Skip to content

O
opencv
Commit 9ba25e9d authored by Andrey Kamaev's avatar Andrey Kamaev Committed by OpenCV Buildbot
Browse files
Options

Merge pull request #742 from bitwangyaoyao:2.4_fix

parents 0c64fc61 8cc5b980
Expand all Show whitespace changes
Inline Side-by-side
Showing with 175 additions and 64 deletions
modules/ocl/include/opencv2/ocl/private/util.hpp
View file @ 9ba25e9d
......@@ -128,7 +128,8 @@ namespace cv
enum DEVICE_INFO
{
WAVEFRONT_SIZE, //in AMD speak
WARP_SIZE = WAVEFRONT_SIZE //in nvidia speak
WARP_SIZE = WAVEFRONT_SIZE, //in nvidia speak
IS_CPU_DEVICE //check if the device is CPU
};
//info should have been pre-allocated
void CV_EXPORTS queryDeviceInfo(DEVICE_INFO info_type, void* info);
......
modules/ocl/src/hog.cpp
View file @ 9ba25e9d
......@@ -44,7 +44,6 @@
//M*/
#include "precomp.hpp"
using namespace cv;
using namespace cv::ocl;
using namespace std;
......@@ -230,7 +229,6 @@ void cv::ocl::HOGDescriptor::computeGradient(const oclMat &img, oclMat &grad, oc
}
}
void cv::ocl::HOGDescriptor::computeBlockHistograms(const oclMat &img)
{
computeGradient(img, grad, qangle);
......@@ -1571,6 +1569,27 @@ void cv::ocl::device::hog::set_up_constants(int nbins, int block_stride_x, int b
cdescr_size = descr_size;
}
static inline int divUp(int total, int grain)
{
return (total + grain - 1) / grain;
}
static void openCLExecuteKernel_hog(Context *clCxt , const char **source, string kernelName,
size_t globalThreads[3], size_t localThreads[3],
vector< pair<size_t, const void *> > &args)
{
size_t wave_size = 0;
queryDeviceInfo(WAVEFRONT_SIZE, &wave_size);
if (wave_size <= 16)
{
char build_options[64];
sprintf(build_options, (wave_size == 16) ? "-D WAVE_SIZE_16" : "-D WAVE_SIZE_1");
openCLExecuteKernel(clCxt, source, kernelName, globalThreads, localThreads, args, -1, -1, build_options);
}
else
openCLExecuteKernel(clCxt, source, kernelName, globalThreads, localThreads, args, -1, -1);
}
void cv::ocl::device::hog::compute_hists(int nbins, int block_stride_x, int block_stride_y,
int height, int width, const cv::ocl::oclMat &grad,
const cv::ocl::oclMat &qangle, float sigma, cv::ocl::oclMat &block_hists)
......@@ -1582,8 +1601,10 @@ void cv::ocl::device::hog::compute_hists(int nbins, int block_stride_x, int bloc
int img_block_width = (width - CELLS_PER_BLOCK_X * CELL_WIDTH + block_stride_x) / block_stride_x;
int img_block_height = (height - CELLS_PER_BLOCK_Y * CELL_HEIGHT + block_stride_y) / block_stride_y;
size_t globalThreads[3] = { img_block_width * 32, img_block_height * 2, 1 };
size_t localThreads[3] = { 32, 2, 1 };
int blocks_total = img_block_width * img_block_height;
int blocks_in_group = 4;
size_t localThreads[3] = { blocks_in_group * 24, 2, 1 };
size_t globalThreads[3] = { divUp(blocks_total, blocks_in_group) * localThreads[0], 2, 1 };
int grad_quadstep = grad.step >> 2;
int qangle_step = qangle.step;
......@@ -1593,14 +1614,15 @@ void cv::ocl::device::hog::compute_hists(int nbins, int block_stride_x, int bloc
int hists_size = (nbins * CELLS_PER_BLOCK_X * CELLS_PER_BLOCK_Y * 12) * sizeof(float);
int final_hists_size = (nbins * CELLS_PER_BLOCK_X * CELLS_PER_BLOCK_Y) * sizeof(float);
int smem = hists_size + final_hists_size;
int smem = (hists_size + final_hists_size) * blocks_in_group;
args.push_back( make_pair( sizeof(cl_int), (void *)&width));
args.push_back( make_pair( sizeof(cl_int), (void *)&cblock_stride_x));
args.push_back( make_pair( sizeof(cl_int), (void *)&cblock_stride_y));
args.push_back( make_pair( sizeof(cl_int), (void *)&cnbins));
args.push_back( make_pair( sizeof(cl_int), (void *)&cblock_hist_size));
args.push_back( make_pair( sizeof(cl_int), (void *)&img_block_width));
args.push_back( make_pair( sizeof(cl_int), (void *)&blocks_in_group));
args.push_back( make_pair( sizeof(cl_int), (void *)&blocks_total));
args.push_back( make_pair( sizeof(cl_int), (void *)&grad_quadstep));
args.push_back( make_pair( sizeof(cl_int), (void *)&qangle_step));
args.push_back( make_pair( sizeof(cl_mem), (void *)&grad.data));
......@@ -1609,7 +1631,7 @@ void cv::ocl::device::hog::compute_hists(int nbins, int block_stride_x, int bloc
args.push_back( make_pair( sizeof(cl_mem), (void *)&block_hists.data));
args.push_back( make_pair( smem, (void *)NULL));
openCLExecuteKernel2(clCxt, &objdetect_hog, kernelName, globalThreads, localThreads, args, -1, -1);
openCLExecuteKernel_hog(clCxt, &objdetect_hog, kernelName, globalThreads, localThreads, args);
}
void cv::ocl::device::hog::normalize_hists(int nbins, int block_stride_x, int block_stride_y,
......@@ -1637,7 +1659,7 @@ void cv::ocl::device::hog::normalize_hists(int nbins, int block_stride_x, int bl
args.push_back( make_pair( sizeof(cl_float), (void *)&threshold));
args.push_back( make_pair( nthreads * sizeof(float), (void *)NULL));
openCLExecuteKernel2(clCxt, &objdetect_hog, kernelName, globalThreads, localThreads, args, -1, -1);
openCLExecuteKernel_hog(clCxt, &objdetect_hog, kernelName, globalThreads, localThreads, args);
}
void cv::ocl::device::hog::classify_hists(int win_height, int win_width, int block_stride_y,
......@@ -1671,7 +1693,7 @@ void cv::ocl::device::hog::classify_hists(int win_height, int win_width, int blo
args.push_back( make_pair( sizeof(cl_float), (void *)&threshold));
args.push_back( make_pair( sizeof(cl_mem), (void *)&labels.data));
openCLExecuteKernel2(clCxt, &objdetect_hog, kernelName, globalThreads, localThreads, args, -1, -1);
openCLExecuteKernel_hog(clCxt, &objdetect_hog, kernelName, globalThreads, localThreads, args);
}
void cv::ocl::device::hog::extract_descrs_by_rows(int win_height, int win_width, int block_stride_y, int block_stride_x,
......@@ -1702,7 +1724,7 @@ void cv::ocl::device::hog::extract_descrs_by_rows(int win_height, int win_width,
args.push_back( make_pair( sizeof(cl_mem), (void *)&block_hists.data));
args.push_back( make_pair( sizeof(cl_mem), (void *)&descriptors.data));
openCLExecuteKernel2(clCxt, &objdetect_hog, kernelName, globalThreads, localThreads, args, -1, -1);
openCLExecuteKernel(clCxt, &objdetect_hog, kernelName, globalThreads, localThreads, args, -1, -1);
}
void cv::ocl::device::hog::extract_descrs_by_cols(int win_height, int win_width, int block_stride_y, int block_stride_x,
......@@ -1734,12 +1756,7 @@ void cv::ocl::device::hog::extract_descrs_by_cols(int win_height, int win_width,
args.push_back( make_pair( sizeof(cl_mem), (void *)&block_hists.data));
args.push_back( make_pair( sizeof(cl_mem), (void *)&descriptors.data));
openCLExecuteKernel2(clCxt, &objdetect_hog, kernelName, globalThreads, localThreads, args, -1, -1);
}
static inline int divUp(int total, int grain)
{
return (total + grain - 1) / grain;
openCLExecuteKernel(clCxt, &objdetect_hog, kernelName, globalThreads, localThreads, args, -1, -1);
}
void cv::ocl::device::hog::compute_gradients_8UC1(int height, int width, const cv::ocl::oclMat &img,
......@@ -1768,7 +1785,7 @@ void cv::ocl::device::hog::compute_gradients_8UC1(int height, int width, const c
args.push_back( make_pair( sizeof(cl_char), (void *)&correctGamma));
args.push_back( make_pair( sizeof(cl_int), (void *)&cnbins));
openCLExecuteKernel2(clCxt, &objdetect_hog, kernelName, globalThreads, localThreads, args, -1, -1);
openCLExecuteKernel(clCxt, &objdetect_hog, kernelName, globalThreads, localThreads, args, -1, -1);
}
void cv::ocl::device::hog::compute_gradients_8UC4(int height, int width, const cv::ocl::oclMat &img,
......@@ -1798,7 +1815,7 @@ void cv::ocl::device::hog::compute_gradients_8UC4(int height, int width, const c
args.push_back( make_pair( sizeof(cl_char), (void *)&correctGamma));
args.push_back( make_pair( sizeof(cl_int), (void *)&cnbins));
openCLExecuteKernel2(clCxt, &objdetect_hog, kernelName, globalThreads, localThreads, args, -1, -1);
openCLExecuteKernel(clCxt, &objdetect_hog, kernelName, globalThreads, localThreads, args, -1, -1);
}
void cv::ocl::device::hog::resize( const oclMat &src, oclMat &dst, const Size sz)
......@@ -1815,14 +1832,16 @@ void cv::ocl::device::hog::resize( const oclMat &src, oclMat &dst, const Size sz
float ifx = (float)src.cols / sz.width;
float ify = (float)src.rows / sz.height;
int src_step = static_cast<int>(src.step);
int dst_step = static_cast<int>(dst.step);
vector< pair<size_t, const void *> > args;
args.push_back( make_pair(sizeof(cl_mem), (void *)&dst.data));
args.push_back( make_pair(sizeof(cl_mem), (void *)&src.data));
args.push_back( make_pair(sizeof(cl_int), (void *)&dst.offset));
args.push_back( make_pair(sizeof(cl_int), (void *)&src.offset));
args.push_back( make_pair(sizeof(cl_int), (void *)&dst.step));
args.push_back( make_pair(sizeof(cl_int), (void *)&src.step));
args.push_back( make_pair(sizeof(cl_int), (void *)&dst_step));
args.push_back( make_pair(sizeof(cl_int), (void *)&src_step));
args.push_back( make_pair(sizeof(cl_int), (void *)&src.cols));
args.push_back( make_pair(sizeof(cl_int), (void *)&src.rows));
args.push_back( make_pair(sizeof(cl_int), (void *)&sz.width));
......@@ -1830,5 +1849,5 @@ void cv::ocl::device::hog::resize( const oclMat &src, oclMat &dst, const Size sz
args.push_back( make_pair(sizeof(cl_float), (void *)&ifx));
args.push_back( make_pair(sizeof(cl_float), (void *)&ify));
openCLExecuteKernel2(clCxt, &objdetect_hog, kernelName, globalThreads, localThreads, args, -1, -1);
openCLExecuteKernel(clCxt, &objdetect_hog, kernelName, globalThreads, localThreads, args, -1, -1);
}
modules/ocl/src/initialization.cpp
View file @ 9ba25e9d
......@@ -397,6 +397,15 @@ namespace cv
}
break;
case IS_CPU_DEVICE:
{
cl_device_type devicetype;
openCLSafeCall(clGetDeviceInfo(impl->devices[impl->devnum],
CL_DEVICE_TYPE, sizeof(cl_device_type),
&devicetype, NULL));
*(bool*)info = (devicetype == CVCL_DEVICE_TYPE_CPU);
}
break;
default:
CV_Error(-1, "Invalid device info type");
break;
......
modules/ocl/src/matrix_operations.cpp
View file @ 9ba25e9d
......@@ -394,7 +394,7 @@ void cv::ocl::oclMat::convertTo( oclMat &dst, int rtype, double alpha, double be
if( rtype < 0 )
rtype = type();
else
rtype = CV_MAKETYPE(CV_MAT_DEPTH(rtype), channels());
rtype = CV_MAKETYPE(CV_MAT_DEPTH(rtype), oclchannels());
//int scn = channels();
int sdepth = depth(), ddepth = CV_MAT_DEPTH(rtype);
......
modules/ocl/src/opencl/objdetect_hog.cl
View file @ 9ba25e9d
......@@ -53,39 +53,48 @@
//----------------------------------------------------------------------------
// Histogram computation
__kernel void compute_hists_kernel(const int width, const int cblock_stride_x, const int cblock_stride_y,
// 12 threads for a cell, 12x4 threads per block
__kernel void compute_hists_kernel(
const int cblock_stride_x, const int cblock_stride_y,
const int cnbins, const int cblock_hist_size, const int img_block_width,
const int blocks_in_group, const int blocks_total,
const int grad_quadstep, const int qangle_step,
__global const float* grad, __global const uchar* qangle,
const float scale, __global float* block_hists, __local float* smem)
{
const int lidX = get_local_id(0);
const int lx = get_local_id(0);
const int lp = lx / 24; /* local group id */
const int gid = get_group_id(0) * blocks_in_group + lp;/* global group id */
const int gidY = gid / img_block_width;
const int gidX = gid - gidY * img_block_width;
const int lidX = lx - lp * 24;
const int lidY = get_local_id(1);
const int gidX = get_group_id(0);
const int gidY = get_group_id(1);
const int cell_x = lidX / 16;
const int cell_x = lidX / 12;
const int cell_y = lidY;
const int cell_thread_x = lidX & 0xF;
const int cell_thread_x = lidX - cell_x * 12;
__local float* hists = smem;
__local float* final_hist = smem + cnbins * 48;
__local float* hists = smem + lp * cnbins * (CELLS_PER_BLOCK_X *
CELLS_PER_BLOCK_Y * 12 + CELLS_PER_BLOCK_X * CELLS_PER_BLOCK_Y);
__local float* final_hist = hists + cnbins *
(CELLS_PER_BLOCK_X * CELLS_PER_BLOCK_Y * 12);
const int offset_x = gidX * cblock_stride_x + (cell_x << 2) + cell_thread_x;
const int offset_y = gidY * cblock_stride_y + (cell_y << 2);
__global const float* grad_ptr = grad + offset_y * grad_quadstep + (offset_x << 1);
__global const uchar* qangle_ptr = qangle + offset_y * qangle_step + (offset_x << 1);
__global const float* grad_ptr = (gid < blocks_total) ?
grad + offset_y * grad_quadstep + (offset_x << 1) : grad;
__global const uchar* qangle_ptr = (gid < blocks_total) ?
qangle + offset_y * qangle_step + (offset_x << 1) : qangle;
// 12 means that 12 pixels affect on block's cell (in one row)
if (cell_thread_x < 12)
{
__local float* hist = hists + 12 * (cell_y * CELLS_PER_BLOCK_Y + cell_x) + cell_thread_x;
__local float* hist = hists + 12 * (cell_y * CELLS_PER_BLOCK_Y + cell_x) +
cell_thread_x;
for (int bin_id = 0; bin_id < cnbins; ++bin_id)
hist[bin_id * 48] = 0.f;
const int dist_x = -4 + cell_thread_x - 4 * cell_x;
const int dist_center_x = dist_x - 4 * (1 - 2 * cell_x);
const int dist_y_begin = -4 - 4 * lidY;
for (int dist_y = dist_y_begin; dist_y < dist_y_begin + 12; ++dist_y)
......@@ -97,32 +106,43 @@ __kernel void compute_hists_kernel(const int width, const int cblock_stride_x, c
qangle_ptr += qangle_step;
int dist_center_y = dist_y - 4 * (1 - 2 * cell_y);
int dist_center_x = dist_x - 4 * (1 - 2 * cell_x);
float gaussian = exp(-(dist_center_y * dist_center_y + dist_center_x * dist_center_x) * scale);
float interp_weight = (8.f - fabs(dist_y + 0.5f)) * (8.f - fabs(dist_x + 0.5f)) / 64.f;
float gaussian = exp(-(dist_center_y * dist_center_y + dist_center_x *
dist_center_x) * scale);
float interp_weight = (8.f - fabs(dist_y + 0.5f)) *
(8.f - fabs(dist_x + 0.5f)) / 64.f;
hist[bin.x * 48] += gaussian * interp_weight * vote.x;
hist[bin.y * 48] += gaussian * interp_weight * vote.y;
}
barrier(CLK_LOCAL_MEM_FENCE);
volatile __local float* hist_ = hist;
for (int bin_id = 0; bin_id < cnbins; ++bin_id, hist_ += 48)
{
if (cell_thread_x < 6) hist_[0] += hist_[6];
if (cell_thread_x < 3) hist_[0] += hist_[3];
if (cell_thread_x < 6)
hist_[0] += hist_[6];
barrier(CLK_LOCAL_MEM_FENCE);
if (cell_thread_x < 3)
hist_[0] += hist_[3];
#ifdef WAVE_SIZE_1
barrier(CLK_LOCAL_MEM_FENCE);
#endif
if (cell_thread_x == 0)
final_hist[(cell_x * 2 + cell_y) * cnbins + bin_id] = hist_[0] + hist_[1] + hist_[2];
}
final_hist[(cell_x * 2 + cell_y) * cnbins + bin_id] =
hist_[0] + hist_[1] + hist_[2];
}
#ifdef WAVE_SIZE_1
barrier(CLK_LOCAL_MEM_FENCE);
#endif
__global float* block_hist = block_hists + (gidY * img_block_width + gidX) * cblock_hist_size;
int tid = (cell_y * CELLS_PER_BLOCK_Y + cell_x) * 16 + cell_thread_x;
if (tid < cblock_hist_size)
int tid = (cell_y * CELLS_PER_BLOCK_Y + cell_x) * 12 + cell_thread_x;
if ((tid < cblock_hist_size) && (gid < blocks_total))
{
__global float* block_hist = block_hists +
(gidY * img_block_width + gidX) * cblock_hist_size;
block_hist[tid] = final_hist[tid];
}
}
//-------------------------------------------------------------
......@@ -133,21 +153,59 @@ float reduce_smem(volatile __local float* smem, int size)
unsigned int tid = get_local_id(0);
float sum = smem[tid];
if (size >= 512) { if (tid < 256) smem[tid] = sum = sum + smem[tid + 256]; barrier(CLK_LOCAL_MEM_FENCE); }
if (size >= 256) { if (tid < 128) smem[tid] = sum = sum + smem[tid + 128]; barrier(CLK_LOCAL_MEM_FENCE); }
if (size >= 128) { if (tid < 64) smem[tid] = sum = sum + smem[tid + 64]; barrier(CLK_LOCAL_MEM_FENCE); }
if (size >= 512)
{
if (tid < 256) smem[tid] = sum = sum + smem[tid + 256];
barrier(CLK_LOCAL_MEM_FENCE);
}
if (size >= 256)
{
if (tid < 128) smem[tid] = sum = sum + smem[tid + 128];
barrier(CLK_LOCAL_MEM_FENCE);
}
if (size >= 128)
{
if (tid < 64) smem[tid] = sum = sum + smem[tid + 64];
barrier(CLK_LOCAL_MEM_FENCE);
}
if (tid < 32)
{
if (size >= 64) smem[tid] = sum = sum + smem[tid + 32];
#if defined(WAVE_SIZE_16) || defined(WAVE_SIZE_1)
}
barrier(CLK_LOCAL_MEM_FENCE);
if (tid < 16)
{
#endif
if (size >= 32) smem[tid] = sum = sum + smem[tid + 16];
#ifdef WAVE_SIZE_1
}
barrier(CLK_LOCAL_MEM_FENCE);
if (tid < 8)
{
#endif
if (size >= 16) smem[tid] = sum = sum + smem[tid + 8];
#ifdef WAVE_SIZE_1
}
barrier(CLK_LOCAL_MEM_FENCE);
if (tid < 4)
{
#endif
if (size >= 8) smem[tid] = sum = sum + smem[tid + 4];
#ifdef WAVE_SIZE_1
}
barrier(CLK_LOCAL_MEM_FENCE);
if (tid < 2)
{
#endif
if (size >= 4) smem[tid] = sum = sum + smem[tid + 2];
#ifdef WAVE_SIZE_1
}
barrier(CLK_LOCAL_MEM_FENCE);
if (tid < 1)
{
#endif
if (size >= 2) smem[tid] = sum = sum + smem[tid + 1];
}
......@@ -224,19 +282,44 @@ __kernel void classify_hists_kernel(const int cblock_hist_size, const int cdescr
if (tid < 64) products[tid] = product = product + products[tid + 64];
barrier(CLK_LOCAL_MEM_FENCE);
volatile __local float* smem = products;
if (tid < 32)
{
volatile __local float* smem = products;
smem[tid] = product = product + smem[tid + 32];
#if defined(WAVE_SIZE_16) || defined(WAVE_SIZE_1)
}
barrier(CLK_LOCAL_MEM_FENCE);
if (tid < 16)
{
volatile __local float* smem = products;
#endif
smem[tid] = product = product + smem[tid + 16];
#ifdef WAVE_SIZE_1
}
barrier(CLK_LOCAL_MEM_FENCE);
if (tid < 8)
{
#endif
smem[tid] = product = product + smem[tid + 8];
#ifdef WAVE_SIZE_1
}
barrier(CLK_LOCAL_MEM_FENCE);
if (tid < 4)
{
#endif
smem[tid] = product = product + smem[tid + 4];
#ifdef WAVE_SIZE_1
}
barrier(CLK_LOCAL_MEM_FENCE);
if (tid < 2)
{
#endif
smem[tid] = product = product + smem[tid + 2];
#ifdef WAVE_SIZE_1
}
barrier(CLK_LOCAL_MEM_FENCE);
if (tid < 1)
{
#endif
smem[tid] = product = product + smem[tid + 1];
}
......
modules/ocl/src/opencl/pyrlk.cl
View file @ 9ba25e9d
This diff is collapsed.
modules/ocl/src/opencl/stereobm.cl
View file @ 9ba25e9d
......@@ -228,7 +228,7 @@ __kernel void stereoKernel(__global unsigned char *left, __global unsigned char
int X = get_group_id(0) * BLOCK_W + get_local_id(0) + maxdisp + radius;
// int Y = get_group_id(1) * ROWSperTHREAD + radius;
#define Y (get_group_id(1) * ROWSperTHREAD + radius)
#define Y (get_group_id(1) * ROWSperTHREAD + radius)
volatile __global unsigned int* minSSDImage = cminSSDImage + X + Y * cminSSD_step;
__global unsigned char* disparImage = disp + X + Y * disp_step;
......@@ -251,9 +251,9 @@ __kernel void stereoKernel(__global unsigned char *left, __global unsigned char
barrier(CLK_LOCAL_MEM_FENCE); //before MinSSD function
uint2 minSSD = MinSSD(col_ssd_cache + get_local_id(0), col_ssd, radius);
if (X < cwidth - radius && Y < cheight - radius)
{
uint2 minSSD = MinSSD(col_ssd_cache + get_local_id(0), col_ssd, radius);
if (minSSD.x < minSSDImage[0])
{
disparImage[0] = (unsigned char)(d + minSSD.y);
......@@ -264,7 +264,7 @@ __kernel void stereoKernel(__global unsigned char *left, __global unsigned char
for(int row = 1; row < end_row; row++)
{
int idx1 = y_tex * img_step + x_tex;
int idx2 = (y_tex + (2 * radius + 1)) * img_step + x_tex;
int idx2 = min(y_tex + (2 * radius + 1), cheight - 1) * img_step + x_tex;
barrier(CLK_GLOBAL_MEM_FENCE);
barrier(CLK_LOCAL_MEM_FENCE);
......@@ -278,10 +278,10 @@ __kernel void stereoKernel(__global unsigned char *left, __global unsigned char
barrier(CLK_LOCAL_MEM_FENCE);
uint2 minSSD = MinSSD(col_ssd_cache + get_local_id(0), col_ssd, radius);
if (X < cwidth - radius && row < cheight - radius - Y)
{
int idx = row * cminSSD_step;
uint2 minSSD = MinSSD(col_ssd_cache + get_local_id(0), col_ssd, radius);
if (minSSD.x < minSSDImage[idx])
{
disparImage[disp_step * row] = (unsigned char)(d + minSSD.y);
......@@ -378,8 +378,8 @@ __kernel void textureness_kernel(__global unsigned char *disp, int disp_rows, in
int beg_row = group_id_y * RpT;
int end_row = min(beg_row + RpT, disp_rows);
// if (x < disp_cols)
// {
// if (x < disp_cols)
// {
int y = beg_row;
float sum = 0;
......
modules/ocl/src/pyrlk.cpp
View file @ 9ba25e9d
This diff is collapsed.
modules/ocl/test/main.cpp
View file @ 9ba25e9d
......@@ -115,10 +115,9 @@ int main(int argc, char **argv)
std::cout << "platform invalid\n";
return -1;
}
if(pid != 0 || device != 0)
{
setDevice(oclinfo[pid], device);
}
cout << "Device type:" << type << endl << "Device name:" << oclinfo[pid].DeviceName[device] << endl;
return RUN_ALL_TESTS();
}
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment