generic-library/opencv
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

fixed some bugs in GPU matrix reductions, removed <functional> into precomp.hpp

Browse Source
This commit is contained in:
Alexey Spizhevoy 2011-01-21 07:43:11 +00:00
parent 0da71a01ff
commit 01dafce1a1
5 changed files with 20 additions and 21 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

18
modules/gpu/src/cuda/matrix_reductions.cu
View File

@ -328,13 +328,13 @@ namespace cv { namespace gpu { namespace mathfunc
__shared__ best_type smaxval[nthreads];
uint tid = threadIdx.y * blockDim.x + threadIdx.x;
uint idx = min(tid, gridDim.x * gridDim.y - 1);
uint idx = min(tid, size - 1);
sminval[tid] = minval[idx];
smaxval[tid] = maxval[idx];
__syncthreads();
findMinMaxInSmem<nthreads, best_type>(sminval, smaxval, tid);
findMinMaxInSmem<nthreads, best_type>(sminval, smaxval, tid);
if (tid == 0)
{
@ -428,7 +428,7 @@ namespace cv { namespace gpu { namespace mathfunc
// Returns required buffer sizes
void getBufSizeRequired(int cols, int rows, int elem_size, int& b1cols,
int& b1rows, int& b2cols, int& b2rows)
int& b1rows, int& b2cols, int& b2rows)
{
dim3 threads, grid;
estimateThreadCfg(cols, rows, threads, grid);
@ -623,7 +623,7 @@ namespace cv { namespace gpu { namespace mathfunc
template <typename T>
void minMaxLocCaller(const DevMem2D src, double* minval, double* maxval,
int minloc[2], int maxloc[2], PtrStep valbuf, PtrStep locbuf)
int minloc[2], int maxloc[2], PtrStep valbuf, PtrStep locbuf)
{
dim3 threads, grid;
estimateThreadCfg(src.cols, src.rows, threads, grid);
@ -671,7 +671,7 @@ namespace cv { namespace gpu { namespace mathfunc
__shared__ uint smaxloc[nthreads];
uint tid = threadIdx.y * blockDim.x + threadIdx.x;
uint idx = min(tid, gridDim.x * gridDim.y - 1);
uint idx = min(tid, size - 1);
sminval[tid] = minval[idx];
smaxval[tid] = maxval[idx];
@ -679,7 +679,7 @@ namespace cv { namespace gpu { namespace mathfunc
smaxloc[tid] = maxloc[idx];
__syncthreads();
findMinMaxLocInSmem<nthreads, best_type>(sminval, smaxval, sminloc, smaxloc, tid);
findMinMaxLocInSmem<nthreads, best_type>(sminval, smaxval, sminloc, smaxloc, tid);
if (tid == 0)
{
@ -1150,7 +1150,7 @@ namespace cv { namespace gpu { namespace mathfunc
const int tid = threadIdx.y * blockDim.x + threadIdx.x;
DstType res = tid < gridDim.x * gridDim.y ? result[tid] : VecTraits<DstType>::all(0);
DstType res = tid < size ? result[tid] : VecTraits<DstType>::all(0);
smem[tid] = res.x;
smem[tid + nthreads] = res.y;
__syncthreads();
@ -1262,7 +1262,7 @@ namespace cv { namespace gpu { namespace mathfunc
const int tid = threadIdx.y * blockDim.x + threadIdx.x;
DstType res = tid < gridDim.x * gridDim.y ? result[tid] : VecTraits<DstType>::all(0);
DstType res = tid < size ? result[tid] : VecTraits<DstType>::all(0);
smem[tid] = res.x;
smem[tid + nthreads] = res.y;
smem[tid + 2 * nthreads] = res.z;
@ -1384,7 +1384,7 @@ namespace cv { namespace gpu { namespace mathfunc
const int tid = threadIdx.y * blockDim.x + threadIdx.x;
DstType res = tid < gridDim.x * gridDim.y ? result[tid] : VecTraits<DstType>::all(0);
DstType res = tid < size ? result[tid] : VecTraits<DstType>::all(0);
smem[tid] = res.x;
smem[tid + nthreads] = res.y;
smem[tid + 2 * nthreads] = res.z;

1
modules/gpu/src/initialization.cpp
View File

@ -41,7 +41,6 @@
//M*/
#include "precomp.hpp"
#include <functional>
using namespace cv;
using namespace cv::gpu;

19
modules/gpu/src/matrix_reductions.cpp
View File

@ -276,11 +276,11 @@ void cv::gpu::minMax(const GpuMat& src, double* minVal, double* maxVal, const Gp
minMaxMaskCaller<double> };
CV_Assert(src.channels() == 1);
CV_Assert(mask.empty() || (mask.type() == CV_8U && src.size() == mask.size()));
bool double_ok = hasGreaterOrEqualVersion(1, 3) &&
hasNativeDoubleSupport(getDevice());
CV_Assert(src.type() != CV_64F || double_ok);
CV_Assert(src.type() != CV_64F || (hasGreaterOrEqualVersion(1, 3) &&
hasNativeDoubleSupport(getDevice())));
double minVal_; if (!minVal) minVal = &minVal_;
double maxVal_; if (!maxVal) maxVal = &maxVal_;
@ -375,11 +375,11 @@ void cv::gpu::minMaxLoc(const GpuMat& src, double* minVal, double* maxVal, Point
minMaxLocMaskCaller<double> };
CV_Assert(src.channels() == 1);
CV_Assert(mask.empty() || (mask.type() == CV_8U && src.size() == mask.size()));
bool double_ok = hasGreaterOrEqualVersion(1, 3) &&
hasNativeDoubleSupport(getDevice());
CV_Assert(src.type() != CV_64F || double_ok);
CV_Assert(src.type() != CV_64F || (hasGreaterOrEqualVersion(1, 3) &&
hasNativeDoubleSupport(getDevice())));
double minVal_; if (!minVal) minVal = &minVal_;
double maxVal_; if (!maxVal) maxVal = &maxVal_;
@ -388,7 +388,7 @@ void cv::gpu::minMaxLoc(const GpuMat& src, double* minVal, double* maxVal, Point
Size valbuf_size, locbuf_size;
getBufSizeRequired(src.cols, src.rows, src.elemSize(), valbuf_size.width,
valbuf_size.height, locbuf_size.width, locbuf_size.height);
valbuf_size.height, locbuf_size.width, locbuf_size.height);
ensureSizeIsEnough(valbuf_size, CV_8U, valBuf);
ensureSizeIsEnough(locbuf_size, CV_8U, locBuf);
@ -459,9 +459,8 @@ int cv::gpu::countNonZero(const GpuMat& src, GpuMat& buf)
CV_Assert(src.channels() == 1);
bool double_ok = hasGreaterOrEqualVersion(1, 3) &&
hasNativeDoubleSupport(getDevice());
CV_Assert(src.type() != CV_64F || double_ok);
CV_Assert(src.type() != CV_64F || (hasGreaterOrEqualVersion(1, 3) &&
hasNativeDoubleSupport(getDevice())));
Size buf_size;
getBufSizeRequired(src.cols, src.rows, buf_size.width, buf_size.height);

1
modules/gpu/src/precomp.hpp
View File

@ -57,6 +57,7 @@
#include <sstream>
#include <exception>
#include <iterator>
#include <functional>
#include "opencv2/gpu/gpu.hpp"
#include "opencv2/imgproc/imgproc.hpp"

2
tests/gpu/src/arithm.cpp
View File

@ -49,7 +49,7 @@ using namespace std;
using namespace gpu;
#define CHECK(pred, err) if (!(pred)) { \
ts->printf(CvTS::LOG, "Fail: \"%s\" at line: %d\n", #pred, __LINE__); \
ts->printf(CvTS::CONSOLE, "Fail: \"%s\" at line: %d\n", #pred, __LINE__); \
ts->set_failed_test_info(err); \
return; }