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

implemented asynchronous call for StereoBM()

Browse Source
This commit is contained in:
Andrey Morozov
2010-07-27 08:29:40 +00:00
parent dc0f313924
commit 12dc52c2e7
4 changed files with 116 additions and 99 deletions
Download Patch File Download Diff File Expand all files Collapse all files

156
modules/gpu/src/cuda/stereobm.cu
View File

@@ -55,13 +55,13 @@ using namespace cv::gpu;
#define ROWSperTHREAD 21 // the number of rows a thread will process
namespace stereobm_gpu
namespace stereobm_gpu
{
#define BLOCK_W 128 // the thread block width (464)
#define N_DISPARITIES 8
#define STEREO_MIND 0 // The minimum d range to check
#define STEREO_MIND 0 // The minimum d range to check
#define STEREO_DISP_STEP N_DISPARITIES // the d step, must be <= 1 to avoid aliasing
__constant__ unsigned int* cminSSDImage;
@@ -71,7 +71,7 @@ __constant__ int cheight;
__device__ int SQ(int a)
{
return a * a;
return a * a;
}
template<int RADIUS>
@@ -82,7 +82,7 @@ __device__ unsigned int CalcSSD(unsigned int *col_ssd_cache, unsigned int *col_s
for(int i = 1; i <= RADIUS; i++)
cache += col_ssd[i];
col_ssd_cache[0] = cache;
__syncthreads();
@@ -101,7 +101,7 @@ __device__ uint2 MinSSD(unsigned int *col_ssd_cache, unsigned int *col_ssd)
{
unsigned int ssd[N_DISPARITIES];
//See above: #define COL_SSD_SIZE (BLOCK_W + 2 * RADIUS)
//See above: #define COL_SSD_SIZE (BLOCK_W + 2 * RADIUS)
ssd[0] = CalcSSD<RADIUS>(col_ssd_cache, col_ssd + 0 * (BLOCK_W + 2 * RADIUS));
ssd[1] = CalcSSD<RADIUS>(col_ssd_cache, col_ssd + 1 * (BLOCK_W + 2 * RADIUS));
ssd[2] = CalcSSD<RADIUS>(col_ssd_cache, col_ssd + 2 * (BLOCK_W + 2 * RADIUS));
@@ -146,7 +146,7 @@ __device__ void StepDown(int idx1, int idx2, unsigned char* imageL, unsigned cha
rightPixel1[4] = imageR[idx1 - 4];
rightPixel1[5] = imageR[idx1 - 5];
rightPixel1[6] = imageR[idx1 - 6];
rightPixel2[7] = imageR[idx2 - 7];
rightPixel2[0] = imageR[idx2 - 0];
rightPixel2[1] = imageR[idx2 - 1];
@@ -155,16 +155,16 @@ __device__ void StepDown(int idx1, int idx2, unsigned char* imageL, unsigned cha
rightPixel2[4] = imageR[idx2 - 4];
rightPixel2[5] = imageR[idx2 - 5];
rightPixel2[6] = imageR[idx2 - 6];
//See above: #define COL_SSD_SIZE (BLOCK_W + 2 * RADIUS)
diff1 = leftPixel1 - rightPixel1[0];
diff2 = leftPixel2 - rightPixel2[0];
//See above: #define COL_SSD_SIZE (BLOCK_W + 2 * RADIUS)
diff1 = leftPixel1 - rightPixel1[0];
diff2 = leftPixel2 - rightPixel2[0];
col_ssd[0 * (BLOCK_W + 2 * RADIUS)] += SQ(diff2) - SQ(diff1);
diff1 = leftPixel1 - rightPixel1[1];
diff2 = leftPixel2 - rightPixel2[1];
col_ssd[1 * (BLOCK_W + 2 * RADIUS)] += SQ(diff2) - SQ(diff1);
diff1 = leftPixel1 - rightPixel1[2];
diff2 = leftPixel2 - rightPixel2[2];
col_ssd[2 * (BLOCK_W + 2 * RADIUS)] += SQ(diff2) - SQ(diff1);
@@ -172,19 +172,19 @@ __device__ void StepDown(int idx1, int idx2, unsigned char* imageL, unsigned cha
diff1 = leftPixel1 - rightPixel1[3];
diff2 = leftPixel2 - rightPixel2[3];
col_ssd[3 * (BLOCK_W + 2 * RADIUS)] += SQ(diff2) - SQ(diff1);
diff1 = leftPixel1 - rightPixel1[4];
diff2 = leftPixel2 - rightPixel2[4];
diff1 = leftPixel1 - rightPixel1[4];
diff2 = leftPixel2 - rightPixel2[4];
col_ssd[4 * (BLOCK_W + 2 * RADIUS)] += SQ(diff2) - SQ(diff1);
diff1 = leftPixel1 - rightPixel1[5];
diff2 = leftPixel2 - rightPixel2[5];
col_ssd[5 * (BLOCK_W + 2 * RADIUS)] += SQ(diff2) - SQ(diff1);
diff1 = leftPixel1 - rightPixel1[6];
diff2 = leftPixel2 - rightPixel2[6];
col_ssd[6 * (BLOCK_W + 2 * RADIUS)] += SQ(diff2) - SQ(diff1);
diff1 = leftPixel1 - rightPixel1[7];
diff2 = leftPixel2 - rightPixel2[7];
col_ssd[7 * (BLOCK_W + 2 * RADIUS)] += SQ(diff2) - SQ(diff1);
@@ -203,7 +203,7 @@ __device__ void InitColSSD(int x_tex, int y_tex, int im_pitch, unsigned char* im
leftPixel1 = imageL[idx];
idx = idx - d;
diffa[0] += SQ(leftPixel1 - imageR[idx - 0]);
diffa[0] += SQ(leftPixel1 - imageR[idx - 0]);
diffa[1] += SQ(leftPixel1 - imageR[idx - 1]);
diffa[2] += SQ(leftPixel1 - imageR[idx - 2]);
diffa[3] += SQ(leftPixel1 - imageR[idx - 3]);
@@ -213,7 +213,7 @@ __device__ void InitColSSD(int x_tex, int y_tex, int im_pitch, unsigned char* im
diffa[7] += SQ(leftPixel1 - imageR[idx - 7]);
y_tex += 1;
}
}
//See above: #define COL_SSD_SIZE (BLOCK_W + 2 * RADIUS)
col_ssd[0 * (BLOCK_W + 2 * RADIUS)] = diffa[0];
col_ssd[1 * (BLOCK_W + 2 * RADIUS)] = diffa[1];
@@ -225,11 +225,11 @@ __device__ void InitColSSD(int x_tex, int y_tex, int im_pitch, unsigned char* im
col_ssd[7 * (BLOCK_W + 2 * RADIUS)] = diffa[7];
}
template<int RADIUS>
template<int RADIUS>
__global__ void stereoKernel(unsigned char *left, unsigned char *right, size_t img_step, unsigned char* disp, size_t disp_pitch, int maxdisp)
{
extern __shared__ unsigned int col_ssd_cache[];
unsigned int *col_ssd = col_ssd_cache + BLOCK_W + threadIdx.x;
unsigned int *col_ssd = col_ssd_cache + BLOCK_W + threadIdx.x;
unsigned int *col_ssd_extra = threadIdx.x < (2 * RADIUS) ? col_ssd + BLOCK_W : 0; //#define N_DIRTY_PIXELS (2 * RADIUS)
//#define X (blockIdx.x * BLOCK_W + threadIdx.x + STEREO_MAXD)
@@ -241,13 +241,13 @@ __global__ void stereoKernel(unsigned char *left, unsigned char *right, size_t i
unsigned int* minSSDImage = cminSSDImage + X + Y * cminSSD_step;
unsigned char* disparImage = disp + X + Y * disp_pitch;
/* if (X < cwidth)
{
{
unsigned int *minSSDImage_end = minSSDImage + min(ROWSperTHREAD, cheight - Y) * minssd_step;
for(uint *ptr = minSSDImage; ptr != minSSDImage_end; ptr += minssd_step )
*ptr = 0xFFFFFFFF;
*ptr = 0xFFFFFFFF;
}*/
int end_row = min(ROWSperTHREAD, cheight - Y);
int y_tex;
int y_tex;
int x_tex = X - RADIUS;
if (x_tex >= cwidth)
@@ -257,7 +257,7 @@ __global__ void stereoKernel(unsigned char *left, unsigned char *right, size_t i
{
y_tex = Y - RADIUS;
InitColSSD<RADIUS>(x_tex, y_tex, img_step, left, right, d, col_ssd);
InitColSSD<RADIUS>(x_tex, y_tex, img_step, left, right, d, col_ssd);
if (col_ssd_extra > 0)
if (x_tex + BLOCK_W < cwidth)
@@ -289,13 +289,13 @@ __global__ void stereoKernel(unsigned char *left, unsigned char *right, size_t i
StepDown<RADIUS>(idx1, idx2, left + BLOCK_W, right + BLOCK_W, d, col_ssd_extra);
y_tex += 1;
__syncthreads(); //before MinSSD function
if (X < cwidth - RADIUS && row < cheight - RADIUS - Y)
{
int idx = row * cminSSD_step;
uint2 minSSD = MinSSD<RADIUS>(col_ssd_cache + threadIdx.x, col_ssd);
{
int idx = row * cminSSD_step;
uint2 minSSD = MinSSD<RADIUS>(col_ssd_cache + threadIdx.x, col_ssd);
if (minSSD.x < minSSDImage[idx])
{
disparImage[disp_pitch * row] = (unsigned char)(d + minSSD.y);
@@ -310,49 +310,57 @@ __global__ void stereoKernel(unsigned char *left, unsigned char *right, size_t i
namespace cv { namespace gpu { namespace impl
{
template<int RADIUS> void kernel_caller(const DevMem2D& left, const DevMem2D& right, const DevMem2D& disp, int maxdisp)
{
template<int RADIUS> void kernel_caller(const DevMem2D& left, const DevMem2D& right, const DevMem2D& disp, int maxdisp, const cudaStream_t & stream)
{
dim3 grid(1,1,1);
dim3 threads(BLOCK_W, 1, 1);
dim3 threads(BLOCK_W, 1, 1);
grid.x = divUp(left.cols - maxdisp - 2 * RADIUS, BLOCK_W);
grid.y = divUp(left.rows - 2 * RADIUS, ROWSperTHREAD);
//See above: #define COL_SSD_SIZE (BLOCK_W + 2 * RADIUS)
size_t smem_size = (BLOCK_W + N_DISPARITIES * (BLOCK_W + 2 * RADIUS)) * sizeof(unsigned int);
//See above: #define COL_SSD_SIZE (BLOCK_W + 2 * RADIUS)
size_t smem_size = (BLOCK_W + N_DISPARITIES * (BLOCK_W + 2 * RADIUS)) * sizeof(unsigned int);
if (stream == 0)
{
stereobm_gpu::stereoKernel<RADIUS><<<grid, threads, smem_size>>>(left.ptr, right.ptr, left.step, disp.ptr, disp.step, maxdisp);
cudaSafeCall( cudaThreadSynchronize() );
}
else
{
stereobm_gpu::stereoKernel<RADIUS><<<grid, threads, smem_size, stream>>>(left.ptr, right.ptr, left.step, disp.ptr, disp.step, maxdisp);
}
stereobm_gpu::stereoKernel<RADIUS><<<grid, threads, smem_size>>>(left.ptr, right.ptr, left.step, disp.ptr, disp.step, maxdisp);
cudaSafeCall( cudaThreadSynchronize() );
};
typedef void (*kernel_caller_t)(const DevMem2D& left, const DevMem2D& right, const DevMem2D& disp, int maxdisp);
typedef void (*kernel_caller_t)(const DevMem2D& left, const DevMem2D& right, const DevMem2D& disp, int maxdisp, const cudaStream_t & stream);
const static kernel_caller_t callers[] =
{
0,
kernel_caller< 1>, kernel_caller< 2>, kernel_caller< 3>, kernel_caller< 4>, kernel_caller< 5>,
kernel_caller< 6>, kernel_caller< 7>, kernel_caller< 8>, kernel_caller< 9>, kernel_caller<10>,
kernel_caller<11>, kernel_caller<12>, kernel_caller<13>, kernel_caller<15>, kernel_caller<15>,
kernel_caller<16>, kernel_caller<17>, kernel_caller<18>, kernel_caller<19>, kernel_caller<20>,
0,
kernel_caller< 1>, kernel_caller< 2>, kernel_caller< 3>, kernel_caller< 4>, kernel_caller< 5>,
kernel_caller< 6>, kernel_caller< 7>, kernel_caller< 8>, kernel_caller< 9>, kernel_caller<10>,
kernel_caller<11>, kernel_caller<12>, kernel_caller<13>, kernel_caller<15>, kernel_caller<15>,
kernel_caller<16>, kernel_caller<17>, kernel_caller<18>, kernel_caller<19>, kernel_caller<20>,
kernel_caller<21>, kernel_caller<22>, kernel_caller<23>, kernel_caller<24>, kernel_caller<25>
//0,0,0, 0,0,0, 0,0,kernel_caller<9>
};
const int calles_num = sizeof(callers)/sizeof(callers[0]);
extern "C" void stereoBM_GPU(const DevMem2D& left, const DevMem2D& right, const DevMem2D& disp, int maxdisp, int winsz, const DevMem2D_<unsigned int>& minSSD_buf)
{
extern "C" void stereoBM_GPU(const DevMem2D& left, const DevMem2D& right, const DevMem2D& disp, int maxdisp, int winsz, const DevMem2D_<unsigned int>& minSSD_buf, const cudaStream_t & stream)
{
int winsz2 = winsz >> 1;
if (winsz2 == 0 || winsz2 >= calles_num)
cv::gpu::error("Unsupported window size", __FILE__, __LINE__);
//cudaSafeCall( cudaFuncSetCacheConfig(&stereoKernel, cudaFuncCachePreferL1) );
//cudaSafeCall( cudaFuncSetCacheConfig(&stereoKernel, cudaFuncCachePreferShared) );
//cudaSafeCall( cudaFuncSetCacheConfig(&stereoKernel, cudaFuncCachePreferShared) );
cudaSafeCall( cudaMemset2D(disp.ptr, disp.step, 0, disp.cols, disp.rows) );
cudaSafeCall( cudaMemset2D(minSSD_buf.ptr, minSSD_buf.step, 0xFF, minSSD_buf.cols * minSSD_buf.elemSize(), disp.rows) );
cudaSafeCall( cudaMemset2D(minSSD_buf.ptr, minSSD_buf.step, 0xFF, minSSD_buf.cols * minSSD_buf.elemSize(), disp.rows) );
cudaSafeCall( cudaMemcpyToSymbol( stereobm_gpu::cwidth, &left.cols, sizeof(left.cols) ) );
cudaSafeCall( cudaMemcpyToSymbol( stereobm_gpu::cheight, &left.rows, sizeof(left.rows) ) );
@@ -361,7 +369,7 @@ namespace cv { namespace gpu { namespace impl
size_t minssd_step = minSSD_buf.step/minSSD_buf.elemSize();
cudaSafeCall( cudaMemcpyToSymbol( stereobm_gpu::cminSSD_step, &minssd_step, sizeof(minssd_step) ) );
callers[winsz2](left, right, disp, maxdisp);
callers[winsz2](left, right, disp, maxdisp, stream);
}
}}}
@@ -381,7 +389,7 @@ extern "C" __global__ void prefilter_kernel(unsigned char *output, size_t step,
if (x < width && y < height)
{
int conv = (int)tex2D(texForSobel, x - 1, y - 1) * (-1) + (int)tex2D(texForSobel, x + 1, y - 1) * (1) +
int conv = (int)tex2D(texForSobel, x - 1, y - 1) * (-1) + (int)tex2D(texForSobel, x + 1, y - 1) * (1) +
(int)tex2D(texForSobel, x - 1, y ) * (-2) + (int)tex2D(texForSobel, x + 1, y ) * (2) +
(int)tex2D(texForSobel, x - 1, y + 1) * (-1) + (int)tex2D(texForSobel, x + 1, y + 1) * (1);
@@ -398,18 +406,18 @@ namespace cv { namespace gpu { namespace impl
extern "C" void prefilter_xsobel(const DevMem2D& input, const DevMem2D& output, int prefilterCap)
{
cudaChannelFormatDesc desc = cudaCreateChannelDesc<unsigned char>();
cudaSafeCall( cudaBindTexture2D( 0, stereobm_gpu::texForSobel, input.ptr, desc, input.cols, input.rows, input.step ) );
cudaSafeCall( cudaBindTexture2D( 0, stereobm_gpu::texForSobel, input.ptr, desc, input.cols, input.rows, input.step ) );
dim3 threads(16, 16, 1);
dim3 grid(1, 1, 1);
grid.x = divUp(input.cols, threads.x);
grid.y = divUp(input.rows, threads.y);
grid.y = divUp(input.rows, threads.y);
stereobm_gpu::prefilter_kernel<<<grid, threads>>>(output.ptr, output.step, output.cols, output.rows, prefilterCap);
cudaSafeCall( cudaThreadSynchronize() );
cudaSafeCall( cudaUnbindTexture (stereobm_gpu::texForSobel ) );
cudaSafeCall( cudaUnbindTexture (stereobm_gpu::texForSobel ) );
}
}}}
@@ -424,8 +432,8 @@ namespace stereobm_gpu
texture<unsigned char, 2, cudaReadModeNormalizedFloat> texForTF;
__device__ float sobel(int x, int y)
{
float conv = tex2D(texForTF, x - 1, y - 1) * (-1) + tex2D(texForTF, x + 1, y - 1) * (1) +
{
float conv = tex2D(texForTF, x - 1, y - 1) * (-1) + tex2D(texForTF, x + 1, y - 1) * (1) +
tex2D(texForTF, x - 1, y ) * (-2) + tex2D(texForTF, x + 1, y ) * (2) +
tex2D(texForTF, x - 1, y + 1) * (-1) + tex2D(texForTF, x + 1, y + 1) * (1);
return fabs(conv);
@@ -453,28 +461,28 @@ __device__ float CalcSums(float *cols, float *cols_cache, int winsz)
return cols[0] + cache + cache2;
}
#define RpT (2 * ROWSperTHREAD) // got experimentally
#define RpT (2 * ROWSperTHREAD) // got experimentally
extern "C" __global__ void textureness_kernel(unsigned char *disp, size_t disp_step, int winsz, float threshold, int width, int height)
{
{
int winsz2 = winsz/2;
int n_dirty_pixels = (winsz2) * 2;
extern __shared__ float cols_cache[];
float *cols = cols_cache + blockDim.x + threadIdx.x;
float *cols_extra = threadIdx.x < n_dirty_pixels ? cols + blockDim.x : 0;
float *cols = cols_cache + blockDim.x + threadIdx.x;
float *cols_extra = threadIdx.x < n_dirty_pixels ? cols + blockDim.x : 0;
int x = blockIdx.x * blockDim.x + threadIdx.x;
int x = blockIdx.x * blockDim.x + threadIdx.x;
int beg_row = blockIdx.y * RpT;
int end_row = min(beg_row + RpT, height);
if (x < width)
{
if (x < width)
{
int y = beg_row;
float sum = 0;
float sum_extra = 0;
float sum = 0;
float sum_extra = 0;
for(int i = y - winsz2; i <= y + winsz2; ++i)
{
sum += sobel(x - winsz2, i);
@@ -486,11 +494,11 @@ extern "C" __global__ void textureness_kernel(unsigned char *disp, size_t disp_s
*cols_extra = sum_extra;
__syncthreads();
float sum_win = CalcSums(cols, cols_cache + threadIdx.x, winsz) * 255;
if (sum_win < threshold)
disp[y * disp_step + x] = 0;
__syncthreads();
for(int y = beg_row + 1; y < end_row; ++y)
@@ -505,12 +513,12 @@ extern "C" __global__ void textureness_kernel(unsigned char *disp, size_t disp_s
}
__syncthreads();
float sum_win = CalcSums(cols, cols_cache + threadIdx.x, winsz) * 255;
float sum_win = CalcSums(cols, cols_cache + threadIdx.x, winsz) * 255;
if (sum_win < threshold)
disp[y * disp_step + x] = 0;
__syncthreads();
}
}
}
}
}
@@ -521,21 +529,21 @@ namespace cv { namespace gpu { namespace impl
{
avgTexturenessThreshold *= winsz * winsz;
stereobm_gpu::texForTF.filterMode = cudaFilterModeLinear;
stereobm_gpu::texForTF.filterMode = cudaFilterModeLinear;
stereobm_gpu::texForTF.addressMode[0] = cudaAddressModeWrap;
stereobm_gpu::texForTF.addressMode[1] = cudaAddressModeWrap;
cudaChannelFormatDesc desc = cudaCreateChannelDesc<unsigned char>();
cudaSafeCall( cudaBindTexture2D( 0, stereobm_gpu::texForTF, input.ptr, desc, input.cols, input.rows, input.step ) );
cudaSafeCall( cudaBindTexture2D( 0, stereobm_gpu::texForTF, input.ptr, desc, input.cols, input.rows, input.step ) );
dim3 threads(128, 1, 1);
dim3 grid(1, 1, 1);
grid.x = divUp(input.cols, threads.x);
grid.y = divUp(input.rows, RpT);
size_t smem_size = (threads.x + threads.x + (winsz/2) * 2 ) * sizeof(float);
grid.y = divUp(input.rows, RpT);
size_t smem_size = (threads.x + threads.x + (winsz/2) * 2 ) * sizeof(float);
stereobm_gpu::textureness_kernel<<<grid, threads, smem_size>>>(disp.ptr, disp.step, winsz, avgTexturenessThreshold, disp.cols, disp.rows);
cudaSafeCall( cudaThreadSynchronize() );