sparse matrix dimensionality" );
cvReadRawData( fs, sizes_node, sizes, "i" );
@ -5578,6 +5609,7 @@ icvWriteImage( CvFileStorage* fs, const char* name,
}
depth = IPL2CV_DEPTH(image->depth);
CV_Assert(depth < 9);
sprintf( dt_buf, "%d%c", image->nChannels, icvTypeSymbol[depth] );
dt = dt_buf + (dt_buf[2] == '\0' && dt_buf[0] == '1');
cvWriteString( fs, "dt", dt, 0 );
@ -5620,7 +5652,7 @@ icvReadImage( CvFileStorage* fs, CvFileNode* node )
elem_type = icvDecodeSimpleFormat( dt );
data_order = cvReadStringByName( fs, node, "layout", "interleaved" );
if( strcmp( data_order, "interleaved" ) != 0 )
if( !data_order || strcmp( data_order, "interleaved" ) != 0 )
CV_Error( CV_StsError, "Only interleaved images can be read" );
data = cvGetFileNodeByName( fs, node, "data" );
@ -5976,6 +6008,7 @@ icvReadSeq( CvFileStorage* fs, CvFileNode* node )
if( header_node )
{
CV_Assert(header_dt);
cvReadRawData( fs, header_node, (char*)seq + sizeof(CvSeq), header_dt );
}
else if( rect_node )
@ -6048,6 +6081,7 @@ icvReadSeqTree( CvFileStorage* fs, CvFileNode* node )
CvSeq* seq;
int level;
seq = (CvSeq*)cvRead( fs, elem );
CV_Assert(seq);
level = cvReadIntByName( fs, elem, "level", -1 );
if( level < 0 )
CV_Error( CV_StsParseError, "All the sequence tree nodes should contain \"level\" field" );
@ -6352,7 +6386,10 @@ icvReadGraph( CvFileStorage* fs, CvFileNode* node )
graph = cvCreateGraph( flags, header_size, vtx_size, edge_size, fs->dststorage );
if( header_node )
{
CV_Assert(header_dt);
cvReadRawData( fs, header_node, (char*)graph + sizeof(CvGraph), header_dt );
}
read_buf_size = MAX( src_vtx_size*3, 1 << 16 );
read_buf_size = MAX( src_edge_size*3, read_buf_size );
@ -7020,17 +7057,20 @@ void* FileNode::readObj() const
return cvRead( (CvFileStorage*)fs, (CvFileNode*)node );
}
static const FileNodeIterator::SeqReader emptyReader = {0, 0, 0, 0, 0, 0, 0, 0};
FileNodeIterator::FileNodeIterator()
{
fs = 0;
container = 0;
reader.ptr = 0;
reader = emptyReader;
remaining = 0;
}
FileNodeIterator::FileNodeIterator(const CvFileStorage* _fs,
const CvFileNode* _node, size_t _ofs)
{
reader = emptyReader;
if( _fs && _node && CV_NODE_TYPE(_node->tag) != CV_NODE_NONE )
{
int node_type = _node->tag & FileNode::TYPE_MASK;
@ -7053,7 +7093,6 @@ FileNodeIterator::FileNodeIterator(const CvFileStorage* _fs,
{
fs = 0;
container = 0;
reader.ptr = 0;
remaining = 0;
}
}
@ -7916,13 +7955,13 @@ class base64::RawDataToBinaryConvertor
{
public:
RawDataToBinaryConvertor(const void* src, int len, const char* dt)
RawDataToBinaryConvertor(const void* src, int len, const std::string & dt)
: beg(reinterpret_cast<const uchar *>(src))
, cur(0)
, end(0)
{
CV_Assert(src);
CV_Assert(dt);
CV_Assert(!dt.empty());
CV_Assert(len > 0);
/* calc step and to_binary_funcs */
@ -7931,7 +7970,7 @@ public:
end = beg;
cur = beg;
step = ::icvCalcStructSize(dt, 0);
step = ::icvCalcStructSize(dt.c_str(), 0);
end = beg + step * static_cast<size_t>(len);
}
@ -7963,7 +8002,7 @@ private:
};
private:
void make_to_binary_funcs(const char* dt)
void make_to_binary_funcs(const std::string &dt)
{
size_t cnt = 0;
size_t offset = 0;
@ -8225,10 +8264,7 @@ base64::Base64Writer::Base64Writer(::CvFileStorage * fs)
void base64::Base64Writer::write(const void* _data, size_t len, const char* dt)
{
check_dt(dt);
RawDataToBinaryConvertor convertor(
_data, static_cast<int>(len), data_type_string.c_str()
);
RawDataToBinaryConvertor convertor(_data, static_cast<int>(len), data_type_string);
emitter->write(convertor);
}

@ -500,7 +500,9 @@ void RNG::fill( InputOutputArray _mat, int disttype,
Mat mat = _mat.getMat(), _param1 = _param1arg.getMat(), _param2 = _param2arg.getMat();
int depth = mat.depth(), cn = mat.channels();
AutoBuffer<double> _parambuf;
int j, k, fast_int_mode = 0, smallFlag = 1;
int j, k;
bool fast_int_mode = false;
bool smallFlag = true;
RandFunc func = 0;
RandnScaleFunc scaleFunc = 0;
@ -553,7 +555,7 @@ void RNG::fill( InputOutputArray _mat, int disttype,
if( depth <= CV_32S )
{
ip = (Vec2i*)(parambuf + cn*2);
for( j = 0, fast_int_mode = 1; j < cn; j++ )
for( j = 0, fast_int_mode = true; j < cn; j++ )
{
double a = std::min(p1[j], p2[j]);
double b = std::max(p1[j], p2[j]);
@ -568,9 +570,9 @@ void RNG::fill( InputOutputArray _mat, int disttype,
int idiff = ip[j][0] = cvFloor(b) - ip[j][1] - 1;
double diff = b - a;
fast_int_mode &= diff <= 4294967296. && (idiff & (idiff+1)) == 0;
fast_int_mode = fast_int_mode && diff <= 4294967296. && (idiff & (idiff+1)) == 0;
if( fast_int_mode )
smallFlag &= idiff <= 255;
smallFlag = smallFlag && (idiff <= 255);
else
{
if( diff > INT_MAX )
@ -596,7 +598,7 @@ void RNG::fill( InputOutputArray _mat, int disttype,
}
}
func = randTab[fast_int_mode][depth];
func = randTab[fast_int_mode ? 1 : 0][depth];
}
else
{
@ -687,9 +689,9 @@ void RNG::fill( InputOutputArray _mat, int disttype,
buf.allocate(blockSize*cn*4);
param = (uchar*)(double*)buf;
if( ip )
if( depth <= CV_32S )
{
if( ds )
if( !fast_int_mode )
{
DivStruct* p = (DivStruct*)param;
for( j = 0; j < blockSize*cn; j += cn )
@ -704,7 +706,7 @@ void RNG::fill( InputOutputArray _mat, int disttype,
p[j + k] = ip[k];
}
}
else if( fp )
else if( depth == CV_32F )
{
Vec2f* p = (Vec2f*)param;
for( j = 0; j < blockSize*cn; j += cn )
@ -732,7 +734,7 @@ void RNG::fill( InputOutputArray _mat, int disttype,
int len = std::min(total - j, blockSize);
if( disttype == CV_RAND_UNI )
func( ptr, len*cn, &state, param, smallFlag != 0 );
func( ptr, len*cn, &state, param, smallFlag );
else
{
randn_0_1_32f(nbuf, len*cn, &state);

@ -76,6 +76,13 @@ struct Sum_SIMD
}
};
template <typename ST, typename DT>
inline void addChannels(DT * dst, ST * buf, int cn)
{
for (int i = 0; i < 4; ++i)
dst[i % cn] += buf[i];
}
#if CV_SSE2
template <>
@ -113,9 +120,7 @@ struct Sum_SIMD<schar, int>
int CV_DECL_ALIGNED(16) ar[4];
_mm_store_si128((__m128i*)ar, v_sum);
for (int i = 0; i < 4; i += cn)
for (int j = 0; j < cn; ++j)
dst[j] += ar[j + i];
addChannels(dst, ar, cn);
return x / cn;
}
@ -143,9 +148,7 @@ struct Sum_SIMD<int, double>
_mm_store_pd(ar, v_sum0);
_mm_store_pd(ar + 2, v_sum1);
for (int i = 0; i < 4; i += cn)
for (int j = 0; j < cn; ++j)
dst[j] += ar[j + i];
addChannels(dst, ar, cn);
return x / cn;
}
@ -174,9 +177,7 @@ struct Sum_SIMD<float, double>
_mm_store_pd(ar, v_sum0);
_mm_store_pd(ar + 2, v_sum1);
for (int i = 0; i < 4; i += cn)
for (int j = 0; j < cn; ++j)
dst[j] += ar[j + i];
addChannels(dst, ar, cn);
return x / cn;
}
@ -220,9 +221,7 @@ struct Sum_SIMD<uchar, int>
unsigned int CV_DECL_ALIGNED(16) ar[4];
vst1q_u32(ar, v_sum);
for (int i = 0; i < 4; i += cn)
for (int j = 0; j < cn; ++j)
dst[j] += ar[j + i];
addChannels(dst, ar, cn);
return x / cn;
}
@ -263,9 +262,7 @@ struct Sum_SIMD<schar, int>
int CV_DECL_ALIGNED(16) ar[4];
vst1q_s32(ar, v_sum);
for (int i = 0; i < 4; i += cn)
for (int j = 0; j < cn; ++j)
dst[j] += ar[j + i];
addChannels(dst, ar, cn);
return x / cn;
}
@ -296,9 +293,7 @@ struct Sum_SIMD<ushort, int>
unsigned int CV_DECL_ALIGNED(16) ar[4];
vst1q_u32(ar, v_sum);
for (int i = 0; i < 4; i += cn)
for (int j = 0; j < cn; ++j)
dst[j] += ar[j + i];
addChannels(dst, ar, cn);
return x / cn;
}
@ -329,9 +324,7 @@ struct Sum_SIMD<short, int>
int CV_DECL_ALIGNED(16) ar[4];
vst1q_s32(ar, v_sum);
for (int i = 0; i < 4; i += cn)
for (int j = 0; j < cn; ++j)
dst[j] += ar[j + i];
addChannels(dst, ar, cn);
return x / cn;
}
@ -748,6 +741,16 @@ struct SumSqr_SIMD
}
};
template <typename T>
inline void addSqrChannels(T * sum, T * sqsum, T * buf, int cn)
{
for (int i = 0; i < 4; ++i)
{
sum[i % cn] += buf[i];
sqsum[i % cn] += buf[4 + i];
}
}
#if CV_SSE2
template <>
@ -796,12 +799,7 @@ struct SumSqr_SIMD<uchar, int, int>
_mm_store_si128((__m128i*)ar, v_sum);
_mm_store_si128((__m128i*)(ar + 4), v_sqsum);
for (int i = 0; i < 4; i += cn)
for (int j = 0; j < cn; ++j)
{
sum[j] += ar[j + i];
sqsum[j] += ar[4 + j + i];
}
addSqrChannels(sum, sqsum, ar, cn);
return x / cn;
}
@ -853,12 +851,7 @@ struct SumSqr_SIMD<schar, int, int>
_mm_store_si128((__m128i*)ar, v_sum);
_mm_store_si128((__m128i*)(ar + 4), v_sqsum);
for (int i = 0; i < 4; i += cn)
for (int j = 0; j < cn; ++j)
{
sum[j] += ar[j + i];
sqsum[j] += ar[4 + j + i];
}
addSqrChannels(sum, sqsum, ar, cn);
return x / cn;
}
@ -1144,6 +1137,8 @@ static bool ipp_sum(Mat &src, Scalar &_res)
#if IPP_VERSION_X100 >= 700
int cn = src.channels();
if (cn > 4)
return false;
size_t total_size = src.total();
int rows = src.size[0], cols = rows ? (int)(total_size/rows) : 0;
if( src.dims == 2 || (src.isContinuous() && cols > 0 && (size_t)rows*cols == total_size) )
@ -1402,6 +1397,9 @@ static bool ipp_mean( Mat &src, Mat &mask, Scalar &ret )
#if IPP_VERSION_X100 >= 700
size_t total_size = src.total();
int cn = src.channels();
if (cn > 4)
return false;
int rows = src.size[0], cols = rows ? (int)(total_size/rows) : 0;
if( src.dims == 2 || (src.isContinuous() && mask.isContinuous() && cols > 0 && (size_t)rows*cols == total_size) )
{
@ -1471,7 +1469,7 @@ static bool ipp_mean( Mat &src, Mat &mask, Scalar &ret )
CV_INSTRUMENT_FUN_IPP(ippiMean, src.ptr(), (int)src.step[0], sz, res);
if( status >= 0 )
{
for( int i = 0; i < src.channels(); i++ )
for( int i = 0; i < cn; i++ )
ret[i] = res[i];
return true;
}
@ -1560,9 +1558,12 @@ static bool ocl_meanStdDev( InputArray _src, OutputArray _mean, OutputArray _sdv
bool haveMask = _mask.kind() != _InputArray::NONE;
int nz = haveMask ? -1 : (int)_src.total();
Scalar mean, stddev;
const int cn = _src.channels();
if (cn > 4)
return false;
{
int type = _src.type(), depth = CV_MAT_DEPTH(type), cn = CV_MAT_CN(type);
int type = _src.type(), depth = CV_MAT_DEPTH(type);
bool doubleSupport = ocl::Device::getDefault().doubleFPConfig() > 0,
isContinuous = _src.isContinuous(),
isMaskContinuous = _mask.isContinuous();
@ -1585,7 +1586,7 @@ static bool ocl_meanStdDev( InputArray _src, OutputArray _mean, OutputArray _sdv
wgs2_aligned <<= 1;
wgs2_aligned >>= 1;
if ( (!doubleSupport && depth == CV_64F) || cn > 4 )
if ( (!doubleSupport && depth == CV_64F) )
return false;
char cvt[2][40];
@ -1638,7 +1639,7 @@ static bool ocl_meanStdDev( InputArray _src, OutputArray _mean, OutputArray _sdv
}
double total = nz != 0 ? 1.0 / nz : 0;
int k, j, cn = _src.channels();
int k, j;
for (int i = 0; i < cn; ++i)
{
mean[i] *= total;
@ -2975,8 +2976,10 @@ static bool ocl_norm( InputArray _src, int normType, InputArray _mask, double &
if (d.isNVidia())
return false;
#endif
int type = _src.type(), depth = CV_MAT_DEPTH(type), cn = CV_MAT_CN(type);
const int cn = _src.channels();
if (cn > 4)
return false;
int type = _src.type(), depth = CV_MAT_DEPTH(type);
bool doubleSupport = d.doubleFPConfig() > 0,
haveMask = _mask.kind() != _InputArray::NONE;
@ -3001,11 +3004,8 @@ static bool ocl_norm( InputArray _src, int normType, InputArray _mask, double &
OCL_OP_SUM_SQR : (unstype ? OCL_OP_SUM : OCL_OP_SUM_ABS), _mask) )
return false;
if (!haveMask)
cn = 1;
double s = 0.0;
for (int i = 0; i < cn; ++i)
for (int i = 0; i < (haveMask ? cn : 1); ++i)
s += sc[i];
result = normType == NORM_L1 || normType == NORM_L2SQR ? s : std::sqrt(s);
@ -3320,7 +3320,10 @@ static bool ocl_norm( InputArray _src1, InputArray _src2, int normType, InputArr
#endif
Scalar sc1, sc2;
int type = _src1.type(), depth = CV_MAT_DEPTH(type), cn = CV_MAT_CN(type);
int cn = _src1.channels();
if (cn > 4)
return false;
int type = _src1.type(), depth = CV_MAT_DEPTH(type);
bool relative = (normType & NORM_RELATIVE) != 0;
normType &= ~NORM_RELATIVE;
bool normsum = normType == NORM_L1 || normType == NORM_L2 || normType == NORM_L2SQR;

@ -835,6 +835,7 @@ void UMat::copyTo(OutputArray _dst) const
if( _dst.isUMat() )
{
UMat dst = _dst.getUMat();
CV_Assert(dst.u);
if( u == dst.u && dst.offset == offset )
return;

Loading…
Cancel
Save