Chiebot-Mirror
/
opencv
mirror of https://github.com/opencv/opencv.git
8
0
Fork 0
Code Issues Projects Releases Wiki Activity

fixed warnings

Browse Source
pull/13383/head
Vladislav Vinogradov 13 years ago
parent 7106513b8a
commit 87f3451ec6
3 changed files with 81 additions and 44 deletions
Whitespace
Show all changes Ignore whitespace when comparing lines Ignore changes in amount of whitespace Ignore changes in whitespace at EOL
Split View
Diff Options
Show Stats Download Patch File Download Diff File
  1. 45
      modules/gpu/src/cuda/bf_knnmatch.cu
  2. 45
      modules/gpu/src/cuda/bf_match.cu
  3. 35
      modules/gpu/src/cuda/bf_radius_match.cu

45
modules/gpu/src/cuda/bf_knnmatch.cu
Unescape View File

@ -43,6 +43,7 @@
#include "internal_shared.hpp"
#include "opencv2/gpu/device/limits.hpp"
#include "opencv2/gpu/device/vec_distance.hpp"
#include "opencv2/gpu/device/datamov_utils.hpp"
using namespace cv::gpu;
using namespace cv::gpu::device;
@ -235,7 +236,15 @@ namespace cv { namespace gpu { namespace bf_knnmatch
{
const int loadX = threadIdx.x + i * BLOCK_SIZE;
s_train[threadIdx.x * BLOCK_SIZE + threadIdx.y] = loadX < train.cols ? train.ptr(min(t * BLOCK_SIZE + threadIdx.y, train.rows - 1))[loadX] : 0;
s_train[threadIdx.x * BLOCK_SIZE + threadIdx.y] = 0;
if (loadX < train.cols)
{
T val;
ForceGlob<T>::Load(train.ptr(min(t * BLOCK_SIZE + threadIdx.y, train.rows - 1)), loadX, val);
s_train[threadIdx.x * BLOCK_SIZE + threadIdx.y] = val;
}
__syncthreads();
@ -402,15 +411,18 @@ namespace cv { namespace gpu { namespace bf_knnmatch
{
const int loadX = threadIdx.x + i * BLOCK_SIZE;
s_query[threadIdx.y * BLOCK_SIZE + threadIdx.x] = 0;
s_train[threadIdx.x * BLOCK_SIZE + threadIdx.y] = 0;
if (loadX < query.cols)
{
s_query[threadIdx.y * BLOCK_SIZE + threadIdx.x] = query.ptr(min(queryIdx, query.rows - 1))[loadX];
s_train[threadIdx.x * BLOCK_SIZE + threadIdx.y] = train.ptr(min(t * BLOCK_SIZE + threadIdx.y, train.rows - 1))[loadX];
}
else
{
s_query[threadIdx.y * BLOCK_SIZE + threadIdx.x] = 0;
s_train[threadIdx.x * BLOCK_SIZE + threadIdx.y] = 0;
T val;
ForceGlob<T>::Load(query.ptr(min(queryIdx, query.rows - 1)), loadX, val);
s_query[threadIdx.y * BLOCK_SIZE + threadIdx.x] = val;
ForceGlob<T>::Load(train.ptr(min(t * BLOCK_SIZE + threadIdx.y, train.rows - 1)), loadX, val);
s_train[threadIdx.x * BLOCK_SIZE + threadIdx.y] = val;
}
__syncthreads();
@ -573,15 +585,18 @@ namespace cv { namespace gpu { namespace bf_knnmatch
{
const int loadX = threadIdx.x + i * BLOCK_SIZE;
s_query[threadIdx.y * BLOCK_SIZE + threadIdx.x] = 0;
s_train[threadIdx.x * BLOCK_SIZE + threadIdx.y] = 0;
if (loadX < query.cols)
{
s_query[threadIdx.y * BLOCK_SIZE + threadIdx.x] = query.ptr(min(queryIdx, query.rows - 1))[loadX];
s_train[threadIdx.x * BLOCK_SIZE + threadIdx.y] = train.ptr(min(t * BLOCK_SIZE + threadIdx.y, train.rows - 1))[loadX];
}
else
{
s_query[threadIdx.y * BLOCK_SIZE + threadIdx.x] = 0;
s_train[threadIdx.x * BLOCK_SIZE + threadIdx.y] = 0;
T val;
ForceGlob<T>::Load(query.ptr(min(queryIdx, query.rows - 1)), loadX, val);
s_query[threadIdx.y * BLOCK_SIZE + threadIdx.x] = val;
ForceGlob<T>::Load(train.ptr(min(t * BLOCK_SIZE + threadIdx.y, train.rows - 1)), loadX, val);
s_train[threadIdx.x * BLOCK_SIZE + threadIdx.y] = val;
}
__syncthreads();

45
modules/gpu/src/cuda/bf_match.cu
Unescape View File

@ -43,6 +43,7 @@
#include "internal_shared.hpp"
#include "opencv2/gpu/device/limits.hpp"
#include "opencv2/gpu/device/vec_distance.hpp"
#include "opencv2/gpu/device/datamov_utils.hpp"
using namespace cv::gpu;
using namespace cv::gpu::device;
@ -110,7 +111,15 @@ namespace cv { namespace gpu { namespace bf_match
{
const int loadX = threadIdx.x + i * BLOCK_SIZE;
s_train[threadIdx.x * BLOCK_SIZE + threadIdx.y] = loadX < train.cols ? train.ptr(min(t * BLOCK_SIZE + threadIdx.y, train.rows - 1))[loadX] : 0;
s_train[threadIdx.x * BLOCK_SIZE + threadIdx.y] = 0;
if (loadX < train.cols)
{
T val;
ForceGlob<T>::Load(train.ptr(min(t * BLOCK_SIZE + threadIdx.y, train.rows - 1)), loadX, val);
s_train[threadIdx.x * BLOCK_SIZE + threadIdx.y] = val;
}
__syncthreads();
@ -258,15 +267,18 @@ namespace cv { namespace gpu { namespace bf_match
{
const int loadX = threadIdx.x + i * BLOCK_SIZE;
s_query[threadIdx.y * BLOCK_SIZE + threadIdx.x] = 0;
s_train[threadIdx.x * BLOCK_SIZE + threadIdx.y] = 0;
if (loadX < query.cols)
{
s_query[threadIdx.y * BLOCK_SIZE + threadIdx.x] = query.ptr(min(queryIdx, query.rows - 1))[loadX];
s_train[threadIdx.x * BLOCK_SIZE + threadIdx.y] = train.ptr(min(t * BLOCK_SIZE + threadIdx.y, train.rows - 1))[loadX];
}
else
{
s_query[threadIdx.y * BLOCK_SIZE + threadIdx.x] = 0;
s_train[threadIdx.x * BLOCK_SIZE + threadIdx.y] = 0;
T val;
ForceGlob<T>::Load(query.ptr(min(queryIdx, query.rows - 1)), loadX, val);
s_query[threadIdx.y * BLOCK_SIZE + threadIdx.x] = val;
ForceGlob<T>::Load(train.ptr(min(t * BLOCK_SIZE + threadIdx.y, train.rows - 1)), loadX, val);
s_train[threadIdx.x * BLOCK_SIZE + threadIdx.y] = val;
}
__syncthreads();
@ -410,15 +422,18 @@ namespace cv { namespace gpu { namespace bf_match
{
const int loadX = threadIdx.x + i * BLOCK_SIZE;
s_query[threadIdx.y * BLOCK_SIZE + threadIdx.x] = 0;
s_train[threadIdx.x * BLOCK_SIZE + threadIdx.y] = 0;
if (loadX < query.cols)
{
s_query[threadIdx.y * BLOCK_SIZE + threadIdx.x] = query.ptr(min(queryIdx, query.rows - 1))[loadX];
s_train[threadIdx.x * BLOCK_SIZE + threadIdx.y] = train.ptr(min(t * BLOCK_SIZE + threadIdx.y, train.rows - 1))[loadX];
}
else
{
s_query[threadIdx.y * BLOCK_SIZE + threadIdx.x] = 0;
s_train[threadIdx.x * BLOCK_SIZE + threadIdx.y] = 0;
T val;
ForceGlob<T>::Load(query.ptr(min(queryIdx, query.rows - 1)), loadX, val);
s_query[threadIdx.y * BLOCK_SIZE + threadIdx.x] = val;
ForceGlob<T>::Load(train.ptr(min(t * BLOCK_SIZE + threadIdx.y, train.rows - 1)), loadX, val);
s_train[threadIdx.x * BLOCK_SIZE + threadIdx.y] = val;
}
__syncthreads();

35
modules/gpu/src/cuda/bf_radius_match.cu
Unescape View File

@ -43,6 +43,7 @@
#include "internal_shared.hpp"
#include "opencv2/gpu/device/limits.hpp"
#include "opencv2/gpu/device/vec_distance.hpp"
#include "opencv2/gpu/device/datamov_utils.hpp"
using namespace cv::gpu;
using namespace cv::gpu::device;
@ -73,15 +74,18 @@ namespace cv { namespace gpu { namespace bf_radius_match
{
const int loadX = threadIdx.x + i * BLOCK_SIZE;
s_query[threadIdx.y * BLOCK_SIZE + threadIdx.x] = 0;
s_train[threadIdx.x * BLOCK_SIZE + threadIdx.y] = 0;
if (loadX < query.cols)
{
s_query[threadIdx.y * BLOCK_SIZE + threadIdx.x] = query.ptr(min(queryIdx, query.rows - 1))[loadX];
s_train[threadIdx.x * BLOCK_SIZE + threadIdx.y] = train.ptr(min(blockIdx.x * BLOCK_SIZE + threadIdx.y, train.rows - 1))[loadX];
}
else
{
s_query[threadIdx.y * BLOCK_SIZE + threadIdx.x] = 0;
s_train[threadIdx.x * BLOCK_SIZE + threadIdx.y] = 0;
T val;
ForceGlob<T>::Load(query.ptr(min(queryIdx, query.rows - 1)), loadX, val);
s_query[threadIdx.y * BLOCK_SIZE + threadIdx.x] = val;
ForceGlob<T>::Load(train.ptr(min(blockIdx.x * BLOCK_SIZE + threadIdx.y, train.rows - 1)), loadX, val);
s_train[threadIdx.x * BLOCK_SIZE + threadIdx.y] = val;
}
__syncthreads();
@ -181,15 +185,18 @@ namespace cv { namespace gpu { namespace bf_radius_match
{
const int loadX = threadIdx.x + i * BLOCK_SIZE;
s_query[threadIdx.y * BLOCK_SIZE + threadIdx.x] = 0;
s_train[threadIdx.x * BLOCK_SIZE + threadIdx.y] = 0;
if (loadX < query.cols)
{
s_query[threadIdx.y * BLOCK_SIZE + threadIdx.x] = query.ptr(min(queryIdx, query.rows - 1))[loadX];
s_train[threadIdx.x * BLOCK_SIZE + threadIdx.y] = train.ptr(min(blockIdx.x * BLOCK_SIZE + threadIdx.y, train.rows - 1))[loadX];
}
else
{
s_query[threadIdx.y * BLOCK_SIZE + threadIdx.x] = 0;
s_train[threadIdx.x * BLOCK_SIZE + threadIdx.y] = 0;
T val;
ForceGlob<T>::Load(query.ptr(min(queryIdx, query.rows - 1)), loadX, val);
s_query[threadIdx.y * BLOCK_SIZE + threadIdx.x] = val;
ForceGlob<T>::Load(train.ptr(min(blockIdx.x * BLOCK_SIZE + threadIdx.y, train.rows - 1)), loadX, val);
s_train[threadIdx.x * BLOCK_SIZE + threadIdx.y] = val;
}
__syncthreads();

Write Preview
Loading…
Cancel
Save