Open Source Computer Vision Library https://opencv.org/
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

688 lines
19 KiB

#include "perf_precomp.hpp"
PERF_TEST_P(DevInfo_Size_MatType, transpose, testing::Combine(testing::ValuesIn(devices()),
testing::Values(GPU_TYPICAL_MAT_SIZES),
testing::Values(CV_8UC1, CV_32SC1, CV_64FC1)))
{
DeviceInfo devInfo = std::tr1::get<0>(GetParam());
Size size = std::tr1::get<1>(GetParam());
int type = std::tr1::get<2>(GetParam());
setDevice(devInfo.deviceID());
Mat src_host(size, type);
declare.in(src_host, WARMUP_RNG);
GpuMat src(src_host);
GpuMat dst(size.width, size.height, type);
declare.time(0.5).iterations(100);
SIMPLE_TEST_CYCLE()
{
transpose(src, dst);
}
Mat dst_host = dst;
SANITY_CHECK(dst_host);
}
PERF_TEST_P(DevInfo_Size_MatType_FlipCode, flip, testing::Combine(testing::ValuesIn(devices()),
testing::Values(GPU_TYPICAL_MAT_SIZES),
testing::Values(CV_8UC1, CV_8UC4),
testing::Values((int)HORIZONTAL_AXIS, (int)VERTICAL_AXIS, (int)BOTH_AXIS)))
{
DeviceInfo devInfo = std::tr1::get<0>(GetParam());
Size size = std::tr1::get<1>(GetParam());
int type = std::tr1::get<2>(GetParam());
int flipCode = std::tr1::get<3>(GetParam());
setDevice(devInfo.deviceID());
Mat src_host(size, type);
declare.in(src_host, WARMUP_RNG);
GpuMat src(src_host);
GpuMat dst(size, type);
declare.time(0.5).iterations(100);
SIMPLE_TEST_CYCLE()
{
flip(src, dst, flipCode);
}
Mat dst_host = dst;
SANITY_CHECK(dst_host);
}
PERF_TEST_P(DevInfo_Size_MatType, LUT, testing::Combine(testing::ValuesIn(devices()),
testing::Values(GPU_TYPICAL_MAT_SIZES),
testing::Values(CV_8UC1, CV_8UC3)))
{
DeviceInfo devInfo = std::tr1::get<0>(GetParam());
Size size = std::tr1::get<1>(GetParam());
int type = std::tr1::get<2>(GetParam());
setDevice(devInfo.deviceID());
Mat src_host(size, type);
Mat lut(1, 256, CV_8UC1);
declare.in(src_host, lut, WARMUP_RNG);
GpuMat src(src_host);
GpuMat dst(size, type);
declare.time(0.5).iterations(100);
SIMPLE_TEST_CYCLE()
{
LUT(src, lut, dst);
}
Mat dst_host = dst;
SANITY_CHECK(dst_host);
}
PERF_TEST_P(DevInfo_Size, cartToPolar, testing::Combine(testing::ValuesIn(devices()),
testing::Values(GPU_TYPICAL_MAT_SIZES)))
{
DeviceInfo devInfo = std::tr1::get<0>(GetParam());
Size size = std::tr1::get<1>(GetParam());
setDevice(devInfo.deviceID());
Mat x_host(size, CV_32FC1);
Mat y_host(size, CV_32FC1);
declare.in(x_host, y_host, WARMUP_RNG);
GpuMat x(x_host);
GpuMat y(y_host);
GpuMat magnitude(size, CV_32FC1);
GpuMat angle(size, CV_32FC1);
declare.time(0.5).iterations(100);
SIMPLE_TEST_CYCLE()
{
cartToPolar(x, y, magnitude, angle);
}
Mat magnitude_host = magnitude;
Mat angle_host = angle;
SANITY_CHECK(magnitude_host);
SANITY_CHECK(angle_host);
}
PERF_TEST_P(DevInfo_Size, polarToCart, testing::Combine(testing::ValuesIn(devices()),
testing::Values(GPU_TYPICAL_MAT_SIZES)))
{
DeviceInfo devInfo = std::tr1::get<0>(GetParam());
Size size = std::tr1::get<1>(GetParam());
setDevice(devInfo.deviceID());
Mat magnitude_host(size, CV_32FC1);
Mat angle_host(size, CV_32FC1);
declare.in(magnitude_host, angle_host, WARMUP_RNG);
GpuMat magnitude(magnitude_host);
GpuMat angle(angle_host);
GpuMat x(size, CV_32FC1);
GpuMat y(size, CV_32FC1);
declare.time(0.5).iterations(100);
SIMPLE_TEST_CYCLE()
{
polarToCart(magnitude, angle, x, y);
}
Mat x_host = x;
Mat y_host = angle;
SANITY_CHECK(x_host);
SANITY_CHECK(y_host);
}
PERF_TEST_P(DevInfo_Size_MatType, addMat, testing::Combine(testing::ValuesIn(devices()),
testing::Values(GPU_TYPICAL_MAT_SIZES),
testing::Values(CV_8UC1, CV_8UC4, CV_32FC1)))
{
DeviceInfo devInfo = std::tr1::get<0>(GetParam());
Size size = std::tr1::get<1>(GetParam());
int type = std::tr1::get<2>(GetParam());
setDevice(devInfo.deviceID());
Mat a_host(size, type);
Mat b_host(size, type);
declare.in(a_host, b_host, WARMUP_RNG);
GpuMat a(a_host);
GpuMat b(b_host);
GpuMat c(size, type);
declare.time(0.5).iterations(100);
SIMPLE_TEST_CYCLE()
{
add(a, b, c);
}
Mat c_host = c;
SANITY_CHECK(c_host);
}
PERF_TEST_P(DevInfo_Size_MatType, addScalar, testing::Combine(testing::ValuesIn(devices()),
testing::Values(GPU_TYPICAL_MAT_SIZES),
testing::Values(CV_32FC1, CV_32FC2)))
{
DeviceInfo devInfo = std::tr1::get<0>(GetParam());
Size size = std::tr1::get<1>(GetParam());
int type = std::tr1::get<2>(GetParam());
setDevice(devInfo.deviceID());
Mat a_host(size, type);
declare.in(a_host, WARMUP_RNG);
GpuMat a(a_host);
Scalar b(1,2,3,4);
GpuMat c(size, type);
declare.time(0.5).iterations(100);
SIMPLE_TEST_CYCLE()
{
add(a, b, c);
}
Mat c_host = c;
SANITY_CHECK(c_host);
}
PERF_TEST_P(DevInfo_Size_MatType, subtractMat, testing::Combine(testing::ValuesIn(devices()),
testing::Values(GPU_TYPICAL_MAT_SIZES),
testing::Values(CV_8UC1, CV_8UC4, CV_16SC1, CV_32FC1)))
{
DeviceInfo devInfo = std::tr1::get<0>(GetParam());
Size size = std::tr1::get<1>(GetParam());
int type = std::tr1::get<2>(GetParam());
setDevice(devInfo.deviceID());
Mat a_host(size, type);
Mat b_host(size, type);
declare.in(a_host, b_host, WARMUP_RNG);
GpuMat a(a_host);
GpuMat b(b_host);
GpuMat c(size, type);
declare.time(0.5).iterations(100);
SIMPLE_TEST_CYCLE()
{
subtract(a, b, c);
}
Mat c_host = c;
SANITY_CHECK(c_host);
}
PERF_TEST_P(DevInfo_Size, multiplyMat, testing::Combine(testing::ValuesIn(devices()),
testing::Values(GPU_TYPICAL_MAT_SIZES)))
{
DeviceInfo devInfo = std::tr1::get<0>(GetParam());
Size size = std::tr1::get<1>(GetParam());
setDevice(devInfo.deviceID());
Mat a_host(size, CV_8UC1);
Mat b_host(size, CV_32FC1);
declare.in(a_host, b_host, WARMUP_RNG);
GpuMat a(a_host);
GpuMat b(b_host);
GpuMat c;
declare.time(0.5).iterations(100);
SIMPLE_TEST_CYCLE()
{
multiply(a, b, c);
}
Mat c_host = c;
SANITY_CHECK(c_host);
}
PERF_TEST_P(DevInfo_Size_MatType, multiplyScalar, testing::Combine(testing::ValuesIn(devices()),
testing::Values(GPU_TYPICAL_MAT_SIZES),
testing::Values(CV_8UC1, CV_32FC1)))
{
DeviceInfo devInfo = std::tr1::get<0>(GetParam());
Size size = std::tr1::get<1>(GetParam());
int type = std::tr1::get<2>(GetParam());
setDevice(devInfo.deviceID());
Mat a_host(size, type);
declare.in(a_host, WARMUP_RNG);
GpuMat a(a_host);
Scalar b(1,2,3,4);
GpuMat c(size, type);
declare.time(0.5).iterations(100);
SIMPLE_TEST_CYCLE()
{
multiply(a, b, c);
}
Mat c_host = c;
SANITY_CHECK(c_host);
}
PERF_TEST_P(DevInfo_Size, exp, testing::Combine(testing::ValuesIn(devices()),
testing::Values(GPU_TYPICAL_MAT_SIZES)))
{
DeviceInfo devInfo = std::tr1::get<0>(GetParam());
Size size = std::tr1::get<1>(GetParam());
setDevice(devInfo.deviceID());
Mat a_host(size, CV_32FC1);
declare.in(a_host, WARMUP_RNG);
GpuMat a(a_host);
GpuMat b(size, CV_32FC1);
declare.time(0.5).iterations(100);
SIMPLE_TEST_CYCLE()
{
exp(a, b);
}
Mat b_host = b;
SANITY_CHECK(b_host);
}
PERF_TEST_P(DevInfo_Size_MatType, pow, testing::Combine(testing::ValuesIn(devices()),
testing::Values(GPU_TYPICAL_MAT_SIZES),
testing::Values(CV_8UC1, CV_8UC4, CV_16UC1, CV_32FC1)))
{
DeviceInfo devInfo = std::tr1::get<0>(GetParam());
Size size = std::tr1::get<1>(GetParam());
int type = std::tr1::get<2>(GetParam());
setDevice(devInfo.deviceID());
Mat src_host(size, type);
declare.in(src_host, WARMUP_RNG);
GpuMat src(src_host);
GpuMat dst(size, type);
declare.time(0.5).iterations(100);
SIMPLE_TEST_CYCLE()
{
pow(src, 2.0, dst);
}
Mat dst_host = dst;
SANITY_CHECK(dst_host);
}
PERF_TEST_P(DevInfo_Size_MatType_CmpOp, compare, testing::Combine(testing::ValuesIn(devices()),
testing::Values(GPU_TYPICAL_MAT_SIZES),
testing::Values(CV_8UC4, CV_32FC1),
testing::Values((int)CMP_NE, (int)CMP_EQ)))
{
DeviceInfo devInfo = std::tr1::get<0>(GetParam());
Size size = std::tr1::get<1>(GetParam());
int type = std::tr1::get<2>(GetParam());
int cmpop = std::tr1::get<3>(GetParam());
setDevice(devInfo.deviceID());
Mat src1_host(size, type);
Mat src2_host(size, type);
declare.in(src1_host, src2_host, WARMUP_RNG);
GpuMat src1(src1_host);
GpuMat src2(src2_host);
GpuMat dst(size, type);
declare.time(0.5).iterations(100);
SIMPLE_TEST_CYCLE()
{
compare(src1, src2, dst, cmpop);
}
Mat dst_host = dst;
SANITY_CHECK(dst_host);
}
PERF_TEST_P(DevInfo_Size_MatType, bitwise_not, testing::Combine(testing::ValuesIn(devices()),
testing::Values(GPU_TYPICAL_MAT_SIZES),
testing::Values(CV_8UC1, CV_16UC1, CV_32SC1)))
{
DeviceInfo devInfo = std::tr1::get<0>(GetParam());
Size size = std::tr1::get<1>(GetParam());
int type = std::tr1::get<2>(GetParam());
setDevice(devInfo.deviceID());
Mat src_host(size, type);
declare.in(src_host, WARMUP_RNG);
GpuMat src(src_host);
GpuMat dst(size, type);
declare.time(0.5).iterations(100);
SIMPLE_TEST_CYCLE()
{
bitwise_not(src, dst);
}
Mat dst_host = dst;
SANITY_CHECK(dst_host);
}
PERF_TEST_P(DevInfo_Size_MatType, bitwise_and, testing::Combine(testing::ValuesIn(devices()),
testing::Values(GPU_TYPICAL_MAT_SIZES),
testing::Values(CV_8UC1, CV_16UC1, CV_32SC1)))
{
DeviceInfo devInfo = std::tr1::get<0>(GetParam());
Size size = std::tr1::get<1>(GetParam());
int type = std::tr1::get<2>(GetParam());
setDevice(devInfo.deviceID());
Mat src1_host(size, type);
Mat src2_host(size, type);
declare.in(src1_host, src2_host, WARMUP_RNG);
GpuMat src1(src1_host);
GpuMat src2(src2_host);
GpuMat dst(size, type);
declare.time(0.5).iterations(100);
SIMPLE_TEST_CYCLE()
{
bitwise_and(src1, src2, dst);
}
Mat dst_host = dst;
SANITY_CHECK(dst_host);
}
PERF_TEST_P(DevInfo_Size_MatType, min, testing::Combine(testing::ValuesIn(devices()),
testing::Values(GPU_TYPICAL_MAT_SIZES),
testing::Values(CV_8UC1, CV_16UC1, CV_32SC1)))
{
DeviceInfo devInfo = std::tr1::get<0>(GetParam());
Size size = std::tr1::get<1>(GetParam());
int type = std::tr1::get<2>(GetParam());
setDevice(devInfo.deviceID());
Mat src1_host(size, type);
Mat src2_host(size, type);
declare.in(src1_host, src2_host, WARMUP_RNG);
GpuMat src1(src1_host);
GpuMat src2(src2_host);
GpuMat dst(size, type);
declare.time(0.5).iterations(100);
SIMPLE_TEST_CYCLE()
{
min(src1, src2, dst);
}
Mat dst_host = dst;
SANITY_CHECK(dst_host);
}
PERF_TEST_P(DevInfo_Size, meanStdDev, testing::Combine(testing::ValuesIn(devices()),
testing::Values(GPU_TYPICAL_MAT_SIZES)))
{
DeviceInfo devInfo = std::tr1::get<0>(GetParam());
Size size = std::tr1::get<1>(GetParam());
setDevice(devInfo.deviceID());
Mat src_host(size, CV_8UC1);
declare.in(src_host, WARMUP_RNG);
GpuMat src(src_host);
Scalar mean;
Scalar stddev;
declare.time(0.5).iterations(100);
SIMPLE_TEST_CYCLE()
{
meanStdDev(src, mean, stddev);
}
SANITY_CHECK(mean);
SANITY_CHECK(stddev);
}
PERF_TEST_P(DevInfo_Size_MatType_NormType, norm, testing::Combine(testing::ValuesIn(devices()),
testing::Values(GPU_TYPICAL_MAT_SIZES),
testing::Values(CV_8UC1, CV_16UC1, CV_32SC1),
testing::Values((int)NORM_INF, (int)NORM_L1, (int)NORM_L2)))
{
DeviceInfo devInfo = std::tr1::get<0>(GetParam());
Size size = std::tr1::get<1>(GetParam());
int type = std::tr1::get<2>(GetParam());
int normType = std::tr1::get<3>(GetParam());
setDevice(devInfo.deviceID());
Mat src_host(size, type);
declare.in(src_host, WARMUP_RNG);
GpuMat src(src_host);
double dst;
GpuMat buf;
declare.time(0.5).iterations(100);
SIMPLE_TEST_CYCLE()
{
dst = norm(src, normType, buf);
}
SANITY_CHECK(dst);
}
PERF_TEST_P(DevInfo_Size_NormType, normDiff, testing::Combine(testing::ValuesIn(devices()),
testing::Values(GPU_TYPICAL_MAT_SIZES),
testing::Values((int)NORM_INF, (int)NORM_L1, (int)NORM_L2)))
{
DeviceInfo devInfo = std::tr1::get<0>(GetParam());
Size size = std::tr1::get<1>(GetParam());
int normType = std::tr1::get<2>(GetParam());
setDevice(devInfo.deviceID());
Mat src1_host(size, CV_8UC1);
Mat src2_host(size, CV_8UC1);
declare.in(src1_host, src2_host, WARMUP_RNG);
GpuMat src1(src1_host);
GpuMat src2(src2_host);
double dst;
declare.time(0.5).iterations(100);
SIMPLE_TEST_CYCLE()
{
dst = norm(src1, src2, normType);
}
SANITY_CHECK(dst);
}
PERF_TEST_P(DevInfo_Size_MatType, sum, testing::Combine(testing::ValuesIn(devices()),
testing::Values(GPU_TYPICAL_MAT_SIZES),
testing::Values(CV_8UC1, CV_16UC1, CV_32FC1)))
{
DeviceInfo devInfo = std::tr1::get<0>(GetParam());
Size size = std::tr1::get<1>(GetParam());
int type = std::tr1::get<2>(GetParam());
setDevice(devInfo.deviceID());
Mat src_host(size, type);
declare.in(src_host, WARMUP_RNG);
GpuMat src(src_host);
Scalar dst;
GpuMat buf;
declare.time(0.5).iterations(100);
SIMPLE_TEST_CYCLE()
{
dst = sum(src, buf);
}
SANITY_CHECK(dst);
}
PERF_TEST_P(DevInfo_Size_MatType, minMax, testing::Combine(testing::ValuesIn(devices()),
testing::Values(GPU_TYPICAL_MAT_SIZES),
testing::Values(CV_8UC1, CV_16UC1, CV_32FC1)))
{
DeviceInfo devInfo = std::tr1::get<0>(GetParam());
Size size = std::tr1::get<1>(GetParam());
int type = std::tr1::get<2>(GetParam());
setDevice(devInfo.deviceID());
Mat src_host(size, type);
declare.in(src_host, WARMUP_RNG);
GpuMat src(src_host);
double minVal, maxVal;
GpuMat buf;
declare.time(0.5).iterations(100);
SIMPLE_TEST_CYCLE()
{
minMax(src, &minVal, &maxVal, GpuMat(), buf);
}
SANITY_CHECK(minVal);
SANITY_CHECK(maxVal);
}
PERF_TEST_P(DevInfo_Size_MatType, minMaxLoc, testing::Combine(testing::ValuesIn(devices()),
testing::Values(GPU_TYPICAL_MAT_SIZES),
testing::Values(CV_8UC1, CV_16UC1, CV_32FC1)))
{
DeviceInfo devInfo = std::tr1::get<0>(GetParam());
Size size = std::tr1::get<1>(GetParam());
int type = std::tr1::get<2>(GetParam());
setDevice(devInfo.deviceID());
Mat src_host(size, type);
declare.in(src_host, WARMUP_RNG);
GpuMat src(src_host);
double minVal, maxVal;
Point minLoc, maxLoc;
GpuMat valbuf, locbuf;
declare.time(0.5).iterations(100);
SIMPLE_TEST_CYCLE()
{
minMaxLoc(src, &minVal, &maxVal, &minLoc, &maxLoc, GpuMat(), valbuf, locbuf);
}
SANITY_CHECK(minVal);
SANITY_CHECK(maxVal);
}
PERF_TEST_P(DevInfo_Size_MatType, countNonZero, testing::Combine(testing::ValuesIn(devices()),
testing::Values(GPU_TYPICAL_MAT_SIZES),
testing::Values(CV_8UC1, CV_16UC1, CV_32FC1)))
{
DeviceInfo devInfo = std::tr1::get<0>(GetParam());
Size size = std::tr1::get<1>(GetParam());
int type = std::tr1::get<2>(GetParam());
setDevice(devInfo.deviceID());
Mat src_host(size, type);
declare.in(src_host, WARMUP_RNG);
GpuMat src(src_host);
int dst;
GpuMat buf;
declare.time(0.5).iterations(100);
SIMPLE_TEST_CYCLE()
{
dst = countNonZero(src, buf);
}
SANITY_CHECK(dst);
}